Datasets:

Hack90

/

europe_pmc_articles_part_1

like 2

Comparison of written reports of mammography, sonography and magnetic resonance mammography for preoperative evaluation of breast lesions, with special emphasis on magnetic resonance mammography

<p>Patients with abnormal breast findings (<italic>n</italic> = 413) were examined by mammography, sonography and magnetic resonance (MR) mammography; 185 invasive cancers, 38 carcinoma <italic>in situ</italic> and 254 benign tumours were confirmed histologically. Sensitivity for mammography was 83.7%, for sonography it was 89.1% and for MR mammography it was 94.6% for invasive cancers. In 42 patients with multifocal invasive cancers, multifocality had been detected by mammography and sonography in 26.2%, and by MR mammography in 66.7%. In nine patients with multicentric cancers, detection rates were 55.5, 55.5 and 88.8%, respectively. Carcinoma <italic>in situ</italic> was diagnosed by mammography in 78.9% and by MR mammography in 68.4% of patients. Combination of all three diagnostic methods lead to the best results for detection of invasive cancer and multifocal disease. However, sensitivity of mammography and sonography combined was identical to that of MR mammography (ie 94.6%).</p>

<contrib id="A1" contrib-type="author"><name><surname>Malur</surname><given-names>Sabine</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Wurdinger</surname><given-names>Susanne</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Moritz</surname><given-names>Andreas</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Michels</surname><given-names>Wolfgang</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Schneider</surname><given-names>Achim</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>aschneider@med.uni-jena.de</email></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>Mammography and sonography are the standard imaging techniques for detection and evaluation of breast disease [<xref ref-type="bibr" rid="B1">1</xref>]. Mammography is the most established screening modality [<xref ref-type="bibr" rid="B2">2</xref>]. Especially in young women and women with dense breasts, sonography appears superior to mammography, and differentiation between solid tumours and cysts is easier. Sensitivity and specificity of sonography or mammography are higher if sonography and mammography are combined [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>It is generally accepted that MR mammography is the most sensitive technique for diagnosis of breast cancer, whereas the reported specificity of MR mammography varies [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. In those studies, MR mammography was performed and evaluated by highly specialized radiologists in a research setting. It was therefore the purpose of the present prospective study to compare the validity of MR mammography with mammography and sonography in clinical routine practice. Findings for the three diagnostic methods documented on routine reports that were available to the surgeon preoperatively formed the basis of this comparison. Special emphasis was placed on the identification of multifocal and multicentric invasive disease.</p></sec><sec sec-type="methods"><title>Patients and methods</title><sec><title>Patients</title><p>Between September 1995 and September 1998, 413 patients with abnormal breast findings were referred for histological evaluation to the Department of Gynecology of the Friedrich-Schiller University, Jena, Germany. Patients had been selected and referred because of the presence of breast lesions detected by palpation and/or mammography and/or sonography. In addition, MR mammography was performed in all patients. We excluded five patients with invasive cancer who had a history of core-needle or fine-needle biopsy cancer within 2 weeks before referral, because the presence of haematoma may mimic false-positive findings on MR mammography. In addition, five patients who did not keep still during MR mammography were excluded.</p></sec><sec><title>Imaging</title><p>Analysis of the sonograms taken in patients with histologically confirmed carcinoma <italic>in situ</italic> were excluded from analysis because the value of sonography for detection of premalignant disease is limited. Mammography was not performed in 32 patients who were younger than 30 years or who had had a mastectomy with suspected local recurrence. For all patients, written reports of mammographic, sonographic and dynamic MR mammographic findings were available preoperatively.</p><p>The majority of mammograms (68%) were performed at the Institute for Diagnostic and Interventional Radiology, Friedrich-Schiller-University, using a senograph DMR (GE Medical Systems, Milwaukee, Wisconsin, USA) with standard craniocaudal and mediolateral oblique projections. Mammograms obtained at other institutions that were considered to match the quality standards of our institution were also accepted for evaluation.</p><p>All sonography and MR examinations were carried out at the Department of Gynecology and the Institute for Diagnostic Interventional Radiology, Friedrich-Schiller-University, respectively. Sonography was done using a 7.5-MHz linear array probe with a Sonoline Versa Pro (Siemens, Erlangen, Germany). MR mammography was performed exclusively at the Institute for Diagnostic Interventional Radiology, using a Gyroscan ACSII (Philips, Nijmegen, The Netherlands) with a field strength of 1.5 T using a double-breast coil. Dynamic T1-weighted images were aquired using a multislice two-dimensional fast-field-echo (FFE) sequence. We used the following parameters: TR 97, TE 5.0, flip angle 80&#x000b0;, slice thickness 4.0 mm, gap 0.4 mm, field of view 350 mm and transverse orientation. In addition T2-weighted images (4000/300/90&#x000b0;/4.0 mm/0.4/350 mm) were obtained. As contrast medium, 0.1 mmol gadolinium-DTPA/kg body weight (Magnevist; Schering, Berlin, Germany) was used and injected as a bolus. One unenhanced and seven enhanced studies were acquired with an acquisition time of 1 min. Criteria for malignancy were signal enhancement of 90% or more within the first 2 min after bolus injection and signal plateau or washout phenomena afterward. Additional criteria were irregular borders of the lesion and low signal intensity in the T2-weighted images.</p><p>Mammograms were read by three different radiologists, sonography was done by three different gynaecologists and MR mammography was interpreted by a total of six different radiologists.</p><p>Definition for multifocal carcinoma was a distance of less than 3 cm and for multicentric carcinoma a distance over 3 cm between various lesions.</p></sec><sec><title>Statistical analysis</title><p>Interpretation of the various diagnostic procedures was compared with the histological examination with regard to sensitivity, specificity, accuracy, and positive and negative predictive value. Criteria for suspected malignancy in the written reports were the terms 'cancer', 'malignant lesion or tumor', or 'suspicious for cancer'. Sensitivity, specificity, negative and positive predicitive value, and accuracy were evaluated as follows:</p><p>Sensitivity = patients with suspected breast cancer/patients with histologically confirmed breast cancer</p><p>Specificity = patients with suspected benign disease/patients with histologically confirmed benign disease</p><p>Positive predictive value = patients with histologically confirmed breast cancer/patients with suspected breast cancer</p><p>Negative predictive value = patients with histologically confirmed benign disease/patients with suspected benign disease</p><p>Accuracy = patients with true-positive and true-negative detected disease/patients with histologically confirmed breast cancer</p><p>A result was classified as false-negative when a diagnostic method classified a histologically confirmed cancer as benign. A result was classified as false-positive when a diagnostic method classified a histologically confirmed benign lesion as cancer. We compared the preformance of all diagnostic methods individually and in combination using the results from all patients. Statistical analysis was performed for all variables with Fisher's exact test and Pearson's &#x003c7;² test.</p></sec></sec><sec><title>Results</title><p>All patients underwent breast surgery and all abnormal lesions identified by mammography, sonography or MR mammography were surgically removed. A total of 477 breast lesions were examined histologically, revealing the presence of 185 invasive cancers, 38 carcinomata <italic>in situ</italic> and 254 benign lesions (fibroadenoma, papilloma, intraductal or adenoid ductal hyperplasia, cystic mastopathia). There were four patients with malignant lesions in both breasts. In 42 patients multifocal tumours and in nine patients multicentric tumors were found on histological examination. Among the 185 invasive lesions, 178 were primary cancers, five were recurrences, one was metastatic and one was an angiosarcoma. The majority of invasive breast cancers were staged as pT1c (44%). Six per cent of tumors were detected in stage pT1a, 18% in stage pT1b, 25% in stage pT2, 3% in stage pT3 and 4% in stage pT4. The distribution of histopathological tumour types is shown in Table <xref ref-type="table" rid="T1">1</xref>. The mean age of patients was 58 years (range 19-85 years).</p><p>The sensitivity of MR mammography was significantly higher than those of mammography and sonography (<italic>P</italic> &#x0003c; 0.005 and <italic>P</italic> &#x0003c; 0.05; Table <xref ref-type="table" rid="T2">2</xref>). The specificity of sonography was significantly higher than those of mammography and MR mammography (<italic>P</italic> &#x0003c; 0.05 and <italic>P</italic> &#x0003c; 0.005; Table <xref ref-type="table" rid="T2">2</xref>). The negative predictive values for sonography and MR mammography were significantly higher than that of mammography (<italic>P</italic> &#x0003c; 0.05 and <italic>P</italic> &#x0003c; 0.005; Table <xref ref-type="table" rid="T2">2</xref>). With regard to accuracy, no significant difference between the three modalities was found (Table <xref ref-type="table" rid="T2">2</xref>). Combining of all three diagnostic methods yielded the best results for detection of cancer (<italic>P</italic> &#x0003c; 0.005; Table <xref ref-type="table" rid="T3">3</xref>). The sensitivity and negative predictive value for the combination of mammography and MR mammography, and the combination of sonography and MR mammography were significantly higher than those for the combination of mammograpy and sonography (<italic>P</italic> &#x0003c; 0.05; Table <xref ref-type="table" rid="T3">3</xref>). The highest result for accuracy was seen for a combination of all three methods (<italic>P</italic> &#x0003c; 0.05; Table <xref ref-type="table" rid="T3">3</xref>).</p><p>Mammography was false-negative in 30 out of 184 invasive cancers, sonography was false-negative in 20 out of 185 cancers, and 10 out of 185 invasive cancers were missed by MR mammography. The majority of false-negative findings was found in stage1 disease, ductal carcinoma and grade 3 tumors (Table <xref ref-type="table" rid="T4">4</xref>). Of 10 invasive cancers missed by MR mammography, eight were found by mammography and sonography. By all three techniques, one invasive ductal carcinoma (pT1b) was misinterpreted as fibroadenoma. In another patient, a microinvasive lobular carcinoma of 5 mm diameter was not detected with mammography and MR mammography, whereas sonography detected a solid, benign tumour. MR mammography identified 10 invasive cancers (5.2%) that were missed by mammography and sonography, whereas one invasive cancer was found by mammography alone. By sonography alone, not a single case of invasive disease was detected when MR mammography or mammography were nonsuspected.</p><p>The highest detection rate for multifocal invasive disease was seen with MR mammography, which identified 28 out of 42 (66.7%) histologically confirmed multifocal invasive cancers, whereas mammography and sonography both identified 11 (26.2%) of these cancers (<italic>P</italic> &#x0003c; 0.05). The combination of all three diagnostic methods leads to the best result for detection of multifocality (76.2%; <italic>P</italic> &#x0003c; 0.05), whereas the detection rate with the combination of mammography and sonography was 35.7%, with the combination of sonography and MR mammography it was 69% (<italic>P</italic> &#x0003c; 0.05 versus mammography + sonography), and with the combination of mammography and MR mammography it was 73.8% (<italic>P</italic> &#x0003c; 0.05 versus mammography + sonography). Multifocal invasive disease was suspected in 12 patients by mammography, in 13 patients by sonography, and in 16 patients by MR mammography, but only unifocal disease was confirmed by histology. Out of nine patients with histologically confirmed multicentric invasive cancer, eight (88.8%) of these cancers were detected by MR mammography and five (55.5%) by mammography or sonography. One patient was diagnosed with multicentric invasive-lobular carcinoma stage pT2G2, which had been misinterpreted as benign tumour by sonography and mammography, and as haematoma by MR mammography.</p><p>Out of 38 patients with carcinoma <italic>in situ</italic>, mammography (suspicious microcalcifications, exclusively) identified 30 cases (78.9%) and MR mammography identified 26 cases (68.4%). When combining mammography and MR mammography, sensitivity for detection of carcinoma <italic>in situ</italic> increased to 87% (not significant).</p></sec><sec><title>Discussion</title><p>When the validity of individual diagnostic methods for detection of invasive breast cancer was analyzed, the sensitivity and specificity of mammography ranged from 79.9 to 89% and from 64 to 93.5%, respectively [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B13">13</xref>]; for sonography from 67.6 to 96% and from 93 to 97.7%, respectively [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]; and for MR mammography from 91 to 98.9 and from 20 to 97.4%, respectively [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>].</p><p>The performance of mammography, sonography and MR mammography was compared in three large series (Table <xref ref-type="table" rid="T5">5</xref>). The present results are similar with regard to sensitivity and specificity for the detection of malignant breast lesions, with MR mammography reaching the highest sensitivity of all imaging procedures.</p><p>When combinations of all three technique were analyzed, mammography and sonography (standard method) had a sensitivity of 83% and a specificity of 92% for detection of malignant disease [<xref ref-type="bibr" rid="B3">3</xref>]. A combination of mammography, sonography and MR mammography (combined method) showed a sensitivity of 95% and a specificity of 64% [<xref ref-type="bibr" rid="B3">3</xref>]. For nonpalpable lesions, sensitivity increased from 73% by the standard method to 82% for the combined method. Specificity for the standard method (89%) was higher than that for the combined method (71%). For palpable lesions a sensitivity of 85% for the standard and 98% for the combined method was achieved, whereas specificity for the standard method was 100% compared with 45% for the combined methods [<xref ref-type="bibr" rid="B3">3</xref>]. The positive predictive value was 94% for the standard and 80% for the combined methods, and the negative predictive values were 78 and 89%, respectively [<xref ref-type="bibr" rid="B3">3</xref>]. In the present study, we also found the highest sensitivity, specificity, and positive and negative predictive values for the combination of all three methods. Combination of mammography and sonography was as sensitive as MR mammography alone (94.6% versus 94.6%).</p><p>The majority of false-positive results for invasive cancer by MR mammography (80 out of 439) were caused by papillomas, intraductal hyperplasia grade 2 or 3, or fibroadenomas in the present series. These lesions have a good blood supply and may mimic invasive cancer [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B18">18</xref>].</p><p>Ten out of 185 (5.4%) malignant lesions were classified as false-negative by MR mammography. On histology, the majority of false-negative invasive cancers were lobular cancers (four out of 10). Bone <italic>et al</italic> [<xref ref-type="bibr" rid="B18">18</xref>] reported false-negative results in 11 out of 155 readings, with the majority being lobular cancers on histology. Lack of tumour-induced neovascularity may explain such findings. In particular, invasive lobular cancers infiltrate the normal tissue with columns of single cells, and receive adequate oxygenation without the requirement for increased vascularization [<xref ref-type="bibr" rid="B19">19</xref>]. Buadu <italic>et al</italic> [<xref ref-type="bibr" rid="B11">11</xref>] found that lobular and mucinous carcinomas had a low microvessel density.</p><p>Multifocality of breast cancers can be recognized adequately by MR mammography [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. Boetes <italic>et al</italic> [<xref ref-type="bibr" rid="B22">22</xref>] reported that all 61 multifocal cancers were detected by MR mammography, compared with 31% by mammography and 38% by sonography. Esserman <italic>et al</italic> [<xref ref-type="bibr" rid="B20">20</xref>] detected multifocality by MR mammography in 100% (10 out of 10) versus 44% (four out of nine) by mammography. Relevant changes in therapy due to additional multicentric and contralateral tumour findings by MR mammography occur in 18% of patients as compared with conventional imaging [<xref ref-type="bibr" rid="B23">23</xref>]. We found a detection rate of multifocality of 66.7% by MR mammography, as compared with 26.2% by mammography and sonography. However, in 16 patients multifocal invasive disease as diagnosed by MR mammography was shown to be unifocal by histology.</p><p>Kramer <italic>et al</italic> [<xref ref-type="bibr" rid="B24">24</xref>] reported that MR mammography yielded the highest sensitivity for detection of multicentricity as compared with mammography and sonography (89, 66 and 79%, respectively) in 38 patients. These findings are comparable with the present results, in which eight out of nine multicentric cancers were diagnosed correctly.</p><p>Carcinoma <italic>in situ</italic> is identified by mammography through the presence of suspicious microcalcifications. Suspicious microcalcifications are more frequent in intraductal than in infiltrating cancers [<xref ref-type="bibr" rid="B25">25</xref>], which was also observed in the present series. Mammography showed a detection rate for carcinoma <italic>in situ</italic> of 78.9%, as compared with 65.8% by MR mammography; the combination of mammography and MR mammography lead to a detection rate of 86.4%. Fischer <italic>et al</italic> [<xref ref-type="bibr" rid="B26">26</xref>] reported that carcinoma <italic>in situ</italic> was identified by MR mammography in 25 out of 35 patients (72%); three ductal carcinomata <italic>in situ</italic> were detected by MR mammography exclusively. Sittek <italic>et al</italic> [<xref ref-type="bibr" rid="B27">27</xref>] reported that 14 out of 20 carcinomata <italic>in situ</italic> (70%) were correctly diagnosed by MR mammography on the basis of focal increase of signal intensity. Those authors concluded that carcinoma <italic>in situ</italic> is not reliably detected by MR mammography because of lack of a uniform pattern of enhancement. Esserman <italic>et al</italic> [<xref ref-type="bibr" rid="B20">20</xref>] reported a detection rate of 43% for ductal carcinoma <italic>in situ</italic> by MR mammography. Among 36 woman with carcinoma <italic>in situ</italic>, Gilles <italic>et al</italic> [<xref ref-type="bibr" rid="B28">28</xref>] demonstrated two cases without early contrast enhancement.</p><p>The present study showed that, for detection of breast cancer, MR mammography is not superior to a combination of sonography and mammography. For identification of multifocal or multicentric disease, MR mammography proved to be the most accurate technique.</p></sec>

BRCA1 and BRCA2 protein expressions in an ovotestis of a 46, XX true hermaphrodite

<p><italic>BRCA1</italic> and <italic>BRCA2</italic> breast cancer susceptibility genes encode proteins, the normal cellular functions of which are complex and multiple, and germ-line mutations in individuals predispose both to breast and to ovarian cancer. There is nevertheless substantial evidence linking BRCA1 and BRCA2 to homologous recombination and DNA repair, to transcriptional control and to tissue proliferation. There is controversy regarding the localization of BRCA1 and BRCA2 proteins to either nucleus or cytoplasm and whether the expression is present in premeiotic germ cells or can still be expressed in mitotic spermatogonia. We report herein an immunohistochemical study of BRCA1 and BRCA2 distribution in a rather unsual tissue (an ovotestis), which addresses this issue.</p>

<contrib id="A1" contrib-type="author"><name><surname>Bernard-Gallon</surname><given-names>Dominique J</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>D&#x000e9;chelotte</surname><given-names>Pierre</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Vissac</surname><given-names>C&#x000e9;cile</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Aunoble</surname><given-names>B&#x000e9;n&#x000e9;dicte</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Cravello</surname><given-names>Laetitia</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Malpuech</surname><given-names>Georges</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Bignon</surname><given-names>Yves-Jean</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>Yves-Jean.Bignon@cjp.u-clermont1.fr</email></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>Hereditary predisposition to breast cancer can be attributed to germline mutations in the <italic>BRCA1</italic> or <italic>BRCA2</italic> breast cancer susceptibility genes [<xref ref-type="bibr" rid="B1">1</xref>]. Germline mutations in the <italic>BRCA1</italic> and <italic>BRCA2</italic> genes are associated with the development of breast and ovarian cancers [<xref ref-type="bibr" rid="B2">2</xref>]. <italic>BRCA2</italic> is associated the development of breast cancer in both women and men [<xref ref-type="bibr" rid="B3">3</xref>], and a moderate increased risk for the development of ovarian cancer.</p><p>Zabludoff <italic>et al</italic> [<xref ref-type="bibr" rid="B4">4</xref>] investigated the tissue distribution of <italic>Brca1</italic> mRNA in adult mouse tissues and reported that <italic>Brca1</italic> mRNA levels were most abundant in the testis and the ovary. They also found that high level <italic>Brca1</italic> mRNA expression in the testis of mice was detected in meiotic cells and postmeiotic round spermatids and, in contrast, little or no <italic>Brca1</italic> mRNA was expressed in premeiotic germ cells. A low level of <italic>Brca1</italic> mRNA was also detected in Sertoli cells. Blackshear <italic>et al</italic> [<xref ref-type="bibr" rid="B5">5</xref>], on the contrary, demonstrated in the mouse that <italic>Brca1</italic> and <italic>Brca2</italic> mRNA are expressed in mitotic spermatogonia in addition to early meiotic prophase spermatocytes; Sertoli cells and Leydig interstitial cells were found consistently negative for <italic>Brca1</italic> and <italic>Brca2</italic> transcripts. In the normal mouse adult ovary, <italic>Brca1</italic> and <italic>Brca2</italic> transcripts were localized specifically to granulosa cells, thecal cells and oocytes of developing follicles as well as luteal cells of recently formed corpora lutea and surface epithelium.</p><p>Considering these results, we further investigated the presence of human BRCA1 and BRCA2 proteins in an ovotestis by immunochemical analysis with a different panel of antibodies against BRCA1 and BRCA2.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><p>True hermaphroditism is a rare cause of atypical genitalia that presents significant diagnostic and management challenges. The patient (male, 6 months old, karyotype of 46, XX; analysis of a sex-determining region of the Y chromosome [SRY] was negative) had a testis on the left side and an ovotestis on the right side. Hematoxylin eosin saffron (HES) demonstrated male and female compartments of the ovotestis (Fig. <xref ref-type="fig" rid="F1">1a</xref>). This paper presents our laboratory findings concerning the BRCA1 and BRCA2 protein expression in this particular gonad.</p><p>All antibodies are described in Table <xref ref-type="table" rid="T1">1</xref>.</p></sec><sec><title>Results and discussion</title><p>The specificity of the polyclonal antibody against BRCA1 (K-18) has been demonstrated elsewhere for BRCA1 [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>], ascertained by Western blotting, and the 220 kDa band corresponding to BRCA1 was detected in both HBL-100 and MCF-7 breast cell lines. Moreover, other major bands appeared around 100 kDa in the two cell lines, which may correspond to different variants to BRCA1 [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>].</p><p>Chen <italic>et al</italic> [<xref ref-type="bibr" rid="B9">9</xref>] previously reported that the monoclonal antibodies anti-BRCA1 8F7 and 17F8 always exhibited a nuclear staining pattern in breast epithelial cells. Lee <italic>et al</italic> [<xref ref-type="bibr" rid="B10">10</xref>] had already reported that nuclear staining of the BRCA1 protein with 17F8 monoclonal antibody was also seen in tumor cells that were not of breast or ovarian origin. Wilson <italic>et al</italic> [<xref ref-type="bibr" rid="B11">11</xref>] also reported recently that 17F8 antibody detected BRCA1 nuclear and cytoplasmic stainings in breast specimens according to antibody concentration. We have also reported a nuclear staining pattern with different tissues, notably in child mammary gland [<xref ref-type="bibr" rid="B12">12</xref>] and in lung from a fetus at 19 weeks' gestation [<xref ref-type="bibr" rid="B13">13</xref>], with the 8F7 and 17F8 anti-BRCA1 monoclonal antibodies raised against a glutathione-<italic>S</italic>-transferase (GST)-BRCA1 fusion protein containing amino acids encoded by a 3&#x000b4; portion of BRCA1 exon 11 and by a 5&#x000b4; portion of BRCA1 exon 11, respectively. These antibodies seemed to remain more specific, and they exhibited nuclear staining.</p><p>We used anti-BRCA1 antibodies raised against amino acids 768-793 (66046N) and amino acids 1847-1863 (66056N). Characterization of these antibodies in MCF-7 human breast tumor cell lysates by Western blotting (data not shown) resulted in the detection of a 220 kDa band corresponding to BRCA1, and one other band around 100 kDa that may correspond to the BRCA1 protein missing exon 11. The 66036N antibodies elicited against amino acids 2-20 of human BRCA1 recognized a 220 kDa BRCA1 protein in HBL-100 breast cell lysates [<xref ref-type="bibr" rid="B14">14</xref>].</p><p>The antibody for BRCA2 (66066E) recognizes epitopes between amino acids 1323-1346 of human BRCA2. The antibody 66076E recognizes epitopes between amino acids 2586-2600. Antibodies were purchased from PharMingen (San Diego, CA, USA) and tested by Western blotting in HBL-100 human breast cells to ensure they recognized the 390 kDa BRCA2 protein. Both BRCA2 antibodies also cross-reacted with smaller proteins, which could be degradation products.</p><p>The 3E6 and 5F6 anti-BRCA2 monoclonal antibodies were generated using two bacterially expressed and purified GST-BRCA2 fusion proteins, containing amino acids 188-563 and 2336-2478 as antigens, respectively [<xref ref-type="bibr" rid="B15">15</xref>]. Using the 3E6 antibodies, we detected a protein of 390 kDa, in MCF-7 human tumor breast cell lysates, which corresponds to the predicted size of the 3418 amino acid BRCA2 sequence. We also detected, in CCL 221 colorectal adenocarcinoma cell lysates, a single band at 390 kDa using the 5F6 antibodies (data not shown).</p><p>As shown in Fig. <xref ref-type="fig" rid="F1">1</xref> and Table <xref ref-type="table" rid="T1">1</xref> for the BRCA1 protein expression study with K-18 antibodies, only cytoplasmic stainings in male germ cells and in oocytes were obtained (Fig. <xref ref-type="fig" rid="F1">1b</xref>). Predominant nuclear stainings of Sertoli cells and oocytes were obtained with 8F7 antibodies (Fig. <xref ref-type="fig" rid="F1">1c</xref>), and cytoplasmic staining of oocytes was also exhibited. Staining patterns with 17F8 antibodies varied from exclusively nuclear staining in Sertoli cells (Fig. <xref ref-type="fig" rid="F1">1d</xref>) to exclusively cytoplasmic staining in oocytes (Fig. <xref ref-type="fig" rid="F1">1e</xref>). Exclusively cytoplasmic staining was seen in male and female germ cells with 66046N (Fig. <xref ref-type="fig" rid="F1">1f</xref>) and 66056N antibodies (data not shown). In contrast, cytoplasmic staining with 66036N antibodies (Fig. <xref ref-type="fig" rid="F1">1g</xref>) was exhibited in male and female germ cells, and low nuclear staining was found in Sertoli cells.</p><p>For BRCA2 protein expression in Sertoli cells and oocytes, exclusively cytoplasmic staining was seen with 66066E antibodies (Fig. <xref ref-type="fig" rid="F1">1h</xref>). Cytoplasmic staining (data not shown) was exhibited with 66076E antibodies in oocytes and in male germ cells. Cytoplasmic staining was obtained in male and female compartments with 3E6 antibodies (Fig. <xref ref-type="fig" rid="F1">1i</xref>). With 5F6 antibodies (data not shown), exclusively cytoplasmic staining was obtained for BRCA2 protein in the male compartment, and low intensive nuclear and cytoplasmic stainings were obtained in oocytes and follicles.</p><p>The differences of staining patterns for the same protein in the same tissue may be explained by the choice of the antibodies. The monoclonal antibodies raised against GST-BRCA1 or GST-BRCA2 fusion proteins seem more specific than antibodies raised against a 20 amino acid peptide.</p></sec><sec><title>Conclusion</title><p>In conclusion, we show using different antibodies that BRCA1 proteins, like BRCA2, are widely expressed in two varieties of non-embryogenic human tissues associated with the cell cycle. BRCA1 and BRCA2 proteins are expressed during growth and differentiation in the ovary. Moreover, they are expressed beyond the spermatogenesis. This is consistent with proposed functions for <italic>BRCA1</italic> and <italic>BRCA2</italic> genes.</p></sec>

Transforming growth factors-&#x003b2; are not good biomarkers of chemopreventive efficacy in a preclinical breast cancer model system

<p>Using a carcinogen-initiated rat model of mammary tumorigenesis, we tested the hypothesis that transforming growth factor (TGF)-&#x003b2;s are useful biomarkers of chemopreventive efficacy in the breast. The chemopreventive agents tested were tamoxifen and the retinoids 9-<italic>cis</italic>-retinoic acid (9cRA) and <italic>N</italic>-(4-hydroxyphenyl)retinamide (4-HPR), because both antiestrogens and retinoids have previously been shown to upregulate TGF-&#x003b2;s <italic>in vitro</italic>. Despite demonstrable chemopreventive efficacy in this model, none of these agents, alone or in combination, had any significant impact on the expression of TGF-&#x003b2;s in the mammary ductal epithelium or periductal stroma as determined by immunohistochemistry. These data suggest that TGF-&#x003b2;s are not likely to be useful biomarkers of chemopreventive efficacy in a clinical setting.</p>

<contrib id="A1" contrib-type="author"><name><surname>Zujewski</surname><given-names>JoAnne</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Vaughn-Cooke</surname><given-names>Anika</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Flanders</surname><given-names>Kathleen C</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Eckhaus</surname><given-names>Michael A</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Lubet</surname><given-names>Ronald A</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Wakefield</surname><given-names>Lalage M</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>wakefiel@dce41.nci.nih.gov</email></contrib>

Breast Cancer Research

<sec><title>Synopsis</title><sec><title>Introduction:</title><p>Chemoprevention has been defined as the use of noncytotoxic nutrients or pharmacologic agents to enhance intrinsic physiologic mechanisms that protect the organism against the development of mutant clones and their progression to malignant cancer. In a recent landmark trial, tamoxifen, a hormonally active selective estrogen receptor modulator (SERM), was shown to decrease the risk of invasive breast cancer by 49% in asymptomatic, but at-risk women [<xref ref-type="bibr" rid="B1">1</xref>]. The search is now on for agents with improved risk-benefit profiles, and for agents that will prevent the subclass of estrogen receptor-negative tumors, the incidence of which was unaffected by the SERMS. Retinoids have already shown potential in this regard [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. Because it will not be possible to test many agents in large randomized clinical trials, efforts are underway to develop useful tissue-based surrogate end-point biomarkers that can be used to select only the most promising agents (and doses) for large-scale trials.</p><p>Provocative mechanistic connections have been made between the steroid hormone superfamily, including the SERMS and retinoids, and the TGF-&#x003b2; family of multifunctional growth factors [<xref ref-type="bibr" rid="B8">8</xref>]. The TGF-&#x003b2; system has tumor suppressor activity, and loss of TGF-&#x003b2; response is associated with advanced disease in many human tumor types, including the breast [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Conversely, experimental overexpression of TGF-&#x003b2; in the mammary gland protects against tumorigenesis [<xref ref-type="bibr" rid="B11">11</xref>]. This strongly suggests that interventions that enhance TGF-&#x003b2; function early in tumorigenesis could delay or prevent the course of the disease. SERMs such as tamoxifen can upregulate TGF-&#x003b2; production and activation by many cell types, including human breast cancer cell lines [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. Similarly, retinoids can upregulate TGF-&#x003b2; production and activation, both in cell culture and in rats <italic>in vivo</italic> [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. It is plausible, therefore, that upregulation of endogenous TGF-&#x003b2; could contribute to the chemopreventive efficacy of SERMs and retinoids.</p><p>In the present study we used a carcinogen-induced rat mammary carcinogenesis model to test the hypothesis that chemoprevention by tamoxifen and retinoids is associated with local upregulation of TGF-&#x003b2;s in the mammary gland, and that TGF-&#x003b2;s might therefore be useful as potential surrogate end-point biomarkers of chemopreventive efficacy in clinical trials.</p></sec><sec><title>Materials and methods:</title><p>A standard protocol for induction of breast cancer in female Sprague-Dawley rats using a single dose of <italic>N</italic>-nitroso-<italic>N</italic>-methylurea (NMU) at 8 weeks of age was used [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. Chemopreventive agents were incorporated into powdered lab chow [<xref ref-type="bibr" rid="B18">18</xref>] and fed <italic>ad libitum</italic>, beginning 1 week after injection with NMU. The rats were fed 9cRA (Kuraray Company, Osaka, Japan) at 120 mg/kg of diet, tamoxifen (Sigma Chemical Co, St Louis, MO, USA) at 1.0 mg/kg of diet, and 4-HPR (RW Johnson Pharmaceutical Research Unit, Spring House, PA, USA) at 782 mg/kg of diet.</p><p>Rats were weighed and palpated for the presence of mammary tumors weekly, and six rats in each experimental group were sacrificed after 6 and 12 weeks of treatment with chemopreventive agent. For experiments to determine the effect of high doses of tamoxifen administered over shorter periods of time, rats were given 10 mg tamoxifen/kg body weight per day intragastrically, or 1 mg tamoxifen/kg in the diet, and were sacrificed after 1 day or 3 weeks of treatment. All palpated tumors were confirmed at necropsy, and mammary glands were fixed in neutral buffered formalin and embedded in paraffin. The number 2 (first thoracic) mammary gland was sectioned for histology and immunohistochemistry.</p><p>Immunohistochemical staining was done using rabbit polyclonal antibodies raised against synthetic peptides that correspond to regions in the mature forms of TGF-&#x003b2;₁, TGF-&#x003b2;₂ and TGF-&#x003b2;₃: anti-TGF-&#x003b2;₁-LC and anti-TGF-&#x003b2;₁-CC [<xref ref-type="bibr" rid="B19">19</xref>], anti-TGF-&#x003b2;₂ (sc-90; Santa Cruz Biotechnologies Inc, Santa Cruz, CA, USA), and anti-50-60-&#x003b2;₃-LC [<xref ref-type="bibr" rid="B20">20</xref>], respectively. Anti-latent TGF-&#x003b2;-binding protein (LTBP; Ab39) was raised against the purified full-length platelet LTBP [<xref ref-type="bibr" rid="B21">21</xref>]. The antibodies were affinity purified against the immunizing peptide (anti-TGF-&#x003b2;₃) or against protein A sepharose (anti-TGF-&#x003b2;₁-LC, anti-TGF-&#x003b2;₁-CC and anti-TGF-&#x003b2;₂). Immunohistochemical staining was performed using an indirect immunoperoxidase detection protocol (Vectastain Elite kit, Vector Laboratories, Burlingame, CA, USA). Staining intensity was scored on a scale of 0-4+, using the mouse embryo control section as a reference standard for each run. Ducts and periductal stroma were scored independently. Staining was scored in a blinded manner by two independent observers, and discrepancies were rescored by consensus. Staining intensity was plotted as the mean &#x000b1; standard deviation for each experimental group.</p></sec><sec><title>Results:</title><p>Palpable mammary tumors were first detected after approximately 35 days following initiation of NMU, and by 70 days incidence had reached 100% in rats not treated with chemopreventive agents (Fig. <xref ref-type="fig" rid="F1">1a</xref>). Tamoxifen, alone or in combination with retinoids, decreased tumor incidence by more than 70% by the end of the study, whereas 9cRA alone decreased it by 50%. 4-HPR alone had a relatively modest effect on tumor incidence in the present study. However, it significantly decreased tumor multiplicity (Fig. <xref ref-type="fig" rid="F1">1b</xref>), indicating that the dose used was efficacious. There was minimal toxicity associated with the chemopreventive intervention, except in the tamoxifen + 4-HPR group, in which mild toxicity was observed, as judged by the weights of the experimental animals (Fig. <xref ref-type="fig" rid="F1">1c</xref>).</p><p>All three TGF-&#x003b2; isoforms and the LTBP (part of the naturally occurring latent TGF-&#x003b2; complex) showed broadly similar immunostaining patterns in the mammary glands of untreated rats at 15 weeks of age (Fig. <xref ref-type="fig" rid="F2">2</xref>). They were present both in the ductal epithelium and in the periductal stroma, suggesting that the TGF-&#x003b2;s are synthesized by the epithelial cells, and possibly stromal cells, and are sequestered in the extracellular matrix. This staining pattern is consistent with a role for the TGF-&#x003b2;s in the maintenance of normal mammary homeostasis.</p><p>None of the chemopreventive agents used, alone or in combination, were found to affect expression of any of the TGF-&#x003b2; isoforms or the LTBP in either ductal epithelium or periductal stroma after 6 weeks of chemopreventive intervention (Fig. <xref ref-type="fig" rid="F3">3</xref>). The 6-week time point was chosen as representative of the period of preneoplasia, as the majority of the animals had no palpable tumors at this time (Fig. <xref ref-type="fig" rid="F1">1</xref>). In the study set, eight out of 36 (22%) of the slides showed histologic evidence of hyperplasia, one out of 36 had a ductal carcinoma <italic>in situ</italic> (mammary intraepithelial neoplasia [<xref ref-type="bibr" rid="B22">22</xref>]), and one out of 36 had a carcinoma. We further investigated the effect of tamoxifen at higher doses and earlier time points. In rats that received tamoxifen at 10 mg/kg per day intragastrically (equivalent to 600 mg/day for a human) or 1 mg/kg per day intragastrically (equivalent to 60 mg/day for a human) for either 1 day or 3 weeks, again no consistent changes were seen in TGF-&#x003b2; expression, using either the TGF-&#x003b2;₁-CC or the TGF-&#x003b2;₂ antibodies (data not shown).</p><p>After 6 weeks of treatment, we noticed that mammary glands from tamoxifen-treated rats were less developed than those of untreated control animals, having fewer tertiary ducts and terminal end buds, and they could consistently be identified from a blind data set (Fig. <xref ref-type="fig" rid="F4">4</xref>). By 12 weeks of treatment, all three chemopreventive agents had a significant effect on glandular histology, with tamoxifen and 9cRA showing the greatest suppression of ductal development and lobule formation, and 4-HPR showing a relatively mild effect.</p></sec><sec><title>Discussion:</title><p>One major goal in the field of prevention is the identification of surrogate biomarkers that might rapidly predict the effect of a given agent on the primary end-point of cancer incidence. The most informative markers are those with modulation that is likely to be directly related to the preventive effect, and a compelling argument can be made that TGF-&#x003b2;s may fall into this category. However, the present data in a well-established preclinical model of breast cancer, employing a variety of highly effective chemopreventive regimens, suggest that this is not the case.</p><p>Most of the previous studies on the regulation of TGF-&#x003b2;s by tamoxifen and retinoids have been done in tissue culture [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. The lack of effect on TGF-&#x003b2; expression in the present <italic>in vivo</italic> study may reflect the dependence of the response on contextual cues that are only present in the artificial <italic>in vitro</italic> environment. In an <italic>in vivo</italic> study [<xref ref-type="bibr" rid="B16">16</xref>], all-<italic>trans</italic>-retinoic acid was shown to cause an upregulation of TGF-&#x003b2; isoforms in rats, with kinetics and isoform selectivity that varied with the target tissue. However, the rats were vitamin A-deficient, and it is not known whether the same effects would be seen in vitamin A-replete animals such as were used in the present study. In a small study in humans [<xref ref-type="bibr" rid="B23">23</xref>] tamoxifen treatment was shown to cause a consistent induction in extracellular TGF-&#x003b2; in breast cancer biopsies, when compared with pretreatment biopsies from the same patients, and complex effects of tamoxifen on induction of TGF-&#x003b2;₂ in the plasma of patients with metastatic breast cancer have been described [<xref ref-type="bibr" rid="B24">24</xref>]. It is possible that tamoxifen is only effective in inducing TGF-&#x003b2; in the context of a tumor, and not in the normal or initiated tissue that was the subject of the present study. However, an optimal surrogate end-point biomarker in a prevention setting needs to be modulated in normal or premalignant tissues. Although we cannot eliminate the possibility of more subtle effects of chemopreventive agents on TGF-&#x003b2; bioavailability or cellular responsiveness, in our preliminary analyses we have seen no effects on the expression of type I and type II TGF-&#x003b2; receptors (data not shown).</p><p>There is considerable evidence to suggest that, at late stages in tumorigenesis, TGF-&#x003b2;s can actually promote the tumorigenic process, particularly if the epithelial cells have lost responsiveness to the growth inhibitory effects of TGF-&#x003b2; by this time [<xref ref-type="bibr" rid="B9">9</xref>]. While the present work was in progress, a study was reported [<xref ref-type="bibr" rid="B25">25</xref>] that showed that loss of the type II TGF-&#x003b2; receptor can already be seen in a significant fraction of hyperplasias without atypia in the human breast. Furthermore, loss of the receptor correlated with increased risk of subsequent development of invasive breast cancer. Thus, loss of TGF-&#x003b2; response may be a very early event in the development of human breast cancer. Because locally elevated TGF-&#x003b2; levels could select for TGF-&#x003b2;-resistant cells, and because TGF-&#x003b2;s can have oncogenic effects on the stroma, it may actually be important for the safety profile of chemopreventive agents to demonstrate that they do not increase TGF-&#x003b2; levels in the at-risk breast. In this regard, this demonstration that the expression of TGF-&#x003b2;s in the preclinical rat model is unaffected by tamoxifen, 9cRA, or 4-HPR may actually have positive implications, because all three agents are already in clinical use.</p><p>The NMU-induced rat model of mammary tumorigenesis is widely used for chemoprevention studies, and yields rapid development of hormonally responsive mammary tumors with 100% incidence [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. To do this, the initiating agent is given at 8 weeks of age and the chemopreventive agent is started a week later, during the period of active development of the mammary gland. We observed that the histology of the tamoxifen-treated mammary glands differed significantly from control glands when examined after 6 weeks of tamoxifen treatment, showing fewer terminal end-buds and less tertiary branching. Part of the chemopreventive efficacy of antiestrogens and retinoids in this model may therefore be due to a generalized decrease in ductal development. Since chemopreventive agents are unlikely to be given to humans during the pubertal period, this form of preclinical model may not accurately reflect the degree of chemopreventive benefit that could be achieved in humans. Although the accelerated time course and high penetrance of disease reduces the costs of this model, it may be advisable to confirm efficacy of promising agents in a model that delays application of the chemopreventive agent until the mammary gland is fully developed.</p><p>In conclusion, we have shown that treatment of rats with tamoxifen or retinoids results in effective chemoprevention of mammary tumorigenesis, without any detectable effect on local expression of TGF-&#x003b2;s. Although we cannot rule out more subtle effects on TGF-&#x003b2; activity, such as the activation of latent forms, the data suggest that TGF-&#x003b2;s are not involved in the underlying molecular mechanism of chemoprevention induced by these agents. This agrees with <italic>in vitro</italic> work [<xref ref-type="bibr" rid="B26">26</xref>] that showed that blockade of TGF-&#x003b2; signaling did not abrogate the growth inhibitory effect of tamoxifen on breast cancer cells. Given the very limited breast tissue available in clinical trials, we do not recommend testing for TGF-&#x003b2;s as a surrogate end-point biomarkers at this time.</p></sec></sec><sec><title>Full article</title><sec><title>Introduction</title><p>Chemoprevention has been defined as the use of noncytotoxic nutrients or pharmacologic agents to enhance intrinsic physiologic mechanisms that protect the organism against the development of mutant clones and their progression to malignant cancer [<xref ref-type="bibr" rid="B27">27</xref>]. Members of the nuclear receptor superfamily are considered to be particularly promising targets for chemoprevention, because of their pivotal role in the regulation of metabolic, developmental, and differentiation pathways [<xref ref-type="bibr" rid="B28">28</xref>]. In a recent landmark trial [<xref ref-type="bibr" rid="B1">1</xref>], tamoxifen, a hormonally active SERM, was shown to decrease the risk of invasive breast cancer by 49% in asymptomatic, but at-risk women. Another SERM, raloxifene, also shows promise [<xref ref-type="bibr" rid="B29">29</xref>]. These studies validate the concept of using pharmacologic agents for prevention of human breast cancer in apparently healthy individuals.</p><p>The search is now on for agents with improved risk-benefit profiles, and for agents that will prevent the subclass of estrogen receptor-negative tumors, the incidence of which was unaffected by the SERMS. Retinoids, a family of compounds structurally related to vitamin A, have already shown potential in this regard [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. Since it will not be possible to test many agents in large randomized clinical trials, efforts are underway to develop useful tissue-based surrogate end-point biomarkers that can be used to select only the most promising agents (and doses) for large-scale trials.</p><p>Provocative mechanistic connections have been made between the steroid hormone superfamily, including the SERMS and retinoids, and the TGF-&#x003b2; family of multifunctional growth factors. TGF-&#x003b2;s are potent inhibitors of the growth of many epithelial cell types [<xref ref-type="bibr" rid="B8">8</xref>]. Recent work has implicated the TGF-&#x003b2; system as an important tumor suppressor pathway, and loss of TGF-&#x003b2; response is associated with advanced disease in many human tumor types, including the breast [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>]. In mouse models, over-expression of TGF-&#x003b2;₁ in the mammary gland protects against tumorigenesis [<xref ref-type="bibr" rid="B11">11</xref>], whereas local inactivation of the type II TGF-&#x003b2; receptor enhances tumorigenesis [<xref ref-type="bibr" rid="B32">32</xref>]. This strongly suggests that interventions that enhance TGF-&#x003b2; function early in tumorigenesis could delay or prevent the course of the disease.</p><p>Antiestrogens such as tamoxifen have been shown to upregulate TGF-&#x003b2; production and activation by many cell types, including human breast cancer cell lines [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. Similarly, retinoids can upregulate TGF-&#x003b2; production and activation, both in cell culture and in rats <italic>in vivo</italic> [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. Therefore, it is reasonable to propose that some of the chemopreventive efficacy of these agents against breast cancer <italic>in vivo</italic> could be mediated via a local upregulation of TGF-&#x003b2;s, with concomitant enhancement of tumor suppressor activity.</p><p>In the present study, we used a carcinogen-induced rat model of mammary carcinogenesis to test whether chemoprevention by tamoxifen and by two different retinoids (4-HPR, also known as fenretinide; and 9-cRA) is associated with local upregulation of TGF-&#x003b2;s in the initiated mammary gland. If this were the case, TGF-&#x003b2;s might be useful as potential surrogate end-point biomarkers in clinical trials. However, the results show that TGF-&#x003b2; levels, as detected immunohistochemically, are not affected by tamoxifen or retinoids in this preclinical model of early-stage breast cancer.</p></sec></sec><sec><title>Materials and method</title><sec><title>Mammary carcinogenesis studies</title><p>A standard protocol for induction of breast cancer in female Sprague-Dawley rats was used [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B18">18</xref>], with initiation induced by a single intravenous dose of NMU (50 mg/kg body weight) at 8 weeks of age. Retinoids and tamoxifen were incorporated into powdered laboratory chow as described previously [<xref ref-type="bibr" rid="B18">18</xref>] and fed <italic>ad libitum</italic>, beginning 1 week after injection with NMU. Rats were fed 9cRA (Kuraray Company, Osaka, Japan) at 120 mg/kg of diet, tamoxifen (Sigma Chemical Co, St Louis, MO, USA) at 1.0 mg/kg of diet, and 4-HPR (RW Johnson Pharmaceutical Research Institute) at 782 mg/kg of diet.</p><p>Rats were weighed weekly and palpated for the presence of mammary tumors. Six rats in each experimental group were sacrificed after 6 and 12 weeks of treatment with chemopreventive agent. The 6-week sacrifice time was chosen for the immunohistochemical studies to represent the period of premalignancy, because the incidence of palpable tumors is less than 20% for all experimental groups at that time. By 12 weeks all rats that have not received a chemopreventive agent have tumors, so the primary purpose of the 12-week sacrifice time was to allow an accurate determination of chemopreventive efficacy for the particular experiment.</p><p>For experiments to determine the effect of high doses of tamoxifen administered over shorter periods of time, rats were given 10 mg tamoxifen/kg body weight per day intragastrically or 1 mg tamoxifen/kg in the diet, and were sacrificed after 1 day or 3 weeks of treatment.</p><p>All palpated tumors were confirmed at necropsy, and mammary glands were fixed in neutral buffered formalin and embedded in paraffin. The number 2 (first thoracic) mammary gland was sectioned for histology and immunohistochemistry.</p></sec><sec><title>Immunohistochemistry of TGF-&#x003b2;s</title><p>Immunohistochemical staining was done using rabbit polyclonal antibodies raised against synthetic peptides that correspond to regions in the mature forms of TGF-&#x003b2;₁, TGF-&#x003b2;₂, and TGF-&#x003b2;₃. Antibodies to TGF-&#x003b2;₁ were raised against a synthetic peptide corresponding to residues 1-30 of the mature protein in either the Laboratory of Chemoprevention (anti-TGF-&#x003b2;₁-LC; NIH, Bethesda, MD, USA) or the Collagen Corporation (anti-TGF-&#x003b2;₁-CC; Palo Alto, CA, USA). These antibodies were raised against different preparations of the 1-30 peptide, and they recognize different epitopes of this peptide [<xref ref-type="bibr" rid="B19">19</xref>]. The LC antibody usually stains intracellular TGF-&#x003b2;₁, whereas the CC antibody stains extracellular TGF-&#x003b2;₁. Anti-TGF-&#x003b2;₂ (sc-90; Santa Cruz Biotechnologies Inc) was raised to a peptide corresponding to residues 72-99 of the mature TGF-&#x003b2;₂. Anti-TGF-&#x003b2;₃ (anti-50-60-&#x003b2;₃-LC) was raised against residues 50-60 of mature TGF-&#x003b2;₃ [<xref ref-type="bibr" rid="B20">20</xref>]. Anti-LTBP (Ab39) was raised against the purified full-length platelet LTBP [<xref ref-type="bibr" rid="B21">21</xref>].</p><p>The antibodies were affinity purified against the immunizing peptide (anti-TGF-&#x003b2;₃) or against protein A sepharose (anti-TGF-&#x003b2;₁-LC, anti-TGF-&#x003b2;₁-CC, and anti-TGF-&#x003b2;₂), and have been assayed for specificity by Western blot analysis [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B33">33</xref>]. All antibodies reacted with the appropriate TGF-&#x003b2; isoform except anti-TGF-&#x003b2;₁-CC, which showed some cross-reactivity with TGF-&#x003b2;₃ on Western blots.</p><p>Immunohistochemical staining was performed using an indirect immunoperoxidase detection protocol (Vectastain Elite kit; Vector Laboratories) following treatment of sections with hyaluronidase to improve antibody penetration. Optimal antibody concentrations were determined by titration on select samples before analysis of the full experimental set. Staining was shown to be specific in control experiments in which either the primary antibody was preincubated with a 50-fold molar excess of immunizing peptide before being applied to the section (anti-TGF-&#x003b2;₂, anti-TGF-&#x003b2;₃), or the section was stained with an equivalent concentration of nonimmune rabbit immunoglobulin (anti-TGF-&#x003b2;₁-LC, anti-TGF-&#x003b2;₁-CC, and anti-LTBP). In analysis of the full experimental set, for any given antibody all sections were stained at the same time so as to be directly comparable, and a normal mouse embryo section was included as a positive control. A normal rabbit immunoglobulin control was also run for the whole set.</p></sec><sec><title>Quantitation of immunostaining</title><p>Two different systems were used to grade immunostaining. For all samples, staining of the ducts and periductal stroma were scored independently. For samples after 6 weeks of chemopreventive treatment, staining intensity was scored on a scale of 0&#x02013;4+, using the mouse embryo control section as a reference standard for each run. Staining was scored in a blinded manner by two independent observers, and scores never differed by more than one point. Discrepancies were rescored by consensus. Staining intensity was plotted as the mean &#x000b1; standard deviation for each experimental group.</p></sec></sec><sec><title>Results</title><sec><title>Chemopreventive efficacy</title><p>Palpable mammary tumors were first detected after approximately 35 days following initation with NMU, and by 70 days incidence had reached 100% in rats not treated with chemopreventive agents (Fig. <xref ref-type="fig" rid="F1">1a</xref>). Tamoxifen, alone or in combination with retinoids, decreased tumor incidence by more than 70% by the end of the study, whereas 9cRA alone decreased it by 50%. 4-HPR alone had a relatively modest effect on tumor incidence in the present study. However, it significantly decreased tumor multiplicity (Fig. <xref ref-type="fig" rid="F1">1b</xref>) indicating that the dose used was efficacious. These incidence and multiplicity data are consistent with previously published studies using this model [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B34">34</xref>]. The mean weights of the rats in the various experimental groups differed by less than 10% from those of rats not treated with a chemopreventive agent, except for the rats treated with a combination of tamoxifen and 4-HPR, in which the mean weight was approximately 15% lower than the that in untreated group by the end of the study (Fig. <xref ref-type="fig" rid="F1">1c</xref>). This suggests that there was minimal toxicity associated with the chemopreventive intervention, except in the tamoxifen + 4-HPR group, in which mild toxicity was observed.</p></sec><sec><title>Effect of chemopreventive agents on TGF-&#x003b2; expression in initiated mammary gland</title><p>Figure <xref ref-type="fig" rid="F2">2</xref> shows the typical immunohistochemical staining pattern for the TGF-&#x003b2;s and the LTBP (part of the naturally occurring latent TGF-&#x003b2; complex) in initiated mammary glands of 15-week-old rats that had not been treated with chemopreventive agents. All three TGF-&#x003b2; isoforms and the LTBP showed broadly similar staining patterns. They were present both in the ductal epithelium and in the periductal stroma, suggesting that the TGF-&#x003b2;s are synthesized by the epithelial cells, and possibly stromal cells, and are sequestered in the extracellular matrix. This staining pattern is consistent with a role for the TGF-&#x003b2;s in the maintenance of normal mammary homeostasis.</p><p>None of the chemopreventive agents used, alone or in combination, were found to affect expression of any of the TGF-&#x003b2; isoforms or the LTBP in either ductal epithelium or periductal stroma after 6 weeks of chemopreventive intervention (Fig. <xref ref-type="fig" rid="F3">3</xref>). The 6-week time point was chosen as representative of the period of preneoplasia, because the majority (&#x0003e;80%) of the rats had no palpable tumors at this stage (Fig. <xref ref-type="fig" rid="F1">1</xref>). Human clinical material for biomarker analysis in primary chemoprevention studies is also likely to comprise normal and initiated, at-risk epithelium with some early preneoplastic changes, but without evidence of major neoplastic change.</p><p>In the 6-week study set, eight out of 36 (22%) of the glands sampled showed histologic evidence of hyperplasia. In addition, one out of 36 glands sampled had a lesion with the appearance of a ductal carcinoma <italic>in situ</italic> (focal necrosis, marked atypia and abundant mitoses, placing it in the category of 'mammary intraepithelial neoplasia' by the Annapolis naming convention [<xref ref-type="bibr" rid="B22">22</xref>]), and an additional one had an invasive carcinoma. Both the <italic>in situ</italic> and the invasive carcinomas were in the control group. However, both samples also had histologically normal ducts on the same slide, which were scored for the analysis. There was no difference in staining between the ducts that were proximal to the tumor and those that were more distal, and neither were there any differences in staining observed between histologically normal and hyperplastic ducts in any of the samples analyzed (data not shown). Since the focus of the present study was on TGF-&#x003b2; expression changes in the preneoplastic gland, staining of the tumors was not scored for the analysis, but in the two cases present, the staining did not differ significantly from that of the surrounding normal-appearing ducts (not shown).</p><p>Assuming that a rat weighs approximately 250 g and eats 10 g of chow each day, the equivalent human doses to those used in this study would be approximately 2.5 mg/day tamoxifen, 2 g/day 4-HPR, and 290 mg/day 9cRA. The human doses that are currently being used in clinical trials are 20 mg/day for tamoxifen, 0.2-0.4 g/day for 4-HPR, and 100-250 mg/day for 9cRA. The different doses of tamoxifen and 4-HPR used in humans and rats probably reflect interspecies differences in pharmacokinetics. However, because the efficacious dose for tamoxifen was 10 times lower in the rat than the dose that has previously been shown to upregulate TGF-&#x003b2;₁ expression in human breast tumor tissue [<xref ref-type="bibr" rid="B23">23</xref>], we looked at the effect of a higher dose of tamoxifen in the rats. Furthermore, because one study [<xref ref-type="bibr" rid="B35">35</xref>] showed that TGF-&#x003b2; may be transiently upregulated early after administration of tamoxifen, we also looked at earlier time points. In rats receiving tamoxifen at 10 mg/kg per day intragastrically (equivalent to 600 mg/day for a human) or 1 mg/kg per day (equivalent to 60 mg/day for a human) for either 1 day or 3 weeks, again no consistent changes were seen in TGF-&#x003b2; expression, using either the TGF-&#x003b2;₁-CC or the TGF-&#x003b2;₂ antibodies (data not shown).</p></sec><sec><title>Effect of chemopreventive agents on the histology of the mammary gland</title><p>While scoring the immunohistochemical slides, we noticed that the mammary histology appeared to be altered in rats treated with chemopreventive agents. Hematoxylin and eosin stained sections from the samples after 6 weeks of treatment were analyzed by a veterinary pathologist (MAE). The mammary glands from tamoxifen-treated rats were less developed than those of untreated control animals, having fewer tertiary ducts and terminal end-buds, and could consistently be identified from a blind data set (Fig. <xref ref-type="fig" rid="F4">4</xref>). By 12 weeks of treatment, all three chemopreventive agents had a significant effect on glandular histology, with tamoxifen and 9cRA showing the greatest suppression of ductal development and lobule formation, and 4-HPR showing a relatively mild effect.</p></sec></sec><sec><title>Discussion</title><sec><title>TGF-&#x003b2;s as candidate biomarkers</title><p>One major goal in the field of prevention is the identification of surrogate biomarkers that might rapidly predict the effect of a given agent on the primary end-point of cancer incidence. There is an extensive literature showing that steroid hormone superfamily members, such as antiestrogens and retinoids, can upregulate TGF-&#x003b2; activity in a variety of systems [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B35">35</xref>]. This suggested that the chemopreventive action of these agents against breast cancer could be mediated in part through enhancing the tumor suppressor activity of the endogenous TGF-&#x003b2; system, and thus that changes in TGF-&#x003b2; expression might serve as useful surrogate end-point biomarkers of chemopreventive efficacy. However, here we used the NMU-induced rat model of mammary carcinogenesis to show that the chemopreventive effect of tamoxifen and two retinoids is not associated with any consistent changes in TGF-&#x003b2; levels, at least as determined immunohistochemically.</p></sec><sec><title>Comparison with earlier studies</title><p>Most of the previous studies on the regulation of TGF-&#x003b2;s by tamoxifen and retinoids [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B17">17</xref>] have been done in tissue culture. Thus, the lack of effect on TGF-&#x003b2; expression in the present <italic>in vivo</italic> study might reflect the dependence of the response on contextual cues that are only present in the artificial <italic>in vitro</italic> environment. In one previous <italic>in vivo</italic> study [<xref ref-type="bibr" rid="B16">16</xref>] all-<italic>trans</italic>-retinoic acid upregulated TGF-&#x003b2; expression in rats, with kinetics and isoform selectivity that varied with the target tissue. However, the rats were vitamin A-deficient, and it is not known whether the same effects would be seen in vitamin A-replete animals, such as were used in the present study, or whether the response would vary with the specific retinoid used.</p><p>There are some data that support an effect of tamoxifen on upregulation of TGF-&#x003b2;s <italic>in vivo</italic> in humans. Three months of tamoxifen treatment was shown to cause a consistent induction in extracellular TGF-&#x003b2; in breast cancer biopsies, when compared with pretreatment biopsies from the same patients [<xref ref-type="bibr" rid="B23">23</xref>]. Furthermore, complex effects of tamoxifen on induction of TGF-&#x003b2;₂ in the plasma of patients with metastatic breast cancer have been described [<xref ref-type="bibr" rid="B24">24</xref>]. It is possible that tamoxifen is only effective in inducing TGF-&#x003b2; in the context of a tumor, and not in normal or initiated tissue, which was the subject of the present study. This issue could be reassessed in preclinical models using the same agents to treat established proliferative intraepithelial neoplasia. However, for ease of tissue acquisition, an optimal surrogate end-point biomarker in a prevention setting needs to be modulated in normal or premalignant tissues.</p></sec><sec><title>Alternative levels of regulation of the bioefficacy of TGF-&#x003b2;s</title><p>Interestingly, in cell culture, both tamoxifen and all-<italic>trans</italic>-retinoic acid have been shown [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B17">17</xref>] to increase the fraction of TGF-&#x003b2; in its biologically active, as opposed to its latent form. The current method for discriminating between active and latent TGF-&#x003b2; in tissue samples requires the use of frozen sections and immunofluorescence techniques, which are not practical for routine clinical use [<xref ref-type="bibr" rid="B36">36</xref>]. As simpler assays become available, however, the issue of possible changes in TGF-&#x003b2; activation status should be readdressed. Retinoids can also affect cellular responsiveness to TGF-&#x003b2;s at the level of receptor expression and downstream events [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>]. To date, expression of TGF-&#x003b2; receptors and downstream signaling components such as the Smads have not been well-characterized in this rat model, but in our preliminary analyses we saw no effect of retinoids on type I and type II TGF-&#x003b2; receptor expression in the mammary gland (data not shown). At this time, however, we certainly cannot rule out the possibility that tamoxifen and retinoids may be having subtle effects on the TGF-&#x003b2; system at levels other than the regulation of TGF-&#x003b2; expression.</p></sec><sec><title>Lack of effect of chemopreventive agents on TGF-&#x003b2; expression may have positive implications</title><p>There is considerable evidence to suggest that, at late stages in tumorigenesis, TGF-&#x003b2;s can actually promote the tumorigenic process, particularly if the epithelial cells have lost responsiveness to the growth regulatory effects of TGF-&#x003b2; by this time [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref>]. Thus, advanced human tumors show increased levels of TGF-&#x003b2; expression [<xref ref-type="bibr" rid="B42">42</xref>,<xref ref-type="bibr" rid="B43">43</xref>,<xref ref-type="bibr" rid="B44">44</xref>], and TGF-&#x003b2;s are known to suppress the immunosurveillance system, to enhance angiogenesis, invasion and metastasis, and to increase drug resistance [<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>,<xref ref-type="bibr" rid="B47">47</xref>,<xref ref-type="bibr" rid="B48">48</xref>].</p><p>In the colon, loss of the type II TGF-&#x003b2; receptor occurs at the late adenoma to carcinoma transition [<xref ref-type="bibr" rid="B49">49</xref>], suggesting that early premalignant lesions retain TGF-&#x003b2; responsiveness and would be amenable to interventions that enhance TGF-&#x003b2; activity. However, while the present work was in progress, a study was reported [<xref ref-type="bibr" rid="B25">25</xref>] showing that loss of the type II TGF-&#x003b2; receptor can already be seen in a significant fraction of hyperplasias without atypia in the human breast. Furthermore, loss of the receptor correlated with increased risk of subsequently developing invasive breast cancer. Thus, unlike in the colon, loss of TGF-&#x003b2; response may be a very early event in the development of human breast cancer.</p><p>Since locally elevated TGF-&#x003b2; levels may select for TGF-&#x003b2;-resistant cells, and because TGF-&#x003b2;s can have oncogenic effects on the stroma, it may actually be important for the safety profile of chemopreventive agents to demonstrate that they do not increase TGF-&#x003b2; levels in the at-risk breast. For example, tamoxifen resistance in a xenograft model of advanced human breast cancer, was recently shown [<xref ref-type="bibr" rid="B50">50</xref>] to be associated with an increase in TGF-&#x003b2;s and concomitant immunosuppressive effects on natural killer cells. In this regard, our demonstration that the expression of TGF-&#x003b2;s in the preclinical rat model is unaffected by tamoxifen, 9cRA, and 4-HPR may actually have positive implications, because these agents are already in clinical use.</p></sec><sec><title>Limitations of the NMU-induced rat model of mammary carcinogenesis</title><p>The NMU-induced rat model of mammary tumorigenesis is widely used for chemoprevention studies and yields rapid development of hormonally responsive mammary tumors with 100% incidence [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. To do this, the initiating agent is given at 8 weeks of age, during early puberty, and the chemopreventive agent is typically given continuously, starting 1 week later. Since sexual maturity is achieved at approximately 11 weeks of age in rats, this means that the chemopreventive agent is given during a period of active development of the mammary gland.</p><p>In the present study we observed that the histology of the tamoxifen-treated mammary gland differed significantly from control glands when examined after 6 weeks of tamoxifen treatment. Specifically, there were fewer terminal end-buds and less tertiary branching, which are indicative of a delay or arrest in normal mammary development. This is consistent with the known requirement for estrogen for proper mammary development [<xref ref-type="bibr" rid="B51">51</xref>]. We saw a lesser effect with 4-HPR, although this type of phenomenon has also been described in the literature for retinoids [<xref ref-type="bibr" rid="B4">4</xref>]. Thus, part of the chemopreventive efficacy of antiestrogens and retinoids in this model may be due to a generalized decrease in ductal development. Because chemopreventive agents are unlikely to be given to humans during the pubertal period, this form of preclinical model may not accurately reflect the degree of chemopreventive benefit that could be achieved in humans. Although the accelerated time course and high penetrance of disease reduces the costs of this model, it may be advisable to confirm efficacy of promising agents in a model that delays application of the chemopreventive agent until the mammary gland is fully developed.</p></sec></sec><sec><title>Conclusion</title><p>We have shown that treatment of rats with tamoxifen or retinoids results in effective chemoprevention of mammary tumorigenesis, without any detectable effect on local expression of TGF-&#x003b2;s. Although we cannot rule out more subtle effects on TGF-&#x003b2; activity, such as the activation of latent forms, the data suggest that the underlying molecular mechanism of chemoprevention by these agents does not involve increases in TGF-&#x003b2; expression. This agrees with <italic>in vitro</italic> work showing that blockade of TGF-&#x003b2; signalling did not abrogate the growth inhibitory effect of tamoxifen on breast cancer cells [<xref ref-type="bibr" rid="B26">26</xref>]. Given the very limited breast tissue available in clinical chemoprevention trials, we do not recommend testing for TGF-&#x003b2;s as surrogate end-point biomarkers at this time.</p></sec>

Smoking and high-risk mammographic parenchymal patterns: a case-control study

<sec><title>Introduction:</title><p>Overall, epidemiological studies [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>] have reported no substantial association between cigarette smoking and the risk of breast cancer. Some studies [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>] reported a significant increase of breast cancer risk among smokers. In recent studies that addressed the association between breast cancer and cigarette smoking, however, there was some suggestion of a decreased risk [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>], especially among current smokers, ranging from approximately 10 to 30% [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Brunet <italic>et al</italic> [<xref ref-type="bibr" rid="B11">11</xref>] reported that smoking might reduce the risk of breast cancer by 44% in carriers of <italic>BRCA1</italic> or <italic>BRCA2</italic> gene mutations. Wolfe [<xref ref-type="bibr" rid="B12">12</xref>] described four different mammographic patterns created by variations in the relative amounts of fat, epithelial and connective tissue in the breast, designated N1, P1, P2 and DY. Women with either P2 or DY pattern are considered at greater risk for breast cancer than those with N1 or P1 pattern [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. There are no published studies that assessed the relationship between smoking and mammographic parenchymal patterns.</p></sec><sec><title>Aims:</title><p>To evaluate whether mammographic parenchymal patterns as classified by Wolfe, which have been positively associated with breast cancer risk, are affected by smoking. In this case-control study, nested within the European Prospective Investigation on Cancer in Norfolk (EPIC-Norfolk) cohort [<xref ref-type="bibr" rid="B16">16</xref>], the association between smoking habits and mammographic parenchymal patterns are examined. The full results will be published elsewhere.</p></sec><sec><title>Methods:</title><p>Study subjects were members of the EPIC cohort in Norwich who also attended the prevalence screening round at the Norwich Breast Screening Centre between November 1989 and December 1997, and were free of breast cancer at that screening. Cases were defined as women with a P2/DY Wolfe's mammographic parenchymal pattern on the prevalence screen mammograms. A total of 203 women with P2/DY patterns were identified as cases and were individually matched by date of birth (within 1 year) and date of prevalence screening (within 3 months) with 203 women with N1/P1 patterns who served as control individuals.</p><p>Two views, the mediolateral and craniocaudal mammograms, of both breasts were independently reviewed by two of the authors (ES and RW) to determine the Wolfe mammographic parenchymal pattern.</p><p>Considerable information on health and lifestyle factors was available from the EPIC Health and Lifestyle Questionnaire [<xref ref-type="bibr" rid="B16">16</xref>]. In the present study we examined the subjects' personal history of benign breast diseases, menstrual and reproductive factors, oral contraception and hormone replacement therapy, smoking, and anthropometric information such as body mass index and waist:hip ratio.</p><p>Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated by conditional logistic regression [<xref ref-type="bibr" rid="B17">17</xref>], and were adjusted for possible confounding factors.</p></sec><sec><title>Results:</title><p>The characteristics of the cases and controls are presented in Table <xref ref-type="table" rid="T1">1</xref>. Cases were leaner than controls. A larger percentage of cases were nulliparous, premenopausal, current hormone replacement therapy users, had a personal history of benign breast diseases, and had had a hysterectomy. A larger proportion of controls had more than three births and were current smokers.</p><p>Table <xref ref-type="table" rid="T2">2</xref> shows the unadjusted and adjusted OR estimates for Wolfe's high-risk mammographic parenchymal patterns and smoking in the total study population and in postmenopausal women separately. Current smoking was strongly and inversely associated with high-risk patterns, after adjustment for concomitant risk factors. Relative to never smokers, current smokers were significantly less likely to have a high-risk pattern (OR 0.37, 95% CI 0.14-0.94). Similar results were obtained when the analysis was confined to postmenopausal women. Past smoking was not related to mammographic parenchymal patterns. The overall effect in postmenopausal women lost its significance when adjusted for other risk factors for P2/DY patterns that were found to be significant in the present study, although the results were still strongly suggestive. There was no interaction between cigarette smoking and body mass index.</p></sec><sec><title>Discussion:</title><p>In the present study we found a strong inverse relationship between current smoking and high-risk mammographic parenchymal patterns of breast tissue as classified by Wolfe [<xref ref-type="bibr" rid="B12">12</xref>]. These findings are not completely unprecedented; Greendale <italic>et al</italic> [<xref ref-type="bibr" rid="B18">18</xref>] found a reduced risk of breast density in association with smoking, although the magnitude of the reduction was unclear. The present findings suggest that this reduction is large.</p><p>Recent studies [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>] have suggested that breast cancer risk may be reduced among current smokers. In a multicentre Italian case-control study, Braga <italic>et al</italic> [<xref ref-type="bibr" rid="B10">10</xref>] found that, relative to nonsmokers, current smokers had a reduced risk of breast cancer (OR 0.84, 95% CI 0.7-1.0). These findings were recently supported by Gammon <italic>et al</italic> [<xref ref-type="bibr" rid="B9">9</xref>], who reported that breast cancer risk in younger women (younger than 45 years) may be reduced among current smokers who began smoking at an early age (OR 0.59, 95% CI 0.41-0.85 for age 15 years or younger) and among long-term smokers (OR 0.70, 95% CI 0.52-0.94 for those who had smoked for 21 years or more).</p><p>The possible protective effect of smoking might be due to its anti-oestrogenic effect [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Recently there has been renewed interest in the potential effect of smoking on breast cancer risk, and whether individuals may respond differently on the basis of differences in metabolism of bioproducts of smoking [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. Different relationships between smoking and breast cancer risk have been suggested that are dependent on the rapid or slow status of acetylators of aromatic amines [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. More recent studies [<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>], however, do not support these findings.</p><p>The present study design minimized the opportunity for bias to influence the findings. Because subjects were unaware of their own case-control status, the possibility of recall bias in reporting smoking status was minimized. Systematic error in the assessment of mammograms was avoided because reading was done without knowledge of the risk factor data. Furthermore, the associations observed are unlikely to be explained by the confounding effect of other known breast cancer risk factors, because we adjusted for these in the analysis. We did not have information on passive smoking status, however, which has recently been reported to be a possible confounder [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B24">24</xref>].</p><p>The present data indicate that adjustment for current smoking status is important when evaluating the relationship between mammographic parenchymal pattern and breast cancer risk. They also indicate smoking as a prominent potential confounder when analyzing effects of other risk factors such as obesity-related variables. It seems that parenchymal patterns may act as an informative biomarker of the effect of cigarette smoking on breast cancer risk.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Sala</surname><given-names>Evis</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>evis.sala@srl.cam.ac.uk</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Warren</surname><given-names>Ruth</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>McCann</surname><given-names>Jenny</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Duffy</surname><given-names>Stephen</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Luben </surname><given-names>Robert</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Day</surname><given-names>Nicholas</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I3">3</xref></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>Overall, epidemiological studies [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>] have reported no substantial association between cigarette smoking and the risk of breast cancer. Some studies [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>] reported a significant increase of breast cancer risk among smokers. It has been suggested [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B24">24</xref>] that passive exposure to cigarette smoking may alter prior associations seen when only active smoking was assessed, with increased risk being observed for passive smoking exposure. Furthermore, there is a possibility of heterogeneity in the response to the carcinogenic effect of smoking, which might explain inconsistent findings for cigarette smoking as a risk factor for breast cancer [<xref ref-type="bibr" rid="B20">20</xref>].</p><p>In recent studies that addressed the association between breast cancer and cigarette smoking, however, there was some suggestion of a decreased risk [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>], especially among current smokers, ranging from approximately 10 to 30% [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Brunet <italic>et al</italic> [<xref ref-type="bibr" rid="B11">11</xref>] reported that smoking might reduce the risk of breast cancer by 44% in carriers of <italic>BRCA1</italic> or <italic>BRCA2</italic> gene mutations.</p><p>Wolfe [<xref ref-type="bibr" rid="B21">21</xref>] described four different mammographic patterns that are created by variations in the relative amounts of fat, epithelial and connective tissue in the breast, designated N1, P1, P2 and DY. Women with either P2 or DY patterns are considered to be at greater risk for breast cancer than those with N1 or P1 pattern [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>].</p><p>There are no published studies that assessed the relationship between smoking and mammographic parenchymal patterns.</p><p>The aim of the present study was to evaluate whether mammographic parenchymal patterns as classified by Wolfe [<xref ref-type="bibr" rid="B12">12</xref>], which have been positively associated with breast cancer risk, are affected by smoking. In the present case-control study, nested within the European Prospective Investigation on Cancer in Norfolk (EPIC-Norfolk) cohort [<xref ref-type="bibr" rid="B16">16</xref>], the association between smoking habits and mammographic parenchymal patterns are examined. The full results will be published elsewhere.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><p>Study subjects were members of the EPIC cohort in Norwich [<xref ref-type="bibr" rid="B16">16</xref>], who also attended the prevalence screening round at the Norwich Breast Screening Centre between November 1989 and December 1997 and were free of breast cancer at that screening. A case-control study was designed within this cohort.</p><p>Cases were defined as women with a P2/DY Wolfe's mammographic parenchymal pattern on the prevalence screen mammogram. Assuming a 2.5-fold increase in risk of P2/DY mammographic patterns from the lowest quintile of a quantitative factor to the highest, 200 cases and 200 controls will yield a power of approximately 90%. A total of 203 women with P2/DY patterns were identified as cases and were individually matched by date of birth (within 1 year) and date of prevalence screening (within 3 months) to 203 women with N1/P1 patterns who served as controls. Additional information regarding case selection is presented elsewhere [<xref ref-type="bibr" rid="B25">25</xref>].</p><p>We examined the screening records of each woman. Mammograms of both breasts were collected. Two views, the mediolateral and craniocaudal mammograms, of both breasts were independently reviewed by two of the authors (ES and RW) to determine the Wolfe mammographic parenchymal pattern. The inter-reader agreement in the classification of mammographic parenchymal patterns was 95% on the four pattern categories, and 99% when the P2 and DY categories were combined, but for the purposes of the present study we used only the films in which we agreed on the patterns.</p><p>Considerable information on health and lifestyle factors was available from the EPIC Health and Lifestyle Questionnaire [<xref ref-type="bibr" rid="B16">16</xref>]. In the present study we examined the subjects' personal and family history of benign breast diseases and cancer, menstrual and reproductive factors, oral contraception and hormone replacement therapy, physical activity, smoking, and anthropometric information such as body mass index and waist:hip ratio.</p><sec><title>Statistical methods</title><p>Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated by conditional logistic regression, which takes into account the matching of controls to cases [<xref ref-type="bibr" rid="B17">17</xref>]. Adjustment was performed for those variables that were previously found to be associated with high-risk mammographic parenchymal patterns [<xref ref-type="bibr" rid="B25">25</xref>].</p></sec></sec><sec><title>Results</title><p>The characteristics of the cases and controls are presented in Table <xref ref-type="table" rid="T1">1</xref>. The mean age of cases and controls was similar (because they were matched on date of birth). Cases were leaner than controls. A larger percentage of cases were nulliparous, similar proportions of cases and controls had between one and three births, and a larger proportion of controls had more than three births. A larger proportion of cases were pre-menopausal, current hormone replacement therapy users, had a personal history of benign breast diseases, and had had a hysterectomy, whereas a larger proportion of controls were current smokers. The cases and controls were similar with respect to age at menarche and age at menopause.</p><p>Table <xref ref-type="table" rid="T2">2</xref> shows the unadjusted and adjusted OR estimates for Wolfe's high-risk mammographic parenchymal patterns and smoking in the total study population and in post-menopausal women separately. Current smoking was strongly and inversely associated with high-risk patterns, after adjustment for concomitant risk factors. Relative to never smokers, current smokers were significantly less likely to have a high-risk pattern (OR 0.37, 95% CI 0.14-0.94). Similar results were obtained when the analysis was confined to postmenopausal women. Past smoking was not related to the mammographic parenchymal patterns. The overall effect in postmenopausal women lost its statistical significance when adjusted for other risk factors for P2/DY patterns that were found to be significant in this study, although the results are still strongly suggestive. There was no interaction between cigarette smoking and body mass index (<italic>P</italic> =0.73 and 0.72 in the whole study population and in postmenopausal women, respectively).</p></sec><sec><title>Discussion</title><p>In the present study, we found a strong inverse relationship between current smoking and mammographic parenchymal patterns of breast tissue as classified by Wolfe [<xref ref-type="bibr" rid="B12">12</xref>]. These findings are not completely unprecedented; Greendale <italic>et al</italic> [<xref ref-type="bibr" rid="B18">18</xref>] found a reduced risk of breast density in association with smoking, although the magnitude of the reduction was unclear. Our findings suggest that this reduction is large.</p><p>Recent studies [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>] suggest that breast cancer risk may be reduced among current smokers. In a multicentre Italian case-control study, Braga <italic>et al</italic> [<xref ref-type="bibr" rid="B10">10</xref>] found that, relative to nonsmokers, current smokers had a reduced risk of breast cancer (OR 0.84, 95% CI 0.7-1.0). These findings were recently supported by Gammon <italic>et al</italic> [<xref ref-type="bibr" rid="B9">9</xref>], who reported that breast cancer risk in younger women (younger than 45 years) may be reduced among current smokers who began smoking at an early age (OR 0.59, 95% CI 0.41-0.85 for age 15 years or younger) and among long-term smokers (OR 0.70, 95% CI 0.52-0.94 for those who had smoked for 21 years or longer).</p><p>The possible protective effect might be due to the anti-oestrogenic effect of smoking [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Exposure to cigarette smoking causes an earlier menopause [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B26">26</xref>]. Smoking appears to alter the metabolism of oestradiol leading to enhanced formation of the inactive catechol estrogens [<xref ref-type="bibr" rid="B1">1</xref>]. Furthermore, smoking increases circulating androgens through adrenal cortical stimulation [<xref ref-type="bibr" rid="B2">2</xref>], but the conversion rates of androgens to oestrogens are lower in those who smoke [<xref ref-type="bibr" rid="B27">27</xref>]. There has been a recent resurgence of interest in the potential effect of smoking on breast cancer risk, and whether individuals may respond differently on the basis of differences in metabolism of bioproducts of smoking [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. Different relationships between smoking and breast cancer risk have been suggested that are dependent on the rapid or slow status of acetylators of aromatic amines [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>], rapid acetylators being better able to inactivate the potential carcinogenic tobacco compounds. More recent studies [<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>] do not support these findings, however.</p><p>The present study design minimized the opportunity for bias to influence the findings. Systematic error in the assessment of mammograms was avoided because reading was done without knowledge of the risk factor data. Because subjects were unaware of their own case-control status, the possibility of recall bias in reporting smoking status was minimized. Furthermore, the associations observed are unlikely to be explained by the confounding effect of other known breast cancer risk factors, because we adjusted for these in the analysis. We did not have information on passive smoking status, however, which has recently been reported as a possible confounder [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B24">24</xref>].</p><p>Although, ideally we would have liked to evaluate the relationship between intensity and duration of smoking and mammographic parenchymal patterns among current smokers, the numbers were too small to perform the analysis. Trends for intensity and duration of smoking were not monotonic, and <italic>P</italic> values were inconclusive (between 0.05 and 0.1). Age at menopause and time since menopause were not related to mammographic patterns in the present study (data not shown). Although current smokers were likely to have an early menopause (70% of current smokers were postmenopausal before age 50 years), there was no difference among mean age at menopause in the three smoking categories (<italic>P</italic> = 0.15). There was no difference in time since menopause among current smokers.</p><p>These data indicate that adjustment for current smoking status is important when evaluating the relationship between mammographic parenchymal patterns and breast cancer risk. They also indicate smoking to be a prominent potential confounder when analyzing effects of other risk factors, such as obesity-related variables. It appears that parenchymal patterns may act as an informative biomarker of the effect of cigarette smoking on breast cancer risk.</p></sec>

Altered expression of estrogen receptor-&#x003b1; variant messenger RNAs between adjacent normal breast and breast tumor tissues

<sec><title>Introduction:</title><p>Estrogen receptor (ER)-&#x003b1; and ER-&#x003b2; are believed to mediate the action of estradiol in target tissues. Several ER-&#x003b1; and ER-&#x003b2; variant messenger RNAs have been identified in both normal and neoplastic human tissues. Most of these variants contain a deletion of one or more exons of the wild-type (WT) ER messenger RNAs. The putative proteins that are encoded by these variant messenger RNAs would therefore be missing some functional domains of the WT receptors, and might interfere with WT-ER signaling pathways. The detection of ER-&#x003b1; variants in both normal and neoplastic human breast tissues raised the question of their possible role in breast tumorigenesis.</p><p>We have previously reported an increased relative expression of exon 5 deleted ER-&#x003b1; variant (ERD5) messenger RNA and of another ER-&#x003b1; variant truncated of all sequences following the exon 2 of the WT ER-&#x003b1; (ERC4) messenger RNA in breast tumor samples versus independent normal breast tissues. In contrast, a decreased relative expression of exon 3 deleted ER-&#x003b1; variant (ERD3) messenger RNA in tumor tissues and cancer cell lines versus independent normal reduction mammoplasty samples has recently been reported. These data were obtained in tissues from different individuals and possible interindividual differences cannot be excluded.</p></sec><sec><title>Aims:</title><p>The goal of this study was to investigate the expressions of ERC4, ERD5 and ERD3 variant messenger RNAs in normal breast tissues and their matched adjacent primary breast tumor tissues.</p></sec><sec><title>Materials and methods:</title><p>Eighteen cases were selected from the Manitoba Breast Tumor Bank, which had well separated and histopathologically characterized normal and adjacent neoplastic components. All tumors were classified as primary invasive ductal carcinomas. Six tumors were ER-negative/progesterone receptor (PR)-negative, nine were ER-positive/PR-positive, two were ER-positive/PR-negative, and one was ER-negative/PR-positive, as measured by ligand-binding assay. For each specimen, total RNA was extracted from frozen normal and tumor tissue sections and was reverse transcribed. The expressions of ERC4, ERD3 and ERD5 messenger RNAs relative to WT ER-&#x003b1; messenger RNA were investigated by previously validated semiquantitative reverse transcription polymerase chain reaction (PCR) assays performed using three different sets of primers.</p></sec><sec><title>Results:</title><p>As shown Figure <xref ref-type="fig" rid="F2">1a</xref>, two PCR products were obtained that corresponded to WT ER and ERC4 messenger RNAs. For each case, the mean of the ratios obtained in at least three independent PCR experiments is shown for both normal and tumor compartments (Fig <xref ref-type="fig" rid="F2">1b</xref>). A statistically higher ERC4 messenger RNA relative expression was found in the neoplastic components of ER-positive/PR-positive tumors, as compared with matched adjacent normal tissues (<italic>n</italic> = 9; <italic>P</italic> = 0.019, Wilcoxon signed-rank test).</p><p>Two PCR products were obtained that corresponded to WT ER and ERD3 messenger RNAs (Fig <xref ref-type="fig" rid="F2">2a</xref>). A significantly higher expression of ERD3 messenger RNA was observed in the normal compared with the adjacent neoplastic components of ER-positive subset (<italic>n</italic> =8; <italic>P</italic> =0.023, Wilcoxon signed-rank test; Fig <xref ref-type="fig" rid="F2">2b</xref>).</p><p>Two PCR products were obtained that corresponded to WT ER and ERD5 complementary DNAs (Fig <xref ref-type="fig" rid="F2">3a</xref>). As shown in Figure <xref ref-type="fig" rid="F3">3b</xref>, a statistically significant higher relative expression of ERD5 messenger RNA was observed in tumor components when this expression was measurable in both normal and adjacent tumor tissues (<italic>n</italic> =15; <italic>P</italic> =0.035, Wilcoxon signed-rank test).</p></sec><sec><title>Discussion:</title><p>A statistically significant higher ERC4 messenger RNA expression was found in ER-positive/PR-positive tumors as compared with matched normal breast tissues. ERC4 variant messenger RNA has previously been demonstrated to be more highly expressed in ER-positive tumors that showed poor as opposed to tumors that showed good prognostic characteristics. Interestingly, we also have reported similar levels of expression of ERC4 messenger RNA in primary breast tumors and their concurrent axillary lymph node metastases. Taken together, these data suggest that the putative role of the ERC4 variant might be important at different phases of breast tumorigenesis and tumor progression; alteration of ERC4 messenger RNA expression and resulting modifications in ER signaling pathway probably occur before breast cancer cells acquire the ability to metastasize. Transient expression assays revealed that the protein encoded by ERC4 messenger RNA was unable to activate the transcription of an estrogen-responsive element-reporter gene or to modulate the wild-type ER protein activity. The biologic significance of the changes observed in ERC4 messenger RNA expression during breast tumorigenesis remains to be determined.</p><p>A higher relative expression of ERD3 messenger RNA in the normal breast tissue components compared with adjacent neoplastic tissue was found in the ER-positive subgroup. These data are in agreement with the recently published report of Erenburg <italic>et al</italic>, who showed a decreased relative expression of ERD3 messenger RNA in neoplastic breast tissues compared with independent reduction mammoplasty and breast tumor. Transfection experiments showed that the activation of the transcription of the pS2 gene by estrogen was drastically reduced in the presence of increased ERD3 expression. The authors hypothesized that the reduction in ERD3 expression could be a prerequisite for breast carcinogenesis to proceed.</p><p>We observed a significantly higher relative expression of ERD5 messenger RNA in breast tumor components compared with matched adjacent normal breast tissue. These data confirm our previous observations performed on unmatched normal and neoplastic human breast tissues. Upregulated expression of this variant has already been reported in ER-negative/PR-positive tumors, as compared with ER-positive/PR-positive tumors, suggesting a possible correlation between ERD5 messenger RNA expression and breast tumor progression. Even though it has been suggested that ERD5 could be related to the acquisition of insensitivity to antiestrogen treatment (ie tamoxifen), accumulating data refute a general role for ERD5 in hormone-resistant tumors. Only ER-positive pS2-positive tamoxifen-resistant tumors have been shown to express significantly higher levels of ERD5 messenger RNA, as compared with control tumors. Taken together, these data suggest that the exact biologic significance of ERD5 variant expression during breast tumorigenesis and breast cancer progression, if any, remains unclear.</p><p>In conclusion, we have shown that the relative expressions of ERC4 and ERD5 variant messenger RNAs were increased in human breast tumor tissue, as compared with normal adjacent tissue, whereas the expression of ERD3 variant messenger RNA was decreased in breast tumor tissues. These results suggest that the expressions of several ER-&#x003b1; variant messenger RNAs are deregulated during human breast tumorigenesis. Further studies are needed to determine whether these changes are transposed at the protein level. Furthermore, the putative role of ER-&#x003b1; variants in the mechanisms that underlie breast tumorigenesis remains to be determined.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Leygue</surname><given-names>Etienne</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>eleygue@cc.umanitoba.ca</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Dotzlaw</surname><given-names>Helmut</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Watson</surname><given-names>Peter H</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Murphy</surname><given-names>Leigh C</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>Estrogen receptor (ER)-&#x003b1; and ER-&#x003b2; are believed to mediate the action of estradiol in target tissues [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. These two receptors, which belong to the steroid/retinoic acid/thyroid receptor superfamily [<xref ref-type="bibr" rid="B3">3</xref>], contain several structural and functional domains [<xref ref-type="bibr" rid="B4">4</xref>] that are encoded by two messenger RNAs that contain eight exons [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. Upon ligand binding, ER-&#x003b1; and ER-&#x003b2; proteins recognize specific estrogen-responsive elements located in DNA in the proximity of target genes, and through interactions with several coactivators modulate the transcription of these genes [<xref ref-type="bibr" rid="B7">7</xref>].</p><p>Several ER-&#x003b1; and ER-&#x003b2; variant messenger RNAs have been identified in both normal and neoplastic human tissues [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. Most of these variants contain a deletion of one or more exons of the wild-type (WT)-ER messenger RNA. The putative proteins encoded by these variant messenger RNAs would therefore be missing some functional domains of the WT receptors and might interfere with WT ER signaling pathways. Indeed, <italic>in vitro</italic> functional studies have shown that some recombinant ER-&#x003b1; variant proteins can affect estrogen-regulated gene transcription. For example, the variant protein encoded by exon 3 deleted ER-&#x003b1; variant (ERD3) messenger RNA, which is missing the second zinc finger of the DNA binding domain, has been shown [<xref ref-type="bibr" rid="B13">13</xref>] to have a dominant negative activity on WT ER-&#x003b1; receptor action. A similar dominant negative activity has been observed for ERD5 variant protein (encoded by an ER-&#x003b1; variant messenger RNA deleted in exon 5 sequences), which is missing a part of the hormone-binding domain of the WT molecule [<xref ref-type="bibr" rid="B14">14</xref>]. Interestingly, a constitutive hormone-independent activity [<xref ref-type="bibr" rid="B15">15</xref>] and a WT enhancing activity [<xref ref-type="bibr" rid="B16">16</xref>] have also been attributed to ERD5 variant protein in different systems. The relevance of the levels achieved in these transfection experiments to <italic>in vivo</italic> expression remains unclear. It should also be noted that these functional activities are likely to be cell-type and promoter specific [<xref ref-type="bibr" rid="B8">8</xref>].</p><p>The discovery that these ER-&#x003b1; variants are expressed in both normal and neoplastic human breast tissues, however, raised the question of their possible role in breast tumorigenesis [<xref ref-type="bibr" rid="B8">8</xref>]. We have previously reported an increased relative expression of ERD5 messenger RNA and of ERC4 messenger RNA, another ER-&#x003b1; variant messenger RNA that is truncated of all sequences following the exon 2 of the WT ER-&#x003b1; [<xref ref-type="bibr" rid="B17">17</xref>], in breast tumor samples versus independent normal breast tissues [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. In contrast, Erenburg <italic>et al</italic> [<xref ref-type="bibr" rid="B20">20</xref>] recently reported a decreased relative expression of ERD3 messenger RNA in tumor tissues and cancer cell lines versus independent normal reduction mammoplasty samples. Those data, which suggested that alteration in ERD5, ERD3 and clone 4 messenger RNA expression might occur during breast tumorigenesis, were obtained in tissues from different individuals, and possible interindividual differences cannot be excluded.</p><p>In order to clarify this issue, we investigated the expression of these three variant messenger RNAs in normal breast tissues and their matched adjacent primary breast tumor tissues.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Human breast tissues and reverse transcription</title><p>In order to investigate the expressions of ERC4, ERD3 and ERD5 messenger RNA relative to WT-ER messenger RNA within matched normal and breast tumor tissues, eighteen cases were selected in the National Cancer Institute of Canada Manitoba Breast Tumor Bank (Winnipeg, Manitoba, Canada), which had well separated and histopathologically characterized normal and adjacent neoplastic components. The Tumor Bank, which serves as a national Tumor Bank and is funded by the National Cancer Institute of Canada, has been reviewed and received approval from the Ethics Review Committee, Faculty of Medicine, University of Manitoba.</p><p>The processing of specimens collected in the Manitoba Breast Tumor Bank has already been described [<xref ref-type="bibr" rid="B21">21</xref>]. Briefly, each specimen had been rapidly frozen as soon as possible after surgical removal. A portion of the frozen tissue block was processed to create a paraffin-embedded tissue block that was matched and oriented relative to the remaining frozen block. These paraffin blocks provide high quality histologic sections, which are used for pathologic interpretation and assessment, and are mirror images of the frozen sections used for RNA extractions.</p><p>For each case, tumor and adjacent normal tissues from the same individual were histologically characterized by observation of paraffin sections. The presence of normal ducts and lobules, as well as the absence of any atypical lesion, were confirmed in all normal tissue specimens. All tumor components were classified as primary invasive carcinomas. Seven tumors were ER-negative (ER &#x0003c; 3 fmol/mg protein), with progesterone receptor (PR) values ranging from 2.2 to 11.2f mol/mg protein, as measured using ligand-binding assay [<xref ref-type="bibr" rid="B22">22</xref>]. Axillary nodal metastases were observed in five of these cases. Eleven tumors were ER-positive (ER values ranged from 3.5 to 159 fmol/mg protein), with PR values ranging from 5.8 to 134 fmol/mg protein. These tumors spanned a wide range of grades (grades 5-9, median 7.5), which were determined using the Nottingham grading system [<xref ref-type="bibr" rid="B23">23</xref>]. Axillary nodal metastases were observed in one of these cases. Patients were from 39 to 86 years old (median 54 years). Total RNA was extracted from frozen tissue sections and reverse-transcribed in a final volume of 25 &#x003bc; l as previously described [<xref ref-type="bibr" rid="B18">18</xref>]. The quality of complementary DNAs obtained was assessed by amplification of the ubiquitously expressed glyceraldehyde-3-phosphate dehydrogenase complementary DNA, as described previously [<xref ref-type="bibr" rid="B18">18</xref>].</p></sec><sec><title>Triple primer polymerase chain reaction</title><p>A previously described triple primer polymerase chain reaction (PCR) assay has been used to coamplify ERC4 and WT-ER-&#x003b1; complementary DNAs [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. Primers used consisted of ERU primer (5' -TGTGCAATGACTATGCTTCA-3', sense, located in WT-ER exon 2, position 792-811), ERL primer (5' -GCTCTTCCTCCTGTTTTTAT-3', antisense, located in WT-ER exon 3, position 940-921), and C4L primer (5' -TTTCAGTCTTCAGATACCCCAG-3', antisense, located in ERC4 sequence, position 1336-1315). The given positions correspond to the published sequences for WT-ER [<xref ref-type="bibr" rid="B1">1</xref>] and ERC4 [<xref ref-type="bibr" rid="B17">17</xref>].</p><p>PCR amplifications were performed as previously described [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. Briefly, 0.2 &#x003bc; l reverse transcription mixture was amplified in a final volume of 15 &#x003bc; l, in the presence of 1.5 &#x003bc; Ci of [&#x003b1;-³²P] deoxycytidine triphosphate (dCTP; 3000 Ci/mmol), 4 ng/&#x003bc;l of each primer and 0.3 unit of Taq DNA polymerase. Each cycle consisted of 1min at 94&#x000b0;C, 30s at 60&#x000b0;C and 1min at 72&#x000b0;C. PCR products were then separated on 6% polyacrylamide gels containing 7mol/l urea (polyacrylamide gel electrophoresis). After electrophoresis, the gels were dried and autoradiographed. Two PCR products were obtained, which were identified by subcloning and sequencing, performed as previously described [<xref ref-type="bibr" rid="B18">18</xref>]. PCR products migrating with the apparent size of 149 and 536 base pairs were shown to correspond to WT-ER and ERC4 complementary DNAs, respectively.</p></sec><sec><title>Polymerase chain reaction</title><p>Two different primer sets, ERD3 and ERD5, were used to coamplify WT-ER and ERD3 complementary DNAs, and WT-ER and ERD5 complementary DNAs, respectively. ERD3 primer set consisted of D3U primer (5' -TGTGCAATGACTATGCTTCA-3', sense, located in WT-ER exon 2, position 792-811) and D3L primer (5' -TGTTCTTCTTAGAGCGTTTGA-3', antisense, located in WT-ER exon 4, position 1145-1125). ERD5 primer set consisted of D5U primer (5' -CAGGGGTGAAGTGGGGTCTGCTG-3', sense, located in WT-ER exon 4, position 1060-1082) and D5L primer (5'-&#x003b1; TGCGGAACCGAGATGATGTAGC-3', anti-sense, located in WT-ER exon 6, position 1542-1520). The given positions correspond to published sequences for WT-ER [<xref ref-type="bibr" rid="B1">1</xref>].</p><p>PCR amplifications were performed and PCR products analyzed as previously described [<xref ref-type="bibr" rid="B18">18</xref>]. Briefly, 0.2 &#x003bc; l reverse transcription mixture was amplified in a final volume of 15 &#x003bc; l, in the presence of 1.5 &#x003bc; Ci of [&#x003b1;-³²P] dCTP (3000 Ci/mmol), 4ng/&#x003bc; l of each primer of the primer set considered (ERD3 or ERD5 primer set) and 0.3 unit of Taq DNA polymerase. Each cycle consisted of 30s at 94&#x000b0;C, 30s at 60&#x000b0;C and 30s at 72&#x000b0;C. PCR products were then separated on 6% polyacrylamide gels containing 7mol/l urea (polyacrylamide gel electrophoresis). Following electrophoresis, the gels were dried and autoradiographed. For each PCR, two PCR products were obtained, which were identified by subcloning and sequencing. PCR products migrating with the apparent size of 354 and 483 base pairs, using ERD3 and ERD5 primer set, respectively, were shown to correspond to WT-ER complementary DNA. PCR products migrating with the apparent size of 237 and 344 base pairs, using ERD3 and ERD5 primer set, were shown to correspond to ERD3 and ERD5 complementary DNAs, respectively.</p></sec><sec><title>Quantitation and statistical analysis</title><p>For each experiment, bands corresponding to the variant messenger RNA (ie ERC4, ERD3 or ERD5) and to WT-ER were excised from the gel and counted in a scintillation counter. For each set of primers (ie ERC4, ERD3 and ERD5 primer set) and for each sample, four independent PCR assays were performed. The ratios between ERC4, ERD3 or ERD5 signals and corresponding WT-ER signal were calculated. For each experiment, in order to correct for overall interassay variations (due to different batches of radiolabelled [&#x003b1; -³²P] dCTP or of Taq DNA polymerase), the ratio observed in the same particular tumor (case number 12) was arbitrarily given the value of one and all other ratios expressed relatively. Under our experimental conditions, some samples did not have measurable levels (ie signal lower than twice the background value) of ERD3 or ERD5 variant messenger RNAs (see Figs <xref ref-type="fig" rid="F2">2a</xref> and <xref ref-type="fig" rid="F3">3a</xref>) in any of the four repetitions performed. Only cases that had detectable levels in at least three of the replicates in both their normal and tumor compartments were included in the statistical analysis. The significance of the differences in the relative levels of expression of ERC4, ERD3 and ERD5 messenger RNAs between matched normal and tumor components was determined using the Wilcoxon signed-rank test.</p></sec></sec><sec><title>Results</title><sec><title>Relative expression of ERC4 messenger RNA in matched normal and breast tumor tissues</title><p>A recently described triple-primer PCR assay was used to compare the relative expressions of ERC4 messenger RNA between adjacent normal and tumor components [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. In this assay, three primers are used simultaneously during the PCR: the upper primer is able to recognize both WT-ER and ERC4 complementary DNA sequences, whereas the two lower primers are specific for each complementary DNA. Competitive amplification of two PCR products occurs, giving a final PCR product ratio related to the initial input of target complementary DNAs. This approach has been validated previously both by competitive amplification of spiked complementary DNA preparations [<xref ref-type="bibr" rid="B19">19</xref>] and by comparison to RNAse protection assays [<xref ref-type="bibr" rid="B24">24</xref>].</p><p>As shown Figure <xref ref-type="fig" rid="F1">1a</xref>, two PCR products were obtained, which migrated at the apparent size of 149 and 536 base pairs. These products have been shown to correspond to WT-ER and ERC4 messenger RNAs, respectively [<xref ref-type="bibr" rid="B24">24</xref>]. One should note the presence, in samples where WT-ER and ERC4 signals are high (Fig <xref ref-type="fig" rid="F1">1a</xref>, lane 5), of minor additional bands, one of which has been previously identified as corresponding to exon 2-duplicated ER-&#x003b1; variant complementary DNA [<xref ref-type="bibr" rid="B24">24</xref>]. The presence of these minor PCR products did not interfere with the quantitative aspect of the triple-primer PCR assay [<xref ref-type="bibr" rid="B24">24</xref>]. For each case, the mean of the ratios obtained in at least three independent PCR experiments, expressed in arbitrary units, is shown for both normal and tumor compartments (Fig <xref ref-type="fig" rid="F1">1b</xref>). A higher clone 4 messenger RNA relative expression in the tumor compartment was observed in 12 out of 18 cases. This difference did not, however, reach statistical significance (<italic>P</italic> = 0.47, Wilcoxon signed-rank test). When considering only the ER-positive/PR-positive subset (<italic>n</italic> = 9), as measured by the ligand-binding assay, a statistically higher ERC4 messenger RNA relative expression was found in the neoplastic components, as compared with matched adjacent normal tissues (<italic>P</italic> = 0.019, Wilcoxon signed-rank test).</p></sec><sec><title>Relative expression of ERD3 messenger RNA in matched normal and breast tumor tissues</title><p>A PCR assay, performed using primers annealing to sequences in exons 2 and 4, was used to investigate ERD3 messenger RNA expression relative to WT-ER in these 18 matched cases. We [<xref ref-type="bibr" rid="B18">18</xref>] and others [<xref ref-type="bibr" rid="B25">25</xref>] have previously shown that the coamplification of WT-ER and an exon-deleted ER-&#x003b1; variant complemetary DNA resulted in the amplification of two PCR products, the relative signal intensity of which provided a previously validated measurement of exon-deleted ER-&#x003b1; variant expression.</p><p>Two PCR products were obtained, that migrated with an apparent size of 354 and 237 base pairs (Fig <xref ref-type="fig" rid="F2">2a</xref>). These fragments were shown by subcloning and sequencing to correspond to WT-ER and ERD3 messenger RNAs (data not shown). The relative ERD3 signal was measurable in the normal and in the tumor compartments of 13 cases (Fig <xref ref-type="fig" rid="F2">2b</xref>). Out of these 13 cases, ERD3 messenger RNA expression was higher in the normal compartment in 10 cases. This difference, however, did not reach statistical significance (<italic>P</italic> = 0.057, Wilcoxon signed-rank test). A significantly higher expression of ERD3 messenger RNA in the normal compared with the adjacent neoplastic components was found when only the ER-positive subset was considered, however (<italic>n</italic> = 8; <italic>P</italic> = 0.023, Wilcoxon signed-rank test).</p></sec><sec><title>Relative expression of ERD5 messenger RNA in matched normal and breast tumor tissues</title><p>Using primers annealing to sequences in exons 4 and 6 of WT-ER, we also investigated the relative expression of ERD5 messenger RNA in these 18 matched cases. Two PCR products were detected, that migrated at an apparent size of 483 and 344 base pairs, and that have previously been shown to correspond to WT-ER and ERD5 complementary DNAs, respectively (Fig <xref ref-type="fig" rid="F3">3a</xref>). As shown in Fig <xref ref-type="fig" rid="F3">3b</xref>, a statistically significant higher relative expression of ERD5 messenger RNA was observed in tumor components when this expression was measurable in both normal and adjacent tumor tissues (<italic>n</italic> = 15; <italic>P</italic> = 0.035, Wilcoxon signed-rank test).</p></sec></sec><sec><title>Discussion</title><p>The expression of ERC4, ERD3 and ERD5 variant messenger RNAs relative to WT-ER messenger RNA expression within adjacent normal and neoplastic human breast tissues was investigated using previously described semi-quantitative reverse transcription PCR assays [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. These assays allow the determination of the expression of ER-&#x003b1; variant messenger RNA relative to WT-ER messenger RNA using a very small amount of starting material, and offer the advantage of allowing investigators to work with histopathologically well characterized human breast tissue regions. It should be noted, however, that the sensitivities of the assays used in this study differed from each other. The triple-primer PCR assay has previously been set up to allow the determination of ERC4 relative expression in tumor samples with very low ER levels, as measured by ligand-binding assay [<xref ref-type="bibr" rid="B24">24</xref>].</p><p>We showed that, in samples with a detectable level of ERC4 messenger RNA using a standardized RNAse protection assay, the relative expression of this variant to WT-ER messenger RNA expression was similar to the relative expression of ERC4 PCR product obtained after triple-primer PCR [<xref ref-type="bibr" rid="B24">24</xref>]. Triple-primer PCR assay applied to the detection of ERC4 messenger RNA in 18 matched normal and tumor breast tissues gave a measurable value of expression in 36 out of the 36 samples studied. This contrasts with the detection of 30 out of 36 and 33 out of 36 obtained using ERD3-specific and ERD5-specific primers, respectively. These differences in sensitivity probably result from different primer set efficiencies under our experimental conditions.</p><p>A higher ERC4 messenger RNA relative expression in tumor components compared with the normal adjacent tissue component has been observed in the ER-positive/PR-positive subgroup. This result is in agreement with our previous data [<xref ref-type="bibr" rid="B19">19</xref>] obtained by comparing ERC4 messenger RNA expression between independent normal reduction mammoplasty samples and a group of ER-positive/PR-positive breast tumors. Even though a higher ERC4 messenger RNA relative expression was observed in the tumor component of 12 out of 18 cases, this difference did not reach statistical significance. This absence of statistically significant differences might result from the low number of matched cases studied or from the different biology of ER-negative cases. Further studies are needed to clarify this issue and to draw any conclusion regarding the expression of ERC4 messenger RNA in ER-negative samples.</p><p>ERC4 variant messenger RNA has previously been shown [<xref ref-type="bibr" rid="B26">26</xref>] to be more highly expressed in ER-positive tumors that show poor prognostic characteristics (presence of more than four axillary lymph nodes, tumor size &#x0003e;2 cm, aneuploid, high percentage S-phase cells) than in ER-positive tumor with good prognostic characteristics (absence of axillary lymph node, tumor size &#x0003c;2 cm, diploid, low percentage S-phase cells). Moreover, in that previous study, a higher ERC4 messenger RNA expression was also observed in ER-positive/PR-negative tumors, as compared with ER-positive/PR-positive tumors. interestingly, we have also recently reported similar levels of expression of ERC4 messenger RNA in primary breast tumors and their concurrent axillary lymph node metastases [<xref ref-type="bibr" rid="B24">24</xref>]. Taken together, these data suggest that the putative role of the ERC4 variant might be important at different phases of breast tumorigenesis and tumor progression; alteration in ERC4 messenger RNA expression and resulting modifications in ER signaling pathway probably occur before breast cancer cells acquire the ability to metastasize. Transient expression assays revealed that the protein encoded by ERC4 messenger RNA was unable to activate the transcription of an estrogen responsive element-reporter gene or to modulate WT-ER protein activity [<xref ref-type="bibr" rid="B17">17</xref>]. The biologic significance of the changes observed in ERC4 messenger RNA expression during breast tumorigenesis and tumor progression therefore remains unclear.</p><p>A trend toward a higher relative expression of ERD3 messenger RNA in the normal breast tissue components compared with adjacent neoplastic tissue was found (10 out of 13 cases), which reached statistical significance when the ER-positive subgroup only was considered. These data are in agreement with the recently published report of Erenburg <italic>et al</italic> [<xref ref-type="bibr" rid="B20">20</xref>] who showed a decreased relative expression of ERD3 messenger RNA in neoplastic breast tissues and breast cancer compared with independent reduction mammoplasty and breast tumor. Transfection experiments performed by those investigators showed that the activation of the transcription of the pS2 gene by estrogen was drastically reduced in the presence of increased ERD3 expression. Moreover, ERD3 transfected MCF-7 human breast cancer cells had a reduced saturation density, exponential growth rate and <italic>in vivo</italic> invasiveness, as compared with control cells. These data led the authors to hypothesize that the reduction of ERD3 expression could be a prerequisite for breast carcinogenesis to proceed. They suggested that if high levels of ERD3 could attenuate estrogenic effects in normal breast tissue, low levels might lead to an excessive and unregulated mitogenic action of estrogen.</p><p>We observed a significantly higher relative expression of ERD5 messenger RNA in breast tumor components compared with matched adjacent normal breast tissue. These data confirm our previous observations [<xref ref-type="bibr" rid="B18">18</xref>] performed on unmatched normal and neoplastic human breast tissues. Upregulated expression of this variant has already been reported in ER-negative/PR-positive tumors, as compared with ER-positive/PR-positive tumors [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B27">27</xref>], suggesting a possible correlation between ERD5 messenger RNA expression and breast tumor progression. Interestingly, ERD5 messenger RNA can be detected in human pituitary adenomas, but not in normal pituitary samples [<xref ref-type="bibr" rid="B28">28</xref>]. This underscores the putative involvement of this ER variant in other tumor systems. Even though it has been suggested that ERD5 could be related to the acquisition of insensitivity to antiestrogen treatment (ie tamoxifen) [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>], accumulating data refute a general role for ERD5 in hormone-resistant tumors [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>]. Only ER-positive pS2-positive tamoxifen resistant tumors have been shown to express significantly higher levels of ERD5 messenger RNA, as compared with control tumors [<xref ref-type="bibr" rid="B33">33</xref>]. Taken together, these data suggest that the exact biologic significance of ERD5 variant expression during breast tumorigenesis and breast cancer progression, if any, remains unclear.</p><p>Among all the articles published so far on ER variants, only one has investigated ER variant expression between normal and neoplastic matched samples. Okada <italic>et al</italic> [<xref ref-type="bibr" rid="B33">33</xref>] recently reported a study performed on 15 cases. They observed an apparent difference in ER variant messenger RNA expression between adjacent normal and tumor samples. That study was performed using a less sensitive PCR approach, however, because PCR products were stained using ethidium bromide, and no attempt was made to quantify ER variant messenger RNA expression relative to WT-ER messenger RNA expression.</p><p>In conclusion, we have shown that the relative expression of ERC4 and ERD5 variant mRNAs was increased in human breast tumor tissue, as compared with normal adjacent tissue, whereas the expression of ERD3 variant messenger RNA was decreased in breast tumor tissues. These results, which confirm previous data obtained on independent human breast tissue samples [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>], suggest that the expressions of several ER-&#x003b1; variant messenger RNAs are deregulated during human breast tumorigenesis. Further studies are needed to determine whether these changes are transposed at the protein level. Only the use of specific antibodies that are able to recognize specifically the different ER variant proteins putatively encoded by these variant messenger RNAs will allow this issue the be addressed. Furthermore, the putative role of ER-&#x003b1; variants in the mechanisms that underlie breast tumorigenesis remain to be determined.</p></sec>

The Million Women Study: design and characteristics of the study population

<sec><title>Objectives:</title><p>To describe the design of the Million Women Study and the characteristics of the study population.</p></sec><sec><title>Study design:</title><p>Population-based cohort study of women aged 50-64 in the UK.</p></sec><sec><title>Setting:</title><p>Women are asked to join the Million Women Study when they are invited to routine screening for breast cancer at 61 of the screening centres of the UK National Health Service Breast Screening Programme (NHSBSP). An estimated 71% of women screened by the NHSBSP return a completed questionnaire.</p></sec><sec><title>Participants:</title><p>800 000 women were recruited between May 1996 and June 1999, and it is planned that an additional 200 000 will be recruited by the year 2000.</p></sec><sec><title>Results:</title><p>The characteristics of the first 121 000 women recruited into the Million Women Study are described here. At recruitment 33% of the study population were currently using hormone replacement therapy and 47% had used it at some time. Over half (54%) had used oral contraceptives, and 18% were current smokers at the time of recruitment. Before they were screened 1.4% of the women had been diagnosed with breast cancer in the past, 6% had a mother with a history of breast cancer and 3.7% had a sister with a history of breast cancer. It is estimated that 1 million women will have been recruited by early in the year 2000, and that by the end of the year 2002 there will be 5000 screen-detected breast cancers and 23 000 deaths in the cohort, the majority of which will be attributed to cancer (12 600 deaths) and circulatory disease (8000 deaths).</p></sec><sec><title>Conclusions:</title><p>By the end of the year 2002, the Million Women Study will have sufficient statistical power to detect relative risks of 0.8 or less, or of 1.2 or more in current users compared with never users of hormone replacement therapy for mortality from breast cancer, colorectal cancer, lung and ovarian cancer, ischaemic heart disease and stroke.</p></sec>

<contrib id="A1" contrib-type="author"><collab>The Million Women Study Collaborative Group</collab><xref ref-type="aff" rid="I1">1</xref></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>The Million Women Study is a nationwide collaborative research project in the UK, the chief aim of which is to describe the relationship between use of hormone replacement therapy (HRT) and the risk of various conditions, particularly breast cancer. The study began in May 1996 and the plan is to recruit and follow-up a cohort of 1 million women invited to attend the UK National Health Service Breast Screening Programme (NHSBSP).</p><p>The NHSBSP was set up in 1988. Once every 3 years each woman in the UK aged between 50 and 64 years who is registered with the NHS is sent a letter by the NHSBSP, offering her routine screening for breast cancer by mammography. About 1 million women are screened annually by the NHSBSP and about 5000 of them have a breast cancer detected on mammography [<xref ref-type="bibr" rid="B1">1</xref>]. The day-to-day organization and screening activities are performed by about 100 separate screening offices throughout the UK, and the work is monitored and statistics gathered centrally by a national co-ordinating centre.</p><p>The present paper describes the design of the Million Women Study and the characteristics of the study population.</p></sec><sec sec-type="methods"><title>Methods</title><p>The Million Women Study is a population-based cohort study. Women are recruited when they are invited for routine breast cancer screening, and the main outcomes to be examined at follow-up are the incidence of screen-detected breast cancer and cause-specific mortality.</p><sec><title>Attendance at screening</title><p>About three-quarters of the women who are invited for screening by the NHSBSP subsequently attend for mammography [<xref ref-type="bibr" rid="B1">1</xref>]. Before the study could be launched, it was necessary to demonstrate that inviting women to join the Million Women Study would not reduce uptake of screening offered by the NHSBSP. During 1994 and 1995 a total of 6000 women who were due to be invited for breast cancer screening in Oxford and West London were randomly divided into two groups. One group was sent the usual invitation for screening and the other group was sent the study questionnaire, accompanying the usual invitation to screening. Attendance rates for screening were similar, at 71%, among those who were and were not sent an accompanying questionnaire [<xref ref-type="bibr" rid="B2">2</xref>].</p></sec><sec><title>Recruitment procedures</title><p>Women are asked to join the Million Women Study by participating NHSBSP screening centres at the time that or just before they are sent their usual invitation for routine breast cancer screening. A questionnaire is included with each woman's invitation and, if the woman wishes to join the study, she is asked to complete the questionnaire, to give signed permission for follow-up, and to return the questionnaire at the time she is screened. A freephone number is provided for women who have any questions or problems filling out the questionnaire. The questionnaire is four pages long (A4 size) and includes questions about lifestyle and sociodemographic factors, reproductive history, past use of oral contraceptives, use of HRT, past medical history and family history of breast cancer. Completed questionnaires are transferred periodically from the participating screening centres to the study co-ordinating centre at the Imperial Cancer Research Fund Cancer Epidemiology Unit (CEU), Oxford, UK.</p></sec><sec><title>Data storage, entry and checking</title><p>The confidential completed questionnaires are stored securely at all times. Once they reach the CEU they are checked and coded by trained staff and then scanned electronically. The scanned data are 'captured' using computerized intelligent character recognition and optical mark reading software (Eyes and Hands^&#x000ae;; Readsoft Inc, Slough, UK). Range and logical checks are performed at the time of data entry. Any inconsistency or information that is not recognised by the data capture software is verified manually by trained data entry staff, who also validate computer-interpreted data and check each questionnaire to confirm whether signed consent for follow up has been granted. Each week the verified data for about 50 individuals are checked against the original questionnaires and the error rate is consistently below 1%. This partially automated process thus permits data to be entered rapidly and with high accuracy.</p></sec><sec><title>Follow up for breast cancer</title><p>Each screening centre of the NHSBSP is required to compile annual statistics on its activities, which include details of all breast cancers detected at mammography [<xref ref-type="bibr" rid="B1">1</xref>]. A list of women enrolled into the Million Women Study at each centre is cross-checked at regular intervals against the list of the women diagnosed with screen-detected breast cancer at that centre. If a breast cancer has been diagnosed at screening in a study participant, routinely recorded details of the cancer are abstracted, including tumour location, histology, size, grade, invasive status and involvement of axillary lymph nodes. Information on hormone receptor status and treatment is abstracted when it is available. Several approaches are being used to identify breast cancers diagnosed subsequent to screening. One will involve record linkage with cancer registry data. Also, women will be contacted directly 2-3 years after they were screened, and asked about new illnesses, including any new breast cancers, that may have been diagnosed (see Additional follow up, below). This will permit the identification of both screen-detected and interval cancers.</p></sec><sec><title>Follow-up for deaths</title><p>Deaths are identified annually by computerized matching of name, date of birth and NHS number of the women who gave signed consent for follow up in the Million Women Study, with the national death files held by the Office of National Statistics. For each death thus identified the date of death and underlying and associated causes of death are provided by the Office of National Statistics.</p></sec><sec><title>Additional follow up</title><p>Participants will be sent a follow-up questionnaire about 2-3 years after recruitment, to ascertain changes in use of HRT and incident morbidity, for example breast cancers, diagnosed outside the screening programme.</p></sec><sec><title>Validation</title><p>The most important variables for this study are the subjects' identification details, their use of HRT, any diagnosis of breast cancer and the recording of deaths. To assess the accuracy of the subjects' identification details and of the recording of deaths, a random sample of 5000 women recruited in 1996 has been selected for flagging on the NHS Central Register (NHSCR). Identification details recorded for the study (name, address, date of birth and NHS number) enabled all but 10 (0.2%) of the 5000 women to be identified on the NHSCR. The completeness and accuracy of the reported deaths will be validated in the future against those recorded in the NHSCR for these 5000 women.</p><p>The reliability of diagnosis of screen-detected breast cancers is monitored by various quality control procedures within the NHSBSP. Screen-detected breast cancers are verified according to defined procedures, and the invasive status, size and type of cancer are recorded for virtually 100% of the cancers.</p><p>The validity of reported information on use of HRT, including the type and dose, is being examined and a full report will be published in due course. Preliminary comparisons with the prescription records from one general practice in Oxfordshire indicate at least 95% agreement for reported current use of HRT, including the hormonal constituents of the preparation used most recently (Banks <italic>et al</italic>, unpublished data).</p></sec></sec><sec><title>Results</title><p>After it was demonstrated that the Million Women Study questionnaire did not alter attendance rates [<xref ref-type="bibr" rid="B2">2</xref>], each screening centre in England, Scotland and Wales was invited to participate in the study. Almost all the centres expressed enthusiasm for the study, although practical problems precluded the involvement of some centres. The most frequent reason for screening centres not participating was that the Million Women Study questionnaire could not readily be packaged together with the letters and other information normally posted to women when they are invited for screening.</p><sec><title>Accrual of the cohort</title><p>Recruitment of women into the study began in May 1996. Most of the participating screening centres began recruitment during 1997, and 61 centres were taking part by late 1998. The locations of these centres are shown in Figure <xref ref-type="fig" rid="F1">1</xref>. Before recruitment could begin at any centre, local ethical committee approval was required, and this often entailed contacting more than one ethical committee for each centre. In total, 126 local ethical committees were approached and approval for the study was obtained without exception.</p><p>Figure <xref ref-type="fig" rid="F2">2</xref> shows the numbers of questionnaires returned to the CEU between May 1996 and June 1999. More than 800000 questionnaires had been returned by the middle of 1999, and according to this accrual rate it is estimated that a cohort of 1 million women will have been recruited by early in the year 2000. </p></sec><sec><title>Response rate</title><p>Statistics presented here are based on the first 227 000 questionnaires, which were printed between May 1996 and February 1997. This represented a convenient point in the accrual of women to assess response rate, because the layout and colour of the questionnaire were modified at this stage. Table <xref ref-type="table" rid="T1">1</xref> shows the numbers of questionnaires dispatched and returned, and whether the respondents also gave signed permission for follow up. Overall 121 000 (53%) of the questionnaires sent out were returned to the CEU. Women who returned a questionnaire are referred to as 'respondents', and it is estimated that they comprise about 71% of the women screened at the participating centres. Not all respondents can be included in the cohort to be followed, however, because 7% of them did not give signed consent or gave insufficient personal details for follow up. The remaining 93% of the respondents who can be followed are referred to as 'the cohort' or as 'participants'.</p></sec><sec><title>Characteristics of 121 000 respondents</title><p>Table <xref ref-type="table" rid="T2">2</xref> summarizes certain characteristics of the first 121 000 respondents, including details of their age, use of HRT, reproductive history, past use of oral contraceptives and consumption of cigarettes. It can be seen that most women are aged between 50 and 64 at recruitment (a small number of women are screened just before their 50th birthday and women aged over 65 can be screened by the NHSBSP if they specifically request it). It can also be seen that for most variables there is little missing data. One-third (33%) of the women reported currently using HRT, and almost half (47%) had used it at some time. More than half (54%) had used oral contraceptives and 18% were current smokers.</p><p>Table <xref ref-type="table" rid="T3">3</xref> summarizes the history and family history of breast disease, including breast cancer, in the respondents: 1.4% of the women had breast cancer diagnosed before recruitment and 9% reported that their mother and/or sister(s) had breast cancer diagnosed in the past. Table <xref ref-type="table" rid="T4">4</xref> summarizes the respondents' history of various other illnesses and operations. It can be seen that a substantial proportion of women have had hypertension diagnosed or are being treated for it, that one in four women have had a hysterectomy, one in five have been sterilized and one in 14 have had a bilateral oophorectomy.</p></sec><sec><title>Comparison of participants and nonparticipants</title><p>The overwhelming reason for nonparticipation in the Million Women Study is not attending for breast cancer screening, having been invited to do so. Women are asked to bring the completed questionnaire with them when they are screened, and thus far over 99% of the respondents were recruited in this way. Although no envelope or pre-paid postage is provided, a small number of the respondents posted their questionnaire back to the screening or co-ordinating centre, and virtually all of them also attended for breast cancer screening.</p><p>A direct comparison of those who agreed to participate in the study with those who did not has been performed in one general practice in Oxfordshire and similar comparisons are planned for other areas. A full report of these findings will be published in the future, but preliminary results suggest that there are few substantial differences between participants and nonparticipants. At this stage the main difference between the groups appears to be that nonparticipants are more likely than participants to be prescribed medications for the treatment of hypertension (Banks <italic>et al</italic>, unpublished data).</p><p>About 7% of the respondents returned the study questionnaire but did not give sufficient information and/or signed permission for follow up. Table <xref ref-type="table" rid="T5">5</xref> compares their characteristics with those of the 93% who can be followed. It can be seen that the main difference between these two groups is that the women who gave consent and sufficient information to be followed were more likely to be current users of HRT (33 versus 25%) and to have ever used oral contraceptives (54 versus 45%) than the women who cannot be followed.</p></sec><sec><title>Expected numbers and statistical power</title><p>At the present accrual rate it is expected that a cohort of 1 million women will have been recruited by the year 2000. Based on national statistics from the NHSBSP [<xref ref-type="bibr" rid="B1">1</xref>], about 5000 screen-detected breast cancers would be expected in this cohort. Given these numbers, and the expected proportion of current and never users of HRT at recruitment, the study should have 80% power to detect a relative risk of 1.1 in both current users and in current users of durations of at least 5 years, compared with never users.</p><p>Another aim of the study is to examine the relationship between use of HRT and mortality from various causes, the objective being to present findings with respect to the most important causes of death within 5 years. Table <xref ref-type="table" rid="T6">6</xref> shows the expected numbers of deaths from various causes by the end of 2002, assuming that 1 million women are recruited by the beginning of the year 2000. As with other cohort studies of women taking hormonal agents [<xref ref-type="bibr" rid="B3">3</xref>], it is likely that mortality in these women will be somewhat lower than that the general population because of self-selection of relatively healthy subjects into the study. The expected numbers in Table <xref ref-type="table" rid="T6">6</xref> have, therefore, been calculated assuming that death rates from causes other than breast cancer are 20% lower than the national rate and that breast cancer death rates are 30% lower than the national rate, thus taking into account the additional expected benefit of screening [<xref ref-type="bibr" rid="B4">4</xref>]. It can be seen that by the end of 2002 about 23 000 deaths will have occurred, with the majority being attributed to cancer (12 600 deaths) or to diseases of the circulatory system (8000).</p><p>Previous studies have suggested that both recency and duration of HRT use are important in determining its effect on breast cancer, and perhaps on other diseases [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. However, because women tend to stop taking HRT when they become ill, there are problems in interpreting differences in mortality according to HRT use at the time of death. One way of overcoming these problems is to examine mortality in relation to use of HRT before diagnosis of any serious illness. Analyses of cause-specific mortality in relation to use of HRT within the Million Women Study will, therefore, exclude women with serious illnesses at the time of recruitment and be based on use as recorded at the time of entry into the cohort.</p><p>Table <xref ref-type="table" rid="T6">6</xref> shows the least extreme detectable relative risks for each of the main causes of death to be examined for various patterns of HRT use as compared with never users. These power calculations show that for a common cancer, such as colorectal cancer, there should be sufficient power to detect an increase or decrease in mortality of as little as about 20% in current users compared with in nonusers, and of about 25% in current users of long duration compared with nonusers. Even for endometrial cancer, which is the least common of the causes listed, it should be possible to detect quite modest increases or decreases in mortality of around 40% in current users compared with never users, and of about 45% in current users of long durations compared with never users.</p><p>By the end of the year 2002, the largest numbers of expected deaths among these women will be due to breast cancer and ischaemic heart disease. Thus, the effect of HRT use on deaths from these two causes will be particularly important in determining the net benefit or risk to mortality in HRT users as compared with nonusers. For both of these conditions relative risks of greater than 1.1 or less than 0.9 would be detectable among current versus never users. The corresponding figures among current users with durations of use of 5 or more years are 1.2 and 0.8, respectively.</p></sec><sec><title>Other questions</title><p>Many other questions about women's health can also be answered by this study. The cohort is sufficiently large to provide reliable data on the health effects of many lifestyle factors, including the consumption of tobacco and alcohol, and on the effects of past use of other hormonal agents, such as oral contraceptives. In addition, the Medical Research Council is supporting an extension of the study to evaluate the effect of HRT on the efficacy of mammography. In that study, women recalled for further assessment after screening and women with interval cancers are being identified. Information on interval cancers is being sought from cancer registries and also directly by sending women a follow-up questionnaire 2-3 years after their initial screen and asking about recent morbidity, including diagnosis of breast cancer. This will allow estimation of how HRT affects the sensitivity and specificity of mammography.</p></sec></sec><sec><title>Discussion</title><p>The main purpose of the Million Women Study is to examine the relationship between breast cancer and use of HRT, in a context where use of hormonal therapy is recorded as reliably as possible and breast cancers are diagnosed as uniformly and consistently as possible. Obtaining details of use of HRT before any breast cancer is diagnosed will minimise possible reporting biases of use of such therapy. Moreover, studying screen-detected breast cancers overcomes the potential bias that women who are taking HRT may be more likely to be screened than women who do not use such therapy. The limitation of examining screen-detected cancers alone, however, is that use of HRT may itself reduce the efficacy of mammographic screening. The plan, therefore, is to follow the women screened for interval breast cancer, and to include those cancers in the analyses of the relation between use of HRT and breast cancer.</p><p>Because the entire cohort is being followed up for deaths, it will also be possible to look at the relationship between use of HRT and mortality from various causes. Women prescribed HRT tend to be healthier than those who are not, however, and so it is crucial that analyses take proper account of the so-called 'healthy user effect' [<xref ref-type="bibr" rid="B7">7</xref>]. In designing the study attention has been given to the recording of detailed information about illnesses present at the time of recruitment. It can be seen in Tables <xref ref-type="table" rid="T3">3</xref> and <xref ref-type="table" rid="T4">4</xref> that a substantial proportion of women recruited have had illnesses such as hypertension and other cardiovascular disease in the past that would affect their risk of death from circulatory disease and other causes. The plan is to analyse results separately according to history of previous illness, and most weight will be given to the findings in women who had no previous illness.</p><p>Randomized clinical trials of HRT are now underway. These trials will have sufficient statistical power to detect a substantial reduction in ischaemic heart disease, but will not be able to pick up important, but modest, changes in the risk of cancer [<xref ref-type="bibr" rid="B7">7</xref>]. Thus, there will be a continued need for observational data to look at the effects of HRT on disease.</p></sec><sec><title>Conclusion</title><p>The Million Women Study is one of the largest cohort studies ever devised. Recruitment is proceeding rapidly and the study is on target to accrue a cohort of 1 million women by the year 2000. Preliminary results indicate that the women joining the Million Women Study do not differ substantially from women of a similar age in the general population.</p><p>It is expected that, within 5 years, the study will have sufficient statistical power to answer questions about the role of HRT in mortality from breast cancer and other specific conditions of interest.</p><p>This cohort may ultimately include about one women in every five aged between 50 and 64 years in the UK. This excellent co-operation at a national level reflects the efficient organization of the NHSBSP. It is also indicative, perhaps, of concern by women at the lack of reliable knowledge about the long-term effects of HRT and the fact that in the UK today there is substantial use of this type of therapy.</p></sec><sec><title>Appendix</title><p>NHS Breast Screening Centres that began recruitment before December 1998 (in alphabetical order) are as follows: Avon, Aylesbury, Barnsley, Basingstoke, Bedfordshire &#x00026; Hertfordshire, Cambridge &#x00026; Huntingdon, Chelmsford &#x00026; Colchester, Chester, Cornwall, Crewe, Cumbria, Doncaster, Dorset, East Berkshire, East Cheshire, East Devon, East of Scotland, East Suffolk, Gateshead, Gloucestershire, Great Yarmouth, Hereford &#x00026; Worcester, Kings Lynn, Leicestershire, Liverpool, Manchester, Milton Keynes, Newcastle, North Birmingham, North East Scotland, North Lancashire, North Middlesex, North Nottingham, North of Scotland, North Tees, North Yorkshire, Nottingham, Oxford, Portsmouth, Rotherham, South Birmingham, South East Scotland, South East Staffordshire, Sheffield, Shropshire, Somerset, South Derbyshire, South Essex, South Lancashire, South West Scotland, Surrey, Warrington Halton St Helens &#x00026; Knowsley, Warwickshire Solihull &#x00026; Coventry, West Berkshire, West Devon, West of London, West Suffolk, West Sussex, Wiltshire, Winchester and Wycombe.</p><p>The Million Women Study Co-ordinating Centre staff are as follows: Emily Banks, Valerie Beral, Anna Brown, Diana Bull, Becky Cameron, Barbara Crossley, Diane Deciacco, Dave Ewart, Laura Gerrard, Julie Hall, Sally Hall, Elizabeth Hilton, Ann Hogg, Carol Keene, Nikki Langley, Nicky Langston, Gillian Reeves, Moya Simmonds.</p><p>The Steering Committee members are Joan Austoker, Emily Banks, Valerie Beral, Ruth English, Julietta Patnick, Richard Peto, Gillian Reeves, Martin Vessey and Matthew Wallis.</p><p>The Writing Committee members are Emily Banks, Valerie Beral and Gillian Reeves.</p></sec>

Glutathione S-transferase M1 null genotype: lack of association with tumour characteristics and survival in advanced breast cancer

<sec><title>Background:</title><p>Glutathione S-transferase (GST)M1, a member of the &#x003bc; class GST gene family, has been shown to be polymorphic because of a partial gene deletion. This results in a failure to express the <italic>GSTM1</italic> gene in 50-60% of individuals. Several studies have demonstrated a possible link with the GSTM1-null genotype and susceptibility to cancer. Furthermore, a GSTM1 isoenzyme has been positively associated with protective effect against mutagenic drugs, such as alkylating agents and anthracyclines.</p></sec><sec><title>Objectives:</title><p>To determine whether GSTM1 polymorphisms are associated with tumour characteristics and survival in advanced breast cancer patients, and whether it may constitute a prognostic factor.</p></sec><sec><title>Methods:</title><p>We genotyped 92 patients receiving primary chemotherapy, which included cyclophosphamide, doxorubicine and 5-fluorouracil. The relationships between allelism at GSTM1 and clinicopathological parameters including age, menopausal status, tumour size, grade hormone receptors, involved nodes and p53 gene mutations were analysed. Of the patients with GSTM1-positive genotype, tissue samples obtained before and after treatment were available from 28 cases, allowing RNA extraction and GSTM1 expression by reverse transcription polymerase chain reaction. Relationships with clinical response to chemotherapy, and disease-free and overall survival were also evaluated. The data obtained was analysed using logistic regression to estimate the odds ratio and 95% confidence interval.</p></sec><sec><title>Results:</title><p>Of 92 patients, 57.6% (<italic>n</italic> = 53) were classified as heritably GSTM1-deficient, and 42.4% (<italic>n</italic> = 39) were of the GSTM1-positive genotype. There were no statistically significant relationships between GSTM1-null genotype and the clinicopathological parameters analysed. No relationship was observed between GSTM1 RNA expression and objective clinical response to chemotherapy. Objective clinical response to chemotherapy was related only to clinical tumour size (<italic>P</italic> = 0.0177) and to the absence of intraductal carcinoma (<italic>P</italic> = 0.0013). GSTM1-null genotype had no effect on disease-free or overall survival. The absence of hormone receptors (<italic>P</italic> = 0.002), the presence of a mutated <italic>p53</italic> gene (<italic>P</italic> = 0.0098) and lack of response to primary chemotherapy (<italic>P</italic> = 0.0086) were the only factors associated with reduced disease-free or overall survival.</p></sec><sec><title>Conclusions:</title><p>GSTM1-null genotype alone had no effect on tumour characteristics and outcome of patients with advanced breast cancers. The lack of correlation of GSTM1 genotype with clinical tumour features, clinical response to chemotherapy and survival exclude a role for GSTM1 polymorphism as a prognostic factor in advanced breast cancer.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Lizard-Nacol</surname><given-names>Sarab</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>Slizard@dijon.fnclcc.fr</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Coudert</surname><given-names>Bruno</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Colosetti</surname><given-names>Pascal</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Riedinger</surname><given-names>Jean-Marc</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Fargeot</surname><given-names>Pierre</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Brunet-Lecomte</surname><given-names>Patrick</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>The human glutathione S-transferases (GSTs) are a multigene, isoenzyme family. Cytosolic GST isoenzymes can be classified by their substrate specificities, isoelectric points and amino acid sequence homologies into major classes termed &#x003b1;, &#x003bc;, &#x003c0; and &#x003b8;, which are encoded by a superfamily of genes located at different loci [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. There are currently five putative &#x003b1; class genes encoding subunits GSTA1, GSTA2, GSTA3, GSTA4 and GST&#x003c9;, whereas the GST &#x003c0; class contains a single gene encoding the GSTP1 protein, and the &#x003b8; class consists of two genes encoding the GSTT1 and GSTT2 proteins.</p><p>The <italic>GSTM1</italic> gene belongs to the GST &#x003bc; class gene family, members of which are clustered on chromosome 1p13, and which contains five genes encoding subunits GSTM1, GSTM2, GSTM3, GSTM4 and GSTM5 [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. The presence or absence of the <italic>GSTM1</italic> gene constitutes the polymorphism, and the lack of the <italic>GSTM1</italic> gene, which is caused by a gene deletion (the GSTM1-null genotype), affects approximately 50-60% of the population [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. Homozygosity for the GSTM1-null genotype has been found to confer risk for many cancers, including those of the breast [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. The GSTM1-null genotype was positively associated with high DNA adduct levels, suggesting it has a role in carcinogenesis [<xref ref-type="bibr" rid="B7">7</xref>]. Smokers with a GSTM1 deficiency had a significantly elevated risk for developing lung, laryngeal and bladder cancer [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. The GSTM1-null genotype has also been associated with higher risk for environmentally related cancer, such as cancers of colon, head and neck, skin, oesophagus and stomach [<xref ref-type="bibr" rid="B9">9,</xref><xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>].</p><p>For breast cancer, the GSTM1-null genotype has been found to confer an increased risk in young post-menopausal women [<xref ref-type="bibr" rid="B14">14</xref>], whereas other studies did not find any association [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. GST &#x003bc; deletions have been reported to be associated with higher grade tumours [<xref ref-type="bibr" rid="B15">15</xref>], however, and to confer accumulation of epoxides, which are mutagenic [<xref ref-type="bibr" rid="B16">16</xref>]. GST isoenzymes catalyze the conjugation of glutathione to several electrophilic compounds, including polyaromatic hydrocarbon, which is lipophilic and stored in adipose tissues, such as those of the breast [<xref ref-type="bibr" rid="B20">20</xref>]. Aromatic adducts were found to be higher in women with breast cancer than in healthy control individuals [<xref ref-type="bibr" rid="B21">21</xref>]. Most polyaromatic compounds are metabolized to reactive epoxide intermediates by the polymorphic cytochrome p450 (CYP1AI) and detoxified by phase II enzymes, including GST. The variation in conjugation of epoxide substrate intermediates has been observed to segregate with inherited loss of the <italic>GSTM1</italic> gene. Therefore, individuals who inherit the homozygous form of the null polymorphism in the <italic>GSTM1</italic> gene will not be capable of conjugating and detoxifying specific substrate epoxide intermediates [<xref ref-type="bibr" rid="B16">16</xref>]. In addition, a wide variety of alkylating chemotherapeutic agents used in the treatment of breast cancer have been postulated to act as substrates for the GSTM1 protein products, thus reducing the effectiveness of these agents as cytotoxins [<xref ref-type="bibr" rid="B22">22</xref>].</p><p>In an attempt to further characterize the clinical features associated with the GSTM1-null genotype, we examined allelism at the GSTM1 locus in 92 locally advanced breast cancer patients undergoing primary chemotherapy. Allelism at the GSTM1 locus was analysed by polymerase chain reaction (PCR) in paired samples of blood and breast tissue. To determine whether levels of GSTM1 expression in patients with a positive genotype has a predictive or a prognostic value, RNA expression was measured by reverse transcription (RT)-PCR in breast tumour samples obtained before and after treatment. Results were then compared with clinicopathological factors of the patients, including characterization for <italic>p53</italic> gene mutations [<xref ref-type="bibr" rid="B23">23</xref>], clinical response to chemotherapy, and disease-free and overall survival.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patients and samples</title><p>After an initial diagnostic biopsy, including characterization for <italic>p53</italic> gene alterations, 92 women who were diagnosed with locally advanced breast carcinoma and who underwent primary chemotherapy were included in this study. The median clinical follow up was 78months (range 10-120months). Three of these patients had bilateral lesions, and in these three cases both lesions were examined. No family history for breast cancer was recorded in the 92 women. The patients received chemotherapy treatment (four or six courses, each lasting 21days) with a regimen containing cyclophosphamide, doxorubicine and 5-fluorouracil. The criteria for inclusion were as follows: inflammatory carcinomas, positive nodes and/or large (T3, T4) tumours. Clinical response to primary chemotherapy was categorized according to World Health Organization criteria and was considered as objective response (complete or partial response) or no response (stabilization or progression). In all cases neither radiotherapy nor hormone therapy were applied before chemotherapy.</p><p>Tumours were characterized before treatment by the clinical tumour size (categorized as T2, T3 or T4); the clinical nodal involvement (categorized as N0, N1 or N2); the histologic grade of Scarff, Bloom and Richardson (categorized as SBR1, SBR2 or SBR3); the hormone receptors (categorized as HR^- or HR⁺, and considered as positive for oestrogen and/or progesterone receptors); the pathological tumour size (categorized as pT0, pT1, pT2 or pT3); and the number of involved axillary nodes (categorized as pN^-, pN¹⁺ and pN³⁺ for none, one or two, and three or more involved nodes, respectively).</p><p>Peripheral blood lymphocytes were obtained from each patient. Tumour samples were frozen in liquid nitrogen and stored at -80&#x000b0;C until analysis. Genomic DNA was extracted using proteinase, followed by phenol extraction and ethanol precipitation according to standard procedures [<xref ref-type="bibr" rid="B24">24</xref>].</p></sec><sec><title>Polymerase chain reaction method</title><p>The GSTM1-null genotype was determined by coamplifi-cation with the interferon-&#x003b2; gene, which served as an internal control. Primers for amplification of the GSTM1 gene corresponding to exon 4, intron 5 and exon 5 were 5' -ctgccctacttgattgatggg-3' and 5' -ctggattgtagcagatcatgc-3' (amplified product size, 271 base pairs). Primers for amplification of a part of the interferon-&#x003b2; gene, producing a constant 170-base pair band in all samples, were 5' -ggcacaacaggtagtaggcg-3' and 5' -gccacaggagcttctgacac-3' . Because the primers for the GSTM1 locus anneal to sites inside the coding region of the gene, the presence of the gene was determined by the presence of the band, whereas the null-genotype was determined by the lack of the band, using agarose gel electrophoresis (2%).</p><p>PCR was performed using 250ng template DNA in 10mmol/l Tris-HCl (pH8.4), 50mmol/l potassium chloride, 1.5mmol/l magnesium chloride (Bioprobe Systems, Illkirch, France), 0.2mmol/l concentrations of each deoxynucleotide triphosphate (dNTP), 500nmol/l concentrations of each primer and 2.5units of Taq DNA polymerase (Bioprobe Systems). The reaction (total volume 50 &#x003bc; l) was amplified on a Omnigene thermal cycler (Hybaid Ltd, Ashford, Kent, UK). After an initial denaturation at 94&#x000b0;C for 3 min the reaction proceeded for 25 cycles of 50s at 94&#x000b0;C, 50s at 55&#x000b0;C and 50s at 72&#x000b0;C, concluded by a final extension step of 10min at 72&#x000b0;C. To test for contamination, negative controls (tubes containing the PCR mixture without the DNA template) were included in every run.</p></sec><sec><title>Reverse transcription polymerase chain reaction</title><p>Total RNA was obtained by the acid guanidine thiocyanate-phenol-chloroform extraction method [<xref ref-type="bibr" rid="B25">25</xref>]. Of the total RNA 1 &#x003bc; g was dissolved in 20 &#x003bc; l reverse transcriptase buffer (Gibco/BRL, Cergy Pontoise, France) containing 200 &#x003bc; mol/l dNTP, 500 ng random hexamer and 200U avian myeloblastosis virus (AMV) reverse transcriptase (Superscript; Gibco/BRL). The reaction was incubated at 42&#x000b0;C for 50min and heated to 70&#x000b0;C for 15min. </p><p>The PCR reaction was carried out in a 50 &#x003bc; l volume of PCR buffer (Promega, Madison, Wisconsin, USA) containing 200 &#x003bc; mol/l dNTP, 2.5 &#x003bc; l complementary DNA template, 2.5U Taq DNA polymerase, and 500nmol/l of both <italic>GSTM1</italic> and <italic>&#x003b2;</italic>₂- microglobulin primers. The reaction was initiated by a heat step at 95&#x000b0;C for 2min, and carried out for 25 cycles of denaturation at 94&#x000b0;C for 50s, annealing at 55&#x000b0;C for 50s and extension at 72&#x000b0;C for 50s. Blanks for each reaction were included with all samples. Of PCR product 10 &#x003bc; l were analysed on a 2% agarose gel with ethidium bromide staining. Band intensities were determined with a gel doc 1000 UV system (Biorad, Ivry-Sur-Seine, France), and the ratio of <italic>GSTM1</italic> to <italic>&#x003b2;</italic>₂- microglobulin was calculated.</p></sec><sec><title>Determination of p53 mutations</title><p>The determination of <italic>p53</italic> mutations was performed as previously described [<xref ref-type="bibr" rid="B23">23</xref>]. Briefly, breast tumour samples were characterized before and after treatment for <italic>p53</italic> gene mutations by PCR single-strand confirmation polymorphism and/or direct sequencing of exons 5-9 of the <italic>p53</italic> gene.</p></sec><sec><title>Hormonal receptor assay</title><p>The oestrogen and progesterone receptor levels were determined in cytosolic tumours using enzyme immunoassay methods (Abbott Laboratories, Rungis, France). The cutoff level used for oestrogen and progesterone was 20fmol/mg cytosolic proteins.</p></sec><sec><title>Statistical analysis</title><p>The Pearson &#x003c7;² test was used as a homogeneity test for proportion. A stepwise logistic regression model with the logoistic regression (LR)-BMDP program (University of California Press, Berkeley, California, USA) [<xref ref-type="bibr" rid="B26">26</xref>] was used to assess the contribution of each independent factor to the GSTM1-null genotype and the probability of response to primary chemotherapy. The log-rank test using the Kaplan-Meier method was used to study the relationship between each factor and the probability of disease-free survival (median 50months) and overall survival (median 78months). A multivariate analysis using the Cox proportional hazards model was performed to assess the contribution of each independent factor to the probability of survival. For logistic regression or Cox regression models the enter and remove limits were 0.1 and 0.15, respectively. Overall significance of each factor (<italic>P</italic> value) was given by the likelihood ratio test. Results are expressed as odds ratios (OR) and corresponding 95% confidence interval (CI), with the significance of ORs being derived from the ratio of the coefficient divided by its standard error (Wald test). Statistical analysis for RNA <italic>GSTM1</italic> expression was performed using the SAS/STAT program (SAS Institute Inc, Cary, North Carolina, USA) by using student <italic>t</italic>-test or analysis of variance, Pearson &#x003c7;² test and log-rank test. For all tests, optimality of the selected models was verified by all-possible-subsets analyses.</p></sec></sec><sec><title>Results</title><sec><title>Clinicopathological data</title><p>Patients (total 92) were characterized by two variables: age and menopausal status at diagnosis. The median age of these patients was 57 years, and status was premenopausal for 51% (<italic>n</italic> = 47) and postmenopausal for 48.9% (<italic>n</italic> = 45). All samples were composed of infiltrating ductal carcinoma with an area of intraductual carcinoma in 35 cases. Of the tumours, 33.7% (<italic>n</italic> = 31) were 2-5cm in size (T2), 27.2% (<italic>n</italic> = 25) were &#x0003e; 5cm (T3) and 39.1% (<italic>n</italic> = 36/92) were &#x0003e; 5cm with skin involvement (T4). Of tumours, 60.9% (<italic>n</italic> = 56) contained significant hormone receptors (HR⁺), and 39.1% (<italic>n</italic> = 36) did not (HR^-). Clinical complete response was found in 19.6% (<italic>n</italic> = 18) of cases. A high histological grade (SBR3) was found in 34.8% (<italic>n</italic> = 32) of tumours. An intermediate level of pathological tumour size (pT2) was found in 42.4% (<italic>n</italic> = 39) of samples. Negative lymph nodes were found in 20.7% (<italic>n</italic> = 19) of the tumours, whereas 23.9% (<italic>n</italic> = 22) of tumours were associated with &#x0003e;1 involved nodes and 55.4% (<italic>n</italic> = 51/92) were associated with &#x0003e; 3 involved nodes. Mutations in <italic>p53</italic> were detected in 30% (<italic>n</italic> = 28) of the cases studied.</p></sec><sec><title>Glutathione S-transferase M1 genotype determination</title><p>The PCR method described above allowed an internal standard controlled classification of GSTM1-deficient (GSTM1-null genotype) individuals. Of the 92 patients, 57.6% (<italic>n</italic> = 53) were classified as heritably GSTM1 deficient, and 42.4% (<italic>n</italic> = 39) were GSTM1-positive genotype. Paired samples of blood and breast tissue were analysed before treatment with primary chemotherapy from the same individual, and GSTM1 genotype was identical for the two samples (Fig. <xref ref-type="fig" rid="F1">1</xref>). Among the 39 patients with GSTM1-positive genotype, tissue samples obtained before and after treatment were available from 28 cases, allowing RNA extraction and <italic>GSTM1</italic> expression using the RT-PCR method. Two of these patients had bilateral lesions, and measurement was determined in the two tumour localizations. Thus, total <italic>GSTM1</italic> expression, as measured by the ratio of <italic>GSTM1</italic> to <italic>&#x003b2;</italic>₂- microglobulin values, were performed on 30 tumour specimens. <italic>GSTM1</italic> RNA signal was detected in all of the tumours analysed before and after treatment. The median <italic>GSTM1</italic> expression was 1.38 (range 0.02-23.27) in the untreated tumours, and 1.16 (range 0.01-6.56) in samples obtained after chemotherapy administration.</p></sec><sec><title>Glutathione S-transferase M1 and clinicopathological characteristics of the patients</title><p>Distribution of GSTM1 genotype and its relation with clinicopathological data of the patients are shown in Table <xref ref-type="table" rid="T1">1</xref>. There were no statistically significant associations between GSTM1-null genotype and the parameters analysed: age, menopausal and hormonal status, clinical and pathological tumour size, grade, involved nodes and p53 gene mutations. <italic>GSTM1</italic> expression measured by RT-PCR in 30 samples (corresponding to 28 cases) before and after treatment with primary chemotherapy was also compared with the clinicopathological characteristics of the patients. None of the parameters tested were related to <italic>GSTM1</italic> expression determined before or after treatment (data not shown).</p><p>Relationship to clinical response to chemotherapy (Table <xref ref-type="table" rid="T2">2</xref>) demonstrated that objective response (complete and partial responses) rate of the group with GSTM1-null genotype (75.5%) did not differ from that in those with GSTM1-positive genotype (76.9%). Thus, no significant relation was found between GSTM1 polymorphism and clinical response to chemotherapy (<italic>P</italic> = 0.8719). Also, no relation was obverved between <italic>GSTM1</italic> RNA expression and clinical response to chemotherapy (<italic>P</italic> = 0.9524 and <italic>P</italic> = 0.5192 for before and after treatment, respectively). In contrast, clinical tumour size (<italic>P</italic> = 0.0177) and intraductal carcinoma (<italic>P</italic> = 0.0013) are strongly associated with clinical response. In multivariate analysis, the clinical tumour size (<italic>P</italic> = 0.0070, OR = 4.83, 95% CI = 1.45-16.10) and the absence of intraductal carcinoma (<italic>P</italic> = 0.0002, OR = 14.1, 95% CI = 2.52-78.50) remained the only factors linked to the clinical response.</p></sec><sec><title>Impact on survival of the patients</title><p>For disease-free survival, no differences were found between individuals with GSTM1-null genotype and those with positive-GSTM1 genotype (<italic>P</italic> =0.8094). Accordingly, no impact for RNA <italic>GSTM1</italic> expression on disease-free survival (<italic>P</italic> = 0.8991 and <italic>P</italic> = 0.9096 for before and after treatment, respectively) was observed (Table <xref ref-type="table" rid="T2">2</xref>). In contrast, the absence of hormone receptors (<italic>P</italic> = 0.0020) and the presence of <italic>p53</italic> gene mutations (<italic>P</italic> = 0.0098) had an impact on disease-free survival. With multivariate analysis, hormone receptors status (<italic>P</italic> = 0.0002, OR=3.99, 95% CI = 1.92-8.29) and <italic>p53</italic> gene mutations (<italic>P</italic> = 0.0138, OR = 2.36, 95% CI=1.22-4.59) remained significantly associated with metastasis recurrence risk.</p><p>No impact was also found (<italic>P</italic> = 0.9729) for GSTM1-null genotype on overall survival (Table <xref ref-type="table" rid="T2">2</xref>), or for RT-PCR RNA expression (<italic>P</italic> = 0.1667 and <italic>P</italic> = 0.9637 for before and after treatment, respectively). Only the absence of hormone receptors (<italic>P</italic> = 0.0018), the presence of <italic>p53</italic> gene mutations (<italic>P</italic> = 0.0071) and no response to primary chemotherapy (<italic>P</italic> = 0.0086) were associated with reduced overall survival of the patients. In multivariate analysis, hormone receptors status (<italic>P</italic> = 0.0003, OR = 5.23, 95% CI = 2.03-13.49) and <italic>p53</italic> gene mutations (<italic>P</italic> = 0.0037, OR = 3.62, 95% CI = 1.53-8.53) were strongly related to the risk for death. Absence of clinical response to chemotherapy was less related to the overall survival (<italic>P</italic> = 0.0530, OR = 2.31, 95% CI = 1.02-5.26).</p></sec></sec><sec><title>Discussion</title><p>In the present study, the frequency of <italic>GSTM1</italic> gene deficiency among breast carcinoma patients (57%) was similar to that previously reported in this lesion [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Among individuals with GSTM1-positive genotype no deletion was observed in somatic tumour cells, suggesting that the deletion of <italic>GSTM1</italic> gene was not a characteristic of breast tumour cells.</p><p>Several series have determined that <italic>GSTM1</italic> deletions were involved in the aetiology of breast cancer [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B17">17</xref>], whereas another studies did not find any such associations [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. In addition, little is known about clinical characteristics that confer <italic>GSTM1</italic> deletion among breast cancer patients. Only one recently published study [<xref ref-type="bibr" rid="B19">19</xref>] reported no relation between clinicopathological parameters and GSTM1 genotype in primary breast tumours. The results obtained in the present study are consistent with these data, because no differences were noted when a variety of known prognostic factors were compared between tumours from patients with and those without GSTM1-null genotype. These factors include age, menopausal status, clinical and pathological tumour size, clinical and pathological nodal involvement, pathological grade, hormone receptors status and <italic>p53</italic> gene mutations. These results suggest that clinical tumour features are not associated with GSTM1-null genotype, not only in primary breast cancers, but also in advanced breast cancers.</p><p>An improved survival has been observed in patients with GSTM1-null genotype and has been explained by a better response to chemotherapeutic agents related to more effective cell killing, which in turn is related to the absence of protective effect of GSTM1 allele [<xref ref-type="bibr" rid="B18">18</xref>]. Among the most active chemotherapeutic agents in the treatment of breast cancer are cyclophosphamide and anthracyclines. These compounds are conjugated with thiols through reactions mediated by GST. The &#x003c0; class of transferase is thought to be the major factor involved in the subsequent prevention of DNA alkylation [<xref ref-type="bibr" rid="B22">22</xref>], although there has been some suggestion that the &#x003bc; class is also involved [<xref ref-type="bibr" rid="B27">27</xref>]. In order to determine whether the absence of GSTM1 locus favours the response to chemotherapeutic agents, and whether, on the contrary, GSTM1-positive genotype may confer a resistance to chemotherapy, we analysed GSTM1 polymorphisms in relation to clinical response to chemotherapy. The results demonstrated that the objective response rate of the group with GSTM1-null genotype did not differ from those with GSTM1-positive genotype. Thus, individuals carrying the GSTM1 locus were no more resistant to chemotherapy than those with GSTM1-null genotype. <italic>GSTM1</italic> RNA expression levels measured in patients with positive genotype were also not associated with clinical response to chemotherapy. Among clinical parameters analysed, only clinical tumour size and the presence of intraductal carcinoma were found to influence clinical response to primary chemotherapy. The results also revealed that GSTM1-null genotype was not related to survival in advanced breast cancer patients, in contrast to the absence of hormone receptors and to the presence of <italic>p53</italic> gene mutations, which are known to be of prognostic value in advanced breast cancer [<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>].</p><p>Although the group studied was not very large (<italic>n</italic> = 92), the clinical characteristics were well known and the results strongly suggest the lack of correlation of GSTM1-null genotype alone with prognostic factors, clinical reponse to chemotherapy and survival. Genotypes of other enzymes (CYP1A1, GSTT1), as well as combinations of genotypes, might be of interest with regard to cancer and carcinogen metabolism.</p><p>Taken together, these results show that advanced breast cancers arising in patients with GSTM1-null genotype have no worse clinical tumour charactaristics and outcome than those of patients without such deletions. Thus, the lack of correlation of GSTM1-null genotype with clinical tumour features, clinical response to chemotherapy and survival exclude a role for GSTM1 polymorphism as a prognostic factor in advanced breast cancer.</p></sec>

Cyclin D₁ expression during rat mammary tumor development and its potential role in the resistance of the Copenhagen rat

<sec><title>Background:</title><p>Resistance to mammary tumorigenesis in Copenhagen rats is associated with loss of early preneoplastic lesions known as intraductal proliferations. The cause of this disappearance, however, is unknown.</p></sec><sec><title>Results:</title><p>There were no differences in the numbers of lesions in mammary whole-mounts prepared from Copenhagen or Wistar-Furth rats at 20 or 30 days after <italic>N-</italic>methyl-<italic>N</italic>-nitrosourea treatment, but at 37 days there were significantly fewer lesions in Copenhagen glands. Furthermore, lesions in Copenhagen glands were exclusively intraductal proliferations, whereas in Wistar-Furth glands more advanced lesions were also present. Immunohistochemical staining showed frequent cyclin D₁ overexpression in Wistar-Furth lesions at 37 days, but not in Copenhagen lesions. There were, however, no differences in p16^{<italic>INK4a</italic>} protein expression, bromodeoxyuridine labeling and apoptotic indices, or mast cell infiltration between Copenhagen and Wistar-Furth lesions at any time.</p></sec><sec><title>Conclusions:</title><p>Overexpression of cyclin D₁ in preneoplastic lesions may be important in the development of mammary tumors in susceptible rats, although this overexpression does not appear to cause significant changes in cell kinetics. Furthermore, the low levels of cyclin D₁ expression in Copenhagen intraductal proliferations may play a role in the resistance of these rats to mammary tumorigenesis.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Korkola</surname><given-names>James E</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Wood</surname><given-names>Geoffrey A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Archer</surname><given-names>Michael C</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>m.archer@utoronto.ca</email></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>Most strains of rats develop multiple mammary tumors when initiated with chemicals or radiation. Several strains, however, are resistant to mammary tumorigenesis induced by both of these means. The Copenhagen rat is the best characterized of these strains [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>], although the mechanism of resistance is still unknown. Recently, linkage analysis has identified at least four loci that modify mammary tumorigenesis in the Copenhagen rat, but the genes have yet to be cloned [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>In order to characterize the phenotype associated with resistance, we recently examined mammary whole-mounts from both Copenhagen and susceptible Wistar-Furth rats at various times after treatment with the mammary carcinogen <italic>N</italic>-methyl-<italic>N</italic>-nitrosourea (MNU) [<xref ref-type="bibr" rid="B4">4</xref>]. At 15 days after MNU treatment, we found that both strains had developed the earliest detectable preneoplastic lesions, known as intraductal proliferations (IDPs). The majority of IDPs from both strains contained activating mutations in the <italic>Ha-ras</italic> oncogene, a common alteration in MNU-induced rat mammary adenocarcinomas [<xref ref-type="bibr" rid="B5">5</xref>]. By 45 days after MNU treatment, in addition to IDPs more advanced lesions such as ductal carcinomas <italic>in situ</italic> (DCIS) and adenocarcinomas were detectable in the glands from Wistar-Furth rats. In contrast, the IDPs from Copenhagen rats failed to progress and instead declined in number, such that by 60 days after MNU treatment the glands were essentially free of lesions.</p><p>To investigate a potential mechanism that could explain the failure of the Copenhagen IDPs to progress and their subsequent disappearance, we have examined the expression of cyclin within IDPs and other lesions from D₁ Copenhagen and Wistar-Furth rats. Cyclin D₁ has been shown to be important in the transition from the G₁ to the S phase of the cell cycle, and perturbations in this control point can lead to neoplastic transformation [<xref ref-type="bibr" rid="B6">6</xref>]. Indeed, cyclin D₁ is frequently overexpressed in both human [<xref ref-type="bibr" rid="B7">7</xref>] and rat mammary tumors [<xref ref-type="bibr" rid="B8">8</xref>], and is thought to be an important factor in their development. This notion was strengthened by studies that showed that mice engineered to overexpress cyclin D₁ in their mammary glands develop hyperplastic lesions and eventually mammary carcinomas [<xref ref-type="bibr" rid="B9">9</xref>]. Overexpression of cyclin D₁ may be an important event in determining whether preneoplastic lesions go on to develop into malignant or benign lesions in humans and is of particular relevance to the present study [<xref ref-type="bibr" rid="B10">10</xref>]. </p><p>In addition to cyclin D₁, we also chose to examine expression of the p16^{<italic>INK4a</italic>} protein in the lesions, because it is a specific inhibitor of the cyclin D₁-cdk4 complex that drives the transition from G₁ to S in the cell cycle [<xref ref-type="bibr" rid="B11">11</xref>]. Expression of p16^{<italic>INK4a</italic>} in normal cells is thought to lead to a growth arrest [<xref ref-type="bibr" rid="B11">11</xref>]. In order to relate changes in the expression of these genes to changes in cell kinetics within the lesions, we used bromodeoxyuridine to label cells during the S phase as a measure of the proliferative index and counted apoptotic cells based on morphology to estimate cell loss. Finally, we stained lesions for mast cells, because they have been implicated in promoting the growth of IDPs [<xref ref-type="bibr" rid="B12">12</xref>].</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Animals and carcinogen treatment</title><p>Copenhagen and Wistar-Furth rats (6-7 weeks old) were purchased from Harlan Sprague Dawley (Indianapolis, Indiana, USA), maintained on a 12h light/dark cycle, fed Harlan Teklad rat chow (6% fat; Harlan Teklad, Madison, Wisconsin, USA), and were given free access to water. After 1 week of acclimatization the rats were given an intraperitoneal injection of 50 mg/kg MNU dissolved in acidified normal saline.</p></sec><sec><title>Bromodeoxyuridine treatment and mammary whole-mount preparation</title><p>At 20, 30, and 37 days after MNU treatment, five rats from each strain selected randomly were given an intraperitoneal injection of 50 mg/kg bromodeoxyuridine (Boehringer, Laval, Canada) dissolved in phosphate-buffered saline. Three hours later, they were killed and mammary whole-mounts prepared, using the technique we described previously [<xref ref-type="bibr" rid="B4">4</xref>].</p></sec><sec><title>Paraffin embedding, staining, and immunohistochemistry</title><p>Putative lesions in the whole-mounts were microdissected from the glands, cleared in xylenes, processed through three changes of paraffin wax, and then embedded in paraffin wax (Fisher, Whitby, Canada) for sectioning. Sections (4&#x003bc; m thick) were placed on poly-L-lysine (Sigma, St Louis, Missouri, USA) coated slides and stained with hematoxylin and eosin. Positive identification of IDPs, DCIS, and adenocarcinomas was based on the criteria we used previously [<xref ref-type="bibr" rid="B4">4</xref>]. Serial sections from confirmed lesions were then used for cyclin D₁, p16^{<italic>INK4a</italic>}, and bromodeoxyuridine immunohistochemistry using established techniques [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. Anti-cycin D₁ and anti-p16^{<italic>INK4a</italic>} antibodies were obtained from Santa Cruz (Santa Cruz, California, USA) and anti-bromodeoxyuridine antibodies from Boehringer. Archival rat mammary tumor tissue was used as a positive control for cyclin D₁, because its overexpression has been reported in these tumors [<xref ref-type="bibr" rid="B8">8</xref>]. The levels of cyclin D₁ in the stained samples were scored as negative (-), low (+), or high (++), based on number of positive cells in the lesion as well as staining intensity. Because the measurement of staining intensity was somewhat subjective, the coded samples were also scored independently by a second individual, with identical results. The percentage of cyclin D₁-positive cells was determined as the number of positive cells divided by total cell number in a lesion.</p><p>The bromodeoxyuridine labeling index was determined by the number of bromodeoxyuridine-positive cells divided by total cells in a lesion. Small intestine from bromodeoxyuridine-treated rats or livers from partially hepatectomized rats were used as positive controls for staining.</p><p>For all immunohistochemistry, the specificity of the staining was ensured by replacing the primary antibody with 1% normal sheep serum. In all cases, no staining was observed.</p><p>For determination of apoptotic indices, bromodeoxyuridine-stained sections were also scored for apoptotic cells based on their morphology (pyknotic nuclei, cell shrinkage) [<xref ref-type="bibr" rid="B16">16</xref>]. Because this is a subjective method, samples were scored independently by two individuals, with similar results.</p><p>For staining of mast cells, samples were deparaffinized in xylenes, rehydrated through acetone and water, stained in 0.025% toluidine blue (Sigma) for 30s. Slides were washed in distilled water, dehydrated in acetone, cleared in xylenes, and mounted using Permount (Fisher). Mast cells were counted per high power (400&#x000d7;) field of view around the lesion.</p></sec><sec><title>Statistical analyses</title><p>For comparison of numbers of lesions, bromodeoxyuridine-labeling indices, and apoptotic indices at 20, 30, and 37 days after MNU treatment, <italic>t</italic>-tests using Bonferroni's correction were used. The data were also analyzed by square root transformation followed by <italic>t</italic>-tests using Bonferroni's correction. For comparison of cyclin D₁ staining in Copenhagen and Wistar-Furth IDPs, a &#x003c7;² test was used, with the groups being IDPs that do not overexpress cyclin D₁(- or +) and IDPs that do overexpress cyclin D₁ (++). For comparison of percentages of cyclin D₁-positive cells in Copenhagen and Wistar-Furth IDPs at day 37, a one-tailed <italic>t</italic>-test was used.</p></sec></sec><sec><title>Results</title><p>To examine the expression of cyclin D₁ and p16^{<italic>INK4a</italic>} proteins within lesions from Copenhagen and Wistar-Furth rats, mammary whole-mounts were prepared at 20, 30, and 37 days after MNU treatment. In order to estimate proliferative indices within the lesions, all rats were administered bromodeoxyuridine before killing. As expected, at 20 and 30 days after MNU treatment the number of lesions in Copenhagen rats was not different from that in Wistar-Furth rats (Fig. <xref ref-type="fig" rid="F1">1</xref>). By 37 days after MNU, however, there were significantly fewer lesions in the glands of Copenhagen rats (Fig. <xref ref-type="fig" rid="F1">1</xref>). Furthermore, we observed only IDPs in the glands of Copenhagen rats, whereas more advanced lesions such as ductal carcinomas <italic>in situ</italic> (DCIS) and small, nonpalpable tumors were also present in the glands of Wistar-Furth rats at 37 days; this is consistent with our previous results [<xref ref-type="bibr" rid="B4">4</xref>]. Figure <xref ref-type="fig" rid="F2">2</xref> shows the same region of the inguinal mammary gland from typical whole-mounts from a Wistar-Furth and a Copenhagen rat, demonstrating the striking difference in development of lesions in the glands at 37 days.</p><p>We determined cyclin D₁ expression immunohistochemically in sections from lesions (Figs <xref ref-type="fig" rid="F3">3a</xref><xref ref-type="fig" rid="F3">3b</xref><xref ref-type="fig" rid="F3">3c</xref>). The staining levels were characterized as negative (-), low (+), or high (++). We observed no staining in either Wistar-Furth or Copenhagen IDPs at 20 or 30 days after MNU treatment, or in any normal mammary tissues. At 37 days, however, there was cyclin D₁ staining in 10 out of 17 Wistar-Furth IDPs, with six of these showing high levels of expression (overexpression), as shown in Figure <xref ref-type="fig" rid="F3">3b</xref>. In contrast, only three out of nine IDPs from Copenhagen rats showed any cyclin D₁ staining, with all of these being at a low level (Fig. <xref ref-type="fig" rid="F3">3a</xref>). A &#x003c7;² analysis showed that overexpression of cyclin D₁ was significantly higher in Wistar-Furth IDPs than in Copenhagen IDPs (<italic>P</italic> &#x0003c; 0.05). Furthermore, we stained the few advanced lesions present in Wistar-Furth glands at 37 days for cyclin D₁, and observed overexpression in four out of five DCIS and in all of three nonpalpable adenocarcinomas (Fig. <xref ref-type="fig" rid="F3">3c</xref>); this is in good agreement with the published observations that approximately 80% of rat mammary tumors overexpress cyclin D₁[<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B17">17</xref>].</p><p>As an additional measure, we determined the mean percentage of cells that expressed cyclin D₁ within lesions at 37 days. We found that 5.9 &#x000b1; 3.4% of cells within Copenhagen IDPs stained for cyclin compared with D₁, 17.5 &#x000b1; 4.0% for Wistar-Furth IDPs, 22.6 &#x000b1; 7.5% for Wistar-Furth DCIS, and 32.1&#x000b1; 2.1% for Wistar-Furth adenocarcinomas (all values are means &#x000b1; standard error of the mean). Statistical analysis of Copenhagen and Wistar-Furth IDPs by <italic>t</italic>-test showed that the percentage of cells that expressed cyclin D₁ was significantly higher in the Wistar-Furth lesions (<italic>P</italic> &#x0003c; 0.05). It should be noted that it is not possible to perform Western analysis to confirm the cyclin D₁ expression levels. Suspected lesions must be microdissected from the whole-mounts, then embedded and sectioned to confirm their identity, leaving insufficient tissue for Western analysis.</p><p>Next, we determined p16^{<italic>INK4a</italic>} expression in the sections by immunohistochemistry (Figs <xref ref-type="fig" rid="F3">3d</xref><xref ref-type="fig" rid="F3">3e</xref><xref ref-type="fig" rid="F3">3f</xref>). We observed similar levels of staining for this protein in all of the samples from both strains, including normal mammary tissue, IDPs, DCIS, and tumors.</p><p>To measure the proliferative index of IDPs from both strains we stained samples using an anti-bromodeoxyuridine antibody as shown in Figures <xref ref-type="fig" rid="F3">3g</xref><xref ref-type="fig" rid="F3">3h</xref><xref ref-type="fig" rid="F3">3i</xref>. The labeling index was determined as the number of bromodeoxyuridine-positive cells divided by the total number of cells in the lesion. The labeling indices in the Copenhagen IDPs were not different from those in Wistar-Furth rats at either 20 or 30 days after MNU treatment when there was no cyclin D₁ overexpression in either strain (Fig. <xref ref-type="fig" rid="F4">4</xref>). At 37 days, when we observed high levels of cyclin D₁ staining in Wistar-Furth lesions but not in Copenhagen lesions, there were also no differences in the bromodeoxyuridine labeling indices between the two strains (Fig. <xref ref-type="fig" rid="F4">4</xref>). Furthermore, there was no significant correlation between labeling indices and cyclin D₁ expression levels in lesions from Wistar-Furth rats at this time.</p><p>In the same lesions that we determined the bromodeoxyuridine labeling indices, we also counted apoptotic cells based on their morphology. The apoptotic indices are shown in Figure <xref ref-type="fig" rid="F5">5</xref>. There were no significant differences between the Copenhagen and Wistar-Furth rats at 20, 30, or 37 days after MNU treatment.</p><p>Toluidine blue, which stains mast cells metachromatically, was used to visualize these cells within sections. Samples were scored for the number of mast cells per high power field of view around each lesion. There were 3.6 &#x000b1; 0.5, 2.8 &#x000b1; 0.3, and 6.4 &#x000b1; 0.7 mast cells around Copenhagen IDPs, and 4.3 &#x000b1; 0.5, 2.4 &#x000b1; 0.5, and 6.6 &#x000b1; 0.8 mast cells around Wistar-Furth IDPs at 20, 30, and 37 days after MNU, respectively (all values are means &#x000b1; standard error of the mean). There were no significant differences in mast cell numbers between the two strains at any of the time points.</p></sec><sec><title>Discussion</title><p>Overexpression of cyclin has been reported in both D₁ human [<xref ref-type="bibr" rid="B7">7</xref>] and rat mammary tumors [<xref ref-type="bibr" rid="B8">8</xref>]. It has recently been shown [<xref ref-type="bibr" rid="B10">10</xref>] that cyclin D₁ overexpression might be a critical early event in human breast tumor development, because overexpression of this gene is common in early lesions that ultimately form malignant breast cancers, but not in those that form benign tumors. It is thought that rat mammary tumorigenesis occurs through the progression of the early IDPs to DCIS and eventually to adenocarcinomas [<xref ref-type="bibr" rid="B18">18</xref>]. Recently, cyclin D₁ expression has been investigated in normal mammary tissue, preneoplastic lesions, and tumors in a susceptible strain of rat [<xref ref-type="bibr" rid="B17">17</xref>]. The percentage of cyclin D₁-positive cells was shown to be very low (approximately 2.4%) in normal mammary tissue. In IDPs, however, approximately 13.6% of cells were positive, and this value increased with each subsequent stage of tumorigenesis such that approximately 40% of cells within adenocarcinomas were positive. We reasoned that if cyclin D₁ overexpression is an early event that is necessary for tumorigenesis in the rat mammary gland, then differences in the expression of this gene in Wistar-Furth and Copenhagen rats could account for their different susceptibilities to mammary tumorigenesis. At 37 days after MNU treatment, when there were significantly more IDPs in Wistar-Furth than in Copenhagen glands, we observed cyclin D₁ overexpression only in Wistar-Furth IDPs. This overexpression was manifested as staining that was both more frequent and more intense than in IDPs from Copenhagen rats. We also found that the percentage of cyclin D₁-positive cells within Wistar-Furth IDPs was significantly higher than in Copenhagen IDPs at day 37. Furthermore, seven out of the eight DCIS and adenocarcinomas that had developed by this time in Wistar-Furth rats showed highly overexpressed cyclin D₁ relative to normal tissue. Both the DCIS and adenocarcinomas had higher percentages of cyclin D₁-positive cells than did IDPs, although the number of advanced lesions present at this time was too few to demonstrate this difference statistically. It should be noted that our values for Wistar-Furth lesions are in good agreement with those reported by Zhu <italic>et al</italic> [<xref ref-type="bibr" rid="B17">17</xref>]. Because cyclin D₁ protein levels are higher in DCIS and adenocarcinomas than in IDPs, overexpression of this gene might be important in the transition from precancerous to cancerous lesions. Furthermore the lack of cyclin D₁ overexpression in Copenhagen IDPs may play a role in their inability to progress to DCIS and tumors, a notion supported by our observation of only a single DCIS in a total of 31 MNU-treated Copenhagen rats from this and our previous study [<xref ref-type="bibr" rid="B4">4</xref>].</p><p>Transition from the G1 to S phase of the cell cycle is tightly regulated within cells. Activity of the cyclin D₁-cdk4 complex that drives this transition can be blocked by the p16^{<italic>INK4a</italic>} protein, leading to growth arrest [<xref ref-type="bibr" rid="B11">11</xref>]. In tumors, a sustained blockage induced by p16^{<italic>INK4a</italic>} may lead to apoptosis [<xref ref-type="bibr" rid="B19">19</xref>]. Loss of the G₁-S checkpoint control can occur by a variety of means, including loss of p16^{<italic>INK4a</italic>} or overexpression of cyclin D₁ [<xref ref-type="bibr" rid="B20">20</xref>]. Our staining showed that p16^{<italic>INK4a</italic>} was expressed in all samples, but, as described above, cyclin D₁ was overexpressed only in Wistar-Furth lesions at day 37. Therefore, we expected that the Wistar-Furth IDPs would have a higher labeling index or lower apoptotic index than those of Copenhagen rats. As expected, at 20 and 30 days after MNU treatment, when there were no differences in cyclin D₁ expression in IDPs or in number of IDPs between the two strains, we found no differences in the bromodeoxyuridine labeling or apoptotic indices. Surprisingly, at 37 days we found no significant difference in the bromodeoxyuridine labeling indices in IDPs from Copenhagen compared with Wistar-Furth rats, indicating that there was no correlation between cyclin D₁ overexpression and cell proliferation. Other studies have also found that cyclin D₁ overexpression does not correlate with the proliferation rate in rat mammary tumors [<xref ref-type="bibr" rid="B8">8</xref>] or in human tumors [<xref ref-type="bibr" rid="B10">10</xref>]. This indicates that cyclin D₁ overexpression may play other roles in tumorigenesis that are unrelated to the cell cycle. Indeed, it has been reported that cyclin D₁ can transactivate the estrogen receptor and influence genomic stability [<xref ref-type="bibr" rid="B21">21</xref>].It has also been found that cyclin D₁ overexpression can inhibit apoptosis [<xref ref-type="bibr" rid="B21">21</xref>]. There was no difference, however, in the apoptotic indices at 37 days between the strains.</p><p>Cyclin D₁ may provide a promotional stimulus for Wistar-Furth IDPs, but we were unable to detect an alteration in cell kinetics. It is possible, however, that small perturbations in the rates of cell loss and/or cell growth may occur that would be undetectable in short-term assays. Such changes could have profound effects over the long period of tumor development. Indeed, our observation that preneoplastic lesions disappear from the glands of Copenhagen rats indicates that cell loss may be occurring, although redifferentiation of preneoplastic cells to a more normal phenotype is also plausible, as we have previously hypothesized [<xref ref-type="bibr" rid="B4">4</xref>].</p><p>It has been postulated by Russo and Russo [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B22">22</xref>] that there are two populations of IDPs. Initiated plus promoted IDPs are able to form more advanced lesions such as DCIS and tumors, whereas the IDPs that are only initiated are unable to progress [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. Those authors distinguished initiated plus promoted IDPs from initiated IDPs by the infiltration of mast cells, which are three times more abundant around the former. They postulated that these mast cells may be involved in promoting the growth of lesions, by the secretion of either mitogenic or angiogenic factors. If mast cells are more abundant surrounding Wistar-Furth than Copenhagen IDPs, then secretion of mitogenic factors could lead to overexpression of cyclin D₁ in the former. We found, however, that there were no differences in the numbers of mast cells surrounding IDPs of the two strains at any time point. It seems unlikely, therefore, that mast cell infiltration plays a role in either cyclin D₁ overexpression or in the resistance of the Copenhagen rat.</p><p>It is unclear what mechanism is responsible for the overexpression of cyclin D₁ we have observed. It has been reported that the <italic>ras</italic> oncogene can induce expression of cyclin D₁ [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>], but it is unlikely that this is involved, because we have previously shown [<xref ref-type="bibr" rid="B4">4</xref>] that similar percentages of Copenhagen and Wistar-Furth IDPs harbor mutant <italic>Ha-ras</italic> alleles. Recently, Tetsu and McCormick [<xref ref-type="bibr" rid="B25">25</xref>] have shown that expression of cyclin D₁ can also be regulated through the actions of transcription factors controlled by the &#x003b2; -catenin and adenomatomous polyposis coli genes in colon carcinoma cells [<xref ref-type="bibr" rid="B25">25</xref>]. Those authors speculated that abnormal levels of &#x003b2; -catenin can contribute to the accumulation and overexpression of the cyclin D₁ protein and hence transformation. The &#x003b2; -catenin pathway, therefore, merits investigation in rat mammary tumorigenesis.</p><p>In conclusion, we measured several parameters that could potentially be involved in the resistance of the Copenhagen rat to mammary tumorigenesis. We found no differences in the number of lesions in Copenhagen compared with Wistar-Furth mammary glands at 20 or 30 days after MNU treatment, but at 37 days there were significantly fewer lesions in the Copenhagen glands. Furthermore, by this time advanced lesions such as DCIS and adenocarcinomas were present in Wistar-Furth glands, whereas no such lesions were observed in Copenhagen rats. Immunohistochemical staining of lesions from both strains indicated that cyclin D₁ was frequently overexpressed in Wistar-Furth lesions at 37 days, but not in Copenhagen lesions from the same time. Expression of p16^{<italic>INK4a</italic>} protein, bromodeoxyuridine labeling and apoptotic indices, and mast cell infiltration around lesions were not significantly different between the two strains at any time. These findings indicate that overexpression of cyclin D₁ might play a fundamental role in the progression of IDPs to DCIS and adenocarcinomas during rat mammary tumorigenesis. Furthermore, this gene might also play a role in the resistance of Copenhagen rats to MNU-induced mammary tumorigenesis.</p></sec>

Increased cell survival by inhibition of BRCA1 using an antisense approach in an estrogen responsive ovarian carcinoma cell line

<sec> <title>Introduction:</title> <p>Germline mutations in the breast and ovarian cancer susceptibility gene <italic>BRCA1</italic>, which is located on chromosome 17q21, are associated with a predisposition to the development of cancer in these organs [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. No mutations in the <italic>BRCA1</italic> gene have been detected in sporadic breast cancer cases, but mutations have been detected in sporadic cases of ovarian cancer [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. Although there is debate regarding the level of cancer risk associated with mutations in <italic>BRCA1</italic> and the significance of the lack of mutations in sporadic tumors, it is possible that alterations in the function of BRCA1 may occur by mechanisms other than mutation, leading to an underestimation of risk when it is calculated solely on the basis of mutational analysis. Such alterations cannot be identified until the function and regulation of BRCA1 are better understood.</p> <p>The <italic>BRCA1</italic> gene encodes a 220-kDa nuclear phosphoprotein that is regulated in response to DNA damaging agents [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>] and in response to estrogen-induced growth [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>]. Germline mutations that cause breast and ovarian cancer predisposition frequently result in truncated and presumably inactive BRCA1 protein [<xref ref-type="bibr" rid="B12">12</xref>].</p> <p>BG-1 cells were derived from a patient with stage III, poorly differentiated ovarian adenocarcinoma [<xref ref-type="bibr" rid="B13">13</xref>]. This cell line, which expresses wild-type BRCA1, is estrogen responsive and withdrawal of estrogen results in eventual cell death. Previous studies suggest that BRCA1 is stimulated as a result of estrogen treatment [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>], and also that BRCA1 may be involved in the cell death process [<xref ref-type="bibr" rid="B14">14</xref>]. Therefore, we examined the effect of reduction of BRCA1 levels in BG-1 cells on the cellular response to hormone depletion as well as estrogen stimulation. The results suggest that reduced levels of BRCA1 correlates with a survival advantage when BG-1 cells are placed under growth-restrictive and hormone-depleted conditions. In optimum growth conditions, significantly reduced levels of BRCA1 correlates with enhanced growth both <italic>in vitro</italic> and <italic>in vivo</italic>.</p> </sec> <sec> <title>Aims:</title> <p>To test the hypothesis that BRCA1 may play a role in the regulation of ovarian tumor cell death as well as in the inhibition of ovarian cell proliferation.</p> </sec> <sec> <title>Materials and methods:</title> <p>The estrogen receptor-positive, BG-1 cell line [<xref ref-type="bibr" rid="B13">13</xref>], which contains an abundant amount of estrogen receptors (600 fmoles/100 &#x003bc;g DNA), was infected using a pLXSN retroviral vector (provided by AD Miller) containing an inverted partial human cDNA 900-base-pair sequence of BRCA1 (from nucleotide 121 in exon 1 to nucleotide 1025 in exon 11, accession #U14680). After 2 weeks of selection in 800 &#x003bc;g/ml of geneticin-G418 (Gibco/Life Technologies, Gaithersburg, MD, USA), BG-1 G418-resistant colonies were pooled, or individually isolated, and assayed for growth in the presence or absence of supplemented estrogen. Virally infected pooled populations of BG-1 cells were examined for BRCA1 message levels by ribonuclease protection assay (Fig. <xref ref-type="fig" rid="F1">1a</xref>). BRCA1 ribonuclease protection probe was made using an <italic>in vitro</italic> transcription kit (Ambion, Inc, Austin, TX, USA) as previously described [<xref ref-type="bibr" rid="B10">10</xref>] and derived clones were tested for protein levels by Western blot analysis using an anti-BRCA1 (Oncogene Research, Ab-1, Cambridge, MA, USA) antibody. Growth curve analysis of Infected populations and were pretreated for 5 days in phenol red-free, Dulbecco's modified eagle medium (DMEM)/F-12 medium (Gibco/Life Technologies) supplemented with 10% charcoal/dextran treated serum (Hyclone, Logan, UT, USA), then plated at 2.5 &#x000d7; 10⁶ cells per 100mm dish in triplicate in the absence or presence of estrogen (10^-8 mol/l; 17&#x003b2;-Estradiol; 1,3,5 (10) - Estratriene 3,17&#x003b2;-diol; Sigma, St Louis, MO, USA). For soft agar assay, clones were plated into 10 60-mm dishes at 1 &#x000d7; 10⁵ cells/dish containing 0.3% bactopeptone agar with or without added estrogen (10^-8 mol/l) in phenol red-free medium with 10% stripped serum in order to test for anchorage independent growth. BG-1 infected clones were tested for tumorigenicity by injection of cells (10⁶ cells in 0.1cm² 50% matrigel; Collaborative Biomedical Products, Bedford, MA, USA) into subcutaneous sites in 6-week-old athymic Ncr-nude mice (NCI Animal Program, Bethesda, MD, USA) that were ovariectomized at approximately 4 weeks of age. Half of the ovariectomized mice received an implanted 0.18mg estrogen 60-day pellet (Innovative Research of America, Sarasota, FL, USA).</p> </sec> <sec> <title>Results:</title> <p>Antisense technology was effective in decreasing both RNA and protein levels of BRCA1 in the BG-1 human ovarian adenocarcinoma cells. BRCA1 antisense-infected populations contained significantly less BRCA1 message than control LXSN-infected pools and selected clones contained varying reduced levels of BRCA1 protein compared with control clones (Figs <xref ref-type="fig" rid="F1">1a</xref> and 1b).</p> <p>Three independent BRCA1 antisense-infected cultures demonstrated a resistance to cell death induced by withdrawal from estrogen over a 6- to 20-day period (Fig. <xref ref-type="fig" rid="F2">2a</xref>). The BRCA1 antisense population also exhibited a threefold to sixfold increase in cell growth compared with control cells in the presence of estrogen treatment. BG-1 BRCA1 antisense clones demonstrated a similar response to pooled population studies, enhanced growth with estrogen, and failure to die upon estrogen depletion (Fig. 2b).</p> <p>The BRCA1 antisense clones were further examined for other associated tumorigenic properties. All of the antisense clones were able to form colonies in soft agar (2-23 colonies per 10⁴ cells plated; data not shown), whereas control clones were deficient in their ability to form colonies (0-0.8 colonies per 10⁴ cells plated). <xref ref-type="table" rid="T1">Table 1</xref> shows, in the presence of estrogen, the clone with the lowest levels of BRCA1 (AS-4) produced significantly more colonies (133 &#x000b1; 17.9 colonies per 10⁴ cells plated) than the control clone (NEO; 6 &#x000b1; 3.1 colonies per 10⁴ cells plated). Clones AS-4 and NEO were also injected with matrigel subcutaneously into ovariectomized athymic mice. Almost twice as many sites were positive for the AS-4 clone (14 out of 14) as for the NEO clone (eight out of 14) 42 days after injection. In addition, BRCA1 antisense tumors averaged twice the size of control tumors. The BRCA1 reduced cells also formed tumors with half the latency of control cells in the presence of implanted estrogen (11 days versus 21 days until tumor formation).</p> </sec> <sec> <title>Discussion:</title> <p>The present studies show that reduction in BRCA1 levels, using an antisense retroviral vector in the estrogen dependent BG-1 ovarian carcinoma cell line, contributes to confirmation of the hypothesis that <italic>BRCA1</italic> plays a pivotal role in the balance between cell death and cell proliferation. <italic>BRCA1</italic> RNA and protein levels were successfully reduced in populations and isolated clones of antisense infected BG-1 cells. Decreased <italic>BRCA1</italic> levels rescued the BG-1 cells from growth arrest or cell death in adverse growth conditions in monolayer or soft agar conditions. Furthermore, a BRCA1 antisense clone that had significantly low levels of BRCA1 protein was able to form twice as many tumors in ovariectomized nude mice with a decreased latency compared with a control clone.</p> <p>In multicellular mammalian organisms, a balance between cell proliferation and cell death is extremely important for the maintenance of normal healthy tissues. In support of this hypothesis, it has been shown that <italic>p53</italic> and <italic>BRCA1</italic> can form stable complexes, and can coactivate <italic>p21</italic> and <italic>bax</italic> genes, which may lead to the activation of the apoptosis pathway [<xref ref-type="bibr" rid="B15">15</xref>]. The present data, which show that cells with a reduction of BRCA1 have a survival advantage in conditions where control cells fail to thrive, also supports this hypothesis. <italic>BRCA1</italic> levels appear to affect the ability of cells to arrest growth or die in the absence of estrogenic growth-inducing conditions. Although mutations in this gene are uncommon in sporadic breast and ovarian tumors, <italic>BRCA1</italic> expression levels and protein levels have been found to be reduced in sporadic human breast carcinomas [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. In addition it has been demonstrated [<xref ref-type="bibr" rid="B20">20</xref>] that hormone-dependent tumors such as breast and ovarian cancers have a decreased ability to undergo apoptosis. Other mechanisms involving gene regulation may allow for decreased expression of <italic>BRCA1</italic> in sporadic tumors. The response of <italic>BRCA1</italic> mRNA and protein levels to mitogens and hormones <italic>in vitro</italic> suggests that <italic>BRCA1</italic> may play a role in regulation of cell growth or maintenance [<xref ref-type="bibr" rid="B21">21</xref>]. The BRCA1 gene product may be involved in the regulation of hormone response pathways, and the present results demonstrate that loss of BRCA1 may result in loss of inhibitory control of these mitogenic pathways. These studies show that reduction in <italic>BRCA1</italic> mRNA and protein can result in increased proliferation of BG-1 ovarian cancer cells in both <italic>in vitro</italic> and <italic>in vivo</italic> conditions, suggesting that <italic>BRCA1</italic> may normally be acting as a growth inhibitor. Low BRCA1 levels found in sporadic cancers may be an important factor in tumorigenesis. The present data suggest that diminished levels of <italic>BRCA1</italic> not only accelerate proliferation in the BG-1 ovarian carcinoma cell line, but also appear to promote tumorigenesis. We propose that the loss or reduction of <italic>BRCA1</italic> may predispose a cell population to neoplastic transformation by altering the balance between cell death and proliferation/survival, rendering it more sensitive to secondary genetic changes.</p> </sec>

<contrib id="A1" contrib-type="author"> <name> <surname>Annab</surname> <given-names>Lois A</given-names> </name> <xref ref-type="aff" rid="I1">1</xref> <email>Annab@NIEHS.NIH.GOV</email> </contrib> <contrib id="A2" contrib-type="author"> <name> <surname>Hawkins</surname> <given-names>Rebecca E</given-names> </name> <xref ref-type="aff" rid="I1">1</xref> </contrib> <contrib id="A3" contrib-type="author"> <name> <surname>Solomon</surname> <given-names>Greg</given-names> </name> <xref ref-type="aff" rid="I1">1</xref> </contrib> <contrib id="A4" contrib-type="author"> <name> <surname>Barrett</surname> <given-names>J Carl</given-names> </name> <xref ref-type="aff" rid="I1">1</xref> </contrib> <contrib id="A5" contrib-type="author"> <name> <surname>Afshari</surname> <given-names>Cynthia A</given-names> </name> <xref ref-type="aff" rid="I1">1</xref> </contrib>

Breast Cancer Research

<sec> <title>Introduction</title> <p>Germline mutations in the breast and ovarian cancer susceptibility gene <italic>BRCA1</italic>, which is located on chromosome 17q21, are associated with a predisposition to the development of cancer in these organs [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Initial analyses [<xref ref-type="bibr" rid="B22">22</xref>] suggested that women with germline mutations in the <italic>BRCA1</italic> gene and a strong family history of breast or ovarian cancer have 85 and 44% lifetime risks of developing breast and ovarian cancer, respectively. Recent studies [<xref ref-type="bibr" rid="B23">23</xref>], however, have suggested that analyses based on women who were not selected for a familial history of cancer indicate that the risk for cancer associated with mutations in these genes is 50 and 16% for breast and ovarian cancers, respectively. No mutations in the <italic>BRCA1</italic> gene have been detected in sporadic breast cancer cases; however, mutations have been detected in sporadic cases of ovarian cancer [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. Although there is debate regarding the level of cancer risk associated with mutations in <italic>BRCA1</italic> and the significance of the lack of mutations in sporadic tumors, it is possible that alterations in the function of BRCA1 may occur by mechanisms other than mutation. This would lead to an underestimation of risk when it is calculated solely on the basis of mutational analysis. Such alterations cannot be identified until the function and regulation of BRCA1 are better understood.</p> <p>The <italic>BRCA1</italic> gene encodes a 220-kDa nuclear protein that may be regulated by phosphorylation through the cell cycle and in response to DNA damaging agents [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. The level of BRCA1 is also regulated in response to estrogen or estrogen-induced growth in breast [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>] and ovarian cell lines. BRCA1 has been shown to colocalize in nuclear dots with other cellular proteins, including BARD-1 [<xref ref-type="bibr" rid="B24">24</xref>], Rad51, PCNA, and BRCA2 [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. In addition, BRCA1 can act as a transcriptional transactivator in yeast reporter assays [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>] and binds the RNA polymerase II holoenzyme, a component of the basal transcription machinery [<xref ref-type="bibr" rid="B25">25</xref>]. The precise mechanism of action and the specific signaling pathway affected by BRCA1 remain unknown, however.</p> <p>Studies of <italic>BRCA1</italic> expression patterns in mouse tissue reveal that BRCA1 is most highly expressed in tissues undergoing rapid proliferation and differentiation, and that expression <italic>in vivo</italic> is also hormone responsive. For example, analyses of mammary gland growth and development show high levels of BRCA1 expression in terminal end buds during puberty and in budding alveoli during pregnancy. In addition, hormonal stimulation in ovariectomized mice results in induction of BRCA1 expression in the breast [<xref ref-type="bibr" rid="B28">28</xref>]. Attempts to develop homozygous, BRCA1-deleted mouse models have resulted in embryonic lethality [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>]. For example, when the <italic>BRCA1</italic> gene deletion was targeted in exons 5 and 6, mutant mice died before day 7.5 of embryogenesis. Analysis of DNA synthesis in the mutant embryos indicated that cell proliferation was impaired, suggesting that BRCA1 may paradoxically play a positive role in the regulation of embryonic cell growth [<xref ref-type="bibr" rid="B29">29</xref>].</p> <p>Most of the mechanistic BRCA1 studies to date have been conducted in breast carcinoma cell lines; therefore, we decided to conduct a study to determine the effect of BRCA1 expression on the cellular phenotype of an ovarian carcinoma cell line, BG-1. BG-1 cells were derived from a patient with stage III, poorly differentiated ovarian adenocarcinoma [<xref ref-type="bibr" rid="B13">13</xref>]. This cell line, which expresses wild-type BRCA1, is estrogen responsive, and withdrawal of estrogen results in eventual cell death. Previous studies suggested that BRCA1 is stimulated as a result of estrogen treatment [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>], and that BRCA1 may be involved in the cell death process [<xref ref-type="bibr" rid="B14">14</xref>]. Therefore, we examined the effect of reduction of BRCA1 levels in BG-1 cells on the cellular response to estrogen stimulation as well as hormone depletion. Our results suggest that when BG-1 cells are subjected to growth restrictive and hormone-depleted conditions, cells that have even moderately reduced levels of BRCA1 protein have a distinct advantage for survival. In addition, significant reduction in BRCA1 protein level correlates with enhanced estrogen proliferation when compared with cells that express moderate to wild-type BRCA1 levels, grown under optimal growth conditions both <italic>in vitro</italic> and <italic>in vivo</italic>.</p> </sec> <sec sec-type="materials|methods"> <title>Materials and methods</title> <sec> <title>Cells and cell culture</title> <p>The estrogen receptor-positive, BG-1 line [<xref ref-type="bibr" rid="B13">13</xref>], which contains an abundant amount of estrogen receptors (600 fmol/100 &#x003bc;g DNA), was provided by J Boyd (Sloan-Kettering Cancer Center, New York, NY, USA). GPE86 and PA317 viral packaging cell lines were provided by AD Miller (Fred Hutchinson Cancer Center, Seattle, WA, USA). BG-1 cells were maintained in Dulbecco's modified eagle medium (DMEM)/F12 medium supplemented with 10% fetal calf serum (Summit, Fort Collins, CO, USA), and 50 units/ml penicillin/streptomycin. BG-1 cells arrest to gamma radiation consistent with a wildtype <italic>p53</italic> phenotype. These cells were tested negative for mycoplasmas.</p> </sec> <sec> <title>Retroviral vector preparation and infection of cells</title> <p>A partial human cDNA sequence of <italic>BRCA1</italic> (from nucleotide 121 in exon 1 to nucleotide 1025 in exon 11, accession #U14680) was inserted in the antisense orientation into the EcoR1 site of the pLXSN retroviral vector (provided by AD Miller). The vector alone, or the antisense <italic>BRCA1</italic> vector, was transfected using the calcium-phosphate precipitation method into the ecotropic packaging cell line GPE86 [<xref ref-type="bibr" rid="B31">31</xref>]. Supernatant, generated from transfected GPE86 cells [<xref ref-type="bibr" rid="B31">31</xref>], was then used to infect the amphotropic packaging cell line PA317 [<xref ref-type="bibr" rid="B31">31</xref>] in the presence of 4 &#x003bc;g/ml polybrene (Abbott Laboratories, Abbott Park, IL, USA). PA317-infected cells were grown in selection media for 2 weeks and pooled for supernatant collection. Supernatants were filtered (0.20 &#x003bc;m filter) and tested for virus-producing cells. Titer efficiencies of the LXSN virus ranged from 10⁴ to 10⁵ colony-forming units/ml on mouse cells (A9). Log phase BG-1 cells were exposed to supernatant containing either the LXSN control retrovirus or retroviruses containing the antisense <italic>BRCA1</italic> cDNA sequence. After 2 weeks of selection in 800 &#x003bc;g/ml of geneticin-G418 (Gibco/Life Technologies, Gaithersburg, MD, USA), BG-1 G418-resistant colonies were pooled or individually isolated and assayed for growth in the presence or absence of supplemented estrogen. Only isolated clones were used in anchorage dependence and tumorigenicity studies.</p> </sec> <sec> <title>BRCA1 ribonuclease protection assay and protein analysis</title> <p>A BRCA1 ribonuclease protection probe was made using an <italic>In Vitro</italic> Transcription Kit (Ambion, Inc, Austin, TX, USA). The DNA template spanned part of exon 22, all of exon 23, and part of exon 24 of the <italic>BRCA1</italic> gene. Template DNA was incubated for 45min at 37&#x000b0;C with (&#x003b1;-³²P)-uridine triphosphate and T7 polymerase in the presence of buffer and nucleotides. DNA template was removed by ribonuclease-free deoxyribonuclease incubation at 37&#x000b0;C for 30min. The reaction was stopped by the addition of 0.5 mol/l ethylenediaminetetra-acetic acid, and the labeled probe was purified on a 5% polyacrylamide gel. Sample RNA (20&#x003bc;g total RNA) was coprecipitated with the BRCA1 probe and the cyclophilin control probe [<xref ref-type="bibr" rid="B32">32</xref>], resuspended in Hyb-speed RPA (Ambion) hybridization buffer at 95&#x000b0;C, and then incubated at 68&#x000b0;C for 10min. Ribonuclease was added and the sample was incubated for 45min at 37&#x000b0;C. Protected fragments were precipitated, and resuspended in loading buffer, followed by separation on a 5% polyacrylamide-urea gel, and exposed to X-ray film.</p> <p>BRCA1 protein was analyzed by Western blot analysis. Whole cell lysate (50 &#x003bc;g) was loaded onto a 6% sodium dodecyl sulfate-polyacrylamide gel electrophoresis gel, transferred to nitrocellulose, and hybridized with an anti-BRCA1 (Ab-1; Oncogene Research, Cambridge, MA, USA) antibody as previously described [<xref ref-type="bibr" rid="B10">10</xref>].</p> </sec> <sec> <title>Estrogen treatment and growth curve analysis</title> <p>G418-resistant colonies from <italic>BRCA1</italic> antisense infected BG-1 cells and control vector LXSN-infected BG-1 cells were pooled, pretreated for 5 days in phenol red-free, DMEM/F-12 medium (Gibco/Life Technologies) supplemented with 10% charcoal/dextran treated serum (Hyclone, Logan, UT, USA), then plated at 2.5 &#x000d7; 10⁶ cells per 100mm dish in triplicate for growth curve analysis in the absence or presence of estrogen (10^-8mol/l; 17&#x003b2;-Estradiol; 1,3,5 (10) - Estratriene 3, 17&#x003b2;-diol; Sigma, St Louis, MO, USA). Extended growth curve analysis was plated at 2.5 &#x000d7; 10⁶ in 100mm dishes for an extended treatment of 20 days without estrogen. Clones were isolated from the <italic>BRCA1</italic> antisense and LXSN BG-1 pooled populations and grown in phenol red-free, DMEM/F-12 medium (Gibco/Life Technologies) supplemented with charcoal/dextran treated serum (Hyclone) for 5 days before plating for growth curve analysis. Cells were then plated in triplicate at 1 &#x000d7; 10⁶ cells per 60mm dish in either the absence or presence of estrogen (10^-8mol/l) and grown for 8-10 days. Cell number was calculated on indicated days using a Coulter counter. The number of dead cells for the extended growth curve experiment was calculated by counting trypan blue incorporated cells using a hemocytometer. Statistical analyses of <italic>P</italic> values were calculated based on the fold differences between growth of the clones by computing a mean ratio and the corresponding standard deviation [<xref ref-type="bibr" rid="B33">33</xref>].</p> </sec> <sec> <title>Anchorage independence analysis and tumorigenicity</title> <p>Selected BG-1 clones were tested for anchorage independent growth in 0.3% bacto-peptone agar with a 0.6% bacto-peptone agar base plus or minus added estrogen (10^-8 mol/l) in phenol red-free medium with 10% stripped serum. Each BG-1, neomycin-resistant clone was plated into 10 60-mm dishes containing the agar at 1 &#x000d7; 10⁵ cells per dish. Colonies (greater than 30 cells) were scored after 14 days. Pairwise comparisons were made by a two-sided Mann-Whitney U test to calculate <italic>P</italic> values [<xref ref-type="bibr" rid="B34">34</xref>].</p> <p>BG-1-infected clones were tested for tumorigenicity in 6-week-old athymic Ncr-nude mice (NCI Animal Program, Bethesda, MD, USA) that were ovariectomized at approximately 4 weeks of age. BRCA1 antisense clone (AS-4) was injected (10⁶ cells in 0.1cm² 50% matrigel; Collaborative Biomedical Products, Bedford, MA, USA) into two subcutaneous sites on one side of 16 mice (32 injection sites), whereas a LXSN control clone was injected on the opposite side of the same mice. Nine of the 16 ovariectomized mice also received an implanted 0.18mg estrogen 60-day pellet (Innovative Research of America, Sarasota, FL, USA). Mice were periodically examined and tumor size was measured during the 3-month period after injection.</p> </sec> </sec> <sec> <title>Results</title> <sec> <title>Effective decrease in <italic>BRCA1</italic> expression using antisense technology</title> <p>Antisense technology was effective in decreasing both RNA and protein levels of BRCA1 in the BG-1 human ovarian adenocarcinoma cells. BG-1 human ovarian adenocarcinoma cells were infected with a retroviral construct composed of an antisense 900 base-pair cDNA sequence of the amino-terminal region of <italic>BRCA1</italic>. Three experiments (two of which were from independently made supernatants) showed that infection of pLXSN (vector alone) and <italic>BRCA1</italic> antisense retroviral constructs into BG-1 cells yielded G418-resistant colonies at similar rates (titers ranged from 0.78 to 4.2 &#x000d7; 10⁴ colony-forming units/ml). The same vectors were also directly transfected into BG-1 cells at an efficiency of 6.3&#x000d7;10^-5 for the anti-sense <italic>BRCA1</italic> or 9.9 &#x000d7; 10^-5 for the control plasmid (pLXSN). Neomycin-resistant colonies were pooled and examined for <italic>BRCA1</italic> message levels by ribonuclease protection assay. <italic>BRCA1</italic> antisense infected cells contained significantly less <italic>BRCA1</italic> message than control LXSN infected cells, whether cultured in the presence or absence of estrogen (Fig. <xref ref-type="fig" rid="F1">1a</xref>). Although there appears to be no detectable amounts of BRCA1 RNA present after estrogen withdrawal, low levels of protein can be detected by western blot analysis [<xref ref-type="bibr" rid="B10">10</xref>].</p> <p>Subclones were also isolated from BRCA1 antisense infected cells, or LXSN infected cells (NEO). Western blot analysis demonstrated that all of the antisense BRCA1 clones had reduced levels of BRCA1 protein compared with the NEO clones, and one antisense clone (AS-4) had very low levels of BRCA1 protein, although it was not totally absent (Fig. <xref ref-type="fig" rid="F1">1b</xref>).</p> </sec> <sec> <title>Effects of reduced <italic>BRCA1</italic> expression on <italic>in vitro</italic> growth</title> <p>Pooled populations of antisense <italic>BRCA1</italic> BG-1 colonies were examined for growth in the absence or presence of supplemented estrogen. Three independently infected cultures of <italic>BRCA1</italic> antisense cells demonstrated a resistance to cell death induced by withdrawal from estrogen over a 6-day period, as well as a threefold to sixfold increase in cell growth compared with control cells in the presence of estrogen treatment (Fig. <xref ref-type="fig" rid="F3">3</xref>). In order to investigate further whether reduction of BRCA1 protein had an effect on hormone-dependent cell growth, BG-1 antisense and control cells were grown in estrogen-deprived conditions for an extended period of time. During the first 5 days, both groups continued to proliferate in the absence of estrogen, but the BRCA1 antisense group continued to grow for the next 10 days, whereas control cells decreased in number (Fig. <xref ref-type="fig" rid="F2">2a</xref>).</p> <p>In order to avoid the problem of a mixed population of cells expressing various levels of BRCA1, subclones were isolated from infected populations of <italic>BRCA1</italic> antisense infected BG-1 cells and control LXSN infected BG-1 cells (NEO). Figure <xref ref-type="fig" rid="F2">2b</xref> shows BG-1 parental and NEO clones exhibited up to a 37% decrease in cell number during a 3-day period of estrogen withdrawal, whereas antisense <italic>BRCA1</italic> clones showed as much as a twofold increase in cell number during the same time period. In an attempt to determine if increased survival of the antisense cells was do to increased proliferation of the antisense cells or decreased death rate, the number of trypan blue positive, non-viable were examined after 14 days without estrogen. There were 5-10-fold more dead cells present in the media of control cells (BG-1 parental and NEO clone) then in the <italic>BRCA1</italic> antisense clone AS-4 (data not shown). It appeared that resistance to cell death plays a significant role in the survival of <italic>BRCA1</italic> antisense cells to estrogen withdrawal. Figure <xref ref-type="fig" rid="F2">2b</xref> again demonstrates the ability of the BRCA1 antisense sub-clones to survive estrogen deprivation. In the presence of estrogen, three out of the four antisense <italic>BRCA1</italic> clones exhibited a growth advantage over NEO clones or the BG-1 parental population (Fig. <xref ref-type="fig" rid="F2">2b</xref>). Antisense <italic>BRCA1</italic> clones 1, 3 and 4 showed a 10-fold to 16-fold increase in cell number between days 5 and 8 after estrogen treatment compared with only a threefold to fivefold increase of cell number in NEO clones and BG-1 parental cells (Fig. <xref ref-type="fig" rid="F2">2b</xref>). The AS-4 clone, which had the lowest levels of BRCA1 protein, showed a highly significant (16-fold; <italic>P</italic>&#x0003c;0.01) stimulation of growth between days 5 and 8 of estrogen induction (Figs <xref ref-type="fig" rid="F2">2b</xref> and <xref ref-type="fig" rid="F4">4</xref>). In summary, although three out of four of the antisense <italic>BRCA1</italic> clones had a growth advantage in the presence of estrogen, all four antisense <italic>BRCA1</italic> clones showed enhanced survival in estrogen-depleted media.</p> </sec> <sec> <title>Anchorage independent growth of BG-1 clones</title> <p>Anchorage independent growth is a common property of many transformed cells. Therefore, the <italic>BRCA1</italic> antisense subclones were also studied for anchorage independent growth in a semisoft agar medium with and without supplemented estrogen. Table1 shows that colony formation efficiencies on plastic of control (NEO) and <italic>BRCA1</italic> anti-sense (AS-4) cells were similar in estrogen-depleted and estrogen-containing media. However, the BG-1 control clone (NEO) was unable to form colonies (fewer than one colony per 10⁴ cells plated) in agar without the addition of estrogen, whereas the BG-1 antisense <italic>BRCA1</italic> clone was able to form soft agar colonies in estrogen depleted conditions (10 &#x000b1; 2.9 colonies per 10⁴ cells plated). In the presence of estrogen, both NEO and AS-4 were able to form colonies; however, there was a significant difference (<italic>P</italic> &#x0003c;0.01) in the ability to form colonies in agar between AS-4 (133 colonies) and the control clone (six colonies). These data suggest a correlation between the loss of BRCA1 protein and an increased survival/growth advantage in anchorage-independent conditions.</p> </sec> <sec> <title>Effects of reduced BRCA1 protein on in vivo tumor cell growth</title> <p>Because the AS-4 clone showed a growth advantage in soft agar, a phenotype that may be correlated with the ability to form tumors <italic>in vivo</italic>, the BRCA1 antisense sub-clone AS-4 was evaluated for its ability to form subcutaneous tumors in ovariectomized athymic mice in the presence or absence of an estrogen pellet. Mice were injected subcutaneously with AS-4 cells in matrigel on one side of each mouse and NEO cells in matrigel on the other side. Of the mice injected, 50% received an implanted estrogen pellet (0.18 mg estrogen) that was designed to release estrogen for 60 days. In the absence of estrogen, a significant difference was detected in tumorigenic growth between AS-4 and NEO cells (Fig. <xref ref-type="fig" rid="F5">5b</xref>). Almost twice as many sites were tumor positive for the AS-4 clones than for NEO injected sites. 100% (14/14) Tumor formation was reached for all AS-4 clones 42 days after injection, compared with 57% (eight out of 14) positive tumor formation of the NEO sites (Fig. <xref ref-type="fig" rid="F5">5b</xref>, upper panel). AS-4 cells also formed tumors that averaged twice the size of NEO control tumors (Fig. <xref ref-type="fig" rid="F5">5b</xref>, lower panel).</p> <p>BG-1 cells without matrigel were nontumorigenic in athymic male or female mice (0 positive sites/20 sites injected at 5&#x000d7;10⁶ cells per site), but these cells formed large, progressively growing tumors when injected with matrigel in the presence of estrogen (Fig. <xref ref-type="fig" rid="F5">5a</xref>). These tumors were very large (&#x0003e;1cm diameter) and did not regress even though the estrogen pellet was effective for only 60 days (Fig. <xref ref-type="fig" rid="F5">5a</xref>, lower panel). Similar to the agar experiments, both the BRCA1 antisense clone and LXSN control clone were positive for tumor formation in the presence of estrogen. AS-4 cells formed tumors with half of the latency of control cells in the presence of implanted estrogen (Fig. <xref ref-type="fig" rid="F5">5a</xref>, upper panel; 11 days versus 21 days until tumor formation). Neither AS-4 nor NEO cells formed progressively growing tumors in the absence of estrogen, however. All tumors in the mice without estrogen pellets had started to regress by 71 days after injection. The observed tumor regression was not surprising, because matrigel is not stable for longer than 14 days in culture, and probably not <italic>in vivo</italic> either (personal communication; Collaborative Biomedical Products, Bedford, MA, USA). By day 71, the matrigel would no longer confer an optimal growth environment for BG-1 cells.</p> </sec> <sec> <title>Discussion</title> <p>The present studies show that reduction of BRCA1 levels, using an antisense retroviral vector in the estrogen dependent BG-1 ovarian carcinoma cell line, may aid in confirmation of the hypothesis that <italic>BRCA1</italic> functions as a tumor suppressor gene by playing a pivotal role in the balance between cell death and cell proliferation. <italic>BRCA1</italic> RNA and protein levels were successfully reduced in pooled and isolated subclones of antisense-infected populations of BG-1 cells. Decreased <italic>BRCA1</italic> levels appeared to affect the ability of BG-1 cells to arrest growth or die in the absence of estrogenic growth-inducing conditions. We found that <italic>BRCA1</italic> antisense cells, both as pooled populations and individual subclones, also exhibited enhanced growth in monolayer culture on plastic in the presence of estrogen compared with control vector-infected colonies. All BRCA1 antisense subclones were able to proliferate as well as exhibit a decreased death rate in estrogen-deprived media, whereas parental and control subclones failed to grow. Death after estrogen withdrawal has been shown in previous studies using BG-1 cells [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B35">35</xref>]. BRCA1 antisense subclones demonstrated other traits associated with a tumorigenic phenotype, such as the ability to grow in soft agar independent of estrogen, whereas control clones could only form colonies with the addition of estrogen. In ovariectomized nude mice, a BRCA1 antisense clone (AS-4) was examined for tumorigenicity compared with a control clone (NEO). The AS-4 clone formed a greater number of and larger tumors than NEO in the absence of estrogen, and in general formed tumors faster in the presence of estrogen. The main conclusion from these studies is that BG-1 clones with reduced levels of BRCA1 protein have a survival advantage over controls in the absence of estrogen both <italic>in vitro</italic> and <italic>in vivo.</italic></p> <p>The response of <italic>BRCA1</italic> mRNA and protein levels to mitogens and hormones <italic>in vitro</italic> suggests that <italic>BRCA1</italic> may play a role in regulation of cell growth or maintenance [<xref ref-type="bibr" rid="B21">21</xref>]. During estrous, many hormones and growth factors interact in a complex manner as survival factors and inducers of cell proliferation, which are then balanced with growth inhibitors [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>,<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. The mechanism by which <italic>BRCA1</italic> can regulate or influence these processes has not yet been identified. It has been shown that <italic>BRCA1</italic> is induced as a result of the mitogenic activity of the estrogen receptor in estrogen receptor-positive cells [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Direct estrogen stimulation is not required for <italic>BRCA1</italic> transcription, however [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B41">41</xref>]. In support of this, <italic>BRCA1</italic> expression has been shown to occur in the mouse ovary (granulosa and thecal cells of small and medium follicles) independent of hormonal status, and even in ovaries from estrogen receptor -/- deficient mice [<xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B42">42</xref>]. In contrast, the tumors from patients with <italic>BRCA1</italic> mutations appear to have downregulation of estrogen receptors [<xref ref-type="bibr" rid="B43">43</xref>,<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>]. Previous experiments in our laboratory showed that another hormone, progesterone, could also cause a modest increase of <italic>BRCA1</italic> mRNA in BG-1 cells after 24 h exposure without an increase in growth (unpublished data). Progesterone has been found [<xref ref-type="bibr" rid="B46">46</xref>] to inhibit cell proliferation and induce apoptosis significantly in two ovarian carcinoma cell lines. Thus, although <italic>BRCA1</italic> may not be regulated directly by hormones, the BRCA1 gene product may be involved in the regulation of hormone response pathways, and the present results may demonstrate that loss of BRCA1 may result in loss of inhibitory control of these mitogenic pathways.</p> <p> <italic>BRCA1</italic> transcription is regulated with the cell cycle, and highest levels correlate with the G1/S-phase boundary [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B47">47</xref>,<xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B49">49</xref>]. The present studies show that reduction of <italic>BRCA1</italic> mRNA and protein can result in increased proliferation of BG-1 ovarian cancer cells <italic>in vitro</italic> and <italic>in vivo</italic>, suggesting that <italic>BRCA1</italic> may normally be acting as a growth inhibitor. Similar to our findings with ovarian carcinoma cells, accelerated growth, anchorage independence and tumorigenicity is associated with <italic>BRCA1</italic> antisense introduction into mouse NIH3T3 cells [<xref ref-type="bibr" rid="B50">50</xref>]. In addition, increased proliferation of mammary cells is induced with antisense oligonucleotides to <italic>BRCA1</italic> [<xref ref-type="bibr" rid="B51">51</xref>]. Conversely, introduction of full-length <italic>BRCA1</italic> by retrovirus-mediated gene transfer inhibited growth of breast and ovarian cancer cell lines in both <italic>in vitro</italic> and <italic>in vivo</italic> experiments [<xref ref-type="bibr" rid="B51">51</xref>], and transfection of <italic>BRCA1</italic> into colon cancer cells inhibited new DNA synthesis by 50% in addition to inhibition of S-phase progression, possibly through direct transactivation of the cell cycle inhibitor <italic>p21</italic> ^WAF1/CIP1 [<xref ref-type="bibr" rid="B49">49</xref>]. </p> <p>In multicellular mammalian organisms, a balance between cell proliferation and cell death is extremely important for the maintenance of normal healthy tissues. This is especially important during early embryonic development as well as in the development and function of adult tissues such as the gonadal cells (ie ovarian and testes) [<xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B48">48</xref>]. For example, <italic>BRCA1</italic> expression is critical during development, as evidenced by the embryonic lethality in transgenic knockout mice [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B52">52</xref>]. Alternatively, overexpression of <italic>BRCA1</italic> may activate apoptosis or cell death [<xref ref-type="bibr" rid="B14">14</xref>]. Human prostate cells with an introduced wild-type <italic>BRCA1</italic> cDNA demonstrated a threefold to sixfold increase in chemosensitivity, as well as an increased susceptibility to drug-induced apoptosis [<xref ref-type="bibr" rid="B53">53</xref>]. We found that clones with even moderately reduced levels of <italic>BRCA1</italic> protein appeared to be relatively resistant to death due to estrogen deprivation. Previous studies in our laboratory showed that response of parental BG-1 cells and antisense clones to gamma radiation were consistent with a <italic>p53</italic> wildtype phenotype, indicating that loss of estrogen dependence is probably not due to a <italic>p53</italic> mutation (unpublished data). Shao <italic>et al</italic> [<xref ref-type="bibr" rid="B14">14</xref>] demonstrated that <italic>BRCA1</italic> transfected into mouse 3T3 fibroblasts resulted in increased programmed cell death. In support of this hypothesis, it has been shown that <italic>p53</italic> and <italic>BRCA1</italic> can form stable complexes, and can coactivate <italic>p21</italic> and <italic>bax</italic> genes, which may lead to the activation of the apoptosis pathway [<xref ref-type="bibr" rid="B15">15</xref>]. The present data, which show that cells with a reduction in BRCA1 have a survival advantage in conditions where control cells fail to thrive, also supports this hypothesis.</p> <p>Like <italic>p53</italic>, <italic>BRCA1</italic> has also been implicated in DNA damage and repair pathways [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B54">54</xref>]. According to this model, cells without normal <italic>BRCA1</italic> activity may accumulate genetic alterations as a result of failure to arrest and repair DNA damage or self-destruct, thereby leading to genomic instability and neoplastic progression. It may not be coincidental that <italic>BRCA1</italic>-mutant breast cancers are preferentially linked to a 'specific' histopathologic pattern that includes a high S-phase fraction of cells, aneuploidy, and hormone receptor-negative status [<xref ref-type="bibr" rid="B45">45</xref>]. In addition, it has been demonstrated [<xref ref-type="bibr" rid="B20">20</xref>] that hormone-dependent tumors such as breast and ovarian cancers have a decreased ability to undergo apoptosis. Although mutations in this gene are uncommon in sporadic breast and ovarian tumors, <italic>BRCA1</italic> expression levels and protein levels have been found to be reduced in sporadic human breast carcinomas [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Other mechanisms that involve gene regulation may allow for decreased expression of <italic>BRCA1</italic> in sporadic tumors. Hypermethylation has been observed in some sporadic breast tumors in the promoter region of <italic>BRCA1</italic>, which may account for decreased <italic>BRCA1</italic> transcription [<xref ref-type="bibr" rid="B55">55</xref>]. Low BRCA1 levels found in sporadic cancers may play an important role in tumorigenesis. The present data suggest that diminished levels of BRCA1 not only accelerate proliferation in the BG-1 ovarian carcinoma cell line, but appear to alter tumorigenesis. The exact mechanism may be unknown, but decreased <italic>BRCA1</italic> levels appear to affect the ability to arrest growth or die in the absence of estrogenic growth-inducing conditions. We propose that the loss or reduction of <italic>BRCA1</italic> may predispose a cell population to neoplastic transformation by altering the balance between cell death and proliferation/survival, rendering it more sensitive to secondary genetic changes.</p> </sec> </sec>

A novel cell culture model for studying differentiation and apoptosis in the mouse mammary gland

<sec><title>Background:</title><p>This paper describes the derivation and characterization of a novel, conditionally immortal mammary epithelial cell line named KIM-2. These cells were derived from mid-pregnant mammary glands of a mouse harbouring one to two copies of a transgene comprised of the ovine &#x003b2;-lactoglobulin milk protein gene promoter, driving expression of a temperature-sensitive variant of simian virus-40 (SV40) large T antigen (T-Ag).</p></sec><sec><title>Results</title><p>KIM-2 cells have a characteristic luminal epithelial cell morphology and a stable, nontransformed phenotype at the semipermissive temperature of 37&#x000b0;C. In contrast, at the permissive temperature of 33&#x000b0;C the cells have an elongated spindle-like morphology and become transformed after prolonged culture. Differentiation of KIM-2 cells at 37&#x000b0;C, in response to lactogenic hormones, results in the formation of polarized dome-like structures with tight junctions. This is accompanied by expression of the milk protein genes that encode &#x003b2;-casein and whey acidic protein (WAP), and activation of the prolactin signalling molecule, signal transducer and activator of transcription (STAT)5. Fully differentiated KIM-2 cultures at 37&#x000b0;C become dependent on lactogenic hormones for survival and undergo extensive apoptosis upon hormone withdrawal, as indicated by nuclear morphology and flow cytometric analysis. KIM-2 cells can be genetically modified by stable transfection and clonal lines isolated that retain the characteristics of untransfected cells.</p></sec><sec><title>Conclusion</title><p>KIM-2 cells are a valuable addition, therefore, to currently available lines of mammary epithelial cells. Their capacity for extensive differentiation in the absence of exogenously added basement membrane, and ability to undergo apoptosis in response to physiological signals will provide an invaluable model system for the study of signal transduction pathways and transcriptional regulatory mechanisms that control differentiation and involution in the mammary gland.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Gordon</surname><given-names>Katrina E</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Binas</surname><given-names>Bert</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Chapman</surname><given-names>Rachel S</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Kurian</surname><given-names>Kathreena M</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Clarkson</surname><given-names>Richard W E</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>John Clark</surname><given-names>A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Birgitte Lane</surname><given-names>E</given-names></name><xref ref-type="aff" rid="I5">5</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Watson</surname><given-names>Christine J</given-names></name><xref ref-type="aff" rid="I5">5</xref><email>cjw53@mole.bio.cam.ac.uk</email></contrib>

Breast Cancer Research

<sec><title>Introduction</title><p>With each successive pregnancy, the mammary gland completes a cycle of growth, functional differentiation and involution. These processes are of great importance in biology in their own right, but they also provide an example of how proliferation, differentiation and apoptosis are integrated into the organization of a complex three-dimensional tissue unit, whose function changes with time. The growth and development of ductal and alveolar structures during pregnancy is dependent on the interaction between the epithelial cells and stromal components of the mammary fat pad and requires the concerted actions of both peptide and steroid hormones, and cell-cell and cell-substratum interactions [<xref ref-type="bibr" rid="B1">1</xref>]. The necessity for these complex structural and hormonal interactions provide a challenge for the development of <italic>in vitro</italic> models for molecular studies that accurately mimic the differentiation and death of mammary epithelial cells.</p><p>A variety of mammary culture systems have been developed to facilitate studies on the regulation of gene transcription in the mammary gland. Whole organ and explant cultures have been of value in identifying the role of specific hormones in both the growth and differentiation of mammary tissue and the induction of milk protein gene expression [<xref ref-type="bibr" rid="B2">2</xref>]. These cultures have a limited lifespan, however, and are not useful for studies at the cellular level. Epithelial cells can be isolated from mammary tissue, maintained in culture and induced to differentiate with lactogenic hormones. The use of such primary cultures has demonstrated the importance of the cellular substratum in the differentiation process [<xref ref-type="bibr" rid="B3">3</xref>]. A major drawback of this system, in addition to the short lifespan of the cells, is the considerable amount of starting material required.</p><p>Spontaneously immortalized cell lines have arisen from prolonged culture of primary epithelial cells in low serum (2%). Many of these established mammary epithelial cell lines have proved to be useful tools in molecular and bio-chemical studies. They include EpH4 cells [<xref ref-type="bibr" rid="B4">4</xref>] and the COMMA-1D cell line, one of the most widely used <italic>in vitro</italic> mammary systems, which exhibits mammary-specific functional differentiation when exposed to lactogenic hormones and extracellular matrix (ECM) [<xref ref-type="bibr" rid="B5">5</xref>]. Subclones of COMMA-1D include HC11 and CID9 [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. HC11 cells have been widely used by us, and others, for studies on transcriptional regulation of milk protein gene expression [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>], whereas CID9 cells have been extensively used to demonstrate the role of ECM in milk protein gene expression [<xref ref-type="bibr" rid="B7">7</xref>]. Furthermore, a pure epithelial population, SCp2, has been derived from CID9 [<xref ref-type="bibr" rid="B10">10</xref>].</p><p>In our laboratory, we are particularly interested in the signalling pathways that regulate gene expression in the differentiating and involuting mouse mammary gland. Despite the undoubted value of these culture systems, expression of transgenes <italic>in vivo</italic> does not always recapitulate expression observed in culture. This reflects the complex requirements for mammary epithelial cell differentiation and apoptosis. There is a need, therefore, for mammary epithelial cell lines that more accurately mimic the developing and involuting gland. Such cells should preferably be immortalized by a conditional (ie reversible) mechanism (in contrast to currently available lines) and be able to be genetically modified.</p><p>In order to achieve this, we adopted a modification of the procedure used to generate 'immortomouse', a line of transgenic mice that harbour a temperature-sensitive variant of an immortalizing gene, SV40 T-Ag, which is expressed from a constitutive promoter H2K^b [<xref ref-type="bibr" rid="B11">11</xref>]. Although the 'immortomouse' shows thymic hyperplasia, these transgenic mice undergo normal development and have proven to be a useful source of material to establish cell lines from tissues that have previously been refractory to culturing [<xref ref-type="bibr" rid="B12">12</xref>]. Our attempts to establish mammary epithelial lines from these mice were unsuccessful, however. This may be due to insufficient levels of T-Ag being expressed in the mammary epithelial compartment to immortalize these cells or the presence of T-Ag in the other mammary compartments, thereby allowing the preferential immortalization of fibroblasts and other stromal cell types.</p><p>We therefore decided to refine this approach and target expression of the thermolabile T-Ag mutant specifically to the mammary epithelium of transgenic mice. Targeted expression can be achieved using either the mouse mammary tumour virus long terminal repeats or a milk protein gene promoter. We chose to use the promoter of the sheep milk protein gene encoding &#x003b2;-lactoglobulin (BLG), because BLG is less dependant than WAP on the transgene genomic integration site for its expression [<xref ref-type="bibr" rid="B13">13</xref>] BLG transgenes are expressed at low levels in the mammary glands of virgin mice, whereas expression is regulated during pregnancy and lactation with a similar expression profile to that of &#x003b2;-casein [<xref ref-type="bibr" rid="B14">14</xref>]. Therefore, it is likely that BLG expression occurs in dividing cells early in the differentiation pathway, a critical factor in establishing cultures from early stages of mammary gland development. Such cultures could contain epithelial stem cells because these are known to be distributed throughout the ductal tree [<xref ref-type="bibr" rid="B15">15</xref>]. We report herein the isolation and characterization of a stable line of mammary epithelial cells, named KIM-2, from mid-pregnant mammary glands of one line of mice with a low transgene copy number. Importantly, this cell line has a phenotypically normal epithelial morphology at 37&#x000b0;C and permits the analysis of the complex processes of differentiation and apoptosis <italic>in vitro</italic>. Moreover, we provide evidence that this line is susceptible to genetic manipulation, thus making available a resource for analysis of genetic function.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Construction of the transgene</title><p>A fusion gene, consisting of 4.2 kb of the 5^'-flanking promoter sequences, including the transcriptional start site, of the ovine BLG gene [<xref ref-type="bibr" rid="B16">16</xref>] and the temperature-sensitive variant of SV40 T-Ag (tsA58) coding sequences [<xref ref-type="bibr" rid="B11">11</xref>], was constructed.</p><p>The 4.2 kb Sall/EcoRV fragment of the BLG promoter was isolated from pBJ39 (provided by Dr CBA Whitelaw) and subcloned into pBluescript vector (Stratagene Europe, Amsterdam, The Netherlands). A 3.9 kb Bgll fragment was isolated from pUC tsA58 [<xref ref-type="bibr" rid="B11">11</xref>], blunt ended using T4 DNA polymerase and digested with BamHI, and a 2.7 kb fragment containing sequences encoding large and small T-Ag was purified. This fragment was fused to the EcoRV site of the BLG promoter.</p></sec><sec><title>Generation and identification of transgenic mice</title><p>For microinjection the transgene was isolated free of vector sequences by double digestion with Sall and Xbal, and the transgene was purified by agarose gel electrophoresis. DNA (1.5 ng/&#x003bc;l) was microinjected into pronuclear mouse eggs (collected from C57BL/6 &#x000d7;CBA F1 mice after mating with F1 male studs) in order to generate transgenic mice. Lines were maintained by mating F1 mice that harboured the transgene, as determined by polymerase chain reaction analysis of tail biopsies [<xref ref-type="bibr" rid="B17">17</xref>]. All DNA manipulations were carried out using standard procedures [<xref ref-type="bibr" rid="B18">18</xref>].</p></sec><sec><title>DNA analysis</title><p>DNA was extracted from tail biopsies of 6-weeks-old mice, and analyzed by polymerase chain reaction and Southern blots. Genomic tail DNA was digested with an appropriate restriction enzyme, subjected to agarose gel electrophoresis and transferred to nylon membrane (Hybond N; Amersham Pharmacia Biotech, Uppsala, Sweden) [<xref ref-type="bibr" rid="B19">19</xref>]. Southern blots were hybridized [<xref ref-type="bibr" rid="B20">20</xref>] with a random oligoprimed probe (Prime-It II kit; Stratagene) [<xref ref-type="bibr" rid="B21">21</xref>] containing sequences from the probe 1 region of the transgene.</p></sec><sec><title>RNA analysis</title><p>RNA was extracted from tissue or from cultured cells using the acid guanidium thiocyanate-phenol chloroform method [<xref ref-type="bibr" rid="B22">22</xref>]. For northern blot analysis, 10 &#x003bc;g total RNA was resolved on 1.0% formaldehyde agarose gels, transferred to nylon membranes (Hybond N; Amersham) and hybridized to [³²P] dCTP-labelled random primed probes. Two probes were used for &#x003b2;-casein [<xref ref-type="bibr" rid="B23">23</xref>] and WAP [<xref ref-type="bibr" rid="B24">24</xref>].</p></sec><sec><title>Explant cultures</title><p>Mammary glands were aseptically removed from a midpregnant (day 14) transgenic mouse carrying the BLG/SV40 T-Ag construct and washed several times in dissection medium (HEPES-buffered M199 with gentamycin at 50 &#x003bc;g/ml and BRL's antibiotic/antimycotic solution; Gibco/BRL, Paisley, Scotland). The tissue was cut with scalpels into pieces of about 1 mm thickness in fresh dissection medium; seeded into collagen type I-coated flasks; and cultured at either the permissive temperature of 33&#x000b0;C or the semipermissive temperature of 37&#x000b0;C for about 2 weeks with daily medium changes (1 ml/25 cm²) in serum-free F12/Dulbecco's modified eagle's medium (1:1), supplemented with bovine insulin, ovine prolactin, cortisol, oestradiol (each 5 &#x003bc;g/ml) and epidermal growth factor (EGF; 10 ng/ml). During this period, an epithelial outgrowth formed around most explants, without fibroblast contamination. Explants were removed by simply shaking them off and washing the flask, and the culture was continued in F12/Dulbecco's modified eagle's medium (1:1) supplemented with 10% foetal calf serum, 5 &#x003bc;g/ml insulin, 10 ng/ml EGF, 5 &#x003bc;g/ml linoleic acid and 5 &#x003bc;g/ml gentamycin.</p><p>Primary cultures were passaged after several weeks when the circumference of the islands stopped growing. Cells were passaged as clumps of about five to 10 cells rather than as single cells, because this appeared to aid survival and maintenance of the epithelial phenotype. In order to produce clumps, trypsinization was shortened and performed at room temperature, the cells scraped off with a cell scraper, and the cell suspension handled with widebore pipettes. Cells are now routinely cultured at 37&#x000b0;C and passaged every three to four days at a 1:4 ratio onto collagen-coated flasks (growth on collagen once in every five passages is sufficient) and maintained for approximately 20 passages. No change in phenotype is observed with careful handling.</p></sec><sec><title>Immunocytochemistry</title><p>Cells were grown subconfluently on collagen-coated glass coverslips in four-well plates or on plastic slide flasks (Nunc/Nalge Europe, Hereford, UK). The cells were fixed in methanol:acetone for 10 min, washed with Tris-buffered saline (TBS) pH 7.6, blocked in TBS+ 20% goat serum for 1 h, and then immunostained with a panel of primary antibodies. Monoclonal antibodies to cytokeratin 18 and 19 were from EB Lane and SV40 T-Ag antibodies were kindly provided by Dr DP Lane (Department of Biochemistry, University of Dundee, Dundee, UK). Murine smooth muscle actin antibody was obtained from Sigma (A2547) as were E-cadherin (U 3254), laminin (L9393) and vimentin (V5255) (Sigma-Aldrich, Gillingham, Dorset, UK). Rat monoclonal antibody to zonula occluden-1 (MAB1520) was from Hemicon (Chemicon International Inc, Temecula, CA, USA). Antibody binding was visualized with fluorescein isothiocyanate (FITC)-labelled secondary antibodies (Sigma). Images were analyzed by fluorescence microscopy and, in some cases, phase and fluorescence images were subsequently merged.</p></sec><sec><title>Electron microscopy</title><p>Differentiated cells were trypsinized, centrifuged and fixed in 3% glutaraldehyde in 0.1 mol/l sodium cacodylate/HCI buffer of pH 7.2-7.4 at 4&#x000b0;C for 48 h. After washing with distilled water (dH₂O) for 20 min, the samples underwent secondary fixation in 1% osmium tetroxide in dH₂O for 45 min at room temperature. Samples were then dehydrated with methylated spirits and absolute ethanol, before linking to propylene oxide for 10 min and impregnation in Emix resin (Fisons, Leicester, UK) overnight at room temperature. After polymerization for 18-24 h at 70&#x000b0;C, 90-nm sections were mounted on 300-mesh copper grids and stained using the uranyl acetate/lead citrate method. Finally, sections were examined and photographs taken using a Jeol 100CXXII transmission electron microscope (Jeol Ltd, Welwyn Garden City, Herts, UK).</p></sec><sec><title>Induction of milk protein gene expression</title><p>Cells were grown on plastic-coated or collagen-coated flasks in growth medium until confluent. After 2 days, hormone induction media was then added, consisting of growth medium without EGF supplemented with dexamethasone and ovine prolactin. Cultures were induced for up to 12 days with media changes every 2 days.</p></sec><sec><title>Western blot analysis</title><p>In order to detect casein in total cell extracts, cells (area 8 cm²) were washed in phosphate-buffered saline and lysed directly in 0.5 ml electrophoresis sample buffer (0.125 mol/l Tris HCI pH 6.8, 2% sodium dodecyl sulphate, 2% &#x003b2;-2-mercaptoethanol, 10% glycerol), shearing the DNA by repetitive pipetting, boiled for 10 min and stored at-20&#x000b0;C. The protein concentrations of the samples were determined using the BCA Protein Assay Reagent kit (Pierce and Warriner, Chester, UK).</p><p>One dimensional sodium dodecyl sulphate-polyacrylamide gel electrophoresis was performed as described by Laemmli [<xref ref-type="bibr" rid="B25">25</xref>] in 15% polyacrylamide gels with a 3% stacking gel. Proteins were transferred from gels to nitrocellulose filters (Schleicher and Schuell, London, UK) at a current of 0.8 mA/cm² for 1 h using a semidry electroblotter [<xref ref-type="bibr" rid="B26">26</xref>]. After blocking nonspecific binding with 1% bovine serum albumin in phosphate-buffered saline/Tween (0.1%), the nitrocellulose was exposed to a polyclonal rabbit antimouse &#x003b2;-casein antibody diluted 1:10 000 in blocking solution. Primary antibodies were visualized by peroxidase-conjugated anti-lg antibodies and ECL detection reagents (Amersham).</p></sec><sec><title>Electrophoretic mobility shift assays</title><p>Electrophoretic mobility shift assays were carried out using the highest affinity STAT binding site (STM) in the BLG promoter, as previously described [<xref ref-type="bibr" rid="B8">8</xref>]. Briefly, 4 &#x003bc;g protein from nuclear extracts were incubated with ³²P-labelled STM oligonucleotide in binding buffer, and analyzed on 6% native acrylamide gels, followed by autoradiography.</p></sec><sec><title>Acridine orange staining and fluorescence microscopy</title><p>KIM-2 cells were grown to confluency, washed in phosphate-buffered saline and incubated in Dulbecco's modified eagle's medium/F12 containing 3% serum (but no additional growth factors) for 48 h. Cells (1&#x000d7;10⁶) were fixed in 70% ethanol and stained with acridine orange (5 &#x003bc;g/ml; Molecular Probes Europe BV, Leiden, The Netherlands). Classical features of apoptosis were identified with fluorescence microscopy.</p></sec><sec><title>Flow cytometry and annexin V assay</title><sec><title>Undifferentiated cells</title><p>KIM-2 cells were grown as above and 24 h later cells from the supernatant and monolayer were harvested. Cells (1&#x000d7;10⁵) were stained with annexin V and propidium iodide using the Apoptosis Detection Kit and following the manufacturer's instructions (R&#x00026;D Systems, Abingdon, Oxford, UK). Cells were analyzed by flow cytometry using a Coulter EPICS XL flow cytometer (Beckman Coulter, High Wycombe, Bucks, UK). Debris and clumps were gated out using forward and orthogonal light scatter. Green fluorescence (525 nm; FITC annexin V) and red fluorescence (613 nm; propidium iodide) of 2000 cells was measured. The experiment was repeated three times.</p></sec><sec><title>Differentiated cells</title><p>KIM-2 cells were differentiated for 12 days with the lactogenic hormones insulin, dexamethasone and prolactin. Apoptosis was induced by removal of these hormones and measured 17 h later, as above.</p></sec></sec></sec><sec><title>Results</title><sec><title>Generation of &#x003b2;-lactoglobulin-tsA58 transgenic mice</title><p>In order to direct expression of the conditional immortalizing gene to mammary epithelium, a hybrid construct containing the promoter of the sheep BLG milk protein gene, which is expressed predominantly in ductal and alveolar cells, was fused to the sequence encoding the temperature-sensitive early region variant of SV40 T antigen (T Ag), tsA58. This construct (Fig. <xref ref-type="fig" rid="F1">1a</xref>) was used to generate transgenic mice by microinjection into fertilized oocytes from (CBA&#x000d7; C57BL/6) F1 mice and in total 13 lines were generated. Several lines developed tumours in a number of anatomical sites as a consequence of ectopic expression of the transgene and leakiness of the temperature-sensitive mutation at the body temperature of the mice, and eight founders died before further analysis could be carried out. The severity of the phenotype in the surviving lines appeared to be copy number-related and was retained through subsequent generations of mice. The remaining five founders were used to derive transgenic lines. The number of copies of the transgene in each line was determined by Southern blotting (Fig. <xref ref-type="fig" rid="F1">1b</xref>) of DNA extracted from liver tissue and probed with a fragment of the BLG promoter and ranged from one to two copies (lines 2 and 8) to approximately 10 copies (line 13). One of the lowest copy number lines, line SV40-2, was used for the isolation of the KIM-2 cell line described herein. This line does not display a tumour phenotype due to ectopic expression of T-Ag.</p></sec><sec><title>Expression of the transgene in simian virus 40 line 2</title><p>The level of expression of T-Ag in the mammary glands of line SV40-2 was determined. Low levels of tsA58 T-Ag mRNA were detected (data not shown). Protein was analyzed by western blotting in extracts from the lactating mammary glands of three sisters from this line (Fig. <xref ref-type="fig" rid="F2">2a</xref>). Levels of protein were consistent within this line. Expression of T-Ag did not perturb the normal development of the mammary gland and no mammary tumours were observed in this, or in any of the lines, despite T-Ag being present. This is shown in Figure <xref ref-type="fig" rid="F2">2b</xref>, which shows sections of tissue from virgin and day 11 lactation mammary glands of line SV40-2 and control F1 mice, showing the normal morphology. No lactational defects were apparent and these mice successfully nursed their offspring. This transgenic line has now been maintained for several years and mammary tumours have not been observed. This contrasts with the observation of tumour development in the WAP-T-Ag mice. This is probably a consequence of wild-type T-Ag being expressed in the WAP transgenics [<xref ref-type="bibr" rid="B27">27</xref>].</p></sec><sec><title>Derivation of conditionally immortal cell lines</title><p>Enrichment procedures have been described for partially purifying epithelial cells from the other types of cells that are present in the mammary gland [<xref ref-type="bibr" rid="B28">28</xref>]. These methods use enzymatic treatment and differential centrifugation to obtain a clean, fibroblast-free population of primary epithelial cells, and did not work well in our hands with the transgenic mice. We therefore developed a novel, simpler method that is based on the unpublished observation of one of us (BB) that, on a collagen-coated surface, fibroblast outgrowth from mammary explants requires the presence of serum, whereas epithelial cells grow out easily in a chemically defined medium. Initially, explant cultures were established from mammary glands excised from several lines of mice, with different transgene copy numbers. This procedure was most successful with tissue from the lowest copy number line SV40-2, and subsequent work was carried out with mammary tissue from this line. Mammary explant cultures were prepared by slicing the glands into 1 mm square pieces, seeding into collagen type 1 coated flasks and culturing for 2 weeks at the permissive (33&#x000b0;C) and semipermissive temperature (37&#x000b0;C) until outgrowths were well established (Fig. <xref ref-type="fig" rid="F3">3a</xref>). The explants were removed and the cells cultured until the islands stopped expanding in size. During this time period dome-like structures began to appear only in the cultures grown at 37&#x000b0;C (Fig. <xref ref-type="fig" rid="F3">3b</xref>).</p><p>The primary cultures were routinely passaged as clumps of five to 10 cells, because single cell passaging results in the appearance of cells with a different morphology that was reminiscent of a fibroblastic phenotype. These cells are now routinely passaged every 3-4 days as clumps and divided 1:3 or 1:4. These cells, which we have designated KIM-2, have been maintained in culture at 37&#x000b0;C for over 60 passages and have a stable phenotype both on plastic and collagen as assessed by immunocytochemistry.</p></sec><sec><title>Characterization of the cell line</title><sec><title>Effects of growth temperature on morphology</title><p>T-Ag immortalizes cells by complexing with the oncosuppressor proteins p53 and retinoblastoma protein [<xref ref-type="bibr" rid="B29">29</xref>]. It was expected that the temperature-sensitive variant of T-Ag used to generate the transgenic animals would be nonfunctional at 37&#x000b0;C. KIM-2 cells can be isolated and grown at 37&#x000b0;C for over 60 passages, however, indicating that sufficient T-Ag is functional at this temperature to effect immortalization (Fig. <xref ref-type="fig" rid="F4">4a</xref>). In contrast, cells isolated from mammary tissue at 33&#x000b0;C show an elongated spindlelike morphology (Fig. <xref ref-type="fig" rid="F4">4b</xref>), grow rapidly and immunostain with an antibody to a mesenchymal specific marker, vimentin (data not shown). These cells also form colonies in soft agar and may therefore be transformed; this is in contrast to the cells at 37&#x000b0;C, which do not form colonies (Fig. <xref ref-type="fig" rid="F4">4c</xref> and <xref ref-type="fig" rid="F4">d</xref>, and Table <xref ref-type="table" rid="T1">1</xref>). A more accurate assessment of ability to form tumours requires growth in mice, and we are awaiting results of injecting these cell types into severe combined immunodeficiency mice. The temperature of 39&#x000b0;C is nonpermissive, and KIM-2 cells grown at this temperature undergo a crisis and die after two or three passages. Because KIM-2 cells are phenotypically 'normal' at 37&#x000b0;C and grow rapidly, dividing approximately once every 24 h, we selected this growth temperature for the experiments described in the present study, unless otherwise indicated.</p></sec><sec><title>Expression of specific cell markers</title><p>The characteristics of the KIM-2 cell line were established by immunocytochemistry. Antibody staining patterns suggest that these cultures are highly enriched for cells of luminal epithelial type, with over 95% of the cells staining strongly positive for keratin 18 (Fig. <xref ref-type="fig" rid="F5">5d</xref>), a luminal cell marker [<xref ref-type="bibr" rid="B30">30</xref>]. Approximately 5% of the cells stained with either smooth muscle actin (Fig. <xref ref-type="fig" rid="F5">5c</xref>), which is characteristic of myoepithelial cells, or vimentin, a stromal and fibroblastic marker (Fig. <xref ref-type="fig" rid="F5">5e</xref>). Most cells show nuclear staining with a T-Ag-specific antibody (Fig. <xref ref-type="fig" rid="F5">5b</xref>). The keratin 14-specific antibody LL002 immunostains many cells. The epitope recognized by this antibody is frequently deregulated in cultured cells, however. The proportion of cells that stain for actin remains constant, even after multiple passages. These are probably of myoepithelial origin and derived from KIM-2 cells that have divided to give two types of cells (myoepithelial and luminal). This notion is supported from observations made on clonal derivatives (discussed below). Confluent cultures of KIM-2 cells deposit laminin, a component of the basement membrane, as shown by strong extracellular staining with an antibody to laminin (Fig. <xref ref-type="fig" rid="F6">6</xref>).</p></sec><sec><title>Induction of differentiation</title><p>Addition of lactogenic hormones to confluent cultures of KIM-2 cells resulted in a morphological change after approximately 2 days. Domes appeared in the confluent monolayer. These were substantial in size, usually containing upwards of 50 cells, and staining of these differentiated cultures with antibodies to &#x003b1;-actin showed the presence of myoepithelial cells around the domes (Fig. <xref ref-type="fig" rid="F7">c</xref> and <xref ref-type="fig" rid="F7">d</xref>). This association of luminal and myoepithelial cells is reminiscent of alveolar structures. Immunocytochemistry with antibodies to the cell adhesion molecule E-cadherin (data not shown) and the junction protein zonula occluden-1 (Fig. <xref ref-type="fig" rid="F7">7e</xref> and <xref ref-type="fig" rid="F7">f</xref>), which is found at the apical and lateral plasma membrane boundaries between epithelial cells, showed that tight junctions are formed between the cells. Confluent undifferentiated monolayers of KIM-2 cells also form junctions as shown by the E-cadherin staining pattern (Fig. <xref ref-type="fig" rid="F7">7a</xref> and <xref ref-type="fig" rid="F7">b</xref>). Using time lapse video microscopy, these domes were observed to pulsate, suggesting that transepithelial fluid transport is taking place into an expanding lumen and confirming that tight junctions have been formed (<ext-link ext-link-type="uri" xlink:href="http://www.breast-cancer-research.com/content/2/3/v1"/>).</p><p>The differentiation of KIM-2 cells was further investigated using electron microscopy, which revealed the presence of milk protein and lipid droplets within the KIM-2 cells. Differentiated cells are polarized, microvilli being present on the apical surface (Fig. <xref ref-type="fig" rid="F8">8a</xref>). We also observed desmosomes between cells (Fig. <xref ref-type="fig" rid="F8">8b</xref>).</p></sec><sec><title>Expression of differentiation markers</title><p>Milk protein synthesis <italic>in vivo</italic> occurs within clusters of differentiated mammary epithelial cells in response to lactogenic hormones, ECM interactions and cell-cell interactions [<xref ref-type="bibr" rid="B1">1</xref>]. In cultures of KIM-2 cells grown at the semipermissive temperature of 37&#x000b0;C or the nonpermissive temperature of 39&#x000b0;C, addition of lactogenic hormones induced partial differentiation as assessed by the synthesis of high levels of &#x003b2;-casein.</p><p>Figure <xref ref-type="fig" rid="F9">9a</xref> shows a western blot of induced and uninduced late passage cells (P31) grown on tissue culture plastic. For induction, the cells are grown to confluence, EGF removed for 2 days, and the lactogenic hormones dexamethasone, insulin and prolactin added for the time period indicated. There appear to be at least two different forms of cytoplasmic &#x003b2;-casein induced. One has an apparent molecular mass of 29 kDa and there is also a lower molecular weight band, compared with 32 kDa for the secreted protein found in milk. The discrepancy in size is probably due to differences in post-translational modifications between intracellular &#x003b2;-casein and the secreted form, and was also observed in HC11 mammary cells (Fig. <xref ref-type="fig" rid="F9">9a</xref>). The level of &#x003b2;-casein induction is not affected by passage number or growth on collagen (data not shown). KIM-2 cells do secrete &#x003b2;-casein but at very low levels.</p><p>Other milk proteins that are expressed later in pregnancy, such as WAP, appear to require other factors in addition to lactogenic hormones for their expression. We were interested in investigating the possibility of WAP expression in KIM-2 cultures, because this is a more appropriate measure of differentiation status. Figure <xref ref-type="fig" rid="F9">9b</xref> shows a time course of WAP mRNA induction with lactogenic hormones in KIM-2 cells grown on tissue culture plastic. Whereas &#x003b2;-casein expression was induced after 2 days, WAP expression was delayed until 4-8 days after hormone stimulation. These kinetics reflect the pattern of milk protein gene expression in the mammary gland, in which &#x003b2;-casein is expressed from day 10 of pregnancy but WAP is not expressed until approximately 4 days later. Recently a STAT5a knockout has been generated that exhibits an impairment of lobuloaveolar development during pregnancy and an inability to lactate [<xref ref-type="bibr" rid="B31">31</xref>]. Although these mice show normal &#x003b2;-casein expression, the levels of WAP expression were reduced considerably. Clearly the induction of WAP expression has more complex requirements that may include contacts with the ECM, expression of specific receptors and/or expression of differentiation-related genes. It is possible that prolonged stimulation of KIM-2 cells with lactogenic hormones induces these changes.</p></sec><sec><title>Signal transduction</title><p>STAT5 is a signalling molecule for prolactin and may also be a survival factor for differentiated mammary epithelia. We therefore investigated the activation of STAT5 in KIM-2 cells stimulated with prolactin. Figure <xref ref-type="fig" rid="F10">10a</xref> shows a time-course of STAT5 induction as assayed by electrophoretic mobility shift assay. Detectable amounts of STAT5 were observed within 5 min of prolactin addition, reaching a peak between 30 and 60 min. Interestingly, the response to prolactin was biphasic. Following the initial peak at 30 min, the response declined to almost basal levels before again reaching a peak at about 24 h that was then sustained for at least 8 days. This kinetic profile is not surprising, because high STAT5 level are maintained in the mammary gland during lactation in the continued presence of prolactin. The immediate downregulation could be in response to phosphatases, which have been shown to be associated with cytokine receptors and may interact directly with STAT factors [<xref ref-type="bibr" rid="B32">32</xref>], or to the upregulation of other genes such as members of the suppressor of cytokine signalling (SOCS) family of negative regulators [<xref ref-type="bibr" rid="B33">33</xref>]. It will be interesting to determine the mechanism of this early regulatory response, which is a feature of STAT activation in other cell systems. The prolonged activation of STAT5 could reflect the need for a survival factor for differentiated cells and may be associated with withdrawal from the cell cycle. Antibodies that specifically recognize either STAT5a or STAT5b supershift the complex observed, suggesting that both forms of STAT5 are activated in KIM-2 cells (data not shown). This parallels observations in the mammary gland [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>].</p><p>These characteristics suggest that KIM-2 cells are an excellent model for mammary epithelial cell differentiation. In order to use this model to investigate the molecular mechanisms of signal transduction and gene expression in mammary cells, it is essential that these cells can be genetically modified. This has been achieved using a variety of transfection methods to mediate DNA transfer, although retroviral transduction is the most efficient. We have now isolated a number of clonal cell lines from KIM-2 cells transfected with green fluorescent protein constructs. These clonal lines are morphologically similar to KIM-2 cells and can be induced to form domes and differentiate [<xref ref-type="bibr" rid="B36">36</xref>]. A similar number of cells staining for &#x003b1;-actin are observed in cultures of clonal lines, suggesting that KIM-2 cells are progenitor cells that can differentiate into myoepithelial or luminal epithelial types. Similar observations have been made for human breast cells [<xref ref-type="bibr" rid="B37">37</xref>].</p></sec></sec><sec><title>Induction of apoptosis in KIM-2 cells</title><p>The mammary gland undergoes a process of remodelling after weaning in which the lobuloalveolar compartment is removed by apoptosis. In order to investigate the molecular basis of involution, it would be invaluable to have available a mammary epithelial cell line that can be induced to undergo apoptosis in response to physiological signals. We therefore tested a variety of growth conditions for their ability to induce apoptosis and found that undifferentiated KIM-2 cells cultured in reduced serum (3%), with no added growth factors, die over a period of 24-28 h.</p><p>Apoptosis was classically defined by the morphological changes that occur, including cell shrinkage, chromatin condensation and nuclear fragmentation [<xref ref-type="bibr" rid="B38">38</xref>]. These changes were visualized using acridine orange staining and fluorescence microscopy and are obvious in dying KIM-2 cells compared with healthy cultures where the cells have diffusely stained nuclei (Fig. <xref ref-type="fig" rid="F11">11</xref>).</p><p>Although this method allows a clear definition of apoptosis, it does not allow rapid and precise quantification of the process. One of the early changes that occur during apoptosis is the translocation of phosphatidylserine from the inner to the outer surface of the plasma membrane, presumably enabling the recognition of apoptotic cells by phagocytes [<xref ref-type="bibr" rid="B39">39</xref>]. This can be detected with annexin V and quantified by flow cytometry. Figure <xref ref-type="fig" rid="F12">12a</xref> shows the flow cytometric analysis of control cells and cells induced to die 24 h after treatment; the percentage of cells in each of the four populations is shown. A small number of control cells stain with propidium iodide and annexin V, probably as a consequence of membrane disruption during the harvesting procedure. An approximately eightfold increase in apoptosis was seen when KIM-2 cells were cultured in reduced serum with no additional growth factors. An increase in the number of cells that stained with both propidium iodide and annexin V was also observed, probably reflecting the late loss of membrane integrity that occurs in cells in culture when they are not removed by phagocytosis.</p><p>Apoptosis of differentiated KIM-2 cultures would be a more useful model of apoptosis during involution <italic>in vivo</italic>. We therefore induced apoptosis in differentiated KIM-2 cultures by removal of the lactogenic hormones prolactin, dexamethasone and insulin. Approximately 30% of the cells underwent apoptosis after 17 h in the absence of these hormones (Fig. <xref ref-type="fig" rid="F12">12b</xref>). Control cells exhibit an apparently high level of apoptosis. This is a consequence of membrane disruption during the harvesting procedure, because these differentiated cells are tightly associated and difficult to disperse. Interestingly, in reduced serum (3%) differentiated KIM-2 cells, in contrast to undifferentiated cells, did not die, suggesting that at least one of the lactogenic hormones is a survival factor for functionally differentiated mammary epithelial cells (Fig. <xref ref-type="fig" rid="F12">12b</xref>).</p><p>Time-lapse video microscopy of differentiated cultures, induced to die by hormone withdrawal, shows cells detaching from the dish and the eventual collapse of the domes (<ext-link ext-link-type="uri" xlink:href="http://www.breast-cancer-research.com/content/2/3/v2"/>).</p><p>The ease of inducing and quantitatively measuring apoptosis, coupled with the ability to manipulate KIM-2 cells genetically, should allow a detailed analysis of the early transcription events that regulate this process. The relevance of these genes to involution in the mammary gland can then be investigated.</p></sec></sec><sec><title>Discussion</title><p>The morphological changes that take place during a mammary gland developmental cycle have been well defined and characterized. However, the molecular mechanisms that are involved in the regulation of lobuloalveolar development during pregnancy and the removal of this compartment by apoptosis during involution are just beginning to be understood, partly because it has been difficult to obtain a suitable <italic>in vitro</italic> model system that accurately mimics mammary gland development. The aim of this work was to establish a mammary cell culture system that would be of value in biochemical and functional studies of the control of differentiation and apoptosis.</p><p>To achieve this aim, we adapted the approach developed by Jat <italic>et al</italic> [<xref ref-type="bibr" rid="B11">11</xref>] and derived a mammary-specific 'immortomouse' using the BLG milk protein milk promoter to drive expression of a temperature-sensitive allelle of SV40 large T-Ag. We also developed a novel culture system to enrich for luminal epithelial cells. These procedures allowed us to successfully derive a novel line of conditionally immortal mammary epithelial cells. These cells retain a stable phenotype that is characteristic of luminal epithelial cells at the normal growth temperature of 37&#x000b0;C as evidenced by immunocytochemical analysis using lineage specific markers. At 39&#x000b0;C, the cells are not immortal and die within a few passages.</p><p>The growth characteristics of the KIM-2 cell strain were dramatically influenced by the culture temperature. Isolating epithelial cells at the permissive temperature of 33&#x000b0;C resulted in apparently transformed cells with a spindle morphology that form colonies in soft agar. This phenotype cannot be reverted by switching up the growth temperature to 37&#x000b0;C. Switching down the growth temperature from 37&#x000b0;C to 33&#x000b0;C, however, caused a gradual change in morphology from cuboidal epithelial to a more fibroblastic appearance over a period of several weeks. Immunocytochemical analysis of these cells revealed the presence of vimentin and &#x003b1;-actin, along with keratin 18, suggesting that conversion of the epithelial cells to a more fibroblastic or myoepithelial cell type had taken place. After a switch down in growth temperature from 37&#x000b0;C to 33&#x000b0;C for 48 h or more, the phenotypic switch became irreversible.</p><p>Of the 13 transgenic founder lines generated, the lowest copy number line SV40-2 was used as the source of mammary cells for these experiments. Surprisingly, expression of T-Ag RNA in the mammary glands of line SV40-2 mice was barely detectable by northern blot analysis (data not shown), although T-Ag could be detected by western blot (Fig. <xref ref-type="fig" rid="F2">2a</xref>), suggesting that the BLG promoter is dysregulated when fused to T-Ag sequences. The reason for this is not clear but it may be a consequence of the integration site of the transgene and the lack of sequences from within the body of the BLG gene and/or in its 3^'-terminal end, which confer integration site independence [<xref ref-type="bibr" rid="B40">40</xref>]. Whatever the mechanism, the low levels of T-Ag messenger RNA encode sufficient protein for the immortalizing role of SV40 T-Ag, which is readily detectable in the nuclei of cells in culture (Fig. <xref ref-type="fig" rid="F5">5b</xref>). The low expression of T-Ag may explain our observation that no mammary tumours were found, even in the animals that developed tumours at ectopic sites. Ectopic expression of BLG transgenes has been observed previously and high levels of T-Ag mRNA were found in the tumours.</p><p>KIM-2 cells have a characteristic epithelial morphology. Immunocytochemical analysis demonstrated the presence of the luminal-specific marker keratin 18 in the majority of the cells and the absence of fibroblastic markers. Some myoepithelial cells were observed that were characterized by their elongated morphology and positive staining with antibodies to &#x003b1;-actin. These cells were frequently found to be associated with the dome-like structures that formed in response to lactogenic hormone induction (Fig. <xref ref-type="fig" rid="F7">7</xref>) and may suggest that both cell types are necessary for the formation of these alveolar-like structures. We have derived a substantial number of clonal lines from KIM-2 cells by transfection of selectable markers. These clonal derivatives also form dome structures and have a myoepithelial component suggesting that KIM-2 cells are a progenitor of both luminal and myoepithelial cells. Selective growth of these cell types [<xref ref-type="bibr" rid="B37">37</xref>] would provide a resource for the discovery of genes, using microarray technology, which are expressed specifically in one type of cell [<xref ref-type="bibr" rid="B41">41</xref>]. The exciting possibility that KIM-2 cells may be stem cells will be tested by their ability to repopulate a cleared mammary fat pad. If this is successful, genetically modified KIM-2 cells could be used to generate transgenic mammary glands [<xref ref-type="bibr" rid="B42">42</xref>].</p><p>The formation of cell-cell contacts was investigated by using antibodies to E-cadherin and zonula occluden-1 and by electron microscopy. Staining at cell contacts was observed for both E-cadherin and zonula occluden-1 (Fig. <xref ref-type="fig" rid="F7">7</xref>). The formation of structures resembling desmosomes was seen by electron microscopy, and the presence of microvilli on the apical surfaces of cell clusters suggests that the cells are polarized, which is a feature of alveolar cells <italic>in vivo</italic>. The synthesis and deposition of laminin by KIM-2 cells that we have observed may be essential for polarization in addition to cell-cell contact.</p><p>Polarization is also essential for the expression and secretion of milk proteins. We showed previously that growth on ECM (principally laminin) is necessary for the activation of STAT5 in response to prolactin in primary cultures [<xref ref-type="bibr" rid="B43">43</xref>]. It is likely, therefore, that the synthesis and deposition of laminin by KIM-2 cells provides the necessary extracellular signals for the activation of sufficient levels of STAT5 for induction of full differentiation. Indeed, it has been shown that the binding of STAT5 to the WAP promoter is essential for maximal expression [<xref ref-type="bibr" rid="B44">44</xref>]. It is likely that STAT5 is a survival factor for differentiated mammary epithelial cells, and this is currently being tested.</p><p>Undifferentiated KIM-2 cells can be induced to undergo apoptosis by the removal of growth factors in serum. Apoptosis is extensive and rapid with approximately 30% of undifferentiated cells dying after 24 h. More relevantly, apoptosis can be induced in fully differentiated KIM-2 cells after removal of lactogenic hormones. Apoptosis has been studied in a variety of mouse mammary epithelial cell lines. KIM-2 cells are an important addition to this repertoire, because they exhibit different features. Differentiated cells do not require addition of exogenous basement membrane for survival (unlike CID9 cells, which die with delayed kinetics in the absence of basement membrane [<xref ref-type="bibr" rid="B45">45</xref>]) and have wild-type p53 (unlike HC11 cells, which harbour mutant p53 allelles [<xref ref-type="bibr" rid="B46">46</xref>]). Apoptosis in individual cells can be observed using fluorescence microscopy, and this can be coupled with detection of other markers such as green fluorescent protein reporter constructs [<xref ref-type="bibr" rid="B36">36</xref>]. Thus, the KIM-2 cell line will be of value in future studies on early transcription events in the apoptotic process.</p><p>The low copy number line of BLG/SV40tsT mice can be crossed with other mice that harbouring either transgenes or gene deletions, thereby allowing the isolation of mammary cell lines from interesting mouse mutants. This will allow us to establish cell lines from knockout mice and should be invaluable for transfection studies that are aimed at identifying functional domains and interactions between signalling pathways in mammary epithelium.</p></sec>

<italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in central and southern Italian patients

<sec><title>Introduction:</title><p>Germline <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations account for most hereditary breast/ovarian cancers and are associated with male breast cancer. Furthermore, constitutional mutations in these genes may occur in breast/ovarian cancer patients that do not meet stringent criteria of autosomal-dominant predisposition. The relevance of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in such patients is still debated.</p></sec><sec><title>Objectives:</title><p>We sought to determine the impact of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in a population of patients from central and southern Italy. We analyzed the <italic>BRCA1</italic> and <italic>BRCA2</italic> coding regions in 136 unrelated probands: 117 females with breast/ovarian cancer and 19 males with breast cancer. This population of patients was mostly representative of cases who are at risk for hereditary susceptibility, but who do not meet stringent criteria of autosomal-dominant predisposition.</p></sec><sec><title>Methods:</title><p>Probands, subclassified as follows, were consecutively recruited depending on informed consent from patients attending breast cancer clinics in Rome and Naples. Selection criteria for females were as follows: breast cancer with breast cancer family history [one to two first-/second-degree relative(s), <italic>n</italic> = 55]; breast cancer diagnosed before age 40 years (no breast/ovarian cancer family history, <italic>n</italic> = 28); bilateral breast cancer (regardless of age and family history, <italic>n</italic> =10); breast cancer associated with gastrointestinal, pancreatic or uterine cancers [synchronous/metachronous or in first-degree relative(s), <italic>n</italic> = 9]; breast or ovarian cancer with family history of breast-ovarian/ovarian cancer (at least 1 first-/ second-degree relative, <italic>n</italic> = 10); and ovarian cancer with no breast/ovarian cancer family history (<italic>n</italic> = 5). Males with breast cancer were recruited regardless of age and family history. <italic>BRCA1</italic> exon 11 and <italic>BRCA2</italic> exons 10 and 11 were screened by PTT. Coding <italic>BRCA1</italic> exons 2, 3, 5-10 and 12-24 and <italic>BRCA2</italic> exons 2-9 and 12-27 were screened by SSCP. Primers are listed in Table <xref ref-type="table" rid="T1">1</xref>. In 27 cases, analyzed by PTT along the entire <italic>BRCA1</italic> coding sequence, <italic>BRCA1</italic> SSCP analysis was limited to exons 2, 5, 20 and 24. Mutations were verified by sequence analysis on two independent blood samples.</p></sec><sec><title>Results:</title><p>Deleterious germline <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations were detected in 11 out of 136 cases (8%). Only three <italic>BRCA2</italic> mutations were novel. One <italic>BRCA2</italic> mutation recurred in two unrelated probands. Table <xref ref-type="table" rid="T2">2</xref> shows the mutations and data concerning carriers and their families. Table <xref ref-type="table" rid="T3">3</xref> shows correlations between <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations and sex, age at disease diagnosis and familial clustering of breast/ovarian cancer in the total patient population. Table <xref ref-type="table" rid="T4">4</xref> shows the proportions of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in females with site-specific breast and breast-ovarian/ovarian cancer. Table <xref ref-type="table" rid="T5">5</xref> shows the frequency of <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations in males. <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations, respectively, accounted for four out of 68 (6%) and one out of 68 (1%) cases diagnosed before age 50 years, and for one out of 68 (1%) and five out of 68 (7%) cases diagnosed after age 50 years. <italic>BRCA1</italic> mutations were found in five out of 117 females (4%) and in none of 19 males (0%), and <italic>BRCA2</italic> mutations were found in four out of 117 females (3%) and in two out of 19 males (10%). The proportions of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations coincided in site-specific female breast cancers (four out of 102; ie 4% each). <italic>BRCA1</italic> and <italic>BRCA2</italic> equally contributed to female breast cancers, with no familial clustering in those diagnosed before age 40 years (one out of 28; 4% each), and to female breast cancers, all ages, with familial clustering in one to two relatives (three out of 55; ie 5% each). In the latter subset of cases, <italic>BRCA1</italic> mostly accounted for tumours diagnosed before age 40 years (two out of eight; 25%), and <italic>BRCA2</italic> for tumours diagnosed after age 50 years (three out of 34; 9%). Regardless of family history, the respective contributions of <italic>BRCA1</italic> and <italic>BRCA2</italic> to site-specific female breast cancers diagnosed before age 40 years were 8% (three out of 36) and 3% (one out of 36). One <italic>BRCA1</italic> mutation was detected among the 15 female probands from breast-ovarian/ovarian cancer families (7%). Among male breast cancers, <italic>BRCA2</italic> mutations were identified in one out of five (20%) cases with family history and in one out of 14 (7%) apparently sporadic cases. No <italic>BRCA1</italic> or <italic>BRCA2</italic> mutations were found in female probands with nonfamilial bilateral breast cancer (10 cases) or in those with breast cancer associated with gastrointestinal, pancreatic or uterine cancers, synchronous/metachronous or in first-degree relative(s) (nine cases). These cases were all diagnosed after age 40 years.</p></sec><sec><title>Discussion:</title><p>Our results indicate a lack of relevant founder effects for <italic>BRCA1-</italic> and <italic>BRCA2</italic>-related disease in the sample of patients studied, which is consistent with other Italian studies and with ethnical and historical data. Overall, the contribution of <italic>BRCA1</italic> and <italic>BRCA2</italic> to breast/ovarian cancer in Italian patients appears to be less significant than in patients from communities with founder mutations. The present study is in agreement with direct estimates on other outbred populations, indicating that 7-10% of all female breast cancers that occur in patients aged under 40 years are due to <italic>BRCA1</italic>/<italic>BRCA2</italic>.</p><p>We found that <italic>BRCA1</italic> and <italic>BRCA2</italic> equally contributed to site-specific breast cancers who had one/two breast cancer-affected first-/second-degree relative(s) or who were diagnosed within age 40 years in the absence of family history. This is consistent with recent data that indicated that the respective frequencies of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations are comparable in early onset breast cancer. Considering the total population of patients analyzed here, however, <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations were mostly found in cases with disease diagnosis before and after age 50 years, respectively. Moreover, in cases with familial clustering of site-specific breast cancer, <italic>BRCA1</italic> mostly accounted for tumours diagnosed before age 40 years, and <italic>BRCA2</italic> for tumours diagnosed after age 50 years. This is in agreement with a trend, which has been observed in other populations, for the proportion of cases with <italic>BRCA2</italic> mutations to increase, and the proportion with mutations in <italic>BRCA1</italic> to decrease, as the age at cancer onset increases.</p><p>As in other studies, the frequency of <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations taken together was lower than the estimated frequencies at comparable ages for all susceptibility alleles derived from the Contraceptive and Steroid Hormones (CASH) study. The discrepancy between direct data deriving from <italic>BRCA1</italic>/<italic>BRCA2</italic> mutational analysis and CASH estimates could be due to several factors, including contribution of gene(s) other than <italic>BRCA1</italic>/<italic>BRCA2</italic>, differences between populations and relative insensitivity of mutational screening. Only <italic>BRCA1</italic> mutations were found in breast/ovarian and site-specific ovarian cancer families. <italic>BRCA2</italic>, but not <italic>BRCA1</italic> mutations were found in the male breast cancers. The overall proportion of males with <italic>BRCA2</italic> mutations was high when compared with data from other studies on outbred populations, but was low compared with data from populations with founder effects.</p><p>The present results should be regarded as an approximation, because the following types of mutation are predicted to escape detection by the screening strategy used: mutations in noncoding regions; missense mutations within <italic>BRCA1</italic> exon 11 and <italic>BRCA2</italic> exons 10 and 11; large gene deletions; and mutations within the first and last 180 nucleotides of the amplicons analyzed by PTT.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Ottini</surname><given-names>Laura</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>D'Amico</surname><given-names>Cristina</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Noviello</surname><given-names>Cristiana</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Lauro</surname><given-names>Salvatore</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Lalle</surname><given-names>Maurizio</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Fornarini</surname><given-names>Giuseppe</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Colantuoni</surname><given-names>Orsola Anna</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Pizzi</surname><given-names>Claudia</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Cortesi</surname><given-names>Enrico</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A10" contrib-type="author"><name><surname>Carlini</surname><given-names>Sandro</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A11" contrib-type="author"><name><surname>Guadagni</surname><given-names>Fiorella</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A12" contrib-type="author"><name><surname>Bianco</surname><given-names>Angelo Raffaele</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A13" contrib-type="author"><name><surname>Frati</surname><given-names>Luigi</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A14" contrib-type="author"><name><surname>Contegiacomo</surname><given-names>Alma</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A15" contrib-type="author"><name><surname>Mariani-Costantini</surname><given-names>Renato</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Breast Cancer Research : BCR

<sec><title>Introduction</title><p>The proportion of breast cancers that are attributable to autosomal-dominant susceptibility genes is estimated to be approximately 7% in the general population [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. Germline mutations of the <italic>BRCA1</italic> and <italic>BRCA2</italic> genes are estimated to contribute to the majority of the breast cancers that have very early disease onset, strong family history and/or association with ovarian cancer [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. <italic>BRCA1</italic> and <italic>BRCA2</italic> also account for a proportion of common breast/ovarian cancer patients that typically do not meet stringent criteria for highly penetrant autosomal-dominant cancer predisposition, but rather report one or two disease-affected relatives and/or manifest an early disease onset [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. Knowledge of the contribution of <italic>BRCA1</italic> and <italic>BRCA2</italic> to breast cancer in these patients is still incomplete [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. A better understanding of the frequencies of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in such moderate risk patients is fundamental to our appreciation of the importance of these genes as a cause of disease in the general population. Furthermore, <italic>BRCA2</italic> and, to a lesser extent, <italic>BRCA1</italic> also appear to be responsible for an important, but still debated proportion of male breast cancers [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>].</p><p>We analyzed the entire <italic>BRCA1</italic> and <italic>BRCA2</italic> coding regions in 136 unrelated probands: 117 female breast/ ovarian cancer patients and 19 male breast cancer patients. This sample, selected from patients attending breast cancer clinics in Rome and Naples, was primarily drawn from moderate-risk families originating from central and southern Italy, a geographic region that is known to be ethnically heterogeneous [<xref ref-type="bibr" rid="B15">15</xref>]. We used a screening strategy based on a combination of PTT and SSCP analyses, techniques that are mutually complementary in sensitivity and that may identify more than 80% of mutations [<xref ref-type="bibr" rid="B16">16</xref>].</p></sec><sec sec-type="methods"><title>Patients and methods</title><sec><title>Patients</title><p>A total of 136 probands, subclassified as listed below according to selection criteria, were consecutively recruited depending on informed consent from among breast/ovarian cancer patients attending the breast cancer clinics participating in the study in Rome and Naples. The patients originated from the regions of Latium and Abruzzo (central Italy) and Campania and Molise (southern Italy). They included 55 patients with female breast cancer, any age, with breast cancer in one or two first-/second-degree relative(s); 28 patients with female breast cancer diagnosed before age 40 years, who reported no family history of breast/ovarian cancer; 10 patients with female breast/ovarian cancer, any age, with a family history of ovarian/breast-ovarian cancer in at least one first-/second-degree relative; 19 patients with male breast cancer, selected regardless of age and family history; five patients with ovarian cancer, any age, who reported no familial history of breast/ovarian cancer; 10 patients with bilateral breast cancer, selected regardless of age, who reported no family history of breast/ovarian cancer; and nine patients with breast cancer associated with gastrointestinal, pancreatic or uterine cancers, synchronous/metachronous, or in first-degree relative(s), selected regardless of age. Analysis of genomic DNA, RNA and cDNA preparations from peripheral blood lymphocytes were performed following standard procedures. The research protocol was approved by the ethical committee of the University 'Gabriele D'Annunzio'.</p></sec><sec><title>BRCA1 and BRCA2 mutational analysis</title><p>All patients were analyzed for constitutional mutations throughout the entire <italic>BRCA1</italic> and <italic>BRCA2</italic> coding regions. <italic>BRCA1</italic> exon 11 and <italic>BRCA2</italic> exons 10 and 11 were screened by PTT from genomic DNA using the primers listed in Table <xref ref-type="table" rid="T1">1</xref>. Coding <italic>BRCA1</italic> exons 2, 3, 5-10 and 12-24, and <italic>BRCA2</italic> exons 2-9 and 12-27 were screened by SSCP analysis using previously reported primers [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. Primers for <italic>BRCA2</italic> exons 14 and 18 and for a 281 bp <italic>BRCA2</italic> fragment that encompassed the intron 10/exon 11 boundary are reported in Table <xref ref-type="table" rid="T1">1</xref>. In 27 cases, who were analyzed by PTT along the entire <italic>BRCA1</italic> coding sequence as described by Friedman <italic>et al</italic> [<xref ref-type="bibr" rid="B17">17</xref>], <italic>BRCA1</italic> SSCP analysis was limited to exons 2, 5, 20 and 24, to identify missense mutations that are reportedly frequent in these exons and to allow a better investigation of the 5' and 3' ends of the coding sequence. PTT and polymerase chain reaction (PCR)-SSCP were performed as described previously [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. When truncated peptides or variant SSCP conformers were identified, genomic DNA was reamplified and directly sequenced with the PCR product sequencing kit (Sequenase Version 2.0, USB-Amersham, Cleveland, OH, USA). Mutations were verified on two independent blood samples.</p></sec></sec><sec><title>Results</title><p>Deleterious germline <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations were detected in 11 out of 136 cases (8%). Table <xref ref-type="table" rid="T2">2</xref> shows the mutations and data concerning carriers and their families. Table <xref ref-type="table" rid="T3">3</xref> shows correlations between <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations and sex, age at disease diagnosis and familial clustering of breast/ovarian cancer in the total patient population. Table <xref ref-type="table" rid="T4">4</xref> shows the proportions of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in females with site-specific breast and breast-ovarian/ovarian cancer. Table <xref ref-type="table" rid="T5">5</xref> shows the frequency of <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations in males.</p><p>The five deleterious <italic>BRCA1</italic> mutations (Table <xref ref-type="table" rid="T2">2</xref>) included four frameshift mutations (<italic>BRCA1</italic> 1479delAG, <italic>BRCA1</italic> 1623del5bp, <italic>BRCA1</italic> 3880delAG, <italic>BRCA1</italic> 5083del19bp) and one missense mutation (<italic>BRCA1</italic> 300TtoG). These mutations were already reported in the literature [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>] or in the Breast Cancer Information Core electronic database (<ext-link ext-link-type="uri" xlink:href="http://nchgr.nih.gov/Intramural_research/Lab_transfer/Bic/"/>). In addition to the above described deleterious mutations, the neutral <italic>BRCA1</italic> coding variants Glu1038Gly and, in homozygosity, Ser1613Gly were found in two patients [<xref ref-type="bibr" rid="B20">20</xref>]. A novel C to G transversion, affecting the conserved C/T residues of the consensus sequence for the 3' -splice site of <italic>BRCA1</italic> intron 22, was also found in a 60-year-old woman with synchronous breast and gastric cancers, but no family history of cancer. Sequence analysis from cDNA and genomic DNA revealed normal exon 23 and exon 24 transcripts. Allele expression analysis was not feasible, because the patient was homozygous at multiple <italic>BRCA1</italic> polymorphisms.</p><p>Five deleterious <italic>BRCA2</italic> mutations, all localized in exon 11 and including three nonsense and two frameshift mutations, were identified in six patients (Table <xref ref-type="table" rid="T2">2</xref>). <italic>BRCA2</italic> 4109TtoA, <italic>BRCA2</italic> 4339CtoT and <italic>BRCA2</italic> 5117CtoG are novel, whereas <italic>BRCA2</italic> 5950delTC and <italic>BRCA2</italic> 6696delTC were previously reported [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. One mutation (<italic>BRCA2</italic> 6696delTC) recurred in two unrelated probands. The haplotype of the mutation-bearing chromosomes could not be reconstructed, because DNA samples from relatives of the two patients were not available.</p><p><italic>BRCA1</italic> and <italic>BRCA2</italic>, respectively, accounted for four out of 68 (6%) and one out of 68 (1%) patients diagnosed before age 50 years, and for one out of 68 (1%) and five out of 68 (7%) cases diagnosed after age 50 years. <italic>BRCA1</italic> mutations were found in five out of 117 females (4%) and in none of 19 males (0%), and <italic>BRCA2</italic> mutations were found in four out of 117 females (3%) and in two out of 19 males (10%). <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations accounted for the same number of cases among patients with family history of breast/ovarian cancer (four out of 66; ie 4% each), whereas <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations, respectively, accounted for one out of 66 (1%) and for two out of 64 (3%) of the cases without a family history (Table <xref ref-type="table" rid="T3">3</xref>).</p><p>In site-specific female breast cancers, the proportions of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations coincided (four out of 102, ie 4% each). <italic>BRCA1</italic> and <italic>BRCA2</italic> equally contributed to female breast cancers with no familial clustering diagnosed before age 40 years (one out 28; 4% each) and to female breast cancers, all ages, with familial clustering in one or two relatives (three out of 55; 5% each). In the latter subset of cases, <italic>BRCA1</italic> mostly accounted for tumours diagnosed before age 40 years (two out of eight; 25%) and <italic>BRCA2</italic> for tumours diagnosed after age 50 years (three out of 34; 9%). Regardless of family history, the respective contributions of <italic>BRCA1</italic> and <italic>BRCA2</italic> to site-specific female breast cancers diagnosed before age 40 years were 8% (three out of 36) and 3% (one out of 36). One <italic>BRCA1</italic> mutation was detected among the 15 female probands from breast-ovarian/ovarian cancer families (7%). This patient was among the 10 who had a family history of breast/ovarian cancer (one out of 10, 10%; Table <xref ref-type="table" rid="T4">4</xref>).</p><p>Among male breast cancers, <italic>BRCA2</italic> mutations were identified in one out of five (20%) patients with a family history and in one out of 14 (7%) apparently sporadic cases. No <italic>BRCA1</italic> or <italic>BRCA2</italic> mutations were found in female probands with nonfamilial bilateral breast cancer (10 cases) and with breast cancer associated with gastrointestinal, pancreatic or uterine cancers, synchronous/metachronous or in first-degree relative(s) (nine cases). These cases were all diagnosed after age 40 years (Table <xref ref-type="table" rid="T5">5</xref>).</p></sec><sec><title>Discussion</title><p>We screened the coding sequences of the <italic>BRCA1</italic> and <italic>BRCA2</italic> genes in 136 breast/ovarian cancer patients, including 102 females with breast cancer who were mostly at moderate risk for mutation-carrier status, 15 females with breast-ovarian/ovarian cancer and 19 males with breast cancer. The sensitivity of the combined PTT/SSCP screening assays is reportedly high [<xref ref-type="bibr" rid="B16">16</xref>]. The present results should be regarded as an approximation, because the following types of mutation are predicted to escape detection: mutations in noncoding regions, which are estimated to account for a minimum of 10% of pathogenetic <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations [<xref ref-type="bibr" rid="B5">5</xref>]; missense mutations in functionally important regions within <italic>BRCA1</italic> and <italic>BRCA2</italic> exons 11; large deletions undetectable by PCR-based assays; and mutations within the first and last 180 nucleotides of the amplicons analyzed by PTT. In spite of these limitations, the present study contributes evidence that is useful for assessing the importance of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in patients who are not in high-risk families from outbred populations.</p><p>In contrast to studies on North and East European populations [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>], the present results indicate a lack of relevant founder effects for <italic>BRCA1-</italic> and <italic>BRCA2</italic>-related disease in the sample of patients analyzed, which is in agreement with other Italian studies [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>] and with ethnical and historical data [<xref ref-type="bibr" rid="B15">15</xref>]. The <italic>BRCA1</italic> mutations detected in the present study were previously reported in families with high cancer incidence of different ethnic or geographic origin, but not in other Italian surveys [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>]. Interestingly, three of the six <italic>BRCA2</italic> mutations were nonsense and three were novel. <italic>BRCA2</italic> 6696delTC, reported in another series of Italian breast cancer patients [<xref ref-type="bibr" rid="B24">24</xref>] but only twice in the Breast Cancer Information Core database, was the only mutation detected more than once. This mutation may represent a candidate frequent mutation in the Italian population. A possible common origin of the mutation detected in two unrelated patients could not be verified, however, because the haplotypes of the mutation-bearing chromosomes could not be reconstructed.</p><p>The contributions of <italic>BRCA1</italic> and <italic>BRCA2</italic> to breast/ ovarian cancer in Italian patients appear to be less significant than in patients from communities with founder mutations [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>]. The present results are in agreement with recent direct estimates on other outbred populations, which indicate that 7-10% of all female breast cancers that occur before age 40 years are due to <italic>BRCA1</italic>/ <italic>BRCA2</italic> [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>]. Moreover, the 8% <italic>BRCA1</italic> mutation rate found in women with site-specific breast cancer diagnosed before age 40 years, regardless of family history, is consistent with an indirect estimate that 5.3% of all female breast cancers among those who are under 40 years old may be due to mutations in <italic>BRCA1</italic> [<xref ref-type="bibr" rid="B34">34</xref>].</p><p>Overall, <italic>BRCA1</italic> and <italic>BRCA2</italic> equally contributed to site-specific breast cancer in patients who reported one or two breast cancer-affected first-/second-degree relative(s) or who were diagnosed before age 40 years in the absence of a family history. This is consistent with recent data [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B7">7</xref>] that indicated that the respective frequencies of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations are comparable in early onset breast cancer. Considering the total population of patients analyzed here, however, <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations were mostly found in cases with disease diagnosis before and after age 50 years, respectively. Moreover, in cases with familial clustering of site-specific breast cancer, <italic>BRCA1</italic> mostly accounted for tumours diagnosed before age 40 years, and <italic>BRCA2</italic> for tumours diagnosed after age 50 years. This is in agreement with a trend, observed in other populations, for the proportion of cases with <italic>BRCA2</italic> mutations to increase, and the proportion with mutations in <italic>BRCA1</italic> to decrease as the age at cancer onset increases [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B35">35</xref>,<xref ref-type="bibr" rid="B36">36</xref>]. As in other studies [<xref ref-type="bibr" rid="B4">4</xref>], the frequency of <italic>BRCA1</italic>/<italic>BRCA2</italic> mutations taken together was lower than the estimated frequencies at comparable ages for all susceptibility alleles derived from the CASH study [<xref ref-type="bibr" rid="B3">3</xref>]. The discrepancy between direct data derived from <italic>BRCA1IBRCA2</italic> mutational analysis and CASH estimates could be due to several factors, including contribution(s) of gene(s) other than <italic>BRCA1/BRCA2</italic>, differences between populations and relative insensitivity of mutational screening.</p><p>In the present study, <italic>BRCA1</italic> mutations were detected in only one out of 10 cases from breast/ovarian and site-specific ovarian cancer families. This is a low proportion compared with studies that suggested that <italic>BRCA1</italic> and <italic>BRCA2</italic> are responsible for the large majority of breast/ovarian cancer families, with the greater proportion due to <italic>BRCA1</italic> [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B31">31</xref>]. In this respect, the limits of mutation detection techniques and the small number of breast/ovarian and site-specific ovarian cancer cases tested should be taken into account, together with the fact that most of the probands examined here were from families with only one case of ovarian cancer.</p><p>As expected, <italic>BRCA2</italic> mutations were detected in male breast cancer patients. <italic>BRCA2</italic> mutations were found in 20% of the males reporting familial clustering of breast cancer. In the males with no family history of breast/ovarian cancer, the proportion of carriers of <italic>BRCA2</italic> mutations (7%) was comparable to that obtained for <italic>BRCA1</italic> and <italic>BRCA2</italic> combined in site-specific female breast cancer patients (8%). The overall proportion of cancer-affected males with <italic>BRCA2</italic> mutations (10%) was high compared with data from other outbred populations, but was lower than that reported for populations with founder effects [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B37">37</xref>].</p><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Oligonucleotide primers used for PTT and SSCP analyses</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Exon screened</td><td align="left">Primer</td></tr></thead><tbody><tr><td align="left">PTT</td><td></td></tr><tr><td align="left">&#x02003;<italic>BRCA1</italic> exon 11</td><td align="left">Br5'-S: 5'-GCTGCTTGTGAATTTTCTGAG-3'</td></tr><tr><td></td><td align="left">Br5'-A: 5'-GCCTGCAGTGATATTAACTGTCTG-3'</td></tr><tr><td></td><td align="left">Br3'-S: 5'-GAAGAAAGTGAACTTGATG-3'</td></tr><tr><td></td><td align="left">Br3'-A: 5'-TAAGTTTGAATCCATGCTTTG-3'</td></tr><tr><td align="left">&#x02003;<italic>BRCA2</italic> exon 10</td><td align="left">B10-S: 5'-TTTGGAAAAACATCAGGGAATT-3'</td></tr><tr><td></td><td align="left">B10-A: 5'-AAACACAGAAGGAATCGTCATC-3'</td></tr><tr><td align="left">&#x02003;<italic>BRCA2</italic> exon 11</td><td align="left">B11-1S: 5'-CATTCTTCTGTGAAAAGAAGCTG-3'</td></tr><tr><td></td><td align="left">B11-1A: 5'-TGGTTTGAATTAAAATCCTGC-3'</td></tr><tr><td></td><td align="left">B11-2S: 5'-TACATGAACAAATGGGCAGGAC-3'</td></tr><tr><td></td><td align="left">B11-2A: 5'-TCCAGTACCAACTGGGACAC-3'</td></tr><tr><td></td><td align="left">B11-3S: 5'-GATCAGAAACCAGAAGAATTGC-3'</td></tr><tr><td></td><td align="left">B11-3A: 5'-TTGGGATATTAAATGTTCTGGAGTA-3'</td></tr><tr><td></td><td align="left">B11-4S: 5'-TCACCTTGTGATGTTAGTTTG-3'</td></tr><tr><td></td><td align="left">B11-4A: 5'-GTTAGCATACCAAGTCTACTG-3'</td></tr><tr><td align="left">SSCP</td><td></td></tr><tr><td align="left">&#x02003;<italic>BRCA1</italic> exons 2, 3, 5-10, 12-24</td><td align="left">Reported by Friedman <italic>et al</italic> [<xref ref-type="bibr" rid="B17">17</xref>]</td></tr><tr><td align="left">&#x02003;<italic>BRCA2</italic> exons 2-9, 12, 13, 15, 16, 19-27</td><td align="left">Reported by Friedman <italic>et al</italic> [<xref ref-type="bibr" rid="B11">11</xref>]</td></tr><tr><td align="left">&#x02003;<italic>BRCA2</italic> exon 14</td><td align="left">B14-S: 5'-GTGTACTAGTCAATAAAC-3'</td></tr><tr><td></td><td align="left">B14-A: 5'-CATCACACAAATTGTCAT-3'</td></tr><tr><td align="left">&#x02003;<italic>BRCA2</italic> exon 18</td><td align="left">B18-S: 5'-GAATTCTAGAGTCACACTTCCT-3'</td></tr><tr><td></td><td align="left">B18-A: 5'-ACTGATTTTTACCAAGAGTGCA-3'</td></tr></tbody></table><table-wrap-foot><p>Sense primers used for PTT contain a T7 promoter and a eukaryotic translation initiation sequence: 5' -GGATCCTAATACGACTCACTATAGGGA-GACCACCATG-3'. A, antisense; S, sense.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Germline <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations detected in selected samples from 136 unrelated probands and clinicopathologic correlations</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Sex</td><td align="center">Type of cancer</td><td align="center">Family history</td><td align="center">Gene</td><td align="center">Exon</td><td align="center">Codon</td><td align="center">Nucleotide</td><td align="center">Effect</td></tr></thead><tbody><tr><td align="left">Female</td><td align="center">Br (26)</td><td align="center">None</td><td align="center"><italic>BRCA1</italic></td><td align="center">11</td><td align="center">454</td><td align="center">1479</td><td align="center">delAG-ter454</td></tr><tr><td align="left">Female</td><td align="center">Ov (45)</td><td align="center">Mo: Ov (57)</td><td align="center"><italic>BRCA1</italic></td><td align="center">11</td><td align="center">502</td><td align="center">1623</td><td align="center">5bpdel-ter505</td></tr><tr><td align="left">Female</td><td align="center">Br (64)</td><td align="center">D: Br (24)</td><td align="center"><italic>BRCA1</italic></td><td align="center">11</td><td align="center">1254</td><td align="center">3880</td><td align="center">delAG-ter1265</td></tr><tr><td></td><td></td><td align="center">D: Br (41)</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left">Female</td><td align="center">Br (38)</td><td align="center">Fa: Ga (64)</td><td align="center"><italic>BRCA1</italic></td><td align="center">5</td><td align="center">61</td><td align="center">300</td><td align="center">Cys/Gly</td></tr><tr><td></td><td></td><td align="center">Pa: Br (50)</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left">Female</td><td align="center">bil Br (40)</td><td align="center">Mo: Br (49)</td><td align="center"><italic>BRCA1</italic></td><td align="center">16</td><td align="center">1655</td><td align="center">5083</td><td align="center">del19bp-ter1670</td></tr><tr><td align="left">Male</td><td align="center">Br (79)</td><td align="center">S: Br (35)</td><td align="center"><italic>BRCA2</italic></td><td align="center">11</td><td align="center">1293</td><td align="center">4109</td><td align="center">TTA-TAA</td></tr><tr><td></td><td></td><td align="center">S: Br (55)</td><td></td><td></td><td></td><td></td><td align="center">(Lys-Stop)</td></tr><tr><td></td><td></td><td align="center">S: Br (50)</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td></td><td></td><td align="center">S: Col (47)</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left">Female</td><td align="center">Br (55)</td><td align="center">S: Br (49)</td><td align="center"><italic>BRCA2</italic></td><td align="center">11</td><td align="center">1370</td><td align="center">4339</td><td align="center">CAG-TAG</td></tr><tr><td></td><td></td><td align="center">S: Pan (52)</td><td></td><td></td><td></td><td></td><td align="center">(Glu-Stop)</td></tr><tr><td></td><td></td><td align="center">B: Col (54)</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left">Female</td><td align="center">Br (62)</td><td align="center">Mo: Br (59)</td><td align="center"><italic>BRCA2</italic></td><td align="center">11</td><td align="center">1629</td><td align="center">5117</td><td align="center">TCA-TGA</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">(Ser-Stop)</td></tr><tr><td align="left">Male</td><td align="center">Br (54)</td><td align="center">None</td><td align="center"><italic>BRCA2</italic></td><td align="center">11</td><td align="center">1906</td><td align="center">5950</td><td align="center">delCT-ter1909</td></tr><tr><td align="left">Female</td><td align="center">Br (57)</td><td align="center">S: Br (50)</td><td align="center"><italic>BRCA2</italic></td><td align="center">11</td><td align="center">2156</td><td align="center">6696</td><td align="center">delTC-ter2174</td></tr><tr><td align="left">2Female</td><td align="center">Br (34)</td><td align="center">None</td><td align="center"><italic>BRCA2</italic></td><td align="center">11</td><td align="center">2156</td><td align="center">6696</td><td align="center">delTC-ter2174</td></tr></tbody></table><table-wrap-foot><p>Numbers in parentheses indicate age at onset. B, brother; bil Br, bilateral breast cancer; Br, breast cancer; Col, colorectal cancer; D, daughter; Fa, father; Ga, gastric cancer; Mo, mother; Ov, ovarian cancer; Pa, paternal aunt; Pan, pancreatic cancer; S, sister.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T3"><label>Table 3</label><caption><p><italic>BRCA1</italic> and <italic>BRCA2</italic> mutations by sex, age at disease diagnosis and presence of breast/ovarian cancer in first-/second-degree relative(s) in the total population of 136 breast/ovarian cancer probands analyzed</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left"></td><td align="center" colspan="2">BRCA1</td><td align="center" colspan="2">BRCA2</td><td align="center" colspan="2">Total</td></tr><tr><td align="left"></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td></tr><tr><td align="left">Characteristics of patients</td><td align="center">Positive (%)</td><td align="center">Negative</td><td align="center">Positive (%)</td><td align="center">Negative</td><td align="center">Positive (%)</td><td align="center">Negative</td></tr></thead><tbody><tr><td align="left">Age</td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left"> 50 years or under</td><td align="center">4 (6)</td><td align="center">64</td><td align="center">1 (1)</td><td align="center">67</td><td align="center">5 (7)</td><td align="center">63</td></tr><tr><td align="left"> Older than 50 years</td><td align="center">1 (1)</td><td align="center">67</td><td align="center">5 (7)</td><td align="center">63</td><td align="center">6 (9)</td><td align="center">62</td></tr><tr><td align="left">Sex</td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left"> Female</td><td align="center">5 (4)</td><td align="center">112</td><td align="center">4 (3)</td><td align="center">113</td><td align="center">9 (8)</td><td align="center">108</td></tr><tr><td align="left"> Male</td><td align="center">0 (0)</td><td align="center">19</td><td align="center">2 (10)</td><td align="center">17</td><td align="center">2 (10)</td><td align="center">17</td></tr><tr><td align="left">Cancer in relative(s)</td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left"> Yes</td><td align="center">4 (6)</td><td align="center">66</td><td align="center">4 (6)</td><td align="center">66</td><td align="center">8 (11)</td><td align="center">62</td></tr><tr><td align="left"> No</td><td align="center">1 (1)</td><td align="center">65</td><td align="center">2 (3)</td><td align="center">64</td><td align="center">3 (4)</td><td align="center">63</td></tr></tbody></table><table-wrap-foot><p>Positive indicates the presence of deleterious mutations, whereas negative indicates the absence of such mutations.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T4"><label>Table 4</label><caption><p>Proportions of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in 117 female breast/ovarian cancer probands, classified by number of cancer patients in the family, age at disease diagnosis, site-specific breast or breast-ovarian/ovarian cancer</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center" colspan="8">Age of proband at disease diagnosis</td></tr><tr><td></td><td colspan="8"><hr></hr></td></tr><tr><td align="left"></td><td align="center" colspan="2">40 years or less</td><td align="center" colspan="2">41-50 years</td><td align="center" colspan="2">51 years or more</td><td align="center" colspan="2">All ages</td><td></td></tr><tr><td align="left"></td><td align="center" colspan="2"><hr></hr></td><td align="center" colspan="2"><hr></hr></td><td align="center" colspan="2"><hr></hr></td><td align="center" colspan="2"><hr></hr></td><td></td></tr><tr><td align="left">Cancer patients</td><td align="center"><italic>BRCA1</italic></td><td align="center"><italic>BRCA2</italic></td><td align="center"><italic>BRCA1</italic></td><td align="center"><italic>BRCA2</italic></td><td align="center"><italic>BRCA1</italic></td><td align="center"><italic>BRCA2</italic></td><td align="center"><italic>BRCA1</italic></td><td align="center"><italic>BRCA2</italic></td></tr><tr><td align="left">in family^&#x02020;</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td></tr></thead><tbody><tr><td align="left"><bold>Breast</bold></td></tr><tr><td align="left"> One</td><td align="center">4 (1/28)</td><td align="center">4 (1/28)</td><td align="center">0 (0/9)^*</td><td align="center">0 (0/9)^*</td><td align="center">0 (0/10)^*</td><td align="center">0 (0/10)^*</td><td align="center">2 (1/47)^*</td><td align="center">2 (1/47)^*</td></tr><tr><td align="left"> Two or three</td><td align="center">25 (2/8)</td><td align="center">0 (0/8)</td><td align="center">0 (0/13)</td><td align="center">0 (0/13)</td><td align="center">3 (1/34)</td><td align="center">9 (3/34)</td><td align="center">5 (3/55)</td><td align="center">5 (3/55)</td></tr><tr><td align="left"> Total</td><td align="center">8 (3/36)</td><td align="center">3 (1/36)</td><td align="center">0 (0/22)</td><td align="center">0 (0/22)</td><td align="center">2 (1/44)</td><td align="center">7 (3/44)</td><td align="center">4 (4/102)</td><td align="center">4 (4/102)</td></tr><tr><td align="left"><bold>Breast/ovary</bold></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left"> None/one</td><td align="center">0 (0/1)</td><td align="center">0 (0/1)</td><td align="center">0 (0/1)</td><td align="center">0 (0/1)</td><td align="center">0 (0/3)</td><td align="center">0 (0/3)</td><td align="center">0 (0/5)</td><td align="center">0 (0/5)</td></tr><tr><td align="left"> None to two/</td><td align="center">-</td><td align="center">-</td><td align="center">25 (1/4)</td><td align="center">0 (0/4)</td><td align="center">0 (0/6)</td><td align="center">0 (0/6)</td><td align="center">10 (1/10)</td><td align="center">0 (0/10)</td></tr><tr><td align="left"> one or two</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td align="left"> Total</td><td align="center">0 (0/1)</td><td align="center">0 (0/1)</td><td align="center">20 (1/5)</td><td align="center">0 (0/5)</td><td align="center">0 (0/9)</td><td align="center">0 (0/9)</td><td align="center">7 (1/15)</td><td align="center">0 (0/15)</td></tr></tbody></table><table-wrap-foot><p>^&#x02020;Including proband; numbers of <italic>BRCA1</italic>- and <italic>BRCA2</italic>-positive cases over number of cases in each age subset are given in parentheses. ^*The 19 breast cancer patients with no familial clustering of breast/ovarian cancer diagnosed above age 40 years included 10 cases with bilateral breast cancer and nine patients with breast cancer associated with gastrointestinal, pancreatic or endometrial cancer (synchronous/metachronous, four cases; in a first-degree relative, five cases).</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T5"><label>Table 5</label><caption><p>Proportions of <italic>BRCA1</italic> and <italic>BRCA2</italic> mutations in 19 male breast cancer probands classified by number of cancer patients in the family and age at disease diagnosis</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center" colspan="6">Age of proband at disease diagnosis</td></tr><tr><td></td><td colspan="6"><hr></hr></td></tr><tr><td></td><td align="center" colspan="2">41-50 years</td><td align="center" colspan="2">51 years or more</td><td align="center" colspan="2">All ages</td></tr><tr><td><hr></hr></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td></tr><tr><td align="left">Cancer patients</td><td align="center"><italic>BRCA1</italic></td><td align="center"><italic>BRCA2</italic></td><td align="center"><italic>BRCA1</italic></td><td align="center"><italic>BRCA2</italic></td><td align="center"><italic>BRCA1</italic></td><td align="center"><italic>BRCA2</italic></td></tr><tr><td align="left">in family^*</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td><td align="center">[% (<italic>n</italic>)]</td></tr></thead><tbody><tr><td align="left">One</td><td align="center">0 (0/1)</td><td align="center">0 (0/1)</td><td align="center">0 (0/13)</td><td align="center">8 (1/13)</td><td align="center">0 (0/14)</td><td align="center">7 (1/14)</td></tr><tr><td align="left">Two or three</td><td align="center">0 (0/3)</td><td align="center">0 (0/3)</td><td align="center">0 (0/2)</td><td align="center">50 (1/2)</td><td align="center">0 (0/5)</td><td align="center">20 (1/5)</td></tr><tr><td align="left">Total</td><td align="center">0 (0/4)</td><td align="center">0 (0/4)</td><td align="center">0 (0/15)</td><td align="center">13 (2/15)</td><td align="center">0 (0/19)</td><td align="center">10 (2/19)</td></tr></tbody></table><table-wrap-foot><p>^*Including proband; numbers of <italic>BRCA1</italic>- and <italic>BRCA2</italic>-positive cases over number of cases in each age subset are given in brackets.</p></table-wrap-foot></table-wrap></sec>

Characterization of fibroblast growth factor receptor 2 overexpression in the human breast cancer cell line SUM-52PE

<sec><title>Introduction:</title><p>The FGFR family of receptor tyrosine kinases includes four members, all of which are highly alternatively spliced and glycosylated. For FGFR2, alternative splicing of the second half of the third Ig-like domain, involving exons IIIb and IIIc, is a mutually exclusive choice that affects ligand binding specificity and affinity [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. It appears that the second half of the third Ig-like domain can dictate high affinity for FGF-2 or keratinocyte growth factor (KGF), whereas affinity for FGF-1 appears to remain the same [<xref ref-type="bibr" rid="B3">3</xref>]. Alternative splicing of the carboxyl terminus has been shown to involve at least two different exons that can produce at least three different variants. The C1-type and C2-type carboxyl termini are encoded by the same exon, and have two different splice acceptor sites, whereas the C3-type carboxyl terminus is encoded by a separate exon [<xref ref-type="bibr" rid="B4">4</xref>]. The biologic significance of the C1 carboxyl terminus, as compared with the shorter C3 variant found primarily in tumorigenic samples, has been studied in NIH3T3 transfection assays, in which C3 variants were able to produce three times more transformed foci in soft agar than C1 variants (both IIIb), whereas full length FGFR2 and FGFR1 (both IIIc variants) showed no transforming activity [<xref ref-type="bibr" rid="B4">4</xref>].</p><p>Previous studies [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>] have found amplification and overexpression of FGFR2 in 5-10% of primary breast cancer specimens. A recent study [<xref ref-type="bibr" rid="B7">7</xref>] done using a tissue array consisting of 372 primary breast cancer specimens found a 5% incidence of FGFR2 amplification. To our knowledge, none of the HBC cell lines studied thus far have an FGFR2 gene amplification, although overexpression of FGFR2 message and protein has been documented for some breast cancer cell lines [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>].</p><p>SUM-52PE is a breast cancer cell line previously isolated in our laboratory that grows under serum-free and epidermal growth factor-free conditions, has high levels of tyrosine-phosphorylated membrane proteins, and has the capacity to invade and grow under anchorage-independent conditions [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. This cell line exhibits all of the important hallmarks of transformed, highly malignant cells. Therefore, SUM-52PE was used as a model to study the diversity of FGFR2 expression in a breast cancer cell line that has true amplification and overexpression of FGFR2.</p></sec><sec><title>Objectives:</title><p>This study was conducted to examine the degree of FGFR2 amplification and overexpression in the breast cancer cell line SUM-52PE. Subsequent sequencing and characterization of individual FGFR2 variants cloned from the SUM-52PE cell line was completed to determine the complexity of FGFR2 alternative splicing in the context of a highly metastatic breast cancer cell line.</p></sec><sec><title>Methods:</title><p>Southern, Northern and Western blot analyses were done in order to determine the degree of FGFR2 amplification and overexpression in the breast cancer cell line SUM-52PE. Individual FGFR2 variants were cloned out of SUM-52PE using FGFR2-specific primers in a reverse transcription (RT) polymerase chain reaction (PCR). FGFR2 cDNAs were characterized by restriction fragment analysis, sequencing and transient transfection into 293 cells to examine the protein expression of each FGFR2 clone.</p></sec><sec><title>Results:</title><p>The results of the Southern blot showed that there was a 12-fold amplification of FGFR2 in the SUM-52PE cell line. Northern blot analysis of SUM-52PE showed FGFR2 transcripts to be highly overexpressed compared with other breast cancer cell lines and normal HME cells. Several overexpressed bands of approximately 6.3, 5.0, 4.0, and 2.8kb were observed in SUM-52PE cells. The most prominent band, at 2.8kb, was so abundant that it was difficult to discern other individual bands clearly. Western blot analysis showed that both normal HME and HBC cells expressed two FGFR2 variants of 95 and 135kDa. The SUM-52PE cell line greatly overexpressed not only these two polypeptides, as compared with HME and HBC cells, but also overexpressed two unique variants of FGFR2 - 85 and 109kDa polypeptides - as well as several smaller polypeptides in the 46-53kDa range. The antibody used in Western blot analysis only recognizes FGFR2 isoforms that express the C1 carboxyl termini, therefore greatly underestimating the actual number of different FGFR2 variants that are overexpressed in this cell line.</p><p>PCR was performed to determine the proportion of C1/C2 variants as compared with C3 variants in the SUM-52PE cell line. Results of this analysis indicated the presence of all three types of variants in this cell line, although the C1/C2 variants were predominant as compared with the C3 variants in SUM-52PE.</p><p>Four different FGFR2-C1 clones were isolated and sequenced from SUM-52PE cells, which differed in their signal sequence, first Ig-like loop, and acid box. Two FGFR2-C2 clones were isolated from the SUM-52PE cell line, which were identical to each other except for the variable expression of the number of Ig-like domains (two or three). Three C3 clones were isolated and sequenced, two of which have not previously been described in the literature. Clone C3-#3 contained two Ig-like domains, but no acid box. C3-#5 was missing the first two Ig-like domains and the acid box, but did contain the third Ig-like domain.</p></sec><sec><title>Discussion:</title><p>There is an extensive amount of evidence implicating <italic>erbB-2</italic>, a gene that is overexpressed in approximately 30% of breast cancer cases, as a breast cancer gene [<xref ref-type="bibr" rid="B13">13</xref>]. The identification of other breast oncogenes that function in the remaining 70% of cases is an ongoing challenge, as is establishing a causal role for such oncogenes in HME cell transformation.</p><p>FGFR1 and FGFR2, previously established oncogenes, were found to be amplified within large amplicons on 8p11 and 10q26, respectively, in the breast cancer cell line SUM-52PE [<xref ref-type="bibr" rid="B14">14</xref>]. Previous studies have shown that the FGFR2 gene is amplified in about 5-10% of breast cancer cases.</p><p>Our results showed that SUM-52PE cells overexpressed many alternatively spliced isoforms of FGFR2 at both the transcript and protein level as compared with normal HME cells. The variability in FGFR2 isoform expression is complex and involves exon IIIb/c, which encodes the second half of the third Ig-like loop; variations in the carboxyl terminal end of the receptor, involving the C1/C2 or C3 domains; and variable expression of the Ig-like loops and acid box in the extracellular portion of the receptor. The characterization of three unique FGFR2 isoforms that were cloned from SUM-52PE may build on the findings of others concerning the transforming potential of FGFR2 variants [<xref ref-type="bibr" rid="B4">4</xref>]. In particular, because it has been demonstrated that expression of C3-IIIb variants may have more transforming activity than C1-IIIb variants, differences between the three C3 clones we have isolated may provide information regarding the influence of particular structural domains on transforming potential.</p><p>Ongoing studies are aimed at characterizing the transforming ability of FGFR2 isoforms obtained from SUM-52PE cells by transducing these genes into normal HME cells. By overexpressing FGFR2 isoforms in a physiologically relevant system, we hope to determine the isoform(s) that acts in a dominant way in the process of cell transformation, and to determine whether different regions present in individual clones drive specific phenotypes associated with transformation.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Tannheimer</surname><given-names>Stacey L</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Rehemtulla</surname><given-names>Alnawaz</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Ethier</surname><given-names>Stephen P</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Breast Cancer Research : BCR

<sec><title>Introduction</title><p>The FGFs are a family of polypeptides consisting of 18 different growth factors that bind with varying specificity and affinity to four different FGFRs. FGFs stimulate proliferation of a wide variety of cells of mesenchymal, neuronal, and epithelial origins [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>], and have been found [<xref ref-type="bibr" rid="B17">17</xref>] to induce, inhibit, and maintain cell differentiation in different experimental systems. In addition, FGFs have been shown to be involved in neuronal survival [<xref ref-type="bibr" rid="B18">18</xref>], and stimulation of angiogenesis [<xref ref-type="bibr" rid="B19">19</xref>] and embryogenesis [<xref ref-type="bibr" rid="B20">20</xref>].</p><p>The FGFRs are a family of receptors that are characterized by the presence of two or three Ig-like domains and an acid box in the extracellular domain, a transmembrane region, and a split kinase domain in the cytoplasmic domain of the molecule [<xref ref-type="bibr" rid="B17">17</xref>]. Binding of FGF to heparin, or cell-surface heparin sulfate proteoglycans, results in high-affinity binding of this complex to FGFRs [219]. FGFRs subsequently undergo dimerization, followed by transphosphorylation on cytoplasmic tyrosine residues [<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>].</p><p>FGFRs can be alternatively spliced into a variety of isoforms that may have different functions. For example, at the amino terminus, which encodes the ectodomain of the receptor, alternative splicing of exon III results in the synthesis of either the IIIb or IIIc versions of FGFR2, which differ in their ligand specificities. FGFR2-IIIb, also referred to as KGF receptor, has been shown to be a high-affinity receptor for KGF, whereas FGFR2-IIIc receptors have a high affinity for FGF-2. At the carboxyl terminus, FGFR2 can be alternatively spliced to produce three different variants from two different exons [<xref ref-type="bibr" rid="B4">4</xref>]. C1/C2 variants are produced from the same exon with two different splice acceptor sites. C3 variants are produced from a separate exon with a different 3' - noncoding region from that of C1/C2 variants, resulting in a shorter form of FGFR2, which has been found [<xref ref-type="bibr" rid="B4">4</xref>] to be the predominant variant in stomach cancer cell lines as compared with normal stomach tissue.</p><p>Overexpression and amplification of growth factor receptors are common alterations observed in HBC cells. The FGFR1 and FGFR2 genes are expressed in both normal and breast cancer tissues [<xref ref-type="bibr" rid="B24">24</xref>], and alterations, including amplification and overexpression of FGFR1 and FGFR2, have previously been reported in 5-10% of primary breast cancer specimens [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. A recent study [<xref ref-type="bibr" rid="B7">7</xref>] that was done using a tissue array consisting of 372 primary breast specimens found a 5% incidence of FGFR2 amplification.</p><p>There are at least four studies [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B24">24</xref>] that have analyzed the presence of FGFR2 message in breast cancer cell lines. These studies show great discrepancy as to the presence or absence of FGFR2 message and level of expression, possibly due to the method of detection (RT-PCR, Northern blot, or ribonuclease protection assay). To our knowledge, none of the HBC cell lines studied thus far have <italic>FGFR2</italic> gene amplification, although overexpression of FGFR2 message and protein has been documented for some breast cancer cell lines [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>].</p><p>SUM-52PE is a breast cancer cell line previously isolated in this laboratory that grows under serum-free and epidermal growth factor-free conditions, has high levels of tyrosine phosphorylated membrane proteins, and has the capacity to invade basement membranes and grow under anchorage-independent conditions [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. Thus, SUM-52PE cells, which have amplifications of both FGFR1 and FGFR2 genes, exhibit all of the important hallmarks of highly malignant cells. In order to understand better the important molecular alterations that lead to this transformed phenotype, we examined the contribution of FGFR2 overexpression to these characteristics. Isolation and characterization of individual FGFR2 isoforms is an important first step in identifying the contribution of this receptor to cell transformation. Therefore, SUM-52PE was used as a model to study the diversity of FGFR2 expression in a breast cancer cell line that has true amplification and overexpression of the FGFR2 gene.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Southern blot analysis</title><p>Genomic DNA (10 &#x003bc; g) was digested with Hind III, and restriction fragments were separated on an agarose gel and transferred by standard methods to a nylon membrane. The Southern blot was probed with either FGFR1 or FGFR2, generated by RT-PCR of MCF-10A cell RNA. The specific primers used in the RT-PCR reaction were as follows: 5' -CAAACCCAACCGTGTGACCAAAGTG-3' and 5' -CGTGCGAGGCCAAAGTCTGCTATCT-3' for FGFR1; and 5' -GGTCGTTTCATCTGCCTGGT-3' and 5' -CCTTCCCGTTTTTCAGCCAC-3' for FGFR2. The RT-PCR reaction was carried out as described below.</p></sec><sec><title>Northern blot analysis</title><p>RNA (10 &#x003bc; g) from a normal HME cell line (MCF-10A) and a panel of HBC cell lines (SUM-52, -149, -185 and -206) was run on a 1% agarose gel containing formaldehyde, and then transferred to a Nytran membrane (Schleicher &#x00026; Schuell, Keene, NH, USA). The membrane was probed with full-length FGFR2 C1-#1 isolated from the SUM-52PE HBC cell line. Probe (50ng) was labeled using a random priming kit (Gibco BRL, Grand Island, NY, USA) with 5 &#x003bc; Cu &#x003b1; -P³²-dCTP at 25&#x000b0;C for 2h. Unlabeled &#x003b1; -P³²-dCTP was removed by centrifugation of the probe through a Quik-Spin (Roche Molecular Biochemicals, Indianapolis, IN, USA) column at 1100 revolutions/min for 4min. Labeled probe (20&#x000d7;10⁶ cpm) was boiled at 100&#x000b0;C for 5min and then used to probe the Northern blot overnight at 42&#x000b0;C. Prehybridization of the membrane occurred for 2.5h at 42&#x000b0;C.</p></sec><sec><title>Fibroblast growth factor receptor-2 Western blots</title><p>Cells were rinsed twice with ice-cold Hanks' balanced salt solution (Gibco BRL) and then lysed on ice with a buffer consisting of 50mmol/l Tris-HCl (pH8.5); 150mmol/l NaCl; 1% Nonidet P-40 (ICN Biomedical, Inc, Aurora, OH, USA); 5mmol/l ethylene diamine tetra-acetic acid supplemented with 5mmol/l sodium orthovanadate; 50 &#x003bc; g/ml phenylmethysulfonyl fluoride; 20 &#x003bc; g/ml aprotinin; and 10 &#x003bc; g/ml leupeptin. Lysates were spun at 20 800 <italic>g</italic> at 4&#x000b0;C for 10min and then analyzed for protein using a modified Lowry's method. Whole-cell lysates were resolved on 7.5% polyacrylamide gels, transferred to PVDF membrane (Millipore Corporation, Bedford, MA, USA), and probed with an anti-FGFR2 antibody (Santa Cruz Biotechnology, Inc, Santa Cruz, CA, USA), an anti-FGFR2 antibody pre-conjugated with a competitive peptide (Santa Cruz Biotechnology, Inc) at room temperature for 2h, or an anti-Flag antibody (M2 antibody; Sigma, St Louis, MO, USA).</p></sec><sec><title>Reverse transcription polymerase chain reaction analysis of exon IIIb/c expression</title><p>FGFR2 variants were amplified using exon III specific primers: 5' -CCCGGGTCTAGATTTATAGTGATGCCCAGCCC-3' for FGF-FB; and 5' -CCCGGGGAATTCACCACCATGCAGGCGATTAA-3' for FGF-RB [<xref ref-type="bibr" rid="B25">25</xref>]. RT-PCR amplification was carried out by use of the SuperScript One Step RT-PCR system (Gibco BRL) with the addition of 1 &#x003bc; l Expand High Fidelity enzyme (Roche Molecular Biochemicals). RT was carried out at 50&#x000b0;C for 30min, followed by 94&#x000b0;C for 3min. The PCR reaction was for 25 cycles of 94&#x000b0;C for 30s, 65&#x000b0;C for 30s, and 72&#x000b0;C for 1min, and final extension at 72&#x000b0;C for 7min. Singly, Ava I and Hinc II digestion of the RT-PCR product was performed overnight at 37&#x000b0;C, and then run on 3% NuSieve 3:1 agarose gel (FMC Bioproducts, Rockland, ME, USA).</p></sec><sec><title>Isolation of fibroblast growth factor receptor 2 cDNA clones</title><p>RNA was isolated from the SUM-52PE cell line using the Trizol isolation technique (Gibco BRL). RT-PCR amplification of FGFR2 was carried out by use of the SuperScript One Step RT-PCR system (Gibco BRL) with the addition of 1 &#x003bc; l Expand High Fidelity enzyme (Roche Molecular Biochemicals). RT was carried out at 50&#x000b0;C for 30min, followed by 94&#x000b0;C for 3min. Amplification of FGFR2 was carried out using gene-specific primers: 5' -ATGCCCGTAGAGGAAGTGTG-3' for FGFR2 upstream; 5' -AACGCACGTCCACCTTGAGTCCT-3' for C1/C2-specific downstream; and 5' -CTATTACTTGTCATCGTCGTCCTT-GTAGTCGATCTCATTGGTTGTGAG-3' for C3-specific downstream. This was done for 40 cycles of denaturation at 94&#x000b0;C for 15s, annealing at 55&#x000b0;C for 30s, and elongation at 72&#x000b0;C for 3min, and final extension at 72&#x000b0;C for 7min. FGFR2 cDNA was then digested with Sal I/Xba I and ligated into the pZ vector at 14&#x000b0;C overnight using T4 DNA ligase. The ligation reaction was transformed into SURE <italic>Escherichia coli</italic> cells (Stratagene, La Jolla, CA, USA) as recommended.</p></sec><sec><title>Transfection of 293 cells</title><p>One-hundred-millimeter dishes of 293 cells at 50% sub-confluence were transfected with 20 &#x003bc; g plasmid containing individual cDNA clones isolated from the SUM-52PE cell line using the calcium phosphate method. Briefly, 293 cells were incubated in 10% Dulbecco's modified eagle medium with 10mmol/l chloroquine for 5-10min. A reaction mix of 20 &#x003bc; g cDNA, 1&#x000d7;Hepes-buffered saline, and CaCl₂ (JT Baker, Phillipsburg, NJ, USA) was put into 293 dishes. Media was changed 6-8h after transfection and whole-cell lysates were prepared 48h after transfection, as described above. Whole-cell lysates (100 &#x003bc; g) were resolved on 7.5% polyacrylamide gels, transferred to a polyvinylidene fluoride membrane, and probed with the anti-FGFR2 antibody (Santa Cruz Biotechnology, Inc) or an anti-Flag antibody (M2 antibody, Sigma).</p></sec><sec><title>Sequencing of fibroblast growth factor receptor 2 variants</title><p>Plasmid cDNA for individually isolated FGFR2 isoforms was prepared and submitted to the University of Michigan DNA Sequencing core with a series of FGFR2-specific primers, as well as vector-specific primers. The sequencing of all FGFR2 variants was performed in both a 5' and a 3' direction, with the exclusion of the transmembrane domain, which was shown to be highly conserved between isoforms.</p></sec></sec><sec><title>Results</title><sec><title>Amplification of fibroblast growth factor receptor 2 in SUM-52PE cells</title><p>In previous experiments, we found by comparative genome hybridization analysis that SUM-52PE cells have large amplifications in the genomic regions of chromosomes 8 and 10, containing the FGFR1 and FGFR2 genes, respectively [<xref ref-type="bibr" rid="B14">14</xref>]. In order to characterize the amplification of these candidate breast cancer oncogenes, Southern blot analysis was performed on the HBC cell line SUM-52PE and other breast cancer cell lines [<xref ref-type="bibr" rid="B14">14</xref>]. The results of the Southern blot showed fivefold amplification of the FGFR1 gene and a 12-fold amplification of the FGFR2 gene in SUM-52PE cells, as compared with other breast cancer cell lines that do not have genomic amplifications in these regions (Fig. <xref ref-type="fig" rid="F1">1</xref>).</p></sec><sec><title>Overexpression of fibroblast growth factor receptor 2 transcript in SUM-52PE cells</title><p>Because gene amplification often involves large genomic regions that contain many genes, Northern blot analysis was performed to determine whether the observed amplifications of FGFR1 and FGFR2 correlated with transcript overexpression. Northern blot analysis of SUM-52PE showed FGFR2 transcripts to be highly overexpressed compared with other breast cancer cell lines (Fig. <xref ref-type="fig" rid="F2">2a</xref>, lanes 2-5) and normal HME cells (Fig. <xref ref-type="fig" rid="F2">2a</xref>, lane 1). Several overexpressed bands of approximately 6.3, 5.0, 4.0, and 2.8kb were observed in SUM-52PE cells. The most prominent band, at 2.8kb, was so abundant that it was difficult to discern other individual bands clearly. FGFR1 transcript overexpression, on the other hand, was not detected in SUM-52PE (data not shown).</p></sec><sec><title>Overexpression of fibroblast growth factor receptor 2 variants in SUM-52PE cells</title><p>FGFR2 has been reported to be alternatively spliced, resulting in translation of multiple FGFR2 isoforms. Because Northern blot experiments demonstrated a number of different FGFR2 isoforms, Western blot analysis was performed to examine the number and level of expression of FGFR2 protein(s). The results showed that both normal HME and HBC cells (SUM-44PE, SUM-52PE, and others not shown) expressed two isoforms of FGFR2 of 135 and 95kDa (Fig. <xref ref-type="fig" rid="F2">2b</xref>, left panel, lanes 1-3, polypeptides labeled common). Interestingly, as compared with HME and other HBC cells, the SUM-52PE cell line greatly overexpressed not only these two polypeptides, but also overexpressed two unique isoforms of FGFR2 - 85 and 109kDa polypeptides - as well as several smaller polypeptides in the 46-53kDa range (Fig. <xref ref-type="fig" rid="F2">2b</xref>, left panel, lane 3, labeled unique). Use of a competitive peptide preconjugated with the anti-FGFR2 antibody confirmed that these bands were specific FGFR2 polypeptides (Fig. <xref ref-type="fig" rid="F2">2b</xref>, right panel). Use of an antibody isotype control also confirmed the same nonspecific bands as those that were identified by use of the competitive peptide (data not shown).</p><p>It is important to note that the polyclonal anti-FGFR2 antibody used in these experiments was created against a peptide derived from the C1 carboxyl terminus of FGFR2. Thus, this antibody only recognizes FGFR2 isoforms that express this carboxyl terminus. Because antibodies that recognize FGFR2 isoforms that express either the C2 or C3 carboxyl termini are not currently available, the results of the Western blot analysis of SUM-52PE cells under-represents the actual number of different FGFR2 proteins expressed in this cell line.</p></sec><sec><title>Preliminary characterization of fibroblast growth factor receptor 2 expression in SUM-52PE cells</title><p>To begin to characterize the FGFR2 transcripts that were present in SUM-52PE cells, PCR-based assays were performed to estimate the relative proportions of transcripts containing exon IIIb versus IIIc, and the proportion of C1/C2 variants as compared with C3 variants.</p><p>RT-PCR analysis using exon III-specific primers was performed on SUM-52PE mRNA, followed by restriction fragment analysis to determine the presence of exon IIIb/c-expressing variants [<xref ref-type="bibr" rid="B25">25</xref>]. Exon IIIb contains one unique Ava I site, whereas exon IIIc contains two Hinc III sites. Therefore, the proportion of Ava I digest fragments to Hinc III digest fragments allows for the determination of the proportion of IIIb to IIIc variants present. Using this method, SUM-52PE cells were found to express only IIIb isoforms, because the PCR product obtained was completely digested by Ava I, whereas Hinc III failed to cut any of the amplified product (Fig. <xref ref-type="fig" rid="F3">3a</xref>).</p><p>C1/C2 variants were amplified using a primer specific to the 3' -noncoding region, whereas C3 variants were amplified using a primer specific for the C3 3' -noncoding region. Results of this analysis indicated the presence of all three types of variants in this cell line, although the C1/C2 variants (Fig. <xref ref-type="fig" rid="F3">3b</xref>; lanes 1 and 2) appeared to be more abundant than the C3 variants in SUM-52PE (Fig. <xref ref-type="fig" rid="F3">3b</xref>; lanes 3 and 4).</p></sec><sec><title>Isolation and analysis of alternatively spliced fibroblast growth factor receptor 2 variants</title><p>The PCR-based approach just described suggested that SUM-52PE cells express exclusively IIIb type receptors, which can contain any of the three carboxyl termini. The data also suggested that C1 and C2 variants are more common than receptors with the C3 terminus. In order to characterize definitively the range of FGFR2 variants expressed by SUM-52PE cells, individual isoforms were cloned and sequenced from SUM-52PE RNA. To isolate specific FGFR2 cDNAs, C1/C2- or C3-specific primers were used in combination with a primer for the 5' end of the gene. RT-PCR amplified FGFR2 products were purified and ligated into the bicistronic vector pZ. Individual clones containing insert were characterized by restriction digest analysis, sequencing, and transient transfection into 293 cells.</p><p>FGFR2 cDNA clones isolated from RNA derived from SUM-52PE cells were highly variable and differed with respect to number of Ig-like loops expressed, the presence or absence of the acid box, and the expression of C1, C2, or C3 carboxyl termini. As predicted from the PCR experiments, all FGFR2 variants isolated from SUM-52PE expressed the IIIb exon.</p><p>Four different FGFR2-C1 clones were isolated and sequenced (Fig. <xref ref-type="fig" rid="F4">4</xref>). The largest clone, C1-#38, was a full-length FGFR2-IIIb isoform, which was previously identified in the literature as KGF receptor [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B26">26</xref>]. C1-#38 contained three Ig-like domains, the characteristic acid box, and the C1 exon in the carboxyl terminus. The other three C1 isoforms contained two Ig-like domains, as well as exon IIIb, but differed with respect to the rest of their extracellular domains. C1-#8 contained two Ig-like domains and the acid box in the extracellular region. C1-#1 expressed two Ig-like domains, but was lacking part of the extracellular region corresponding to the acid box. This type of deletion has previously been identified as an FGFR2-IIIc variant, Bek103 [<xref ref-type="bibr" rid="B27">27</xref>]. C1-#10 was a unique variant that expressed two Ig-domains and the acid box, but contained a unique deletion of the 5' region of the receptor corresponding to the signal sequence. The functional significance of this type of deletion has yet to be determined.</p><p>Two FGFR2-C2 clones were isolated from the SUM-52PE cell line (Fig. <xref ref-type="fig" rid="F4">4</xref>). Clone C2-#19 contained three Ig-like domains, the acid box, and the C2 exon in the carboxyl termini. Clone C2-#5 contained two Ig-like domains, but was missing part of the extracellular domain, which is similar to, but smaller than the deletion characterized for clone C1-#1. A clone identical to C2-#5 has previously been identified in the literature as human K-sam C2, a variant isolated from the KATO-III human stomach cancer-derived cell line [<xref ref-type="bibr" rid="B4">4</xref>]. Both K-sam and clone C2-#5 are missing the same 89 amino acid sequence in the 5' region of FGFR2 corresponding to the first Ig-like domain region.</p><p>Three C3 clones were isolated and sequenced, two of which have not previously been described in the literature (Fig. <xref ref-type="fig" rid="F4">4</xref>). C3-#4, a previously described FGFR2 variant, contained three Ig-like domains and the acid box, and was considered a full-length C3 clone [<xref ref-type="bibr" rid="B4">4</xref>]. Clone C3-#3 contained two Ig-like domains, but the lack of the acid box makes this a unique C3 variant. C3-#5 was missing the first two Ig-like domains and the acid box, but did contain the third Ig-like domain. This FGFR2 isoform also has not previously been reported.</p><p>Because FGFR2 transcripts from HME cells were not detectable by Northern blot, and because HME cells express FGFR2 protein that is detectable by Western blot, experiments were performed to isolate FGFR2 clones from RNA derived from normal HME cells. After two cycles of RT-PCR and subsequent cloning into pZ, three different FGFR2 isoforms were obtained. As with the SUM-52PE cells, all three isoforms cloned from HME cells contained exon IIIb. Two clones with C1 carboxyl termini were isolated, one of which was full length (the same as clone C1-#38 from SUM-52PE) and the second was missing the first Ig loop (the same as clone C1-#8). One clone containing the C2 terminus was isolated that was otherwise a full-length isoform (the same as clone C2-#19).</p></sec><sec><title>Fibroblast growth factor receptor 2 protein expression in transfected 293 whole cell lysates</title><p>To assess the ability of the various FGFR2 isoforms isolated from SUM-52PE cell RNA to synthesize protein, and to determine which FGFR2 clones direct the synthesis of protein isoforms detected in Western blots, 293 cells were transiently transfected with individual FGFR2 iso-forms using the bicistronic pZ expression vector and analyzed by Western blot. As can be seen in the left panel of Figure <xref ref-type="fig" rid="F5">5a</xref>, all four C1-containing isoforms expressed protein in 293 cells and each clone gave rise to an FGFR2 isoform with distinct molecular size. Similarly, all three C3-containing clones expressed protein in 293 cells (Fig. <xref ref-type="fig" rid="F5">5b</xref>). C3-containing isoforms were visualized using a flag antibody that detects an epitope tag incorporated into the design of the C3-specific primer. As can be seen from the right panel of Figure <xref ref-type="fig" rid="F5">5b</xref>, the molecular sizes of the C3 variants correspond to what would be predicted on the basis of the size of the individual clones. C3-#4 is full length at the amino terminus, C3-#3 is missing the first Ig loop and the acid box, and C3-#5 is missing the first Ig loop, the acid box and the second Ig loop.</p><p>Unfortunately, we cannot assess protein synthesis from any of the C2-containing clones at this time. Because C1 and C2 are derived from the same exon, and because we used a single primer to clone both C1- and C2-containing variants, which did not contain an epitope tag, we are currently unable to detect FGFR2-C2 proteins.</p><p>As shown in Figure <xref ref-type="fig" rid="F2">2b</xref>, two FGFR2-C1 proteins (of 95 and 135kDa) are commonly detected in SUM-52PE cells, HME cells, and other breast cancer cell lines. This is in accord with the observation that both C1 variants cloned from HME cells were also cloned from SUM-52PE cells. Thus, the data suggest that the common band that migrates at 135kDa represents the glycosylated version of the full-length FGFR2-C1 (C1-#38: Fig. <xref ref-type="fig" rid="F5">5a</xref>; lanes 1 and 4). Similarly, the second common band that migrates at 95kDa is likely derived from clone C1-#8, which is missing the first Ig loop (Fig. <xref ref-type="fig" rid="F5">5a</xref>; lanes 1 and 3). Clone C1-#1 was highly expressed in 293 cells and resulted in the detection of multiple immunoreactive proteins with a range of sizes (Fig. <xref ref-type="fig" rid="F5">5a</xref>; lane 2). It is possible that this isoform, which is missing both the first Ig loop and the acid box, is responsible for some of the high- and low-molecular-weight bands detected in SUM-52PE Western blots, although more work will be required to definitively demonstrate this. The contribution of clone C1-#10, which was only weakly expressed in 293 cells, to FGFR2 protein expression in SUM-52PE cells cannot be determined at this time (Fig. <xref ref-type="fig" rid="F5">5a</xref>; lane 5).</p><p>As described above, clone C1-#38 encodes the full-length FGFR2-IIIb/C1 form of the receptor and is expressed at the message level in normal cells and all breast cancer cells examined, including SUM-52PE. However, SUM-52PE cells also have an overexpressed band at 109kDa, which was also detected in 293 cells transfected with clone C1-#38 (Fig. <xref ref-type="fig" rid="F5">5a</xref>; lane 4) and in HME cells transduced with this clone (not shown). This band was not detected in HME cells or any other breast cancer cells (Fig. <xref ref-type="fig" rid="F2">2b</xref>; left panel). The molecular size of this band corresponds to the size of the mature, non-processed, full-length protein, suggesting that the 135kDa band corresponds to the fully glycosylated form of the protein. The ability to detect this 109kDa non-processed form of full-length FGFR2-IIIb/C1 only in SUM-52PE cells and cells transduced with clone C1-#38 suggests that the protein is rapidly glycosylated in cells that express physiologic levels of FGFR2. By contrast, in cells that overexpress FGFR2, the 109-kDa nonglycosylated protein accumulates in the cell and is detectable by Western blot as one of the unique bands shown in Figure <xref ref-type="fig" rid="F2">2b</xref>. Thus, the detection of unique bands in Western blots derived from SUM-52PE cells reflects not only differences in isoform synthesis at the message level, but also differences in protein processing that occurs when the protein is highly overexpressed. At present, the functional consequences of the accumulation of nonglycosylated FGFR2 are not known.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Amplification of FGFR1 and FGFR2 in SUM-52PE cells. <bold>(a)</bold> Genomic DNA from four breast cancer cells lines (SKBR3, SUM-52PE, SUM-44PE, and T47D) were compared by Southern blot for FGFR1 expression. <bold>(b)</bold> Genomic DNA from three breast cancer cell lines (SKBR3, SUM-52PE, and SUM-44PE) were compared by Southern blot for FGFR2 expression.</p></caption><graphic xlink:href="bcr-2-4-311-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>Analysis of FGFR2 expression in HBC and HME cells. <bold>(a)</bold> A Northern blot probed with full-length FGFR2 is shown. A glyceraldehyde 3' -phosphate dehydrogenase (GAPDH) probe was used to normalize RNA levels. Cell lines shown are as follows: normal HME cell line (MCF-10A, lane 1) and HBC cell lines (lanes 2-5). Size of visualized bands are marked according to kilobase. <bold>(b)</bold> Whole-cell lysates were immunoblotted with an anti-FGFR2 antibody in the absence (left panel) or presence (right panel) of a competitive peptide. Breast cancer cell lines are in lanes 2 and 3, and a normal HME sample (MCF-10A) is in lane 1. All cell lines expressed two 'common' FGFR2 variants (of 95 and 135 kDa), whereas SUM-52PE expressed at least three 'unique' variants (of 46-53, 85, and 109kDa), as determined by comparison with nonspecific bands (NS; right panel).</p></caption><graphic xlink:href="bcr-2-4-311-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p>Preliminary characterization of FGFR2 expression in SUM-52PE cells. <bold>(a)</bold> Exon III-specific primers were used in RT-PCR of SUM-52PE RNA. RT-PCR product was then digested with Ava I or Hinc II at 37&#x000b0;C overnight and then resolved on a 3% NuSieve gel. Exon IIIb contains one unique Ava I site, whereas exon IIIc contains two Hinc II sites, and therefore the proportion of Ava I digest fragments to Hinc II digest fragments determines the proportion of IIIb to IIIc variants present. The presence of 269 and 188 bp fragments generated by Ava I digestion (lane 2) and lack of Hinc II digested products (lane 3) confirms the exclusive presence of exon IIIb in FGFR2 variants in the SUM-52PE cell line. <bold>(b)</bold> SUM-52PE mRNA was reverse transcribed using an oligo dT primer, and then amplified using a 5' -FGFR2-specific primer and a 3' -specific primer for C1/C2 or C3. Equimolar amounts of primer were used in the PCR reaction, and then 2 or 5 &#x003bc; l of PCR product were compared on a 0.8% agarose gel. Lane 1, 2 &#x003bc; l C1/C2 product; lane 2,5 &#x003bc; l C1/C2 product; lane 3, 2 &#x003bc; l C3 product; lane 4, 5 &#x003bc; l C3 product.</p></caption><graphic xlink:href="bcr-2-4-311-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>Isolated and sequenced FGFR2 isoforms from SUM-52PE. Isolated FGFR2 variants had variable expression of two to three Ig-like domains, as well as the characteristic acid box in the extracellular portion of the molecule. Variability between isolated clones also existed in the intracellular portion of the molecule, where alternative splicing of exons C1/C2 or C3 created either a full-length carboxyl termini (C1), or truncated versions (C2 or C3).</p></caption><graphic xlink:href="bcr-2-4-311-4"/></fig><fig position="float" id="F5"><label>Figure 5</label><caption><p>Transient expression of FGFR2 clones isolated from SUM-52PE. <bold>(a)</bold> Whole-cell lysates from 293 cells that had been transiently transfected with individual FGFR2 clones were immunoblotted with an anti-FGFR2 antibody (lanes 2-5). Whole-cell lysate from the SUM-52PE breast cancer cell line was loaded in lane 1 as a positive control, whereas whole-cell lysates from a mock transfection of 293 cells was loaded in lane six as a negative control. <bold>(b)</bold> Whole-cell lysates from 293 cells that were transiently transfected with individual FGFR2-C3 clones containing a Flag sequence were immunoblotted with an anti-Flag antibody (lanes 1, 2, and 4). Whole-cell lysates from a mock transfection of 293 cells were loaded in lane 3 as a negative control.</p></caption><graphic xlink:href="bcr-2-4-311-5"/></fig></sec></sec><sec><title>Discussion</title><p>The progression of cells from the normal to neoplastic state is a multistep process that involves alterations in multiple signaling pathways. Both epidermal growth factor receptor and erbB-2 have been identified as signaling molecules that play a dominant role in breast cell transformation [<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>]. There is an extensive amount of evidence for <italic>erbB-2</italic> as a breast cancer gene that is overexpressed in approximately 30% of breast cancer cases [<xref ref-type="bibr" rid="B13">13</xref>]. The identification of other breast oncogenes that function in the remaining 70% of cases is an ongoing challenge, as is establishing a causal role for such oncogenes in HME cell transformation.</p><p>Large regions of gene amplification in cancer cells can be detected by techniques such as comparative genomic hybridization and high-density arrays, which helps to localize areas that may contain functional oncogenes. FGFR1 and FGFR2, which were previously established as candidate breast cancer oncogenes, were found to be amplified within large amplicons on 8p11 and 10q26, respectively, in the breast cancer cell line SUM-52PE [<xref ref-type="bibr" rid="B14">14</xref>]. Previous studies [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>] have shown that the FGFR2 gene is amplified in about 5-10% of cases. Because genes can be amplified without being overexpressed [<xref ref-type="bibr" rid="B14">14</xref>], we chose to examine whether FGFR2 may be an important oncogene in this breast cancer cell line by examining its expression at the mRNA and protein levels. Our results showed that SUM-52PE cells overexpressed many alternatively spliced forms of FGFR2 at both the transcript and protein level, as compared with normal mammary epithelial cells. By contrast, FGFR1 is not expressed in SUM-52 cells.</p><p>In contrast to SUM-52PE cells, FGFR2 expression at the message level is very low in HME cells. Indeed, even prolonged exposure of Northern blots to film did not allow the visualization of FGFR2 message in normal cells. However, Western blots did indicate the presence of FGFR2 protein in HME cells. To resolve this apparent paradox, two rounds of RT-PCR were performed using HME cell-derived RNA, which resulted in the isolation of three alternatively spliced forms of FGFR2 message, each of which expressed the IIIb exon. The predicted protein products of these clones correspond to that which was observed in Western blots.</p><p>The variability in FGFR2 isoform expression is complex and involves exon IIIb/c, which encodes the second half of the third Ig-like loop, variations in the carboxyl terminal end of the receptor involving the C1/C2 or C3 domains, and variable expression of the Ig-like loops and acid box in the extracellular portion of the receptor.</p><p>Alternative splicing of the FGFR2 mRNA that encodes the carboxyl terminus has been shown to involve at least two different exons, which can produce at least three different variants. The C1- and C2-type carboxyl termini are encoded by the same exon, having two different splice acceptor sites, whereas the C3-type carboxyl terminus is encoded by a separate exon [<xref ref-type="bibr" rid="B4">4</xref>]. The biologic significance of the full-length carboxyl terminus (C1), as compared with the truncated variant found primarily in tumorigenic samples (C3), has been studied in NIH3T3 transfection assays. The IIIb variants KGF receptor (C1) and K-sam C3 were both able to produce transformed foci, growth in soft agar and tumorigenicity in nude mice as compared with full-length IIIc variants of FGFR2 and FGFR1, which were not transforming [<xref ref-type="bibr" rid="B4">4</xref>]. The question of whether C3 variants are more transforming than C1 variants remains to be determined conclusively, because the number of transformed foci obtained using K-sam C3 was only threefold greater than that obtained using KGF receptor (C1) variants. A significant difference between the C3 and C1 termini is that the former does not contain the binding site for phospholipase C&#x003b3;. Thus, the ability of the variants of FGFR2 containing the C3 terminus to transform 3T3 cells suggests that signaling through this pathway is not necessary for FGFR2 IIIb to act as an oncogene.</p><p>The panel of FGFR2 isoforms isolated from SUM-52PE includes several unique and previously unreported isoforms. The first of these unique variants, C1-#10, contains a large deletion of the 5' region that includes the first Ig-like domain as well as the signal sequence, which could have interesting cellular localization and cell signaling properties due to the absence of part of the signal sequence. Clones C3-#3 and C3-#5 have not previously been reported and are missing the first Ig-like domain and acid box. C3-#5 is also missing the second Ig-like domain. The characterization of these three unique isoforms may build upon the findings of others concerning the transforming potential of FGFR2 variants [<xref ref-type="bibr" rid="B4">4</xref>]. In particular, because it has been demonstrated that C3-IIIb variants may have more transforming activity than C1-IIIb variants, differences between the three C3 clones we have isolated may provide information on the influence of particular structural domains on transforming potential.</p><p>Previous studies that examined FGFR2 expression in prostate cancer have suggested that a change in the expression from the exon IIIb to IIIc isoform correlates with a progression from an androgen-sensitive to an androgen-insensitive state. RT-PCR analysis on the SUM-52PE breast cancer cell line showed that this cell line exclusively expressed the IIIb FGFR2 isoform (Fig. <xref ref-type="fig" rid="F3">3</xref>). Exon IIIb expression was also exclusively found in normal luminal HME cells (data not shown). This suggests that exon IIIb to IIIc switching is not necessary for FGFR2 to act as an oncogene when the gene is amplified. Rather, overexpression of one of the common IIIb isoforms or one of the novel variants may be important in driving transformation of HME cells. Ongoing studies are aimed at characterizing the transforming ability of individual FGFR2 isoforms obtained from SUM-52PE cells. These studies will directly test the hypothesis that specific FGFR2 isoforms have transforming activity towards HME cells and will compare variants with the different carboxyl termini. Overexpression of the C1-#38 and C3-#5 FGFR2 clones has been successfully accomplished in both the MCF-10A and H16N2 HME cell lines, and these cells have acquired phenotypes that distinguish them from parental cells (to be described in detail in a separate paper that is in preparation). Thus, by overexpressing FGFR2 isoforms in a physiologically relevant system, we hope to determine the isoform(s) that acts in a dominant way in the process of cell transformation, as well as to determine whether different regions present in individual clones drive specific phenotypes associated with transformation.</p></sec>

Polymorphic repeat in <italic>AIB1</italic> does not alter breast cancer risk

<sec><title>Introduction:</title><p>A causal association between endogenous and exogenous estrogens and breast cancer has been established. Steroid hormones regulate the expression of proteins that are involved in breast cell proliferation and development after binding to their respective steroid hormone receptors. Coactivator and corepressor proteins have recently been identified that interact with steroid hormone receptors and modulate transcriptional activation [<xref ref-type="bibr" rid="B1">1</xref>]. AIB1 (amplified in breast 1) is a member of the steroid receptor coactivator (SRC) family that interacts with estrogen receptor (ER)&#x003b1; in a ligand-dependent manner, and increases estrogen-dependent transcription [<xref ref-type="bibr" rid="B2">2</xref>]. Amplification and overexpression of <italic>AIB1</italic> has been observed in breast and ovarian cancer cell lines and in breast tumors [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. A polymorphic stretch of glutamine amino acids, with unknown biologic function, has recently been described in the carboxyl-terminal region of AIB1 [<xref ref-type="bibr" rid="B4">4</xref>]. Among women with germline <italic>BRCA1</italic> mutations, significant positive associations were observed between <italic>AIB1</italic> alleles with 26 or fewer glutamine repeats and breast cancer risk [<xref ref-type="bibr" rid="B5">5</xref>]</p></sec><sec><title>Aim:</title><p>To establish whether <italic>AIB1</italic> repeat alleles are associated with breast cancer risk and specific tumor characteristics among Caucasian women.</p></sec><sec><title>Patients and methods:</title><p>We evaluated associations prospectively between <italic>AIB1</italic> alleles and breast cancer risk in the Nurses' Health Study using a nested case-control design. The Nurses' Health Study was initiated in 1976, when 121 700 US-registered nurses between the ages of 30 and 55 years returned an initial questionnaire reporting medical histories and baseline health-related exposures. Between 1989 and 1990 blood samples were collected from 32 826 women. Eligible cases in this study consisted of women with pathologically confirmed incident breast cancer from the subcohort who gave a blood specimen. Cases with a diagnosis anytime after blood collection up to June 1, 1994, with no previously diagnosed cancer except for nonmelanoma skin cancer were included. Controls were randomly selected participants who gave a blood sample and were free of diagnosed cancer (except nonmelanoma skin cancer) up to and including the interval in which the cases were diagnosed, and were matched to cases on year of birth, menopausal status, postmenopausal hormone use, and time of day, month and fasting status at blood sampling. The nested case-control study consisted of 464 incident breast cancer cases and 624 matched controls. The protocol was approved by the Committee on Human Subjects, Brigham and Womens' Hospital, Boston, Massachusetts USA. Information regarding breast cancer risk factors was obtained from the 1976 baseline questionnaire, subsequent biennial questionnaires, and a questionnaire that was completed at the time of blood sampling. Histopathologic characteristics, such as stage, tumor size and ER and progesterone receptor (PR) status, were ascertained from medical records when available and used in case subgroup analyses.</p><p><italic>AIB1</italic> repeat alleles were determined by automated fluorescence-based fragment detection from polymerase chain reaction (PCR)-amplified DNA extracted from peripheral blood lymphocytes. Fluorescent 5' -labeled primers were utilized for PCR amplification, and glutamine repeat number discrimination was performed using the ABI Prism 377 DNA Sequencer (Perkin-Elmer, Foster City, CA, USA). Genotyping was performed by laboratory personnel who were blinded to case-control status, and blinded quality control samples were inserted to validate genotyping identification procedures (<italic>n</italic> = 110); concordance for the blinded samples was 100%. Methods regarding plasma hormone assays have previously been reported [<xref ref-type="bibr" rid="B6">6</xref>]. Conditional and unconditional logistic regression models, including terms for the matching variables and other potential confounders, were used to assess the association of <italic>AIB1</italic> alleles and breast cancer characterized by histologic subtype, stage of disease, and ER and PR status. We also evaluated whether breast cancer risk associated with <italic>AIB1</italic> genotype differed within strata of established breast cancer risk factors, and whether repeat length in <italic>AIB1</italic> indirectly influenced plasma hormone levels.</p></sec><sec><title>Results:</title><p>The case-control comparisons of established breast cancer risk factors among these women have previously been reported [<xref ref-type="bibr" rid="B7">7</xref>], and are generally consistent with expectation. The mean age of the women was 58.3 (standard deviation [SD] 7.1) years, ranging from 43 to 69 years at blood sampling. There were 188 premenopausal and 810 postmenopausal women, with mean ages of 48.1 (SD 2.8) years and 61.4 (SD 5.0) years, respectively, at blood sampling. Women in this study were primarily white; Asians, African-Americans and Hispanics comprised less than 1% of cases or controls.</p><p>The distribution of <italic>AIB1</italic> glutamine repeat alleles and <italic>AIB1</italic> genotypes for cases and controls are presented in Table <xref ref-type="table" rid="T1">1</xref>. Women with <italic>AIB1</italic> alleles of 26 glutamine repeats or fewer were not at increased risk for breast cancer (odds ratio [OR] 1.01, 95% confidence interval [CI] 0.75-1.36; Table <xref ref-type="table" rid="T2">2</xref>). Results were also similar by menopausal status and in analyses additionally adjusting for established breast cancer risk factors. Among premenopausal women, the OR for women with at least one allele with 26 glutamine repeats or fewer was 0.82 (95% Cl 0.37-1.81), and among postmenopausal women the OR was 1.09 (95% Cl 0.78-1.52; Table <xref ref-type="table" rid="T2">2</xref>). We did not observe evidence of a positive association between shorter repeat length and advanced breast cancer, defined as women with breast cancer having one or more involved nodes (OR 1.07, 95% Cl 0.64-1.78), or with cancers with a hormone-dependent phenotype (ER-positive: OR 1.16, 95% Cl 0.81-1.65; Table <xref ref-type="table" rid="T3">3</xref>). No associations were observed among women who had one or more alleles with 26 glutamine repeats or fewer, with or without a family history of breast cancer (family history: OR 1.09; 95% Cl 0.46-2.58; no family history: OR 0.94; 95% Cl 0.68-1.31; test for interaction <italic>P</italic> = 0.65). We also did not observe associations with breast cancer risk to be modified by other established breast cancer risk factors. Among postmenopausal controls not using postmenopausal hormones, geometric least-squared mean plasma levels of estrone sulfate and estrone were similar among carriers and noncarriers of <italic>AIB1</italic> alleles with 26 glutamine repeats or fewer (both differences: &#x02264; +3.5%; <italic>P</italic> &#x0003e;0.50). Mean levels of estradiol were slightly, but nonsignificantly elevated among carriers of alleles with 26 glutamine repeats or fewer (+11.6%; <italic>P</italic> = 0.08).</p></sec><sec><title>Discussion:</title><p>In this population-based nested case-control study, women with at most 26 repeating glutamine codons (CAG/CAA) within the carboxyl terminus of AIB1 were not at increased risk for breast cancer. We did not observe shorter repeat alleles to be positively associated with breast cancer grouped by histologic subtype, stage of disease, or by ER and PR status. These data suggest that <italic>AIB1</italic> repeat length is not a strong independent risk factor for postmenopausal breast cancer, and does not modify the clinical presentation of the tumor among Caucasian women in the general population.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Haiman</surname><given-names>Christopher A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Hankinson</surname><given-names>Susan E</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Spiegelman</surname><given-names>Donna</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Colditz</surname><given-names>Graham A</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Willett</surname><given-names>Walter C</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Speizer</surname><given-names>Frank E</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Brown</surname><given-names>Myles</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Hunter</surname><given-names>David J</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib>

Breast Cancer Research : BCR

<sec><title>Introduction</title><p>A causal association between endogenous and exogenous estrogens and breast cancer development has been established [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. Variant alleles of candidate genes involved in steroid hormone production (eg <italic>CYP17</italic>, <italic>CYP19</italic> and <italic>17&#x003b2; -HSD</italic>) [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>] and polymorphic variants in steroid hormone receptors (eg <italic>ER&#x003b1;</italic>, <italic>PR</italic> and <italic>AR</italic>) [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>] are currently being evaluated as potential high prevalence, low penetrance markers of breast cancer risk. Women with biologically functional polymorphisms in these genes may be exposed to increased levels of circulating steroid hormones or enhanced activity of hormone-responsive genes, and thus have altered susceptibility to steroid hormone-associated cancers.</p><p>Steroid hormones regulate the expression of proteins that are involved in breast cell proliferation and development after binding to their respective steroid hormone receptors. Coactivator and corepressor proteins have recently been identified that interact with steroid hormone receptors and modulate transcriptional activation [<xref ref-type="bibr" rid="B1">1</xref>]. AIB1 is a member of the SRC family that interacts with ER&#x003b1; in a ligand-dependent manner and increases estrogen-dependent transcription [<xref ref-type="bibr" rid="B2">2</xref>]. Amplification and over-expression of AIB1 has been observed in breast and ovarian cancer cell lines and in breast tumors [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. A polymorphic stretch of glutamine amino acids, with unknown biologic function, has recently been described in the carboxyl-terminal region of <italic>AIB1</italic> [<xref ref-type="bibr" rid="B4">4</xref>]. The androgen receptor has a similar polymorphic glutamine region, and shorter repeat lengths have been associated with increased risk of prostate cancer [<xref ref-type="bibr" rid="B18">18</xref>]. Among women with germline <italic>BRCA1</italic> mutations, significant positive associations were observed between <italic>AIB1</italic> alleles of 26 or fewer glutamine repeats and breast cancer risk [<xref ref-type="bibr" rid="B5">5</xref>].</p><p>The goal of the present study was to establish whether <italic>AIB1</italic> repeat alleles are associated with breast cancer risk among Caucasian women. We evaluated associations prospectively between <italic>AIB1</italic> alleles and breast cancer risk in the Nurses' Health Study using a nested case-control design. We investigated potential interactions between <italic>AIB1</italic> alleles and established breast cancer risk factors, and whether specific repeat alleles may provide a growth advantage for the developing malignancy and thus be associated with histopathologic tumor characteristics. We also examined whether repeat length in <italic>AIB1</italic> indirectly influences plasma hormone levels.</p></sec><sec sec-type="methods"><title>Patients and methods</title><sec><title>Study population</title><p>The Nurses' Health Study was initiated in 1976, when 121 700 US-registered nurses between the ages of 30 and 55 years returned an initial questionnaire reporting medical histories and baseline health-related exposures. Updated information has been obtained by questionnaire every 2 years, including data on reproductive variables, and oral contraceptive and postmenopausal hormone use. Incident breast cancers are identified by self-report and confirmed by medical record review. Between 1989 and 1990, blood samples were collected from 32 826 women. Approximately 97% of the blood samples were returned within 26 h of being drawn, immediately centrifuged, aliquoted into plasma, red blood cells and buffy coat fractions, and stored in liquid nitrogen freezers. Subsequent follow up has been greater than 98% for this subcohort.</p><p>Eligible cases in this study consisted of women with pathologically confirmed incident breast cancer from the subcohort who gave a blood specimen. Cases with a diagnosis anytime after blood collection up to June 1, 1994, with no previously diagnosed cancer except for nonmelanoma skin cancer were included. Controls were randomly selected participants who gave a blood sample and were free of diagnosed cancer (except nonmelanoma skin cancer) up to and including the interval in which the cases were diagnosed. Controls were matched to cases on year of birth, menopausal status, postmenopausal hormone use, and time of day, month and fasting status at blood sampling. Women were defined as postmenopausal at the time of a bilateral oophorectomy or after having no menstrual cycle within the past 12 months before blood sampling. Women who had had a hysterectomy with one or both ovaries left intact were classified as pre-menopausal until the age at which 10% of the cohort had undergone natural menopause (46 years for smokers and 48 years for nonsmokers), and as postmenopausal at the age at which 90% of the cohort had undergone natural menopause (54 for smokers and 56 for nonsmokers). During the intervening years these women were classified as being of uncertain menopausal status.</p><p>For postmenopausal cases not using postmenopausal hormones within 3 months before blood sampling, we matched a second control to increase our statistical power in plasma steroid hormone analyses. The nested case-control study consisted of 464 incident breast cancer cases and 624 matched controls. The study sample for the plasma hormone analysis was comprised of 298 postmenopausal controls not using hormone replacement therapy within 3 months of blood sampling.</p><p>The protocol was approved by the Committee on Human Subjects, Brigham and Womens' Hospital, Boston, Massachusetts, USA.</p></sec><sec><title>Exposure data</title><p>Information regarding breast cancer risk factors was obtained from the 1976 baseline questionnaire, subsequent biennial questionnaires, and a questionnaire completed at the time of blood sampling. Menopausal status and use of postmenopausal hormones was assessed at blood sampling, and was updated until date of diagnosis for cases and the equivalent date for matched controls. Histopathologic characteristics such as stage, tumor size, and ER and PR status were ascertained from medical records when available, and used in case subgroup analyses.</p></sec><sec><title>Genotyping analysis</title><p>DNA was extracted from buffy coat fractions using the Qiagen QIAamp^&#x000ae; Blood Kit (Qiagen Incorporated, Chatsworth, CA, USA). <italic>AIB1</italic> genotyping was performed as follows. PCR amplification of the polymorphic fragment was generated using the following primers: 5'-TTCCGACAACAGAGGGTGG-3' (forward) and 5'-AGTCA CATTAGGTGGGC-3' (reverse). Forty nanograms of genomic DNA was used per 22 &#x003bc; l reaction with 1.7 &#x003bc; l of each 10 &#x003bc; mol/l primer, 4 &#x003bc; l of 10 &#x003bc; mol/l dNTPs, 2.2 &#x003bc; l 10 &#x000d7; PCR buffer, 9.0 &#x003bc; l water and 1.50 U <italic>Taq</italic> polymerase (Qiagen Incorporated). Amplification conditions were 2 min of initial denaturation at 94&#x000b0;C followed by 35 cycles of 30 s at 94&#x000b0;C, 90 s at 60&#x000b0;C and 30 s at 72&#x000b0;C, followed by a final extension at 72&#x000b0;C for 8 min. Two fluorescent 5' -labeled primers were utilized, allowing two samples per lane. A 5% Long Ranger/6 mol/l urea gel (Biowhittaker Molecular Applications, Rockland, ME, USA) was used for rapid fragment detection using the ABI Prism 377 DNA Sequencer (Perkin-Elmer, Foster City, CA, USA) at the Dana Farber Cancer Institute Molecular Biology Core Facility and the Harvard Center for Cancer Prevention, Boston, MA, USA. The size of the amplified products was determined relative to an internal size standard using Genescan and Genotyper Analysis software (Perkin-Elmer). Genotyping was performed by laboratory personnel who were blinded to case-control status, and blinded quality control samples were inserted to validate genotype identification procedures (<italic>n</italic> =110); concordance for the blinded samples was 100%.</p></sec><sec><title>Hormone analysis</title><p>Steroid hormone fractions of estrone sulfate, estrone and estradiol were assayed in up to three separate batches. Estrone sulfate from batches 1 and 2 were assayed in the laboratory of Dr C Longcope (University of Massachusetts Medical Center, Worcester, MA, USA). All other analyses were performed by at the Nichols Institute (San Juan Capistrano, CA, USA). Methods for plasma hormone assays and information regarding laboratory precision and reproducibility have previously been published [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Within-batch laboratory coefficients of variation were 13.6% or less.</p></sec><sec><title>Statistical analysis</title><p>A Mantel-Haenszel &#x003c7;²-test across matched case-control sets was used to evaluate differences in the frequency of alleles with 26 repeats or fewer. ORs and 95% CIs were calculated using conditional and unconditional logistic regression. In addition to the matching variables, we adjusted for the following breast cancer risk factors: body mass index (kg/m²) at age 18 years (continuous); weight gain since age 18 years (&#x0003c; 5, 5-19.9, or &#x02265; 20 kg); age at menarche (&#x0003c; 12, 12, 13, or &#x0003e; 13 years); parity/age at first birth (nulliparous, one to two children/age at first birth &#x02264; 24 years, one to two children/age at first birth &#x0003e; 24 years, three or more children/age at first birth &#x02264; 24 years, three or more children/age at first birth &#x0003e; 24 years); first-degree family history of breast cancer (yes/no); history of benign breast disease (yes/no); and duration of postmenopausal hormone use (never; in the past for &#x0003c; 5 or &#x02265; 5 years; or current for &#x0003c; 5 years or &#x02265; 5 years). We also adjusted for age at menopause (years) in analyses limited to post-menopausal women.</p><p>To be consistent with prior studies [<xref ref-type="bibr" rid="B5">5</xref>], indicator variables for three <italic>AIB1</italic> genotypes were created: two alleles with 26 repeats or fewer, one allele with 26 repeats or fewer, and two alleles with more than 26 repeats (reference group). We also evaluated breast cancer risk for specific <italic>AIB1</italic> genotypes (28/29, 28/28, 26/29, 26/28 and 26/26) relative to 29/29 homozygotes. Unconditional logistic regression models, including terms for the matching variables and other potential confounders, were used to assess the association between <italic>AIB1</italic> alleles and breast cancer characterized by histologic subtype, stage of disease, and ER and PR status. To increase statistical power, all controls were utilized when cases were limited to clinical parameters and histologic subtypes. We also evaluated whether breast cancer risk associated with <italic>AIB1</italic> genotype differed between strata of established breast cancer risk factors. Because of the low prevalence of genotypes with both alleles with 26 glutamine repeats or fewer, in stratified analyses carriers of alleles with 26 repeats or fewer were compared with noncarriers. Multiplicative interactions were evaluated by including interaction terms between genotype and risk factor variables in logistic regression models. The likelihood ratio test was used to assess the statistical significance of these interactions.</p><p>Mixed regression models for clustered data were used to evaluate the association between genotype and circulating hormone levels among postmenopausal controls, controlling for body mass index at blood sampling and the matching variables [<xref ref-type="bibr" rid="B20">20</xref>]. The natural logarithms of the plasma hormone values were used in the analyses to reduce the skew of the regression residuals. Information regarding outlying values and exclusions were previously reported [<xref ref-type="bibr" rid="B7">7</xref>]. We used the Statistical Analysis System for all analyses [<xref ref-type="bibr" rid="B21">21</xref>].</p></sec></sec><sec><title>Results</title><sec><title><italic>AIB1</italic> genotype and breast cancer risk</title><p>The case-control comparisons of established breast cancer risk factors among the women studied have previously been published [<xref ref-type="bibr" rid="B7">7</xref>], and are generally consistent with expectation. The mean age of the women was 58.3 (SD 7.1) years, ranging from 43 to 69 years at blood sampling. There were 188 premenopausal and 810 post-menopausal women, with mean ages of 48.1 (SD 2.8) years and 61.4 (SD 5.0) years, respectively, at blood sampling. Women in this study were primarily white; Asians, African-Americans and Hispanics comprised less than 1% of cases or controls.</p><p>The distribution of <italic>AIB1</italic> glutamine repeat alleles and <italic>AIB1</italic> genotypes for cases and controls are presented in Table <xref ref-type="table" rid="T1">1</xref>. In this study we observed alleles of 19, 22, 25, 26, 27, 28, 29, 30, 31 and 34 repeats, and the expected genotype frequencies, based on the allele frequencies among the cases and controls, were similar to the observed distributions. The allele frequencies were nearly identical for cases and controls; the most prevalent alleles were of 29 (cases 46.4%, controls 47.8%), 28 (cases 40.4%, controls 38.9%), and 26 (cases 12.0%, controls 11.9%) repeats. Average repeat lengths were also the same for cases and controls (cases 28.2, controls 28.2). Twenty-six-repeat allele homozygotes and heterozygotes were evenly distributed among cases and controls (26/26 homozygotes: cases 1.9%, controls 1.6%; 26/X heterozygotes: cases 20.0%; controls 20.5%). A cutoff point of 26 repeats or fewer was used to allow comparison with previously published data [<xref ref-type="bibr" rid="B5">5</xref>]. Using this cutoff point, allele frequencies were almost identical for cases and controls (one or more alleles of 26 repeats or fewer: 12.1% versus 12.2%; <italic>P</italic> = 0.82).</p><p>Women with 26 repeats or fewer were not at increased risk for breast cancer; compared with noncarriers, the adjusted OR for carriers of alleles with 26 repeats or fewer was 1.01 (95% Cl 0.75-1.36; Table <xref ref-type="table" rid="T2">2</xref>). Compared with noncarriers, the adjusted ORs for heterozygotes and homozygotes for alleles with 26 repeats or fewer were 0.99 (95% Cl 0.73-1.35) and 1.29 (95% Cl 0.52-3.23), respectively. Results were also similar by menopausal status and when adjusting for established breast cancer risk factors. Among premenopausal women, the OR for women with at least one allele with 26 repeats or fewer was 0.82 (95% Cl 0.37-1.81), and among post-menopausal women the corresponding OR was 1.09 (95% Cl 0.78-1.52). Compared with 29/29 homozygotes, no significant associations were observed for specific genotypes with alleles of 26 repeats (Table <xref ref-type="table" rid="T2">2</xref>).</p><table-wrap position="float" id="T1"><label>Table 1</label><caption><p><italic>AIB1</italic> allele^* and genotype frequency among cases and controls, Nurses' Health Study 1989-1994</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left"><italic>AIB1</italic> allele</td><td align="center">Cases (<italic>n</italic> [%])</td><td align="center">Controls (<italic>n</italic> [%])</td></tr></thead><tbody><tr><td align="left">19</td><td align="center">0 (0.0)</td><td align="center">1 (0.1)</td></tr><tr><td align="left">22</td><td align="center">0 (0.0)</td><td align="center">1 (0.1)</td></tr><tr><td align="left">25</td><td align="center">1 (0.1)</td><td align="center">1 (0.1)</td></tr><tr><td align="left">26</td><td align="center">111 (12.0)</td><td align="center">148 (11.9)</td></tr><tr><td align="left">27</td><td align="center">3 (0.3)</td><td align="center">3 (0.2)</td></tr><tr><td align="left">28</td><td align="center">375 (40.4)</td><td align="center">485 (38.9)</td></tr><tr><td align="left">29</td><td align="center">431 (46.4)</td><td align="center">597 (47.8)</td></tr><tr><td align="left">30</td><td align="center">4 (0.4)</td><td align="center">10 (0.8)</td></tr><tr><td align="left">31</td><td align="center">2 (0.2)</td><td align="center">2 (0.2)</td></tr><tr><td align="left">34</td><td align="center">1 (0.1)</td><td align="center">0 (0.0)</td></tr><tr><td align="left">Total</td><td align="center">928</td><td align="center">1248</td></tr><tr><td colspan="3"><hr></hr></td></tr><tr><td align="left"><italic>AIB1</italic> genotype</td><td align="center">Cases (<italic>n</italic> [%])</td><td align="center">Controls (<italic>n</italic> [%])</td></tr><tr><td colspan="3"><hr></hr></td></tr><tr><td align="left">19/29</td><td align="center">0 (0)</td><td align="center">1 (0.2)</td></tr><tr><td align="left">22/29</td><td align="center">0 (0)</td><td align="center">1 (0.2)</td></tr><tr><td align="left">25/28</td><td align="center">0 (0)</td><td align="center">1 (0.2)</td></tr><tr><td align="left">25/29</td><td align="center">1 (0.2)</td><td align="center">0 (0.0)</td></tr><tr><td align="left">26/26</td><td align="center">9 (1.9)</td><td align="center">10 (1.6)</td></tr><tr><td align="left">26/28</td><td align="center">45 (9.7)</td><td align="center">55 (8.8)</td></tr><tr><td align="left">26/29</td><td align="center">47 (10.1)</td><td align="center">71 (11.4)</td></tr><tr><td align="left">26/30</td><td align="center">1 (0.2)</td><td align="center">2 (0.3)</td></tr><tr><td align="left">27/28</td><td align="center">2 (0.4)</td><td align="center">1 (0.2)</td></tr><tr><td align="left">27/29</td><td align="center">1 (0.2)</td><td align="center">2 (0.3)</td></tr><tr><td align="left">28/28</td><td align="center">64 (13.8)</td><td align="center">98 (15.7)</td></tr><tr><td align="left">28/29</td><td align="center">196 (42.2)</td><td align="center">228 (36.5)</td></tr><tr><td align="left">28/30</td><td align="center">2 (0.4)</td><td align="center">3 (0.5)</td></tr><tr><td align="left">28/31</td><td align="center">1 (0.2)</td><td align="center">1 (0.2)</td></tr><tr><td align="left">28/34</td><td align="center">1 (0.2)</td><td align="center">0 (0.0)</td></tr><tr><td align="left">29/29</td><td align="center">92 (19.8)</td><td align="center">144 (23.1)</td></tr><tr><td align="left">29/30</td><td align="center">1 (0.2)</td><td align="center">5 (0.8)</td></tr><tr><td align="left">29/31</td><td align="center">1 (0.2)</td><td align="center">1 (0.2)</td></tr><tr><td align="left">Total</td><td align="center">464</td><td align="center">624</td></tr></tbody></table><table-wrap-foot><p>^*There was no significant difference in frequency of alleles with 26 repeats or fewer between cases and controls (<italic>P</italic> = 0.82).</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Association of <italic>AIB1</italic> genotype and breast cancer risk, Nurses' Health Study 1989-1994</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Premenopausal</td><td align="center">Cases</td><td align="center">Controls</td><td align="center">OR^*</td><td align="center">95% CI</td></tr></thead><tbody><tr><td align="left">No of alleles with 26 or fewer repeats</td><td></td><td></td><td></td><td></td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">0</td><td align="center">49</td><td align="center">51</td><td align="center">1.00</td><td></td></tr><tr><td align="left">1</td><td align="center">14</td><td align="center">16</td><td align="center">0.91</td><td align="center">0.40-2.09</td></tr><tr><td align="left">2</td><td align="center">1</td><td align="center">3</td><td align="center">0.32</td><td align="center">0.03-3.25</td></tr><tr><td align="left">1 or 2</td><td align="center">15</td><td align="center">19</td><td align="center">0.82</td><td align="center">0.37-1.81</td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">Postmenopausal</td><td align="center">Cases</td><td align="center">Controls</td><td align="center">OR^&#x02020;</td><td align="center">95% CI</td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">No of alleles with 26 or fewer repeats</td><td></td><td></td><td></td><td></td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">0</td><td align="center">275</td><td align="center">397</td><td align="center">1.00</td><td></td></tr><tr><td align="left">1</td><td align="center">75</td><td align="center">106</td><td align="center">1.04</td><td align="center">0.74-1.46</td></tr><tr><td align="left">2</td><td align="center">8</td><td align="center">7</td><td align="center">1.90</td><td align="center">0.67-5.42</td></tr><tr><td align="left">1 or 2</td><td align="center">83</td><td align="center">113</td><td align="center">1.09</td><td align="center">0.78-1.52</td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">Combined analysis</td><td align="center">Cases</td><td align="center">Controls^&#x02021;</td><td align="center">OR^&#x000a7;</td><td align="center">95% CI</td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">No of alleles with 26 or fewer repeats</td><td></td><td></td><td></td><td></td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">0</td><td align="center">361</td><td align="center">476</td><td align="center">1.00</td><td></td></tr><tr><td align="left">1</td><td align="center">94</td><td align="center">129</td><td align="center">0.99</td><td align="center">0.73-1.35</td></tr><tr><td align="left">2</td><td align="center">9</td><td align="center">10</td><td align="center">1.29</td><td align="center">0.52-3.23</td></tr><tr><td align="left">1 or 2</td><td align="center">103</td><td align="center">139</td><td align="center">1.01</td><td align="center">0.75-1.36</td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left"><italic>AIB1</italic> genotype^&#x000b6;</td><td></td><td></td><td></td><td></td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">29/29</td><td align="center">92</td><td align="center">141</td><td align="center">1.00</td><td></td></tr><tr><td align="left">28/29</td><td align="center">196</td><td align="center">227</td><td align="center">1.27</td><td align="center">0.91-1.77</td></tr><tr><td align="left">28/28</td><td align="center">64</td><td align="center">95</td><td align="center">0.99</td><td align="center">0.65-1.50</td></tr><tr><td align="left">26/29</td><td align="center">47</td><td align="center">69</td><td align="center">1.04</td><td align="center">0.66-1.64</td></tr><tr><td align="left">26/28</td><td align="center">45</td><td align="center">55</td><td align="center">1.23</td><td align="center">0.75-2.02</td></tr><tr><td align="left">26/26</td><td align="center">9</td><td align="center">10</td><td align="center">1.48</td><td align="center">0.57-3.81</td></tr></tbody></table><table-wrap-foot><p>^*Unconditional logistic regression adjusted for the matching variables: age, date of blood sampling, time of blood sampling and fasting status. ^&#x02020;Unconditional logistic regression adjusted for the matching variables: age, postmenopausal hormone use, date of blood sampling, time of blood sampling and fasting status. ^&#x02021;Nine controls were excluded due to incomplete matching. ^&#x000a7;Conditional logistic regression adjusted for the matching variables: age, menopausal status, postmenopausal hormone use, date of blood sampling, time of blood sampling and fasting status. ^&#x000b6;Genotypes combined and modeled together, but not included in table (cases, controls): 19/29 (0, 1), 22/29 (0, 1), 25/28 (0, 1), 25/29 (1, 0), 26/30 (1, 2), 27/28 (2, 1), 27/29 (1, 2), 28/30 (2, 3), 28/31 (1, 1), 28/34 (1, 0), 29/30 (1, 5), and 29/31 (1, 1).</p></table-wrap-foot></table-wrap></sec><sec><title><italic>AIB1</italic> genotype and tumor characteristics</title><p>In analyses by histologic subtype, results were similar when limiting cases to those with invasive disease, or advanced stage of disease as characterized by number of involved lymph nodes, tumor size, and type of carcinoma (ductal versus lobular; Table <xref ref-type="table" rid="T3">3</xref>). Results were also similar after adjustment for established breast cancer risk factors. A positive association was observed between short repeat alleles and poorly differentiated tumors, but this was not statistically significant (OR 1.54, 95% Cl 0.90-2.64). We did not observe evidence that shorter <italic>AIB1</italic> alleles were associated with a hormone-dependent phenotype. In case-case analyses, alleles with 26 repeats or fewer were not over-represented among cases that were positive for the ER or PR (ER-positive versus ER-negative: 24% [61/250] versus 21% [12/58], two-tailed Fisher's Exact Test [FET] <italic>P</italic> = 0.61; PR-positive versus PR-negative: 24% [45/187] versus 22% [24/110], two-tailed FET <italic>P</italic> = 0.78).</p><table-wrap position="float" id="T3"><label>Table 3</label><caption><p>Associations between <italic>AIB1</italic> genotypes and breast cancer risk by histologic subtype and receptor status</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center" colspan="2">No of <italic>AIB1</italic> alleles</td><td></td><td></td></tr><tr><td></td><td align="center" colspan="2">with &#x02264;26 repeats</td><td></td><td></td></tr><tr><td/><td colspan="2"><hr></hr></td><td/><td/></tr><tr><td align="left">Controls/cases</td><td align="center">0</td><td align="center">1 or 2</td><td align="center">OR^*</td><td align="center">95% Cl</td></tr></thead><tbody><tr><td align="left">Controls</td><td align="center">483</td><td align="center">141</td><td></td><td></td></tr><tr><td align="left">Cases^&#x02020;</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Invasive</td><td align="center">295</td><td align="center">93</td><td align="center">1.13</td><td align="center">0.83-1.54</td></tr><tr><td align="left">&#x02003;Ductal</td><td align="center">252</td><td align="center">79</td><td align="center">1.10</td><td align="center">0.81-1.53</td></tr><tr><td align="left">&#x02003;Lobular</td><td align="center">33</td><td align="center">9</td><td align="center">1.05</td><td align="center">0.48-2.30</td></tr><tr><td align="left">&#x02003;Involved nodes</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;&#x02265; 1</td><td align="center">83</td><td align="center">24</td><td align="center">1.07</td><td align="center">0.64-1.78</td></tr><tr><td align="left">&#x02003;&#x02265; 4</td><td align="center">30</td><td align="center">6</td><td align="center">0.72</td><td align="center">0.29-1.81</td></tr><tr><td align="left">&#x02003;Receptor status</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;ER⁺</td><td align="center">189</td><td align="center">61</td><td align="center">1.16</td><td align="center">0.81-1.65</td></tr><tr><td align="left">&#x02003;ER^-</td><td align="center">46</td><td align="center">12</td><td align="center">0.94</td><td align="center">0.47-1.85</td></tr><tr><td align="left">&#x02003;PR⁺</td><td align="center">142</td><td align="center">45</td><td align="center">1.14</td><td align="center">0.77-1.69</td></tr><tr><td align="left">&#x02003;PR^-</td><td align="center">86</td><td align="center">24</td><td align="center">1.03</td><td align="center">0.63-1.71</td></tr><tr><td align="left">&#x02003;Tumor size</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;&#x02264; 2 cm</td><td align="center">207</td><td align="center">64</td><td align="center">1.13</td><td align="center">0.80-1.60</td></tr><tr><td align="left">&#x02003;&#x0003e; 2 cm</td><td align="center">74</td><td align="center">25</td><td align="center">1.21</td><td align="center">0.73-2.00</td></tr><tr><td align="left">&#x02003;Degree of differentiation</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;High or</td><td align="center">98</td><td align="center">37</td><td align="center">1.36</td><td align="center">0.89-2.10</td></tr><tr><td align="left">&#x02003;moderate</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Poor</td><td align="center">55</td><td align="center">23</td><td align="center">1.54</td><td align="center">0.90-2.64</td></tr></tbody></table><table-wrap-foot><p>^*Unconditional logistic regression adjusted for matching variables: age, menopausal status, postmenopausal hormone use, date of blood sampling, time of blood sampling and fasting status. ^&#x02020;Numbers do not add up to total for invasive cancer because of missing data.</p></table-wrap-foot></table-wrap></sec><sec><title><italic>AIB1</italic> genotype-breast cancer risk factor interactions</title><p>Because the data of Rebbeck <italic>et al</italic> [<xref ref-type="bibr" rid="B5">5</xref>] suggested that breast cancer risk associated with carrying a germline <italic>BRCA1</italic> mutation may be modified by <italic>AIB1</italic> allele status, we examined the potential interaction between genotype and a first-degree family history of breast cancer. In case-case analyses, women with at least one repeat allele of 26 repeats or fewer were nonsignificantly over-represented among cases with a family history of breast cancer (family history versus no family history: 23/83 [27.7%] versus 80/381 [21.0%]; two-tailed FET <italic>P</italic> = 0.19). However, in the nested case-control study there was no difference in association between women with one or more repeat alleles of 26 or fewer repeats with or without a family history of breast cancer (family history: OR 1.09, 95% Cl 0.46-2.58; no family history: OR 0.94, 95% Cl 0.68-1.31; test for interaction <italic>P</italic> = 0.58).</p><p>We evaluated other potential interactions between breast cancer risk factors and <italic>AIB1</italic> alleles with 26 repeats or fewer. We observed no interactions between short repeat alleles and the following variables (<italic>P</italic> values represent tests for interaction): body mass index among post-menopausal women (<italic>P</italic> = 0.89), weight gain since age 18 years (<italic>P</italic> = 0.41), history of benign breast disease (<italic>P</italic> = 0.20), age at menarche (<italic>P</italic> = 0.24), age at first birth (<italic>P</italic> = 0.89) and parity (<italic>P</italic> = 0.18). We also stratified by oral contraceptive and postmenopausal hormone use status. The adjusted ORs for women who had ever and never used oral contraceptives were 0.75 (95% Cl 0.48-1.18) and 1.22 (95% Cl 0.79-1.89), respectively (test for interaction <italic>P</italic> = 0.14). No significant interaction was seen between carriers of alleles with 26 repeats or fewer and postmenopausal hormone use (test for interaction <italic>P</italic> = 0.41). The OR for alleles with 26 repeats or fewer among postmenopausal women who had never used hormones was 0.87 (95% Cl 0.45-1.66). Similar results were observed among past users (OR 1.18, 95% Cl 0.56-2.47) and current users of short duration (&#x0003c; 5 years; OR 1.16, 95% Cl 0.44-3.07). A borderline significant association was observed among current users who used postmenopausal hormones for 5 or more years (OR 1.98, 95% Cl 0.96-4.07).</p></sec><sec><title><italic>AIB1</italic> genotype and hormone levels</title><p>Geometric least-squared mean plasma levels of estrone sulfate and estrone were similar among carriers and non-carriers of <italic>AIB1</italic> alleles of 26 repeats or fewer (both differences &#x02264; +3.5%, <italic>P</italic> &#x0003e; 0.50). Mean levels of estradiol were slightly, but not significantly elevated among carriers of alleles with 26 repeats or fewer (+11.6%; <italic>P</italic> = 0.08). In analyses among cases the absolute differences were greater for carriers of alleles with 26 repeats or fewer (estrone sulfate: +18.5%, <italic>P</italic> = 0.21; estrone: +14.9%, <italic>P</italic> = 0.10; estradiol: +25.0%, <italic>P</italic> = 0.02).</p></sec></sec><sec><title>Discussion</title><p>In the present population-based nested case-control study, women with 26 or fewer repeating glutamine codons (CAG/CAA) within the carboxyl terminus of <italic>AIB1</italic> were not at increased risk for breast cancer. We did not observe shorter repeat alleles to be positively associated with breast cancer, as defined by histologic subtype or stage of disease, or by ER and PR status. We also did not observe evidence that risk associated with <italic>AIB1</italic> allele status is modified by established breast cancer risk factors.</p><p>Coregulators of gene transcription modulate transactivation of hormone-responsive genes through interactions with transcription factors and steroid hormone receptors [<xref ref-type="bibr" rid="B1">1</xref>]. Members of the SRC family share similar functional domains, including a glutamine-rich domain within the carboxyl-terminus. Shorter CAG repeat lengths in the androgen receptor are associated with increased transcriptional activation of that receptor [<xref ref-type="bibr" rid="B22">22</xref>], and may be a model for the potential biologic significance of the polymorphic repeat in <italic>AIB1</italic>. The biologic function of this glutamine-rich region is currently unknown, however, and to our knowledge no studies have correlated this <italic>AIB1</italic> repeat polymorphism with coactivator activity. Ainzick <italic>et al</italic> [<xref ref-type="bibr" rid="B2">2</xref>] first reported <italic>AIB1</italic> gene amplification in 9.5% (10/105) of primary breast cancer specimens, and <italic>AIB1</italic> mRNA over-expression in 58% of tumors without <italic>AIB1</italic> gene amplification relative to normal mammary epithelium. AIB1 was also found to interact with ER&#x003b1; in a ligand-dependent manner, and transfection of <italic>AIB1</italic> increased estrogen-dependent transcription. Bautista <italic>et al</italic> [<xref ref-type="bibr" rid="B3">3</xref>] also observed <italic>AIB1</italic> gene amplification in 4.8% (56/1157) of breast and 7.4% (9/122) of ovarian tumors. In that study gene amplification was correlated with larger ER-positive and PR-positive tumors. In a relatively large case-control study (cases <italic>n</italic> = 581; controls <italic>n</italic> = 786), Platz <italic>et al</italic> [<xref ref-type="bibr" rid="B23">23</xref>] observed no association between germline <italic>AIB1</italic> glutamine repeat length and prostate cancer incidence or stage.</p><p>To our knowledge, only one study has examined the association between <italic>AIB1</italic> repeat alleles and breast cancer risk. Among a cohort of 366 <italic>BRCA1</italic> mutation carriers, Rebbeck <italic>et al</italic> [<xref ref-type="bibr" rid="B5">5</xref>] observed a significant association between <italic>AIB1</italic> alleles with 26 repeats or fewer and breast cancer risk (OR 1.9, 95% Cl 1.5-2.9). These results suggest that <italic>AIB1</italic> genotypes may accentuate <italic>BRCA1</italic>-associated breast cancer risk through steroid hormone pathways. We were unable to examine this particular question because the study population we studied was not comprised of women with a high likelihood of carrying a highly penetrant mutation in <italic>BRCA1</italic> (ie early-onset cases with an extensive family history of breast cancer - only 18 women had more than one first-degree relative with breast cancer). In the present study the majority of cases were postmenopausal at diagnosis (approximately 77%) and were diagnosed as having breast cancer at a later age (mean 62.7 years). However, the relative risk for shorter allele carriers was actually nonsignificantly reduced among premenopausal women (ie those more likely to carry a <italic>BRCA1</italic> mutation). In addition, shorter repeat alleles were not associated with increased risk for breast cancer among women with a first-degree family history of breast cancer.</p><p>The strengths of the present study are the relatively large sample size and prospective design. We had high power (99.5% at the 0.05 significance level) to detect a significant relative risk of 2.0 among postmenopausal women, but we had low power (47.1%) to detect the same magnitude of association among premenopausal women. <italic>AIB1</italic> genotype may be an important predictor of breast cancer risk among the small proportion of predisposed women with highly penetrant mutations in breast cancer susceptibility genes. However, the present data do not provide evidence to support the hypothesis that <italic>AIB1</italic> genotype is involved in the development of postmenopausal breast cancer among Caucasian women in the general population.</p></sec>

P53 autoantibodies in 1006 patients followed up for breast cancer

<sec><title>Introduction:</title><p>Dysfunction of the tumour-suppressor protein, p53, may be due to either mutational or epigenetic factors, each of which may lead to accumulation of cytoplasmic p53. Abnormal accumulation of p53 in breast cancer tissue is predictive of poor prognosis [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Humoral studies [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>] have shown that cancer patients may develop immunity to abnormally expressed p53, as revealed by p53 autoantibodies in the blood. Again, prognostic correlates have been noted, with presence of circulating p53 autoantibodies at diagnosis of breast cancer being associated with reduced overall survival [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>] and with poor prognostic factors such as high histological grade and the absence of hormone receptors [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>].</p><p>Little is known of the potential value of p53 autoantibody in follow up of cancer. In lung cancer there is evidence that autoantibodies to p53 may provide a useful tool to monitor response to therapy [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>], whereas serial measurements of autoantibodies to p53 in 40 patients with advanced ovarian cancer were not found to be clinically useful [<xref ref-type="bibr" rid="B11">11</xref>]. In breast cancer some 30% of node-negative patients will relapse within 5 years, but there is no current means to predict those who are at risk.</p><p>We performed the present study to ask if the presence of autoantibodies to p53 has any association with breast cancer progression.</p></sec><sec><title>Materials and method:</title><p>A library of plasma samples were collected from all patients attending one general oncology clinic for postoperative follow up of breast cancer. The clinical status of each patient at the time of sampling was summarized. An average of eight plasma samples were cryopreserved for each patient over a period of 15 years.</p><p>The enzyme-linked immusorbent assay (ELISA) for p53 autoantibodies was developed in-house, based on the ELISA procedure of Lubin <italic>et al</italic> [<xref ref-type="bibr" rid="B3">3</xref>]. Our in-house method is detailed in the full text of this article. In one assay series we compared a commercial ELISA kit for p53 autoantibodies with our in-house ELISA. A total of 20 patients' samples were tested, representing a range of positive and negative readings. Two samples scored as strongly positive with the in-house assay, but only one of these two scored positive with the commercial assay. Having established sensitivity, specificity and reproducibility of the in-house assay, we judged that this was superior to the commercial assay both in terms of sensitivity and of cost (&#x000a3;1 per test compared with &#x000a3;23 per test). The in-house assay was thus used throughout the present study.</p></sec><sec><title>Results:</title><p>Serial plasma samples from 1006 patients with breast cancer revealed the following: (i) no correlation of p53 autoantibody status with disease status at the time of sample collection (Table <xref ref-type="table" rid="T1">1</xref>), or with menopausal status at time of primary diagnosis of breast cancer (Table <xref ref-type="table" rid="T2">2</xref>); (ii) 155 out of 1006 (15%) of patients were positive for p53 autoantibodies, and these patients tended to have a persistent autoantibody status throughout follow up, irrespective of disease behaviour; and (iii) where a negative autoantibody status was found at primary diagnosis of breast cancer, this negative status persisted throughout follow up, irrespective of later disease behaviour (Table <xref ref-type="table" rid="T3">3</xref>).</p></sec><sec><title>Discussion:</title><p>As a working hypothesis, we proposed that levels of autoantibodies to p53 would reflect tumour behaviour. However, we found that the presence or absence of p53 autoantibodies was not predictive of presence or absence of recurrent disease. There was an equivalent incidence of active disease at the time of sampling in both the autoantibody-negative and autoantibody-positive groups, these being 25.2 and 28.7%, respectively. Thus, humoral immune activity against p53 appeared to be relatively restricted to a subgroup of patients in whom, once an autoantibody response had been generated, antibody was likely to persist regardless of tumour behaviour. Conversely, where no detectable p53 autoantibody was present at the time of primary diagnosis, these patients remained similarly negative for antibody, irrespective of subsequent disease activity (Table <xref ref-type="table" rid="T3">3</xref>).</p><p>In contrast to shed markers that correlate with tumour mass, such as CA15.3 for cancer of the breast, any tumour-related immune response will be subject to complex regulation. Autoantibody responses to p53 will require appropriate primary immunization; initial low-dose antigen exposure may induce immune tolerance and lack of response. Higher antigen doses may activate either antibody-mediated immunity, or cellular immunity.</p><p>In breast cancer patients, our results suggest that, once an active humoral response against p53 is established, then this remains active. This persistent humoral reaction may be driven by persistent antigenic stimulation by p53 protein derived from overexpression of p53 at distant metastatic sites; alternatively, irradiated normal tissue may be a source of continued antigenic stimulation, because a long-term side effect of radiation therapy is an increased expression of p53 in normal breast tissue that persists for several years [<xref ref-type="bibr" rid="B12">12</xref>]. Since the great majority of our total patient cohort had received radiotherapy, humoral immunity to p53 associated with primary disease might persist, even in those patients who enter remission, due to tumour-independent antigenic stimulation.</p><p>Loss of p53 function is known to correlate with loss of efficacy of cancer therapy <italic>in vivo</italic> [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. This raised the possibility that autoantibodies to p53 that develop during follow up might indicate those patients whose tumor has become resistant to therapy. However, the present results show that, if no immunity has been generated at the time of primary diagnosis, then later immunity is unlikely to occur. This corresponds to the finding that expression of p53 antigen in biopies of locally advanced breast cancer did not correlate with drug resistance [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. Overall, the present observations show that screening for p53 autoantibody status is not informative on residual tumour activity, or on therapeutic responsiveness. We conclude that the potential value of p53 autoantibody screening in patients with breast cancer is limited to the prognostic information obtained at diagnosis.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Metcalfe</surname><given-names>Su</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>smm1001@cam.ac.uk</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Wheeler</surname><given-names>Terence K</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Picken</surname><given-names>Sheila</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Negus </surname><given-names>Susanne</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Jo Milner</surname><given-names>A</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib>

Breast Cancer Research : BCR

<sec><title>Introduction</title><p>The tumour-suppressor protein p53 is a nuclear transcription factor that is autoregulatory in terms of expression, this being low in normal cells. Loss of p53 function is often associated with high accumulation of p53 and its retention in the cytoplasm. Abnormal accumulation of p53 in breast cancer tissue is predictive for poor prognosis [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Humoral studies [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>] have shown that cancer patients may develop immunity to abnormally expressed p53, as revealed by p53 autoantibodies in the blood. Again, prognostic correlates have been noted, with presence of circulating p53 autoantibodies at diagnosis of breast cancer being associated with reduced overall survival [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>] and with poor prognostic factors such as high histological grade and the absence of hormone receptors [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. P53 dysfunction may be due to either mutational or epigenetic factors, each of which may lead to accumulation of cytoplasmic p53.</p><p>Little is known of the potential value of p53 autoantibody in follow-up of cancer. In lung cancer there is evidence that autoantibodies to p53 may provide a useful tool to monitor response to therapy [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>], whereas serial measurements of autoantibodies to p53 in 40 patients with advanced ovarian cancer were not found to be clinically useful [<xref ref-type="bibr" rid="B11">11</xref>]. In breast cancer, some 30% of node-negative patients will relapse within 5 years, but there is no current means to predict those who are at risk. We have asked if the presence of autoantibodies to p53 have any association with disease progression by testing plasma samples taken from 1006 patients with breast cancer with a median follow-up period of 4 years.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patient plasma samples</title><p>A library of plasma samples were collected from all patients attending one general oncology clinic for postoperative follow up of breast cancer; samples were taken without any patient selection and irrespective of clinical stage, menopausal status, histological type or degree of differentiation. The operative management of each patient was according to the tenets of their surgeons at the time of presentation; the breast was conserved wherever possible. The majority of patients were treated with adjuvant radio-therapy. Adjuvant cytotoxic chemotherapy, with or without tamoxifen, was the subject of clinical trial at the time.</p><p>The clinical status of each patient at time of sampling had been summarized on a database using Helix Express (version 1.0.1; Helix Technologies, Northbrook, IL, USA). The plasma library had been established over a 15-year period, with an average of eight plasma samples per patient being cryopreserved. It is known that cryopreserved p53 autoantibodies are stable in either sera or plasma, and that several cycles of freezing/thawing do not cause loss of titre (Soussi T, personal communication). Although it is theoretically possible that shed p53 protein might partially block autoantibody in some patients, it has already been shown [<xref ref-type="bibr" rid="B17">17</xref>] that autoantibody to p53 is normally in excess, and is thus detectable by ELISA.</p></sec><sec><title>P53 autoantibody detection by enzyme-linked immunosorbent assay</title><p>The ELISA assay for p53 autoantibodies was developed in-house, based on the ELISA procedure of Lubin <italic>et al</italic> [<xref ref-type="bibr" rid="B3">3</xref>] in which the amount of specific p53 autoantibody recorded by ELISA was confirmed by Western blot and immunoprecipitation analyses. Briefly, 96-well assay plates (Falcon 3912; Becton Dickinson and Co, Oxnard, CA, USA) were precoated with baculoviral-derived, purified, wild-type, full-length, human recombinant p53 (0.1 ng/ml phosphate-buffered saline [PBS]) using 30 &#x003bc;l per well. After overnight incubation at 4&#x000b0;C, the plates were washed three times with PBS-0.1% Tween at room temperature. Wells were then blocked for 2 h with 250 &#x003bc;l PBS-0.5% Tween. After five washes with PBS-0.1% Tween, 30 &#x003bc;l of each plasma sample was incubated on ice for 2 h. The plate was again washed five times as above before addition of horse radish peroxidase-linked goat antiserum to human immunoglobulin (Sigma A8667; Sigma Aldrich Co Ltd, Poole, Dorset, UK) and incubation on ice for 1 h. After a further five washes, bound horse radish peroxidase was detected by conventional methods using <italic>O</italic>-phenylenediamine dihydrochloride (30 min), stopping with 3 mol/l H₂SO₄, and reading the optical density (OD) at 492 nm.</p><p>An OD reading of less than 0.4 was taken as a negative result. This cutoff was determined by titration of positive and negative samples (1/10, 1/50, 1/100 and 1/500) and gave good discrimination between positive and negative samples at 1/500 dilution. All assay runs included the same internal standards of a sample known to be positive at 1/500 dilution and a sample known to be negative at 1/10 dilution, in order to confirm assay reproducibility throughout the series.</p><p>The intial screen of all patients was at a plasma dilution of 1/10 in PBS-0.1% Tween. Samples from all patients with a positive OD reading at the 1/10 dilution were then re-assayed at 1/500 (see Sample processing, below). The positive and negative internal standards ensured that all results were directly comparable. Background controls received no p53 protein. Specificity controls used soluble p53 protein added to the positive plasma sample before assay; this resulted in loss of signal.</p><p>In one assay series we compared a commercial ELISA kit for p53 autoantibodies (produced by Dianova GmbH; licenced to CalBiochem-Novabiochem Corp; distributed by Oncogene Research Products, Cambridge, MA, USA; cat no. QIA 16) with our in-house ELISA. A total of 20 patients' samples were tested, representing a range of positive and negative readings. Two samples were scored strongly positive by the in-house assay, but only one of these two scored positive by the Dianova assay. Having established sensitivity, specificity and reproducibility of the in-house assay, we judged that this was superior to the commercial assay both in terms of sensitivity and of cost (&#x000a3;1 per test compared with &#x000a3;23 per test). The in-house assay was thus used throughout the study presented here.</p></sec><sec><title>Sample processing</title><p>The last sample obtained from each of the 1006 patients was screened for anti-p53 antibodies at a 1/10 dilution. Those patients who proved to be positive for p53 autoantibodies at this low dilution were then tested for autoantibody status using all samples from each patient; here a 1/500 dilution of plasma was used because this was found to give good specificity and sensitivity for known positive and negative controls. Of those patients who were antibody negative in the initial screen, 60 had had a plasma sample obtained around the time of their primary diagnosis of breast cancer; these primary samples were assayed (1/10 dilution) to look for any positive to negative switches in autoantibody levels during clinical follow up.</p></sec><sec><title>Clinical fields</title><p>The presence or absence of autoantibodies to p53 was compared with the clinical status of each patient, which was classified: (i) primary remission; (ii) secondary remission following a previous relapse; (iii) active relapsed disease; or (iv) continuous active disease since first diagnosis. For those patients who were anti-p53 positive, their clinical history was compared with levels of measured p53 autoantibodies throughout their follow-up period.</p></sec></sec><sec><title>Results</title><p>Of the 1006 breast cancer patients screened, 155 (15%) had autoantibodies to p53 in the most recently obtained plasma sample. Of the total patient cohort, the disease status corresponding to the time of collection of this sample was known in 960 patients. Autoantibody status was compared with disease status (Table <xref ref-type="table" rid="T1">1</xref>). There was no correlation with those patient groups who had known active disease (either continuous active, or recurrent) or with those in primary or secondary clinical remission; the incidence of p53 autoantibody positivity was approximately 16% in all groups. The possibility that patient age may influence humoral immunity was tested by comparing premenopausal with postmenopausal patients; there was no correlation with the patients' menopausal status at diagnosis with autoantibodies to p53 (Table <xref ref-type="table" rid="T2">2</xref>).</p><p>For those patients whose most recent sample contained p53 autoantibodies, all previous samples were screened. This longitudinal review showed that autoantibody tended to be persistent. Although preoperative plasma samples were not available, some patients had been included in this study from the time of their first appointment at the oncology clinic; these samples were those that were most likely to contain any residual p53 autoantibody associated with the primary tumour. For those who had antibody present at early follow up, levels tended to persist throughout follow up. This was in contrast to those patients who had no autoantibody detectable within 30 days of surgery (see below). Overall, any fluctuations in autoantibody levels within the positive patient cohort gave no consistent pattern when compared with the clinical history of each patient. Although there was no correlation between p53 autoantibody status and disease behaviour, in one patient there was a strong correlation with prednisolone therapy and a fall in p53 autoantibodies, presumably as a result of steroid-induced immunosuppression. Shortly after steroid treatment, the patient developed cerebral metastases which were marked by a rapid rise in CA15.3.</p><p>Of the 851 patients who were negative for p53 autoantibodies at a 1/10 plasma dilution, 60 had had a plasma sample taken within 30 days of diagnosis of breast cancer. Of these, 22 had current active disease and 38 had current inactive disease, at the time when the autoantibody status of all 60 patients was negative. The first sample (taken less than 30 days after diagnosis) was screened and showed that 59 out of the 60 proved to have also been negative for p53 autoantibodies around the time of diagnosis (Table <xref ref-type="table" rid="T3">3</xref>). This suggests that, if a patient is negative for p53 autoantibodies at diagnosis of active primary disease, then that patient is highly unlikely to develop humoral immunity to p53, even in the presence of recurrent disease.</p></sec><sec><title>Discussion</title><p>This large, single-centre study was designed to explore a possible relationship between p53 autoantibody status and breast cancer activity. One thousand and six patients with breast cancer attending a single referral center were included in the study, irrespective of age or disease characteristics at the time of primary diagnosis. The median patient follow up was 4 years. An average of eight serial plasma samples per patient had been cryopreserved, and corresponding clinical information at the time of each sample's collection had been entered into the database. This fully documented library of over 8000 samples was used to look for correlates between tumour behaviour and autoantibodies to p53 during the clinical follow-up period.</p><sec><title>Anti-p53 levels were independent of changes in tumour status</title><p>As a working hypothesis we proposed that levels of autoantibodies to p53 would reflect tumour behaviour. Thus, for those patients who were positive for p53 autoantibodies at diagnosis, we reasoned that surgical removal of primary tumour might result in reduced p53 autoantibody levels. Should these levels then show a secondary increase associated with relapsed disease, then increasing levels of p53 autoantibodies might act as a biochemical marker of tumour progression. For those patients who were negative for p53 autoantibodies, then, development of recurrent disease may be associated with changes in p53 expression within the metastatic tumour, leading to a switch to an autoantibody-positive status.</p><p>We found that the presence or absence of p53 autoantibodies was not predictive for presence or absence of recurrent disease. There was an equivalent incidence of active disease at the time of sampling in the autoantibody-negative and autoantibody-positive groups, these being 25.2% and 28.7%, respectively.</p><p>We found that humoral immune activity against p53 appeared to be relatively restricted to a subgroup of patients in whom, once an autoantibody response had been generated, antibody was likely to persist regardless of tumour behaviour. Thus, antibody-positive patients without clinical recurrence remained antibody positive throughout the follow-up period. Conversely, where no detectable p53 autoantibody was present at the time of primary diagnosis, these patients remained similarly negative for antibody irrespective of subsequent disease activity (Table <xref ref-type="table" rid="T3">3</xref>).</p></sec><sec><title>Immune regulation and potential responsiveness to breast cancer</title><p>In contrast to shed markers that correlate with tumour mass, such as CA15.3 for cancer of the breast, any tumour-related immune response will be subject to complex regulation. Autoantibody responses to p53 will require appropriate primary immunization. Initial low-dose antigen exposure may induce immune tolerance and lack of response. Higher antigen doses may activate either antibody-mediated immunity, or cellular immunity.</p><p>In breast cancer patients, the present results suggest that, once an active humoral response against p53 is established, this response remains active. This persistent humoral reaction may be driven by persistent antigenic stimulation by p53 protein derived from overexpression of p53 at distant metastatic sites. Alternatively, irradiated normal tissue may be a source of continued antigenic stimulation, because a long-term side effect of radiation therapy is an increased expression of p53 in normal breast tissue, which persists for several years [<xref ref-type="bibr" rid="B12">12</xref>]. Since the great majority of our total patient cohort had received radiotherapy, humoral immunity to p53 associated with primary disease might persist, even in those patients who enter remission, due to tumour-independent antigenic stimulation.</p><p>Loss of p53 function is known to correlate with loss of efficacy of cancer therapy <italic>in vivo</italic> [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. This raised the possibility that autoantibodies to p53 that develop during follow up might indicate those patients whose tumor has become resistant to therapy. However, the present results show that if no immunity has been generated at the time of primary diagnosis, then later immunity is unlikely to occur. This corresponds to the finding that expression of p53 antigen in biopsies of locally advanced breast cancer did not correlate with drug resistance [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. Overall, our observations show that screening for p53 autoantibody status is not informative on residual tumour activity, nor on therapeutic responsiveness. We conclude that the potential value of p53 autoantibody screening in patients with breast cancer is limited to the prognostic information obtained at diagnosis.</p><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Incidence of autoantibodies to p53 compared with disease status at last clinic attendance</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Disease status</td><td align="center">% Anti-p53 positive</td></tr></thead><tbody><tr><td align="left">Primary remission</td><td align="center">14.7% (82/557)</td></tr><tr><td align="left">Secondary remission</td><td align="center">17.4% (27/155)</td></tr><tr><td align="left">Secondary recurrent</td><td align="center">19.4% (20/103)</td></tr><tr><td align="left">Continuous active</td><td align="center">16.5% (24/145)</td></tr><tr><td align="left">Total</td><td align="center">15.9% (153/96)^*</td></tr></tbody></table><table-wrap-foot><p>Pearson &#x003c7;²: <italic>P</italic> = 0.606. ^*Information on disease status at last sample time was available for 960 out of 1006 patients</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Incidence of autoantibodies to p53 compared with menopausal status at diagnosis</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Menopausal status</td><td align="center">Anti-p53 positive (%)</td></tr></thead><tbody><tr><td align="left">Premenopausal</td><td align="center">15.0% (104/693)</td></tr><tr><td align="left">Postmenopausal</td><td align="center">16.8% (51/313)</td></tr><tr><td align="left">Total</td><td align="center">15.4% (155/1006)</td></tr></tbody></table><table-wrap-foot><p>Pearson &#x003c7;²: <italic>P</italic> = 0.788.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T3"><label>Table 3</label><caption><p>Anti-p53 negative patients do not become positive with recurrent disease</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Current</td><td align="center">Anti-p53 status</td><td align="center">Current anti-p53</td></tr><tr><td align="left">disease status</td><td align="center">at diagnosis</td><td align="center">status</td></tr></thead><tbody><tr><td align="left">Nonactive disease</td><td align="center">38/38 negative</td><td align="center">38/38 negative</td></tr><tr><td align="left">Active disease</td><td align="center">21/22 negative</td><td align="center">22/22 negative</td></tr></tbody></table><table-wrap-foot><p>Sixty patients who were negative for p53 autoantibodies had also had a plasma sample taken within 30 days of their primary diagnosis of breast cancer. To determine whether antibody status at diagnosis might have been predictive of later disease behaviour (ie independent of the current negative status), we compared two patient subgroups: patients with current nonactive disease and patients with current active disease. With one exception, all patients were antibody negative within 30 days of initial diagnosis. This showed that recurrent disease is highly unlikely to induce a humoral anti-p53 response in those patients who were initially antibody negative.</p></table-wrap-foot></table-wrap></sec></sec>

Estrogen receptor of primary breast cancers: evidence for intracellular proteolysis

<sec><title>Introduction:</title><p>We previously reported that about two-thirds of [¹²⁵I]oestradiol-labelled cytosolic ERs from breast cancer samples eluted as low-molecular-weight isoforms (&#x02264; 37 kDa, size-exclusion fast pressure liquid chromatography [FPLC]). These isoforms failed to adsorb strongly to hydroxylapatite at high ionic strength, a property that was ascribed to receptors devoid of amino-terminal ABC domains. In view of recent data concerning intracellular proteolysis of several transcriptional regulators, the possibility of such behaviour for ER was assessed.</p><p>The clinical significance of ER measurement in breast cancer cytosols is well established; approximately 50% of ER-positive cases respond to endocrine therapy. Whether such a poor correlation is related to a high proportion of cleaved ER is a question of prime importance. Failure of routine ER assays to discriminate between full-length and cleaved receptors led us to develop an oestradiol-binding assay based on hydroxylapatite adsorption.</p><p>The aims of the present study were to demonstrate that hydroxylapatite adsorption assay easily identifies cleaved cytosolic ER forms and to assess the origin of such ER forms.</p></sec><sec><title>Method:</title><p>Breast cancer cytosols classified as ER-positive according to [³H]oestradiol-binding assay (dextran-coated charcoal [DCC]) were subjected to hydroxylapatite adsorption. ER isoforms covalently labeled with [¹²⁵I]tamoxifen aziridine (TAZ) released from this matrix with 0.5 mol/l KCl were subsequently immunoprecipitated with a panel of monoclonal antibodies raised against various domains of ER (H222 [E], H226 [C] or ER1D5 [AB]) before being subjected to SDS-gel electrophoresis.</p><p>Three approaches were used to identify the origins of the cleaved ER forms: potential truncated ER-&#x003b1; messenger RNAs that may encode ER isoforms of low molecular weights (Northern blot assay) were sought by using ER-&#x003b1; full-length probe; heat treatment of tumour cytosols in the absence or presence of a cocktail of protease inhibitors was performed; and the molecular weight of intracellular ER molecules was determined by <italic>in situ</italic> [¹²⁵I]TAZ-labelling, which minimizes ER proteolysis.</p><p>Breast cancer samples classified as ER-positive according to both biochemical (cytosolic DCC assay) and histochemical (ER1D5 monoclonal antibody) criteria were labelled with [³H]oestradiol and were subsequently subjected to hydroxylapatite adsorption. Hydroxylapatite extraction index (EI) is defined as a ratio of the specifically bound [³H]oestradiol released from the hydroxylapatite matrix with KCl to the total amount of the specifically bound [³H]oestradiol extracted successively with KCl and ethanol: EI= ([³H]oestradiol) [KCl] &#x000d7; 100/([³H]oestradiol) [KCl] + ([³H]oestradiol) [EtOH]. The EI was calculated for each cytosol in order to evaluate the amount of cleaved ER forms present. Persistence of adsorption ER to hydroxylapatite in the presence of KCl (low EI) and ER1D5 positivity established by immunohistochemistry are two independent criteria for the presence of amino-terminal ABC domains. We therefore assessed whether hydroxylapatite determinations performed on cytosols are related to immuno-histochemistry data.</p></sec><sec><title>Results:</title><p>Cytosol pools labelled with [¹²⁵I]TAZ gave different electrophoretic patterns depending on the nature of the anti-ER monoclonal antibody used in the immunoprecipitation step preceding electrophoresis. The carboxyl-terminal-specific antibody H222 precipitated all ER isoforms (full-length 67 kDa ER, and cleavage products of 50 and 37-28 kDa), whereas the amino-terminal-specific antibodies H226 and ER1D5 precipitated only the full-length and a partially truncated isoform. Adsorption of this labelled cytosol pool onto hydroxylapatite with subsequent KCl extraction yielded ER isoforms with molecular weights between 37 and 28 kDa when immunoprecipitation of the elutes was carried out using H222. The absence of these isoforms after exposure of the elutes to H226 or ER1D5 demonstrated truncation of these isoforms at a site(s) downstream of ABC domains.</p><p>Total RNA from 46 tumours was exposed to ER-&#x003b1; full-length probe (Northern blot). All tumours expressed a full-length 6.6-kb ER mRNA; small-sized isoforms were not recorded. A good correlation resulted when amounts of 6.6-kb ER mRNA estimated by densitometry were compared with corresponding [³H]oestradiol-binding capacities (DCC assay), thereby rejecting the concept that low-molecular-weight isoforms were encoded by truncated ER mRNA.</p><p>We next investigated whether such isoforms might be generated by proteolysis. Cytosol samples of a series of breast tumours were labelled with [¹²⁵I]TAZ in the presence of a cocktail of protease inhibitors. These inhibitors failed to maintain the full-length 67 kDa ER by SDS-PAGE. <italic>In situ</italic> [¹²⁵I]TAZ-labelling of receptors associated with a protein extraction procedure minimizing their proteolysis displayed multi-bands electrophoretic patterns, almost identical to those found under conventional methods. Hence, ER molecular heterogeneity appears to result from an intracellular proteolysis. ER1D5 immunostaining scores (ISs) of a series of 15 tumours were significantly correlated with ER levels, as measured by hydroxylapatite assay of corresponding cytosols (total number of binding sites). Sequential extraction of bound [³H]oestradiol from hydroxylapatite with KCl and ethanol revealed an EI of over 30% in the large majority of these cytosols, indicating a high frequency of cleaved ER isoforms. Of note, no significant correlation between IS and EI data was recorded, suggesting that ABC and E domains are separated at high ionic strength, but are apparently held together within the cell nucleus in oligomeric structures.</p></sec><sec><title>Discussion:</title><p>Endogenous proteolysis is a regulatory mechanism in many cellular processes, such as cell cycle progression and transcriptional regulation. The present data extend this concept to ER. Indeed, proteolysis-generated ER fragments appear to be held together within the cell in oligomeric structures. Because ER proteolysis is probably relevant to several oestrogen target tissues, we suggest that the protein environment, which differs among tissues, may be a factor of major importance in the formation of distinct oligomeric structures, which elicit specific biological responses. The possibility of heterogeneous association between cleaved ER and regulatory proteins might perhaps result in a spectrum of such transcriptional activities. In this context, we propose that a complementary hydroxylapatite extraction assay (EI assessment) should be added to the usual tests to identify ER-positive tumours. Such a complementary test would provide an estimate of the level of cleaved ER forms, which may have biological and/or clinical relevance.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Maaroufi</surname><given-names>Younes</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>lcanmamm@ulb.ac.be</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Lacroix</surname><given-names>Marc</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Lespagnard</surname><given-names>Laurence</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Journ&#x000e9;</surname><given-names>Fabrice</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Larsimont</surname><given-names>Denis</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Leclercq</surname><given-names>Guy</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Breast Cancer Research : BCR

<sec><title>Introduction</title><p>Assessment of the ER status in breast cancer samples is currently used to select patients for endocrine (tamoxifen) therapy [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Biochemical determinations of receptor concentration are based on the measurement of the tritiated oestradiol-binding capacity of cytosol samples; immunoenzymatic measurements (Abbott's ER enzyme immunoassay) usually give similar data because they use monoclonal antibodies against epitopes that are localized at both edges of the hormone-binding domain (E domain) [<xref ref-type="bibr" rid="B4">4</xref>]. Histochemical data established with various anti-ER monoclonal antibodies are in agreement with these biochemical assays, at least on a qualitative basis [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. Of note, histochemical determinations identify ER essentially in the cell nucleus, leading to the concept of a possible dissociation of nuclear oligomers with concomitant release of the receptor into the cytosol at the time of homogenization.</p><p>Approximately 50% of ER-positive breast cancers respond to endocrine therapy [<xref ref-type="bibr" rid="B8">8</xref>]. A defect in the synthesis and/or turnover of ER that leads to the emergence of altered receptors has been suggested as an explanation for this relatively poor correlation between treatment response and ER status [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. In agreement with this, we found that about two-thirds of [¹²⁵I]oestradiol-labelled cytosolic ERs elute as low-molecular-weight isoforms (&#x02264; 37 kDa, by size exclusion FPLC) that fail to adsorb strongly to hydroxylapatite at high ionic strength [<xref ref-type="bibr" rid="B11">11</xref>], a property ascribed to receptors that are devoid of amino-terminal ABC domains [<xref ref-type="bibr" rid="B12">12</xref>] required for transcription activity. ER proteolysis may lead to such a situation. Thus, a differential susceptibility of the hinge region (D domain) that separates such ABC domains from the hormone-binding domain to endoproteolytic cleavage has long been recognized [<xref ref-type="bibr" rid="B13">13</xref>]; whole hormone-binding domain itself also contains a few specific motifs that are accessible to proteolytic attacks [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. Of note, an oestradiol-inductible protease activity, which is detected in mouse reproductive tissues, has been reported to produce a cleaved ER that is devoid of its amino-terminal region, but which has a nuclear localization capacity [<xref ref-type="bibr" rid="B16">16</xref>]. Hence, cleaved ER isoforms may maintain a regulatory activity at the genomic level.</p><p>The relevance of the molecular heterogeneity of ER in breast cancer is still a matter of controversy [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. Our observation that a given ER cleavage pattern is relevant to the whole tumour mass (ie various samples taken from a given patient usually display identical radiolabelled SDS-PAGE profiles when their receptor isoforms were labelled with [³H]TAZ [<xref ref-type="bibr" rid="B19">19</xref>]) accredit the analysis of this question. Whether ER degradation occurs within the cell or is a product of homogenization or assay is a question of prime interest, in view of recent data regarding the biological significance of intracellular proteolysis of transcriptional regulators, such as the nuclear factor-&#x003ba;B, p53, c-JUN, sterol-regulated element-binding proteins and MAT&#x003b1;2 [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. Of note, activation of nuclear factor-&#x003ba;B requires proteasome-dependent proteolysis, leading to the processing of a precursor (nuclear factor-&#x003ba;B1) into the mature form [<xref ref-type="bibr" rid="B22">22</xref>]. Association of nuclear factor-&#x003ba;B with a specific inhibitory protein (I&#x003ba;B), which masks its nuclear localization signal, constitutes a functional system in which DNA binding/activation and regulatory/inhibitory functions are performed by two separate proteins that can interact [<xref ref-type="bibr" rid="B23">23</xref>]. Similar association of distinct functional domains may be applicable to cleaved ER isoforms, because the native form is also subject to proteasomal degradation [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>].</p><p>In the present paper we demonstrate that ER isoforms of low molecular weight could not be ascribed to the expression of truncated mRNAs, or to the proteolysis of the native peptide at the time of assay, suggesting a degradative mechanism at the intracellular level. Moreover, we establish systematic immunohistochemical identification of ER amino-terminal domains within the cell nucleus with an anti-ER monoclonal antibody (ER1D5) that is raised against this part of the receptor molecule. All of these findings support the concept of a potential intranuclear association between cleaved ER and other regulatory proteins.</p></sec><sec sec-type="materials|methods"><title>Material and methods</title><sec><title>Reagents</title><p>[³H]oestradiol (approximately 100 Ci/mmol) was obtained from Amersham (Buckinghamshire, UK); [¹²⁵I]TAZ (approximately 2000 Ci/mmol) was obtained from Trigone Ligands/Biocode (Li&#x000e8;ge, Belgium); and unlabelled oestradiol was obtained from Sigma (St Louis, MO, USA). Reagents for SDS-PAGE as well as Bio-Gel HTP hydroxylapatite were from Bio-Rad (Richmond, CA, USA); and H222 and H226 anti-ER monoclonal antibodies were a gift from Dr D Cotter (Abbott Laboratories, North Chicago, IL, USA). ER1D5 anti-ER monoclonal antibody was from Immunotech (Marseille-Luminy, France). Protease inhibitors (AEBSF, 4- [2-aminoethyl]-benzenesulphonyl fluoride; PMSF, phenylmethylsulphonylfluoride; antipain; chymostatin; leupeptin; and calpastatin] were from Euro Biochem (Bierges, Belgium).</p></sec><sec><title>Oestrogen receptor preparations</title><p>Primary breast cancer samples (for steroid hormone receptor measurements) were obtained from our surgery department. For ER measurement, they were homogenized in 10 mmol/l phosphate (pH 7.4) containing 1 mmol/l EDTA and 1 mmol/l monothioglycerol using a whole glass homogenizer. Homogenates were subsequently centrifuged for 1 h at 100 000 <bold><italic>g</italic></bold>. Supernatants were stored in liquid nitrogen until they were labelled, which was within the following week.</p><p>Human full-length ER produced in yeast [<xref ref-type="bibr" rid="B27">27</xref>] was provided by Dr P Sj&#x000f6;holm (Karo Bio, Huddinge, Sweden). This preparation was at a concentration of approximately 1 pmol/ml phosphate buffer (pH 7.5), and was stored at -70&#x000b0;C. For experiments, aliquots were diluted with 10 mmol/l Tris-HCl buffer (pH 8), containing bovine serum albumin (fraction V; Sigma) to reach a final protein concentration of approximately 1 mg/ml.</p><p>Cytosols from MCF-7 cells growing in monolayer culture were prepared as described previously [<xref ref-type="bibr" rid="B28">28</xref>].</p></sec><sec><title>Oestrogen receptor assays</title><sec><title>Dextran-coated charcoal assay</title><p>ER preparations were labelled with increasing concentrations of [³H]oestradiol (range 0.25-5 nmol/l) in the absence and presence of a 200-fold excess of unlabelled oestradiol (overnight incubation at 0&#x000b0;C) and unbound ligands were removed using the DCC adsorption method (0.5% charcoal, 0.05% dextran) [<xref ref-type="bibr" rid="B29">29</xref>]. Specific binding was calculated from the difference between radioactivity levels measured in the absence and presence of unlabelled oestradiol. Binding capacities were expressed per milligram of protein, the latter being measured using the Bio-Rad assay. Parameters of the binding reaction (<italic>n</italic>, number of binding sites; K<italic>d</italic>, dissociation constant) were estimated using the LIGAND v4.5 program (Dr PJ Munson, National Institutes of Health, Bethesda, MD, USA), and the data were plotted according to the method of Scatchard [<xref ref-type="bibr" rid="B30">30</xref>]. The same procedure was applied for [¹²⁵I]TAZ, except that a 5 nmol/l saturating concentration was used.</p></sec><sec><title>Hydroxylapatite assay and extraction index</title><p>ER preparations adsorbed on hydroxylapatite pellets (prepared in 10 mmol/l Tris-HCl buffer, pH 8) were labelled with increasing concentrations of [³H]oestradiol (range 0.25-5 nmol/l) in the presence and absence of a 200-fold excess of unlabelled oestradiol (overnight incubation at 0&#x000b0;C) [<xref ref-type="bibr" rid="B12">12</xref>]. Bound [³H]oestradiol complexes were then successively extracted with 0.5 mol/l KCl or absolute ethanol. Specific binding was calculated from the difference between radioactivity levels measured in the absence and presence of unlabelled oestradiol and the data were analyzed according to the method of Scatchard [<xref ref-type="bibr" rid="B30">30</xref>]. EI was defined as a ratio of the specifically bound [³H]oestradiol released with KCl to the total amount of the specifically bound [³H]oestradiol extracted successively with KCl and ethanol: EI= ([³H]oestradiol) [KCl]&#x000d7;100/([³H]oestradiol) [KCl]+ ([³H]oestradiol) [EtOH]. High values of EI were considered to indicate the absence of ABC ER domains [<xref ref-type="bibr" rid="B11">11</xref>].</p><p>The same procedure was applied for [¹²⁵I]TAZ, except that a 5 nmol/l saturating concentration was used. Because ethanol was unable to elute the covalently bound [¹²⁵I]TAZ from the ER in the final step of the conventional hydroxylapatite assay, extraction of labelled receptor was carried out with 0.5 mol/l phosphate. Bound [³H]oestradiol was measured with Ecoscint H scintillation fluid (National Diagnostics, Atlanta, GA, USA) in a Wallac 1409 liquid scintillation counter (Wallac Oy, Turku, Finland) with 45-60% counting efficiency; bound [¹²⁵I]TAZ was measured in a &#x003b3;-detector (Crystal 5400 Series, United Technologies Packard, IL, USA).</p></sec></sec><sec><title>Heat treatment of breast tumour cytosols</title><p>Tumour cytosols were heated at 37&#x000b0;C for 2 min in the presence or absence of a cocktail of protease inhibitors (final concentrations: 1 mmol/l for AEBSF, antipain and chymostatin; 5 mmol/l for PMSF; 0.1 mmol/l for leupeptin; and 0.1 mg/ml for calpastatin). After treatment, samples were labelled with 1 nmol/l [¹²⁵I]TAZ (1 h at 0&#x000b0;C) in the presence or absence of a 200-fold excess of radioinert oestradiol, immunoprecipitated with H222 anti-ER monoclonal antibody, and then analyzed using SDS-PAGE.</p></sec><sec><title>SDS-PAGE analysis for cytosolic oestrogen receptors</title><p>ER preparations (cytosol/KCl extract from hydroxylapatite) were covalently labelled with 1 nmol/l [¹²⁵I]TAZ (1 h incubation at 0&#x000b0;C) in the absence or presence of a 200-fold excess of unlabelled oestradiol; unbound ligand was then removed with DCC treatment. Labelled ERs were subsequently immunoprecipitated with H222, H226, or ER1D5 anti-ER monoclonal antibodies, and then antirat IgG agarose was added. Resulting suspensions were centrifuged and the pellets washed before being solubilized in lysis buffer (4% SDS, 20% glycerol, 10% mercaptoethanol, 0.05% bromophenol blue in 500 mmol/l Tris-HCl; pH 6.8). Lysates were subjected to electrophoresis on a 10% polyacrylamide gel in buffer (25 mmol/l Tris-HCl, 0.1% SDS, 192 mmol/l glycine; pH 8.2). Gels were stained, dried and finally exposed to Kodak X-OMAT films (Eastman Kodak Cy, Rochester, NY, USA) for 1 day for autoradiographical visualization of radioactive bands [<xref ref-type="bibr" rid="B19">19</xref>]. The molecular weights of the labelled peptides were estimated according to the migration of 14-94 kDa protein standards (Amersham Pharmacia Biotech, Uppsala, Sweden).</p></sec><sec><title>SDS-PAGE analysis for tissue oestrogen receptors</title><p>Sliced breast cancer tissues (thickness varied among samples, but was always less than 0.5 mm) were incubated for 1 h at 37&#x000b0;C with [¹²⁵I]TAZ (final concentration 1 nmol/l) in Krebs-Ringer phosphate buffer (pH 7.4), with 2% bovine albumin and 0.25% glucose [<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>]. Then, all tissues were washed with Krebs-Ringer phosphate buffer and mixed with the same solution, but containing 1% SDS, 1.6 mmol/l EDTA and 2% &#x003b2;-mercaptoethanol.Tissue slices were briefly and delicately homogenized, and heated to 100&#x000b0;C for 2 min. Protein extraction was carried out by phenol and precipitation by acetone containing 0.1 mol/l acetic acid [<xref ref-type="bibr" rid="B32">32</xref>]. After 2 h at -20&#x000b0;C, the extracts were centrifuged. After washing, the pellets were solubilized in lysis buffer and subjected to electrophoresis as described above.</p></sec><sec><title>Oestrogen receptor mRNA measurement</title><p>Total RNA was extracted with TRIzol reagent (Boehringer-Mannheim, Mannheim, Germany), dissolved in RNase-free water and quantified by spectrophotometry at 260-280 nm. Aliquots of 30 &#x003bc;g RNA/15 &#x003bc;l were electrophoresed through a 1% agarose formaldehyde gel, capillary transferred to a Hybond-N membrane (Amersham), and treated according to the manufacturer's instructions. An EcoRI fragment (1300 pb) of pOR3 used as an ER-&#x003b1; mRNA probe was from the American Type Culture Collection (Rockville, MD, USA). Blots were hybridized sequentially with [³²P]-labelled ER cDNA probe (10⁹ cpm/mg cDNA, produced by random priming [Boehringer Mannheim]). Prehybridization (4 h) and hybridization (18 h) were performed as described previously [<xref ref-type="bibr" rid="B33">33</xref>]. The membranes were then washed with sodium citrate solutions of increasing stringency, the last wash being performed in 0.3 &#x000d7; sodium citrate solutions containing 0.1% SDS. Blots were visualized by exposure of the membranes for 1 day to Kodak XAR-5 film in an autoradiography cassette with an intensifying screen.</p></sec><sec><title>Immunohistochemistry</title><p>Breast cancer sections submitted to formalin fixation and paraffin embedding were immunostained with ER1D5 monoclonal antibody (Immunotech, Marseille-Luminy, France) using microwave pretreatment for antigen retrieval. Briefly, the sections were soaked in buffer (sodium citrate 10 mmol/l; pH 6) and heated for three cycles, each for 5 min at 600 W. After microwave treatment, sections were allowed to cool in room temperature and incubated with ER1D5 (used at 1:150 dilution) for 1 h. Sections were processed using a multilink supersensitive streptavidin-biotin detection system (Biogenex, San Ramon, CA, USA), employing diaminobenzidine as chromogen and haematoxylin counterstaining.</p><p>The nuclear staining was classified in intense, moderate, weak or nonstained categories. The immunostaining score (IS) was based on the staining intensity among these categories and on the percentage of stained cells. The IS was defined as a weighted sum of the percentages in the following manner: IS= 4a + 2b + c, where a is the percentage of intensely stained cells, b is the percentage of moderately stained cells and c is the percentage of weakly stained cells [<xref ref-type="bibr" rid="B34">34</xref>]. This weighted distribution of positive cells is based on the previously demonstrated linear relationship between the subjectively determined optical density and the concentration of histochemical concentration product measured by microspectrophotometry [<xref ref-type="bibr" rid="B35">35</xref>].</p></sec></sec><sec><title>Results</title><sec><title>Adsorption of oestrogen isoforms onto hydroxylapatite at low ionic strength</title><p>Previous studies from our laboratory [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B36">36</xref>] revealed that ER preparations (human ER expressed in yeast, cytosols from MCF-7 breast cancer cells and uterus) adsorbed onto hydroxylapaatite display a lower [³H]oestradiol-binding capacity than those provided by the conventional DCC assay; interference of the phosphocalcic matrix of the hydroxylapatite with ligand binding was advocated to explain this discrepancy. Assessment of cytosols from human primary breast tumours revealed that this property also holds for their receptors.</p><p>In 97 cases out of a series of 102 cytosols, assessment of the binding parameters of [³H]oestradiol (multipoint Scatchard plot analysis) gave K_d values of the same order of magnitude for both assays (median Kd values of 0.30 for DCC and 0.40 nmol/l for hydroxylapatite; Table <xref ref-type="table" rid="T1">1</xref>). Binding capacities established by these two methods were also significantly correlated (<italic>r</italic> = 0.79; <italic>P</italic> &#x0003c; 0.001), with systematically higher values for the DCC assay (slope of the regression line= 1.47; Fig <xref ref-type="fig" rid="F1">1</xref>). The remaining five outlayers bound [³H]oestradiol with a relatively lower affinity (Kd about ten times higher; Table <xref ref-type="table" rid="T1">1</xref>). These outliers were also characterized by a weak [³H]oestradiol-binding capacity by the DCC assay and a wide range of values by the hydroxylapatite procedure (Fig. <xref ref-type="fig" rid="F1">1</xref>; inset). They did not differ from the others with regard to their protein contents, thus refuting the hypothesis of false-negative or false-positive cases associated with a low or high protein level [<xref ref-type="bibr" rid="B37">37</xref>]. The nature of these low affinity-binding sites remains unknown.</p><p>Using [¹²⁵I]TAZ instead of [³H]oestradiol as the labelling agent in another series of 36 cytosols, we established a similar correlation (<italic>r</italic> = 0.83; <italic>P</italic> &#x0003c; 0.001; Fig. <xref ref-type="fig" rid="F2">2</xref>). The slope of the regression line was similar to that established with [³H]oestradiol (1.40), indicating a similar interference of the phosphocalcic matrix of the hydroxylapatite in the binding of [¹²⁵I]TAZ. Hence, at low ionic strength, ligand-binding capacities of cytosolic ER from breast tumours are significantly correlated when assessed by DCC and hydroxylapatite assays.</p></sec><sec><title>Release of cleaved oestrogen receptor isoforms from hydroxylapatite with KCl</title><p>Data reported above suggest that the ER domains that are required for hydroxylapatite adsorption (ABC domains) and ligand-binding (E domain) are present in most tumour cytosols. In view of the high frequency of cleaved ER isoforms in breast cancer [<xref ref-type="bibr" rid="B11">11</xref>], such domains should be dissociated at high ionic strength, as demonstrated here using [¹²⁵I]TAZ as a labelling agent.</p><p>ER isoforms labelled with radiolabelled TAZ were easily detected by successive immunoprecipitation and SDS-PAGE (full-length 67 kDa and cleavage products of 50 and 37-28 kDa) [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B38">38</xref>,<xref ref-type="bibr" rid="B39">39</xref>]. Using this approach we observed that a pool of breast cancer cytosols labelled with [¹²⁵I]TAZ gave different electrophoretic patterns, depending on the nature of the anti-ER monoclonal antibody used in the immunoprecipitation step preceding electrophoresis (H222, H226, ER1D5; Fig. <xref ref-type="fig" rid="F3">3a</xref>, left). The carboxyl-terminal-specific antibody H222 precipitated all ER isoforms, whereas the amino-terminal-specific antibodies H226 and ER1D5 precipitated only the full-length and a partially truncated 50 kDa isoform, indicating a lack of corresponding antigenic sites in the 37-28 kDa isoforms. Of note, the 50 kDa band was more intensely labelled with H226. The [¹²⁵I]TAZ-labelling intensity of all of these bands was suppressed with a 200-fold excess of unlabelled oestradiol, establishing their specificity.</p><p>Part of this [¹²⁵I]TAZ-labelled cytosol pool was adsorbed onto hydroxylapatite and subsequently subjected to KCl extraction (Fig. <xref ref-type="fig" rid="F3">3a</xref>; right). Elutes were then immunoprecipitated before being subjected to SDS-PAGE. Absence of 67 and 50 kDa isoforms under all immunoprecipitation conditions confirmed their adherence onto hydroxylapatite due to their strong interaction (ABC domains) with the matrix. On the contrary, and as expected, ER isoforms with a molecular weight of between 37 and 28 kDa were detected in the elutes when the immunoprecipitation was carried out by H222; their absence after exposure to H226 or ER1D5 confirmed the cleavage of these isoforms at a site(s) downstream of ABC domains. Hence, hydroxylapatite extraction assay easily identifies ER isoforms that lack amino-termini. Figure <xref ref-type="fig" rid="F3">3b</xref> illustrates the presumed structure of these various ER isoforms, as well as their sizes as determined by SDS-PAGE.</p></sec><sec><title>Assessment of oestrogen receptor mRNA size in breast cancer</title><p>Total RMA from 46 breast tumours was qualitatively and quantitatively analyzed by hybridization with an ER-&#x003b1; full-length probe (Northern blotting). All tumours expressed a full-length 6.6-kb ER mRNA (small-sized species were not recorded). Moreover, a good correlation was obtained when the amount of 6.6-kb ER mRNA estimated by densitometry was compared with corresponding [³H]oestradiol-binding capacities (DCC values; Fig. <xref ref-type="fig" rid="F4">4</xref>). Hence, ER isoforms of low molecular weight did not appear to be encoded by truncated ER mRNAs, suggesting that they were generated by proteolysis. Whether such a phenomenon is an intracellular process is analyzed below.</p></sec><sec><title>Origin of cleaved oestrogen receptor forms</title><p>Cytosol samples of a series of breast tumours were labelled with [¹²⁵I]TAZ in the presence of a cocktail of compounds that are known to inhibit the action of a wide range of proteolytic activities (final concentrations: 1 mmol/l for AEBSF, antipain and chymostatin; 5 mmol/l for PMSF; 0.1 mmol/l for leupeptin; and 0.1 mg/ml for calpastatin); this cocktail was added before or shortly after the homogenization of the samples. These inhibitors failed to maintain the native 67 kDa nature of the receptor as demonstrated by SDS-PAGE and autoradiography (Fig. <xref ref-type="fig" rid="F5">5</xref>); the ER electrophoretic pattern was not significantly modified, still showing bands of low molecular weight.</p><p>In order to determine the molecular weight of intracellular ER molecules, we applied an <italic>in situ</italic> labelling approach along with an extraction procedure to minimize ER proteolysis [<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>]; tumour slices were incubated with [¹²⁵I]TAZ at 37&#x000b0;C before homogenization and lysis at 100&#x000b0;C in the presence of SDS. Labelled proteins extracted with phenolic solution were precipitated by acetone and finally identified by SDS-PAGE. ER electrophoretic patterns remained almost identical to those usually found with cytosols, still showing bands of truncated receptors (mainly 50 and 37-28 kDa; Fig. <xref ref-type="fig" rid="F6">6</xref>). Hence, the high proportion of the low-molecular-weight isoforms appeared already to be present within tumour samples.</p><p>Finally, breast cancer cytosols were heated at 37&#x000b0;C for 2 min in the absence or presence of a cocktail of protease inhibitors in order to determine whether they possess proteolytic activities that are able to cleave native ER. Samples were subsequently adsorbed onto hydroxylapatite and successively subjected to KCl and ethanol extraction. No significant increase in EI was recorded (Table <xref ref-type="table" rid="T2">2</xref>); inhibitors also failed to induce any drastic change in EI values, refuting the hypothesis that isoforms devoid of ABC domains may emerge during assay procedures. Of note, treatment at 37&#x000b0;C of control preparations of native ER (human recombinant ER, MCF-7 cells) resulted in similar behaviour. Hence, major proteolysis did not occur at the time of manipulation. Taken together, these data suggest that dominant ER proteolytic cleavage is an intracellular process.</p></sec><sec><title>Hydroxylapatite extraction index and heterogeneity of oestrogen receptors</title><p>Data reported here clearly show that low-molecular-weight ER isoforms extracted from hydroxylapatite matrix with KCl were not recognized by the ER1D5 monoclonal antibody. Because this antibody is often used in immunohistochemical assessment of ERs, we assessed whether immunohistochemical data are related to hydroxylapatite ER adsorption characteristics measured in cytosolic preparations from the corresponding tumours. For this purpose, cytosols from a set of 15 ER-positive tumours (by DCC assay), for which nuclear ERs had been detected by immunohistochemistry (IS cutoff &#x02265; 5), were labelled with [³H]oestradiol and were then subjected to hydroxylapatite assay (Table <xref ref-type="table" rid="T3">3</xref>). A significant correlation between the two sets of measurement was recorded (IS versus total number of binding sites assayed by hydroxylapatite, <italic>r</italic> = 0.71; <italic>P</italic> &#x0003c; 0.001; Table <xref ref-type="table" rid="T3">3</xref>). Sequential extraction of bound [³H]oestradiol from hydroxylapatite with KCl and ethanol revealed an EI of over 30% in the large majority of these cytosols (11/15), indicating a high frequency of cleaved ER. Of note, no significant correlation between IS and EI data (<italic>r</italic> = 0.2; <italic>P</italic> &#x0003e; 0.05) was detected, clearly establishing that identification of ABC domains within the cell (indicated by IS) does not imply the presence of (native) full-length ER in the corresponding cytosol.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Comparison of [³H]oestradiol (E₂)-binding capacities of a series of human breast cancer cytosols simultaneously measured by DCC and hydroxylapatite (HAP) assays (ethanolic extraction). The ordinate corresponds to values established by DCC, and the abscissa to those by hydroxylapatite (data established by Scatchard plot analysis). A significant correlation was obtained between the two assays in 97 out of 102 (95%) samples; five outlayers are represented by closed symbols in the insert.</p></caption><graphic xlink:href="bcr-2-6-444-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>Comparison of [¹²⁵I]TAZ-binding capacities (5 nmol/l) of a series of human breast cancer cytosols simultaneously measured by DCC and hydroxylapatite (HAP) assays (phosphate extraction).</p></caption><graphic xlink:href="bcr-2-6-444-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p><bold>(A)</bold> Molecular weight and monoclonals anti-ER recognition of [¹²⁵I]TAZ-labelled ER isoforms extracted from hydroxylapatite (HAP) with KCl. (left) Part of human breast cancer cytosol pool, after labelling with 1 nmol/l [¹²⁵I]TAZ for 1 h at 0&#x000b0;C in the presence or absence of a 200-fold excess of radioinert oestradiol, was immunoprecipitated with H222, H226 or ER1D5 anti-ER monoclonal antibodies, and then analyzed by SDS-PAGE and autoradiography. (right) Another part of this cytosol pool was adsorbed onto HAP, labelled with 1 nmol/l [¹²⁵I]TAZ, extracted with 0.5 mol/l KCl, and immunoprecipitated before being subjected to electrophoresis. <bold>(B)</bold> Presumed structure of ER isoforms extracted from HAP with KCl. Potential sites of covalent attachment of TAZ [<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref>] are indicated by open circles; antigenic sites for anti-ER monoclonal antibodies are shown above ER structure. The predicted ER isoforms extracted from HAP as well as their sizes determined by SDS-PAGE are shown below.</p></caption><graphic xlink:href="bcr-2-6-444-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>Correlation between ER mRNA and ER (by DCC assay). Breast tumours mRNA abundance (intensity of the 6.6-kb band) was expressed relatively to mRNA levels in MCF-7 cells (MCF-7 = 1).</p></caption><graphic xlink:href="bcr-2-6-444-4"/></fig><fig position="float" id="F5"><label>Figure 5</label><caption><p>Effect of heat treatment on the relative expression of ER isoforms. Human breast cancer cytosols were heated at 37&#x000b0;C for 2 min in the presence or absence of protease inhibitors. They were then labelled with [¹²⁵I]TAZ in the presence or absence of a 200-fold excess of radioinert oestradiol, immunoprecipitated with H-222 anti-ER monoclonal antibodies, and then subjected to SDS-PAGE. Lane 1, unheated control; lane 2, plus an excess of oestradiol; lane 3, 2 min heating in the absence of protease inhibitors; lane 4, 2 min heating in the presence of protease inhibitors.</p></caption><graphic xlink:href="bcr-2-6-444-5"/></fig><fig position="float" id="F6"><label>Figure 6</label><caption><p><italic>In situ</italic> labelling of ER with [¹²⁵I]TAZ. Breast tissues slices (samples 1-4, ER positive; sample 5, ER-negative) were incubated with 1 nmol/l [¹²⁵I]TAZ for 1 h at 0&#x000b0;C and the unbound ligand was removed. Then, all tissues were mixed with Krebs-Ringer phosphate buffer containing 1% SDS, 1.6 mmol/l EDTA and 2% &#x003b2;-mercaptoethanol, and briefly homogenized. After lysis at 100&#x000b0;C, proteins were extracted with phenol, precipitated by acetone and were finally analyzed by SDS-PAGE. The figure shows the electrophoretic patterns of these tissue ERs.</p></caption><graphic xlink:href="bcr-2-6-444-6"/></fig><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Comparison of K_d values estimated by DCC and hydroxylapatite (HAP) assays in the whole series of human breast cancer cytosols(102 cases) as well as in the five cases not included in the correlation</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center" colspan="2">102 Cases</td><td align="center" colspan="2">97 Cases</td><td align="center" colspan="2">5 Cases</td></tr><tr><td></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td></tr><tr><td></td><td align="center">Range</td><td align="center">Median</td><td align="center">Range</td><td align="center">Median</td><td align="center">Range</td><td align="center">Median</td></tr></thead><tbody><tr><td align="left" colspan="7">Dissociation constants (nmol/l)</td></tr><tr><td align="center">DCC</td><td align="center">0.001-02.74</td><td align="center">0.31</td><td align="center">0.001-2.74</td><td align="center">0.30</td><td align="center">0.010-02.70</td><td align="center">0.45</td></tr><tr><td align="center">HAP</td><td align="center">0.001-17.90</td><td align="center">0.42</td><td align="center">0.001-2.92</td><td align="center">0.40</td><td align="center">0.080-17.90</td><td align="center">4.26</td></tr><tr><td align="left" colspan="7">Protein (mg/ml)</td></tr><tr><td></td><td align="center">0.80-11.00</td><td align="center">4.52</td><td align="center">0.80-11.00</td><td align="center">4.41</td><td align="center">0.80-10.11</td><td align="center">4.01</td></tr></tbody></table></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Comparison of EIs of adsorbed ER onto hydroxylapatite matrix in the absence and presence of protease inhibitors</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center" colspan="4">EI (%)</td></tr><tr><td></td><td colspan="4"><hr></hr></td></tr><tr><td></td><td align="center" colspan="2">Without protease inhibitors</td><td align="center" colspan="2">With protease inhibitors</td></tr><tr><td></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td></tr><tr><td align="left">ER</td><td align="center">0&#x000b0;C</td><td align="center">37&#x000b0;C</td><td align="center">0&#x000b0;C</td><td align="center">37&#x000b0;C</td></tr></thead><tbody><tr><td align="left">Cytosol from tumors</td><td align="center">69 &#x000b1; 9</td><td align="center">82 &#x000b1; 16</td><td align="center">61 &#x000b1; 7</td><td align="center">59 &#x000b1; 13</td></tr><tr><td align="left">Recombinant hER</td><td align="center">&#x0003c;10</td><td align="center">&#x0003c;10</td><td align="center">&#x0003c;10</td><td align="center">12 &#x000b1; 5</td></tr><tr><td align="left">MCF-7 cells</td><td align="center">&#x0003c;10</td><td align="center">11 &#x000b1; 2</td><td align="center">13 &#x000b1; 02</td><td align="center">16 &#x000b1; 4</td></tr></tbody></table><table-wrap-foot><p>hER, human full-length ER.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T3"><label>Table 3</label><caption><p>Comparative analysis of ER by immunohistochemistry (ER1D5), DCC ([³H]oestradiol-binding) and hydroxylapatite ([³H]oestradiol-binding and EI)</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center">Tissular ER</td><td align="center" colspan="3">Cytosolic ER</td></tr><tr><td></td><td><hr></hr></td><td colspan="3"><hr></hr></td></tr><tr><td></td><td align="center">IHC</td><td align="center">DCC</td><td align="center" colspan="2">Hydroxylapatite</td></tr><tr><td></td><td><hr></hr></td><td><hr></hr></td><td colspan="2"><hr></hr></td></tr><tr><td align="left">Case no.</td><td align="center">IS</td><td align="center"><italic>n</italic> (fmol/mg protein)</td><td align="center"><italic>n</italic> (fmol/mg protein)^*</td><td align="center">EI (%)^&#x02020;</td></tr></thead><tbody><tr><td align="left">1</td><td align="center">10</td><td align="center">39</td><td align="center">27 (20 + 7)</td><td align="center">74.5</td></tr><tr><td align="left">2</td><td align="center">20</td><td align="center">69</td><td align="center">62 (54 + 8)</td><td align="center">86.8</td></tr><tr><td align="left">3</td><td align="center">20</td><td align="center">32</td><td align="center">67 (21 + 46)</td><td align="center">31.5</td></tr><tr><td align="left">4</td><td align="center">40</td><td align="center">38</td><td align="center">35 (12 + 23)</td><td align="center">34.4</td></tr><tr><td align="left">5</td><td align="center">130</td><td align="center">23</td><td align="center">130 (91 + 39)</td><td align="center">70.0</td></tr><tr><td align="left">6</td><td align="center">130</td><td align="center">83</td><td align="center">151 (62 + 89)</td><td align="center">41.1</td></tr><tr><td align="left">7</td><td align="center">160</td><td align="center">163</td><td align="center">166 (30 + 136)</td><td align="center">18.0</td></tr><tr><td align="left">8</td><td align="center">160-200</td><td align="center">257</td><td align="center">209 (58 + 151)</td><td align="center">27.5</td></tr><tr><td align="left">9</td><td align="center">260</td><td align="center">23</td><td align="center">103 (64 + 39)</td><td align="center">62.1</td></tr><tr><td align="left">10</td><td align="center">360</td><td align="center">265</td><td align="center">230 (114 + 116)</td><td align="center">49.6</td></tr><tr><td align="left">11</td><td align="center">360-400</td><td align="center">400</td><td align="center">360 (190 + 170)</td><td align="center">52.6</td></tr><tr><td align="left">12</td><td align="center">360-400</td><td align="center">354</td><td align="center">337 (82 + 255)</td><td align="center">24.5</td></tr><tr><td align="left">13</td><td align="center">360-400</td><td align="center">178</td><td align="center">193 (60 + 133)</td><td align="center">31.0</td></tr><tr><td align="left">14</td><td align="center">360-400</td><td align="center">92</td><td align="center">174 (36 + 138)</td><td align="center">20.8</td></tr><tr><td align="left">15</td><td align="center">360-400</td><td align="center">312</td><td align="center">390 (156 + 234)</td><td align="center">40.0</td></tr></tbody></table><table-wrap-foot><p>^*Hydoxylapatite corresponds specifically to [3H]oestradiol extracted successively from hydroxylapatite with KCl and ethanol (ie 27 = 20 + 7). ^&#x02020;EI = ([³H] oestradiol) [KCl] &#x000d7; 100/([³H]oestradiol) [KCl] + ([³H]oestradiol) [EtOH]. IHC, immunohistochemistry; IS, immunostaining score.</p></table-wrap-foot></table-wrap></sec></sec><sec><title>Discussion</title><p>Cytosolic ER contains several sites that are sensitive to proteolytic cleavage. On limited digestion, ER fragments of 50 and 37 kDa and lower molecular weight were recorded in uterine cytosol preparations [<xref ref-type="bibr" rid="B42">42</xref>]. The presence of such isoforms has generally been attributed to proteolysis of the native ER [<xref ref-type="bibr" rid="B43">43</xref>,<xref ref-type="bibr" rid="B44">44</xref>]. Regarding ER prepared from breast cancer samples, we previously suggested [<xref ref-type="bibr" rid="B11">11</xref>] that proteolytic events that operate at the time of tissue processing might explain the observed extreme ER molecular heterogeneity. Data presented here clearly refute this hypothesis. A high proportion of these isoforms is derived from an intracellular proteolytic activity, as shown by electrophoresis of <italic>in situ</italic> [¹²⁵I]TAZ-labelled ER and addition of protease inhibitors to cytosols. In this regard, the present results are in agreement with those reported recently by Navarro <italic>et al</italic> [<xref ref-type="bibr" rid="B45">45</xref>] who demonstrated that the inclusion of PMSF, aprotinin and leupeptin to human uterine cytosolic ER did not prevent the 8S to 4S transformation known to be associated with ER cleavage [<xref ref-type="bibr" rid="B46">46</xref>]. Hence, truncated ER isoforms are probably produced before contact with protease inhibitors.</p><p>On breast tumour slices, staining intensities (ISs) established with the ER1D5 monoclonal antibody used here have often been reported to be correlated with [³H]oestradiol binding capacities of corresponding cytosolic extracts [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B47">47</xref>]. Hence, a consensus seems to exist between immunohistochemical detection of the amino-terminal region of ER (ABC domains) and biochemical measurement of its carboxyl-terminal region (E domain). In agreement with this, we reported a significant correlation between immunohistochemical data and [³H]oestradiol-binding capacities measured by hydroxylapatite assay; ABC domains are indeed required for hydroxylapatite adsorption [<xref ref-type="bibr" rid="B12">12</xref>]. For both [³H]oestradiol- and [¹²⁵I]TAZ labelled-ER isoforms, we found an excellent correlation between DCC values and hydroxylapatite adsorption data, suggesting the concomitant presence of free ABC and E domains in most cytosols (proteolytic fragments physically separated from each other at high ionic strength; high EI values). Hence, these domains would be held together within the cell nucleus because of the tendency of the receptor to be included into oligomeric structures. Corroborating these views of intracellular association between ER fragments, Fritsch <italic>et al</italic> [<xref ref-type="bibr" rid="B42">42</xref>] described large oligomeric complexes containing proteolysis-generated ER fragments in rat uterus.</p><p>When expressed individually as separate polypeptides in ER-negative breast cancer cells (MDA-MB-231), neither ABCD (AF-1 containing) nor EF (AF-2 containing) domains activated transcription from a hormone-inducible reporter gene (<italic>3ERE-pS2-CAT</italic>) [<xref ref-type="bibr" rid="B48">48</xref>]. Coexpression of the steroid receptor coactivator-1 protein with either ABCD or EF alone was also ineffective [<xref ref-type="bibr" rid="B49">49</xref>]. On the contrary, when expressed together, amino- and carboxyl-terminal ER regions interacted in an oestradiol-dependent manner to reconstitute ER transcriptional activity [<xref ref-type="bibr" rid="B48">48</xref>]. In this context, it should be emphasized that steroid receptor coactivator-1 has also been suggested to promote the association between separate amino- and carboxyl-terminal regions of ER, allowing full ER activation [<xref ref-type="bibr" rid="B49">49</xref>]. In view of these data, one may speculate that the secondary and tertiary structures of the receptor peptide required for its transcriptional ability would not be lost when it has been subjected to endogenous proteolytic activity. Hence, intracellular ER fragmentation should not necessarily imply lack of transcriptional activity.</p><p>Endogenous proteolysis is a mechanism of regulation of many cellular processes, such as cell cycle progression and transcriptional regulation [<xref ref-type="bibr" rid="B50">50</xref>,<xref ref-type="bibr" rid="B51">51</xref>]. Recent data concerning cell-specific regulation of oestrogen target gene expression in various rat tissues [<xref ref-type="bibr" rid="B52">52</xref>] and cell lines [<xref ref-type="bibr" rid="B53">53</xref>] demonstrated different levels of coactivators and corepressors. Hence, protein environment would be a factor of major importance in the ability of ER to transcribe genes. Although not presently demonstrated in breast cancer, the possibility of heterogeneous association between cleavage products of ER and other regulatory proteins could be proposed, generating a situation that may mimic the behaviour of chimaeric receptors. Thus, domain-swapping experiments in which the ER amino-terminal domain was switched with that of the glucocorticoid receptor yielded a receptor that upregulates transcription of glucocorticoid-responsive target genes when treated with oestradiol [<xref ref-type="bibr" rid="B54">54</xref>]. Investigation of homogenous association and/or heterogeneous association between cleaved ER regions and regulatory proteins would be informative with regard to the validity of this hypothesis.</p><p>In view of these considerations, it seems that analysis of the biological role and potential clinical relevance of endogenous ER proteolysis is of prime importance. The hydroxylapatite extraction assay described here, which is extremely simple, would provide a criteria for the detection of tumours characterized by a high amount of full-length ER (low EI). Such tumours may perhaps be considered an indication for adjuvant hormone therapy if we refer to earlier sucrose gradient sedimentation data concerning hormone dependency of tumours that express native and/or cleaved ER forms [<xref ref-type="bibr" rid="B17">17</xref>]; very high EI may perhaps be an index of poor prognosis. In practice, for clinical correlation studies devoted to analysis of these hypotheses, we propose that a hydroxylapatite extraction assay should be performed as a complement to the usual immunohistochemical test (or Abbott's ER enzyme immunoassay [<xref ref-type="bibr" rid="B4">4</xref>]), in view of the small size of most tumours that restrict the cytosol amount. If our speculations concerning the biological/clinical relevance of the detection of cleaved ER isoforms are verified, then the introduction of such a hydroxylapatite assay into routine practice would be helpful to orientate therapy.</p></sec>

Mechanisms of suberoylanilide hydroxamic acid inhibition of mammary cell growth

<p>The mechanism of suberoylanilide hydroxamic acid in cell growth inhibition involved induction of pRb-2/p130 interaction and nuclear translocation with E2F-4, followed by significant repression in E2F-1 and PCNA nuclear levels, which led to inhibition in DNA synthesis in mammary epithelial cell lines.</p>

<contrib id="A1" corresp="yes" contrib-type="author"><name><surname>Said</surname><given-names>Thenaa K</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>tsaid@bcm.tmc.edu</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Moraes</surname><given-names>Ricardo CB</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Sinha</surname><given-names>Raghu</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Medina</surname><given-names>Daniel</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Breast Cancer Research

<sec><title>Synopsis</title><sec><title>Background:</title><p>Hybrid polar compounds (HPCs) have induced cell growth arrest, terminal differentiation and/or apoptosis in various transformed cell lines. We have previously reported that the prototype HPC (hexamethylene bisacetamide [HMBA]) was able to arrest the growth of transformed mammary (TM) 2H cells (p53 null), a highly tumorigenic mouse mammary epithelial cell line, by inhibiting G1 kinase activities, concomitant with an increase in the cyclin D2 protein level and hypophosphorylated isoforms of the three pRb pocket proteins, which led to the formation of stable cyclin D2/pRb complexes and G1 cell arrest. It has been reported that the second generation of HPCs (suberoylanilide hydroxamic acid [SAHA]), structurally related to but 2000-fold more potent than HMBA, was an inhibitor of histone deacetylase activity and caused accumulation of hyperacetylated histone H4 in murine erythroleukemia.</p></sec><sec><title>Objectives:</title><p>To determine the mechanism of SAHA in cell growth inhibition in TM10 (p53 wt) and TM2H (p53 null) hyperplastic mouse mammary cell lines.</p></sec><sec><title>Methods:</title><p>TM10 and TM2H cells were examined in the presence or absence of 2.5 &#x003bc;M SAHA for cell growth rate by [³H]-thymidine uptake, DNA synthesis by flow cytometry after cells were labeled with BrdU, G1/S cyclin-dependent kinase (cdk) activities, phosphorylation levels of pRb pocket proteins, protein levels of E2F-1, PCNA and p21, pRb-2/p130 interaction, and nuclear localization with E2F-4 by western blot, immunoprecipitation and immunostaining assays.</p></sec><sec><title>Results:</title><p>SAHA was able to arrest cell growth at G1, and inhibited DNA synthesis in both TM10 and TM2H cell lines. Cell growth arrest was accompanied by increases in histone H3 and H4 protein and acetylation levels, a profound increase in the interaction and nuclear localization of pRb-2/p130&#x02013;E2F-4 complexes, significant reductions in E2F-1 and PCNA protein levels, inhibition in G1/S cdk activities and increases in the levels of hypophosphorylated isoforms of three pRb pocket proteins.</p></sec><sec><title>Conclusion:</title><p>A novel mechanism of SAHA mediated growth inhibition through significant increases in the formation and nuclear localization of pRb-2/p130&#x02013;E2F-4 complexes, which resulted in cell growth arrest and significant repression in the levels of two key molecules, E2F-1 and PCNA, essential for DNA synthesis in two mouse mammary epithelial cell lines. These responses to SAHA were independent of the p53 status of the cell; however, reversibility of SAHA-mediated growth correlated with the wild type p53 status.</p></sec></sec><sec><title>Introduction</title><p>Progression through the mammalian cell cycle requires that gene expression is coordinated with the activity of cell cycle control proteins. A critical period is the transition from the G1 into the S phase, as cells become committed to the division cycle. Binding of free E2F/DP heterodimers to E2F sites generally activates transcription of proteins required for G1 &#x02192; S transition and DNA synthesis [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>], whereas complex formation with pRb or other pocket proteins including p107 and pRb-2/p130 silences transcriptional activities of downstream target genes [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. The resulting retinoblastoma protein (Rb)&#x02013;E2F interaction not only blocks transcriptional activation by E2F, but also forms an active transcriptional repressor complex at the promoter of cell cycle genes that can block transcription by recruiting histone deacetylase (HDAC) and remodeling chromatin [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>].</p><p>Several HDAC inhibitors mediate cell growth arrest and/or differentiation [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. We chose to examine the effect of HPCs, which have been reported to induce terminal differentiation and/or apoptosis [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>] in many transformed cells. Although treatment with HMBA induces remission in patients with myelodysplastic syndrome and acute myelogenoud leukemia, it is not currently used therapeutically because of the high dosage required (millimolar blood levels) and the accompanying toxic side effects (thrombocytopenia) [<xref ref-type="bibr" rid="B21">21</xref>].</p><p>In this study, we report a novel mechanism of cell growth inhibition by the second generation of HPCs, named SAHA, which is 2000-fold more potent than HMBA and bears at least one hydroxamide in place of the amides in HMBA [<xref ref-type="bibr" rid="B17">17</xref>]. SAHA was reported to be a histone deacetylase inhibitor and caused accumulation of hyperacetylated histone H4 in murine erythroleukemia [<xref ref-type="bibr" rid="B17">17</xref>]. Very little is known about the anticancer mechanism of SAHA in epithelial cells; however, a recent study demonstrated that SAHA diet, at 900 parts per million (ppm), fed to rats reduced methylenitrosouren-induced mammary tumor incidence by 40%, total tumors by 66% and tumor volume by 78% [<xref ref-type="bibr" rid="B22">22</xref>]. In this study, we tested whether SAHA has similar potency to inhibit cell growth in two mouse mammary epithelial cell lines, TM10 (p53 wt) and TM2H (p53 null). We identified a novel mechanism for SAHA in cell growth arrest through inhibition in DNA synthesis, concomitant with significant increases in the nuclear localization of pRb-2/p130 associated with E2F-4, decreases in key molecules in DNA synthesis (E2F-1, PCNA and p21), and increases in histone H3 and H4 protein and acetylation levels. This study discusses the difference in recovery from cell growth inhibition in two mammary epithelial cell lines, TM10 and TM2H, after SAHA removal from cultures.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Development of cell lines and cell culture</title><p>The TM10 and TM2H cell lines chosen for this study were isolated from two different mouse mammary hyperplastic outgrowths: TM10 and TM2H, respectively, as described earlier [<xref ref-type="bibr" rid="B23">23</xref>]. The parental TM10 outgrowth is a moderately tumorigenic outgrowth line <italic>in vivo</italic> (time for 50% of the transplants to produce tumors, 11 months) that is karyotypically diploid and maintains wild type p53 expression. TM2H, in contrast, is a highly tumorigenic outgrowth line <italic>in vivo</italic> (time for 50% of the transplants to produce tumors, &#x02264; 4 months), karyotypically aneuploid (DNA index = 1.69) and contains a p53 mutation resulting in a null phenotype [<xref ref-type="bibr" rid="B24">24</xref>]. Exponentially growing TM10 and TM2H cell lines in DMEM/F12 media buffered with 10 mM HEPES at pH 7.6 with 2% adult bovine serum, 10 &#x003bc;g/ml insulin, 5 ng/ml epidermal growth factor (EGF) and 5 &#x003bc;g/ml gentamycin at 60-70% confluence were treated with SAHA (courtesy of Dr Paul Marks and Dr Victoria Richon, Memorial Sloan Kettering Cancer Center, New York, NY, USA). Cells were examined at the time points indicated for cell growth and cell cycle activities.</p></sec><sec><title>Analysis of cell growth</title><p>Cell growth rates of TM10 and TM2H lines were determined using a [³H]-thymidine uptake assay, as described earlier [<xref ref-type="bibr" rid="B25">25</xref>]. In initial studies, both cell lines were cultured in the absence or presence of 0.1, 2.5, 5.0, and 10 &#x003bc;M SAHA for 6 days. Both cell lines were also cultured for 2 days in the presence of SAHA at the concentrations already mentioned, followed by removing SAHA from the media for a subsequent 4 days. Based on the results of these studies, subsequent experiments examined asynchronously growing TM10 and TM2H cell lines in the presence or absence of 2.5 &#x003bc;M SAHA for 24 h.</p></sec><sec><title>FACS analysis</title><p>Exponentially growing cells at 60-70% confluence at control or treated with 2.5 &#x003bc;M SAHA for 24 h were pulse-labeled for 1 h with 10 &#x003bc;M BrdU (Sigma, St Louis, MO, USA). The cell cultures were rinsed once with phosphate buffered saline (PBS), trypsinized for 3 min and rinsed three times with PBS. Cells were resuspended in 200 &#x003bc;l PBS and fixed in 5 ml cold 70% ethanol overnight. Fixed cells were counted and, generally, 4 &#x000d7; 10⁶ cells were transferred to 15 ml polypropylene tubes, centrifuged at 3000 rpm for 5 min and the supernatant removed. Cells were stained for the newly incorporated BrdU for DNA synthesis using BrdU monoclonal antibodies conjugated to fluorescein isothiocyanate (FITC) and stained with propidium iodide for DNA content following the protocol described in Becton and Dickinson's (San Diego, CA, USA) instructions for flow cytometric analysis.</p></sec><sec><title>Nuclear and cytoplasmic extracts</title><p>To obtain nuclei, (4&#x02013;6) &#x000d7; 10⁷ cells of each cell line grown in the presence and absence of 2.5 &#x003bc;M SAHA were washed twice in PBS, followed by suspension in 0.3 ml nuclear buffer consisting of 2 mM MgCl₂, 5 mM K₂HPO₄, 0.1 mM EDTA, 1 mM PMSF, 20 &#x003bc;g/ml aprotinin, 20 &#x003bc;g/ml leupeptin, 0.1 mM Na₃VO₄ and 5 mM &#x003b2;-glycerophosphate. An additional 0.3 ml nuclear buffer containing 0.7% Triton X-100 was then added, after standing on ice for 8-10 min. The suspensions were examined for cell lysis microscopically, centrifuged at 800 rpm for 10 min at 4&#x000b0;C, and the supernatant designated the cytoplasmic fraction. The pellets were washed once with nuclear buffer, and the nuclear extracts were prepared by resuspending the pellets in 0.3 ml buffer containing 20 mM HEPES (pH 7.8), 25% glycerol, 0.42 M NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.5 mM PMSF, 0.5 mM DTTl, 0.1 mM Na₃VO₄, 50 mM NaF, 20 &#x003bc;g/ml leupeptin, and 5 mM &#x003b2;-glycerophosphate. Sonication was carried out on ice using an ultrasonicator processor (PGC Scientific, Caithersburg, MA, USA), and the mixtures were examined microscopically for complete break of the nuclei. The supernatants were designated as nuclear extracts after centrifugation of the mixtures, and the total protein was determined in the nuclear and cytoplasmic fractions.</p></sec><sec><title>Western blot and immunoprecipitation analysis</title><p>Histones were isolated and lyophilized from nuclear extracts in SAHA treated and untreated TM10 and TM2H cell lines following a protocol described earlier [<xref ref-type="bibr" rid="B26">26</xref>]. Histone samples were assessed for purification quality on 15% SDS acrylamide gel including calf thymus histones as controls before western blot analysis was carried out. Histone samples were resolved by electrophoresis using 15% acid&#x02013;urea gel containing 36%w/v urea, 5%v/v acetic acid, 600 &#x003bc;l TEMED and 0.7 ml of 10% ammonium persulfate prepared as described elsewhere [<xref ref-type="bibr" rid="B27">27</xref>]. Gels were either stained by Coomassie Brilliant Blue or equilibrated to be transferred into transfer buffer (0.7% acetic acid). The gel sandwich was set up as usual except for the placement of the blotting membrane (because proteins in this type of gel will migrate toward the negative electrode), and the procedure was continued as described earlier [<xref ref-type="bibr" rid="B27">27</xref>]. Each histone was resolved into multiple bands and were visualized in Coomassie blue gel. The acetylated histone isoforms were detected by immunoblot against acetylated histone H3 isoforms using anti-H3 antibodies raised and characterized by Dr Sharon Roth at MD Anderson Cancer Center (personal communication) or against acetylated histone H4 isoforms using anti-H4 polyclonal antibody (Upstate Biotechnology Inc, Lake Placid, NY, USA).</p><p>Western blot analysis for all the proteins examined in this study was carried out on equal amounts of cellular fraction protein extracts (100 &#x003bc;g/sample) following a protocol described earlier [<xref ref-type="bibr" rid="B25">25</xref>]. Staining the gel with Coomassie Brilliant Blue for each experiment assessed equal loading control. TM10 and TM2H cells (0.65 &#x000d7; 10³ cells/cm²) seeded in 75 T flasks grew for 2 days, and were treated with 2.5 &#x003bc;M SAHA for 24 h. Cellular fraction protein extracts were prepared after SAHA treatment [<xref ref-type="bibr" rid="B25">25</xref>]. Primary antibodies used at 1 or 2 &#x003bc;g/ml were p21/Cip (Pharmingen Inc, San Diego, CA, USA), pRb (IF8), p107 (SD9), p130 (C-20), E2F-1 (C-20), E2F-4(C-20) and PCNA (C-20) (Santa Cruz Biotechnology Inc, Santa Cruz, CA, USA). All antibodies were examined for specificity prior to use. The secondary antibodies conjugated to horseradish peroxidase (1 : 5000-1 : 15,000 dilution) were used followed by enhanced chemiluminescence detection reaction as described by the manufacturer (Amersham Pharmacia Biotechnology, Amersham, Bucks, UK).</p><p>The immunoprecipitation assay followed by western blot analysis was as described previously [<xref ref-type="bibr" rid="B25">25</xref>]. Briefly, equal nuclear cell extract (200 &#x003bc;g/sample) was mixed with 3 &#x003bc;g anti-E2F-4 antibodies at 4&#x000b0;C followed by the addition of 50 &#x003bc;l protein A-sepharose beads (Amersham Biotechnology). The immune complex was centrifuged, and the proteins in the immune complex were resolved by 10% SDS-PAGE followed by western blot analysis using anti-pRb-2/p130 antibodies as already described for western blot analysis.</p></sec><sec><title>Cyclin-dependent kinase assay</title><p>The TM10 and TM2H cell cultures at control and treated with 2.5 &#x003bc;M SAHA for 24 h were examined for cyclin D1, E, and A associated kinase activities as described previously [<xref ref-type="bibr" rid="B28">28</xref>]. Briefly, cellular protein extracts (25 &#x003bc;g) were precleared with 10 &#x003bc;l of 10% preimmune normal rabbit serum followed by immunoprecipitation with either 3 &#x003bc;g anti-cyclin D1, E polyclonal antibodies (Upstate Biotechnology), 2 &#x003bc;g anti-cyclin A polyclonal antibody antibodies (C-19)-G (Pharmingen), or normal rabbit preimmune serum as a negative control. All antibodies were examined for specificity [<xref ref-type="bibr" rid="B28">28</xref>]. Immunoprecipitate complexes were examined for kinase assay following a procedure described previously [<xref ref-type="bibr" rid="B28">28</xref>]. The substrates utilized in the kinase assays were either Histone-H1 (Sigma) or RB protein (Santa Cruz Biotechnology). The phosphorylated H1 and pRb bands were scanned and quantitated densitometrically using a Phosphoimager (Molecular Dynamics, Sunnyvle, CA, USA).</p></sec><sec><title>Immunofluorescent staining</title><p>Exponentially growing cells on slides were fixed in 4% paraformaldehyde in PEM buffer (80 mM Pipes [pH 7.0], 5 mM EGTA and 2 mM MgCl₂) for 30 min, permeabilized in 0.5% Triton X-100 in PEM buffer at room temperature for 15 min and rinsed three times with TBS + 0.1% Tween 20, followed by incubation in 0.5% nonfat dry milk in TBS + 0.1% Tween 20 for 2 h at room temperature. For BrdU and DAPI double immunostaining, cells were incubated for 30 min in media supplemented with 5 &#x003bc;M BrdU. After 30 min, cells were incubated in 2N HCl for 5 min at room temperature prior to incubation with the primary antibody (this step was omitted for pRb-2/p130 immunostaining). Following washing three times, cells were incubated in a (1 : 50) dilution of mouse anti-BrdU monoclonal antibody (Boehringer Mannheim, Indianapolis, IN, USA) for 2 h at 37&#x000b0;C. After washing three times, cells were then incubated in FITC-conjugated anti-mouse secondary antibody (1 : 400) dilution for 1 h at 37&#x000b0;C, followed by washing three times with anti-fade equilibrating buffer and mounting in anti-fade mounting medium (Molecular Probe, Eugene, OR, USA). The same sequential steps were followed for pRb-2/p130 immunostaining using anti-pRb-2/p130 monoclonal antibodies (Santa Cruz Biotechnology).</p></sec><sec><title>Apoptosis</title><p>Apoptotic activities in TM10 and TM2H cells in the absence and presence of 2.5 &#x003bc;M SAHA were examined by two procedures: the TACS 2TdT In Situ Apoptosis Detection assay following the manufacturer's instructions (Trevingen, Gaithersburg, MD, USA), and the DNA Fragmentation Assay described elsewhere [<xref ref-type="bibr" rid="B29">29</xref>].</p></sec></sec><sec><title>Results</title><sec><title>Effect of SAHA on TM10 and TM2H cell proliferation and morphology</title><p>The TM10 and TM2H cell lines were used to investigate whether the status of p53 influences the outcome of SAHA treatment. The origin of TM10 and TM2H cell lines was described earlier [<xref ref-type="bibr" rid="B30">30</xref>]. The rate of cell proliferation at concentrations of 0, 1.0, 2.5, 5, and 10 &#x003bc;M SAHA was examined by [³H]-thymidine assay at the indicated time points. The inhibitory effect of SAHA at concentrations of 2.5 &#x003bc;M and higher on cell proliferation was profound on both cell lines after 24 h. The TM10 cell proliferation continued to be inhibited during the following 5 days in the presence of SAHA, whereas minimal growth in TM2H cell growth on days 3 and 4 was observed (Figs <xref ref-type="fig" rid="F1">1A</xref>,<xref ref-type="fig" rid="F1">C</xref>). After removing SAHA from the cultures following 2 days of exposure to all SAHA concentrations, TM10 cells continued to be inhibited at concentrations of 2.5 &#x003bc;M and higher during the following 4 days of SAHA free cultures (Fig. <xref ref-type="fig" rid="F1">1B</xref>), whereas TM2H cells resumed proliferation (Fig. <xref ref-type="fig" rid="F1">1D</xref>). These results suggest that SAHA induced a dose-dependent block in DNA replication on both cell lines. A 2.5 &#x003bc;M dose of SAHA was used in further experiments because this concentration seems to be unequivocally effective on cell growth inhibition on both cell lines after 24 h of treatment. Furthermore, no apoptotic activity was observed after 24 and 48 h of treatment with 2.5 &#x003bc;M SAHA in both cell lines (data not shown).</p><p>The effect of SAHA on blocking cell proliferation through blocking DNA synthesis was examined by BrdU labeling, as determined by immunofluorescence staining (Fig. <xref ref-type="fig" rid="F2">2A</xref>), and the cell cycle profile by flow cytometric analysis (Fig. <xref ref-type="fig" rid="F2">2B</xref>). The BrdU indices in the control TM10 and TM2H cells, including all types of BrdU staining patterns, were 16 and 34%, respectively, indicating that the percentage of TM2H cells synthesizing DNA or in the 'S phase' were more than double that of TM10 cells (Fig. <xref ref-type="fig" rid="F2">2A</xref>). Upon 2.5 &#x003bc;M SAHA treatment for 24 h, 800 cells per sample were counted stained with BrdU against cells stained with DAPI, and BrdU labeling indices dropped dramatically to 4 and 6% in TM10 and TM2H cells, respectively. Furthermore, the BrdU staining revealed the size and pattern of replication clusters, which are related to the general patterns of DNA replication in mammalian cell nuclei at early, mid and late S phase [<xref ref-type="bibr" rid="B31">31</xref>]. These data suggested that 2.5 &#x003bc;M SAHA was capable of inhibiting cell proliferation in both TM10 and TM2H cell lines by blocking DNA synthesis, regardless of their differences in the number of cells synthesizing DNA.</p><p>These results were confirmed by flow cytometric analysis after staining cells with BrdU for newly synthesized DNA and with propidium iodide for DNA content. The percentages of cells in the G1 and S phases in control TM10 cells at the time of treatment were 57 and 18%, respectively, and those for TM2H cells were 39 and 28%, respectively (Fig. <xref ref-type="fig" rid="F2">2B</xref>). The number of TM2H cells growth arrested in the G1 phase upon SAHA treatment for 24 h increased by 21%, compared with the 8% increase in the TM10 cell line, concomitant with similar differences in the percentage of cells inhibited in the S and G2/M phases (Fig. <xref ref-type="fig" rid="F2">2B</xref>). This difference in G1 cell growth arrest between the two cell lines may be attributed to their differences in the initial number of cells distributed in each cell cycle phase before treatment.</p><p>The shape and morphology of TM10 and TM2H cells grown on slides in the presence and absence of 2.5 &#x003bc;M SAHA were examined. The dimensions of 50 randomly selected cells of each cell line were measured using a microscope scale. The TM10 and TM2H mammary epithelial cells measured 5.4 &#x000b1; 2.1 and 3.6 &#x000b1; 1.4 &#x003bc;m in width and 13.7 &#x000b1; 5.9 and 18.6 &#x000b1; 8.0 &#x003bc;m in length, respectively. TM10 cells generally had five to seven extensions, whereas TM2H cells had three to five extensions. TM10 and TM2H cells became flattened and increased in both nuclear and cytoplasmic volume after 24 h of 2.5 &#x003bc;M SAHA treatment. Both cells exhibited less distinct intracellular borders (Fig. <xref ref-type="fig" rid="F3">3</xref>). The morphology of both cell lines was similar to cells committed to cell differentiation; however, these mammary cells grown on plastic dishes were unable to differentiate by SAHA.</p></sec><sec><title>Effect of SAHA on histone protein and acetylation levels in TM10 and TM2H cell lines</title><p>It is not known whether SAHA alters the protein acetylation levels of various histones in the mammary epithelial cells. Changes in the acetylation or phosphorylation of histones result in alterations that can be visualized as changes in protein mobility on an acid-urea gel following western blot analysis [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>]. These gels separate histones on the basis of charges as well as size, resulting in multiple bands for each histone that correspond to differently modified isoforms. These isoforms can be identified by their characteristic mobility and by staining with antibodies specific to each histone isoform [<xref ref-type="bibr" rid="B26">26</xref>]. Protein and acetylation levels of histone H3 and H4 isoforms on histone samples isolated from SAHA treated and untreated TM10 and TM2H cells stained by Coomassie Brilliant Blue (Fig. <xref ref-type="fig" rid="F4">4A</xref>) and by western blots (Fig. <xref ref-type="fig" rid="F4">4B</xref>) were examined to identify the potency of SAHA as a histone deacetylase inhibitor on mammary epithelial cells. We focused on H3 and H4 because antibodies are available and acetylation events in those histones have been studied. These results suggested that SAHA not only increased histone acetylation, as shown in H4 (Fig. <xref ref-type="fig" rid="F4">4A</xref>), but also increased histone protein levels in TM10 and TM2H cells (Fig. <xref ref-type="fig" rid="F4">4B</xref>). The increases in protein and acetylation levels of H3 and H4 in SAHA treated TM10 and TM2H cell lines were indicated by the strong staining with the acetylation specific antibodies of H3 and H4, while the acetylated slow migrated isoforms showed smearing due to the increases in their protein level (Fig. <xref ref-type="fig" rid="F5">5B</xref>). We did not, however, observe differences in the initial acetylation levels between untreated TM10 and TM2H cell lines.</p></sec><sec><title>SAHA induced hypophosphorylation of all the three pocket proteins</title><p>Phosphorylation of three major pocket proteins, Rb, p107 and Rb2/p130, were examined after SAHA treatment, as each pocket protein affects the G1 and S phases of the cell cycle. Upon treatment with 2.5 &#x003bc;M SAHA for 24 h, a profound increase in hypophosphorylation of the three pocket proteins was observed in both TM10 and to a lesser extent in TM2H cells as compared with their control counterparts, judged by the faster mobility of the Rb pocket protein isoforms on SDS gel (Fig. <xref ref-type="fig" rid="F5">5</xref>). The difference in the level of hypophosphorylated pRb pocket protein isoforms between the two cell lines treated with SAHA may be attributed to differences in G1 and S phase cdk activity (examined in the following section). Results suggested that SAHA induced cell growth arrest through dephosphorylation of three pRb pocket proteins, which may be related to inhibition in G1/S phase cdk activities.</p></sec><sec><title>SAHA inhibits multiple cyclin-associated kinase activities</title><p>There are extensive studies on the relation between specific phosphorylation sites on pRb and its potential transcription repression [<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>,<xref ref-type="bibr" rid="B34">34</xref>]. It is well known that pRb phosphorylation sites are recognized by specific cdks [<xref ref-type="bibr" rid="B33">33</xref>], and that most of the 16 pRb-phosphorylation sites are sequentially phosphorylated throughout the cell cycle (reviewed in [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>]). SAHA treatment resulted in G1 cell growth arrest in TM10 and TM2H cell lines, concomitant with the profound increase in dephosphorylation of the three pRb pocket proteins. Therefore, we examined cyclin D1, E and A cdk2 activities in TM10 and TM2H cell cultures after 24 h treatment with SAHA because they reflect kinase activities before, during and after the G1 &#x02192; S checkpoint, respectively. The three cyclin/cdk2 complexes directly reflect the sequential events in DNA synthesis [<xref ref-type="bibr" rid="B36">36</xref>]. The initial levels of cyclin D1, E and A associated kinase activities were 3-, 2.5- and 4-fold higher in untreated TM2H compared with TM10 cells, respectively (Fig. <xref ref-type="fig" rid="F6">6A</xref><xref ref-type="fig" rid="F6">B</xref><xref ref-type="fig" rid="F6">C</xref>). These results were predicted based on their differences in cell distribution throughout the cell cycle, as appeared in the cell cycle profile analysis, and may reflect the cell line differences in known p53 status. Cyclin D1, E and A associated kinase activities decreased by four-, four-, and fivefold, respectively, in TM10 cells compared with control on SAHA treatment, whereas only cyclin D1 and A associated kinase activities decreased (2.2- and 2.6-fold, respectively) in SAHA treated TM2H cells compared with control (Fig. <xref ref-type="fig" rid="F6">6</xref>). These data demonstrate that SAHA inhibits multiple cdk activities, and the degree of inhibition was more profound in TM10 verses TM2H cells.</p></sec><sec><title>Effect of SAHA on p21, PCNA, E2F-1 and E2F-4 proteins as a function of cell growth inhibition</title><p>The difference in the protein level of p21/Cip1 between TM10 and TM2H cell lines was predicted based on their differences in p53 status (Fig. <xref ref-type="fig" rid="F7">7A</xref>). Perinuclear localization of p21 was detected in only few TM2H cells (p53 null) when both cell lines were immunostained for p21, whereas p21 was localized in both cytoplasmic and nuclear compartments in TM10 cells (p53 wt) (data not shown). The p21 expression level decreased by 50% in TM10 cells upon SAHA treatment (Fig. <xref ref-type="fig" rid="F7">7A</xref>). These results suggest that SAHA inhibited DNA synthesis in the presence or absence of p21 protein.</p><p>PCNA [<xref ref-type="bibr" rid="B37">37</xref>] and E2F-1 [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B7">7</xref>] are two key mediators of the G1 &#x02192; S transition and DNA synthesis. Because such regulatory proteins can be present at different levels and in different molecular complexes in nuclei as compared with the cytoplasm, separation of the two cellular compartments allowed an assessment of the site(s) at which these proteins exert their primary activity unimpeded by artifacts that may result from intermixing the two compartments. In the untreated cells, the cytoplasmic PCNA protein levels in TM10 and TM2H cell lines were similar, but the level of PCNA in the nuclear fraction was fourfold higher in TM2H cells compared with TM10 cells (Fig. <xref ref-type="fig" rid="F7">7B</xref>). The effect of SAHA was profound on the nuclear PCNA fraction in both cell lines, but minimal on the cytoplasmic PCNA fraction. The pronounced difference in the levels of PCNA in growing versus arrested cells is in accordance with the determinant role PCNA plays in DNA synthesis [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>].</p><p>To further investigate the effect of SAHA on inhibition of DNA synthesis, we examined E2F-1 as a key mediator for G1 &#x02192; S transition and DNA synthesis. The initial nuclear protein level of E2F-1 in the untreated TM2H cells was 4.8-fold higher compared with the TM10 cell line (Fig. <xref ref-type="fig" rid="F7">7C</xref>). Nuclear E2F-1 protein was completely abolished in TM10 cells after 24 h of SAHA treatment, and decreased 2.2-fold in TM2H cells (Fig. <xref ref-type="fig" rid="F7">7C</xref>). These results demonstrate that SAHA treatment also inhibits E2F-1 protein levels.</p><p>It has been reported that expression of <italic>E2F-4</italic> gene does not change in relation to cell growth, although a modest increase can sometimes be observed in growing cells [<xref ref-type="bibr" rid="B39">39</xref>]. E2F-4 protein levels were examined in the nuclear and cytoplasmic fractions in both cell lines with and without SAHA, and the protein levels did not change upon SAHA treatment in either fractions of either cell line (Fig. <xref ref-type="fig" rid="F7">7D</xref>). It is noteworthy that the mechanism of SAHA in G1 cell growth arrest may target the expression of specific proteins involved in DNA synthesis as it inhibited E2F-1 and PCNA, but not E2F-4, suggesting that SAHA may cause different levels of histone acetylation at distinct regions of the genome.</p><p>Rb2/p130 subcellular localization and interaction with E2F-4 was examined in both cell lines in the absence and presence of SAHA because Rb2/p130 is an important cytoplasmic partner of E2F-4 and is capable of inducing nuclear localization of the complex upon cell growth arrest and differentiation [<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref>]. Immunostaining analysis using monoclonal antibody specific for pRb2/p130 revealed that SAHA enhanced pRb2/ p130 nuclear localization in both cell lines treated with SAHA for 24 h compared with their controls (Fig. <xref ref-type="fig" rid="F8">8A</xref>). To examine whether the increase in hypophosphorylation of pRb2/p130 (Fig. <xref ref-type="fig" rid="F5">5</xref>) and its nuclear localization after SAHA treatment was associated with interaction to E2F-4, equal amounts of nuclear protein extracts were immunoprecipitated with antibodies against E2F-4 followed by western blot analysis against pRb2/p130 antibodies. Interestingly, the results revealed an unequivocal increase in the nuclear pRb2/p130 protein level associated with E2F-4 in both cell lines treated with SAHA (Fig. <xref ref-type="fig" rid="F8">8B</xref>). These intriguing results indicate that the interaction of hypophosphorylated pRb2/p130 with E2F-4 followed by enhancement in nuclear location after SAHA treatment is a novel mechanism in SAHA-mediating cell growth arrest.</p></sec></sec><sec><title>Discussion</title><p>In this study, we present novel data on the mechanism of SAHA in cell growth arrest on two mouse mammary epithelial cell lines, TM10 (p53 wt) and TM2H (p53 null). SAHA was able to increase histone H3 and H4 protein and acetylation levels, and caused a profound decrease in the protein levels of key molecules, PCNA and E2F-1, essential for DNA synthesis. Furthermore, SAHA significantly enhanced the interaction of pRb2/p130 to E2F-4 and the nuclear localization of the pRb2/p130&#x02013;E2F-4 complex. SAHA also resulted in the inhibition of G1/S kinase activities and, consequently, hypophosphorylation of the three pRb pocket proteins, which led to G1 cell growth arrest and dramatic decreases in DNA synthesis in both cell lines. TM10 cells continued to be inhibited for 4 days upon removing SAHA after 2 days of treatment, whereas TM2H cells were able to recover their proliferation potentials. A summary of the differences in the molecular status and cell cycle profile between TM10 and TM2H cell lines before SAHA treatment are summarized in Table <xref ref-type="table" rid="T1">1</xref>. These differences in p21 protein and BrdU index were predicted based on p53 status of these two cell lines and are in parallel with other reports on mammary tumors, where the absence of p53 results primarily in greater proliferation response in the affected cell [<xref ref-type="bibr" rid="B42">42</xref>]. No differences in histone H3 and H4 acetylation levels were, however, observed in relation to p53 status between the two cell lines. It is not known at this point whether specific mutations of p53 alter the degree of histone acetylation in cells. Histone acetylation and p53 mutation appear not to be correlated in this study; nevertheless, it is necessary and more sensitive to examine histone acetylase and deacetylase activities in correlation to p53 status.</p><p>The mechanism of SAHA in blocking DNA synthesis, as demonstrated by flow cytometric analysis, appeared similar in both TM10 and TM2H cell lines, and implicated several events. First, SAHA increased histones H3 and H4 protein levels after 24 h of treatment, which could result in cell cycle arrest. It has been reported that upregulation of KIAA0128 gene expression, which has been implicated in activation of histone mRNA synthesis, was related to cell cycle arrest in MCF-7 cells after treatment with SAHA [<xref ref-type="bibr" rid="B43">43</xref>]. As SAHA treatment increased histone (H3 and H4) protein and acetylation levels in both cell lines, this may have altered the association of histones with DNA, thereby altering nucleosomal conformation and stability [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B44">44</xref>]. Local perturbations of chromatin structure can specifically affect the accessibility and/or function of transcriptional regulatory proteins that bind DNA sequences in the region where histone acetylation or deacetylation took place [<xref ref-type="bibr" rid="B44">44</xref>]. HDAC inhibitors, such as trichostatin A and trapoxin, modulate gene expression in either a positive, negative or neutral fashion [<xref ref-type="bibr" rid="B45">45</xref>]. Ample studies have demonstrated the implication of histone hyperacetylation in gene transcription but also in silencing gene expression of others [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>].</p><p>It is well known that E2F-1 regulates transcription of genes predominantly expressed during the G1 &#x02192; S transition such as cyclins [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B5">5</xref>], cdks [<xref ref-type="bibr" rid="B1">1</xref>], <italic>E2F-1</italic> gene [<xref ref-type="bibr" rid="B46">46</xref>], the <italic>RB1</italic> tumor suppressor gene [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B47">47</xref>], and genes for DNA replication and repair enzymes and factors [<xref ref-type="bibr" rid="B4">4</xref>]. It appears that SAHA has a profound inhibitory impact on the protein levels of key molecules, E2F-1, PCNA and p21, essential for DNA synthesis. Based on previous reports, disintegration of the cyclin/cdk complexes important for DNA synthesis is correlated to E2F-1 expression level [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B5">5</xref>]. It is thus conceivable to interpret that the profound inhibition in E2F-1 and PCNA protein levels after 24 h of exposure to 2.5 &#x003bc;M SAHA may result in disintegration and deactivation of D1, E and A cdk2 complexes, which consequently leads to hypophosphorylation of the three Rb pocket proteins. It is plausible to suggest that the inhibition in E2F-1 protein levels by SAHA was either at the transcription level or induction of the ubiquitin-protein ligase responsible for E2F-1 degradation, but not E2F-4, and this resulted in blocked DNA synthesis. Further work is necessary to prove whether the effect of SAHA is at the RNA transcription level or on stability of E2F-1 protein.</p><p>The inhibition in nuclear p21 in SAHA treated TM2H as well as TM10 cells underscores that SAHA-arrested cell growth is through a p53-independent pathway. A recent report indicated that the transcription of p21^Cip1 and accumulation of acetylated histones associated with the promoter and coding regions of that gene were induced after 2 h in 7.5 &#x003bc;M SAHA and fall by 24 h in T24 bladder carcinoma cells [<xref ref-type="bibr" rid="B48">48</xref>]. Although we have utilized 2.5 &#x003bc;M SAHA, our results are in agreement with their data on the fall in p21 level after 24 h of treatment.</p><p>A more intriguing and novel mechanism of SAHA-mediated cell growth arrest was the enhanced interaction and nuclear localization of Rb2/p130&#x02013;E2F-4 complexes in both cell lines after 24 h of treatment. It is well documented that E2F-1 possesses an intrinsic nuclear localization signal whereas E2F-4 is devoid of such signal [<xref ref-type="bibr" rid="B49">49</xref>,<xref ref-type="bibr" rid="B50">50</xref>], and that the mechanism of E2F-4 nuclear localization has been documented to be through its interaction with Rb2/p130 pocket protein, which impedes cell cycle progression [<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. Furthermore, recent studies demonstrated that Rb2/p130 in complexes with E2F-4 actively represses E2F-1 transcription in cell differentiation and growth arrest, and that this complex was considered the main E2F-1 regulator during the early G1 phase [<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref>]. Other reports suggest that Rb recruitment of HDAC1 activity repressed E2F-1 [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. Although SAHA increases acetylation of histones H3 and H4, it is not known whether SAHA is able to inhibit all HDAC activities of all types of histones, including HDAC1, or whether the profound increase in Rb2/p130&#x02013;E2F-4 nuclear complex after SAHA treatment may have an alternative pathway other than recruitment of HDAC1 activity.</p><p>The difference between TM10 (p53 wt) and TM2H (p53 null) cell cultures in response to removing 2.5 &#x003bc;M SAHA following 2 or 3 days of treatment was significant. We suggest that it might reflect their difference in p53 status. The TM10 cells did not exhibit signs of cell proliferation or 'recovery' during the following 3-4 days. TM2H cells, in contrast, recovered by 88% after 2 days of treatment. Longer treatment may be necessary to inhibit TM2H (p53 null) mammary epithelial cells as preliminary results indicate TM2H cells did not recover after 3 days of SAHA treatment (data not shown). We suggest that the difference in growth recovery between TM10 (p53 wt) and TM2H (p53 null) cell lines after 2 days in 2.5 &#x003bc;M SAHA may be attributed to two factors, both related to their p53 status. Firstly, the p53 in TM10 (p53 wt) cells might have been acetylated upon treatment with 2.5 &#x003bc;M SAHA for 2 or 3 days. Recent studies demonstrated acetylation of p53 in the C-terminal domain increased the DNA-binding capacity of the protein [<xref ref-type="bibr" rid="B53">53</xref>,<xref ref-type="bibr" rid="B54">54</xref>,<xref ref-type="bibr" rid="B55">55</xref>]. This event is obviously not present in TM2H (p53 null). Secondly, although TM10 cells (p53 wt) have lost 64% of their nuclear p21 during SAHA treatment, the remaining 36% of the nuclear p21 plus the continued synthesis of p21 by p53 activity [<xref ref-type="bibr" rid="B55">55</xref>,<xref ref-type="bibr" rid="B56">56</xref>,<xref ref-type="bibr" rid="B57">57</xref>] during the recovery period would maintain TM10 cells in the inhibited state for several days without SAHA. TM2H cells (p53 null), in contrast, lack both negative regulatory potentials of acetylated p53 and the availability of nuclear p21.</p><p>We conclude that the mechanisms of SAHA inhibition of DNA synthesis and cell growth arrest at G1 were similar in both TM10 (p53 wt) and TM2H (p53 null) mouse mammary epithelial cell lines. A proposed mechanism of SAHA stresses the involvement of pRb2/p130&#x02013;E2F-4 interaction and nuclear localization, which ultimately results in cell growth arrest and repression in nuclear E2F-1 and PCNA protein levels, and the subsequent inhibition of DNA synthesis in both cell lines. However, p53 status was critical in maintaining growth arrest in TM10 cells 4 days after removing SAHA treatment, whereas TM2H cells (p53 null) recovered growth arrest under the same conditions. We therefore suggest that the dose and time regimen for histone deacetylase inhibitors, such as SAHA, may have to consider the p53 status of breast cancers.</p></sec><sec><title>Abbreviations</title><p>CI = confidence interval; GPRD = General Practice Research Database; NA-NSAID = nonaspirin nonsteroidal anti-inflammatory drug; RR = relative risk; UGIC = upper gastrointestinal complications.CI = confidence interval; GPRD = General Practice Research Database; NA-NSAID = nonaspirin nonsteroidal anti-inflammatory drug; RR = relative risk; UGIC = upper gastrointestinal complications.</p></sec>

An investigation of soy intake and mammographic characteristics in Hawaii

<p>This cross-sectional investigation in Hawaii explored the relation between soy foods and mammographic characteristics using two food frequency questionnaires and a computer-assisted density assessment method. Japanese and Chinese women reported significantly greater soy food intake than Caucasian women. Whereas soy intake and the size of the dense areas were not related, soy intake and percent mammographic densities were positively associated. The size of the entire breast and the nondense area (ie the fatty part of the breast) were inversely related to soy intake. These results suggest the hypothesis that soy foods by themselves or as part of an Asian dietary pattern may affect the growth of the female breast before adulthood, but the possible mechanisms of action have to be explored in future studies.</p>

<contrib id="A1" corresp="yes" contrib-type="author"><name><surname>Maskarinec</surname><given-names>Gertraud</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>gertraud@crch.hawaii.edu</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Meng</surname><given-names>Lixin</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Breast Cancer Research

<sec><title>Synopsis</title><sec><title>Introduction:</title><p>Based on ethnic [<xref ref-type="bibr" rid="B1">1</xref>] and international [<xref ref-type="bibr" rid="B2">2</xref>] differences in breast cancer risk, and on results from cell and animal studies [<xref ref-type="bibr" rid="B3">3</xref>], a role for soy foods in breast cancer prevention has been proposed. The epidemiologic investigations on soy and breast cancer risk have so far produced conflicting results [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Mammographic densities that refer to the distribution of fat, connective, and epithelial tissue in the female breast have been shown to be related to breast cancer risk [<xref ref-type="bibr" rid="B5">5</xref>]. A high percentage of dense parenchyma on mammographic images appears to confer a fourfold risk of developing breast cancer [<xref ref-type="bibr" rid="B11">11</xref>].</p></sec><sec><title>Aims:</title><p>The purpose of this report is to investigate the hypothesis that soy foods are related to lower mammographic densities among a population of women with Caucasian, Chinese, Japanese, and Native Hawaiian ancestry living in Hawaii.</p></sec><sec><title>Methods:</title><p>In a cross-sectional study approved by the institutional review boards of all participating institutions, women with different ethnic backgrounds were recruited at five mammography facilities on the island of Oahu. All study participants signed informed consent and completed a validated food frequency questionnaire (FFQ) [<xref ref-type="bibr" rid="B12">12</xref>] especially designed for a multiethnic population [<xref ref-type="bibr" rid="B13">13</xref>]. Because the FFQ included only two soy food items, we designed a soy questionnaire (SQ) with 10 categories of soy foods that are commonly consumed in Hawaii: tofu, fried tofu, pressed tofu, green soybeans, miso, soybean sprouts, tofu skin, soy milk/drink, vegetarian meats, and fermented soybeans. Intake of soy foods according to the SQ in a validation study among 102 women was highly correlated with urinary isoflavone excretion [<xref ref-type="bibr" rid="B14">14</xref>].</p><p>Both cranio-caudal views of the mammogram were obtained from the mammography clinics after the radiologic evaluation had been completed and any malignancy or suspicious lesions ruled out. Women who reported a history of breast cancer or augmentation surgery were excluded from the study. Computer-assisted mammographic density assessment was performed using a method that was first developed in Toronto [<xref ref-type="bibr" rid="B15">15</xref>] and later modified at the University of Southern California in Los Angeles [<xref ref-type="bibr" rid="B16">16</xref>]. The reader first draws the outline of the breast (using an outlining tool) and then searches for the best threshold gray level value <italic>X</italic> where all pixels with values above <italic>X</italic> are considered to represent mammographic densities. The pixel count corresponding to the area colored within the outline of the breast is determined by the computer, as is the total area within the outline of the breast. We used four mammographic parameters for analysis: breast area (total area within the outline of the breast), dense area (colored area), percent densities (the ratio of dense area and breast area), and the nondense area (the difference between the breast area and the dense area). Body mass index (BMI) was computed using the formula weight (kg) divided by the height (m²). We calculated intraclass correlation coefficients (icc) and weighted Kappa statistics (&#x003ba;_w) using quartile ranking to compare dietary intake from the two questionnaires. After adjustment for total energy intake using the residual method [<xref ref-type="bibr" rid="B17">17</xref>], we computed means for the mammographic parameters by quartile of soy intake. Trend tests regressing the mean of each category on soy intake were performed to investigate a possible relation between soy intake and the mammographic parameters. Multiple linear regression with a stepwise selection procedure was applied to explore the relation of soy intake, BMI, and reproductive factors with mammographic parameters.</p></sec><sec><title>Results:</title><p>Complete data were available for 514 women, of whom more than one-half were postmenopausal (Table <xref ref-type="table" rid="T1">1</xref>). Mean soy food intake was considerably higher according to the SQ than the FFQ, but the estimates from the two instruments were highly correlated. The icc for all soy foods combined was 0.70 (95% confidence interval [CI], 0.59-0.77) and &#x003ba;_w was 0.71 (95% CI, 0.66-0.76). The same was true for tofu intake from the SQ and the FFQ: the icc was 0.82 (95% CI, 0.79-0.85) and &#x003ba;_w was 0.77 (95% CI, 0.73-0.81). The means of tofu intake for each ethnic group, as estimated from the two dietary instruments, were very similar. We observed considerable ethnic differences in the consumption of meat, rice, alcohol, and soy foods (Table <xref ref-type="table" rid="T1">1</xref>). Counting all soy foods, Chinese and Japanese women reported nearly twice as high an intake as Caucasian women. Native Hawaiian women had the highest BMI, the greatest caloric intake, and consumed more soy foods than the Asian women. Tofu and soy milk/drinks were the most commonly reported soy foods. Green soybeans and soybean sprouts were eaten in moderate amounts, whereas the other products, miso, tofu skin, Western vegetarian meats, and fermented soybeans, were reported infrequently.</p><p>All mammographic parameters (Table <xref ref-type="table" rid="T1">1</xref>) except the size of the dense area were statistically significantly different by ethnicity (<italic>P</italic> &#x0003c; 0.001). Breast area, nondense area, and percent densities showed a significant trend by calorie adjusted quartiles of soy intake (Table <xref ref-type="table" rid="T2">2</xref>). The average breast area and the nondense area were approximately 25% smaller among women in the highest soy intake category than among women in the lowest category, whereas percent densities were 11% greater with higher soy intake. The dense area appeared to be slightly smaller with increasing soy intake, but the trend test was not significant. We observed no significant trends in the Chinese and Japanese women after we stratified by ethnic group. Dense area and percent densities among Caucasian and Native Hawaiian women showed a positive relation with soy intake. In a linear regression model, we observed a weak association between calorie adjusted soy intake and percent densities (partial regression coefficient <italic>b</italic> = 0.009, <italic>P</italic> = 0.004) after adjustment for age, BMI, parity, and family history of breast cancer. On the contrary, the size of the nondense area (ie the fatty part of the breast) was inversely related to soy intake after adjustment for BMI, age, and parity (<italic>b</italic> = -0.06, <italic>P</italic> = 0.002).</p></sec><sec><title>Discussion:</title><p>This study among healthy pre- and post-menopausal women who participated in mammography screening detected an inverse relation between self-reported soy food intake and the size of the breast, in particular the nondense area, as measured in mammographic images. Contrary to our hypothesis, we detected a positive association between soy intake and percent densities, and also higher levels of percent densities among Chinese and Japanese women who are at lower risk for breast cancer. This could be interpreted as evidence that soy increases breast cancer risk as a result of its estrogenic effects [<xref ref-type="bibr" rid="B3">3</xref>]. However, an alternate explanation may be that the size of the nondense area, which is inversely related to soy, is important in determining breast cancer risk. Mammary gland mass [<xref ref-type="bibr" rid="B18">18</xref>] has been proposed as a risk factor for breast cancer because of the higher number of ductal stem cells that develop as a result of intrauterine influences and high energy intake early in life. Although we did not collect information on childhood soy consumption, we wish to propose that soy foods, either by themselves or as part of a Asian dietary pattern characterized by low fat, high cereal, low dairy and low red meat intake, may contribute to the ethnic differences in mammographic characteristics. The importance of childhood nutrition for adult anthropometric measures is supported by comparative growth studies [<xref ref-type="bibr" rid="B19">19</xref>].</p><p>The finding of a nutritional effect on breast development before puberty agrees with the results from migrant studies demonstrating that it takes two or more generations to increase breast cancer risk. The time lag indicates that only descendants of the migrants who were born and who grew up in the host country will develop a similar risk as the host population, and that possible causes act early in life or <italic>in utero</italic> [<xref ref-type="bibr" rid="B20">20</xref>]. Evidence exists in addition to suggestions from migration studies that exposures during the perinatal period and adolescence are important for breast cancer risk later in life. A Western diet, as opposed to an Asian diet, may lead to increased body fat mass [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B21">21</xref>] during preadolescence mediated through more available sex steroids. Alternatively, the high soy content of Asian diets may inhibit the secretion of sex steroids and its effects on body fat through the antiestrogenic effects of isoflavones [<xref ref-type="bibr" rid="B22">22</xref>].</p><p>We purposefully did not adjust our analyses for ethnicity because we wished to explore nutritional factors that may determine the ethnic differences. The advantage of including several ethnic groups with different dietary habits was a greater variability in soy intake than could ever be found in an ethnically homogeneous population. The higher percent densities among Asian women do not necessarily contradict the mammographic density hypothesis. It appears from a case-control study [<xref ref-type="bibr" rid="B23">23</xref>] that the relation between percent densities and breast cancer is of similar strength in women of different ethnicities, although density levels vary by ethnicity.</p><p>Because of the lack of a lifetime history of soy intake in our study, we do not know whether soy consumption during early life when breast development occurred was similar to the current intake. Caucasian women who reported soy intake probably started eating these foods later in life and not during childhood. Soy may therefore have not had the opportunity to influence the appearance of their mammograms. Soy consumption, however, may be a marker for Chinese and Japanese ethnicity; an indicator for childhood nutritional patterns that include, but are not limited to, soy foods. These dietary patterns may explain some of the anthropometric characteristics found in adult women.</p><p>In conclusion, this examination of mammographic characteristics detected some associations with soy foods that suggest an influence of nutritional patterns on the development of the female breast. What implications, if any, these findings have for breast cancer risk has to be determined in prospective or interventions studies that will investigate possible mechanisms of actions, such as hormone levels, differences in estrogen metabolism, other dietary parameters, and genetic polymorphisms.</p></sec></sec><sec><title>Introduction</title><p>Breast cancer risk differs greatly by ethnicity. The US breast cancer incidence rates per 100,000 women in 1988-1992 [<xref ref-type="bibr" rid="B1">1</xref>] (invasive cases only, age-adjusted to 1970 US population) were 112 for Caucasian women, 106 for Native Hawaiian women, 82 for Japanese women, and 55 for Chinese women. Incidence rates between 23 and 31 per 100,000 (age-adjusted to the World Standard population) have been reported in Japan [<xref ref-type="bibr" rid="B2">2</xref>] for the same period. The role of isoflavones contained in soy products, based on these risk differences and on results from cell and animal studies [<xref ref-type="bibr" rid="B3">3</xref>], has been investigated in breast cancer prevention and produced conflicting results. A case-control study among Asian-American women [<xref ref-type="bibr" rid="B4">4</xref>] detected a 30% decreased risk of pre- and post-menopausal women breast cancer for women who reported eating tofu more than once a week as compared with women who ate tofu less than once a month. In a study from Singapore [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>], consumption of 55 g or more soy products per day was protective in premenopausal but not in postmenopausal women. Studies from China [<xref ref-type="bibr" rid="B7">7</xref>] and Japan [<xref ref-type="bibr" rid="B8">8</xref>] did not detect a protective effect of soy intake. However, two recent studies reported a lower breast cancer risk with increasing isoflavonoid excretion in urine [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>].</p><p>Mammographic densities that refer to the distribution of fat, connective, and epithelial tissue in the female breast have been shown to be related to breast cancer risk [<xref ref-type="bibr" rid="B11">11</xref>]. These densities are not abnormalities, but variations of healthy breast tissue. Fat appears dark on mammograms, whereas the radiographically light areas representing epithelial and connective tissue are relevant to breast cancer risk. A high percentage of dense parenchyma on mammographic images, 50-75% as compared with 0-10% of densities depending on the study, appears to confer a fourfold risk of developing breast cancer [<xref ref-type="bibr" rid="B11">11</xref>]. The purpose of this report is to investigate the hypothesis that soy foods are related to lower mammographic densities among a population of women with Caucasian, Chinese, Japanese, and Native Hawaiian ancestry living in Hawaii.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Recruitment and data collection</title><p>In a cross-sectional study design, women receiving annual screening mammograms were recruited at five mammography facilities on the island of Oahu. These clinics together provide more than one-half of all mammograms on the island (Cancer Research Center of Hawaii, unpublished data). As a result of greater than 90% insurance coverage among Hawaii's population [<xref ref-type="bibr" rid="B24">24</xref>] and a 1991 legislative mandate for health plans to cover annual screening mammography for women aged 40 years and older, self-reported mammography utilization [<xref ref-type="bibr" rid="B25">25</xref>] has been greater than 70%. Participation has been similar among Caucasian, Native Hawaiian, and Japanese women [<xref ref-type="bibr" rid="B25">25</xref>]. The Committee on Human Subjects at the University of Hawaii and the research boards of all participating institutions approved the study protocol. Women who did not speak English, women who reported a history of breast cancer or augmentation surgery, and women with suspicious lesions on the mammograms were excluded from the study, but women taking oral contraceptives and hormone replacement therapy were eligible to participate.</p></sec><sec><title>Dietary data collection</title><p>All study participants signed informed consent and completed a validated FFQ [<xref ref-type="bibr" rid="B12">12</xref>] especially designed for a multiethnic population [<xref ref-type="bibr" rid="B13">13</xref>]. The FFQ was self-administered and processed by optical scanning. The goal, as described in detail elsewhere [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>], was to include a sufficient number of food items so that at least 85% of the intake for nutrients of interest was captured. Analyses were performed to estimate intakes of 130 macro- and micronutrients for every study participant. Basic nutrient data from the US Department of Agriculture [<xref ref-type="bibr" rid="B26">26</xref>] has been supplemented and continuously updated with information from local recipes that are consumed by the different ethnic groups in Hawaii. Unpublished data also found acceptable validity for Native Hawaiian women, in addition to a published calibration study [<xref ref-type="bibr" rid="B12">12</xref>] that included Caucasian and Japanese women. Because the FFQ included only two soy food items, one each for tofu and Western vegetarian meats, we designed an additional SQ with 10 categories of soy foods commonly consumed in Hawaii: tofu, fried tofu, pressed tofu, green soybeans, miso, soybean sprouts, tofu skin, soy milk/drink, vegetarian meats, and fermented soybeans. Intake of soy foods according to the SQ in a validation study among 102 women was highly correlated with urinary isoflavone excretion [<xref ref-type="bibr" rid="B14">14</xref>].</p></sec><sec><title>Mammogram density assessment</title><p>All participating mammography clinics were accredited by the Food and Drug Administration and by the American College of Radiology. Mammography was performed on dedicated mammographic units with high contrast mammographic film and dedicated film processing. All mammograms for this study were therefore of comparable quality. Both cranio-caudal views of the mammogram were obtained from the mammography clinics after the radiologic evaluation had been completed and any malignancy ruled out. The films were scanned into a PC using an X-ray digitizer (Cobrascan CX-612-T, Radiographic Digital Imaging, Compton, CA, USA) at a resolution of 300 dots per inch. Computer-assisted mammographic density assessment was performed using a method that was first developed in Toronto [<xref ref-type="bibr" rid="B15">15</xref>] and later modified at the University of Southern California in Los Angeles [<xref ref-type="bibr" rid="B16">16</xref>]. The reader first draws the outline of the breast (using an outlining tool) and then searches for the best threshold gray level value <italic>X</italic> where all pixels with values above <italic>X</italic> are considered to represent mammographic densities. The pixel count corresponding to the area colored within the outline of the breast is determined by the computer, as is the total area within the outline of the breast. The readers were blinded to the identity of the subjects and to the clinic where the mammogram was taken. The correlation coefficient between the two readers for the readings of the size of the dense areas was 0.92. Four mammographic parameters were used for analysis: breast area (total area within the outline of the breast), dense area (colored area), percent densities (the ratio of dense area and breast area), and the nondense area (the difference between the breast area and the dense area).</p></sec><sec><title>Statistical analysis</title><p>We transformed non-normally distributed variables such as breast area, dense area, and calories using their natural logarithms. BMI was computed using the formula weight (kg) divided by the height (m²). Because breast cancer risk is low for Japanese and Chinese women and high in Caucasian and Native Hawaiian women, we divided the population into these two groups for analysis. The small number of women did not allow analysis for every ethnic group individually. The reason for including different ethnic groups into this study was, more importantly, to increase variability in exposure (ie soy intake) not to study the effect of soy foods in each ethnic group separately. We calculated icc and &#x003ba;_w values [<xref ref-type="bibr" rid="B27">27</xref>] using quartile ranking to compare dietary intake from the two questionnaires. A categorical soy intake variable was created based on quartiles of the calorie adjusted soy intake as measured by the SQ. Adjustment for total energy intake used the residual method [<xref ref-type="bibr" rid="B17">17</xref>]. After entering soy intake into a regression model as the dependent variable and calories as the independent variable, we added the residuals from this model to the median caloric intake and to the intercept to obtain adjusted intake estimates. We then computed arithmetical means for the mammographic parameters by quartile of soy intake. Trend tests regressing the mean of each category on soy intake were performed to investigate a possible relation between soy intake and the mammographic parameters. Multiple linear regression with a stepwise selection procedure [<xref ref-type="bibr" rid="B28">28</xref>] was applied to explore the relation of soy intake, BMI, and reproductive factors with mammographic parameters. All analyses were performed using PC-SAS^&#x000ae;, release 6.12 (SAS Institute, Cary, NC, USA).</p></sec></sec><sec><title>Results</title><sec><title>Dietary intake</title><p>Complete data were available for 514 women, of whom more than one-half were postmenopausal. Caucasian and Japanese women represented the largest groups in this study (Table <xref ref-type="table" rid="T1">1</xref>). Educational achievement was fairly high across all ethnic groups. BMI and daily caloric intake were highest among Native Hawaiian women, and lowest among Chinese and Japanese women. Mean soy food intake was considerably higher according to the SQ than the FFQ, but the estimates from the two instruments were highly correlated. The icc for all soy foods combined was 0.70 (95% CI, 0.59-0.77) and &#x003ba;_w was 0.71 (95% CI, 0.66-0.76). The same was true for tofu intake from the SQ and the FFQ: the icc was 0.82 (95% CI, 0.79-0.85) and &#x003ba;_w was 0.77 (95% CI, 0.73-0.81). The means of tofu intake as estimated from the two dietary instruments were very similar for each ethnic group.</p><p>We observed considerable ethnic differences in the consumption of meat, rice, alcohol, and soy foods. Whereas Caucasians consumed more alcohol than all other groups, they reported the lowest rice and soy intake. Interestingly, meat intake was equally low in Caucasian as in Chinese women. Chinese and Japanese women reported nearly twice as high an intake as Caucasian women, counting all soy foods. The Native Hawaiian group, however, consumed even more soy foods than did the Asian women. Tofu and soy milk/drinks were the most commonly reported soy foods. Green soybeans and soybean sprouts were eaten in moderate amounts, whereas the other products, miso, tofu skin, Western vegetarian meats, and fermented soybeans, were reported infrequently.</p></sec><sec><title>Mammographic characteristics</title><p>All mammographic parameters (Table <xref ref-type="table" rid="T1">1</xref>) except the size of the dense area were statistically significantly different by ethnicity (<italic>P</italic> &#x0003c; 0.001). Breast area and nondense area were considerably larger among Caucasian and Native Hawaiian women than among Chinese and Japanese women. Although the dense areas were slightly smaller among women of Asian descent, their percent densities were greater than for Caucasian and Native Hawaiian women because of the smaller breast area among women of Asian descent.</p></sec><sec><title>Soy intake and mammographic characteristics</title><p>Breast area, nondense area, and percent densities showed a significant trend by calorie adjusted quartiles of soy intake (Table <xref ref-type="table" rid="T2">2</xref>). The average breast area and the non-dense area were approximately 25% smaller among women in the highest soy intake category than among women in the lowest category, whereas percent densities were 11% greater with higher soy intake. The dense areas appeared to be slightly smaller with increasing soy intake, but the trend test was not significant. We observed no significant trends in the Chinese and Japanese women after we stratified by ethnic group. Dense areas and percent densities among Caucasian and Native Hawaiian women showed a weak positive relation with soy intake. Excluding women who reported less than 800 calories or more than 4000 calories did not change these results substantially. We obtained similar results in a linear regression model with adjustment for confounding variables. Whereas we did not observe a relation between soy intake and the size of the dense areas, we found an association between calorie adjusted soy intake and percent densities (<italic>b</italic> = 0.009, <italic>P</italic> = 0.004) after adjustment for age, BMI, parity, and family history of breast cancer. On the contrary, the size of the nondense area (ie the fatty part of the breast) was inversely related to soy intake after adjustment for BMI, age, and parity (<italic>b</italic> = - 0.06, <italic>P</italic> = 0.002). The effects were in the same direction for pre- and postmenopausal women, but they were stronger for postmenopausal women.</p></sec></sec><sec><title>Discussion</title><p>This study among healthy pre- and postmenopausal women who participated in mammography screening detected an inverse relation between self-reported soy food intake and the size of the breast, in particular the non-dense area, as measured in mammographic images. Contrary to our hypothesis, we detected a positive association between soy intake and percent densities, and also higher levels of percent densities among Chinese and Japanese women who are at lower risk for breast cancer. This could be interpreted as evidence that soy increases breast cancer risk as a result of its estrogenic effects [<xref ref-type="bibr" rid="B3">3</xref>]. An alternative explanation, however, may be that the size of the nondense areas, which is inversely related to soy, is important in determining breast cancer risk. Mammary gland mass [<xref ref-type="bibr" rid="B18">18</xref>] has been proposed as a risk factor for breast cancer because of the higher number of ductal stem cells that develop as a result of intrauterine influences and high energy intake early in life. Although we did not collect information on childhood soy consumption, we wish to propose that soy foods, either by themselves or as part of an Asian dietary pattern characterized by low fat, high cereal, low dairy and low red meat intake, may contribute to the ethnic differences in mammographic characteristics. Dietary patterns, as shown in several publications [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>], can be identified from nutritional information using factor analysis. We purposefully did not adjust our analyses for ethnicity because we wished to explore nutritional factors that may determine the ethnic differences.</p><p>We hypothesize, as a possible mechanism of how soy intake may affect mammographic characteristics, that soy intake during developmental years may be responsible for the smaller fat component of the breast through an effect of soy on the growth of the female breast during that period. The importance of childhood nutrition for adult anthropometric measures is supported by comparative growth studies [<xref ref-type="bibr" rid="B19">19</xref>]. A nutritional effect on breast development before puberty agrees with the results from migrant studies demonstrating that it takes two or more generations to increase breast cancer risk [<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>]. The time lag indicates that only descendants of the migrants who were born and who grew up in the host country will develop a similar risk as the host population, and that possible environmental and behavioral causes act early in life or <italic>in utero</italic> [<xref ref-type="bibr" rid="B20">20</xref>]. In addition to suggestions from migration studies, evidence exists that exposures during the perinatal period and adolescence are important for breast cancer risk later in life. The occurrence of high risk mammographic patterns, for example, was associated with the weight of the placenta, the main estrogen-producing organ during pregnancy [<xref ref-type="bibr" rid="B33">33</xref>]. A Western diet, as opposed to an Asian diet, may lead to increased body fat mass [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B21">21</xref>] during preadolescence, mediated through more available sex steroids. The high soy content of Asian diets may alternatively inhibit the secretion of sex steroids and its effects on body fat through the antiestrogenic effects of isoflavones [<xref ref-type="bibr" rid="B22">22</xref>].</p><p>The higher percent densities among Asian women do not necessarily contradict the mammographic density hypothesis. It appears from a recent case-control study [<xref ref-type="bibr" rid="B23">23</xref>] that the relation between percent densities and breast cancer is of similar strength in women of different ethnicities although density levels vary by ethnicity. Although the relation of soy intake and mammographic densities has not been investigated previously, several dietary studies have shown weak associations between micro- and macronutrients and mammographic densities. Energy-adjusted saturated fat intake was positively related, and fiber and carotenoids were inversely related to mammographic densities in a case-control study [<xref ref-type="bibr" rid="B34">34</xref>]. Areas of density decreased by 6.1% in the intervention group, as compared with 2.1% in the control group, during a low fat intervention [<xref ref-type="bibr" rid="B35">35</xref>]. Lower cholesterol and saturated fat intake, in particular, predicted a decrease in mammographic density [<xref ref-type="bibr" rid="B36">36</xref>]. Higher fat intake [<xref ref-type="bibr" rid="B37">37</xref>] and high density lipoprotein cholesterol levels [<xref ref-type="bibr" rid="B38">38</xref>] were related to denser mammographic patterns, suggesting that breast cancer risk may be related to blood lipid levels. Small differences in mammographic densities across levels of vitamin B₁₂, C, and E as well as polyunsaturated fat intake were described recently [<xref ref-type="bibr" rid="B39">39</xref>] from a large cohort of Caucasian women.</p><p>Differences in the quality of films between clinics and subjects may have introduced some measurement error. To minimize the subjective component of the assessment method, we spent much effort in training the readers and compared the results between the readers frequently. The high correlation coefficient indicates a high level of standardization in mammographic assessment. Given the measurement errors in dietary and mammographic density assessment, the multi-factorial determination of mammographic densities, and the recall bias inherent to all dietary studies, it is difficult to identify a small effect. Because the soy food relation was the primary aim in the proposal for this study, we did not search for positive results among a multitude of food items, but focused on the soy hypothesis.</p><p>The lack of a lifetime history of soy intake and the cross-sectional design limit our study's ability to establish causal relations. The dietary assessment asked only about intake during the past year. We therefore do not know whether soy consumption during earlier life when breast development occurred was similar to the current intake. Because Caucasian women who reported soy intake probably started eating these foods later in life and not during childhood, soy may have not had the opportunity to influence the appearance of their mammograms. Soy consumption may, however, be a marker for Chinese and Japanese ethnicity; an indicator for childhood nutritional patterns that include, but are not limited to, soy foods. These dietary patterns may explain some of the anthropometric characteristics found in adult women.</p><p>One strength of this study was the opportunity to investigate women with different ethnicity living in a similar environment that has a variety of foods available to persons with all ancestries. The advantage of including several ethnic groups with different dietary habits was a greater variability in soy intake than could ever be found in an ethnically homogeneous population. Because 91% of study participants were born in the United States, the majority of women also lived through adolescence under similar conditions. The comparable socioeconomic status and equal access to mammography screening among women with Caucasian and Asian ancestry also offered a unique opportunity to compare mammographic parameters. In conclusion, this examination of mammographic characteristics detected some associations with soy foods that suggest an influence of nutritional patterns on the development of the female breast. What implications, if any, these findings have for breast cancer risk has to be determined in prospective or interventions studies that will investigate possible mechanisms of actions, such as hormone levels, differences in estrogen metabolism, other dietary parameters, and genetic polymorphisms.</p></sec>

Regulation of Mitotic Inhibitor Mik1 Helps to Enforce the DNA Damage Checkpoint

<p>The protein kinase Chk1 enforces the DNA damage checkpoint. This checkpoint delays mitosis until damaged DNA is repaired. Chk1 regulates the activity and localization of Cdc25, the tyrosine phosphatase that activates the cdk Cdc2. Here we report that Mik1, a tyrosine kinase that inhibits Cdc2, is positively regulated by the DNA damage checkpoint. Mik1 is required for checkpoint response in strains that lack Cdc25. Long-term DNA damage checkpoint arrest fails in <italic>&#x00394;mik1</italic> cells. DNA damage increases Mik1 abundance in a Chk1-dependent manner. Ubiquitinated Mik1 accumulates in a proteasome mutant, which indicates that Mik1 normally has a short half-life. Thus, the DNA damage checkpoint might regulate Mik1 degradation. Mik1 protein and mRNA oscillate during the unperturbed cell cycle, with peak amounts detected around S phase. These data indicate that regulation of Mik1 abundance helps to couple mitotic onset to the completion of DNA replication and repair. Coordinated negative regulation of Cdc25 and positive regulation of Mik1 ensure the effective operation of the DNA damage checkpoint.</p>

<contrib contrib-type="author"><name><surname>Baber-Furnari</surname><given-names>Beth A.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Rhind</surname><given-names>Nick</given-names></name></contrib><contrib contrib-type="author"><name><surname>Boddy</surname><given-names>Michael N.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Shanahan</surname><given-names>Paul</given-names></name></contrib><contrib contrib-type="author"><name><surname>Lopez-Girona</surname><given-names>Antonia</given-names></name></contrib><contrib contrib-type="author"><name><surname>Russell</surname><given-names>Paul</given-names></name><xref ref-type="author-notes" rid="FN150">&#x0002a;</xref></contrib>

Molecular Biology of the Cell

<sec><title>INTRODUCTION</title><p>In response to DNA damage or incomplete DNA synthesis, eukaryotic cells delay the onset of mitosis by activation of mitotic checkpoints (<xref ref-type="bibr" rid="B20">Hartwell and Weinert, 1989</xref>; <xref ref-type="bibr" rid="B11">Elledge, 1996</xref>; <xref ref-type="bibr" rid="B39">Rhind and Russell, 1998a</xref>). These checkpoints enhance genome integrity by ensuring that chromosomes are fully replicated and repaired before nuclear division. Genomic instability arising from checkpoint defects may lead to cancer (<xref ref-type="bibr" rid="B18">Hartwell, 1992</xref>; <xref ref-type="bibr" rid="B19">Hartwell and Kastan, 1994</xref>). Moreover, checkpoints influence the response of tumor cells to radiotherapy and chemotherapy protocols that damage DNA or inhibit DNA replication. Thus, understanding checkpoint mechanisms is a major priority of current studies that investigate cell cycle control or cancer.</p><p>Fundamental insights into DNA damage checkpoints have arisen from studies of the fission yeast <italic>Schizosaccharomyces pombe</italic> (<xref ref-type="bibr" rid="B41">Russell, 1998</xref>). Chk1, a protein kinase that is essential for DNA damage checkpoint arrest, was discovered in fission yeast (<xref ref-type="bibr" rid="B52">Walworth <italic>et al.</italic>, 1993</xref>; <xref ref-type="bibr" rid="B2">al-Khodairy <italic>et al.</italic>, 1994</xref>). Chk1 functions downstream of a group of &#x0201c;checkpoint Rad&#x0201d; proteins that includes Rad1, Rad3, Rad9, Rad17, Rad26, and Hus1. The functions of checkpoint Rad proteins are poorly understood, as is the regulation of Chk1. However, it is known that DNA damage stimulates Chk1 phosphorylation by a process that requires the checkpoint Rad proteins Rad4/Cut5 and Rhp9/Crb2 (<xref ref-type="bibr" rid="B46">Saka and Yanagida, 1993</xref>; <xref ref-type="bibr" rid="B53">Walworth and Bernards, 1996</xref>; <xref ref-type="bibr" rid="B45">Saka <italic>et al.</italic>, 1997</xref>; <xref ref-type="bibr" rid="B54">Willson <italic>et al.</italic>, 1997</xref>). Cdc2, the cdk that triggers mitosis, is the ultimate target of checkpoint regulation (<xref ref-type="bibr" rid="B38">Rhind <italic>et al.</italic>, 1997</xref>). Cdc2 is inhibited by phosphorylation on tyrosine 15 catalyzed by the protein kinases Wee1 and Mik1. Cdc25, the tyrosine phosphatase that activates Cdc2, is an important substrate of Chk1 (<xref ref-type="bibr" rid="B15">Furnari <italic>et al.</italic>, 1997a</xref>; <xref ref-type="bibr" rid="B37">Peng <italic>et al.</italic>, 1997</xref>). Chk1 regulates Cdc25 by two mechanisms. One mechanism is direct inhibition of Cdc25 phosphatase activity directed toward Cdc2 (<xref ref-type="bibr" rid="B3">Blasina <italic>et al.</italic>, 1999</xref>; <xref ref-type="bibr" rid="B14">Furnari <italic>et al.</italic>, 1999</xref>). The second mechanism involves stimulated association with 14-3-3 proteins, such as Rad24 in fission yeast, which leads to net nuclear export of Cdc25 and consequent exclusion from the nuclear pool of Cdc2/cyclin B (<xref ref-type="bibr" rid="B27">Lopez-Girona <italic>et al.</italic>, 1999</xref>).</p><p>The checkpoint Rad proteins, but not Chk1 or Crb2, are required for S&#x02013;M replication checkpoint. This checkpoint prevents mitosis when DNA replication is slowed, e.g., with the drug hydroxyurea, an inhibitor of ribonucleotide reductase (<xref ref-type="bibr" rid="B1">al-Khodairy and Carr, 1992</xref>; <xref ref-type="bibr" rid="B52">Walworth <italic>et al.</italic>, 1993</xref>; <xref ref-type="bibr" rid="B2">al-Khodairy <italic>et al.</italic>, 1994</xref>; <xref ref-type="bibr" rid="B45">Saka <italic>et al.</italic>, 1997</xref>; <xref ref-type="bibr" rid="B54">Willson <italic>et al.</italic>, 1997</xref>). The protein kinase Cds1, instead of Chk1, functions downstream of checkpoint Rad proteins to enforce the S&#x02013;M checkpoint (<xref ref-type="bibr" rid="B33">Murakami and Okayama, 1995</xref>; <xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B26">Lindsay <italic>et al.</italic>, 1998</xref>). Like the DNA damage checkpoint (<xref ref-type="bibr" rid="B38">Rhind <italic>et al.</italic>, 1997</xref>), the S&#x02013;M replication checkpoint prevents mitosis by maintaining Cdc2 in an inhibited, tyrosine phosphorylated state (<xref ref-type="bibr" rid="B12">Enoch and Nurse, 1990</xref>; <xref ref-type="bibr" rid="B40">Rhind and Russell, 1998b</xref>). In a manner similar to Chk1, Cds1 phosphorylates and thereby inhibits Cdc25 (<xref ref-type="bibr" rid="B57">Zeng <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B14">Furnari <italic>et al.</italic>, 1999</xref>). Cds1 is also required for the large accumulation of Mik1 protein that occurs in hydroxyurea-treated cells (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>). It is thought that inhibition of Cdc25 function and accumulation of Mik1 both contribute to enforcement of the replication checkpoint.</p><p>Checkpoint mechanisms uncovered in studies of <italic>S. pombe</italic> appear to be conserved among most eukaryotes, including humans. DNA database searches have identified human homologues of <italic>chk1</italic>^&#x0002b; and <italic>cds1</italic>^&#x0002b; (<xref ref-type="bibr" rid="B47">Sanchez <italic>et al.</italic>, 1997</xref>; <xref ref-type="bibr" rid="B29">Matsuoka <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B3">Blasina <italic>et al.</italic>, 1999</xref>; <xref ref-type="bibr" rid="B8">Brown <italic>et al.</italic>, 1999</xref>). These homologues inhibit Cdc25 and become activated or hyperphosphorylated in response to DNA damage (<xref ref-type="bibr" rid="B47">Sanchez <italic>et al.</italic>, 1997</xref>; <xref ref-type="bibr" rid="B29">Matsuoka <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B3">Blasina <italic>et al.</italic>, 1999</xref>). Moreover, DNA damage in human cells leads to inhibition of Cdc25 activity by a process that requires ATM, a kinase related to Rad3 in fission yeast (<xref ref-type="bibr" rid="B3">Blasina <italic>et al.</italic>, 1999</xref>). In mammalian or <italic>Xenopus</italic> cells grown in culture, nuclear exclusion of Cdc25 requires association with 14-3-3 proteins, an observation that suggests a potential checkpoint role for 14-3-3 proteins (<xref ref-type="bibr" rid="B10">Dalal <italic>et al.</italic>, 1999</xref>; <xref ref-type="bibr" rid="B24">Kumagai and Dunphy, 1999</xref>; <xref ref-type="bibr" rid="B56">Yang <italic>et al.</italic>, 1999</xref>). In <italic>Xenopus</italic> egg extracts, activation of the DNA replication checkpoint induces stabilization of exogenous Wee1 added to the extract (<xref ref-type="bibr" rid="B30">Michael and Newport, 1998</xref>), an observation similar to that of Mik1 accumulation in fission yeast cells treated with hydroxyurea (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>).</p><p>Regulation of Cdc25 by Chk1 appears to be very important for the DNA damage checkpoint. However, this fact does not exclude the possibility that the Cdc2-directed kinases Wee1 and Mik1 might be regulated in a positive manner by the DNA damage checkpoint. Indeed, Wee1 appears to be hyperphosphorylated in UV-irradiated cells, although the significance of this phosphorylation is unknown (<xref ref-type="bibr" rid="B35">O'Connell <italic>et al.</italic>, 1997</xref>). Here we report that Mik1 is regulated by the DNA damage checkpoint. Genetic and physiological evidence shows that Mik1 is important for maintenance of the DNA damage checkpoint and that Mik1 protein abundance increases in cells arrested at the checkpoint. Mik1 also accumulates in the nucleus during an unperturbed S phase. These data indicate that Mik1 regulation helps to ensure that the onset of mitosis is coupled to the completion of DNA replication during the normal cell cycle or completion of DNA repair in cells that have suffered DNA damage.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><sec><title>General Methods</title><p>Genetic and biochemical methods for studying fission yeast were used as described (<xref ref-type="bibr" rid="B31">Moreno <italic>et al.</italic>, 1991</xref>). All strains were grown in yeast extract&#x02013;glucose medium at 30&#x000b0;C unless indicated otherwise. Cells were synchronized by centrifugal elutriation at 30&#x000b0;C with a Beckman (Fullerton, CA) JE-5.0 elutriation rotor. DNA damage was inflicted from a ¹³⁷C source at 3 Gy min^&#x02212;1 for 67 min or by the addition of 2.5 mU ml^&#x02212;1 bleomycin sulfate. Cells were scored for progression through mitosis by microscopic observation (<xref ref-type="bibr" rid="B38">Rhind <italic>et al.</italic>, 1997</xref>). Indirect immunofluorescence studies were performed as described (<xref ref-type="bibr" rid="B27">Lopez-Girona <italic>et al.</italic>, 1999</xref>). Northern blot analysis was performed with 10 &#x003bc;g of total RNA with random primed ³²P-labeled <italic>mik1</italic>^&#x0002b; or <italic>leu1</italic>^&#x0002b; probe as described (<xref ref-type="bibr" rid="B16">Furnari <italic>et al.</italic>, 1997b</xref>). Immunoblot analysis was performed with 50 &#x003bc;g of total cell lysate. Samples were electrophoresed on a 10&#x00025; SDS-PAGE gel and wet transferred to Immobilon (Millipore, Bedford, MA). Blots were probed with mouse mAbs to the Myc epitope (9E10; Santa Cruz Biotechnology, Santa Cruz, CA) or the PSTAIR peptide derived from Cdc2. Primary antibodies were detected with the use of an HRP-conjugated anti-mouse immunoglobulin G antibody (Promega, Madison, WI) and Luminol reagents (<named-content content-type="company" xlink:href="pierce">Pierce</named-content>, Rockford, IL). Proteins that were covalently linked to His₆-ubiquitin were purified in denaturing conditions as described (<xref ref-type="bibr" rid="B49">Shiozaki and Russell, 1997</xref>).</p></sec><sec><title>Yeast Strains</title><p><italic>S. pombe</italic> strains of the following genotypes were used in this study: PR109, wild type; GL192, <italic>cdc2-3w cdc25::ura4</italic>^&#x0002b;; NR1976, <italic>cdc2-3w cdc25::ura4</italic>^&#x0002b; <italic>chk1::ura4</italic>^&#x0002b;; NR1977, <italic>cdc2-3w cdc25::ura4</italic>^&#x0002b; <italic>mik1::ura4</italic>^&#x0002b;; PR712, <italic>mik1::ura4</italic>^&#x0002b;; BF1758, <italic>nmt1:GST-chk1:leu1</italic>^&#x0002b;; BF2404, <italic>wee1-50 cdc25::ura4</italic>^&#x0002b; <italic>nmt1:GST-chk1:leu1</italic>^&#x0002b;; NR1365, <italic>mik1::ura4</italic>^&#x0002b; <italic>nmt1:GST-chk1:leu1</italic>^&#x0002b;; NR1291, <italic>wee1-50 nmt1:GST-chk1:leu1</italic>^&#x0002b;; AL2405, <italic>mik1:13Myc:kan</italic>; BF2406, <italic>rad3::ura4</italic>^&#x0002b; <italic>mik1:13Myc:kan</italic>; BF2407, <italic>chk1::ura4</italic>^&#x0002b; <italic>mik1:13Myc:kan</italic>; BF2408, <italic>cds1::ura4</italic>^&#x0002b; <italic>mik1:13Myc:kan</italic>; BF2409, <italic>cdc25-22 mik1:13Myc:kan</italic>; BF2410, <italic>cdc25-22 chk1::ura4</italic>^&#x0002b; <italic>mik1:13Myc:kan</italic>; BF2442, <italic>cdc25-22 rad3::ura4</italic>^&#x0002b; <italic>mik1:13Myc:kan</italic>; and NB2411, <italic>mts3-1 mik1:13Myc:kan</italic>. All strains were <italic>leu1-32 ura4-D18</italic>.</p></sec></sec><sec><title>RESULTS</title><sec><title>Cdc25-independent Checkpoint Delay</title><p>To determine if Cdc25 is the sole target of Chk1, a careful analysis of the DNA damage checkpoint was performed with <italic>cdc2-3w &#x00394;cdc25</italic> cells. Cdc25 is normally essential for division, but the <italic>cdc2-3w</italic> mutation, which is a dominant activating allele, bypasses the requirement for Cdc25 (<xref ref-type="bibr" rid="B42">Russell and Nurse, 1986</xref>). Cells that have the <italic>cdc2-3w</italic> mutation divide at a cell length of &#x0223c;8 &#x003bc;m, whereas <italic>cdc2-3w &#x00394;cdc25</italic> cells divide at &#x0223c;18 &#x003bc;m. These facts are consistent with the observation that DNA damage, which leads to inhibition of Cdc25, causes a substantial checkpoint delay in <italic>cdc2-3w</italic> cells (<xref ref-type="bibr" rid="B48">Sheldrick and Carr, 1993</xref>). Importantly, Cdc2 protein encoded by <italic>cdc2-3w</italic> is responsive to changes in the activity of kinases that phosphorylate Cdc2 on tyrosine 15 (<xref ref-type="bibr" rid="B44">Russell and Nurse, 1987b</xref>). A synchronous population of <italic>cdc2-3w &#x00394;cdc25</italic> cells in G₂ phase was collected by centrifugal elutriation and immediately exposed to 200 Gy of ionizing radiation or mock treated. Irradiation caused an &#x0223c;45-min mitotic delay relative to the unirradiated control (Figure <xref ref-type="fig" rid="F1">1</xref>A). These findings showed that the DNA damage checkpoint is partially retained in <italic>cdc2-3w &#x00394;cdc25</italic> cells. To determine if Chk1 is required for this mitotic delay, a <italic>cdc2-3w &#x00394;cdc25 &#x00394;chk1</italic> culture was analyzed by the same experimental protocol (Figure <xref ref-type="fig" rid="F1">1</xref>B). The <italic>&#x00394;chk1</italic> mutation eliminated the mitotic delay induced by irradiation. The DNA damage checkpoint is abolished in cells that are unable to phosphorylate Cdc2 on tyrosine 15 (<xref ref-type="bibr" rid="B38">Rhind <italic>et al.</italic>, 1997</xref>). Therefore, these data suggested that Chk1 must regulate Wee1 or Mik1, the kinases that phosphorylate Cdc2 on tyrosine 15. </p></sec><sec><title>Mik1 Is Important for Division Arrest Induced by GST&#x02013;Chk1 Overproduction</title><p>Expression of large amounts of GST&#x02013;Chk1 fusion protein causes cell cycle arrest, mimicking a damage-induced checkpoint arrest (<xref ref-type="bibr" rid="B38">Rhind <italic>et al.</italic>, 1997</xref>). This phenotype also occurs in a <italic>&#x00394;wee1</italic> strain (<xref ref-type="bibr" rid="B15">Furnari <italic>et al.</italic>, 1997a</xref>). To confirm that Chk1 is able to delay mitosis independently of Cdc25 and Wee1, GST&#x02013;Chk1 was overproduced under the control of the thiamine-repressible <italic>nmt1</italic> promoter in a <italic>wee1-50 &#x00394;cdc25</italic> strain. These cells were grown at 35&#x000b0;C, a restrictive temperature that inactivates <italic>wee1-50</italic> gene product. Remarkably, GST&#x02013;Chk1 overexpression caused cell cycle arrest in <italic>wee1-50 &#x00394;cdc25</italic> cells (Figure <xref ref-type="fig" rid="F2">2</xref>A). These data implicated Mik1 as a target of Chk1 regulation. </p><p>To test the possibility that Mik1 is regulated by Chk1, a copy of the <italic>nmt1:GST-chk1</italic>^&#x0002b; construct was integrated in a <italic>&#x00394;mik1</italic> strain. Induction of GST&#x02013;Chk1 expression induced cell elongation but failed to cause cell cycle arrest in <italic>&#x00394;mik1</italic> cells (Figure <xref ref-type="fig" rid="F2">2</xref>B). In fact, <italic>&#x00394;mik1</italic> cells formed viable colonies in medium that induces GST&#x02013;Chk1 expression (Figure <xref ref-type="fig" rid="F2">2</xref>C). In contrast, overproduction of GST&#x02013;Chk1 caused cell cycle arrest in wild-type and <italic>wee1-50</italic> cells incubated at 32&#x000b0;C (Figure <xref ref-type="fig" rid="F2">2</xref>) and in <italic>&#x00394;wee1</italic> cells (<xref ref-type="bibr" rid="B15">Furnari <italic>et al.</italic>, 1997a</xref>). These findings demonstrated that Mik1 is important for the cell cycle arrest induced by GST&#x02013;Chk1 overproduction.</p></sec><sec><title>Damage Checkpoint Impaired by &#x00394;mik1 Mutation</title><p>The contribution of Mik1 to the G₂&#x02013;M damage checkpoint was determined by performing a long-term DNA damage checkpoint experiment with a <italic>&#x00394;mik1</italic> strain. A synchronous population of cells in G₂ phase collected by centrifugal elutriation was exposed to the radiomimetic drug bleomycin, which causes DNA double-strand breaks (<xref ref-type="bibr" rid="B23">Kostrub <italic>et al.</italic>, 1997</xref>). Bleomycin-treated <italic>&#x00394;mik1</italic> cells exhibited a mitotic delay of &#x0223c;100 min relative to mock-treated <italic>&#x00394;mik1</italic> or wild-type cells (Figure <xref ref-type="fig" rid="F3">3</xref>A). Unlike wild-type cells, <italic>&#x00394;mik1</italic> cells were unable to maintain a checkpoint arrest. In the continuous presence of bleomycin, all of the <italic>&#x00394;mik1</italic> cells completed mitosis by 240 min, whereas fewer than 10&#x00025; of wild-type cells had undergone mitosis by 320 min (Figure <xref ref-type="fig" rid="F3">3</xref>A). These findings demonstrated that Mik1 is required for a prolonged DNA damage checkpoint arrest. </p><p>Independent confirmation of the importance of Mik1 in the DNA damage checkpoint was provided by examination of the checkpoint response in a <italic>cdc2-3w &#x00394;cdc25 &#x00394;mik1</italic> strain. A synchronized culture <italic>cdc2-3w &#x00394;cdc25 &#x00394;mik1</italic> cells was exposed to ionizing radiation or mock treated (Figure <xref ref-type="fig" rid="F3">3</xref>B). The irradiated and mock-treated cultures completed mitosis with similar kinetics. These findings contrast with the behavior of <italic>cdc2-3w &#x00394;cdc25</italic> cells (Figure <xref ref-type="fig" rid="F1">1</xref>), in which irradiation caused a substantial delay of mitosis by a Chk1-dependent process. These studies supported the proposition that Mik1 is in some way positively regulated by Chk1.</p></sec><sec><title>Cell Cycle Regulation of Nuclear Mik1</title><p>Cellular localization of the mitotic activator, Cdc25, changes in response to DNA damage (<xref ref-type="bibr" rid="B27">Lopez-Girona <italic>et al.</italic>, 1999</xref>). This precedence prompted investigations of Mik1 localization in cells that have suffered DNA damage. Before undertaking these studies, we examined the localization of Mik1 during the normal cell cycle with the use of a strain that expressed a myc-tagged form of Mik1 (<italic>mik1:13Myc</italic>) from the <italic>mik1</italic>^&#x0002b; genomic locus. These cells underwent division at a normal size, and <italic>wee1-50 mik1:13Myc</italic> cells were viable at 36&#x000b0;C (our unpublished data); therefore, the myc-tagged form of Mik1 appeared to be fully functional. Mik1 was detected in the nucleus of binucleate cells and small cells that had recently completed cell division (Figure <xref ref-type="fig" rid="F4">4</xref>A). G₁ phase is normally very short in fission yeast; thus, DNA replication ensues almost immediately after nuclear division and is essentially complete when daughter cells have detached. Therefore, the nuclear staining pattern of Mik1 corresponds to S phase and early G₂. </p><p>The cytoplasmic signal detected in the myc-tagged Mik1 strain was comparable to the background signal observed in the untagged control strain (Figure <xref ref-type="fig" rid="F4">4</xref>A). This observation suggested that the periodic nuclear detection of Mik1 was determined by changes in protein abundance, as opposed to regulation of Mik1 subcellular localization. This hypothesis was tested by immunoblot measurements of Mik1 abundance in a synchronous cell culture. A temperature-sensitive <italic>cdc25-22</italic> strain was arrested in late G₂ phase by incubation at restrictive temperature, followed by a shift to permissive temperature, which caused synchronous resumption of cell cycle progression. Immunoblot analysis showed that Mik1 protein abundance oscillated during the cell cycle, with the peak signal coinciding with maximum septation index (Figure <xref ref-type="fig" rid="F4">4</xref>B). This pattern corresponds most closely to S phase in a <italic>cdc25-22</italic> arrest-and-release experiment. Northern blot analysis demonstrated that <italic>mik1</italic>^&#x0002b; mRNA abundance also oscillated during the cell cycle, with the peak signal occurring immediately before the maximum immunoblot signal for Mik1 (Figure <xref ref-type="fig" rid="F4">4</xref>B). Thus, the nuclear localization of Mik1 during S and early G₂ phases correlated with the appearance of <italic>mik1</italic>^&#x0002b; mRNA and protein.</p></sec><sec><title>Mik1 Accumulation Induced by Checkpoints</title><p>Immunofluorescence studies were performed to monitor Mik1 localization in cells arrested at DNA replication or damage checkpoints. These experiments used <italic>mik1:13Myc</italic> cells incubated for 4 h in 10 mM hydroxyurea (HU) or 2.5 mU of bleomycin (BL). Activation of the DNA replication checkpoint by hydroxyurea caused accumulation of Mik1 in the nucleus (Figure <xref ref-type="fig" rid="F5">5</xref>A). Essentially all cells treated with hydroxyurea displayed a nuclear Mik1 signal, whereas in asynchronous cultures only binucleate or septated cells in S, or short cells in early G₂, presented a Mik1 signal in the nucleus. The Mik1 signal in hydroxyurea-arrested cells was much more intense than that observed in asynchronous cells that were in S phase. The intense nuclear signal of Mik1 in hydroxyurea-arrested cells accords with previous immunoblot studies that demonstrated dramatic accumulation of Mik1 protein in cells arrested at the S&#x02013;M replication checkpoint (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>). These studies also showed that hydroxyurea-induced accumulation was largely dependent on Rad3 and Cds1 (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>). In agreement with the previous immunoblot studies, the Mik1 nuclear signal was substantially reduced in hydroxyurea-treated <italic>&#x00394;rad3</italic> or <italic>&#x00394;cds1</italic> cells relative the wild-type counterparts (Figure <xref ref-type="fig" rid="F5">5</xref>, B and C). The <italic>&#x00394;rad3</italic> cells failed to arrest in hydroxyurea and instead entered mitosis with incompletely replicated DNA. This checkpoint defect accounts for the large number of septated <italic>&#x00394;rad3</italic> cells in which DNA, visualized with the stain DAPI, is unequally segregated to the daughter cells (Figure <xref ref-type="fig" rid="F5">5</xref>B). As predicted (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>), the hydroxyurea-induced nuclear accumulation of Mik1 was undiminished in <italic>&#x00394;chk1</italic> cells (Figure <xref ref-type="fig" rid="F5">5</xref>D) </p><p>The asynchronous culture of <italic>&#x00394;rad3</italic> cells presented a Mik1 staining pattern that was generally similar to that in wild-type cells, i.e., the Mik1 nuclear signal was strongest in binucleate or septated cells (Figure <xref ref-type="fig" rid="F5">5</xref>B, left panels). This pattern indicated that the normal periodicity of Mik1 mRNA and protein accumulation during S and early G₂ was not dependent on Rad3. Likewise, unperturbed <italic>&#x00394;cds1</italic> or <italic>&#x00394;chk1</italic> cells displayed a Mik1 staining pattern that was very similar to that in wild-type cells. Interestingly, it appeared that an increased fraction of the uninucleate cells in the asynchronous <italic>&#x00394;rad3</italic> culture displayed a detectable Mik1 nuclear signal, relative to uninucleate wild-type cells. This observation might indicate that <italic>&#x00394;rad3</italic> cells spend a longer period of the cell cycle in S phase, perhaps as a result of heretofore unrecognized problems with DNA replication.</p><p>Upon activation of the DNA damage checkpoint by bleomycin, wild-type cells arrest in G₂, a period in the cell cycle in which Mik1 protein is not detected by immunofluorescence (Figure <xref ref-type="fig" rid="F4">4</xref>). Interestingly, Mik1 nuclear staining was clearly visible in cells treated with bleomycin (Figure <xref ref-type="fig" rid="F5">5</xref>A). This signal was quite evident but was substantially less strong than that observed with wild-type cells arrested with hydroxyurea. A large fraction of the <italic>&#x00394;rad3</italic> or <italic>&#x00394;chk1</italic> cells treated with bleomycin had no nuclear Mik1 signal, the exceptions being mostly septated, binucleate, or shorter uninucleate cells that were presumably in S or early G₂ (Figure <xref ref-type="fig" rid="F5">5</xref>, B and D). This pattern was particularly evident in the <italic>&#x00394;chk1</italic> cells. The <italic>&#x00394;cds1</italic> cells treated with bleomycin appeared very similar to wild-type counterparts, a finding that is consistent with the notion that Cds1 has no significant role in the DNA damage checkpoint. An immunoblot experiment was performed to confirm that bleomycin induces the accumulation of Mik1 to a level greater than that found in asynchronous cells but less than the amount found in cells treated with hydroxyurea (Figure <xref ref-type="fig" rid="F5">5</xref>E)</p><p>Thus, DNA damage inflicted by bleomycin induced the accumulation of Mik1 by a process dependent on Rad3 and Chk1. These findings substantially strengthened evidence that Mik1 is specifically regulated by the DNA damage checkpoint.</p></sec><sec><title>Mik1 Nuclear Accumulation Induced by DNA Damage in Prearrested G₂ Cells</title><p>An experiment was performed to determine if the accumulation of Mik1 induced by bleomycin was a specific consequence of the DNA damage checkpoint, as opposed to prolonged arrest in G₂. A culture of <italic>cdc25-22 mik1:13Myc</italic> cells was arrested in G₂ by incubation at restrictive temperature. Nuclear accumulation of Mik1 was not detected in these cells (Figure <xref ref-type="fig" rid="F6">6</xref>). However, when bleomycin was added after the shift to restrictive temperature, <italic>cdc25-22 mik1:13Myc</italic> cells exhibited Mik1 nuclear staining. Cells of the same genetic background that also contained the <italic>&#x00394;chk1</italic> or <italic>&#x00394;rad3</italic> allele failed to accumulate Mik1 when treated with bleomycin after a shift to restrictive temperature (Figure <xref ref-type="fig" rid="F6">6</xref>). These results demonstrated that nuclear accumulation of Mik1 induced by bleomycin was caused by activation of the damage checkpoint and was not simply a result of the G₂ arrest. Thus, nuclear accumulation of Mik1 appeared to be a cause instead of a consequence of the G₂ arrest induced by DNA damage. </p></sec><sec><title>Increased mik1 mRNA during Replication but Not Damage Checkpoint</title><p>Nuclear accumulation of Mik1 during S and early G₂ of the unperturbed cell cycle correlated with the appearance of <italic>mik1</italic> mRNA (Figure <xref ref-type="fig" rid="F4">4</xref>). Northern blot experiments were performed to determine if checkpoint-induced accumulation of Mik1 protein also correlated with changes in the abundance of <italic>mik1</italic> mRNA. These experiments were performed with wild-type or <italic>&#x00394;rad3</italic> cells exposed to hydroxyurea or bleomycin. A substantial increase in <italic>mik1</italic>^&#x0002b; mRNA was detected in wild-type cells treated with hydroxyurea (Figure <xref ref-type="fig" rid="F7">7</xref>). These cells were arrested in S, the phase of the cell cycle in which <italic>mik1</italic>^&#x0002b; mRNA is most abundant in cycling cells (Figure <xref ref-type="fig" rid="F4">4</xref>). Earlier unpublished studies indicated that <italic>mik1</italic>^&#x0002b; mRNA was unchanged in hydroxyurea-arrested cells (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>), but this finding was erroneous, perhaps because of the paucity of <italic>mik1</italic>^&#x0002b; mRNA. Importantly, <italic>mik1</italic>^&#x0002b; mRNA was unchanged in cells arrested in G₂ with bleomycin (Figure <xref ref-type="fig" rid="F7">7</xref>). Thus, the bleomycin-induced accumulation of nuclear Mik1 was apparently not caused by enhanced expression or stabilization of <italic>mik1</italic>^&#x0002b; mRNA. </p></sec><sec><title>Accumulation of Polyubiquitinated Mik1 in a Proteasome Mutant</title><p>The rapid disappearance of Mik1 upon exit from S phase in cycling cells (Figure <xref ref-type="fig" rid="F4">4</xref>) suggested that Mik1 protein might normally have a short half-life. As a first step in the exploration of this possibility, an experiment was performed to determine if Mik1 is stabilized in cells that have the temperature-sensitive <italic>mts3-1</italic> mutation. Mts3 is an essential component of the 26S proteasome (<xref ref-type="bibr" rid="B17">Gordon <italic>et al.</italic>, 1996</xref>). Incubation of this strain at the restrictive temperature led to a large increase of Mik1:13Myc protein (Figure <xref ref-type="fig" rid="F8">8</xref>A). To determine if Mik1 protein turnover might be mediated by polyubiquitination, this experiment was repeated in a strain that produced a hexahistidine-tagged form of ubiquitin. Hexahistidine proteins were purified with Ni^2&#x0002b;-nitrilotriacetic acid agarose and immunoblotted with antibodies to the myc epitope. Samples prepared from <italic>mts3-1</italic> cells yielded a substantial amount of Mik1:13Myc protein, whereas only a very weak Mik1:13Myc signal was detected in the sample from wild-type cells (Figure <xref ref-type="fig" rid="F8">8</xref>B). Most of this protein migrated in a broad range that is substantially larger than that of unmodified Mik1:13Myc protein, which indicated that Mik1:13Myc was most likely polyubiquitinated. These data indicated that Mik1 is rapidly degraded by ubiquitin-mediated proteolysis and suggested that checkpoints might induce Mik1 accumulation by stabilization of Mik1 protein. </p></sec></sec><sec><title>DISCUSSION</title><sec><title>Role of Mik1 in the DNA Damage Checkpoint</title><p>Genetic and biochemical studies established that Cdc25 is a target of negative regulation by Chk1. Mutational inactivation of Cdc25 prevents the onset of mitosis. Chk1, therefore, could theoretically enforce a DNA damage checkpoint solely by inhibition of Cdc25. Our goal was to determine if Cdc25 is the only important target of the DNA damage checkpoint. Our findings argue against this hypothesis. These studies strongly suggest that positive regulation of Mik1 plays a significant role in the DNA damage checkpoint.</p><p>Cdc25 is normally essential for mitosis, but some strains are viable without Cdc25. The genotype of one such strain is <italic>cdc2-3w &#x00394;cdc25</italic>. How <italic>cdc2-3w</italic> suppresses <italic>&#x00394;cdc25</italic> is unknown; Cdc2 encoded by <italic>cdc2-3w</italic> is inhibited by tyrosine 15 phosphorylation catalyzed by Wee1 or Mik1. In fact, <italic>cdc2-3w wee1-50</italic> cells undergo premature lethal mitosis (&#x0201c;mitotic catastrophe&#x0201d;) at restrictive temperature (<xref ref-type="bibr" rid="B44">Russell and Nurse, 1987b</xref>). We found that DNA damage causes a substantial mitotic delay in <italic>cdc2-3w &#x00394;cdc25</italic> cells. This defect was not obvious with static analysis of asynchronous cultures (<xref ref-type="bibr" rid="B15">Furnari <italic>et al.</italic>, 1997)</xref>, but it was readily detected by careful analysis of synchronous cultures. This delay was abolished by <italic>&#x00394;chk1</italic> or <italic>&#x00394;mik1</italic> mutations, a result consistent with a model in which positive regulation of Mik1 by Chk1 helps to enforce the DNA damage checkpoint.</p><p>Deletion of <italic>cdc25</italic>^&#x0002b; is also suppressed by <italic>wee1-50</italic> (<xref ref-type="bibr" rid="B42">Russell and Nurse, 1986</xref>). We found that GST&#x02013;Chk1 overproduction arrests division in <italic>wee1-50 &#x00394;cdc25</italic> cells. This observation provided further indications that Chk1 regulates Mik1, because GST&#x02013;Chk1 overproduction cannot arrest division in <italic>wee1-50 &#x00394;mik1</italic> cells or in cells that express Cdc2-F15, a form of Cdc2 that cannot be phosphorylated by Wee1 or Mik1 (<xref ref-type="bibr" rid="B38">Rhind <italic>et al.</italic>, 1997a</xref>). The hypothesis that Chk1 regulates Mik1 was explored in more detail by determining if the <italic>&#x00394;mik1</italic> mutation impairs division arrest induced by GST&#x02013;Chk1 overproduction. Surprisingly, division arrest induced by GST&#x02013;Chk1 was suppressed by <italic>&#x00394;mik1</italic> but not by <italic>wee1-50</italic> (as shown here) or <italic>&#x00394;wee1</italic> (<xref ref-type="bibr" rid="B15">Furnari <italic>et al.</italic>, 1997a</xref>). These findings are remarkable because Wee1 is presumed to contribute the bulk of kinase activity that phosphorylates Cdc2 on tyrosine 15 (<xref ref-type="bibr" rid="B28">Lundgren <italic>et al.</italic>, 1991</xref>). Mutational inactivation of Wee1 but not Mik1 causes a wee phenotype and suppresses <italic>cdc25</italic> loss-of-function mutations. These findings suggested that Mik1 but not Wee1 is a target of Chk1-mediated division arrest.</p><p>How does <italic>&#x00394;mik1</italic> rescue division arrest caused by enzymatic inactivation of Cdc25 catalyzed by GST&#x02013;Chk1 but not mutational inactivation of Cdc25, whereas <italic>wee1</italic> mutations have the opposite effect? An answer is provided by a model in which Chk1 increases Mik1 abundance. If Chk1 simultaneously inhibits Cdc25 function and increases Mik1 abundance, then <italic>wee1</italic>^&#x02212; mutations could be insufficient to suppress enzymatic inactivation of Cdc25. Mutational inactivation of Cdc25 is not accompanied by enhanced Mik1 abundance. A related question is the following: why is the <italic>&#x00394;mik1</italic> mutation sufficient to suppress enzymatic inactivation of Cdc25 catalyzed by GST&#x02013;Chk1 but not mutational inactivation of Cdc25? Two explanations come readily to mind. Cdc25 might be more severely inhibited by mutations compared with GST&#x02013;Chk1 overproduction. Hence, <italic>&#x00394;mik1</italic> might be sufficient to rescue Chk1-mediated inhibition of Cdc25 but not mutational inactivation of Cdc25. Alternatively, suppression of GST&#x02013;Chk1 overproduction in <italic>&#x00394;mik1</italic> cells might involve checkpoint adaptation. This adaptation process would depend on intact Cdc25 protein that might be subject to positive regulation that counteracts the effect of Chk1. The notion of adaptation is consistent with the observation that <italic>&#x00394;mik1</italic> cells appear to temporarily cease division for a period after induction of GST&#x02013;Chk1 but eventually recover (our unpublished data). Checkpoint adaptation is unexplored in fission yeast but has been proposed to operate in budding yeast (<xref ref-type="bibr" rid="B51">Toczyski <italic>et al.</italic>, 1997</xref>; <xref ref-type="bibr" rid="B25">Lee <italic>et al.</italic>, 1998</xref>),</p><p>The hypothesis that Mik1 is important for the DNA damage checkpoint was confirmed in studies that examined the effect of <italic>&#x00394;mik1</italic> in the checkpoint response elicited by continuous exposure to bleomycin. In wild-type cells, bleomycin causes a prolonged arrest that lasts for at least 240 min. In contrast, <italic>&#x00394;mik1</italic> cells exposed to bleomycin initially arrest normally but undergo division &#x0223c;100 min after the mock-treated cells. The fact that <italic>&#x00394;mik1</italic> cells undergo a substantial but abbreviated mitotic delay may explain why this defect has remain undiscovered until now.</p></sec><sec><title>Regulation of Mik1 by the DNA Damage Checkpoint</title><p>Having established that Mik1 is important for the DNA damage checkpoint, we then investigated how Mik1 might be regulated by the checkpoint apparatus. We found that Mik1 protein abundance was substantially increased in cells treated with bleomycin. This fact was evident from both immunoblot and immunolocalization studies. The latter studies indicated that Mik1 accumulation requires Rad3 and Chk1; therefore, the Mik1 accumulation appears be a consequence of DNA damage checkpoint activation. Indeed, G₂ arrest caused by the <italic>cdc25-22</italic> mutation did not lead to Mik1 accumulation, a result that indicates that Mik1 accumulation is a cause rather than a consequence of G₂ arrest induced by the DNA damage checkpoint. Mik1 is a dose-dependent inhibitor of mitosis (<xref ref-type="bibr" rid="B28">Lundgren <italic>et al.</italic>, 1991</xref>); thus, it is easy to understand how increased abundance of Mik1 would help to enforce the DNA damage checkpoint.</p><p>Mik1 abundance increases in cells arrested at the DNA damage checkpoint with bleomycin, but the magnitude of Mik1 accumulation is less than that observed in cells treated with hydroxyurea. The bleomycin effect is mediated through the damage checkpoint involving Chk1, whereas hydroxyurea leads to Cds1 activation as part of the replication checkpoint response (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B26">Lindsay <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B7">Brondello <italic>et al.</italic>, 1999</xref>). The difference in the magnitude of Mik1 protein accumulation is probably attributable, at least in part, to the difference in <italic>mik1</italic> mRNA accumulation. As we have shown here, there is substantially more <italic>mik1</italic> mRNA in cells arrested with hydroxyurea compared with bleomycin. The effect of hydroxyurea on <italic>mik1</italic> mRNA appears to be part of the S&#x02013;M replication checkpoint response, because it depends on Rad3.</p><p>We have presented initial studies aimed at determining the mechanism by which DNA damage induces the accumulation of Mik1 protein. We found that Mik1 accumulated in a <italic>mts3-1</italic> mutant that is defective in a subunit of the 26S proteasome (<xref ref-type="bibr" rid="B17">Gordon <italic>et al.</italic>, 1996</xref>). Moreover, our studies revealed that Mik1 is ubiquitinated. These findings, and the fact that Mik1 abundance oscillates in a very transitory manner during the normal cell cycle, suggest that Mik1 protein is normally very unstable. These findings prompt speculation that the DNA damage checkpoint stabilizes Mik1 protein. There is precedence for checkpoint-induced stabilization of proteins in the Wee1/Mik1 family. In the budding yeast <italic>Saccharomyces cerevisiae</italic>, the Wee1/Mik1 homologue Swe1 is stabilized in cells arrested at the morphogenesis checkpoint that coordinates mitosis with bud formation (<xref ref-type="bibr" rid="B50">Sia <italic>et al.</italic>, 1998</xref>). In cell extracts, exogenous Swe1 is polyubiquitinated by a process that is impaired by <italic>cdc34</italic> or <italic>met30</italic> mutations (<xref ref-type="bibr" rid="B21">Kaiser <italic>et al.</italic>, 1998</xref>). Cdc34 is an E2 ubiquitin-conjugating enzyme, whereas Met30 is an &#x0201c;F box&#x0201d; protein that is a component of an E3 ubiquitin ligase enzyme. Swe1 protein is also stabilized by <italic>cdc34</italic> or <italic>met30</italic> mutations in vivo (<xref ref-type="bibr" rid="B21">Kaiser <italic>et al.</italic>, 1998</xref>). In <italic>Xenopus</italic> oocyte extracts, activation of the replication checkpoint stabilizes exogenous Wee1 protein (<xref ref-type="bibr" rid="B30">Michael and Newport, 1998</xref>). In these assays, Wee1 was stabilized by the addition of a dominant negative form of Cdc34. Together, these studies and our results suggest that the induced stabilization of Cdc2-directed tyrosine kinases plays an important role in checkpoint mechanisms that couple the onset of mitosis to DNA replication, DNA repair, or morphogenetic events required for successful cell division.</p></sec><sec><title>Regulation of Mik1 during S Phase</title><p>We have also explored Mik1 regulation during the normal cell cycle. Studies performed with synchronous cultures showed that <italic>mik1</italic>^&#x0002b; mRNA is expressed periodically during the cell cycle, with peak signals detected during S phase. The appearance of Mik1 protein was quite similar, being slightly delayed relative to the detection of <italic>mik1</italic>^&#x0002b; mRNA. These findings were confirmed by immunolocalization studies that showed that Mik1 is detected in binucleate cells and short uninucleate cells, a pattern that corresponds to S phase and perhaps early G₂.</p><p>It is remarkable that expression of Mik1, a mitotic inhibitor, is enhanced during S phase. It is tempting to speculate that this pattern of Mik1 expression helps to couple the onset of mitosis to the completion of DNA replication in cell cycles that are unperturbed by DNA replication or damage checkpoints. This could be an intrinsic mechanism of inhibiting mitosis during S phase. DNA replication is normally completed very quickly after nuclear division in fission yeast, long before cells have satisfied the size requirement for the initiation of mitosis. Thus, in these conditions, the Mik1-independent cell size control is sufficient to ensure that mitosis occurs after the completion of DNA replication. This cell size control requires Wee1, as indicated by the phenotype of <italic>wee1</italic>^&#x02212; mutants, which undergo mitosis at approximately half the size of wild-type cells (<xref ref-type="bibr" rid="B34">Nurse, 1975</xref>; <xref ref-type="bibr" rid="B44">Russell and Nurse, 1987b</xref>). In fact, simultaneous inactivation of Wee1 and Mik1 causes a mitotic catastrophe in which cells undergo mitosis at a very small size and apparently before the completion of DNA replication (<xref ref-type="bibr" rid="B28">Lundgren <italic>et al.</italic>, 1991</xref>). Thus, in a <italic>wee1</italic>^&#x02212; mutant, the periodic expression of Mik1 that occurs during S phase is essential to couple the onset of mitosis to the completion of DNA replication.</p><p>It remains to be determined if there are situations that do not involve the inhibition of DNA replication in which periodic expression of Mik1 during S phase is required to couple mitosis to the completion of DNA replication. In some conditions of poor nutrient availability, wild-type cells divide at a small cell size that approaches the size of <italic>wee1</italic>^&#x02212; mutants (<xref ref-type="bibr" rid="B13">Fantes and Nurse, 1977</xref>). It is possible that Wee1 is inactivated in poor nutrient conditions, thereby accounting for the contraction of G₂ and the extension of G₁. Future experiments will determine if Mik1 is important for proper mitotic control in these circumstances.</p></sec><sec><title>Strategies for Dual Enforcement of Checkpoints</title><p>This and previous studies provide strong evidence that replication and damage checkpoint arrests are maintained through two general mechanisms: negative regulation of Cdc25 and positive regulation of Mik1 (Figure <xref ref-type="fig" rid="F9">9</xref>). In each case, the effect is the same, namely, to maintain tyrosine 15 phosphorylation of Cdc2 and thereby prevent the onset of mitosis. The question arises regarding the relative importance of the two modes of regulation. The <italic>&#x00394;mik1</italic> mutation clearly causes a damage checkpoint defect, but this defect is modest relative to the absence of checkpoint arrest seen in <italic>&#x00394;chk1</italic> cells. Likewise, the hydroxyurea-induced replication checkpoint fails gradually in a population of <italic>&#x00394;mik1</italic> cells (<xref ref-type="bibr" rid="B57">Zeng <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B14">Furnari <italic>et al.</italic>, 1999</xref>), whereas the checkpoint is abolished in <italic>&#x00394;rad3</italic> or <italic>&#x00394;cds1 &#x00394;chk1</italic> cells (<xref ref-type="bibr" rid="B4">Boddy <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B26">Lindsay <italic>et al.</italic>, 1998</xref>). This comparison suggests that positive regulation of Mik1 may be of secondary import relative to negative regulation of Cdc25. However, this view of the checkpoint mechanism may be incorrect, because the equivalent experiment cannot be performed with a strain that lacks Cdc25 but is otherwise wild type. The closest approximation are studies with <italic>&#x00394;cdc25 cdc2-3w</italic> cells, which have shown that the damage checkpoint is largely but not completely eliminated. However, in these studies, the effect of the <italic>cdc2-3w</italic> mutation cannot be ascertained with certainty. Attempts to specifically abrogate the checkpoint regulation of Cdc25 by mutation of Cds1- or Chk1-directed phosphorylation sites have produced only partial checkpoint defects (<xref ref-type="bibr" rid="B57">Zeng <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B14">Furnari <italic>et al.</italic>, 1999</xref>), although only a subset of phosphorylation sites have been eliminated in these experiments. Therefore, an accurate evaluation of the relative importance of Cdc25 and Mik1 regulation in the checkpoint responses awaits a more detailed understanding of the modes of regulation. These questions will be more fully addressed with the invention of a method to specifically abrogate checkpoint regulation of Cdc25. </p><p>An alternative viewpoint is that Mik1 has specialized importance in maintaining long-term checkpoint arrests. In our studies with bleomycin treatment, we observed that the <italic>&#x00394;mik1</italic> mutation caused a dramatic failure of the checkpoint arrest, but this failure occurred &#x0223c;100 min after mock-treated cells underwent mitosis. Two factors may be important in these circumstances. One factor is checkpoint adaptation. Some types of DNA damage may be unrepairable but not necessarily lethal to both daughter cells. In this case, the best survival strategy is to relieve the checkpoint arrest and undergo division. This postulated mechanism of active checkpoint override has been termed &#x0201c;adaptation&#x0201d; (<xref ref-type="bibr" rid="B51">Toczyski <italic>et al.</italic>, 1997</xref>). The behavior of <italic>&#x00394;mik1</italic> cells in bleomycin might be viewed as premature adaptation.</p><p>Size control of mitosis is a second factor that may explain the significance of Mik1 regulation by the DNA damage checkpoint. Cells must reach a certain size threshold before initiating mitosis, but after this threshold is reached, it is likely that mitosis-promoting &#x0201c;forces&#x0201d; increase as cell growth continues. These forces might be influenced by the continued accumulation of Cdc25 or cyclin B (<xref ref-type="bibr" rid="B5">Booher and Beach, 1988</xref>; <xref ref-type="bibr" rid="B32">Moreno <italic>et al.</italic>, 1990</xref>) or the increased activity of the kinases Nim1/Cdr1 or Cdr2 that inhibit Wee1 (<xref ref-type="bibr" rid="B43">Russell and Nurse, 1987a</xref>; <xref ref-type="bibr" rid="B9">Coleman <italic>et al.</italic>, 1993</xref>; <xref ref-type="bibr" rid="B36">Parker <italic>et al.</italic>, 1993</xref>; <xref ref-type="bibr" rid="B55">Wu and Russell, 1993</xref>; <xref ref-type="bibr" rid="B6">Breeding <italic>et al.</italic>, 1998</xref>; <xref ref-type="bibr" rid="B22">Kanoh and Russell, 1998</xref>), to name only some possibilities. Hence, Chk1-mediated inhibition of Cdc25 may be sufficient to delay mitosis in cells that are somewhat above the size threshold, but reliance on this regulation alone may fail as cell size increases further. Thus, regulation of Mik1 by the DNA damage checkpoint may be particularly important in late G₂ cells that suffer a large amount of DNA damage or damage that requires more time to repair.</p></sec></sec>

'Gene shaving' as a method for identifying distinct sets of genes with similar expression patterns

<sec><title>Background:</title><p>Large gene expression studies, such as those conducted using DNA arrays, often provide millions of different pieces of data. To address the problem of analyzing such data, we describe a statistical method, which we have called 'gene shaving'. The method identifies subsets of genes with coherent expression patterns and large variation across conditions. Gene shaving differs from hierarchical clustering and other widely used methods for analyzing gene expression studies in that genes may belong to more than one cluster, and the clustering may be supervised by an outcome measure. The technique can be 'unsupervised', that is, the genes and samples are treated as unlabeled, or partially or fully supervised by using known properties of the genes or samples to assist in finding meaningful groupings.</p></sec><sec><title>Results:</title><p>We illustrate the use of the gene shaving method to analyze gene expression measurements made on samples from patients with diffuse large B-cell lymphoma. The method identifies a small cluster of genes whose expression is highly predictive of survival.</p></sec><sec><title>Conclusions:</title><p>The gene shaving method is a potentially useful tool for exploration of gene expression data and identification of interesting clusters of genes worth further investigation.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Hastie</surname><given-names>Trevor</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref><email>tibs@stat.stanford.edu</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Tibshirani</surname><given-names>Robert</given-names></name><xref ref-type="aff" rid="I2">2</xref><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Eisen</surname><given-names>Michael B</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Alizadeh</surname><given-names>Ash</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Levy</surname><given-names>Ronald</given-names></name><xref ref-type="aff" rid="I5">5</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Staudt</surname><given-names>Louis</given-names></name><xref ref-type="aff" rid="I6">6</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Chan</surname><given-names>Wing C</given-names></name><xref ref-type="aff" rid="I7">7</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Botstein</surname><given-names>David</given-names></name><xref ref-type="aff" rid="I8">8</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Brown</surname><given-names>Patrick</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib>

Genome Biology

<sec><title>Background</title><p>Through the use of recently developed DNA arrays, it is now possible to obtain accurate, quantitative (relative) measurements of a large proportion of the mRNA species present in a biological sample. DNA arrays have been used to monitor changes in gene expression during important biological processes (for example, cellular replication and the response to changes in the environment), and to study variation in gene expression across collections of related samples (such as tumor samples from patients with cancer). A major challenge in interpreting these results is to understand the structure of the data produced by such studies, which often consist of millions of measurements. A variety of clustering techniques have been applied to such data, and have proved useful for identifying biologically relevant groupings of genes and samples [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. Although the underlying principles and computational details of these methods differ, they share the goal of organizing the elements under consideration (such as genes) into groups (clusters) with coherent behavior across relevant measurements (such as samples). Generally absent is any consideration of the nature of the coherent variation. For example, one might want to identify groups of genes that have coherent patterns of expression with large variance across samples, or groups of genes that optimally separate samples into predefined classes (such as different clinical response groups in tumor samples). Here, we introduce a new statistical method, which we call gene shaving, that attempts to identify groups of elements (genes) that have coherent expression and are optimal for various properties of the variation in their expression.</p><p>Figure <xref ref-type="fig" rid="F1">1</xref> shows the dataset used in our study, which consisted of 4673 gene expression measurements on 48 patients with diffuse large B-cell lymphoma (DLCL). These data have been described in detail previously [<xref ref-type="bibr" rid="B14">14</xref>]. The column labels refer to different patients, and the rows correspond to genes. The order of rows and columns is arbitrary.</p><p>Some authors have recently explored the use of clustering methods to arrange the genes in some systematic way, with similar genes placed close together (see [<xref ref-type="bibr" rid="B2">2</xref>] for developments and [<xref ref-type="bibr" rid="B15">15</xref>] for an overview). In Figure <xref ref-type="fig" rid="F2">2</xref>, we have applied hierarchical clustering to the genes and samples separately. Each clustering produces a (non-unique) ordering, one that ensures that the branches of the corresponding dendrogram do not cross. Figure <xref ref-type="fig" rid="F2">2</xref> displays the original data, with rows and columns ordered accordingly.</p><p>Some structure is evident in Figure <xref ref-type="fig" rid="F2">2</xref>, and this method can be used to recognize relationships among the genes and samples.With any method that reduces the dimension of the data, however, finer structure can be lost. For example, suppose the expression of some subset of genes divides the samples in an informative way, correlating with the rate of proliferation of tumor cells, for example, whereas another subset of genes divides the samples a different way, representing the immune response, for example. Then methods such as two-way hierarchical clustering, which seek a single reordering of the samples for all genes, cannot find such structure.</p><p>The method of gene shaving we describe here is designed to extract coherent and typically small clusters of genes that vary as much as possible across the samples. Figure <xref ref-type="fig" rid="F3">3</xref> shows three gene clusters for the DLCL data, found using shaving. Some of the genes within each cluster lie close to each other in the hierarchical clustering of Figure <xref ref-type="fig" rid="F2">2</xref>, but others, and the clusters themselves, are quite far apart.</p><p>In Figure <xref ref-type="fig" rid="F3">3</xref> the samples have been ordered by values of the average gene expression. This average gene is a good representative of the cluster, as all the members are so similar. The variance measures at the top of each cluster are discussed in more detail later. The clusters are all of different sizes. We use an automatic method for determining the size of the clusters, based on a randomization procedure that protects us from looking too hard in the large sea of genes and finding spurious structure. The three cluster-average genes, one from each cluster, are reasonably uncorrelated (see below and Figure <xref ref-type="fig" rid="F6">6</xref>). This is another aspect of the shaving process - it seeks different clusters, where difference is measured by correlation of the cluster mean. Figure <xref ref-type="fig" rid="F4">4</xref> shows the results of a hierarchical clustering applied to the three column-average genes. Whereas hierarchical clustering suggests two main gene groupings, the shaving process may suggest more useful groupings.</p><p>This article is organized as follows. In the section 'Gene shaving' we describe the method itself. The section entitled 'The gap estimate of cluster size' outlines the gap test for choosing the cluster size. In the section 'Predicting patient survival' we try to predict patient survival from gene cluster averages. 'Supervised shaving' is discussed in the following section. Finally, in the 'Conclusions' we propose some further generalizations. A more statistical treatment of gene shaving is given in [<xref ref-type="bibr" rid="B16">16</xref>].</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>The DLCL expression matrix, in no particular row or column order. The display is a heat map, ranging from bright green (negative, underexpressed) to bright red (positive, overexpressed). The gray cells indicate missing measurements.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>The DLCL expression matrix with rows and columns ordered according to a hierarchical clustering applied separately to the rows and columns.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p>The first three gene clusters found for the DLCL data. Each is a collection of genes showing similar and strong (high variance) expression behavior.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>The top panel shows the three signed-mean genes together, and ordered by a hierarchical clustering in this three-dimensional space. The lower panel is similar, except here we show all the genes in each cluster, 33 in all.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-4"/></fig></sec><sec><title>Results</title><sec><title>Gene shaving</title><p>In this section we describe in detail our technique for finding clusters like the example in Figure <xref ref-type="fig" rid="F3">3</xref>. A gene expression matrix is an <italic>N</italic> &#x000d7; <italic>p</italic> matrix of real-valued measurements <italic>X </italic> = x_{<italic>ij</italic>}. The rows are genes, the columns are samples, and x_{<italic>ij</italic>} is the measured (log) expression, relative to a baseline. Typically there are missing entries in <italic>X.</italic> We use a technique described in [<xref ref-type="bibr" rid="B17">17</xref>], an iterative algorithm based on the singular value decomposition, for imputing missing expression values;our analysis here assumes that <italic>X</italic> has no missing values.</p><p>Let <italic>S</italic>_{<italic>k</italic>} be the indices of a cluster of <italic>k</italic> genes, and</p><graphic xlink:href="gb-2000-1-2-research0003-I1.gif"/><p>be the collection of <italic>p</italic> column averages of the expression values for this cluster. Then for each cluster size <italic>k,</italic> gene shaving seeks a cluster <italic>S</italic>_{<italic>k</italic>} having the highest variance of the column averages:</p><graphic xlink:href="gb-2000-1-2-research0003-I2.gif"/><p>The important question of how to choose the cluster size <italic>k</italic> is addressed in the next section.</p><p>Our procedure generates a sequence of nested clusters <italic>S</italic>_{<italic>k</italic>}, in a top-down manner, starting with <italic>k = N,</italic> the total number of genes, and decreasing down to <italic>k</italic> = 1 gene. At each stage the largest principal component of the current cluster of genes is computed. This eigen gene is the normalized linear combination of genes with largest variance across the samples. We then compute the inner product (essentially the correlation) of each gene with the eigen gene, and discard ('shave off') a fraction of the genes having lowest (absolute) inner product. The process is repeated on the reduced cluster of genes. The shaving algorithm is depicted in Figure <xref ref-type="fig" rid="F5">5</xref> and given in detail in Box <xref ref-type="fig" rid="F15">1</xref>.</p><p>There are a number of duplicate genes in the dataset. In some cases the sequence for a given gene appears on the microarray more than once, either by design or by accident. In other cases, more than one different EST (expressed sequence tag) is present for the same gene. Gene shaving can be affected by duplicate genes, since they are highly correlated with each other. We therefore averaged expression profiles for the duplicate genes, leaving 3624 unique gene profiles.</p><p>The sequence of operations 1-5 in Box <xref ref-type="fig" rid="F15">1</xref> gives the first cluster of rows - the first ribbon in Figure <xref ref-type="fig" rid="F3">3</xref>. Step 6 orthogonalizes the data to encourage discovery of a different (uncorrelated) second cluster. Note that although we work with the orthogonalized matrix in the shaving process for the second and subsequent clusters, the derived clusters and their averages involve the original genes.</p><p>The shaving process requires repeated computation of the largest principal component of a large set of variables. If naively implemented, this requires the construction of a <italic>N</italic> &#x000d7; <italic>N</italic> sample covariance matrix &#x02211; of the genes, and the computation of its largest eigenvector. We can avoid the computational burden by working in the dual space, where the matrices have dimension <italic>p</italic> &#x000d7; <italic>p.</italic> Furthermore, as we require only the largest eigenvector, the computations can be reduced even further by using the power method, using the eigenvector of the previous cluster as a starting value.</p><p>The three resulting clusters are shown in Figure <xref ref-type="fig" rid="F3">3</xref> and again in Figure <xref ref-type="fig" rid="F4">4</xref>. Figure <xref ref-type="fig" rid="F6">6</xref> shows the pairwise scatterplots of each of the three column averages ('super genes') from the clusters. The absolute correlations range from 0.27 to 0.68. The full gene names for the members of the first three clusters are given in Table <xref ref-type="table" rid="T1">1</xref>.</p><p>It is useful to evaluate how much of the dimensionality of the gene expression variation is captured by the clusters derived from gene shaving. We can approximate the expression profile for each gene in the complete dataset as a linear combination of the three super genes from each cluster (which are simple averages of the genes in each cluster). The percent variance explained by the first <italic>j =</italic> 1,2, ... 10 super genes is shown in Figure <xref ref-type="fig" rid="F7">7</xref>.</p><p>Thus the three gene clusters, involving a total of 33 genes, explain about 20% of the variation. The percent variance explained by the first <italic>j</italic> principal components (broken curve) exceeds that from gene shaving. Each principal component gives a non-zero weight to all 3624 genes, however.</p></sec><sec><title>The gap estimate of cluster size</title><p>Each shaving sequence produces a nested set of gene clusters <italic>S</italic>_{<italic>k</italic>}, starting with the entire expression matrix and then proceeding down to some minimum cluster size (typically 1). If we applied this procedure to null data, in which the rows and columns were independent of each other, we could still find some interesting-looking patterns in the resulting blocks. Hence, we need to calibrate this process so that we can differentiate real patterns from spurious ones. Even in the case of real structure, it is unlikely that a distinct set of genes is correct for a cluster, and the rest not. More likely there is a graduation of membership eligibility, and we have to decide where to draw the line. Here we describe a procedure based on randomization that helps us select a reasonable cluster size.</p><p>Our method requires a quality measure for a cluster. We favor both high-variance clusters, and high coherence between members of the cluster. As the generation of the cluster sequence was driven strongly by the former, we focus on the latter in selecting a good cluster. By analogy with the analysis of variance for grouped data, we define the following measures of variance for a cluster <italic>S</italic>_{<italic>k</italic>}:</p><graphic xlink:href="gb-2000-1-2-research0003-I3.gif"/><graphic xlink:href="gb-2000-1-2-research0003-I4.gif"/><graphic xlink:href="gb-2000-1-2-research0003-I5.gif"/><p>The between variance is the variance of the (signed) mean gene. The within variance measures the variability of each gene about the cluster average, also averaged over samples. As this can be small if the overall variance is small, a more pertinent measure is the between-to-within variance ratio <italic>V</italic>_{<italic>B</italic>}/<italic>V</italic>_{<italic>W</italic>}, or alternatively, the percent variance explained</p><graphic xlink:href="gb-2000-1-2-research0003-I6.gif"/><p>A large value of <italic>R</italic>² implies a tight cluster of coherent genes. This is the quality measure we use to select a cluster from the shaving sequence <italic>S</italic>_{<italic>k</italic>}.</p><p>Let <italic>S</italic>_{<italic>k</italic>} index the clusters of a given shaving sequence (with <italic>k</italic> being the number of genes). Let <italic>D</italic>_{<italic>k</italic>} be the <italic>R</italic>² measure for the <italic>k</italic>th member of sequence. We wish to know whether <italic>D</italic>_{<italic>k</italic>} is larger than we would expect by chance, if the rows and columns of the data were independent.</p><p>Let <italic>X</italic>^{*<italic>b</italic>} be a permuted data matrix, obtained by permuting the elements within each row <italic>of X.</italic> We form <italic>B</italic> such matrices, indexed by <italic>b</italic> = 1,2,... <italic>B</italic>. Let <italic>D</italic>_{<italic>k</italic>}^{*<italic>b</italic>} be the <italic>R</italic>² measure for cluster S_{<italic>k</italic>}^{*<italic>b</italic>}. Denote by <inline-graphic xlink:href="gb-2000-1-2-research0003-I10.gif"/>_{<italic>k</italic>}^* the average of <italic>D</italic>_{<italic>k</italic>}^{*<italic>b</italic>} over <italic>b</italic>. The <italic>Gap</italic> function is defined by</p><graphic xlink:href="gb-2000-1-2-research0003-I7.gif"/><p>We then select as the optimal number of genes that value of <italic>k</italic> producing the largest gap:</p><graphic xlink:href="gb-2000-1-2-research0003-I8.gif"/><p>The idea is that at the value <inline-graphic xlink:href="gb-2000-1-2-research0003-I11.gif"/> the observed variance is the most ahead of expected. Multiple clusters are produced for the randomized data just like for the original data, and the gap test is used repeatedly to select the cluster size at each stage.</p><p>For the DLCL data, the maximum for the first cluster occurs at eight genes. Figure <xref ref-type="fig" rid="F8">8</xref> shows the percent-variance curves, <italic>D</italic>_{<italic>k</italic>}, for both the original and randomized tumor data as a function of size, and the gap curves used to select the specific cluster sizes in Figure <xref ref-type="fig" rid="F3">3</xref>.</p></sec><sec><title>Predicting patient survival</title><p>One important motivation for developing gene shaving was the wish to identify distinct sets of genes whose variation in expression could be related to a biological property of the samples. In the present example, finding genes whose expression correlates with patient survival is an obvious challenge. Group factors <italic>g</italic>₁, <italic>g</italic>₂, <italic>g</italic>₃ were created by splitting each gene cluster in Figure <xref ref-type="fig" rid="F3">3</xref> into two groups of 24 patients. We used each of these groupings as a factor in Cox's proportional hazards model for predicting overall survival [<xref ref-type="bibr" rid="B18">18</xref>]. Of the group factors only <italic>g</italic>₂ was significant, at the 0.05 level (<italic>p</italic> = 0.04).</p><p>In [<xref ref-type="bibr" rid="B14">14</xref>], a cluster of 380 genes was chosen on the basis of their large variation over samples, and their 'germinal center B-like' or 'activated B-like' expression profiles. Using these 380 genes, a hierarchical clustering produced two groups of patients which were (just) statistically different in survival. Close inspection shows that 18 of the 23 genes in the second cluster above also fall into this cluster of 380 genes. Hence, gene shaving can find clinically and biologically relevant subdivisions in gene expression data in an unsupervised fashion.</p><p>It may be fortuitous that one of these groupings correlates with survival, as the clusters were not chosen with survival in mind. We next describe a modification of gene shaving that explicitly looks for clusters that are related to patient survival.</p></sec><sec><title>Supervised shaving</title><p>The methods discussed so far have not used information about the columns to 'supervise' the shaving of rows. Here we generalize gene shaving to incorporate full or partial supervision.</p><p>As in Equation (1), we consider a cluster of genes <italic>S</italic>_{<italic>k</italic>} having column average vector <inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/>. Let <italic>y</italic> = (<italic>y</italic>₁, <italic>y</italic>₂, ... <italic>y</italic>_p) be a set of auxiliary measurements available for the samples. For example each <italic>y</italic>_{<italic>j</italic>}, might be a survival time for the patient corresponding to sample <italic>j</italic> or a class label for each sample, such as a diagnosis category. Supervised shaving maximizes a weighted combination of column variance and an information measure <italic>J</italic>(<inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/>, <italic>y</italic>):</p><graphic xlink:href="gb-2000-1-2-research0003-I9.gif"/><p>for fixed 0 &#x02264; &#x003b1; &#x02264; 1. The value &#x003b1; = 1 gives full supervision; values between o and 1 provide partial supervision.</p><p>Choice of the measure <italic>J</italic>(<inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/>, <italic>y</italic>) depends on the nature of the auxiliary information <italic>y</italic>. If the <italic>y</italic> codes class labels, <italic>J</italic>(<inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/>, <italic>y</italic>) can be taken as the sum of squared differences between the category averages <inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/>. For censored survival times <italic>y</italic>, think of <inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/> as a covariate in a Cox (proportional hazards) model. If the score vector from this model is <italic>g,</italic> we set <italic>J</italic>(<inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/>, <italic>y</italic>) = <italic>gg</italic>^{<italic>T</italic>}, a <italic>p</italic> &#x000d7; <italic>p</italic> matrix. Computationally we first scale the genes so that the within-risk set variance is 1.</p><p>When fully supervised, the shaving procedure reduces to simply ranking the genes from largest to smallest Cox model score test. Thus there is no role for principal components or top-down shaving in this case. However, to encourage coherence within the gene clusters, it can be better to use a partially supervised criterion, which does use principal components and top-down shaving. This is demonstrated in the example below. One can also include other prognostic factors in the model, and direct shaving to find a gene cluster whose column average is a strong predictor in the model. This can be done with other models, for example a linear regression model for a quantitative measurement. Operationally, all of these choices for <italic>J</italic> are quadratic functions of the column averages <inline-graphic xlink:href="gb-2000-1-2-research0003-I13.gif"/>, and gene shaving can be carried out via principal components of a suitably modified variance matrix.</p><p>We applied supervised shaving to the set of 3624 genes from the DLCL samples. Figure <xref ref-type="fig" rid="F9">9</xref> examines the effect of different values of the supervision weight &#x003b1;, showing the average (absolute) gene correlation and Cox model <italic>p</italic> value for each choice. From this plot we chose the value &#x003b1; = 0.10, which gives a good mix of high gene correlation and low <italic>p</italic> value. Partially supervised gene shaving produced a cluster with 234 genes, shown in Figure <xref ref-type="fig" rid="F10">10</xref> and in Table <xref ref-type="table" rid="T2">2</xref>.</p><p>Some of the genes are close together in the hierarchical clustering order (indicated by the first number in Table <xref ref-type="table" rid="T2">2</xref>), many are not. Some genes have a negative sign, and others have no sign. We will call these 'negative' and 'positive' genes respectively. The negative genes have their expression values flipped before being averaged with other gene expression profiles. Figure <xref ref-type="fig" rid="F11">11a</xref> shows the gap curve, suggesting a cluster size of 35. However, further analysis below suggests the better cluster size of 234.</p><p>The cluster of 234 genes contains many of the strongest individual genes for predicting survival. For example, 130 of the strongest 200 genes are in the cluster. Some weaker genes are, however, also in the cluster, the weakest being the 1332nd strongest gene among the full list of 3624. Figure <xref ref-type="fig" rid="F11">11b</xref> shows the survival curves stratified by the low and high expression of this gene cluster, using the median of the cutoff point. The two resulting groups are shown in Figure <xref ref-type="fig" rid="F12">12</xref>.</p><p>Using this grouping as a predictor in the Cox model for survival gave a highly significant <italic>p</italic> value (0.00042). However, this <italic>p</italic> value must be scrutinized. Figure <xref ref-type="fig" rid="F13">13a</xref>,<xref ref-type="fig" rid="F13">b</xref> shows the Cox model <italic>p</italic> value as a function of the cluster size, for both partially and fully supervised shaving. We will call these the 'training set <italic>p</italic> values'. As the gene shaving process was supervised by the survival times, the training set <italic>p</italic> values will be biased downward, and it is important to adjust them appropriately. We permuted the survival times, re-ran the shaving process and computed the corresponding <italic>p</italic> values. This was repeated 100 times, and for each cluster size we computed the proportion of times the permutation <italic>p</italic> values were less than or equal to the training set <italic>p</italic> values. These proportions are shown in Figure <xref ref-type="fig" rid="F13">13c</xref>,<xref ref-type="fig" rid="F13">d</xref>, and are unbiased estimates of the true <italic>p</italic> values. Fully supervised shaving is too aggressive, and the upward adjustment of the <italic>p</italic> values is large. As a result the <italic>p</italic> value is around 0.05 for the full set of genes, but none of the smaller clusters is significant at the 0.05 level. For partially supervised shaving, many of the <italic>p</italic> values are below 0.05, and from this we chose the cluster size of 234 near the minimum.</p><p>Using the full set of genes, flipping each to have positive correlation with survival, averaging their expression values and finally cutting at the median, gave a grouping nearly the same as Groups 1 and 2 in Figure <xref ref-type="fig" rid="F12">12</xref>. The only change was a swap between DLCL-oo14 and DLCL-oo18, and these two samples are right at the median cutpoint between the two groups in Figure <xref ref-type="fig" rid="F10">10</xref>.</p><p>The international prognostic index (IPI) A score was also available for these patients. Components of the IPI include age, level of the enzyme lactate dehydrogenase (LDH) and the number of extranodal sites. As in [<xref ref-type="bibr" rid="B14">14</xref>], we dichotomized IPI scores into low (0, 1 or 2) and high (3, 4 or 5). The resulting grouping seems to be about as predictive as the IPI index, and is quite independent from it, as Table <xref ref-type="table" rid="T3">3</xref> indicates.</p><p>When added to a Cox model containing IPI, this grouping had a training set <italic>p</italic> value of 0.0006. Figure <xref ref-type="fig" rid="F14">14</xref> shows the Kaplan-Meier survival curves for each group, stratified by low and high IPI.</p><p>In [<xref ref-type="bibr" rid="B14">14</xref>], two patient groups were defined from a hierarchical clustering tree grown from a 38o-gene cluster. As a predictor, the grouping was just significant in the low IPI group only, at the 0.05 level. Table <xref ref-type="table" rid="T4">4</xref> gives a cross-tabulation of that grouping with the one used in this paper in Figure <xref ref-type="fig" rid="F10">10</xref>.</p><p>Thus 25/36 = 69% of the patients are classified the same way by both groupings. The patient grouping of Alizadeh <italic>et al.</italic> [<xref ref-type="bibr" rid="B14">14</xref>] was based on a cluster of 380 genes, chosen for their large variation over the samples. Our cluster of 234 genes has 38 genes in common with this cluster of 380, and they are indicated by an asterisk in Table <xref ref-type="table" rid="T2">2</xref>. Five of the 234 genes also appear in the unsupervised clusters found earlier, in the second of the three clusters.</p><fig position="float" id="F5"><label>Figure 5</label><caption><p>Schematic of the gene shaving process.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-5"/></fig><fig position="float" id="F6"><label>Figure 6</label><caption><p>Scatterplot matrix of the three column averages, or `super genes', from each cluster.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-6"/></fig><fig position="float" id="F7"><label>Figure 7</label><caption><p>Percent of gene variance explained by first <italic>j</italic> gene shaving column averages (<italic>j</italic> = 1,2,... 0) (solid curve), and by first <italic>j</italic> principal components (broken curve). For the shaving results, the total number of genes in the first <italic>j</italic> clusters is also indicated.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-7"/></fig><fig position="float" id="F8"><label>Figure 8</label><caption><p><bold>(a)</bold> Variance plots for real and randomized data. The percent variance explained by each cluster, both for the original data, and for an average over three randomized versions. <bold>(b)</bold> Gap estimates of cluster size. The gap curve, which highlights the difference between the pair of curves, is shown.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-8"/></fig><fig position="float" id="F9"><label>Figure 9</label><caption><p>Average (absolute) gene correlation and Cox model <italic>p</italic> value, for clusters of size 200 from supervised shaving and for different values of &#x003b1;. The value of Qa = 0.1 seems best, and is used in the gene shaving procedure.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-9"/></fig><fig position="float" id="F10"><label>Figure 10</label><caption><p>Cluster of 234 genes from supervised shaving.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-10"/></fig><fig position="float" id="F11"><label>Figure 11</label><caption><p><bold>(a)</bold> Gap curve for supervised shaving. <bold>(b)</bold> Survival curves in the two groups defined by the low or high expression of the 234 genes. Group I has high expression of positive genes, and low expression of negative genes; group 2 has low expression of positive genes, and high expression of negative genes. Negative genes are those preceded by a minus sign in Table <xref ref-type="table" rid="T2">2</xref>.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-11"/></fig><fig position="float" id="F12"><label>Figure 12</label><caption><p>The two groups of samples that showed highest and lowest expression of the gene cluster associated with survival.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-12"/></fig><fig position="float" id="F13"><label>Figure 13</label><caption><p>Supervised gene shaving from full gene set. <bold>(a,c)</bold> Partially supervised with &#x003b1; = 0.10; (b,d) fully supervised (&#x003b1; = 1). (a,b) Training set <italic>p</italic> values; (c,d) permutation <italic>p</italic> values for the cluster average as a function of cluster size. The chosen cluster size of 234 is indicated.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-13"/></fig><fig position="float" id="F14"><label>Figure 14</label><caption><p>Kaplan-Meier survival curves in the two groups defined by the cluster of 234 genes shown in Figure <xref ref-type="fig" rid="F10">10</xref>, stratified by IPI. Group I has high expression of positive genes and low expression of negative genes in Figure <xref ref-type="fig" rid="F9">9</xref>, and vice-versa for Group 2.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-14"/></fig><fig position="float" id="F15"><label>Box 1</label><caption><p>The Shaving algorithm.</p></caption><graphic xlink:href="gb-2000-1-2-research0003-15"/></fig><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>The three gene clusters from unsupervised shaving</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Gene number</td><td align="left">Clone ID</td><td align="left">Description</td></tr></thead><tbody><tr><td align="left">Cluster 1</td><td></td><td></td></tr><tr><td align="left">2866</td><td align="left">"139009"</td><td align="left">"Fibronectin I"</td></tr><tr><td align="left">2867</td><td align="left">"358168"</td><td align="left">"Unknown UG Hs. 106127 ESTs, Highly similar to (defline not available 4689136) [H. sapiens]"</td></tr><tr><td align="left">2868</td><td align="left">"323656"</td><td align="left">"MMP-2=Matrix metalloproteinase 2=72 kD type IV collagenase precursor=72 kD gelatinase=gelatinase A=TBE-1"</td></tr><tr><td align="left">2907</td><td align="left">"897910"</td><td align="left">"OSF-2os=osteoblast-specific factor=putative bone adhesion protein with homology with the insect protein fasciclin 1"</td></tr><tr><td align="left">2869</td><td align="left">"359412"</td><td align="left">"Cyclin D2/KIAK0002=overlaps with middle of KIAK0002 cDNA"</td></tr><tr><td align="left">2871</td><td align="left">"754106"</td><td align="left">"TIMP-3=Tissue inhibitor of metalloproteinase 3"</td></tr><tr><td align="left">2865</td><td align="left">"526335"</td><td align="left">"MMP-9=Matrix metalloproteinase 9=92 kD Gelatinase B=92 KD type IV collagenase"</td></tr><tr><td align="left">2870</td><td align="left">"487878"</td><td align="left">"osteonectin=SPARC=basement membrane protein"</td></tr><tr><td align="left">Cluster 2</td><td></td><td></td></tr><tr><td align="left">2820</td><td align="left">"753794"</td><td align="left">"BLC=BCA-1=B lymphocyte chemoattractant BLC=CXC chemokine"</td></tr><tr><td align="left">785</td><td align="left">"1334260"</td><td align="left">"Unknown UG Hs. 120716 ESTs"</td></tr><tr><td align="left">2521</td><td align="left">"713158"</td><td align="left">"Unknown UG Hs.89104 ESTs"</td></tr><tr><td align="left">801</td><td align="left">"701361"</td><td align="left">"Similar to FXI-TI =FX-induced thymoma transcript"</td></tr><tr><td align="left">2720</td><td align="left">"814655"</td><td align="left">"Similar to retinol dehydrogenase type 1 (RODH 1)"</td></tr><tr><td align="left">2721</td><td align="left">"701122"</td><td align="left">"Unknown UG Hs.119410 Homo sapiens cytokine receptor related protein 4 (CYTOR4) mRNA, complete cds"</td></tr><tr><td align="left">2522</td><td align="left">"1272196"</td><td align="left">"IRF-4=LSIRF=Mum l=homologue of Pip=Lymphoid-specific interferon regulatory factor =Multiple myeloma oncogene 1"</td></tr><tr><td align="left">2659</td><td align="left">"685177"</td><td align="left">"PTP-1 B=phosphotyrosyl-protein phosphatase"</td></tr><tr><td align="left">774</td><td align="left">"701606"</td><td align="left">"CD 10=CALLA=Neprilysin=enkepalinase"</td></tr><tr><td align="left">771</td><td align="left">"1305913"</td><td align="left">"Unknown UG Hs.106771 ESTs"</td></tr><tr><td align="left">432</td><td align="left">"417048"</td><td align="left">"Similar to human endogenous retrovirus-4"</td></tr><tr><td align="left">781</td><td align="left">"1367994"</td><td align="left">"myb-related gene A=A-myb"</td></tr><tr><td align="left">2539</td><td align="left">"182764"</td><td align="left">"EB12=Epstein-Barr virus induced G-protein coupled receptor=Putative chemokine receptor"</td></tr><tr><td align="left">757</td><td align="left">"683405"</td><td align="left">"SA3=nuclear protein"</td></tr><tr><td align="left">793</td><td align="left">"1353041"</td><td align="left">"Unknown 166"</td></tr><tr><td align="left">2494</td><td align="left">"1357360"</td><td align="left">"Cyclin D2/KIAK0002=3\325 end of KIAK0002 cDNA"</td></tr><tr><td align="left">2929</td><td align="left">"469297"</td><td align="left">"DECI=basic helix-loop-helix protein"</td></tr><tr><td align="left">728</td><td align="left">"1338981"</td><td align="left">"Unknown UG Hs.l 37038 EST"</td></tr><tr><td align="left">2656</td><td align="left">"814768"</td><td align="left">"Unknown UG Hs.l 93857 ESTs"</td></tr><tr><td align="left">787</td><td align="left">"1338448"</td><td align="left">"Unknown UG Hs.224323 ESTs, Moderately similar to alternatively spliced product using exon 13A [H. sapiens]"</td></tr><tr><td align="left">720</td><td align="left">"815539"</td><td align="left">"JAWI=lymphoid-restricted membrane protein"</td></tr><tr><td align="left">772</td><td align="left">"700718"</td><td align="left">"Unknown UG Hs.202588 ESTs"</td></tr><tr><td align="left">777</td><td align="left">"1352112"</td><td align="left">"FMR2=Fragile X mental retardation 2=putative transcription factor=LAF-4 and AF-4 homologue"</td></tr><tr><td align="left">Cluster 3</td><td></td><td></td></tr><tr><td align="left">546</td><td align="left">"725263"</td><td align="left">"immunoglobulin kappa light chain"</td></tr><tr><td align="left">547</td><td align="left">"1172268"</td><td align="left">"HKG7=cell surface protein in NK and T cells=G-CSF-induced gene"</td></tr></tbody></table><table-wrap-foot><p>The first value given is the gene number in the set of 3624. The second value is the ClonelD. Cross-referencing of this Clone ID with the Accession number is available in the data tables at http://llmpp.nih.gov/lymphoma/data.shtml</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Cluster from supervised shaving applied to full set of 3624 genes</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Position</td><td align="left">ClonelD</td><td align="left">Description</td></tr></thead><tbody><tr><td align="left">"-685"</td><td align="left">"712937"</td><td align="left">"hPMSI=DNA mismatch repair protein=mutL homologue"</td></tr><tr><td align="left">"-3531"</td><td align="left">"1186043"</td><td align="left">"Unknown UG Hs.134746 ESTs,"</td></tr><tr><td align="left">"1661"</td><td align="left">"352820"</td><td align="left">"Unknown UG Hs.231825 ESTs"</td></tr><tr><td align="left">"-2667"</td><td align="left">"1356433"</td><td align="left">"Unknown 645"</td></tr><tr><td align="left">^*"798"</td><td align="left">"814622"</td><td align="left">"Unknown UG Hs.49614 ESTs"</td></tr><tr><td align="left">"-3545"</td><td align="left">"713080"</td><td align="left">"CLK-2=cdc2/CDC28-like protein kinase-2"</td></tr><tr><td align="left">^*"-153"</td><td align="left">"1339106"</td><td align="left">"XE7=B-lymphocyte surface protein"</td></tr><tr><td align="left">^*"824"</td><td align="left">"1356501"</td><td align="left">"Unknown UG Hs.130721 ESTs"</td></tr><tr><td align="left">"-3414"</td><td align="left">"1319801"</td><td align="left">"Similar to non-erythropoietic porphobilinogen deaminase (hydroxymethylbilane synt EC4.3.1.8)"</td></tr><tr><td align="left">"-1577"</td><td align="left">"1353785"</td><td align="left">"Unknown UG Hs.119769 ESTs"</td></tr><tr><td align="left">"-3242"</td><td align="left">"376942"</td><td align="left">"Ro ribonucleoprotein autoantigen (Ro/SS-A)=autoantigen calreticulin"</td></tr><tr><td align="left">^*"-3535"</td><td align="left">"1336373"</td><td align="left">"Similar to High mobility group (nonhistone chromosomal) protein isoforms I and Y"</td></tr><tr><td align="left">"-3412"</td><td align="left">"344219"</td><td align="left">"5'-terminal region of UMK"</td></tr><tr><td align="left">"-673"</td><td align="left">"279363"</td><td align="left">"Adenosine kinase"</td></tr><tr><td align="left">"920"</td><td align="left">"1355987"</td><td align="left">"Unknown UG Hs.180836 EST"</td></tr><tr><td align="left">^*"800"</td><td align="left">"1358163"</td><td align="left">"Phosphatidylinositol 3-kinase p1 10 catalytic, gamma isoform"</td></tr><tr><td align="left">^*"823"</td><td align="left">"1319062"</td><td align="left">"WIP/HS PRPL-2=WASP interacting protein"</td></tr><tr><td align="left">^*"799"</td><td align="left">"1339726"</td><td align="left">"Unknown 168"</td></tr><tr><td align="left">^*"788"</td><td align="left">"825199"</td><td align="left">"Unknown 164"</td></tr><tr><td align="left">"-3544"</td><td align="left">"1285581"</td><td align="left">"Similar to myb-related gene A-myb 5'-region"</td></tr><tr><td align="left">"-68"</td><td align="left">"589589"</td><td align="left">"homolog of Drosophila splicing regulator suppressor-of-white-apricot"</td></tr><tr><td align="left">^*"759"</td><td align="left">"1333557"</td><td align="left">"Unknown 161"</td></tr><tr><td align="left">"339"</td><td align="left">"1336946"</td><td align="left">"Unknown 80"</td></tr><tr><td align="left">"-178"</td><td align="left">"1354703"</td><td align="left">"Unknown UG Hs.150458 ESTs"</td></tr><tr><td align="left">"-933"</td><td align="left">"1184133"</td><td align="left">"CASPASE-3=CPP32 isoform alpha=yama=cysteine protease"</td></tr><tr><td align="left">"-2714"</td><td align="left">"149994"</td><td align="left">"B12 protein=tumor necrosis factor-alpha-induced endothelial primary response gene</td></tr><tr><td align="left">"-3364"</td><td align="left">"271976"</td><td align="left">"ACYI =aminoacylase-I"</td></tr><tr><td align="left">"-118"</td><td align="left">"145409"</td><td align="left">"Low-affinity IgG Fc receptor II-B and C isoforms (multiple orthologous genes)"</td></tr><tr><td align="left">^*"-671"</td><td align="left">"1317098"</td><td align="left">"tyrosine kinase (TnkI)"</td></tr><tr><td align="left">"-2623"</td><td align="left">"324973"</td><td align="left">"9G8 splicing factor"</td></tr><tr><td align="left">^*"783"</td><td align="left">"814601"</td><td align="left">"Unknown UG Hs.161905 EST"</td></tr><tr><td align="left">"2421"</td><td align="left">"1370055"</td><td align="left">"Unknown 602"</td></tr><tr><td align="left">"1855"</td><td align="left">"1358160"</td><td align="left">"Unknown 428"</td></tr><tr><td align="left">^*"813"</td><td align="left">"23173"</td><td align="left">"JNK3=Stress-activated protein kinase"</td></tr><tr><td align="left">"-1412"</td><td align="left">"22438"</td><td align="left">"RYK receptor-like tyrosine kinase"</td></tr><tr><td align="left">"1104"</td><td align="left">"1336779"</td><td align="left">"Unknown 221"</td></tr><tr><td align="left">"1521"</td><td align="left">"1670861"</td><td align="left">"Unknown UG Hs.32533 ESTs"</td></tr><tr><td align="left">"2568"</td><td align="left">"1184568"</td><td align="left">"Unknown UG Hs.120785 ESTs"</td></tr><tr><td align="left">"-3161"</td><td align="left">"365358"</td><td align="left">"pM5 protein=homology to conserved regions of the collagenase gene family"</td></tr><tr><td align="left">"279"</td><td align="left">"1367883"</td><td align="left">"KIAA0430"</td></tr><tr><td align="left">"338"</td><td align="left">"1336591"</td><td align="left">"Unknown UG Hs.180644 ESTs"</td></tr><tr><td align="left">^*"63"</td><td align="left">"746300"</td><td align="left">"Unknown UG Hs.136345 ESTs"</td></tr><tr><td align="left">^*"-2661"</td><td align="left">"1302032"</td><td align="left">"Deoxycytidylate deaminase"</td></tr><tr><td align="left">^*"787"</td><td align="left">"1338448"</td><td align="left">"Unknown UG Hs.224323 ESTs, Moderately similar to alternatively spliced product exon 13A [H.sapiens]"</td></tr><tr><td align="left">"2567"</td><td align="left">"1354788"</td><td align="left">"Unknown 627"</td></tr><tr><td align="left">^*"758"</td><td align="left">"1333558"</td><td align="left">"Unknown 160"</td></tr><tr><td align="left">"-3264"</td><td align="left">"704732"</td><td align="left">"Unknown 699"</td></tr><tr><td align="left">"-2654"</td><td align="left">"724397"</td><td align="left">"lymphopain=C 1 peptidase expressed in natural killer and cytotoxic T cells"</td></tr><tr><td align="left">"1132"</td><td align="left">"1354522"</td><td align="left">"Unknown UG Hs.125285 ESTs, Highly similar to (defline not available 4200446) [Mlus]"</td></tr><tr><td align="left">^*"-1595"</td><td align="left">"1186040"</td><td align="left">"Unknown UG Hs.136589 ESTs"</td></tr><tr><td align="left">"-2320"</td><td align="left">"241481"</td><td align="left">"CASPASE-10=Mch4=FLICE2"</td></tr><tr><td align="left">"-3345"</td><td align="left">"502761"</td><td align="left">"Phosphoribosylglycinamide formyltransferase, phosphoribosylglycinamide synthetase phoribosylaminoimidazole synthetase"</td></tr><tr><td align="left">"-33"</td><td align="left">"268727"</td><td align="left">"MYH=DNA mismatch repair protein=mutY homologue"</td></tr><tr><td align="left">^*"774"</td><td align="left">"701606"</td><td align="left">"CDIO=CALLA=Neprilysin=enkepalinase"</td></tr><tr><td align="left">"-533"</td><td align="left">"276483"</td><td align="left">"(2'-5') oligoadenylate synthetase E"</td></tr><tr><td align="left">"1388"</td><td align="left">"1350824"</td><td align="left">"Unknown UG Hs.163773 ESTs"</td></tr><tr><td align="left">"-3244"</td><td align="left">"488754"</td><td align="left">"DAP-1 =putative mediator of the gamma interferon-induced cell death"</td></tr><tr><td align="left">"3097"</td><td align="left">"686331"</td><td align="left">"DCHT=Similar to rat pancreatic serine threonine kinase"</td></tr><tr><td align="left">"-2641"</td><td align="left">"1355868"</td><td align="left">"Unknown 643"</td></tr><tr><td align="left">"-3135"</td><td align="left">"199018"</td><td align="left">"P120=proliferating-cell nudeolar protein"</td></tr><tr><td align="left">"-1578"</td><td align="left">"713301"</td><td align="left">"Unknown UG Hs.32218 ESTs,</td></tr><tr><td align="left">"-2502"</td><td align="left">"153355"</td><td align="left">"LD78 beta=almost identical to MIP-1 alpha=chemokine"</td></tr><tr><td align="left">"2328"</td><td align="left">"1341026"</td><td align="left">"yotiao=protein of neuronal and neuromuscular synapses that interacts with specific variants of NMDA receptor subunit NR 1"</td></tr><tr><td align="left">"1863"</td><td align="left">"1357676"</td><td align="left">"Unknown UG Hs.191211 ESTs"</td></tr><tr><td align="left">"1399"</td><td align="left">"1356420"</td><td align="left">"Unknown UG Hs.207995 ESTs"</td></tr><tr><td align="left">"-3401"</td><td align="left">"844479"</td><td align="left">"Pig8=p53 inducible gene=etoposide-induced mRNA=Similar to El 24 = p53 responsive (sculus)"</td></tr><tr><td align="left">"-3040"</td><td align="left">"1368740"</td><td align="left">"Unknown UG Hs.125307 EST"</td></tr><tr><td align="left">"-3193"</td><td align="left">"152653"</td><td align="left">"C-I-Tetrahydrofolate Synthase, cytoplasmic"</td></tr><tr><td align="left">"-3437"</td><td align="left">"814765"</td><td align="left">"kinase A anchor protein"</td></tr><tr><td align="left">"1387"</td><td align="left">"1318821"</td><td align="left">"Unknown UG Ms. 108614 Homo sapiens mRNAfor KIAA0627 protein, partial cds"</td></tr><tr><td align="left">"-2527"</td><td align="left">"1357085"</td><td align="left">"Acidic 82 kDa protein"</td></tr><tr><td align="left">^*"1400"</td><td align="left">"682995"</td><td align="left">"Unknown 298"</td></tr><tr><td align="left">^*"724"</td><td align="left">"1286796"</td><td align="left">"Unknown UG Hs.61506 ESTs"</td></tr><tr><td align="left">"413"</td><td align="left">"1334297"</td><td align="left">"Unknown 98"</td></tr><tr><td align="left">^*"789"</td><td align="left">"825217"</td><td align="left">"Unknown UG Hs.169565 ESTs,</td></tr><tr><td align="left">"-2754"</td><td align="left">"1318136"</td><td align="left">"5'-AMP-activated protein kinase, gamma-I subunit"</td></tr><tr><td align="left">"1052"</td><td align="left">"1240803"</td><td align="left">"Unknown 211"</td></tr><tr><td align="left">"278"</td><td align="left">"815671"</td><td align="left">"Unknown UG Hs.101340 ESTs"</td></tr><tr><td align="left">"-2501"</td><td align="left">"346550"</td><td align="left">"MIP-1 alpha=LD78 alpha=pAT464=Small inducible cytokine A3=macrophage inflammatory in (GOS 19-1)=chemokine"</td></tr><tr><td align="left">"1988"</td><td align="left">"1320268"</td><td align="left">"Unknown 480"</td></tr><tr><td align="left">"-903"</td><td align="left">"704637"</td><td align="left">"Unknown UG Hs.5354 ESTs"</td></tr><tr><td align="left">"-2649"</td><td align="left">"181998"</td><td align="left">"NFAT3=NFATc4"</td></tr><tr><td align="left">"-2648"</td><td align="left">"171693"</td><td align="left">"Lst-1 =IC7=interferon-gamma-inducible gene present in lymphoid tissues, T cells, macrophages, and histiocyte cell lines</td></tr><tr><td></td><td></td><td align="left">encoding a transmembrane protein"</td></tr><tr><td align="left">"2373"</td><td align="left">"1338072"</td><td align="left">"Unknown 592"</td></tr><tr><td align="left">"223"</td><td align="left">"1352327"</td><td align="left">"Unknown 52"</td></tr><tr><td align="left">"1269"</td><td align="left">"1339210"</td><td align="left">"Unknown 261"</td></tr><tr><td align="left">"-3004"</td><td align="left">"1289545"</td><td align="left">"Unknown UG Hs.187869 ESTs"</td></tr><tr><td align="left">"1177"</td><td align="left">"700949"</td><td align="left">"Similar to myosin-IXb"</td></tr><tr><td align="left">^*"779"</td><td align="left">"703735"</td><td align="left">"Unknown UG Hs.28355 ESTs"</td></tr><tr><td align="left">^*"464"</td><td align="left">"685761"</td><td align="left">"Unknown III"</td></tr><tr><td align="left">"1229"</td><td align="left">"700643"</td><td align="left">"Unknown UG Hs.104492 ESTs"</td></tr><tr><td align="left">"-3482"</td><td align="left">"51058"</td><td align="left">"E2F-4=pRB-binding transcription factor"</td></tr><tr><td align="left">"-3584"</td><td align="left">"1358191"</td><td align="left">"Similar to DNA polymerase beta=DNA alkylation repair protein"</td></tr><tr><td align="left">^*"-429"</td><td align="left">"35356"</td><td align="left">"Neurotrophic tyrosine kinase, receptor, type 3 (TrkC)"</td></tr><tr><td align="left">"-3136"</td><td align="left">"265590"</td><td align="left">"NFI =Neurofibromin"</td></tr><tr><td align="left">"956"</td><td align="left">"1289384"</td><td align="left">"Unknown 198"</td></tr><tr><td align="left">"2491"</td><td align="left">"814251"</td><td align="left">"SLAM=signaling lymphocytic activation molecule"</td></tr><tr><td align="left">"2083"</td><td align="left">"1353083"</td><td align="left">"Unknown UG Hs.136972 EST"</td></tr><tr><td align="left">"1102"</td><td align="left">"1372068"</td><td align="left">"KIAA0603=Similar to TBCI"</td></tr><tr><td align="left">"-1010"</td><td align="left">"595474"</td><td align="left">"Pak 1 =p21-activated protein kinase"</td></tr><tr><td align="left">"-3594"</td><td align="left">"1269836"</td><td align="left">"BCL-7B"</td></tr><tr><td align="left">"-2270"</td><td align="left">"265267"</td><td align="left">"HSP70"</td></tr><tr><td align="left">"-944"</td><td align="left">"1337124"</td><td align="left">"Unknown UG Hs.81248 CUG triplet repeat, RNA-binding protein 1"</td></tr><tr><td align="left">"-3330"</td><td align="left">"1301224"</td><td align="left">"Elongin B=RNA polymerase II transcription factor SIII pi 8 subunit"</td></tr><tr><td align="left">"1658"</td><td align="left">"1241118"</td><td align="left">"Unknown 346"</td></tr><tr><td align="left">"-3140"</td><td align="left">"841361"</td><td align="left">"GR02=GRO beta=MIP2 alpha=macrophage inflammatory protein-2 alpha=chemokine"</td></tr><tr><td align="left">"-2651"</td><td align="left">"525540"</td><td align="left">"BCL-3"</td></tr><tr><td align="left">"-3350"</td><td align="left">"1186114"</td><td align="left">"Unknown UG Hs.116447 EST"</td></tr><tr><td align="left">"-2990"</td><td align="left">"1289569"</td><td align="left">"Unknown UG Hs.146165 ESTs"</td></tr><tr><td align="left">^*"809"</td><td align="left">"1270618"</td><td align="left">"Unknown UG Hs.208970 EST, Weakly similar to neuronal thread protein AD7c-NTP [ens]"</td></tr><tr><td align="left">"-3160"</td><td align="left">"703707"</td><td align="left">"Protein disulfide isomerase-related protein (PDIR)"</td></tr><tr><td align="left">"874"</td><td align="left">"1320313"</td><td align="left">"Unknown UG Hs.132458 ESTs"</td></tr><tr><td align="left">"-3390"</td><td align="left">"1339763"</td><td align="left">"Unknown 710"</td></tr><tr><td align="left">"1343"</td><td align="left">"1318717"</td><td align="left">"LOK=lymphocyte oriented kinase=STE20-like protein kinase that is expressed predominantly in lymphocytes"</td></tr><tr><td align="left">"-179"</td><td align="left">"301551"</td><td align="left">"Integrin, alpha V (vitronectin receptor, alpha polypeptide, antigen CD51)"</td></tr><tr><td align="left">^*"723"</td><td align="left">"824754"</td><td align="left">"Unknown UG Hs.145058 ESTs"</td></tr><tr><td align="left">"-3406"</td><td align="left">"1300230"</td><td align="left">"Unknown UG Hs.56421 ESTs, Weakly similar to Similarity to H.influenza ribonucl H [C.elegans]"</td></tr><tr><td align="left">"-573"</td><td align="left">"1341161"</td><td align="left">"Similar to rhoGap protein"</td></tr><tr><td align="left">^*"722"</td><td align="left">"1341225"</td><td align="left">"Unknown UG Hs.186709 ESTs,! [H.sapiens]"</td></tr><tr><td align="left">"2212"</td><td align="left">"1350784"</td><td align="left">"Unknown UG Hs.163202 EST"</td></tr><tr><td align="left">"-3478"</td><td align="left">"417897"</td><td align="left">"cleavage stimulation factor 77kDa subunit=polyadenylation factor subunit=homolog the Drosophila suppressor of forked protein"</td></tr><tr><td align="left">"-887"</td><td align="left">"756965"</td><td align="left">"RGS14=regulator of G protein signaling"</td></tr><tr><td align="left">"1344"</td><td align="left">"825333"</td><td align="left">"Unknown UG Hs.193017 ESTs, Highly similar to (defline not available 4220898) [ens]"</td></tr><tr><td align="left">^*"743"</td><td align="left">"1358192"</td><td align="left">"Unknown UG Hs.228205 EST,</td></tr><tr><td align="left">"1850"</td><td align="left">"1353072"</td><td align="left">"Unknown 426"</td></tr><tr><td align="left">"-3391"</td><td align="left">"1340604"</td><td align="left">"Unknown UG Hs.127121 ESTs"</td></tr><tr><td align="left">"-236"</td><td align="left">"686771"</td><td align="left">"tubulin-gamma"</td></tr><tr><td align="left">"-3343"</td><td align="left">"293934"</td><td align="left">"CAS=chromosome segregation gene homolog"</td></tr><tr><td align="left">"2566"</td><td align="left">"1350728"</td><td align="left">"Unknown 626"</td></tr><tr><td align="left">"-2984"</td><td align="left">"955354"</td><td align="left">"putative cell surface ligand for T1/ST2 receptor (related to IL-1 receptors)"</td></tr><tr><td align="left">"-3149"</td><td align="left">"366713"</td><td align="left">"GSK3=glycogen synthase kinase 3"</td></tr><tr><td align="left">^*"720"</td><td align="left">"815539"</td><td align="left">"JAWI=lymphoid-restricted membrane protein"</td></tr><tr><td align="left">"-3177"</td><td align="left">"378364"</td><td align="left">"PRODH=proline dehydrogenase/proline oxidase=p53-induced gene"</td></tr><tr><td align="left">"1268"</td><td align="left">"1339305"</td><td align="left">"Unknown 260"</td></tr><tr><td align="left">"-3616"</td><td align="left">"1302092"</td><td align="left">"Unknown UG Hs.214428 ESTs"</td></tr><tr><td align="left">"1210"</td><td align="left">"685368"</td><td align="left">"Unknown 243"</td></tr><tr><td align="left">"2330"</td><td align="left">"1240688"</td><td align="left">"Unknown 577"</td></tr><tr><td align="left">"259"</td><td align="left">"1369262"</td><td align="left">"KIAAOO 19=similar to transforming protein tre"-2528" "1184411" "MINOR=mitogen induced nuclear orphan receptor=NOR-l=Nur77 orphan nuclear receptor family member"</td></tr><tr><td align="left">"-3586"</td><td align="left">"1309295"</td><td align="left">"Unknown UG Hs.136985 ESTs"</td></tr><tr><td align="left">"2045"</td><td align="left">"1352570"</td><td align="left">"Unknown 494"</td></tr><tr><td align="left">"2067"</td><td align="left">"1320316"</td><td align="left">"Unknown 508"</td></tr><tr><td align="left">"-3533"</td><td align="left">"298303"</td><td align="left">"TECK chemokine"</td></tr><tr><td align="left">"-3530"</td><td align="left">"1355240"</td><td align="left">"Unknown UG Hs.130849 ESTs"</td></tr><tr><td align="left">^*"-2469"</td><td align="left">"417226"</td><td align="left">"c-myc"</td></tr><tr><td align="left">"1784"</td><td align="left">"1355354"</td><td align="left">"Unknown 394"</td></tr><tr><td align="left">"-3023"</td><td align="left">"700772"</td><td align="left">"Smad2=Madr2=JV 18-1 =Homologue of Mothers Against Decapentaplegic (MAD)=Activated beta"</td></tr><tr><td align="left">^*"793"</td><td align="left">"1353041"</td><td align="left">"Unknown 166"</td></tr><tr><td align="left">"-3162"</td><td align="left">"1289546"</td><td align="left">"Similar to arginine/aspartate-rich 37.3 K protein"</td></tr><tr><td align="left">^*"-2669"</td><td align="left">"1186215"</td><td align="left">"Unknown UG Hs.190288 EST"</td></tr><tr><td align="left">"-113"</td><td align="left">"1337185"</td><td align="left">"KIAA0037"</td></tr><tr><td align="left">"-3434"</td><td align="left">"1338032"</td><td align="left">"CPR2=cell cycle progression 2"</td></tr><tr><td align="left">"-2621"</td><td align="left">"1338456"</td><td align="left">"c-myc binding protein"</td></tr><tr><td align="left">"1333"</td><td align="left">"824376"</td><td align="left">"Similar to (AFO 16450) Similar to acyltransferase"</td></tr><tr><td align="left">"-3405"</td><td align="left">"1334813"</td><td align="left">"Unknown UG Hs.17883 protein phosphatase IG (formerly 2C), magnesium-dependent, isoform"</td></tr><tr><td align="left">"2301"</td><td align="left">"300051"</td><td align="left">"myosin light chain-2"</td></tr><tr><td align="left">"1144"</td><td align="left">"1372011"</td><td align="left">"Unknown UG Hs.209146 ESTs"</td></tr><tr><td align="left">"-3436"</td><td align="left">"485171"</td><td align="left">"methionine adenosyltransferase alpha subunit"</td></tr><tr><td align="left">"1339"</td><td align="left">"1355713"</td><td align="left">"Unknown 277"</td></tr><tr><td align="left">"1156"</td><td align="left">"1351290"</td><td align="left">"Similar to (Z49125) C47G2.4"</td></tr><tr><td align="left">^*"721"</td><td align="left">"1353015"</td><td align="left">"Unknown 154"</td></tr><tr><td align="left">"-3125"</td><td align="left">"86040"</td><td align="left">"Cytochrome P450, subfamily 1, polypeptide 2 (aromatic compound-inducible)"</td></tr><tr><td align="left">"258"</td><td align="left">"1367988"</td><td align="left">"Unknown 61"</td></tr><tr><td align="left">"-3258"</td><td align="left">"1304523"</td><td align="left">"APRT=adenine phosphoribosyltransferase"</td></tr><tr><td align="left">"-3548"</td><td align="left">"1340120"</td><td align="left">"Unknown 733"</td></tr><tr><td align="left">"151 1"</td><td align="left">"1351701"</td><td align="left">"Unknown UG Hs.124230 ESTs"</td></tr><tr><td align="left">"-3280"</td><td align="left">"826594"</td><td align="left">"replication factor C"</td></tr><tr><td align="left">"-3363"</td><td align="left">"293035"</td><td align="left">"APEX=apurinic endonuclease=DNA alkylation repair protein"</td></tr><tr><td align="left">"1190"</td><td align="left">"1371313"</td><td align="left">"Similar to G-protein coupled receptor pH218"</td></tr><tr><td align="left">"1321"</td><td align="left">"1309301"</td><td align="left">"Unknown UG Hs.136987 EST"</td></tr><tr><td align="left">"-3180"</td><td align="left">"591683"</td><td align="left">"GADD45 alpha=growth arrest and DNA-damage-inducible protein alpha"</td></tr><tr><td align="left">"1748"</td><td align="left">"1371159"</td><td align="left">"Unknown 377"</td></tr><tr><td align="left">"-2781"</td><td align="left">"1288183"</td><td align="left">"BAK=BCL-2 family member"</td></tr><tr><td align="left">"108"</td><td align="left">"1370125"</td><td align="left">"Unknown 22"</td></tr><tr><td align="left">"-2941"</td><td align="left">"742132"</td><td align="left">"Interferon-induced 17 KD protein"</td></tr><tr><td align="left">"-2994"</td><td align="left">"1271685"</td><td align="left">"Unknown UG Hs.176669 ESTs"</td></tr><tr><td align="left">"1287"</td><td align="left">"1353226"</td><td align="left">"Unknown UG Hs.30209 Homo sapiens mRNA for KIAA0854 protein, complete cds"</td></tr><tr><td align="left">"1039"</td><td align="left">"1671442"</td><td align="left">"Unknown UG Hs.171096 ESTs, Weakly similar to (defline not available 4456988) [ens]"</td></tr><tr><td align="left">^*"83"</td><td align="left">"52408"</td><td align="left">"ABR=guanine nucleotide regulatory protein"</td></tr><tr><td align="left">"3624"</td><td align="left">"1355859"</td><td align="left">"Similar to myosin IE heavy chain"</td></tr><tr><td align="left">"-2746"</td><td align="left">"1350736"</td><td align="left">"IRF-3=interferon regulatory factor-3"</td></tr><tr><td align="left">"1303"</td><td align="left">"665682"</td><td align="left">"Jnkk2=JNK kinase 2=MAP kinase kinase"</td></tr><tr><td align="left">"877"</td><td align="left">"1367968"</td><td align="left">"Unknown UG Hs.105072 ESTs"</td></tr><tr><td align="left">"-3344"</td><td align="left">"1341245"</td><td align="left">"CD73=5' nudeotidase"</td></tr><tr><td align="left">"1191"</td><td align="left">"1371317"</td><td align="left">"Similar to arylacetyltransferase"</td></tr><tr><td align="left">^*"-310"</td><td align="left">"154493"</td><td align="left">"HNPP=nuclear phosphoprotein"</td></tr><tr><td align="left">"1976"</td><td align="left">"1334933"</td><td align="left">"Unknown UG Hs.144684 ESTs"</td></tr><tr><td align="left">"-2609"</td><td align="left">"1670958"</td><td align="left">"SRF=c-fos serum response element-binding transcription factor"</td></tr><tr><td align="left">"405"</td><td align="left">"701689"</td><td align="left">"putative tumor suppressor (LUCA15)"</td></tr><tr><td align="left">"-3319"</td><td align="left">"1307997"</td><td align="left">"Similar to bromodeoxyuridine-sensitive transcript protein=px 19"</td></tr><tr><td align="left">"-3255"</td><td align="left">"810743"</td><td align="left">"MLF2=myelodysplasia/myeloid leukemia factor 2"</td></tr><tr><td align="left">"2150"</td><td align="left">"1353466"</td><td align="left">"Unknown UG Hs.124360 EST"</td></tr><tr><td align="left">"-2650"</td><td align="left">"511407"</td><td align="left">"69 kDa 2'5' oligoadenylate synthetase (P69 2-5A synthetase)"</td></tr><tr><td align="left">"252"</td><td align="left">"1356345"</td><td align="left">"Unknown UG Hs.49500 Homo sapiens mRNA for KIAA0746 protein, partial cds"</td></tr><tr><td align="left">"1337"</td><td align="left">"1367875"</td><td align="left">"Unknown UG Hs.128127 ESTs"</td></tr><tr><td align="left">"1302"</td><td align="left">"1351266"</td><td align="left">"Unknown UG Hs.134197 ESTs, Moderately similar to FAM [M.musculus]"</td></tr><tr><td align="left">"1386"</td><td align="left">"815165"</td><td align="left">"Unknown UG Hs.188732 ESTs"</td></tr><tr><td align="left">"-3147"</td><td align="left">"549277"</td><td align="left">"cell cycle protein p38-2G4 homolog (hG4-l)"</td></tr><tr><td align="left">"-3349"</td><td align="left">"1355524"</td><td align="left">"Similar to rapamycin-binding protein (FKBP25)"</td></tr><tr><td align="left">"-173"</td><td align="left">"1287032"</td><td align="left">"Similar to Drosophila female sterile homeotic (FSH) homologue"</td></tr><tr><td align="left">^*"777"</td><td align="left">"1352112"</td><td align="left">"FMR2=Fragile X mental retardation 2=putative transcription factor=LAF-4 and AF-4 ogue"</td></tr><tr><td align="left">"-3334"</td><td align="left">"346948"</td><td align="left">"nm23-H2=NDP kinase B=Nucleoside dephophate kinase B"</td></tr><tr><td align="left">"-3256"</td><td align="left">"1303575"</td><td align="left">"Unknown UG Hs.123304 ESTs"</td></tr><tr><td align="left">"1289"</td><td align="left">"704690"</td><td align="left">"Dyrk6=Ser/Thr protein kinase"</td></tr><tr><td align="left">"1133"</td><td align="left">"1351498"</td><td align="left">"Unknown UG Hs.189063 ESTs"</td></tr><tr><td align="left">"2058"</td><td align="left">"1339890"</td><td align="left">"Unknown 503"</td></tr><tr><td align="left">"-2927"</td><td align="left">"342647"</td><td align="left">"MAPKAP kinase (3pK)"</td></tr><tr><td align="left">"1324"</td><td align="left">"687198"</td><td align="left">"Unknown UG Hs.125860 ESTs"</td></tr><tr><td align="left">"-3047"</td><td align="left">"203704"</td><td align="left">"flavin-containing monooxygenase (FMO 1)"</td></tr><tr><td align="left">"-2662"</td><td align="left">"1288102"</td><td align="left">"Similar to nuclear-encoded mitochondrial NADH-ubiquinone reductase 24Kd subunit"</td></tr><tr><td align="left">"1852"</td><td align="left">"1371200"</td><td align="left">"Similar to (Z78012) C52E4.6"</td></tr><tr><td align="left">"1383"</td><td align="left">"1319529"</td><td align="left">"Unknown 293"</td></tr><tr><td align="left">"-3360"</td><td align="left">"1671396"</td><td align="left">"Similar to friend of GATA-1 (FOG)=zinc finger GATA-I coactivator in erythroid and megakaryocyte lineages"</td></tr><tr><td align="left">"1228"</td><td align="left">"1336501"</td><td align="left">"Unknown 249"</td></tr><tr><td align="left">"1353"</td><td align="left">"1356762"</td><td align="left">"Unknown UG Hs.127480 ESTs"</td></tr><tr><td align="left">^*"-575"</td><td align="left">"490387"</td><td align="left">"zinc finger protein 42 MZF-1"</td></tr><tr><td align="left">"1242"</td><td align="left">"1031754"</td><td align="left">"Protein-tyrosine phosphatase 2C"</td></tr><tr><td align="left">"1211"</td><td align="left">"1372274"</td><td align="left">"Unknown UG Hs.208983 ESTs,</td></tr><tr><td align="left">"-2759"</td><td align="left">"489438"</td><td align="left">"MyD88=myeloid differentiation primary response protein=death domain-containing p</td></tr><tr><td align="left">"1227"</td><td align="left">"1334962"</td><td align="left">"Similar to KIAA0437"</td></tr><tr><td align="left">"260"</td><td align="left">"1341211"</td><td align="left">"Unknown UG Hs.191209 ESTs"</td></tr><tr><td align="left">"-3137"</td><td align="left">"1250770"</td><td align="left">"Purine nucleoside phophorylase=lnosine phosphorylase=PNP"</td></tr><tr><td align="left">"1385"</td><td align="left">"1371029"</td><td align="left">"Unknown 295"</td></tr><tr><td align="left">"1808"</td><td align="left">"1372833"</td><td align="left">"Unknown 403"</td></tr><tr><td align="left">"-2762"</td><td align="left">"1184153"</td><td align="left">"Unknown UG Hs.230206 EST"</td></tr><tr><td align="left">"1046"</td><td align="left">"1352940"</td><td align="left">"Unknown 208"</td></tr><tr><td align="left">"-2766"</td><td align="left">"756452"</td><td align="left">"tyk2=non-receptor protein tyrosine kinase"</td></tr><tr><td align="left">"1204"</td><td align="left">"1370570"</td><td align="left">"Lamin B receptor (LBR)"</td></tr><tr><td align="left">"1201"</td><td align="left">"1241671"</td><td align="left">"Similar to (AE000860) conserved protein [Methanobacterium thermoautotrophicum]"</td></tr><tr><td align="left">^*"735"</td><td align="left">"686893"</td><td align="left">"Unknown UG Hs.226955 ESTs"</td></tr><tr><td align="left">"1338"</td><td align="left">"1370103"</td><td align="left">"Unknown 276"</td></tr><tr><td align="left">"255"</td><td align="left">"1338624"</td><td align="left">"Unknown UG Hs.192864 ESTs"</td></tr><tr><td align="left">"1200"</td><td align="left">"1352335"</td><td align="left">"Unknown UG Hs.99701 ESTs"</td></tr><tr><td align="left">"2133"</td><td align="left">"1340880"</td><td align="left">"Cancer associated surface antigen (RCASI)"</td></tr></tbody></table><table-wrap-foot><p>Genes are ordered from strongest to weakest correlation with survival. The first number is the position in the hierarchical clustering ordering (a minus sign indicates the sign of the gene is to be flipped before averaging); ^* indicates a gene that also falls in the 380 gene cluster from Alizadeh <italic>et al.</italic> [<xref ref-type="bibr" rid="B14">14</xref>].</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T3"><label>Table 3</label><caption><p>Cross-tabulation of gene shaving groups with IPI index</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center" colspan="2">IPI</td></tr><tr><td></td><td colspan="2"><hr></hr></td></tr><tr><td></td><td align="center">Low</td><td align="center">High</td></tr></thead><tbody><tr><td align="left">Gene shaving groups</td><td></td><td></td></tr><tr><td align="left">1</td><td align="center">7</td><td align="center">7</td></tr><tr><td align="left">2</td><td align="center">11</td><td align="center">7</td></tr></tbody></table></table-wrap><table-wrap position="float" id="T4"><label>Table 4</label><caption><p>A comparison of the patient groups obtained by gene shaving with those obtained previously [<xref ref-type="bibr" rid="B14">14</xref>]</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="left">Patient groups of Alizadeh <italic>et al.</italic> [<xref ref-type="bibr" rid="B14">14</xref>]</td><td></td></tr><tr><td></td><td colspan="2"><hr></hr></td></tr><tr><td></td><td align="center">Low</td><td align="center">High</td></tr></thead><tbody><tr><td align="left">Gene shaving groups</td><td></td><td></td></tr><tr><td align="left">1</td><td align="center">13</td><td align="center">5</td></tr><tr><td align="left">2</td><td align="center">6</td><td align="center">12</td></tr></tbody></table></table-wrap></sec></sec><sec><title>Conclusions</title><p>We have proposed a set of 'shaving' methods for isolating interesting clusters of genes from a set of DNA microarray experiments. The methods may be unsupervised, or may be supervised - that is, use information available about the samples such as a class label or survival time. The proposed shaving methods search for clusters of genes showing both high variation across the samples, and coherence (correlation) across the genes. Both of these aspects are important and cannot be captured by simple clustering of the genes, or thresholding of individual genes based on the variation over samples.</p><p>With our model-based approach for supervised shaving, one can incorporate other prognostic factors in the search for interesting gene clusters. If an outcome such as survival time is available for each sample, the method searches for a gene cluster whose column average gene <inline-graphic xlink:href="gb-2000-1-2-research0003-I12.gif"/> has a significant effect, possibly the presence of other prognostic factors, for predicting the outcome.</p><p>The microarray data <italic>x_ij</italic> we have considered are real-valued expression levels. However, other kinds of arrays produce different kinds of data. In particular, some arrays detect the presence or absence of single-nucleotide polymorphisms (SNPs), so that the <italic>x_ij</italic> values take on one of <italic>k</italic> &#x02265; 2 unordered values. The shaving methods described can be easily modified to handle this kind of data. In detail, we construct <italic>k</italic> data matrices <italic>X</italic>₁, <italic>X</italic>₂ ... <italic>X_k</italic>, each of size <italic>n</italic> &#x000d7; <italic>m</italic>. The <italic>ij</italic>th element of <italic>X</italic>_{<italic>j</italic>} is 1 if x_{<italic>ij</italic>} falls in class <italic>j,</italic> and zero otherwise. Letting &#x02211;_{<italic>j</italic>}<italic>j</italic> = 1, 2, ... <italic>k</italic> be the <italic>n</italic> &#x000d7; <italic>n</italic> covariance matrices of the genes in each <italic>X</italic>_{<italic>j</italic>}, we simply apply principal component shaving, using <inline-graphic xlink:href="gb-2000-1-2-research0003-I14.gif"/> &#x02211;_{<italic>j</italic>}, as the variance matrix for the penalty. This can be done unsupervised, or a supervision term can also be added.</p></sec>

Exclusion of <italic>EDNRB</italic> and <italic>KIT</italic> as the basis for white spotting in Border Collies

<sec><title>Background:</title><p>White spotting patterns in mammals can be caused by mutations in the genes for the endothelin B receptor and c-Kit, whose protein products are necessary for proper migration, differentiation or survival of the melanoblast population of cells. Although there are many different dog breeds that segregate white spotting patterns, no genes have been identified that are linked to these phenotypes.</p></sec><sec><title>Results:</title><p>An intercross was generated from a female Newfoundland and a male Border Collie and the white spotting phenotypes of the intercross progeny were evaluated by measuring percentage surface area of white in the puppies. The Border Collie markings segregated as a simple autosomal recessive (7/25 intercross progeny had the phenotype). Two candidate genes, for the endothelin B receptor (<italic>EDNRB</italic>) and c-Kit (<italic>KIT</italic>), were evaluated for segregation with the white spotting pattern. Polymorphisms between the Border Collie and Newfoundland were identified for <italic>EDNRB</italic> using Southern analysis after a portion of the canine gene had been cloned. Polymorphisms for <italic>KIT</italic> were identified using a microsatellite developed from a bacterial artificial chromosome containing the canine gene.</p></sec><sec><title>Conclusions:</title><p>Both <italic>EDNRB</italic> and <italic>KIT</italic> were excluded as a cause of the white spotting pattern in at least two of the intercross progeny. Although these genes have been implicated in white spotting in other mammals, including horses, pigs, cows, mice and rats, they do not appear to be responsible for the white spotting pattern found in the Border Collie breed of dog.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Metallinos</surname><given-names>Danika</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Rine</surname><given-names>Jasper</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib>

Genome Biology

<sec><title>Background</title><p>The genetics of coat color has been studied for many years in a variety of mammals, and the inheritance patterns of many of the relevant genes have been determined. Both scientists and breeders have studied the range of color and pattern that can be found in mammals. Dogs are uniquely suited for the investigation of the inheritance of coat color and patterns as the more than 200 different dog breeds are defined in part by a specific set of colors and patterns. 'White spotting' in mice, rats, dogs and horses is characterized by irregular white patches of skin and hair that are devoid of pigment-producing melanocytes. White spotting in domestic dog breeds has been postulated to be controlled by one locus, called <italic>s</italic>, which has at least three alleles [<xref ref-type="bibr" rid="B1">1</xref>]. The three alleles segregate in some breeds, while other breeds are homozygous and show a standard spotting pattern. Although the variation in coat colors and patterns in dogs is far greater than in other species, the genes responsible for coat color in dogs have yet to be identified at the molecular level, with the exception of that for the yellow coat color [<xref ref-type="bibr" rid="B2">2</xref>].</p><p>Genes responsible for white spotting in mice and horses and for hypopigmentation defects in humans have been identified. One of these, <italic>EDNRB</italic>, encoding the endothelin B receptor, causes white spotting in the <italic>Ednrb</italic> mouse [<xref ref-type="bibr" rid="B3">3</xref>]. Some alleles of <italic>Ednrb</italic> are associated with more severe defects, such as deafness and aganglionic megacolon. Mutations in the endothelin B receptor gene are also responsible for Hirshsprung disease in humans [<xref ref-type="bibr" rid="B4">4</xref>]. This disease is characterized by intestinal aganglionosis, which is occasionally associated with hypopigmentation and/or deafness. A similar syndrome in horses, called lethal white foal syndrome, is due to a mutation in the horse endothelin B receptor gene [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>].</p><p>Mutations in a second gene, <italic>cKIT</italic>, which encodes a receptor tyrosine kinase, cause the piebald trait in humans [<xref ref-type="bibr" rid="B8">8</xref>]. Piebaldism is characterized by white patches on skin and hair. In horses, a similar phenotype, called Tobiano, is caused by a mutation in either <italic>KIT</italic> or a closely linked gene [<xref ref-type="bibr" rid="B9">9</xref>]. A duplication of <italic>KIT</italic> causes white spotting in pigs [<xref ref-type="bibr" rid="B10">10</xref>]. <italic>EDNRB</italic> and <italic>KIT</italic> are therefore two likely candidates for white spotting in dogs.</p></sec><sec><title>Results and discussion</title><p>As a resource for building a dog genetic map and as a tool to study the genes responsible for behavioral and morphological differences in the dog, an intercross was created between a male Border Collie and a female Newfoundland. The Newfoundland parent had a small patch of white on the chest and was otherwise completely black (Figure <xref ref-type="fig" rid="F1">1a</xref>). The Border Collie used in this cross had markings characteristic for the breed - black with white markings on the face, chest, neck, tail tip, ventral abdomen, all four digits, and extending up the front legs to the carpals (Figure <xref ref-type="fig" rid="F1">1b</xref>). These markings have many similarities to the white spotting patterns of other mammals. The Border Collie's sire and dam had the same markings, consistent with homozygosity for the causative loci. This cross provides the opportunity for analyzing the inheritance of the white spotting pattern exhibited by the Border Collie. Six F1 animals were produced which had medium-sized white patches on their chests. These six dogs were intercrossed to produce 25 F2 progeny. In the F2 generation, 7/25 had markings like the Border Collie parent, consistent with the phenotype being caused by a recessive allele of a single locus.</p><p>To determine if <italic>EDNRB</italic> was responsible for the white spotting pattern of the Border Collie, a portion of canine cDNA was cloned. The amino-acid sequence of canine endothelin B receptor was highly homologous to that of other mammalian endothelin B receptors (Figure <xref ref-type="fig" rid="F2">2</xref>). The canine cDNA clone was used as a hybridization probe for DNA hybridization blot experiments. A <italic>Pvu</italic>II restriction fragment length polymorphism (RFLP) was identified between the Border Collie grandparent and the Newfoundland grandparent of the cross. Spotted F2 progeny carried at least one allele of this polymorphism that was different from those carried by the Border Collie parent, indicating that this locus was not linked to white spotting pattern in this cross (Figure <xref ref-type="fig" rid="F3">3</xref>).</p><p>To evaluate the segregation of the second candidate for white spotting, <italic>KIT</italic>, in this cross, a polymorphic simple-sequence repeat was developed from bacterial artificial chromosome (BAC) clones containing the <italic>KIT</italic> gene. Three of the four BACs containing the <italic>KIT</italic> gene also contained this simple-sequence repeat, as ascertained by PCR amplification of a product from the BACs using primers flanking the simple-sequence repeat. In addition, positive PCR amplification of a product using both <italic>KIT</italic> genomic primers and the simple-sequence repeat primers occurred in the same subset of radiation hybrid cell lines from a dog/hamster panel from Research Genetics, Inc. (Alabama, USA). This simple-sequence repeat marker was polymorphic in the intercross and segregated independent of the white spotting phenotype (Figure <xref ref-type="fig" rid="F4">4</xref>).</p><p>By analogy with mice and humans, mutations in both <italic>EDNRB</italic> and <italic>KIT</italic> were excellent candidates for white spotting in Border Collies; however, these data taken together excluded both genes from being responsible for the distinctive Border Collie coat color markings. Although it has been postulated that all white spotting in dogs is due to the same major locus, allelism tests between the spotting phenotypes in all the different dog breeds have not been performed. It is possible that either <italic>EDNRB</italic> or <italic>KIT</italic> could be responsible for white spotting in other breeds. Extreme white spotting, an additional allele of the <italic>s</italic> locus, segregates in the Boxer breed. The markers developed in this study were not polymorphic in available individuals of this breed, so we were unable to test for linkage to this presumptive allele of the <italic>s</italic> locus. As more genes are placed on the canine genetic map, it will be possible to test other candidates for white spotting derived from studies of coat color genetics in mice. White markings are often associated with deafness in dogs, cats, and even humans. They can also be associated with aganglionic megacolon in humans and horses. The link between white spotting and lack of innervation to the distal colon is the neural crest, from which melanocytes and enteric ganglia are both derived. The exclusion of <italic>EDNRB</italic> as a candidate gene for white spotting in Border Collies may explain why aganglionic megacolon is not associated with this trait in dogs. The gene responsible for white spotting in dogs may shed light on new genes involved in the differentiation and survival of the neural crest.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><p>A female Newfoundland dog was bred to a male Border Collie. All dogs were placed in people's homes as pets. Blood was drawn for DNA extraction and photographs were taken of the dogs. F1 progeny from this cross were intercrossed to produce 25 F2 progeny. Seven of those 25 F2 progeny had white markings similar to the Border Collie parent. To evaluate the white spotting, the percentage of white based on surface area was compared between parents, F1 and F2 progeny. Photographs of the dogs were taken and the weight of the white markings in the image was compared to the total weight of the dog image after cutting, on a percent basis.</p><p>A portion of the <italic>EDNRB</italic> cDNA was cloned as a 1,314 bp product using the tri-clone kit (Invitrogen Corp., California, USA) using primers designed from mouse and human sequences (E1(ATG)F 5'-CAG GTA GCA GCA TGC AGC-3' and E3 5'-GGA ACG GAA GTT GTC ATA TCC-3'). The clone was sequenced (GenBank AF276427) and the deduced amino-acid sequence was compared to that of the published mammalian sequences using the 'pretty' program (GCG Wisconsin Package; Genetics Computer Group, Wisconsin, USA). The 1,314 bp fragment was radioactively labeled and used as a hybridization probe against genomic DNA digested with <italic>Pvu</italic>II. Segregation of <italic>EDNRB</italic> was followed in the intercross progeny by this <italic>Pvu</italic>II RFLP (Figure <xref ref-type="fig" rid="F3">3</xref>).</p><p>To follow segregation of <italic>KIT</italic> in this pedigree, a simple-sequence repeat in close proximity to <italic>KIT</italic> was isolated. A portion of <italic>KIT</italic> was cloned by PCR using published primers [<xref ref-type="bibr" rid="B11">11</xref>] and used to screen a canine BAC library (BAC-PAC Resources, BACPAC Resource Center at the Children's Hospital Oakland Research Institute, California, USA). A GAAA (18) repeat microsatellite marker was developed from one of the BACs (KIT forward 5'-GCA TGG AGC CTG CTT CTC-3', KIT reverse 5'-AGA GCA TCC TTG GTC TGT CC-3'). PCR amplification of the simple-sequence repeat from the intercross animals is shown in Figure <xref ref-type="fig" rid="F4">4</xref>. PCR amplification products of the <italic>KIT</italic> microsatellite were separated by electrophoresis on a 10% polyacrylamide gel and stained with ethidium bromide before photography.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>A Newfoundland female <bold>(a)</bold> was bred to a Border Collie male <bold>(b)</bold> to produce animals for the intercross.</p></caption><graphic xlink:href="gb-2000-1-2-research0004-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>Amino-acid sequence alignment of mammalian endothelin B receptors using the 'pretty' function from the GCG Wisconsin package. Accession numbers for sequences used for comparison: bovine EDNRB, S63513; human EDNRB, JQ1042; rat EDNRB, I57950; canine EDNRB, AF276427.</p></caption><graphic xlink:href="gb-2000-1-2-research0004-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p>Segregation of <italic>EDNRB</italic> compared with the segregation of the white spotting phenotype. Genomic DNA digested with <italic>Pvu</italic>II was probed with a radioactively labeled fragment of <italic>EDNRB</italic> DNA which detected a <italic>Pvu</italic>II restriction fragment length polymorphism (RFLP) between the Border Collie and Newfoundland grandparents. BC, DNA from the Border Collie grandparent; N, DNA from the Newfoundland grandparent. F1, DNA from the first generation hybrids of the Newfoundland &#x000d7; Border Collie cross; F2, DNA from the progeny of intercrosses between the F1 hybrids; s, white spotted; ns, not white spotted. The presumed genotypes for the <italic>EDNRB</italic> locus are shown below the lanes.</p></caption><graphic xlink:href="gb-2000-1-2-research0004-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>Segregation of <italic>KIT</italic> compared with segregation of the white spotting phenotype. Segregation of <italic>KIT</italic> was followed by detection of a simple-sequence repeat polymorphic difference between the Newfoundland and Border Collie grandparents. BC, DNA from the Border Collie grandparent; N, DNA from the Newfoundland grandparent; F1, DNA from the first generation hybrids; F2, DNA from the second generation intercrosses. s, white spotted; ns, not white spotted. The presumed genotypes for the <italic>KIT</italic> locus are shown below the lanes.</p></caption><graphic xlink:href="gb-2000-1-2-research0004-4"/></fig></sec>

Gene expression changes during murine postnatal brain development

<sec><title>Background:</title><p>For most vertebrate organs and tissues, the majority of development occurs during embryogenesis, and postnatal changes are primarily concerned with growth. The central nervous system is unusual in that a considerable amount of morphological development, cell differentiation and acquisition of function, takes place during postnatal development. As yet, the molecular mechanisms underlying these complex developmental processes are not well understood. In order to identify markers for these developmental processes, we have analyzed the expression profiles, during postnatal murine brain development, of approximately 25,000 transcripts. This analysis, performed at day 1, day 10, day 20 and day 42 of postnatal development, identified a large number of developmentally regulated genes which we have assigned into three broad expression categories.</p></sec><sec><title>Results:</title><p>Expression levels at four timepoints during postnatal murine brain development were established for approximately 25,000 gene transcripts. Approximately 1% of the genes examined displayed a developmentally regulated pattern of expression and we provide all the necessary information required to easily obtain molecular markers for a subset of these developmentally regulated transcripts. Of this subset, 61 showed increasing expression during development, 61 showed decreasing expression during development, and 9 exhibited a peak of expression during this period.</p></sec><sec><title>Conclusions:</title><p>A small percentage of the genes expressed in the postnatal developing brain show changes in expression level during the newborn to adult phase of development. It is likely that these developmentally regulated transcripts represent molecular markers for the complex developmental process occurring in the postnatal brain.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Clinton</surname><given-names>M</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Manson</surname><given-names>J</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>McBride</surname><given-names>D</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Miele</surname><given-names>G</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>For most vertebrate organs and tissues, the majority of development occurs during embryogenesis, and postnatal changes are primarily concerned with growth. The central nervous system (CNS) is unusual in that a considerable amount of morphological development, cell differentiation and acquisition of function takes place during postnatal development [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. As yet, the molecular mechanisms underlying these complex developmental processes are not well understood.</p><p>We have recently completed a differential display (DDRT-PCR) [<xref ref-type="bibr" rid="B3">3</xref>] analysis of genes expressed in the murine postnatal developing brain. This analysis formed the baseline for an ongoing study and produced a series of RNA fingerprints representing genes transcribed at specific stages of brain development. For our analysis we compared RNA fingerprints at four different timepoints during postnatal development: newborn (day 1), day 10, day 20 and adult (day 42). Incidental to our experimental objective, we also identified a large number of genes which were clearly developmentally regulated in the wild-type mouse brain. We submit this data principally to allow others to isolate specific transcripts with developmentally regulated expression in the postnatal brain without the necessity of performing extensive screening procedures. From a total of approximately 25,000 transcripts displayed, around 200-300 (approximately 1%) exhibited developmentally regulated expression profiles. Here we present groups of RNA fingerprints displaying a subset of these developmentally regulated genes, along with sufficient technical information to replicate specific fingerprints and recover selected cDNAs.</p><p>The particular fingerprints presented here were selected on the basis that identical DDRT-PCR profiles were obtained with at least two separate batches of RNA, and that each fingerprint displayed at least three developmentally regulated transcripts. On each group of fingerprints we have indicated the position of at least three bands representing the more obviously developmentally regulated genes. We have assigned these marked transcripts into three broad categories as follows: genes for which mRNA levels increase during brain development; genes for which mRNA levels decrease during development; and genes exhibiting a peak in mRNA levels during this developmental period.</p><p>To ensure that changes in DDRT-PCR profiles represent genuine changes in expression levels, two cDNA fragments were recovered and used in downstream expression analyses. The overall procedure followed is illustrated in Figure <xref ref-type="fig" rid="F1">1</xref> (and see [<xref ref-type="bibr" rid="B4">4</xref>]). The northern blot expression profiles of both transcripts accurately replicated the original DDRT-PCR expression profiles, confirming the validity of our approach.</p></sec><sec><title>Results</title><p>In this study, we have observed a large number of transcripts which apparently increase and/or decrease in expression level during postnatal murine brain development. The period of development under study spanned from day 1 to day 42 post-partum. Figure <xref ref-type="fig" rid="F2">2</xref> shows a selection of RNA fingerprints containing bands representing such developmentally regulated transcripts. Thirty-four groups of RNA fingerprints were selected on the basis that identical DDRT-PCR profiles were obtained with at least two separate batches of RNA, and that each fingerprint displayed at least three developmentally regulated transcripts. The gel positions of a subset of developmentally regulated transcripts are indicated in Figure <xref ref-type="fig" rid="F2">2</xref> and individual transcripts are assigned into one of three broad expression categories: A (+), genes for which mRNA levels increase during brain development; B (-), genes for which mRNA levels decrease during development; and C (^), genes exhibiting a peak in mRNA levels during this developmental period. Of a total of 131 transcripts indicated, 61 are assigned to category A, 61 to category B, and 9 to category C. The sequence details of the primers used to produce the RNA fingerprints shown in Figure <xref ref-type="fig" rid="F2">2</xref> are contained in Table <xref ref-type="table" rid="T1">1</xref>. A detailed description of the various procedures employed and the animals used in this study are presented in the Materials and methods section.</p><p>To ensure that developmentally dependent changes in intensity of individual DDRT-PCR bands reflected changes in levels of individual mRNAs, two apparently up-regulated transcripts were chosen for further study. The cDNA fragments represented by these bands were recovered, purified from co-migrating DNA species by mSSCP [<xref ref-type="bibr" rid="B4">4</xref>], cloned and used in northern hybridization studies (Figure <xref ref-type="fig" rid="F3">3</xref>). For northern analyses, whole brain RNA was isolated from a separate batch of animals to those used in the original DDRT-PCR. As can be seen from Figure <xref ref-type="fig" rid="F3">3</xref>, for both transcripts northern analyses produced expression profiles similar to those observed by DDRT-PCR, thereby validating this approach to identifying developmentally regulated transcripts.</p><p>Sequence analysis revealed that these cDNA fragments represented portions of the prion Protein (PrP) and of the Thy-1.2 gene transcripts [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. The cloned DNA used as a probe in Figure <xref ref-type="fig" rid="F3">3a</xref> is identical to nucleotides 588-1,414 of the PrP transcript while that used in Figure <xref ref-type="fig" rid="F3">3b</xref> is identical to nucleotides 1,599-1,700 of the Thy-1.2 transcript (Figure <xref ref-type="fig" rid="F4">4</xref>). This shows that the original DDRT-PCR reactions have amplified an internal fragment of the PrP transcript and the terminal 3' untranslated region (UTR) of the Thy-1.2 transcript. The PrP fragment was amplified by a combination of arbitrary primer R4 and anchored primer T₁₂CC while the Thy-1.2 fragment was amplified by a combination of arbitrary primer P3 and the anchored primer T₁₂CA. For the PrP fragment, the arbitrary primer showed a 100% match while the anchored primer showed 86% homology (12/14) with the published sequence. In contrast, for the Thy 1.2 fragment, while the anchored primer matched with 100% homology, the arbitrary primer showed only 70% homology (7/10) with the corresponding region of the published sequence.</p></sec><sec><title>Conclusions</title><p>In the course of analyzing gene expression in the postnatal developing murine brain we have observed that approximately 1% of genes transcribed are developmentally regulated. The objective of this report is to permit those interested in such transcripts to selectively isolate specific developmentally regulated transcripts without the necessity of performing an extensive screen. Here we present the RNA fingerprints containing a subset of these developmentally regulated transcripts and we include all the necessary information to permit individual transcripts to be isolated and identified. We have highlighted a total of 131 developmentally regulated transcripts in three broad categories of expression profiles. Of these, approximately 7% (9) fell into category C while the remainder were equally divided between categories A (61) and B (61). We also present confirmatory evidence that changing DDRT-PCR expression profiles represent genuine alterations in expression levels during brain development. It should be noted that changes in expression profile reflect changes in RNA level per microgram total RNA in whole brain and, given that the postnatal brain is not a uniform structure, expression profiles should be interpreted accordingly.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Collection of tissues</title><p>Male animals from an inbred line of 129/Ola mice were used. Animals were humanely sacrificed and whole brains from six individual animals at P1, P10, P20 and P42 stages of postnatal development were collected, flash frozen in liquid nitrogen and stored at -80&#x000b0;C.</p></sec><sec><title>RNA extraction</title><p>Total RNA was extracted using the RNAzol^&#x02122; B method (AMS Biotechnology, Oxon, UK), based on the guanidinium thiocyanate/phenol/chloroform extraction method of Chomczynski and Sacchi [<xref ref-type="bibr" rid="B7">7</xref>]. Following the initial precipitation step, the RNA pellet was washed once with 85% ethanol, resuspended in water, and 0.1 volumes of 3M sodium acetate (pH 5.2) and 2.5 volumes ethanol were added, the tubes mixed briefly and RNA stored as ethanol precipitates at -80&#x000b0;C. RNA from six individuals at each timepoint was pooled for DDRT-PCR analysis. For quantitation of total RNA, a small aliquot was removed, pelleted by centrifugation, resuspended in water and the concentration estimated by spectrophotometry. Stock RNA solutions were adjusted to a concentration of approximately 1 &#x003bc;g/&#x003bc;l in ethanol suspension. The use of RNAzol B and homogenization of tissue with a Polytron homogenizer resulted in the isolation of total RNA with OD_260/280 ratios consistently greater than 1.95.</p></sec><sec><title>Synthesis of first-strand cDNA</title><p>For each sample, a volume of RNA stock solution containing approximately 8 &#x003bc;g RNA was transferred to a fresh tube. Samples were centrifuged at 13,000 rpm for 25 minutes and the resulting RNA pellet was washed with 85% ethanol, dried at 45&#x000b0;C for 2 minutes and resuspended in 7 &#x003bc;l RNAse-free water. A small portion (2 &#x003bc;l) of each sample was re-quantitated by spectrophotometery and 5 &#x003bc;g total RNA (in 5 &#x003bc;l volume) was used to synthesize first-strand cDNA (First-strand cDNA synthesis kit; Amersham Pharmacia Biotech, Hertfordshire, UK). Reactions contained 1 &#x003bc;l DTT (200 mM), 5 &#x003bc;l of bulk first-strand mix (containing Moloney murine leukemia virus reverse transcriptase) and 4 &#x003bc;l of either d(T)₁₂MA, d(T)₁₂MG, d(T)₁₂MC, d(T)₁₂MT primer (24 &#x003bc;M; M = A, G or C). Reactions were mixed and incubated at 37&#x000b0;C for 1 hour and then heated to 95&#x000b0;C for 10 minutes to inactivate reverse-transcriptase. The synthesized cDNA was then dispensed into 1 &#x003bc;l aliquots and stored at -20&#x000b0;C.</p></sec><sec><title>Differential display PCR</title><p>For differential display PCR reactions, 1 &#x003bc;l aliquoted cDNA was diluted to 133 &#x003bc;l with water and 10 &#x003bc;l of this solution was used for each display PCR (equivalent to the amount of cDNA produced from 25 ng RNA). To each 10 &#x003bc;l cDNA on ice, 2 &#x003bc;l arbitrary 10-mer primer (5 &#x003bc;M) was added and the solution overlaid with 30 &#x003bc;l mineral oil (Sigma, Dorset, UK). Master mix (8 &#x003bc;l) containing 2 &#x003bc;l 10X PCR buffer (Boehringer Mannheim, East Sussex, UK), 2 &#x003bc;l dNTPs (20 &#x003bc;M; Amersham Pharmacia Biotech, Hertfordshire, UK), 0.3 &#x003bc;l Taq polymerase (1.5 U, Boehringer Mannheim, East Sussex, UK), 2 &#x003bc;l (dT)₁₂MN (25 &#x003bc;M), 1 &#x003bc;l [&#x003b1;-³⁵S]dATP (1000 Ci/mmol) and 0.7 &#x003bc;l dH₂O was added to each tube. Tubes were centrifuged briefly and incubated in a Biometra Unoblock PCR machine at 94&#x000b0;C (2 minutes), followed by 40 cycles of 94&#x000b0;C denaturation (30 seconds), 40&#x000b0;C annealing (2 minutes) and 72&#x000b0;C extension (30 seconds), followed by a final extension step at 72&#x000b0;C for 5 minutes. Type III [<xref ref-type="bibr" rid="B8">8</xref>] loading dye (4 &#x003bc;l) was added to each tube and 8 &#x003bc;l of each sample loaded onto a 6% non-denaturing HR-1000 GenomyxLR polyacrylamide gel (Beckman Instruments Ltd., Buckinghamshire, UK). Samples were run for 2 hours 15 minutes at 2,700V (50&#x000b0;C) on a GenomyxLR DNA analyzer (Beckman Instruments Ltd., Bucks, UK). The gels were transferred to 3 MM blotting paper, dried and exposed to BiomaxMR autoradiography film for 16 hours.</p></sec><sec><title>Recovery of cDNA from dried polyacrylamide gels</title><p>Gel regions corresponding to bands representing candidate cDNAs were excised from all four lanes using sterile scalpels and transferred to sterile 0.5 ml Eppendorf tubes. Glogos^&#x02122; autoradiograph markers (Stratagene, Cambridgeshire, UK) were used to align the gel with the autoradiograph and identical regions were excised from all lanes. The gel fragments were rehydrated at room temperature for 15 minutes in 100 &#x003bc;l water, and cDNA eluted at 99&#x000b0;C for 15 minutes before transfer of the liquid phase to fresh 0.5 ml tubes. cDNA was precipitated by the addition of 2.5 volumes ethanol, 0.1 volumes of 3 M sodium acetate (pH 5.2) and 1 &#x003bc;l See-DNA (Amersham Pharmacia Biotech, Buckinghamshire, UK), and stored on dry ice for 1 hour. Following centrifugation (13,000 rpm, 25 minutes, 4&#x000b0;C) the pellet was washed in 85% ethanol and resuspended in 4 &#x003bc;l of water.</p></sec><sec><title>Modified single-strand conformation polymorphism (mSSCP)</title><p>For the mSSCP-PCR reaction, 4 &#x003bc;l 10X PCR buffer (Boehringer Mannheim, East Sussex, UK), 3.2 &#x003bc;l dNTPs (2.5 mM dGTP, dCTP, dTTP; 0.025 mM dATP; Amersham Pharmacia Biotech, Buckinghamshire, UK), 2.5 &#x003bc;l anchored primer (20 &#x003bc;M), 2.5 &#x003bc;l arbitrary primer (20 &#x003bc;M), 0.3 &#x003bc;l Taq polymerase (1.5 U; Boehringer Mannheim, East Sussex, UK) and 0.5 &#x003bc;l [&#x003b1;-³³P] dATP (3,000 Ci/mmol) was added to the DNA and the reaction volume adjusted to 40 &#x003bc;l with water. PCR conditions were similar to those used for display-PCR, with the exception that only five cycles were performed. After removal of mineral oil, PCR products were purified by phenol/chloroform/isoamyl alcohol extraction and precipitated for 1 hour on dry ice. Pellets were washed, resuspended in 8 &#x003bc;l of mSSCP loading buffer (80% deionized formamide, 0.01% bromophenol blue, 0.01% xylene cyanol, 1 mM EDTA, 10 mM NaOH) and denatured at 95&#x000b0;C for 10 minutes before loading onto a 0.5X MDE gel (Flowgen, Staffordshire, UK). Samples were electrophoresed typically for 18 hours at 8W (25&#x000b0;C) in 0.6X TBE buffer using the Genomyx LR system. Following autoradiography, areas of the gel corresponding to candidate cDNAs were excised and the cDNA eluted and precipitated as described above.</p></sec><sec><title>PCR re-amplification of mSSCP-purified candidate cDNAs</title><p>A final PCR reamplification of the recovered cDNA was performed using modified versions of the arbitrary and anchored primers used in the DDRT-PCR and mSSCP steps. For both the anchored and arbitrary primers, oligonucleotides were synthesized with an additional 5' sequence containing an <italic>Eco</italic>RI restriction site to aid cloning (referred to as extended primers [<xref ref-type="bibr" rid="B9">9</xref>]; see Table <xref ref-type="table" rid="T1">1</xref>). Reactions contained 4 &#x003bc;l mSSCP purified cDNA, 4 &#x003bc;l 10X buffer, 3.2 &#x003bc;l dNTPs (10 mM each dATP, dCTP, dTTP, dGTP), 2.5 &#x003bc;l extended anchored primer (20 &#x003bc;M), 2.5 &#x003bc;l extended arbitrary primer (20 &#x003bc;M) and 0.5 &#x003bc;l Taq polymerase (Boehringer Mannheim, East Sussex, UK) in a 40 &#x003bc;l volume.</p><p>Samples were denatured at 94&#x000b0;C for 2 minutes and a single round of PCR performed at 94&#x000b0;C for 30 seconds, 40&#x000b0;C for 2 minutes and 72&#x000b0;C for 30 seconds. A further five cycles of PCR were then performed at 94&#x000b0;C for 30 seconds, 58&#x000b0;C for 1 minute and 72&#x000b0;C for 30 seconds, followed by a final extension step of 72&#x000b0;C for 5 minutes.</p><p>PCR products were phenol/chloroform extracted, precipitated and then subjected to <italic>Eco</italic>RI restriction endonuclease digestion, prior to further purification and cloning into the <italic>Eco</italic>RI site of pBluescript SKII⁺ or KSII⁺ (Stratagene, Cambridgeshire, UK).</p><p>Cloned cDNAs were sequenced and the sequences compared to non-redundant GENBANK/EMBL protein and nucleotide databases, using either the GCG [<xref ref-type="bibr" rid="B10">10</xref>] or NCBI worldwide web [<xref ref-type="bibr" rid="B11">11</xref>] implementation of the BLAST algorithm [<xref ref-type="bibr" rid="B12">12</xref>]. Under standard procedures, sequences would also be compared with the GENBANK/EMBL expressed sequence tags (EST) databases and matching I.M.A.G.E. ESTs obtained from the Human Genome Mapping (HGMP) Resource Centre, Cambridge [<xref ref-type="bibr" rid="B13">13</xref>]. The I.M.A.G.E. clones are then fully sequenced in both directions and the resulting sequence re-checked against the databases to confirm their identity. The expression profiles of transcripts represented by isolated clones were confirmed by standard northern procedures [<xref ref-type="bibr" rid="B13">13</xref>].</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Differential display procedures; mSSCP: modified single-stranded conformational polymorphism [<xref ref-type="bibr" rid="B4">4</xref>].</p></caption><graphic xlink:href="gb-2000-1-3-research0005-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>Gene expression profiles in the postnatal developing mouse brain. RNA fingerprints produced by specific DDRT-PCR primer combinations at four stages of brain development run in parallel. Anchored primer/arbitrary primer combinations are shown at the top for each group of RNA fingerprints and stages of development are indicated in days (1, 10, 20 and 42). Gel positions of developmentally regulated transcripts in expression category A (+), expression category B (-) and expression category C (^) are highlighted.</p></caption><graphic xlink:href="gb-2000-1-3-research0005-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p>Verification of expression profiles. Downstream analysis of two developmentally regulated transcripts - <bold>(a)</bold> and <bold>(b)</bold> are the two transcripts. Candidate cDNAs were selected on the basis of the original DDRT-PCR profile (I), purified from co-migrating sequences by mSSCP (II) and used as probes in northern analysis (III). To correct for variation in RNA loading, northern blot membranes were re-probed for expression of the mouse beta actin gene (IV).</p></caption><graphic xlink:href="gb-2000-1-3-research0005-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>Origin of DDRT-PCR fragments. Schematic illustrating the position of DDRT-PCR cDNA fragments in relation to full-length transcripts of <bold>(a)</bold> PrP and <bold>(b)</bold> Thy-1.2. Black boxes represent transcripts, with open reading frames in white. Areas of transcripts amplified by DDRT-PCR are shown as gray boxes. Sequences of regions matching arbitrary and anchored primers are detailed with mismatched pairs in bold.</p></caption><graphic xlink:href="gb-2000-1-3-research0005-4"/></fig><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Oligonucleotide sequences of primers used</p></caption><table frame="hsides" rules="groups"><tbody><tr><td align="left">Anchored primer</td></tr><tr><td align="left">&#x02003;dT₁₂AA 5'-d(TTTTTTTTTTTTAA)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂GA 5'-d(TTTTTTTTTTTTGA)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂CA 5'-d(TTTTTTTTTTTTCA)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂AG 5'-d(TTTTTTTTTTTTAG)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂GG 5'-d(TTTTTTTTTTTTGG)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂CG 5'-d(TTTTTTTTTTTTCG)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂AC 5'-d(TTTTTTTTTTTTAC)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂GC 5'-d(TTTTTTTTTTTTGC)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂CC 5'-d(TTTTTTTTTTTTCC)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂AT 5'-d(TTTTTTTTTTTTAT)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂GT 5'-d(TTTTTTTTTTTTGT)-3'</td></tr><tr><td align="left">&#x02003;dT₁₂CT 5'-d(TTTTTTTTTTTTCT)-3'</td></tr><tr><td align="left">Random primer</td></tr><tr><td align="left">(arbitrary but defined sequence)</td></tr><tr><td align="left">&#x02003;P3 5'-d(GCCGTTCCAT)-3'</td></tr><tr><td align="left">&#x02003;R1 5'-d(GGAACTCCGT)-3'</td></tr><tr><td align="left">&#x02003;R2 5'-d(GGCAAGTCAC)-3'</td></tr><tr><td align="left">&#x02003;R4 5'-d(AGGACCGCTA)-3'</td></tr><tr><td align="left">&#x02003;R5 5'-d(CGGACCCCGG)-3'</td></tr><tr><td align="left">&#x02003;R7 5'-d(TACAACGAGG)-3'</td></tr><tr><td align="left">&#x02003;R8 5'-d(TGGATTGGTC)-3'</td></tr><tr><td align="left">&#x02003;R9 5'-d(TGGTAAAGGG)-3'</td></tr><tr><td align="left">&#x02003;R10 5'-d(TCGGTCATAG)-3'</td></tr><tr><td align="left">&#x02003;R11 5'-d(TACCTAAGCG)-3'</td></tr><tr><td align="left">&#x02003;R12 5'-d(CTGCTTGATG)-3'</td></tr><tr><td align="left">&#x02003;R14 5'-d(GATCGCATTG)-3'</td></tr><tr><td align="left">&#x02003;R18 5'-d(GGAACCAATC)-3'</td></tr><tr><td align="left">&#x02003;M2 5'-d(CACAGTTTGC)-3'</td></tr><tr><td align="left">&#x02003;M3 5'-d(CCACAGAGTA)-3'</td></tr><tr><td align="left">&#x02003;S 5'-d(GCGACCCATG)-3'</td></tr><tr><td align="left">&#x02003;A1 5'-d(ACAGAGCACA)-3'</td></tr><tr><td align="left">&#x02003;A2 5'-d(ACGTATCCAG)-3'</td></tr><tr><td align="left">Oligo(dT) primer mix</td></tr><tr><td align="left">(poly(A) anchored)</td></tr><tr><td align="left">&#x02003;MA Equal mix of dT₁₂AA, dT₁₂GA and dT₁₂CA</td></tr><tr><td align="left">&#x02003;MG Equal mix of dT₁₂AG, dT₁₂GG and dT₁₂CG</td></tr><tr><td align="left">&#x02003;MC Equal mix of dT₁₂AC, dT₁₂GC and dT₁₂CC</td></tr><tr><td align="left">&#x02003;MT Equal mix of dT₁₂AT, dT₁₂GT and dT₁₂CT</td></tr><tr><td align="left">Extended random primer</td></tr><tr><td align="left">&#x02003;5'-d(GTCAGAATTC-random primer)-3'</td></tr><tr><td align="left">Extended oligo(dT) primer</td></tr><tr><td align="left">&#x02003;5'-d(GTCAGAATTCT12)-3'</td></tr></tbody></table></table-wrap></sec></sec>

Genomic structure of the gene for mouse germ cell nuclear factor (GCNF)

<sec><title>Background:</title><p>The germ cell nuclear factor (GCNF, also known as retinoid acid receptor-related testis-associated receptor, neuronal cell nuclear receptor or NR6A1) is an orphan receptor in the nuclear receptor superfamily found in mammals, amphibians and fish. The mouse <italic>Gcnf</italic> gene is expressed in the placenta and the developing nervous system and germ cells, and responds to retinoic acid.</p></sec><sec><title>Results:</title><p>We have defined the intron-exon structure of the mouse <italic>Gcnf</italic> gene and found that it contains 11 exons. Exons 1-4 encode the 75 amino acid amino-terminal domain and exon 4 also encodes the core DNA-binding domain. The carboxy-terminal extension is encoded by exon 5, exons 6 and 7 encode the hinge region, and exons 7-11 encode the putative ligand-binding domain. Unusually, the two zinc-finger motifs in the DNA-binding domain are encoded by separate exons.</p></sec><sec><title>Conclusions:</title><p>The protein-coding region of GCNF is contained in 11 exons. The genomic structure of this nuclear receptor gene will be useful for further studies.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>S&#x000fc;sens</surname><given-names>Ute</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Borgmeyer</surname><given-names>Uwe</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>The germ cell nuclear factor (GCNF, NR6A1) is a member of the nuclear receptor superfamily [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Originally isolated from mouse cDNA libraries, homologs of GCNF have been identified in humans, frogs and fish [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. As no ligand has been identified, GCNF is designated an orphan receptor. Also known as RTR (retinoid acid receptor-related testis-associated receptor) or NCNF (neuronal cell nuclear receptor), evolutionary studies have defined GCNF as the only known member of a sixth subfamily of nuclear receptors [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. The mouse <italic>Gcnf</italic> gene is highly expressed in the developing nervous system, in the labyrinthine layer of the placenta and in the developing germ cells [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. Two transcripts of approximately 7.5 kb and 2.4 kb are present in testis, but only the larger transcript is found in somatic cells. Hybridization experiments reveal that the size difference is at least partially due to the use of different polyadenylation sites [<xref ref-type="bibr" rid="B13">13</xref>]. Interestingly, GCNF expression is transiently up-regulated and later down-regulated again when embryonal carcinoma cells are triggered to differentiate by retinoic acid [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. </p></sec><sec><title>Results and discussion</title><p>We have isolated genomic clones encompassing the mouse <italic>Gcnf</italic> gene, and have defined the intron-exon structure of the gene. Sequence analysis reveals that the coding region of <italic>Gcnf</italic> comprises 11 exons and 10 introns (Table <xref ref-type="table" rid="T1">1</xref>). A bacteriophage lambda library and a cosmid library of genomic DNA of the mouse 129 strain were screened with the full-length <italic>Gcnf</italic> cDNA. The DNA from colonies that hybridized was cloned into pBluescript (SK) for further sequence analysis. Exons 3 and 4 were identified from bacteriophage subclones, and exons 6-11 were identified in cosmid-derived subclones. Additional intron-exon boundaries and the 5'-untranslated region (5'-UTR) were identified by genome walking analysis following the manufacturer s instructions (Clontech). DNA sequencing was performed on an ABI 377-sequencer using the dye terminator protocol (Perkin Elmer) and on a DNA sequencer model 400 (Li-Cor). The DNA sequences were processed using the Wisconsin Package Version 10.0 of the Genetics Computer Group (GCG), Madison, Wisconsin. </p><p>All intron-exon junctions obeyed the GT/AG rule ([<xref ref-type="bibr" rid="B17">17</xref>] and Table <xref ref-type="table" rid="T1">1</xref>). The location of the intron-exon junctions relative to the peptide sequence is shown in Figure <xref ref-type="fig" rid="F1">1</xref>. The translational start and stop codons are on exons 1 and 11, respectively. Exon 1 contains the 244 bp untranslated sequence at the 5' end of the cDNA and codes for the first 33 amino acids (Figure <xref ref-type="fig" rid="F2">2</xref>). This cDNA, isolated by Hirose <italic>et al</italic>. ([<xref ref-type="bibr" rid="B7">7</xref>]; GenBank entry MMU09563), starts with an <italic>Eco</italic>RI site that is present in the genomic DNA. The T at position 174 is a G in our genomic isolate, which could represent a genomic variant. As no promoter has been identified for <italic>Gcnf</italic>, the sequence preceding the <italic>Eco</italic>RI site may contain promoter elements. It is also possible, however, that the promoter precedes a not-yet-identified additional exon in the 5'-UTR of <italic>Gcnf</italic>. </p><p>The amino-terminal domain of 75 amino acids is encoded by exons 1-4. Exon 4 also codes for the core DNA-binding domain (DBD) of 66 amino acids and for three additional amino acids (Figure <xref ref-type="fig" rid="F1">1</xref>). The DBD consists of two zinc-finger motifs that are encoded by separate exons in most vertebrate nuclear receptor genes, except for those of the COUP transcription factor subfamily. Evolutionary studies do not provide further evidence that these receptors are closely related to GCNF. A further domain important for DNA binding and for homodimeric interactions, and known as the DBD carboxy-terminal extension, is encoded by the 56 bp of exon 5. The sizes of intron 2 and intron 4 were determined by PCR amplification of mouse genomic DNA. Exons 6 and 7 code for the hinge region, whereas exons 7-11 code for the putative ligand-binding domain. A variant of the typical AUAAA polyadenylation signal (AGUAAA) and the cleavage site that is used in the testis are part of the eleventh exon [<xref ref-type="bibr" rid="B13">13</xref>].</p></sec><sec><title>Conclusions</title><p>The protein-coding region of GCNF is contained in 11 exons. Additional studies will be required to define the regulatory/promoter region. We think the genomic structure of this first, and at present only, member of the sixth subfamily of nuclear receptors will be useful for further studies of this unique receptor.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>The location of the different exons in the GCNF amino-acid sequence. The core DNA-binding domain is underlined.</p></caption><graphic xlink:href="gb-2000-1-3-research0006-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>Sequence of exon 1 of <italic>Gcnf</italic>. The location of the <italic>Eco</italic>RI site (GAATTC) marking the 5'-end of the <italic>Gcnf</italic> cDNA (GenBank entry MMU09563) and the putative translational start codon (ATG) are underlined.</p></caption><graphic xlink:href="gb-2000-1-3-research0006-2"/></fig><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Organization of the mouse <italic>Gcnf</italic> gene</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Exon</td><td align="center">Exon</td><td align="center">cDNA</td><td align="center">5' splice donor</td><td align="center">Intron</td><td align="center">Intron</td><td align="center">3' splice acceptor</td></tr><tr><td align="left">number</td><td align="center">size (bp)</td><td align="center">position^*</td><td></td><td align="center">number</td><td align="center">size (kbp)</td><td></td></tr></thead><tbody><tr><td align="left">1</td><td align="center">&#x0003e;344</td><td align="center">1-344</td><td align="center">CCGCGCAACGgtgggta</td><td align="center">1</td><td align="center">ND</td><td align="center">ctattgttctctctttagGTTTCT</td></tr><tr><td align="left">2</td><td align="center">42</td><td align="center">345-386</td><td align="center">CCAGGCACTAgtaagttc</td><td align="center">2</td><td align="center">&#x0003e;12</td><td align="center">gttctttttgctttgcagATGGAG</td></tr><tr><td align="left">3</td><td align="center">45</td><td align="center">387-431</td><td align="center">CATATACCTGgtaagtgg</td><td align="center">3</td><td align="center">ND</td><td align="center">tgacttatccatgtttagTTTCCG</td></tr><tr><td align="left">4</td><td align="center">243</td><td align="center">432-674</td><td align="center">AACAGGAAGGgtgagttg</td><td align="center">4</td><td align="center">&#x0003e;12</td><td align="center">gtctacatttccttctagCTATCA</td></tr><tr><td align="left">5</td><td align="center">56</td><td align="center">675-730</td><td align="center">ACCAGTCCAGgtgagtcc</td><td align="center">5</td><td align="center">ND</td><td align="center">atccatttcttgccaaagATATCA</td></tr><tr><td align="left">6</td><td align="center">155</td><td align="center">731-885</td><td align="center">TATCATCCAGgtgagcta</td><td align="center">6</td><td align="center">ND</td><td align="center">tgaagtttttctctccagTAGGTC</td></tr><tr><td align="left">7</td><td align="center">228</td><td align="center">886-1113</td><td align="center">TTGAAGATGGgtgagtta</td><td align="center">7</td><td align="center">1.238</td><td align="center">tcctgtccctgcccccagGTATGC</td></tr><tr><td align="left">8</td><td align="center">255</td><td align="center">1114-1368</td><td align="center">AACTCCACAGgtgagagc</td><td align="center">8</td><td align="center">ND</td><td align="center">cctgtatctgttctccagATTTAG</td></tr><tr><td align="left">9</td><td align="center">122</td><td align="center">1369-1490</td><td align="center">CTGAATCAAGgtgagtag</td><td align="center">9</td><td align="center">1.408</td><td align="center">ttttgtttttgttttcagATATCA</td></tr><tr><td align="left">10</td><td align="center">153</td><td align="center">1491-1643</td><td align="center">TACATCGCAGgtaatatt</td><td align="center">10</td><td align="center">1.567</td><td align="center">tctcttccctttacctagGCAAGA</td></tr><tr><td align="left">11</td><td align="center">&#x0003e;869</td><td align="center">1644</td><td></td><td></td><td></td><td></td></tr></tbody></table><table-wrap-foot><p>Lower-case letters are used for the intron sequence and capital letters for the exon sequence. The GenBank accession numbers for the exons and the flanking sequences are AF254575S1-AF254575S8. ^*Relative to GenBank entry MMU09563.</p></table-wrap-foot></table-wrap></sec>

The alpha/beta fold uracil DNA glycosylases: a common origin with diverse fates

<sec><title>Background:</title><p>Uracil DNA glycosylases (UDGs) are major repair enzymes that protect DNA from mutational damage caused by uracil incorporated as a result of a polymerase error or deamination of cytosine. Four distinct families of UDGs have been identified, which show very limited sequence similarity to each other, although two of them have been shown to possess the same structural fold. The structural and evolutionary relationships between the rest of the UDGs remain uncertain.</p></sec><sec><title>Results:</title><p>Using sequence profile searches, multiple alignment analysis and protein structure comparisons, we show here that all known UDGs possess the same fold and must have evolved from a common ancestor. Although all UDGs catalyze essentially the same reaction, significant changes in the configuration of the catalytic residues were detected within their common fold, which probably results in differences in the biochemistry of these enzymes. The extreme sequence divergence of the UDGs, which is unusual for enzymes with the same principal activity, is probably due to the major role of the uracil-flipping caused by the conformational strain enacted by the enzyme on uracil-containing DNA, as compared with the catalytic action of individual polar residues. We predict two previously undetected families of UDGs and delineate a hypothetical scenario for their evolution.</p></sec><sec><title>Conclusions:</title><p>UDGs form a single protein superfamily with a distinct structural fold and a common evolutionary origin. Differences in the catalytic mechanism of the different families combined with the construction of the catalytic pocket have, however, resulted in extreme sequence divergence of these enzymes.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Aravind</surname><given-names>L</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Koonin</surname><given-names>Eugene V</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>Mutagenic uracil appears in DNA opposite to guanine as a result of misincorporation or of deamination of cytosine. Similarly, the deamination process generates thymine opposite guanine in those organisms that undergo cytosine methylation [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. DNA is safeguarded from the consequences of these events by the activity of uracil DNA glycosylases (UDGs), which remove uracil (and sometimes thymine) from the sugar backbone of DNA without breaking the phosphodiester bonds in the backbone. There are different types of these enzymes in the three superkingdoms of life. The best studied family of UDGs, typified by the <italic>Escherichia coli</italic> Ung protein, is largely specific for uracil and is present in a variety of bacteria, eukaryotes and large eukaryotic DNA viruses [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. The mismatch-specific uracil DNA glycosylases (MUGs) have been identified in eukaryotes and several bacteria and, unlike the Ung-family enzymes, are additionally active on G:T mismatches [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. Comparison of the crystal structures of these two enzymes has shown that they are structurally very similar, despite the low sequence similarity [<xref ref-type="bibr" rid="B5">5</xref>]. Subsequently, two other classes of UDGs have been characterized, one from thermophilic archaea and several bacteria [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>] (hereinafter called AUDG) and the other from vertebrates (SMUG) [<xref ref-type="bibr" rid="B9">9</xref>]. The latter enzyme has a high specificity for uracil and for single-stranded substrates. The single-strand-specific UDGs (ssUDGs) are believed to be functionally similar to the UNGs and MUGs because they possess motifs similar to the catalytic motifs of the latter enzymes despite supposedly lacking significant sequence similarity to them [<xref ref-type="bibr" rid="B9">9</xref>]. In contrast, the structural and evolutionary affinities of the AUDGs are uncertain [<xref ref-type="bibr" rid="B8">8</xref>]. Thus, considerable structural diversity appears to exist among the UDGs, their generally similar catalytic activities notwithstanding.</p><p>Here, using sequence profile searches, multiple alignment analysis and structural comparisons, we unify all known UDGs into a single protein superfamily and predict a common &#x003b1;/&#x003b2; fold for them. We additionally detect several new probable UDGs that are distinct from the already characterized families, and explore the evolutionary scenarios that could have resulted in the observed phyletic distribution of these enzymes.</p></sec><sec><title>Results and discussion</title><sec><title>Characterization of the UDG superfamily using iterative database searches</title><p>An iterative PSI-BLAST search [<xref ref-type="bibr" rid="B10">10</xref>] (cut-off for inclusion of sequences into the position-specific scoring matrix <italic>e</italic> &#x0003c; 0.01) initiated with the sequence of the TM0511 protein, the prototype member of the AUDG family, retrieved, with statistically significant <italic>e</italic> values, not only its orthologs and highly conserved paralogs from a variety of organisms, but also the classical MUGs and the <italic>Drosophila</italic> ssUDG. In addition, these searches resulted in the detection of uncharacterized UDG homologs from the bacteria <italic>Deinococcus radiodurans, Campylobacter jejuni</italic> and <italic>Neisseria meningitidis.</italic> The next round of iterative searches initiated with the sequences of the newly detected UDG homologs resulted in the retrieval of the Ung family of UDGs without any false positives. Thus, by using multiple profile searches, it was possible to connect all known UDGs, as well as several putative new ones through statistically significant sequence similarity. Clustering of the proteins of the emerging UDG superfamily using reciprocal retrieval in BLASTP searches as a criterion led to the identification of six distinct families. These are: UNG (orthologs of <italic>E. coli</italic> Ung); MUG (orthologs of <italic>E. coli</italic> Mug); AUDG; ssUDG; a previously undetected family that includes members from the genus <italic>Neisseria, Mycobacterium leprae, C. jejuni</italic> and <italic>Zymomonas mobilis</italic> (UDGX); and another new family including members from <italic>D. radiodurans</italic> and <italic>Rhodococcus erythropolis</italic> (DRUDG). Proteins from each of these families were aligned separately, and the regions corresponding to conserved secondary-structure elements were identified. The available three-dimensional structures of Ung and Mug were superimposed, and the resulting structural alignment was used to combine the multiple alignments of all six UDG families (Figure <xref ref-type="fig" rid="F1">1</xref>). Comparison of the multiple alignment with the available structures showed conservation of the principal structural elements (Figure <xref ref-type="fig" rid="F1">1</xref>), indicating that all proteins of the UDG superfamily adopt the same &#x003b1;/&#x003b2; fold as Ung and Mug. This predicted structural unity of the UDGs, along with the subtle but significant sequence similarity, suggests a common evolutionary origin for the entire superfamily.</p><p>The sequence conservation in the UDG superfamily is concentrated primarily in three motifs, with the two motifs located near the amino and carboxyl termini corresponding to the substrate-binding pocket (Figures <xref ref-type="fig" rid="F1">1</xref>,<xref ref-type="fig" rid="F2">2</xref>). The ancestral core fold of the UDG superfamily consists of a central parallel four-strand &#x003b2; sheet with a 2-1-3-4 topology, which is associated with four helices; the substrate-binding pocket is formed by the regions located after strand 1 and strand 4 (Figure <xref ref-type="fig" rid="F2">2</xref>). The central conserved motif corresponds to a sharp turn between strand 3 and the adjacent helix 3, which is one of the most characteristic structural features of the UDG super-family and is probably required to support the enzyme conformation needed to accommodate the DNA. In both structurally characterized members of this superfamily (Udg and Mug), a conserved aromatic residue located in the loop preceding the conserved helix 1 (Figures <xref ref-type="fig" rid="F1">1</xref>,<xref ref-type="fig" rid="F2">2</xref>) mediates binding of the attacked uracil from the DNA double helix via stacking interactions [<xref ref-type="bibr" rid="B11">11</xref>]. This aromatic residue is replaced by an aliphatic residue in a small subset of the UDGs, and the loop may contain poorly conserved short helices in some of the UDGs, such as Mug. This position is highly conserved in the entire UDG superfamily (Figure <xref ref-type="fig" rid="F1">1</xref>), which suggests that a similar mechanism of uracil binding is universal in the UDGs.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Multiple alignment of the UDG superfamily. The secondary-structure elements of the core UDG fold are shown in color above the multiple alignment. Some nonconserved elements in the MUG structure from <italic>E. coli</italic> are indicated in gray. The coloring of the alignment positions is according to the 85% consensus that includes the following categories of amino acid residues: h, hydrophobic, l, aliphatic, a, aromatic, shaded yellow (YFWLIVMA); s, small, individual letters colored green (SAGTVPNHD); p, polar, colored purple (STQNEDRKH); u, tiny, shaded green (GAS); and b, big, shaded gray (KREQWFYLMI). Af, <italic>Archaeoglobus fulgidus</italic>; Bb, <italic>Borrelia burgdorferi</italic>; Bs, <italic>Bacillus subtilis</italic>; Cj, <italic>Campylobacter jejuni</italic>; Ct, <italic>Chlamydia trachomatis</italic>; Dm, <italic>Drosophila melanogaster</italic>; Dr, <italic>Deinococcus radiodurans</italic>; Ec, <italic>Escherichia coli</italic>; Hi, <italic>Haemophilus influenzae</italic>; Hp, <italic>Helicobacter pylori</italic>; Hs, <italic>Homo sapiens</italic>; Mtu, <italic>Mycobacterium tuberculosis</italic>; Ph, <italic>Pyrococcus horikoshii</italic>; Rp, <italic>Rickettsia prowazekii</italic>; Sc, <italic>Saccharomyces cerevisiae</italic>; Scoel, <italic>Streptomyces coelicolor</italic>; Sp, <italic>Schizosaccharomyces pombe</italic>; Ssp, <italic>Synechocystis</italic> sp.; Tp, <italic>Treponema pallidum</italic>; Uu, <italic>Ureaplasma urealyticum</italic>; Yp, <italic>Yersinia pestis</italic>. The numbers at each end of each sequence are amino-acid positions and indicate the extent of the domain in each protein. The numbers within the alignment indicate inserts that have not been shown. The conserved motifs discussed in the text are designated I, II and III; the conserved aromatic (aliphatic) residue involved in the stacking interaction with uracil is indicated by an asterisk.</p></caption><graphic xlink:href="gb-2000-1-4-research0007-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>The topology of the UDG superfamily core fold, with the conserved and unique features of different families. The core secondary-structure elements of the UDG fold are colored as in Figure <xref ref-type="fig" rid="F1">1</xref> and numbered according to their order in the sequence. The elements observed only in the MUGs are shown in gray. The conserved motif I occurs after strand 1 and motif II occurs after strand 4 and forms the active-site pocket in the three-dimensional structure.</p></caption><graphic xlink:href="gb-2000-1-4-research0007-2"/></fig></sec><sec><title>Catalytic mechanism</title><p>The experimental determination of a similar catalytic activity in diverse members of the UDG superfamily and the conservation of the substrate-binding site suggest a generally conserved catalytic mechanism. However, several family-specific features predict interesting differences in the catalytic properties of the individual families. On the basis of studies on Ung-family enzymes such as those from herpesviruses, it has been suggested that protonation of the O2 of the flipped-out uracil is carried out by the conserved histidine in motif III, which acts as a general acid [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. Studies on the <italic>E. coli</italic> Ung enzyme, however, have shown that this conserved histidine does not act as a general acid, but instead is neutral and acts as an electrophile [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. On this basis, it has been proposed that the electrophilic interaction stabilizes the developing enolate on the uracil O2 in course of its excision [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. This reaction is assisted by the conserved aspartate in motif I that acts as a general base and directs a water molecule for the nucleophilic attack [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. The MUGs and the new family of bacterial UDGs (UDGX) identified here lack both the conserved electrophile (histidine) and the general base (aspartate) (Figure <xref ref-type="fig" rid="F1">1</xref>), which suggests that these are less efficient enzymes [<xref ref-type="bibr" rid="B5">5</xref>]. The remaining UDG families typically contain the electrophilic histidine, but not the general base aspartate in motif I (Figure <xref ref-type="fig" rid="F1">1</xref>). A subset of the AUDGs and the newly identified DRUDG family, however, contain a glutamate one position upstream of the aspartate present in motif I of the UNGs (Figure <xref ref-type="fig" rid="F1">1</xref>); this glutamate could act as an alternative general base for this subset of UDGs. Additionally, the loop formed by motif III also helps in clamping on the phosphate backbone to allow recognition of the target nucleotide by the active site [<xref ref-type="bibr" rid="B11">11</xref>]. The discrimination of uracil over cytosine in enzymes of the UNG family depends on the asparagine located near the end of the core strand 2 (Figures <xref ref-type="fig" rid="F1">1</xref>,<xref ref-type="fig" rid="F2">2</xref>). An asparagine or aspartate is conserved in the majority of the UDG superfamily enzymes in this position, with the exception of some members of the MUG, AUDG and UDGX families. Both <italic>E. coli</italic> Mug and its human ortholog, TDG, have been shown to act on powerfully mutagenic alkylated bases such as etheno-cytosine [<xref ref-type="bibr" rid="B15">15</xref>]. Mutational replacement of asparagine by aspartate in the human uracil DNA glycosylase (UNG) results in its acquiring cytosine glycosylase activity [<xref ref-type="bibr" rid="B16">16</xref>]. This substitution, which occurs naturally in several UDGX family proteins, along with other replacements of asparagine in this position in different members of this superfamily (Figure <xref ref-type="fig" rid="F1">1</xref>), is probably an adaptation for removal of mutagenic alkylated bases such as etheno-cytosine.</p></sec><sec><title>Evolution</title><p>On the basis of the conservation of functionally important residues in the UDG superfamily, a parsimonious, although speculative, scheme for the evolution of these enzymes can be proposed (Figure <xref ref-type="fig" rid="F3">3</xref>). The ancestral uracil DNA glycosylase probably possessed the core fold with an asparagine at the end of strand 2 and a histidine at the end strand 4 and most closely resembed the AUDG and DRUDG families. From this ancestral form, the high-activity forms such as the UNG class could have evolved by acquiring the general base in motif I. The acquisition of the glutamate in motif I of the AUDGs and DRUDGs could represent independent evolution of the same type of high-activity enzyme. The lower-activity forms, such as the MUGs and the UDGXs, could have evolved by replacement of the ancestral electrophilic histidine that stabilized the reaction intermediate by another polar residue such as serine or asparagine. The localization of the active site formed by long loops on the same side of the UDG molecules probably resulted in relaxation of the selective constraints on their sequences beyond the maintenance of the general shape of the binding pocket. Moreover, even the charged residues in the binding pocket are not entirely constrained, because the enzyme mechanism seems to depend more critically on the steric strain caused by base-flipping than on the base or other residues that stabilize the intermediate. These features of the UDGs probably contributed to the evolution of a very high level of sequence divergence, without a single residue conserved throughout the superfamily, which is unusual for homologous enzymes that catalyze essentially the same reaction.</p><p>The phyletic distribution of the UDGs shows partial complementarity between different families, which suggests that they perform at least partially overlapping functions in different organisms (Table <xref ref-type="table" rid="T1">1</xref>). Each completely sequenced genome, with the apparent exception of the archaea <italic>Methanococcus jannaschii</italic> and <italic>Methanobacterium thermoautotrophicum,</italic> encodes at least one member of the UDG superfamily, with a maximum of four members in the case of the radioresistant bacterium <italic>D. radiodurans</italic> (Table <xref ref-type="table" rid="T1">1</xref>). Each of these families, with the exception of the ssUDGs, which so far are limited to animals, shows a patchy spread over a wide phylogenetic range, which suggests important roles for horizontal gene transfer and lineage-specific gene loss in the evolution of the UDGs. The presence of AUDG in at least one bacteriophage and of UNGs in large eukaryotic DNA viruses (Table <xref ref-type="table" rid="T1">1</xref>) point to one possible type of vehicle for horizontal dissemination of these enzymes. The phyletic distribution of the UDGs suggests that the AUDGs could be the ancestral form, possibly inherited from the last common ancestor of all extant life forms. This seems to be compatible with the apparent ancestral layout of the active center of these enzymes (see above). The UNGs appear to be a primitive bacterial form, whereas the MUG-UDGX group could have been derived at a later stage of bacterial evolution. The separation between UNGs and MUG-UDGX could have been driven by selection for distinct functional niches, a uracil-specific enzyme in the case of the former and a G:U/T mismatch repair enzyme in the latter. The UDGX and MUG families show a closer relationship to each other than to other families of the superfamily, suggesting a relatively recent divergence and a similar mismatch repair function. The DRUDGs appear to be specialized derivatives that emerged within one bacterial lineage followed by limited dispersal, at least in the currently sampled bacterial taxa. Under this scenario, AUDGs have been displaced in some of the bacteria, and possibly in the ancestral eukaryotes, by the UNGs and MUGs.</p><p>The AUDGs are fused to two distinct DNA polymerases - a DNA polymerase III&#x003b1; subunit in <italic>Yersinia pestis</italic> and a polymerase of the A family (homolog of bacterial Pol I) in bacteriophage SP01. This fusion is the cause of many erroneous annotations of AUDG family members as 'putative phage-type DNA polymerases' that are found in current sequence databases. The fusion with the polymerases suggests that the functioning of the AUDGs, and possibly other UDGs, could be tightly coupled to that of the DNA replication apparatus. This may be particularly important in the archaea, whose polymerases stall at uridines in the template strand [<xref ref-type="bibr" rid="B17">17</xref>]. Given this possible function of AUDGs in replication and the fundamental role of UDGs in repair, the apparent absence of these enzymes in two archaeal methanogens is unexpected. Although these archaeal genomes could encode extremely divergent members of the UDG superfamily that escaped detection even in the present detailed analysis, it seems more likely that in these archaea the UDGs have been displaced by unrelated enzymes of the &#x003b1;-helical MutY super-family [<xref ref-type="bibr" rid="B18">18</xref>].</p><p>Eukaryotes encode UNG- and MUG-family enzymes that are not found in archaea and are closely related to their bacterial orthologs. This strongly suggests acquisition from bacterial endosymbionts (including mitochondria), followed by displacement of the UDG inherited from the common ancestor with archaea (probably AUDG). The MUG-family enzymes from animals have low-complexity segments on either side of the DNA glycosylase domain. In the case of <italic>Drosophila</italic> these are particularly expanded and are associated with two minor groove DNA-binding motifs, the AT hooks [<xref ref-type="bibr" rid="B19">19</xref>]. This motif is found in many chromosomal proteins and could help in the translocation of the enzyme to specific sites in chromatin, such as matrix attachment regions. Interestingly, different eukaryotic lineages show notable differences in their repertoires of UDGs, with only an UNG-family enzyme so far detected in the nematode <italic>Caenorhabditis elegans.</italic> The phylogenetic affinity of the ssUDGs, which have been detected up to now only in coelomates, is hard to discern, because they have evolved distinct structural features, such as long inserts, that are not seen in the other members of the UDG superfamily. The presence of a histidine in motif III suggests that the ssUDGs could have evolved from a UNG-like enzyme by rapid divergence. The evolutionary divergence and the origin of acquisition of a distinct DNA glycosylase may correlate with the need for an enzyme that can meet the particular DNA repair needs of multicellular animals, such as the repair of frequently transcribed DNA.</p><fig position="float" id="F3"><label>Figure 3</label><caption><p>A hypothetical evolutionary scenario for the UDG superfamily. The different families are shown in different colors and potential order and lineage of derivation is indicated on the standard phylogenetic model for the three domains of life. The representation of the active-site pocket residues typical of that set is shown next to each class at the point of derivation. The first position is the general base represented by an aspartate in the UNGs, the second position is the uracil/cytosine discrimination site occurring after the core strand 2, and the third position is typically represented by a histidine that acts as an electrophile. The X at a given position denotes lack of conservation. In some of the AUDGs and the DRUDGs, a glutamate could function as alternative general base.</p></caption><graphic xlink:href="gb-2000-1-4-research0007-3"/></fig><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Phyletic distributions of the six families of UNGs</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Species/family</td><td align="center">UNG^*</td><td align="center">AUDG^*</td><td align="center">MUG</td><td align="center">SsUDG^* DRUDG^*</td></tr><tr><td></td><td></td><td></td><td align="center">+ UDGX^*</td><td></td></tr></thead><tbody><tr><td align="left">Bacteria</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Escherichia coli</italic></td><td align="center">1</td><td></td><td align="center">1 (MUG)</td><td></td></tr><tr><td align="left">&#x02003;<italic>Haemophilus influenzae</italic></td><td align="center">1</td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Neisseria meningitidis</italic></td><td align="center">1</td><td align="center">1</td><td align="center">1 (UDGX)</td><td></td></tr><tr><td align="left">&#x02003;<italic>Rickettsia prowazekii</italic></td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Campylobacter jejuni</italic></td><td align="center">1</td><td align="center">1</td><td align="center">1 (UDGX)</td><td></td></tr><tr><td align="left">&#x02003;<italic>Helicobacter pylori</italic></td><td align="center">1</td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Bacillus subtilis</italic></td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Mycoplasma genitalium</italic></td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Mycoplasma pneumoniae</italic></td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Ureaplasma urealyticum</italic></td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Deinococcus radiodurans</italic></td><td align="center">1</td><td align="center">1</td><td align="center">1 (MUG)</td><td align="center">1</td></tr><tr><td align="left">&#x02003;<italic>Mycobacterium tuberculosis</italic> 1</td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Streptomyces coelicolor</italic></td><td align="center">1</td><td align="center">2</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Synechocystis</italic> sp.</td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Chlamydia trachomatis</italic></td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Chlamydophila pneumoniae</italic> 1</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Treponema pallidum</italic></td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Borrelia burgdorferi</italic></td><td align="center">1</td><td align="center">1(d)^&#x02020;</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Aquifex aeolicus</italic></td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Thermotoga maritima</italic></td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">Archaea</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Aeropyrum pernix</italic></td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Archaeoglobus fulgidus</italic></td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Pyrococcus horikoshii</italic></td><td></td><td align="center">1</td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Methanobacterium</italic></td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;&#x02003;<italic>thermoautotrophicum</italic></td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Methanococcus jannaschii</italic></td><td></td><td></td><td></td><td></td></tr><tr><td align="left">Eukaryota</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Saccharomyces cerevisiae</italic></td><td align="center">1</td><td></td><td align="center">(r)^&#x02021;</td><td></td></tr><tr><td align="left">&#x02003;<italic>Schizosaccharomyces pombe</italic>1</td><td></td><td></td><td align="center">1(MUG)</td><td></td></tr><tr><td align="left">&#x02003;<italic>Caenorhabditis elegans</italic></td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;<italic>Drosophila melanogaster</italic></td><td align="center">(?)^&#x000a7;</td><td></td><td align="center">1(MUG)</td><td align="center">1</td></tr><tr><td align="left">&#x02003;<italic>Homo sapiens</italic></td><td align="center">1</td><td></td><td align="center">1(MUG)</td><td align="center">1</td></tr><tr><td align="left">Large DNA viruses</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Poxviruses</td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Herpesviruses</td><td align="center">1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Bacteriophages SPO1</td><td></td><td align="center">1</td><td></td><td></td></tr></tbody></table><table-wrap-foot><p>^*The number of detected representatives of each family is indicated for each species. Note that duplication is uncharacteristic of the UNGs. ^&#x02020;(d) indicates a possibly disrupted version in which the amino-terminal conserved motifs are not detectable; ^&#x02021;(r) indicates an apparent recent loss in <italic>S. cerevisiae</italic>, as the gene is retained in the related yeast <italic>Candida albicans</italic>; ^&#x000a7;(?) indicates the unusual lack of a detectable UNG in both the genome and EST sequences.</p></table-wrap-foot></table-wrap></sec></sec><sec><title>Conclusions</title><p>Using sequence profile searches, multiple alignment analysis and protein structure comparisons, we have shown that all known UDGs form a single protein superfamily with a distinct structural fold and a common evolutionary origin. The extreme sequence divergence of different families of UDGs is probably due to differences in their biochemistry, with only the general shape of the protein molecule and the binding pocket being essential for the DNA glycosylase reaction <italic>per se.</italic> Although the UDG superfamily is nearly ubiquitous among cellular life forms, the individual families show limited and distinct phyletic distributions. The emerging evolutionary scenario for the UDGs involves multiple events of lateral gene transfer and lineage-specific gene loss. In addition, we predict two previously undetected families of UDGs; the experimental investigation of their functions is expected to broaden the current perspective on these critical repair enzymes.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><p>The databases used in this study were the Non-redundant Nucleotide and Protein and the Expressed Sequence Tags (EST) databases (National Center for Biotechnology Information) and the individual protein sequence databases of completely and partially sequenced genomes [<xref ref-type="bibr" rid="B20">20</xref>]. Local alignment searches were performed using the gapped version of the BLAST programs (BLASTPGP for proteins and TBLASTNGP for translating searches of nucleotide databases) [<xref ref-type="bibr" rid="B10">10</xref>]. Sequence profile searches were performed using the PSI-BLAST program [<xref ref-type="bibr" rid="B10">10</xref>] or using the HMMSEARCH program, with input hidden Markov models generated from multiple alignments using the HMMBUILD program [<xref ref-type="bibr" rid="B21">21</xref>]. The multiple alignments were generated using a combination of PSI-BLAST and CLUSTALW [<xref ref-type="bibr" rid="B22">22</xref>]. The statistically significant motifs were detected using the Gibbs sampling option of the MACAW program [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. The three-dimensional structure visualization, alignment and modeling were carried out using the SWISS-PDB-Viewer program [<xref ref-type="bibr" rid="B25">25</xref>].</p></sec>

Analysis of prolactin-modulated gene expression profiles during the Nb2 cell cycle using differential screening techniques

<sec><title>Background:</title><p>Rat Nb2-11C lymphoma cells are dependent on prolactin for proliferation and are widely used to study prolactin signaling pathways. To investigate the role of this hormone in the transcriptional mechanisms that underlie prolactin-stimulated mitogenesis, five different techniques were used to isolate differentially expressed transcripts: mRNA differential display, representational difference analysis (RDA), subtractive suppressive hybridization (SSH), analysis of weakly expressed candidate genes, and differential screening of an organized library.</p></sec><sec><title>Results:</title><p>About 70 transcripts were found to be modulated in Nb2 cells following prolactin treatment. Of these, approximately 20 represent unknown genes. All cDNAs were characterized by northern blot analysis and categorized on the basis of their expression profiles and the functions of the known genes. We compared our data with other cell-cycle-regulated transcripts and found several new potential signaling molecules that may be involved in Nb2 cell growth. In addition, abnormalities in the expression patterns of several transcripts were detected in Nb2 cells, including the constitutive expression of the immediate-early gene <italic>EGR-1.</italic> Finally, we compared the differential screening techniques in terms of sensitivity, efficiency and occurrence of false positives.</p></sec><sec><title>Conclusions:</title><p>Using these techniques to determine which genes are differentially expressed in Nb2 lymphoma cells, we have obtained valuable insight into the potential functions of some of these genes in the cell cycle. Although this information is preliminary, comparison with other eukaryotic models of cell-cycle progression enables identification of expression abnormalities and proteins potentially involved in signal transduction, which could indicate new directions for research.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Bole-Feysot</surname><given-names>Christine</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Perret</surname><given-names>Eric</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Roustan</surname><given-names>Paul</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Bouchard</surname><given-names>Brigitte</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Kelly</surname><given-names>Paul A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Introduction</title><p>Prolactin is a pleiotropic hormone whose numerous actions are associated with reproduction, growth and development, water and electrolyte balance, metabolism, behavior and immunoregulation [<xref ref-type="bibr" rid="B1">1</xref>]. The prolactin-dependent rat Nb2 lymphoma (Nb2-11C) is widely used as a model in which to study signal transduction and transcriptional mechanisms that underlie prolactin-stimulated mitogenesis (reviewed in [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]). Deprivation of lactogen induces a block in the early G1 phase of the Nb2 cell cycle [<xref ref-type="bibr" rid="B4">4</xref>]. The addition of physiological concentrations of prolactin to synchronized G0/G1-arrested cultures reinitiates cell-cycle progression [<xref ref-type="bibr" rid="B5">5</xref>], which is characterized by the induction of growth-related genes such as those for c-Myc, &#x003b2;-actin and ornithine decarboxylase (ODC) and <italic>hsp70</italic>-like genes [<xref ref-type="bibr" rid="B6">6</xref>]. This list has recently been extended to include the genes for interferon regulatory factor-1 (IRF-1) [<xref ref-type="bibr" rid="B7">7</xref>], cyclins D2 and D3 [<xref ref-type="bibr" rid="B8">8</xref>], proto-oncogene Pim-1 [<xref ref-type="bibr" rid="B9">9</xref>], growth factor independence-1 (Gfi-1) [<xref ref-type="bibr" rid="B10">10</xref>], and rat nuclear distribution c (Rnudc) [<xref ref-type="bibr" rid="B11">11</xref>]. The identification of other prolactin-regulated genes in proliferating Nb2 cells would help to elucidate the relationships between prolactin-activated proteins and genes induced or repressed by prolactin, and lead to a better understanding of the role of prolactin in proliferation regulatory mechanisms. In this study, five differential screening techniques were applied at different stages of the Nb2 cell cycle. Differential display [<xref ref-type="bibr" rid="B12">12</xref>], representational difference analysis (RDA [<xref ref-type="bibr" rid="B13">13</xref>]) and suppressive subtractive hybridization (SSH [<xref ref-type="bibr" rid="B14">14</xref>]) consist of selective and/or suppressive cycles of PCR using cDNA prepared from the cell populations or tissues to be compared. The two other techniques used were the screening of an organized library [<xref ref-type="bibr" rid="B15">15</xref>] and the analysis of weakly expressed candidate genes. These two methods are based on the hybridization of DNA macro- or microarrays on nylon filters using complex probes generated from radiolabeled transcribed cDNA isolated from the cell populations to be compared.</p><p>We have characterized known and unknown transcripts identified by these five techniques, adding information relative to their expression peak or expression variations during Nb2 cell proliferation. Whenever possible, prolactin-induced transcripts were compared with those in other eukaryotic models of cell-cycle progression such as <italic>Saccharomyces cerevisiae</italic> and normal human fibroblasts. This comparison allowed us to establish non-exhaustive lists of cell-cycle-regulated transcripts. Regulated mRNAs were classified with respect to their functional characteristics and to their conservation from yeast to vertebrates. On the basis of this analysis, new signaling molecules presumably involved in Nb2 proliferation are proposed. Furthermore, we have detected expression profile abnormalities in Nb2 lymphoma cells, and we discuss the consequence of one, the constitutive expression of the immediate-early gene <italic>EGR-1.</italic></p></sec><sec><title>Results</title><sec><title>Application of the different screening techniques to Nb2 cells</title><p>When deprived of lactogen, 80-85% of an Nb2 cell culture is synchronized in growth arrest [<xref ref-type="bibr" rid="B5">5</xref>] (Figure <xref ref-type="fig" rid="F1">1</xref>). Addition of prolactin to the culture reinitiates cell-cycle progression and cell proliferation. Using differential display, we first compared RNAs from synchronized Nb2 cells stimulated for various times with prolactin. In addition, three different RDA and SSH subtractive libraries were prepared. One RDA library allowed the identification of transcripts expressed at a higher level during proliferation (12 hours prolactin-stimulated) compared with growth arrest, and two SSH subtractive libraries were used to compare expression profiles in growth arrest and G1 (mix of cells stimulated with prolactin for 2, 4, 6 and 8 hours) and <italic>vice versa.</italic> Messenger RNAs from Nb2 cells were used to differentially screen an organized library of rat brain cDNA. Finally, the expression of 91 weakly expressed candidate genes was also compared in growth-arrested, early (mix of 2, 4, 6 and 8 hours prolactin-stimulated Nb2 cells), intermediate (10, 12 and 14 hours) or late (20, 22 and 24 hours) proliferative phase and unsynchronized Nb2 cells.</p><p>Most of the potential positive clones isolated by differential display, RDA, SSH and screening of rat brain organized library were analyzed by northern blot to eliminate false positives and to evaluate variations in the expression of each clone during the Nb2 proliferative response. The remaining cDNA clones were tested using reverse northern blot to rapidly eliminate false-positive cDNAs. Briefly, PCR products corresponding to potential positive clones were screened by hybridization with complex probes generated from the populations tested. This step was necessary because of the high rate of false-positive clones generated by the earlier protocols used for differential display. The method has since been improved and may now have a better readout. Sequencing of these clones enabled us to identify known transcripts and to determine which ones were of unknown genes.</p><p>Together, the different techniques enabled the isolation of about 70 known or unknown differentially expressed transcripts potentially involved in the resumption of cell proliferation by quiescent cells. A summary of the data is presented in Table <xref ref-type="table" rid="T1">1</xref>. Examples of expression profiles obtained by northern blot using the isolated cDNAs as probes are shown in Figure <xref ref-type="fig" rid="F2">2</xref>. We did not isolate transcripts for known molecules such as histones or cyclins; it is, however, of interest to note that the expression of the rat homolog of Cdc21, the adenosine nucleotide translocator Ant-2, the nuclear export factor CRM-1 (exportin), and unknown transcripts DD3, 4-16 and 4-15 (Figure <xref ref-type="fig" rid="F2">2</xref>) are induced during Nb2 cell proliferation. In contrast, expression of the unknown transcript 6-4 is decreased during G1 phase, but this transcript is much more abundant in unsynchronized Nb2 cells. Northern blots indicate that some of these cDNA probes identify several distinct transcripts, probably generated by alternative splicing (ANT-2, CRM-1, CD45 (leukocyte common antigen), DD3, 4-15), which are not necessarily all induced in the same manner. Interestingly, two opposite expression profiles are observed for the two transcripts recognized by the cDNA probe identified as CD45 (clone from the SSH library G1 phase growth arrest). Indeed, the longer transcript is progressively repressed during Nb2 cell-cycle progression, whereas the shorter form is induced. These examples emphasize that northern blot analysis provided new information that could not be obtained using other methods (such as reverse northern blot, arrays or DNA chips). As shown in Table <xref ref-type="table" rid="T1">1</xref>, about 20 of the differentially expressed cDNAs that were isolated correspond to unknown transcripts whose expression is modulated during Nb2 proliferation. Most of these unknown cDNAs share significant homology with several mouse and human expressed sequence tags (ESTs) isolated from various libraries, suggesting that the corresponding transcripts are ubiquitously expressed and have a role in cell proliferation in one of the different functional categories described below.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Cell-cycle analysis of synchronized Nb2 cells stimulated by prolactin. Nb2 cells were serum deprived for 24 hours, then incubated with no PRL or PRL at 20 ng/ml before cells were collected for analysis. DNA content (FL2-A) <italic>versus</italic> cell number is presented in each panel. <bold>(a)</bold> Profiles obtained with control cells; <bold>(b)</bold> profiles obtained with cells incubated with PRL. From left to right, profiles correspond to cells in apoptosis (Apo, area below 400 on the x axis), in G0/G1 (peak centered on 400 on the x axis), or in S/M phase (area above 400 on the x axis).</p></caption><graphic xlink:href="gb-2000-1-4-research0008-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>Expression profiles of various known and unknown transcripts during Nb2 cell cycle progression. Samples of total RNA (10 &#x003bc;g) were loaded per lane and blots were hybridized with the indicated cDNA probes. Ethidium bromide staining (EtBr) of the gels is shown as a control (18S and 28S rRNA). <bold>(a)</bold> Cdc21 homolog; <bold>(b)</bold> Ant-2; <bold>(c)</bold> CRM-1; <bold>(d)</bold> CD45; <bold>(e)</bold> unknown DD3; <bold>(f)</bold> unknown 4-16; <bold>(g)</bold> unknown 4-15; <bold>(h)</bold> unknown 6-4.</p></caption><graphic xlink:href="gb-2000-1-4-research0008-2"/></fig></sec><sec><title>Putative signaling molecules potentially involved in Nb2 cell survival and/or cell-cycle progression</title><p>The list of differentially expressed transcripts has been completed with Nb2 cell transcripts described in previous reports (Table <xref ref-type="table" rid="T2">2</xref>). These differentially expressed genes can be found in almost all the subclasses listed in Table <xref ref-type="table" rid="T2">2</xref>, including, for example, those for receptors (such as the prolactin receptor [<xref ref-type="bibr" rid="B16">16</xref>], the T-cell receptor &#x003b3; chain [<xref ref-type="bibr" rid="B17">17</xref>], the vitamin D3 receptor, the thromboxane A2 and the prostaglandin F2 receptors (our present results)), transcription factors (IRF-1 [<xref ref-type="bibr" rid="B7">7</xref>], c-Myc [<xref ref-type="bibr" rid="B18">18</xref>], Zfx (our present results)), and T-cell survival and apoptosis molecules (Bag-1 [<xref ref-type="bibr" rid="B19">19</xref>], Bcl-2, Bax [<xref ref-type="bibr" rid="B20">20</xref>], Ant-2 (our present results)). Although all the signaling molecules are not necessarily regulated at the transcriptional level, it can be hypothesized that those translated from cell-cycle-modulated transcripts are involved in Nb2 cell-cycle progression. This hypothesis seems to be confirmed, as we have found that several molecules previously described as transducers in Nb2 cells are encoded by cell-cycle-modulated transcripts (for example, phosphatidylinositide 3-kinase (PI 3-kinase), phospholipase C&#x003b3;1 (PLC&#x003b3;1), and focal adhesion kinase (FAK) p125). Consequently, it is tempting to speculate that at least some of the cell-cycle-modulated transcripts may encode transducers of Nb2 cell proliferation. For example, the stress kinase p38 mitogen-activated protein kinase (p38 MAP kinase), whose transcript is induced in Nb2 cells upon prolactin stimulation [<xref ref-type="bibr" rid="B21">21</xref>], maybe involved in prolactin-induced signaling pathways. This hypothesis is reinforced by the fact that p38 MAP kinase seems to be required for the optimal activation of T cells by interleukin (IL)-12 and IL-2 and for the regulation of serine phosphorylation of STAT transcription factors [<xref ref-type="bibr" rid="B22">22</xref>]. The GD3 ganglioside synthase, which mediates the propagation of CD95-generated apoptotic signals in hematopoietic cells [<xref ref-type="bibr" rid="B23">23</xref>], may also be involved in the regulation of survival and apoptosis in Nb2 cells. It is also possible that receptors, such as the prostaglandin F2, thromboxane A2 and vitamin D3 receptors, galectin-8 and CD45 and their ligands, may be involved in the signaling pathways required for Nb2 cell survival and proliferation.</p></sec><sec><title>Functional classification of cell-cycle-regulated transcripts</title><p>We identified 70 differentially expressed genes in proliferating Nb2 cells. Although this number is not negligible, it is of course not exhaustive, as the number of genes involved in cell-cycle modifications could be as high as several hundred. However, estimations of the number of genes modulated using other proliferation and cell-cycle models, such as yeast [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>] or human fibroblasts [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>], are equally limited. We compared these different models, adding new information generated from large-scale differential screening techniques (microarrays and DNA chips). On the basis of these analyses, approximately 7% of transcripts from yeast (416 out of 6,220) and 6% from normal human fibroblasts (517 out of 8,600) display cell-cycle-dependent fluctuations. All the yeast cell-cycle-regulated transcripts are not, however, regulated in vertebrates, and <italic>vice versa.</italic></p><p>These transcripts were classified into 10 different functional categories as shown in Table <xref ref-type="table" rid="T2">2</xref>: cell cycle (cyclins and cell-cycle regulators); nucleotide metabolism, and DNA replication and repair; chromatin structure; cytoskeleton, cell surface antigens, adhesion molecules and signaling molecules (involved in apoptosis, survival and/or proliferation); heat shock, stress response and chaperones; metabolism (energy); protein and RNA synthesis, modifications and degradation; inter-compartment transport and trafficking; and unknown function(s).</p><p>These functional categories agree with previous observations concerning cell-cycle-regulated transcripts in various eukaryotic models. Indeed, cell-cycle-dependent mRNA fluctuations have been observed for genes involved in many cellular processes, including control of mRNA transcription [<xref ref-type="bibr" rid="B29">29</xref>], responsiveness to external stimuli [<xref ref-type="bibr" rid="B30">30</xref>] and subcellular localization of proteins [<xref ref-type="bibr" rid="B31">31</xref>]. Genetic studies have revealed that the activity of cell-cycle regulatory proteins is required for normal DNA repair [<xref ref-type="bibr" rid="B32">32</xref>], meiosis [<xref ref-type="bibr" rid="B33">33</xref>] and multicellular development [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>]. These observations suggest that, in eukaryotic cells, diverse biological events depend on maintenance of this periodicity.</p></sec><sec><title>Expression abnormalities</title><p>The loss of appropriate cell-cycle regulation leads to genomic instability [<xref ref-type="bibr" rid="B36">36</xref>] and is believed to have a role in the etiology of both hereditary and spontaneous cancers [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>,<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. In Nb2-11C cells, several growth-related genes that display abnormalities in their expression patterns were observed. These abnormalities may be the cause or the consequence of the tumor phenotype of Nb2 cells.</p><p>Using the candidate gene approach (Figure <xref ref-type="fig" rid="F3">3a</xref>), striking expression abnormalities were observed. For example, Nb2 cells display an abnormal response to heat shock. Indeed, whereas the <italic>hsp70</italic>-like mRNA is upregulated following lactogen stimulation [<xref ref-type="bibr" rid="B41">41</xref>], no expression of the inducible <italic>hsp70</italic> (GenBank X74271) gene was detected. As components of the heat-shock response are involved in normal cell-cycle-progression, the abnormalities observed in Nb2 cells may have important consequences for their growth. Furthermore, in comparison with mammalian models of cell-cycle progression, expression abnormalities of immediate-early genes are observed in Nb2 cells. Indeed, the expression of c-<italic>fos</italic> remains undetectable in our model (Table <xref ref-type="table" rid="T2">2</xref>) as well as in starved Nb2 cells, which resume proliferation after prolactin stimulation [<xref ref-type="bibr" rid="B18">18</xref>]. In contrast, the expression of <italic>EGR-1</italic> (also termed <italic>Zif268 / KROX24 / ETR103 / NGFIA / TIS8 / GOS30</italic>) is constitutive in Nb2 cells. This peculiarity, observed using the candidate gene approach (Figures <xref ref-type="fig" rid="F3">3b</xref>,<xref ref-type="fig" rid="F4">4</xref>), was confirmed by northern blot (Figure <xref ref-type="fig" rid="F3">3c</xref>). The gene has been shown in numerous model systems to have induction kinetics similar to c-<italic>fos</italic>, characterized by a rapid transient expression requiring <italic>de novo</italic> transcription between 15 and 30 minutes after the mitogenic stimulus [<xref ref-type="bibr" rid="B42">42</xref>,<xref ref-type="bibr" rid="B43">43</xref>,<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>,<xref ref-type="bibr" rid="B47">47</xref>].</p><fig position="float" id="F3"><label>Figure 3</label><caption><p>Analysis of candidate gene expression. <bold>(a)</bold> General principles. Messenger RNAs from the different cell populations (cells A and B) are reverse transcribed. Multiplex PCR is then performed using specific primer pairs to amplify the cDNAs of interest. The resulting mixture of PCR products is radiolabeled and these complex probes are used to hybridize identical membranes spotted with the candidate gene cDNA targets. After autoradiography, the intensities of the hybridization signals are compared and quantified. Arrows indicate the positions of differentially expressed genes. The absence of hybridization (open circles) indicates that the candidate gene is not expressed. <bold>(b)</bold> Efficiency of the technique and examples of differentially expressed genes. The expression of different candidate genes was compared in either unsynchronized (UN), growth-arrested (GA), G1 phase (G1), G1/S transition (G1/S) or G2 phase (G2) cultures of Nb2 cells. The efficiency of the technique was controlled using equivalent amounts of rabbit &#x003b1; and &#x003b2; globin cDNAs, which were included on the nylon membranes along with the candidate gene targets. The two globin cDNAs were added in different amounts (50 or 150 ng) to each cDNA population before co-amplification. For each population tested, filters were hybridized with both globin probes, but only representative hybridization signals are shown, for either &#x003b1; (Panel A) or &#x003b2; (Panel B) globin. Numbers 1 and 3 represent the relative amount of the control rabbit globin cDNAs added, and are reflected in the differences in the intensity of the hybridization signals. Thus, a threefold difference in the quantity of a particular transcript in the initial population generates a clear difference in the intensity of the corresponding hybridization signals. Rows C, D, E and F are examples of the results obtained with ganglioside synthase GD3, EGR-1, FAK p125 and Stat3, respectively. Except for Stat3, which is not differentially expressed in probes UN, GA, G1, G1/S and G2, the three other genes showed a clear differential expression. <bold>(c)</bold> Northern blot analysis showing the constitutive expression of EGR-1 during Nb2 cell-cycle progression. Growth-arrested Nb2 cells were stimulated with ovine prolactin and collected after various periods of stimulation corresponding to different stages of the cell cycle (G1, G1/S and G2). The expression of EGR-1 was evaluated by northern blot using 10 &#x003bc;g of total RNA from the various times following prolactin stimulation. Ethidium bromide (EtBr) staining of the gel is shown as a control (18S and 28S rRNA).</p></caption><graphic xlink:href="gb-2000-1-4-research0008-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>Schematic representation of rat candidate genes on a nylon filter. Squares with names and accession numbers represent the places where the cDNAs were spotted. The solid gray boxes correspond to the controls (rabbit &#x003b1; and &#x003b2; globin). The boxes enclosed in a thick black square represent differentially expressed genes in Nb2 cells; the boxes enclosed in a thin black square represent genes that are repressed, but not differentially in Nb2 cells; and those enclosed in an oval correspond to expression abnormalities in Nb2 cells.</p></caption><graphic xlink:href="gb-2000-1-4-research0008-4"/></fig><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Differentially expressed transcripts found in Nb2 cells during cell-cycle progression using five different screening techniques</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Identity</td><td></td><td align="center">Accession number</td><td align="left">Expression variations</td></tr></thead><tbody><tr><td align="left">Differential display</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Unknown DD3</td><td></td><td align="center">No EST</td><td align="left">Three transcripts (2, 2.8, 3.7 kb)</td></tr><tr><td></td><td></td><td></td><td align="left">expressed in G1 and G1/S,</td></tr><tr><td></td><td></td><td></td><td align="left">(&#x0003e;4-fold induction)</td></tr><tr><td align="left">Representational difference analysis: G1/S-GA</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;ATP synthase &#x003b2; subunit</td><td></td><td align="center">M19044</td><td align="left">Induced in G1/S, G2 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Aldehyde dehydrogenase</td><td></td><td align="center">U79118</td><td align="left">Induced in late G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Enolase &#x003b1;1</td><td></td><td align="center">NM012554</td><td align="left">Induced in late G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Dynein heavy chain</td><td></td><td align="center">D13896</td><td align="left">Induced in G1, G1/S (3-fold induction)</td></tr><tr><td align="left">&#x02003;TCP-1 &#x003b5;</td><td></td><td align="center">D43950(h)</td><td align="left">Induced in G1/S, G2 (3-fold induction)</td></tr><tr><td align="left">&#x02003;&#x003b1;COP (COPA)</td><td></td><td align="center">U24105</td><td align="left">Induced in late G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Cdc21 homolog</td><td></td><td align="center">D26089 (m)</td><td align="left">Peaks in G1/S (&#x0003e;8-fold induction)</td></tr><tr><td align="left">&#x02003;Ribosomal protein L26</td><td></td><td align="center">A1716351</td><td align="left">Induced in late G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Itm1</td><td></td><td align="center">A1956728 (h)</td><td align="left">Induced in G1/S (2-fold induction)</td></tr><tr><td align="left">&#x02003;Unknown T22</td><td></td><td align="center">A1053031 (h)</td><td align="left">6 kb, 2-fold induction in G1/S</td></tr><tr><td align="left">&#x02003;Unknown T34</td><td></td><td align="center">A1235326</td><td align="left">2 kb, 2-fold induction in G1/S</td></tr><tr><td align="left">Subtractive suppressive hybridization: G1-GA</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Galectin-8</td><td></td><td align="center">U09824</td><td align="left">Induced in G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Hsp86</td><td></td><td align="center">X16857</td><td align="left">Induced in late G1 (&#x0003e;3-fold induction)</td></tr><tr><td align="left">&#x02003;TCP-1 &#x003b7;</td><td></td><td align="center">AA900460</td><td align="left">Induced in late G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Ribosomal protein L13A</td><td></td><td align="center">X68282</td><td align="left">Induced in late G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Ribosomal protein L12</td><td></td><td align="center">AA900142</td><td align="left">Induced in late G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Ribosomal protein L3</td><td></td><td align="center">A1687295</td><td align="left">Induced in G1 (2-fold induction)</td></tr><tr><td></td><td></td><td align="center">Y00441</td><td align="left">Induced in G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;&#x003b2;₂-microglobulin</td><td></td><td align="center">D12771</td><td align="left">Induced from late G1 to G2 (&#x0003e;5 fold induction)</td></tr><tr><td align="left">&#x02003;Adenine nucleotide translocator ANT-2</td><td></td><td align="center">NM011342 (m)</td><td align="left">Induced in G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;Sec-22</td><td></td><td align="center">NM012562</td><td align="left">Induced in G1 (2-fold induction)</td></tr><tr><td align="left">&#x02003;L-fucosidase</td><td></td><td align="center">U91538</td><td align="left">Induced in G1/S (&#x0003e;3-fold induction)</td></tr><tr><td align="left">&#x02003;CRM-1 homolog/exportin 1</td><td></td><td align="center">X81839</td><td align="left">Induced in G1/S (&#x0003e;3-fold induction)</td></tr><tr><td align="left">&#x02003;Ubiquitin/ribosomal protein S27a</td><td></td><td align="center">A47416</td><td align="left">Induced in G1/S (&#x0003e;3-fold induction)</td></tr><tr><td align="left">&#x02003;Ubiquitin/ribosomal protein S30 (FAU)</td><td></td><td align="center">AF195142 (m)</td><td align="left">3-fold induction in G1 (1.5 kb)</td></tr><tr><td align="left">&#x02003;Unknown 4-2 (mouse selenoprotein R mRNA)</td><td></td><td align="center">H35219</td><td align="left">2-fold induction in G1/S (2 and 4 kb)</td></tr><tr><td align="left">&#x02003;Unknown 4-4 (human KIAA0081)</td><td></td><td align="center">No EST</td><td align="left">2-fold induction in G1 (1 kb)</td></tr><tr><td align="left">&#x02003;Unknown 4-11</td><td></td><td align="center">AU035826</td><td align="left">4-fold induction in G1 (1 and 1.5 kb)</td></tr><tr><td align="left">&#x02003;Unknown 4-15</td><td></td><td align="center">AF046001</td><td align="left">4-fold induction in G1 (2.5 kb)</td></tr><tr><td align="left">&#x02003;Unknown 4-16 (human ZNF207 or mouse Zep)</td><td></td><td align="center">No EST</td><td align="left">3-fold induction in G1 (1.5 kb)</td></tr><tr><td align="left">&#x02003;Unknown 4-20</td><td></td><td align="center">AW435432</td><td align="left">2-fold induction in G1</td></tr><tr><td align="left">&#x02003;Unknown 4-27 (new ribosomal protein L15 type)</td><td></td><td align="center">A1121996 (m)</td><td align="left">2-fold induction in G1</td></tr><tr><td align="left">&#x02003;Unknown 4-49</td><td></td><td align="center">AW246248 (h)</td><td align="left">2-fold induction in G1</td></tr><tr><td align="left">&#x02003;Unknown 4-58 (SH3, Rab GAP, TBC domain)</td><td></td><td align="center">No EST</td><td align="left">2-fold induction in G1</td></tr><tr><td align="left">&#x02003;Unknown 4-59</td><td></td><td></td><td></td></tr><tr><td align="left">Subtractive suppressive hybridization: GA-G1</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Spermidine/spermine N-acetyl transferase (SSAT)</td><td></td><td align="center">AA955996</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Leukocyte common antigen (alternative splicing) CD45</td><td></td><td align="center">Y00065</td><td align="left">Switch between two transcripts (one</td></tr><tr><td></td><td></td><td></td><td align="left">repressed, the other induced in G1)</td></tr><tr><td align="left">&#x02003;ZFX</td><td></td><td align="center">X75171</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Ribosomal protein S8</td><td></td><td align="center">AA874997</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Ribosomal protein S13</td><td></td><td align="center">L01123</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Unknown 6-2</td><td></td><td align="center">No EST</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Unknown 6-3 (hypothetical protein expressed in thymocytes)</td><td></td><td align="center">AJ237585 (m)</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Unknown 6-4</td><td></td><td align="center">No EST</td><td align="left">Repressed in G1</td></tr><tr><td align="left">&#x02003;Unknown 6-9</td><td></td><td align="center">No EST</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Unknown 6-10</td><td></td><td align="center">No EST</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Unknown 6-12 (human protein KIAA0710)</td><td></td><td align="center">AB014610 (h)</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">&#x02003;Unknown 6-45 (homolog to mouse PARP-2)</td><td></td><td align="center">NM009632 (m)</td><td align="left">Repressed transiently in G1</td></tr><tr><td align="left">Differential screening of a rat organized library: G2-GA</td><td></td><td></td><td></td></tr><tr><td align="left">&#x02003;Prothymosin &#x003b1;</td><td></td><td align="center">M86564</td><td align="left">Induced in G2</td></tr><tr><td align="left">&#x02003;Cyclophilin</td><td></td><td align="center">M19533</td><td align="left">Induced in G2</td></tr><tr><td align="left">&#x02003;ATP synthase &#x003b2; subunit</td><td></td><td align="center">M19044</td><td align="left">Induced in G1/S and G2</td></tr><tr><td align="left">&#x02003;Tubulin &#x003b1;2</td><td></td><td align="center">AA686718</td><td align="left">Induced in G1/S and G2</td></tr><tr><td align="left">&#x02003;GaPDH</td><td></td><td align="center">X02231</td><td align="left">Induced in G2</td></tr><tr><td align="left">&#x02003;Phosphoglycerate kinase</td><td></td><td align="center">M31788</td><td align="left">Induced in G1/S and G2</td></tr><tr><td align="left">&#x02003;MRG1 related protein</td><td></td><td align="center">U65093</td><td align="left">Induced in G1/S and G2</td></tr><tr><td align="left">&#x02003;Unknown BO1</td><td></td><td align="center">No EST</td><td align="left">Induced in G2</td></tr><tr><td align="left" colspan="4"><bold>Analysis of weakly expressed candidate genes: UN, A, GI, G1/S, G2</bold></td></tr><tr><td align="left">Name</td><td align="center">Accession number</td><td align="center">Fold induction</td><td align="left">Kinetics</td></tr><tr><td colspan="4"><hr></hr></td></tr><tr><td align="left">Ganglioside synthase (GD3)</td><td align="center">D84068</td><td align="center">&#x0003e; 4</td><td align="left" rowspan="3"><graphic xlink:href="gb-2000-1-4-research0008-I1.jpg"/></td></tr><tr><td align="left">P13 kinase</td><td align="center">D64045</td><td align="center">&#x0003e; 3</td></tr><tr><td align="left">Phospholipase C&#x003b3;1</td><td align="center">M34667</td><td align="center">&#x0003e; 3</td></tr><tr><td align="left">Bax</td><td align="center">S76511</td><td align="center">&#x0003e; 2</td><td align="left" rowspan="4"><graphic xlink:href="gb-2000-1-4-research0008-I2.jpg"/></td></tr><tr><td align="left">P53</td><td align="center">X13058</td><td align="center">&#x0003e; 2</td></tr><tr><td align="left">FAKp125</td><td align="center">AF020777</td><td align="center">&#x0003e; 3</td></tr><tr><td align="left">14-3-3 &#x003b5;</td><td align="center">M84416</td><td align="center">&#x0003e; 2</td></tr><tr><td align="left">14-3-3 &#x003b7;</td><td align="center">D17445</td><td align="center">&#x0003e; 2</td><td align="left" rowspan="2"><graphic xlink:href="gb-2000-1-4-research0008-I3.jpg"/></td></tr><tr><td align="left">Vitamin D3 receptor</td><td align="center">J09838</td><td align="center">3</td></tr><tr><td align="left">Glycine transporter</td><td align="center">M88595</td><td align="center">3</td><td align="left"><graphic xlink:href="gb-2000-1-4-research0008-I4.jpg"/></td></tr><tr><td align="left">Thromboxane A2 receptor</td><td align="center">D32080</td><td align="center">2</td><td align="left"><graphic xlink:href="gb-2000-1-4-research0008-I5.jpg"/></td></tr><tr><td align="left">Phosphatidylinositol transfer protein</td><td align="center">D17445</td><td align="center">2</td><td align="left"><graphic xlink:href="gb-2000-1-4-research0008-I6.jpg"/></td></tr><tr><td align="left">RexB/NSP</td><td align="center">U17604</td><td align="center">&#x0003e; 2</td><td align="left"><graphic xlink:href="gb-2000-1-4-research0008-I7.jpg"/></td></tr><tr><td align="left">Glucocorticoid receptor</td><td align="center">M14053</td><td align="center">2</td><td align="left"><graphic xlink:href="gb-2000-1-4-research0008-I8.jpg"/></td></tr><tr><td align="left">Ga3PDH</td><td align="center">J04147</td><td align="center">2</td><td align="left"><graphic xlink:href="gb-2000-1-4-research0008-I9.jpg"/></td></tr><tr><td align="left">Zif268 = EGR1</td><td align="center">U75398</td><td align="center">2</td><td align="left"><graphic xlink:href="gb-2000-1-4-research0008-I10.jpg"/></td></tr></tbody></table><table-wrap-foot><p>Nb2 cells: UN, unsynchronized; GA, growth arrested; G1, G1 phase; G1/S, G1/S transition; G2, G2 phase. PC12, rat pheochromocytoma PC12 cells. 18S and 28 S, ribosomal RNA. Whenever possible, rat accession numbers (GenBank) are written in the second column. When this sequence is not in known for the rat, (h) or (m) indicates that the accession number corresponds, respectively, to a human and a mouse cDNA homologous to those of the rat. `No EST' means that our rat sequence does not correspond to any previously described EST in mammals. For some of the unknown cDNAs, an estimation of the size of the corresponding transcript(s) as well as the fold induction is given in the right column `Expression variations', and for weakly expressed candidate genes, the expression kinetics are shown.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Functional classification of cell-cycle-regulated transcripts found in Nb2 cells</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Category</td><td align="left">Transcript</td></tr></thead><tbody><tr><td align="left">Cell cycle: cyclins and cell-cycle regulators</td><td></td></tr><tr><td align="left">&#x02003;Transcripts that have a cell-cycle-modulated homolog in yeast</td><td align="left">Cyclin E1 [<xref ref-type="bibr" rid="B78">78</xref>], peaks in G1</td></tr><tr><td></td><td align="left">EGR-1, immediate-early gene (constitutive expression in Nb2 cells)</td></tr><tr><td></td><td align="left">Cdc5-like protein [<xref ref-type="bibr" rid="B79">79</xref>], peaks in M</td></tr><tr><td></td><td align="left">Cdk2, Cdk5 [<xref ref-type="bibr" rid="B78">78</xref>], peak in G1</td></tr><tr><td align="left">&#x02003;Cell-cycle-modulated transcripts with a yeast homolog (not modulated)</td><td align="left">Cyclin B1, peaks in G2/M</td></tr><tr><td align="left">&#x02003;Cell-cycle-modulated transcripts with no yeast homolog</td><td align="left">Cyclin B2, peaks in G2</td></tr><tr><td></td><td align="left">Cyclin D2 [<xref ref-type="bibr" rid="B8">8</xref>], peaks in G1</td></tr><tr><td></td><td align="left">Cyclin D3 [<xref ref-type="bibr" rid="B8">8</xref>], peaks in G1</td></tr><tr><td align="left">Nucleotide metabolism, DNA replication and repair</td><td align="left">Cdc21 homolog</td></tr><tr><td></td><td align="left">Spermidine/spermine N-acetyl transferase (SSAT)</td></tr><tr><td></td><td align="left">Ornithine decarboxylase (ODC) [<xref ref-type="bibr" rid="B80">80</xref>]</td></tr><tr><td></td><td align="left"><italic>S</italic>-adenosylmethionine decarboxylase [<xref ref-type="bibr" rid="B80">80</xref>]</td></tr><tr><td></td><td align="left">Prothymosin &#x003b1;</td></tr><tr><td></td><td align="left">PARP-2 (Unknown 6-45)</td></tr><tr><td align="left">Chromatin structure</td><td align="left">Histones H2A, H2B, peak in S in mammals and in yeast</td></tr><tr><td align="left">Cytoskeleton</td><td align="left">Myosin heavy chain</td></tr><tr><td></td><td align="left">Tubulin &#x003b1;, &#x003b2;</td></tr><tr><td></td><td align="left">&#x003b2;-actin [<xref ref-type="bibr" rid="B6">6</xref>]</td></tr><tr><td></td><td align="left">Clone 15 = rNUDC [<xref ref-type="bibr" rid="B81">81</xref>]</td></tr><tr><td></td><td align="left">FAK p125</td></tr><tr><td align="left">Cell-surface antigens, adhesion molecules and signaling molecules</td><td></td></tr><tr><td align="left">(involved in apoptosis, survival and/or proliferation)</td><td></td></tr><tr><td align="left">&#x02003;Growth factor</td><td align="left">FGF-2 [<xref ref-type="bibr" rid="B82">82</xref>]</td></tr><tr><td align="left">&#x02003;Surface molecules</td><td align="left">Prolactin receptor Nb2 form [<xref ref-type="bibr" rid="B16">16</xref>]</td></tr><tr><td></td><td align="left">T-cell receptor &#x003b3; chain [<xref ref-type="bibr" rid="B17">17</xref>]</td></tr><tr><td></td><td align="left">T-cell receptor &#x003b1; chain [<xref ref-type="bibr" rid="B17">17</xref>]</td></tr><tr><td></td><td align="left">GnRH receptor [<xref ref-type="bibr" rid="B83">83</xref>]</td></tr><tr><td></td><td align="left">Glucocorticoid receptor</td></tr><tr><td></td><td align="left">Galectin-8</td></tr><tr><td></td><td align="left">Leukocyte membrane glycoprotein, CD45</td></tr><tr><td></td><td align="left">Vitamin D3 receptor</td></tr><tr><td></td><td align="left">Thromboxane A2 receptor</td></tr><tr><td></td><td align="left">&#x003b2;₂-microglobulin</td></tr><tr><td align="left">Cytoplasmic and/or nuclear signaling molecules</td><td align="left">p38 Map kinase [<xref ref-type="bibr" rid="B21">21</xref>]</td></tr><tr><td></td><td align="left">Pim-1 [<xref ref-type="bibr" rid="B9">9</xref>]</td></tr><tr><td></td><td align="left">Gfi-1 [<xref ref-type="bibr" rid="B10">10</xref>]</td></tr><tr><td></td><td align="left">Stathmin [<xref ref-type="bibr" rid="B84">84</xref>]</td></tr><tr><td></td><td align="left">P13 kinase p110 &#x003b1;</td></tr><tr><td></td><td align="left">Phospholipase C&#x003b3;1</td></tr><tr><td></td><td align="left">14-3-3 &#x003b7; and &#x003b5;</td></tr><tr><td></td><td align="left">Bax</td></tr><tr><td></td><td align="left">p53</td></tr><tr><td></td><td align="left">RexB/NSP</td></tr><tr><td></td><td align="left">Phosphatidylinositol transfer protein</td></tr><tr><td></td><td align="left">&#x003b1; 4 phosphoprotein [<xref ref-type="bibr" rid="B79">79</xref>]</td></tr><tr><td align="left">Transcription factors</td><td align="left">IRF-1 [<xref ref-type="bibr" rid="B7">7</xref>]</td></tr><tr><td></td><td align="left">c-Myc [<xref ref-type="bibr" rid="B18">18</xref>]</td></tr><tr><td></td><td align="left">c-Fos (not in Nb2 cells [<xref ref-type="bibr" rid="B18">18</xref>])</td></tr><tr><td></td><td align="left">MRG1-related protein</td></tr><tr><td></td><td align="left">E2F-1 [<xref ref-type="bibr" rid="B78">78</xref>]</td></tr><tr><td></td><td align="left">Zfx</td></tr><tr><td align="left">Heat shock, stress response and chaperones</td><td align="left">Cyclophilin (B)</td></tr><tr><td></td><td align="left">TCP-1 &#x003b5; and &#x003b7;</td></tr><tr><td></td><td align="left">Hsp70 (not expressed in Nb2 cells)</td></tr><tr><td></td><td align="left">Hsp70-like = Nb29 [<xref ref-type="bibr" rid="B41">41</xref>]</td></tr><tr><td></td><td align="left">Hsp27</td></tr><tr><td></td><td align="left">Hsp86</td></tr><tr><td></td><td align="left">&#x003b2;-actin</td></tr><tr><td></td><td align="left">&#x003b1;₂-tubulin</td></tr><tr><td></td><td align="left">Rdnuc (Golgi-associated protein) [<xref ref-type="bibr" rid="B11">11</xref>]</td></tr><tr><td></td><td align="left">Myosin heavy chain</td></tr><tr><td></td><td align="left">Focal adhesion kinase (FAK)</td></tr><tr><td align="left">Metabolism (energy)</td><td align="left">Phosphoglycerate kinase</td></tr><tr><td></td><td align="left">Enolase &#x003b1;</td></tr><tr><td></td><td align="left">Aldehyde dehydrogenase</td></tr><tr><td></td><td align="left">ATP synthase &#x003b2; subunit</td></tr><tr><td align="left">Protein and RNA synthesis, modifications and degradation</td><td></td></tr><tr><td align="left">&#x02003;Ribosomal proteins</td><td align="left">L3, L12, L13A, new ribosomal protein L15 type (unknown 4-27)</td></tr><tr><td></td><td align="left">S8, S13</td></tr><tr><td align="left">&#x02003;Glycosylation factor</td><td align="left">Itm1</td></tr><tr><td align="left">&#x02003;Elongation factor</td><td align="left">EF-2 [<xref ref-type="bibr" rid="B77">77</xref>]</td></tr><tr><td align="left">Inter-compartment transport and trafficking</td><td align="left">CRM-1 = exportin 1</td></tr><tr><td></td><td align="left">Sec-22</td></tr><tr><td></td><td align="left">Glycine transporter</td></tr><tr><td></td><td align="left">Unknown 4-58 (SH3, Rab GAP, TBC domain: putative nuclear pore protein)</td></tr><tr><td align="left">Unknown function(s)</td><td align="left">FGF-responsive Non/p54nrb</td></tr><tr><td></td><td align="left">Unknowns T22, T34, 4-2, 4-4, 4-11, 4-15, 4-20, 4-49, 4-59, BO1</td></tr><tr><td></td><td align="left">Unknowns 6-2, 6-3, 6-4, 6-9, 6-10, 6-12</td></tr></tbody></table></table-wrap></sec></sec><sec><title>Discussion</title><sec><title>New putative signaling molecules in Nb2 cells</title><p>In this study we have identified new regulated genes encoding potential signaling molecules. Genes encoding proteins involved in cell survival, apoptosis, proliferation and/or cell-cycle progression may or may not have cell-cycle-dependent expression. Most of these proteins (listed in Tables <xref ref-type="table" rid="T1">1</xref>,<xref ref-type="table" rid="T2">2</xref>) are known to be activated by post-translational mechanisms; little is known, however, about their regulation at the transcriptional level.</p><p>In our experiments, for example, PI 3-kinase, PLC&#x003b3;1 and ganglioside synthase GD3 share identical expression profiles, characterized by lower expression in unsynchronized than in synchronized Nb2 cells. This indicates that they may share similar regulation mechanisms. In high-density Nb2 cell cultures [<xref ref-type="bibr" rid="B48">48</xref>], secreted growth factors maybe involved in this negative regulation. PI 3-kinase and/or PLC&#x003b3;1 have been implicated in cell-cycle progression, proliferation, survival, transformation and apoptosis in different cellular models [<xref ref-type="bibr" rid="B49">49</xref>,<xref ref-type="bibr" rid="B50">50</xref>,<xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref>]. Thus, ganglioside synthase may also take part in similar processes in immune cells. Indeed, ganglioside synthase GD3 is highly expressed in various human cancer cell lines, is upregulated in activated T lymphocytes [<xref ref-type="bibr" rid="B53">53</xref>], and has been implicated in Fas-mediated apoptosis [<xref ref-type="bibr" rid="B23">23</xref>]. It may therefore be of interest to determine whether prolactin is able to activate, directly or indirectly, the activity of GD3 and the Fas signaling pathways.</p><p>Expression profiles of genes for Bax, p53, 14-3-3 &#x003b5; and FAK are characterized by increased mRNA expression during the G1, G1/S and G2 phases in comparison to the growth-arrested or unsynchronized Nb2 cells (Table <xref ref-type="table" rid="T1">1</xref>). These kinetics suggest that prolactin may have a direct effect on the transcription of these genes. Indeed, in myeloid cells [<xref ref-type="bibr" rid="B54">54</xref>] and in proliferating prostate cells [<xref ref-type="bibr" rid="B55">55</xref>], FAK expression is induced by various cytokines. This molecule is located at the signaling crossroads of cell growth and attachment, and is involved in dynamic cytoskeletal rearrangements [<xref ref-type="bibr" rid="B56">56</xref>]. In Nb2 cells, prolactin has been shown to increase <italic>bax</italic> mRNA expression in 8 hours [<xref ref-type="bibr" rid="B20">20</xref>]. The thromboxane A2 receptor, which is highly expressed in immature thymocytes, has also been shown to mediate DNA fragmentation and apoptosis [<xref ref-type="bibr" rid="B57">57</xref>]. It is possible that, in Nb2 cells, prolactin could also counteract thromboxane-induced apoptosis, as is the case for glucocorticoids.</p><p>It is not known at present whether these expression profiles are common to all dividing mammalian cells or only to a particular subclass of immune-system cells. Alternatively, these profiles could be the consequence of the genetic abnormalities displayed by Nb2 cells. It is of interest that the expression of the p38 MAP kinase gene is modulated in Nb2 cells [<xref ref-type="bibr" rid="B21">21</xref>] but not in normal human fibroblasts [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>], suggesting that this regulation is specific to the T-cell lineage. Different arguments exist in favor of the existence of cross-talk between the JAK/STAT and p38 MAP kinase pathways, at both the translational and transcriptional levels. As the expression of ganglioside synthase GD3 is restricted to the brain and the hematopoietic lineage, the regulation of the transcript and the involvement of the protein in the regulation of survival and apoptosis may also be shared by these tissues. In contrast, the cell-cycle-dependent expression of Ant-2, observed in both Nb2 cells (our results) and human fibroblasts [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>], suggests that this feature is common to all dividing mammalian cells. These speculations require further experimentation at the protein level for confirmation.</p></sec><sec><title>Expression abnormalities</title><p>We report for the first time, to our knowledge, the constitutive expression of the transcription factor EGR-1 in synchronized proliferating cells. EGR-1 has a role in differentiation and development, in normal growth and in virus-induced growth and immortalization (reviewed in [<xref ref-type="bibr" rid="B58">58</xref>]). Many of these effects may be related to complex cooperative and competitive mechanisms between the three transcription factors Sp1, EGR-1 and Wt1, which often have overlapping binding sites in target promoters.</p><p>Several arguments suggest that EGR-1 may act as a tumor suppressor [<xref ref-type="bibr" rid="B59">59</xref>,<xref ref-type="bibr" rid="B60">60</xref>,<xref ref-type="bibr" rid="B61">61</xref>], and that its anti-oncogenic function could be due to the transcriptional induction of the gene for transforming growth factor &#x003b2;1 (TGF&#x003b2;1), which suppresses growth by an autocrine mechanism in the late G1 phase of the cell cycle [<xref ref-type="bibr" rid="B62">62</xref>,<xref ref-type="bibr" rid="B63">63</xref>]. Exogenous TGF&#x003b2; inhibits Nb2 cell growth [<xref ref-type="bibr" rid="B64">64</xref>], suggesting that these cells are still sensitive to the anti-proliferative action of TGF&#x003b2;, but are unable to synthesize or activate TGF&#x003b2; on their own, despite their constitutive EGR-1 expression. It is possible that, in Nb2 cells, the anti-proliferative effect induced by the constitutive expression of EGR-1 is suppressed by other abnormalities such as a deficiency in the production of active TGF&#x003b2;.</p><p>Other studies, however, argue in favor of the existence of anti-apoptotic and/or pro-proliferative properties of EGR-1 [<xref ref-type="bibr" rid="B65">65</xref>,<xref ref-type="bibr" rid="B66">66</xref>,<xref ref-type="bibr" rid="B67">67</xref>,<xref ref-type="bibr" rid="B68">68</xref>]. In this context, the constitutive expression of EGR-1 in Nb2 cells suggests that this transcription factor may have both anti- and pro-proliferative effects, as previously described for other proteins such as p53 [<xref ref-type="bibr" rid="B69">69</xref>]. These dual and antagonistic functions may constitute a protective mechanism against tumor formation. In this model, three oncogenic abnormalities would have to occur in order to generate continuous tumor growth: immortalization; activation of all the transduction pathways required for proliferation; and suppression of all the anti-proliferative and apoptotic properties resulting from proto/anti-oncogene modifications.</p><p>Further studies are required to confirm the integrity of the EGR-1 protein and its constitutive expression in Nb2 cells and to understand the relationships between the EGR-1 target genes and their signaling pathways.</p></sec><sec><title>Comparison of various differential screening techniques</title><p>We have compared the advantages and drawbacks of four differential screening techniques. Of the four approaches, differential display analysis presents several advantages: it is easy, rapid, does not require large amounts of biological material, and it allows the comparison of multiple transcriptomes in a single experiment. The occurrence of false-positive clones is, however, quite high. In our experimental conditions, we isolated 20 potential positive cDNAs; only two, however, presented a relatively distinct cell-cycle-modulated expression profile. The problem with this technique lies in the low reproducibility of the PCR reaction and the occurrence of non-differential PCR products, which are recovered together with differentially expressed transcripts from the acrylamide gel. Several methods (such as reverse northern blots) have been proposed to circumvent this problem, albeit with rather limited success. Another disadvantage is that the identification of cDNA clones may be difficult if the model system studied has not previously been used in an extended EST sequencing program.</p><p>RDA and SSH are based on the same principle. These techniques are easy to use and enable the rapid generation of RDA or SSH subtractive libraries. The proportion of false-positive cDNAs can be less than 10% (our data), but if the differences between the two libraries are discrete, this number is increased. RDA or SSH cDNA clones correspond to sequences positioned in the middle portions of transcripts (generally coding regions), and the sequencing of each cDNA clone enables their identification independently of the model used. These techniques have two principal disadvantages: only two different transcriptomes can be compared in one experiment, and these approaches are far from being exhaustive. Although they should facilitate the detection of low-level transcripts, this is often not the case, as non-optimal conditions appear preferentially to select cDNAs corresponding to highly expressed transcripts.</p><p>The screening of an organized library can be compared with the use of DNA arrays, and the detection of a wide range of differential transcripts. This approach theoretically allows the screening of transcripts corresponding to unknown genes. False-positive cDNAs are relatively rare (10-20%) if the amount of DNA fixed on the high-density nylon filters is strictly controlled. The detection threshold for these techniques, however, does not allow the detection of weakly expressed differential transcripts and remains a major limitation.</p><p>The principle of the analysis of weakly expressed candidate genes is derived from that of macroarrays. In that case, the detection threshold is increased by a step of moderate PCR amplification for each candidate gene in the complex probes, but allows only a semi-quantitative detection of differentially expressed candidate genes. Internal controls are included to monitor the efficiency of the technique.</p></sec><sec><title>Cell-cycle-regulated transcripts in mammals</title><p>To facilitate the comparative analysis of cell-cycle-regulated transcripts, we classified them into ten different functional categories (Table <xref ref-type="table" rid="T2">2</xref>). This classification is in agreement with the target genes of growth-related transcription factors. Indeed, putative c-Myc target genes are involved in the cell cycle, apoptosis, DNA metabolism and dynamics, energy metabolism and macromolecular synthesis [<xref ref-type="bibr" rid="B70">70</xref>]. Nevertheless, as transcriptional gene activation or inhibition result from a complex multifactorial <italic>cis</italic> and <italic>trans</italic> regulation, it is necessary to integrate the various components of this regulation and of post-transcriptional modifications to explain the origin of differential expression.</p><p>The current functional classification is too restrictive and does not take into account the interactions between these functional categories. For example, the regulation of proto-oncogene and cytokine mRNAs and proteins is particularly complex. The regulatory processes involved include transcriptional control, nuclear export and import of transcripts and proteins, translation, heat-shock pathways (Hsc70-Hsp70), and the ubiquitin- and proteasome-mediated degradation of mRNA and proteins [<xref ref-type="bibr" rid="B71">71</xref>]. Interestingly, transcripts of many proteins involved in these processes (such as Hsp proteins, translation factors, and CRM-1) were found to be induced by prolactin in Nb2 cells in this study, as well as in other models. Moreover, the abnormal heat-shock response described in Nb2 cells may perturb this regulation and have important consequences for tumor progression.</p></sec><sec><title>Comparison of cell-cycle regulated transcripts in yeast and higher eukaryotes</title><p>Data are still lacking for an extensive comparison of cell-cycle-modulated transcripts in unicellular (yeast) and multicellular organisms (essentially mammals) and for deduction of evolutionarily conserved features. Several conclusions can be drawn, however. In almost all the functional subclasses detailed in Table <xref ref-type="table" rid="T2">2</xref>, homologous proteins are regulated at the transcriptional level in both yeast and mammals. For example, the yeast cyclins, Cdc proteins, histones and their mammalian homologs are usually regulated in the same fashion during the cell cycle. These observations have been well documented in numerous studies [<xref ref-type="bibr" rid="B72">72</xref>] and suggest that the functions of the corresponding proteins and the regulation of their transcripts are both conserved. A list of yeast cell-cycle-regulated transcripts that have known mammalian homologs has been established [<xref ref-type="bibr" rid="B73">73</xref>]. In this list, 26 out of 99 of the yeast cell-cycle-regulated transcripts correspond to proteins involved in nucleotide synthesis, DNA replication and repair. Within the cytoskeleton subclass, the transcripts for yeast MYO3 and its mammalian homolog, myosin I heavy chain, both peak at the G2/M phase of the cell cycle, whereas TUB3 and mammalian tubulin &#x003b3; peak in G1. Proteins in several classes and subclasses do not have yeast homologs. For example, 'adhesion molecules, extracellular matrix and cellular surface antigens' are specific to multicellular organisms.</p><p>There are numerous cellular and physiological differences between yeast and multicellular organisms, and some of the molecular divergences observed may reflect these differences. In unicellular organisms the growth priorities are to proliferate as long as enough nutrients are available. In contrast, in multicellular organisms the integrity of the organism is paramount, and individual cell behavior is highly controlled. Therefore proliferation is basically prohibited, and occurs in a cell lineage only if the environment sends the appropriate combination of signals to unlock all the growth-inhibition mechanisms. Despite these differences, cell-cycle checkpoints located just before and at the end of mitosis are essential and similar in all organisms. These two controls ensure that all the DNA has been correctly replicated before the cell enters mitosis and that the condensed chromosomes are properly aligned on the division spindle before anaphase. Thus, all proteins potentially involved in such controls may be regulated similarly from yeasts to animals and plants.</p><p>The present study illustrates how data from various large-scale differential screening analyses can be integrated into specific and/or global biological studies. Such comparisons of gene expression profiles are of value in understanding general expression profiles of all dividing cells and in analyzing differences between unicellular and multicellular organisms. They can also identify new signaling molecules and explain how different signals and transduction pathways could regulate the proliferation of different cell types. They can help elucidate the function of proteins, and finally they can identify abnormal patterns of gene expression in transformed and tumor cells. With the increasing amount of information being generated from microarrays and DNA chips, the potential value of comparative analyses will be all the greater.</p></sec></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Cell culture</title><p>Suspension cultures of lactogen-dependent Nb2-11C lymphoma cells were grown in RPMI 1640 medium containing 10% fetal calf serum (FCS), 10% lactogen-free horse serum, 0.1 mM &#x003b2;-mercaptoethanol, 2 mM L-glutamine, 5 mM HEPES pH 7.3, and penicillin-streptomycin (50 lU/ml and 50 &#x003bc;g/ml, respectively) at 37&#x000b0;C with 5% CO₂. Cultures were rendered quiescent by transferring cells into starvation medium (cell density about 1.5 &#x000d7; 10⁵ cells/ml), deficient in FCS and &#x003b2;-mercaptoethanol, for 24 hours. Under these conditions, about 80-85% of the cells were arrested in the G0/G1 phase (Figure <xref ref-type="fig" rid="F1">1a</xref>). Ovine prolactin (Sigma) was added at a concentration of 20 ng/ml to starvation medium to reinitiate growth (Figure <xref ref-type="fig" rid="F1">1b</xref>). Cells were collected after various periods of prolactin stimulation and total RNA was extracted. Unsynchronized Nb2 cells were collected from high-density cultures (10⁶ cells/ml).</p></sec><sec><title>RNA extraction, poly(A)⁺ RNA preparation, cDNA synthesis, northern blot and reverse northern blot analyses</title><p>For each cell population to be compared, RNA was prepared by acid guanidinium-thiocyanate-phenol-chloroform extraction [<xref ref-type="bibr" rid="B74">74</xref>] and poly(A)⁺ RNA was isolated using magnetic oligodT (Dynabeads, Dynal). Double-stranded cDNA was transcribed using a commercial kit (Boehringer). Northern blots were performed using the formaldehyde/formamide procedure and reverse northern blots as described in [<xref ref-type="bibr" rid="B75">75</xref>]. The Vacugene transfer system (Pharmacia), nylon filters (Hybond N⁺, Amersham) and the hybridization solution (ExpressHyb, Clontech) were used following the manufacturer's instructions. The cDNAs were radiolabeled using the Readyprime kit (Amersham).</p></sec><sec><title>mRNA differential display, representational difference analysis (RDA), subtractive suppressive hybridization (SSH) and organized library screening</title><p>Total RNA treated with DNase I was prepared from Nb2 cells treated for 0, 2, 4, 6, 8, 12 or 24 h with prolactin, and used in mRNA differential display using the GenHunter kit (GeneHunter Corporation, TN, USA). Poly(A)⁺ mRNA extracted from Nb2 cells treated with prolactin for 12 h was used for RDA according to the protocol described in [<xref ref-type="bibr" rid="B13">13</xref>]. SSH [<xref ref-type="bibr" rid="B14">14</xref>] was performed using poly(A)⁺ RNA obtained by mixing mRNA from Nb2 cells stimulated with ovine prolactin for 2, 4, 6 and 8 h. An organized rat brain library was screened as described in [<xref ref-type="bibr" rid="B15">15</xref>].</p></sec><sec><title>Preparation of complex probes, multiplex PCR and differential screening of candidate genes</title><p>In this technique, nylon filters dotted with candidate genes are hybridized with complex cDNA probes from different cell populations, to compare expression levels (see, for example [<xref ref-type="bibr" rid="B76">76</xref>]). Originally developed and validated by SANOFI-Recherche (unpublished data), the nylon filters contain 91 rat candidate genes (PCR products), including signaling molecules and transcription factors (Figure <xref ref-type="fig" rid="F4">4</xref>). These cDNAs were selected with the idea of defining a panel of genes whose expression is likely to be modulated in response to a proliferative signal. To increase the sensitivity of the approach, the mRNAs of interest are co-amplified by reverse transcription polymerase chain reaction (RT-PCR), using primers specific for the 91 candidate genes (the multiplex PCR step), before their use as hybridization probes (Figure <xref ref-type="fig" rid="F3">3a</xref>). Under these conditions, moderate PCR amplification allows the detection of weakly expressed genes and the evaluation of their differential expression in a semi-quantitative manner. mRNAs are amplified by RT-PCR, and the number of PCR cycles is adapted to their relative abundance in the population tested (16, 21, 24 or 26 cycles are performed for the detection of relatively abundant, moderately expressed, weakly expressed and very weakly expressed transcripts, respectively). For each cell population analyzed, the multiplex PCR products are then mixed, radiolabeled by random priming with &#x003b1;-³²P-labeled dCTP, and used as hybridization probes against the 91 candidate genes. Hybridization signals are quantified by densitometry (Visiomic) for each of the different populations. During the development of this technique, control studies involving repeated hybridizations with replicate filters showed minimum variation of signal response (data not shown).</p><p>The efficiency of the technique was controlled by including rabbit &#x003b1; and &#x003b2; globin cDNAs on the nylon membranes along with the candidate genes. Defined amounts (50 or 150 ng) of the globin cDNAs were also added to the cDNAs mixtures before the multiplex PCR step. As expected, when these different amounts of rabbit globin cDNAs were added to the complex cDNAs, spots of different intensities were obtained (Figure <xref ref-type="fig" rid="F3">3b</xref>), indicating that this approach could detect at least a threefold difference in mRNA expression between two cell populations.</p><p>Figure <xref ref-type="fig" rid="F4">4</xref> shows the list of the candidate genes analyzed and indicates their expression pattern in Nb2 cells.</p></sec><sec><title>Sequencing</title><p>The potential positive clones isolated by differential display, RDA, SSH or screening of the organized library were sequenced with a dye terminator kit using the ABI Prism system (Perkin-Elmer).</p></sec><sec><title>Bioinformatics</title><p>To identify the sequenced cDNAs, BLAST and UniGene from NCBI were used [<xref ref-type="bibr" rid="B77">77</xref>]. For comparison, we have also consulted databases of transcripts differentially expressed during cell-cycle progression in human fibroblasts [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>] and in <italic>S. cerevisiae</italic> [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>].</p></sec></sec>

Towards understanding the first genome sequence of a crenarchaeon by genome annotation using clusters of orthologous groups of proteins (COGs)

<sec><title>Background:</title><p>Standard archival sequence databases have not been designed as tools for genome annotation and are far from being optimal for this purpose. We used the database of Clusters of Orthologous Groups of proteins (COGs) to reannotate the genomes of two archaea, <italic>Aeropyrum pernix</italic>, the first member of the Crenarchaea to be sequenced, and <italic>Pyrococcus abyssi</italic>.</p></sec><sec><title>Results:</title><p><italic>A. pernix</italic> and <italic>P. abyssi</italic> proteins were assigned to COGs using the COGNITOR program; the results were verified on a case-by-case basis and augmented by additional database searches using the PSI-BLAST and TBLASTN programs. Functions were predicted for over 300 proteins from <italic>A. pernix</italic>, which could not be assigned a function using conventional methods with a conservative sequence similarity threshold, an approximately 50% increase compared to the original annotation. <italic>A. pernix</italic> shares most of the conserved core of proteins that were previously identified in the Euryarchaeota. Cluster analysis or distance matrix tree construction based on the co-occurrence of genomes in COGs showed that <italic>A. pernix</italic> forms a distinct group within the archaea, although grouping with the two species of Pyrococci, indicative of similar repertoires of conserved genes, was observed. No indication of a specific relationship between Crenarchaeota and eukaryotes was obtained in these analyses. Several proteins that are conserved in Euryarchaeota and most bacteria are unexpectedly missing in <italic>A. pernix</italic>, including the entire set of <italic>de novo</italic> purine biosynthesis enzymes, the GTPase FtsZ (a key component of the bacterial and euryarchaeal cell-division machinery), and the tRNA-specific pseudouridine synthase, previously considered universal. <italic>A. pernix</italic> is represented in 48 COGs that do not contain any euryarchaeal members. Many of these proteins are TCA cycle and electron transport chain enzymes, reflecting the aerobic lifestyle of <italic>A. pernix.</italic></p></sec><sec><title>Conclusions:</title><p>Special-purpose databases organized on the basis of phylogenetic analysis and carefully curated with respect to known and predicted protein functions provide for a significant improvement in genome annotation. A differential genome display approach helps in a systematic investigation of common and distinct features of gene repertoires and in some cases reveals unexpected connections that may be indicative of functional similarities between phylogenetically distant organisms and of lateral gene exchange.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Natale</surname><given-names>Darren A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Shankavaram</surname><given-names>Uma T</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Galperin</surname><given-names>Michael Y</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Wolf</surname><given-names>Yuri I</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Aravind</surname><given-names>L</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Koonin</surname><given-names>Eugene V</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>Functional annotation of genomes is a critical aspect of the genomics enterprise. Without reliable assignment of gene function at the appropriate level of specificity, new genome sequences are plainly useless. The primary methodology used for genome annotation is the sequence database search, the results of which allow transfer of functional information from experimentally characterized genes (proteins) to their uncharacterized homologs in newly sequenced genomes [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. However, general-purpose, archival sequence databases are not particularly suited for the purpose of genome annotation. The quality of the annotation of a new genome produced using a particular database critically depends on the reliability and completeness of the annotations in the database itself. As far as annotation is concerned, the purpose of primary sequence databases is to faithfully preserve the description attached to each sequence by its submitter. In their capacity as sequence archives, such databases include no detailed documentation in support of the functional annotations. Furthermore, primary sequence databases are not explicitly structured by either evolutionary or functional criteria. These features, which are inevitable in archival databases, seriously impede their utility as resources for genome annotation, particularly when an automated or semi-automated approach is attempted [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>]. At its worst, this situation results in a notorious vicious circle of error amplification - an inadequately annotated database is used to produce an error-ridden and incomplete annotation of a new genome, which in turn makes the database even less useful [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>].</p><p>One way out of this 'Catch-22' situation is to use a different type of database for genome annotation, namely databases in which sequence information is organized by structural, functional or phylogenetic criteria, or a combination thereof. For example, the KEGG [<xref ref-type="bibr" rid="B9">9</xref>] and WIT [<xref ref-type="bibr" rid="B10">10</xref>] databases are primarily function-oriented and organize protein sequences from completely and partially sequenced genomes according to their known or predicted roles in biochemical pathways, although WIT also provides a phylogenetic classification. In contrast, the SMART database [<xref ref-type="bibr" rid="B11">11</xref>] is organized on a structural principle and provides a searchable collection of common protein domains. All these databases share a fundamental common feature - they encapsulate carefully verified knowledge on protein structure, function and/or evolutionary relationships, and therefore, at least in principle, provide for a more robust mode of genome annotation than general-purpose databases and may serve as a stronger foundation for partially automated approaches to genome analysis.</p><p>The database of Clusters of Orthologous Groups of proteins (COGs) is a phylogenetic classification of proteins encoded in completely sequenced genomes [<xref ref-type="bibr" rid="B12">12</xref>]. An attempt has been made to organize these proteins into groups of orthologs, direct evolutionary counterparts related by vertical descent [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. Because of lineage-specific duplications, orthologous relationships in many cases exist between gene (protein) families, rather than between individual proteins, hence 'orthologous groups' (including only lineage-specific duplications in a COG is the principle of this analysis; in practice, because of insufficient resolution of sequence comparisons, certain COGs may include ancestral duplications). The principal phylogenetic classification in the COG database is overlaid with functional classification and annotation based on detailed sequence and structure analysis and published experimental data. The COG system has been designed as a platform for evolutionary analyses and for phylogenetic and functional annotation of genomes. The COGNITOR program associated with the COGs allows one to fit new proteins into existing COGs. The central tenet of this analysis is that, if it can be shown that the protein under analysis is an ortholog of functionally characterized proteins from other genomes, this functional information can be transferred to the analyzed protein with considerable confidence. In addition to COGNITOR, the COG system includes certain higher-level functionalities, such as analysis of phylogenetic patterns and co-occurrence of genomes in COGs. The current (as of 1 June, 2000) system consists of 2,112 COGs that encompass about 27,000 proteins from 21 completely sequenced genomes [<xref ref-type="bibr" rid="B15">15</xref>].</p><p>Here we describe the application of the COGs to the systematic annotation and evolutionary analysis of two recently sequenced archaeal genomes, those of the euryarchaeon <italic>Pyrococcus abyssi</italic> [<xref ref-type="bibr" rid="B16">16</xref>] and the crenarchaeon <italic>Aeropyrum pernix</italic> [<xref ref-type="bibr" rid="B17">17</xref>]. These genomes were selected to compare the utility of the COGs for the annotation of two types of genomes - one that is closely related to another genome already included in the system, as <italic>Pyrococcus abyssi</italic> is to <italic>P. horikoshii</italic>, and one that represents a group previously not covered by the COGs, the Crenarchaeota. We show here the relatively low error rate of the COG-assisted analysis and its contribution to a significant number of new functional predictions. Emphasis is on using the COG approach to identify features of the <italic>A. pernix</italic> genome that are shared among all Archaea and those that distinguish Crenarchaeota from Euryarchaeota. Thus this work had a dual focus: first, to explore the potential of the COG system for genome annotation; and second, to use the COG approach to reveal important trends in archaeal genome evolution. It should not be construed as a comprehensive analysis of any particular genome or a comprehensive comparative and evolutionary study; addressing each of these tasks would require the use of several additional methodologies.</p></sec><sec><title>Results and discussion</title><sec><title>The protocol for genome annotation using the COG database</title><p>Figure <xref ref-type="fig" rid="F1">1</xref> depicts the steps of the procedure used for the COG-based genome annotation. This protocol is not limited to straightforward COGNITOR analysis but also takes advantage of the phylogenetic information encapsulated in the COGs, primarily in the form of phylogenetic patterns, which can be used to guide the search for missing COG members (described in detail in [<xref ref-type="bibr" rid="B18">18</xref>]). Briefly, whenever one of the analyzed genomes was unexpectedly not represented in a COG, additional analysis was undertaken to identify possible diverged members by using an iterative database search with the PSI-BLAST program, or to detect members that could have been missed in the original genome annotation by using translating searches with the TBLASTN program. In the present analysis of two archaeal genomes, such unexpected absences involved COGs represented in all or most of the other species or in all other archaea. Conversely, unexpected occurrences of the analyzed genomes in COGs, for example the first archaeal member of a purely bacterial COG, was examined case by case to detect likely horizontal gene transfer events and novel functions in archaeal genomes.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>A flow chart of the genome annotation process using COGs. NR is the Non-Redundant sequence database at the National Center for Biotechnology Information.</p></caption><graphic xlink:href="gb-2000-1-5-research0009-1"/></fig></sec><sec><title>Assessment of computational assignment of proteins to COGs</title><p>Proteins were assigned to COGs by two rounds of automated comparison using COGNITOR, each followed by manual checking of the assignments. The first round attempts to assign proteins to existing COGs; typically, &#x0003e;90% of the assignments are made in this step. The second round serves two purposes: first, to assign paralogs that could have been missed in the first round to existing COGs; and second, to create new COGs from those proteins that remained unassigned. With the goal of determining the optimal level of automation for such tasks, we assessed the performance of the automated procedure for annotating the <italic>A. pernix</italic> genome, which belongs to a major taxon, Crenarchaeota, that so far has not been represented in the COG database. For comparative purposes, the performance of the automated procedure for annotating proteins from <italic>Pyrococcus abyssi</italic> was also evaluated. <italic>P. abyssi</italic> is a member of the Euryarchaeota and is closely related to <italic>P. horikoshii</italic>, which is currently represented in the COGs. The data are shown in Table <xref ref-type="table" rid="T1">1</xref>. Three main classes of protein assignments are considered: true positives, false positives and false negatives.</p><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Assignment of proteins to COGs</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Category</td><td align="center"><italic>A. pernix</italic></td><td align="center"><italic>P. abyssi</italic></td></tr></thead><tbody><tr><td align="left">Proteins assigned by COGNITOR</td><td align="center">1,123</td><td align="center">1,421</td></tr><tr><td align="left">Proteins included in COGs^*</td><td align="center">1,102</td><td align="center">1,404</td></tr><tr><td align="left">True positives</td><td align="center">1,062</td><td align="center">1,381</td></tr><tr><td align="left">&#x02003;Pre-existing COGs</td><td align="center">1,011</td><td align="center">1,339</td></tr><tr><td align="left">&#x02003;New COGs</td><td align="center">27</td><td align="center">3</td></tr><tr><td align="left">&#x02003;Divided^&#x02020;</td><td align="center">24</td><td align="center">39</td></tr><tr><td align="left">False positives</td><td align="center">44</td><td align="center">31</td></tr><tr><td align="left">&#x02003;Not accepted</td><td align="center">21</td><td align="center">17</td></tr><tr><td align="left">&#x02003;Reassigned to a related COG</td><td align="center">21</td><td align="center">14</td></tr><tr><td align="left">&#x02003;Reassigned to an unrelated COG</td><td align="center">2</td><td align="center">0</td></tr><tr><td align="left">False negatives^&#x02021;</td><td align="center">17</td><td align="center">9</td></tr></tbody></table><table-wrap-foot><p>^*Includes true positives, reassigned false positives, and false negatives. ^&#x02020;Not included in 'To preexisting COG' or 'To new COG' numbers. ^&#x02021;Proteins added during manual checking.</p></table-wrap-foot></table-wrap><p>True positives are proteins that were correctly assigned either to an existing COG or to a COG that was created as a result of adding the new species. After a detailed examination of the COGNITOR results, 95% of the automatically assigned <italic>A. pernix</italic> proteins and 97% of the automatically assigned <italic>P. abyssi</italic> proteins were classified as true positives. As expected, the number of COGs created as a result of adding each species significantly differed. <italic>P. abyssi</italic>, which belongs to a previously represented clade, contributed only three new COGs, each representing a conserved family missing in <italic>P. horikoshii.</italic> In contrast, 27 new COGs were created as a result of adding <italic>A. pernix</italic> proteins.</p><p>False positives are proteins that were incorrectly assigned to a COG, and these fall into two classes. The first class are those proteins that needed to be removed altogether (that is, not included in any COG). In such cases, although the criterion that the query protein had at least three genome-specific best hits to members of the given COG was formally met, a detailed examination showed that these hits most probably arose by chance (see the Materials and methods section). The second class are those proteins that were assigned to one COG by COGNITOR, but subsequently moved to another COG. Most often (21 of 23 cases for <italic>A. pernix</italic>, and all of the cases for <italic>P. abyssi</italic>), the proteins were moved to a related COG (for example, between two COGs that include distinct, but related families of ATPases).</p><p>False negatives are proteins that were not assigned to any COG by COGNITOR because they failed the three-best-hit criterion (see the Materials and methods section), but were included subsequently as a result of additional sequence comparisons initiated upon examination of unexpected phylogenetic patterns. Again, the occurrence of false negatives for <italic>P. abyssi</italic> was about half that of <italic>A. pernix</italic> (1% versus 2%). Of the 17 such omissions identified among <italic>A. pernix</italic> proteins, 11 occurred as a result of pre-processing the protein sequences for low-complexity regions using the SEG program. When COGNITOR was run without filtering, these proteins were automatically assigned to the COGs. The remainder, including all false negatives seen for <italic>P. abyssi</italic>, failed the three-best-hit criterion because they showed only weak similarity to the members of the respective COG. However, given that they were the best candidates for filling unexpected gaps in phylogenetic patterns, and also because they contained the typical sequence motifs of the respective families, these proteins were included in the COGs.</p></sec><sec><title>Annotation of <italic>Aeropyrum pernix</italic> and <italic>Pyrococcus abyssii</italic> protein sets</title><p><italic>Aeropyrum pernix</italic> has been reported to encode 2,694 putative proteins in a 1.67 megabase (Mb) genome [<xref ref-type="bibr" rid="B17">17</xref>]. Of these, 633 proteins were assigned a function or partial characterization in the original report, on the basis of sequence comparison with proteins in the GenBank, SWISS-PROT, EMBL, PIR and Owl databases. Each of these databases contains individually annotated proteins. In contrast, the COG database annotates protein families rather than individual proteins, and the method used for the construction of the COGs often allows distant relationships to be discerned. Thus, use of the COG database for the annotation of a newly sequenced genome would probably increase the number of functional assignments. Indeed, we have assigned 1,102 <italic>A. pernix</italic> proteins to COGs. Some of these proteins (154) are members of COGs belonging to the uncharacterized (S) group, about which little is known except that they form a conserved family [<xref ref-type="bibr" rid="B12">12</xref>]. Subtracting these, annotation has been added to 315 proteins - an increase of about 50% compared with the original annotation. These include, among others, the key glycolytic enzymes glucose-6-phosphate isomerase (APE0768, COG0166) and triosephosphate isomerase (APE1538, COG0149), and the pyrimidine biosynthetic enzymes orotidine-5'-phosphate decarboxylase (APE2348, COG0284), uridylate kinase (APE0401, COG0528), cytidylate kinase (APE0978, COG1102), and thymidylate kinase (APE2090, COG0125). Similarly, important functions in DNA replication and repair were confidently assigned to a significant number of <italic>A. pernix</italic> proteins that in the original annotation were described simply as a 'hypothetical protein'. Examples include the bacterial-type DNA primase (COG0358), the large subunit of the archaeal-eukaryotic-type primase (COG2219), a second ATP-dependent DNA ligase (COG1423), three paralogous photolyases (COG1533), and several helicases and nucleases of different specificities.</p><p>The case of the large subunit of the archaeal-eukaryotic primase illustrates well the contribution of different types of inference to genome annotation. COGNITOR failed to assign an <italic>A. pernix</italic> protein to this COG. Given the ubiquity of this subunit in euryarchaea and eukaryotes [<xref ref-type="bibr" rid="B19">19</xref>], however, and the presence of a readily detectable small primase subunit in <italic>A. pernix</italic> (COG1467), a more detailed analysis was undertaken by running PSI-BLAST searches against the NR database with all members of the original COG as starting queries. When the <italic>Archaeoglobus fulgidus</italic> primase sequence (AF0336) was used to initiate the search, the <italic>A. pernix</italic> counterpart (APE0667) was indeed detected at a statistically significant level.</p><p>An interesting case of reannotation of a protein with a critical function, which also resulted in more general conclusions, is the archaeal uracil DNA glycosylase (UDG; COG1573). The members of this COG are currently annotated either as a putative DNA polymerase (APE0427 from <italic>A. pernix</italic> and AF2277 from <italic>A. fulgidus</italic>) or as a hypothetical protein. However, UDG activity has been experimentally shown for the respective proteins from <italic>Thermotoga maritima</italic> [<xref ref-type="bibr" rid="B20">20</xref>] and <italic>A. fulgidus</italic> [<xref ref-type="bibr" rid="B21">21</xref>]. The reason for the erroneous annotation as a DNA polymerase is the independent fusion of the uracil DNA glycosylase with DNA polymerases in bacteriophage SPO1 and in <italic>Yersinia pestis.</italic> Although these fusions hampered the correct annotation in the original analysis of the archaeal genomes, they seem to be functionally informative, suggesting that this type of UDG functions in conjunction with the replicative DNA polymerase. This is consistent with a recent report that archaeal DNA polymerases stall in the presence of uracil before misincorporating adenine [<xref ref-type="bibr" rid="B22">22</xref>].</p></sec><sec><title>Additions, subtractions and changes to the <italic>A. pernix</italic> protein set</title><p>In all, 1,102 of the predicted 2,694 <italic>A. pernix</italic> proteins (41%) were included in the COGs, whereas 1,404 of the predicted 1,765 <italic>P. abyssi</italic> proteins (79%) were included. The percentage of <italic>A. pernix</italic> proteins included in the COGs was significantly less than the average (72%) for the other five archaeal protein sets currently included in the COG database (Table <xref ref-type="table" rid="T2">2</xref>). It seems likely that this is due to an overestimate of the total number of ORFs in the <italic>A. pernix</italic> genome. Many of the ORFs with no similarity to proteins in sequence databases (1,538, or 57.1% [<xref ref-type="bibr" rid="B17">17</xref>]) overlap with ORFs from conserved families, including COG members. On the basis of the average representation of all genomes in the COGs (67%) and the average for the other archaea (72%), one could estimate the total number of <italic>A. pernix</italic> proteins to be between 1,550 and 1,700. This range is consistent with the size of the <italic>A. pernix</italic> genome (1.67 Mb) given the gene density of about one gene per kilobase that is typical of bacteria and archaea. Considering that <italic>A. pernix</italic> is the first crenarchaeon sequenced, and is also the only archaeal aerobe sequenced so far, one might expect that the representation of <italic>A. pernix</italic> proteins in COGs could be somewhat lower than the average for the Euryarchaeota. Taking 60% as a conservative estimate, this puts the upper limit of protein-coding genes in <italic>A. pernix</italic> at about 1,900. Complete reconstruction of the <italic>A. pernix</italic> proteome is beyond the scope of this work, but 849 ORFs, originally annotated as proteins, that significantly overlapped with COG members could be confidently excluded, which brings the number of genes to a maximum of 1,873.</p><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Comparison of proteins in COGs for archaeal species</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Species^* COGs</td><td align="center">Genome size (Mb)</td><td align="center" colspan="2">ORFs</td><td align="center">Percentage of ORFs in COGs</td></tr><tr><td></td><td></td><td colspan="2"><hr></hr></td><td></td></tr><tr><td align="left"></td><td align="center"></td><td></td><td></td><td align="center"></td></tr><tr><td></td><td></td><td align="center">Total</td><td align="center">in COGs</td><td></td></tr></thead><tbody><tr><td align="left">Af</td><td align="center">2.18</td><td align="center">2,411</td><td align="center">1,755</td><td align="center">73</td></tr><tr><td align="left">Mj</td><td align="center">1.74</td><td align="center">1,747</td><td align="center">1,252</td><td align="center">72</td></tr><tr><td align="left">Mth</td><td align="center">1.75</td><td align="center">1,871</td><td align="center">1,339</td><td align="center">72</td></tr><tr><td align="left">Ph</td><td align="center">1.74</td><td align="center">2,072</td><td align="center">1,333</td><td align="center">64</td></tr><tr><td align="left">Pa</td><td align="center">1.77</td><td align="center">1,765</td><td align="center">1,404</td><td align="center">79</td></tr><tr><td align="left">Ap</td><td align="center">1.67</td><td align="center">2,694</td><td align="center">1,102</td><td align="center">41</td></tr><tr><td align="left">Ap^&#x02020;</td><td align="center">1.67</td><td align="center">1,873</td><td align="center">1,129</td><td align="center">60</td></tr></tbody></table><table-wrap-foot><p>^*For abbreviations see the Materials and methods section. ^&#x02020;After adjusting for new ORFs and removal of likely false ORFs. This adjusted number of genes is the upper estimate of the actual total number of genes in <italic>A. pernix</italic>.</p></table-wrap-foot></table-wrap><p>Despite the apparent over-representation of ORFs in <italic>A. pernix</italic>, we nonetheless added 28 previously unidentified ORFs that represent conserved protein families, including such functionally indispensable proteins as chorismate mutase (APE0563a, COG1605), translation initiation factor IF-1 (APE_IF-1, COG0361), and seven ribosomal proteins (APE_rpl21E, COG2139; APE_rps14, COG0199; APE_rpl29, COG0255; APE_rplX, COG2157; APE_rpl39E, COG2167; APE_rpl34E, COG2174; APE_rps27AE, COG1998). These missed genes were identified by searching the genome sequence translated in all six frames for possible members of COGs with unexpected phylogenetic patterns. For example, the translation initiation factor IF-1 COG0361 contained exactly one protein from each of the species represented in COGs, except for <italic>A. pernix.</italic> Considering the importance of this protein in translation and its conservation across all species in the COGs, it seemed unlikely that it would be missing from <italic>A. pernix</italic>, and, indeed, a highly conserved IF-1 ortholog was readily identified in translating searches with the respective COG members as queries. Not unexpectedly, all newly identified <italic>A. pernix</italic> genes encode small proteins.</p></sec><sec><title>Conservation of the core of archaeal COGs shows that <italic>A. pernix</italic> is a typical archaeon</title><p>Because <italic>A. pernix</italic> is the first crenarchaeal genome to be completely sequenced, it was important to investigate whether or not the conserved core of archaeal genes previously identified by comparative analysis of euaryarchaeal genomes [<xref ref-type="bibr" rid="B19">19</xref>] is shared by the crenarchaea. The data in Table <xref ref-type="table" rid="T3">3</xref> indicate that this is indeed the case - in all functional categories of COGs, the majority of COGs that contain representatives from all five euryarchaeal species also include <italic>A. pernix.</italic> A very similar conclusion has been independently reached in a recent cluster analysis of archaeal proteins [<xref ref-type="bibr" rid="B23">23</xref>]. The fraction of the <italic>A. pernix</italic> gene set that belongs to this conserved core, approximately 30%, is very similar to those in each of the euryarchaea if the number of predicted <italic>A. pernix</italic> genes is adjusted as described above (Figure <xref ref-type="fig" rid="F2">2</xref>). Furthermore, the breakdown pattern of the proteins into members of COGs including all archaeal species, those in COGs with a subset of archaeal species, those in COGs with no other archaeal species, and those not included COGs, appeared to be conserved in <italic>A. pernix</italic> and each of the Euryarchaeota, indicating common evolutionary trends (Figure <xref ref-type="fig" rid="F2">2</xref>).</p><fig position="float" id="F2"><label>Figure 2</label><caption><p>The main phylogenetic patterns for the predicted proteins encoded in six archaeal genomes. Af, <italic>Archaeoglobus fulgidus</italic>; Mt, <italic>Methanobacterium thermoautotrophicum</italic>, Pa, <italic>Pyrococcus abyssi</italic>; Mj, <italic>Methanococcus jannaschii</italic>; Ph, <italic>Pyrococcus horikoshii</italic>; Ap, <italic>Aeropyrum pernix</italic>. 1, members of COGs including all archaeal species; 2, members of COGs including a subset of archaeal species; 3, members of COGs that include no archaeal species other then the analyzed one; 4, not in COGs. The percentage of proteins in each category is indicated.</p></caption><graphic xlink:href="gb-2000-1-5-research0009-2"/></fig><table-wrap position="float" id="T3"><label>Table 3</label><caption><p>Phyletic distribution of the archaeal COGs</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Functional category</td><td align="center">Number of COGs including all five euryarchaeal species</td><td align="center">Number of COGs including all five euryarchaeal species and <italic>A. pernix</italic></td><td align="center">Number of COGs including a subset of euryarchaeal species and <italic>A. pernix</italic></td><td align="center">Number of COGs including <italic>A. pernix</italic> but none of the euryarchaeal species</td></tr></thead><tbody><tr><td align="left">Translation and ribosome biogenesis</td><td align="center">113</td><td align="center">109</td><td align="center">17</td><td align="center">0</td></tr><tr><td align="left">Transcription</td><td align="center">26</td><td align="center">25</td><td align="center">11</td><td align="center">0</td></tr><tr><td align="left">Replication, recombination, repair</td><td align="center">38</td><td align="center">27</td><td align="center">15</td><td align="center">2</td></tr><tr><td align="left">Cell division and chromosome partitioning</td><td align="center">3</td><td align="center">1</td><td align="center">0</td><td align="center">0</td></tr><tr><td align="left">Post-translational modification,</td><td align="center">19</td><td align="center">15</td><td align="center">4</td><td align="center">2</td></tr><tr><td align="left">&#x02003;protein turnover, chaperones</td><td></td><td></td><td></td><td></td></tr><tr><td align="left">Cell envelope biogenesis, outer membrane</td><td align="center">8</td><td align="center">7</td><td align="center">8</td><td align="center">0</td></tr><tr><td align="left">Cell motility and secretion</td><td align="center">8</td><td align="center">8</td><td align="center">6</td><td align="center">0</td></tr><tr><td align="left">Inorganic ion transport and metabolism</td><td align="center">13</td><td align="center">9</td><td align="center">27</td><td align="center">2</td></tr><tr><td align="left">Signal transduction</td><td align="center">4</td><td align="center">4</td><td align="center">5</td><td align="center">1</td></tr><tr><td align="left">Carbohydrate transport and metabolism</td><td align="center">20</td><td align="center">18</td><td align="center">13</td><td align="center">2</td></tr><tr><td align="left">Energy production and conversion</td><td align="center">36</td><td align="center">22</td><td align="center">33</td><td align="center">18</td></tr><tr><td align="left">Amino acid transport and metabolism</td><td align="center">34</td><td align="center">29</td><td align="center">58</td><td align="center">5</td></tr><tr><td align="left">Nucleotide transport and metabolism</td><td align="center">34</td><td align="center">25</td><td align="center">6</td><td align="center">3</td></tr><tr><td align="left">Coenzyme metabolism</td><td align="center">25</td><td align="center">21</td><td align="center">30</td><td align="center">2</td></tr><tr><td align="left">Lipid metabolism</td><td align="center">8</td><td align="center">6</td><td align="center">15</td><td align="center">2</td></tr><tr><td align="left">General functional prediction only</td><td align="center">75</td><td align="center">59</td><td align="center">60</td><td align="center">8</td></tr><tr><td align="left">Uncharacterized</td><td align="center">67</td><td align="center">45</td><td align="center">82</td><td align="center">4</td></tr><tr><td align="left">Total</td><td align="center">514</td><td align="center">416</td><td align="center">273</td><td align="center">50</td></tr></tbody></table></table-wrap><p>The matrix of co-occurrence of genomes in the COGs (Table <xref ref-type="table" rid="T4">4</xref>) shows that the gene repertoire of <italic>A. pernix</italic> overlaps to a much greater extent with those of Euryarchaeota than with those of bacteria or yeast. Typically, there are fewer common COGs between <italic>A. pernix</italic> and euryarchaeal species than there are among the latter, although some preferential co-occurrence of <italic>A. pernix</italic> with the two species of <italic>Pyrococcus</italic> is notable (Table <xref ref-type="table" rid="T4">4</xref>). To further assess the relationships between genomes, we used the co-occurrence data to construct a distance matrix (see the Materials and methods section), which in turn was used for generating a cluster dendrogram and neighbor-joining and least-square trees (Figure <xref ref-type="fig" rid="F3">3</xref>). This analysis not only unequivocally placed <italic>A. pernix</italic> within the archaeal domain, but even grouped it with the <italic>Pyrococcus</italic> species in the cluster dendrogram (Figure <xref ref-type="fig" rid="F3">3a</xref>) and the neighbor-joining tree (Figure <xref ref-type="fig" rid="F3">3b</xref>), although not in the least-square tree where it was positioned at the base of the archaeal branch and outside of the Euryarchaeota (data not shown). The outcome of this type of analysis, which is conceptually similar to recent attempts at constructing 'gene content' evolutionary trees [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>], is a mixed reflection of phylogenetic relationships and similarities or differences in gene repertoires related to the lifestyles of the respective organisms. The contribution of the latter non-phylogenetic factors is well illustrated by the clustering of parasitic bacteria such as, for example, <italic>Haemophilus influenzae</italic> and <italic>Helicobacter pylori</italic>, which contradicts the obvious phylogenetic affinity of the former with <italic>Escherichia coli</italic>, and the deep branching of the mycoplasmas, the most degraded bacterial parasites, instead of the phylogenetically justified grouping with <italic>Bacillus subtilis</italic> (Figure <xref ref-type="fig" rid="F3">3</xref>). By the same token, it appears most likely that clustering of <italic>A. pernix</italic> with the pyrococci primarily reflects some common aspects of their metabolism which remain to be identified. A contribution of preferential lateral gene exchange to this grouping also seems possible. Some genes shared by <italic>A. pernix</italic> and the pyrococci, to the exclusion of the rest of the Euryarchaeota, are discussed below. <italic>A. pernix</italic> did not show any closer relationship to yeast than did the euryarchaea (Table <xref ref-type="table" rid="T4">4</xref> and Figure <xref ref-type="fig" rid="F3">3a</xref>,<xref ref-type="fig" rid="F3">b</xref>). Thus, at least at the level of co-occurrence in COGs, or in other words, the fraction of shared orthologs, we see no support for the hypothesis of the origin of eukaryotes from crenarchaea [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>].</p><fig position="float" id="F3"><label>Figure 3</label><caption><p>Classification of genomes by co-occurrence in the COGs. <bold>(a)</bold> A cluster dendrogram. <bold>(b)</bold> A neighbor-joining unrooted tree. For abbreviations, see the Materials and methods section.</p></caption><graphic xlink:href="gb-2000-1-5-research0009-3"/></fig><table-wrap position="float" id="T4"><label>Table 4</label><caption><p>Co-occurrence of genomes in COGs: <italic>A. pernix</italic> groups within the archaeal domain</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center"><bold>Ap</bold></td><td align="center">Mj</td><td align="center">Mth</td><td align="center">Af</td><td align="center">Ph</td><td align="center">Pa</td><td align="center">Tm</td><td align="center">Ec</td><td align="center">Bs</td><td align="center">Ssp</td><td align="center">Sc</td></tr></thead><tbody><tr><td align="left"><bold>Ap</bold></td><td align="center">-</td><td align="center">275</td><td align="center">273</td><td align="center">151</td><td align="center">203</td><td align="center">168</td><td align="center"><bold>424</bold></td><td align="center"><bold>321</bold></td><td align="center"><bold>349</bold></td><td align="center"><bold>386</bold></td><td align="center"><bold>387</bold></td></tr><tr><td></td><td align="center">836</td><td align="center">561</td><td align="center">563</td><td align="center">685</td><td align="center">634</td><td align="center">677</td><td align="center"><bold>412</bold></td><td align="center"><bold>515</bold></td><td align="center"><bold>487</bold></td><td align="center"><bold>450</bold></td><td align="center"><bold>449</bold></td></tr><tr><td></td><td align="center">-</td><td align="center">408</td><td align="center">424</td><td align="center">411</td><td align="center">299</td><td align="center">297</td><td align="center"><bold>647</bold></td><td align="center"><bold>955</bold></td><td align="center"><bold>878</bold></td><td align="center"><bold>752</bold></td><td align="center"><bold>402</bold></td></tr><tr><td align="left">Mj</td><td></td><td align="center">-</td><td align="center">157</td><td align="center">164</td><td align="center">252</td><td align="center">258</td><td align="center">507</td><td align="center">460</td><td align="center">496</td><td align="center">481</td><td align="center">524</td></tr><tr><td></td><td></td><td align="center">969</td><td align="center">812</td><td align="center">805</td><td align="center">717</td><td align="center">722</td><td align="center">462</td><td align="center">509</td><td align="center">473</td><td align="center">488</td><td align="center">445</td></tr><tr><td></td><td></td><td align="center">-</td><td align="center">175</td><td align="center">291</td><td align="center">249</td><td align="center">252</td><td align="center">597</td><td align="center">961</td><td align="center">892</td><td align="center">714</td><td align="center">406</td></tr><tr><td align="left">Mth</td><td></td><td></td><td align="center">-</td><td align="center">177</td><td align="center">325</td><td align="center">296</td><td align="center">509</td><td align="center">441</td><td align="center">478</td><td align="center">465</td><td align="center">515</td></tr><tr><td></td><td></td><td></td><td align="center">987</td><td align="center">810</td><td align="center">662</td><td align="center">700</td><td align="center">478</td><td align="center">546</td><td align="center">509</td><td align="center">522</td><td align="center">472</td></tr><tr><td></td><td></td><td></td><td align="center">-</td><td align="center">286</td><td align="center">271</td><td align="center">274</td><td align="center">581</td><td align="center">924</td><td align="center">856</td><td align="center">680</td><td align="center">379</td></tr><tr><td align="left">Af</td><td></td><td></td><td></td><td align="center">-</td><td align="center">355</td><td align="center">330</td><td align="center">569</td><td align="center">483</td><td align="center">526</td><td align="center">540</td><td align="center">593</td></tr><tr><td></td><td></td><td></td><td></td><td align="center">1096</td><td align="center">741</td><td align="center">780</td><td align="center">527</td><td align="center">613</td><td align="center">570</td><td align="center">556</td><td align="center">503</td></tr><tr><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">192</td><td align="center">194</td><td align="center">532</td><td align="center">857</td><td align="center">795</td><td align="center">646</td><td align="center">348</td></tr><tr><td align="left">Ph</td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">46</td><td align="center">460</td><td align="center">436</td><td align="center">453</td><td align="center">495</td><td align="center">487</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td align="center">933</td><td align="center">894</td><td align="center">473</td><td align="center">497</td><td align="center">480</td><td align="center">438</td><td align="center">446</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">80</td><td align="center">586</td><td align="center">973</td><td align="center">885</td><td align="center">764</td><td align="center">405</td></tr><tr><td align="left">Pa</td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">470</td><td align="center">431</td><td align="center">451</td><td align="center">501</td><td align="center">505</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">974</td><td align="center">504</td><td align="center">543</td><td align="center">523</td><td align="center">473</td><td align="center">469</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">577</td><td align="center">964</td><td align="center">898</td><td align="center">742</td><td align="center">396</td></tr><tr><td align="left">Tm</td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">223</td><td align="center">216</td><td align="center">325</td><td align="center">592</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">1059</td><td align="center">836</td><td align="center">843</td><td align="center">734</td><td align="center">467</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">634</td><td align="center">522</td><td align="center">468</td><td align="center">384</td></tr><tr><td align="left">Ec</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">339</td><td align="center">480</td><td align="center">813</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">1470</td><td align="center">1331</td><td align="center">990</td><td align="center">657</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">234</td><td align="center">212</td><td align="center">194</td></tr><tr><td align="left">Bs</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">464</td><td align="center">750</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">1365</td><td align="center">901</td><td align="center">615</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">301</td><td align="center">236</td></tr><tr><td align="left">Ssp</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">632</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">1202</td><td align="center">570</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td><td align="center">281</td></tr><tr><td align="left">Sc</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">851</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td align="center">-</td></tr></tbody></table><table-wrap-foot><p>^*In each cell, the middle line is the number of COGs in which the given two species co-occur; the top line is the number of COGs in which the genome in the corresponding row, but not the one in the corresponding column, is represented; conversely, the bottom line is the number of COGs in which the genome in the corresponding column, but not the one in the corresponding row, is represented. The diagonal cells show the total number of COGs that include representatives from the given genome. The cells that show the co-occurrence data among archaea show the numbers of COGs in red, and the cells that show the co-occurrence data for archaea and yeast show the numbers of COGs in blue. For abbreviations, see the Materials and methods section.</p></table-wrap-foot></table-wrap><p>Within the conserved archaeal core, there is a considerable number of genes that either comprise unique archaeal COGs (examples of these are shown in Table <xref ref-type="table" rid="T5">5</xref>) or are sporadically present in certain bacterial species, but not in eukaryotes. These COGs are of particular interest from the viewpoint of the 'standard model' of evolution of the three domains of life, which applies primarily to the information-processing systems of the cell and places the tree root between bacteria and archaea-eukaryotes [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>], because they can be considered synapomorphies defining the archaeal state (the sporadic representation of some such COGs in bacteria is most likely explained by horizontal gene transfer). Such characteristic archaeal proteins include, for example, the archaeal-type Holliday junction resolvase, the ATPase subunit of the archaeal-specific TopoVI (Table <xref ref-type="table" rid="T5">5</xref>) and the archaeal DnaG-like primase with its unique domain organization (COG0358 [<xref ref-type="bibr" rid="B19">19</xref>]). The finding that these proteins are shared by Euryarchaeota and Crenarchaeota is important because it suggests that the respective genes were most probably present in the common ancestor of archaea and eukaryotes, but have been displaced in the eukaryotic lineage.</p><table-wrap position="float" id="T5"><label>Table 5</label><caption><p>COGs represented in all archaea but not in other species: probable archaeal synapomorphies</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">COG number</td><td align="left">(Predicted) function</td><td align="left">Comments</td></tr></thead><tbody><tr><td align="left">2511</td><td align="left">Glu-tRNA^Gln amidotransferase B subunit</td><td align="left">This protein is homologous to bacterial B subunits and an archaeal paralog but contains a synapomorphic insert, the so-called GAD domain, shared with bacterial aspartyl-tRNA synthetases [49]</td></tr><tr><td align="left">2016</td><td align="left">Predicted RNA-binding protein, contains PUA domain</td><td align="left">PUA domain is most common in archaea and is found also in pseudouridine synthases, archaeosine synthases and glutamate kinases [50]</td></tr><tr><td align="left">1370</td><td align="left">Predicted RNA-binding protein, contains PUA domain</td><td align="left">This form of the PUA domain is present as a stand-alone protein in <italic>A. pernix</italic> and <italic>A. fulgidus</italic> but is fused with the archaeosine synthase in the other euryarchaea [50]</td></tr><tr><td align="left">1746</td><td align="left">tRNA nucleotidyltransferase (CCA-adding enzyme)</td><td align="left">Archaeal CCA-adding enzyme is only very distantly related to other members of the Pol&#x003b2; superfamily of nucleotidyltransferases [51]</td></tr><tr><td align="left">1395</td><td align="left">Predicted transcription regulators</td><td align="left">The proteins of this family do not share similarity with other proteins beyond the DNA-binding helix-turn-helix domain [32]</td></tr><tr><td align="left">1389</td><td align="left">DNA topoisomerase VI, subunit B</td><td align="left">These proteins contain an ATPase domain of the TopoII/MutL/HSP90/histidine kinase fold, but do not show a specific relationships to any other proteins of this class</td></tr><tr><td align="left">1591</td><td align="left">Holliday junction resolvase, archaeal-type</td><td align="left">Distant homologs seen in some bacteria (L.A., K.S. Makarova and E.V.K., unpublished.observations)</td></tr><tr><td align="left">1571</td><td align="left">Predicted DNA-binding proteins, possibly nucleotidyl transferase or nuclease</td><td align="left">These proteins consist of two distinct, predicted DNA-binding domains (OB-fold and Zn-ribbon) and an uncharacterized, probably enzymatic domain that is unique for archaea (see text)</td></tr><tr><td align="left">1491</td><td align="left">Predicted DNA-binding protein</td><td align="left">These proteins contain the helix-hairpin-helix module, but otherwise, do not show significant similarity to any other proteins</td></tr><tr><td align="left">1938</td><td align="left">Predicted ATP-grasp-domain-containing enzymes</td><td align="left">Only distantly related to other ATP-grasp proteins; predicted to possess ATP-dependent carboligase or similar activity [19]</td></tr><tr><td align="left">1407</td><td align="left">Predicted calcineurin-type phosphoesterase</td><td align="left">Only distantly related to other phosphohydrolases of the calcineurin fold [37]</td></tr><tr><td align="left">1782</td><td align="left">Predicted metal-dependent RNase of the metallo-&#x003b2;-lactamase fold</td><td align="left">In spite of significant similarity to other families of metallo-&#x003b2;-lactamases, this family shows a clear synapomorphy, the presence of the RNA-binding KH domain [52]</td></tr><tr><td align="left">1608</td><td align="left">Predicted kinase related to acetylglutamate kinase</td><td align="left">Only distantly related to other kinases of the same fold</td></tr><tr><td align="left">1829</td><td align="left">Predicted kinase of the actin/HSP70/sugar kinase fold</td><td align="left">Only distantly related to other kinases of the same fold</td></tr><tr><td align="left">1907</td><td align="left">Predicted kinase of the actin/HSP70/sugar kinase fold</td><td align="left">Only distantly related to other kinases of the same fold</td></tr><tr><td align="left">1831</td><td align="left">Predicted metal-dependent hydrolase of the urease superfamily</td><td align="left">Only distantly related to other hydrolases of the same superfamily [53]</td></tr><tr><td align="left">1571</td><td align="left">Predicted DNA-binding protein containing the Zn-ribbon module</td><td></td></tr><tr><td align="left">2034</td><td align="left">Conserved membrane protein</td><td></td></tr><tr><td align="left">2064</td><td align="left">Conserved membrane protein</td><td></td></tr><tr><td align="left">1339, 2090, 1581,</td><td align="left">Uncharacterized proteins unique to archaea</td><td></td></tr><tr><td align="left">1460, 1786, 1701, 1931, 1909, 1888, 1382, 1849, 1630, 1303, 1325, 1679</td><td></td><td></td></tr></tbody></table></table-wrap><p>Detailed sequence analysis of the core archaeal genes, which included comparison of the protein sequences from the respective COGs to pre-computed profiles for specific protein domains, resulted in the prediction of previously uncharacterized potential roles in conserved systems for some of them. For example, proteins in COG1571 contain two previously recognized domains, namely a nucleic-acid-binding OB-fold domain similar to those found in the ssDNA-binding protein RPA and in DNA polymerase subunits from archaea and bacteria [<xref ref-type="bibr" rid="B31">31</xref>] and a metal-binding Zn ribbon [<xref ref-type="bibr" rid="B32">32</xref>]. The amino-terminal portion of these proteins comprises a predicted globular domain that also occurs as a stand-alone protein in some of the euryarchaea and contains a conserved signature GxDDXD preceded by a predicted &#x003b2; strand. The combination of this potential enzymatic domain with two predicted DNA-binding domains suggests that members of this protein family are likely to be enzymes with an important role in archaeal DNA metabolism, most probably nucleotidyl transferases or nucleases. The members of COG1444 are multidomain proteins that combine an amino-terminal superfamily I helicase-like ATPase domain with a carboxy-terminal acetyltransferase domain. This domain organization is suggestive of a role in the basal transcription system as a protein-modifying acetyltransferase. COG1094 (KH + S1 domains) and COG1096 (Si+Zn-ribbon domain) are predicted to encode RNA-binding proteins that could be involved in RNA processing or in a translation-related role. COG1293 includes uncharacterized, conserved proteins whose probable ortholog from <italic>Streptococcus</italic> has been annotated as a fibronectin-binding protein [<xref ref-type="bibr" rid="B33">33</xref>]. Their conservation and phyletic distribution is, however, more consistent with a basic core function. Consistent with this, we detected in these proteins a specific version of the helix-hairpin-helix nucleic-acid-binding module, which is specifically similar to those found in ribosomal proteins of the S13/S18 family (L.A., unpublished observations) and suggests a function for the members of this COG as a ribosome-associated, RNA-binding protein or, less likely, an uncharacterized DNA repair system component.</p><p>A tally of the COGs that are shared by <italic>A. pernix</italic> with each of the five euryarchaeal species, to the exclusion of the rest of the euryarchaea, shows a clear prevalence of the association with pyrococci and <italic>A. fulgidus</italic> (data not shown). Given the larger number of genes in the latter, the relationship with the pyrococci is most notable, in agreement with the clustering data presented above. <italic>A. pernix</italic> and the pyrococci share some typically bacterial proteins, for example ribonuclease E/G, an RNA-processing enzyme (COG1530), and the glycine cleavage system (COGs 0403, 0404, 0509 and 1003). This is compatible with horizontal gene transfer between these two archaeal lineages subsequent to the acquisition of the respective gene from a bacterium.</p><p><italic>A. pernix</italic> shows a notable paucity of signaling proteins, resembling in this respect <italic>Methanococcus jannaschii</italic> and parasitic bacteria. <italic>A. pernix</italic> and <italic>M. jannaschii</italic> have no detectable histidine kinase, PAS or GAF domains, unlike <italic>Methanobacterium thermoautotrophicum</italic> and <italic>A. fulgidus</italic>, in which these domains comprise the basis of the signaling system. These domains are present in the pyrococci, but in much smaller numbers than in <italic>M. thermoautotrophicum</italic> and <italic>A. fulgidus</italic>. <italic>A. pernix</italic>, <italic>M. jannaschii</italic> and the pyrococci also encode very few serine/threonine kinases, and those that are present are highly conserved representatives of the Ri01 family, which are probably involved in transcription regulation rather than in typical signaling [<xref ref-type="bibr" rid="B34">34</xref>]. It appears therefore that conventional phosphorylation-mediated signaling is selected against in <italic>A. pernix</italic> and other hyperthermophiles. In contrast, all these archaea, including <italic>A. pernix</italic>, possess comparable numbers of predicted transcription factors, mainly those of the helix-turn-helix class [<xref ref-type="bibr" rid="B32">32</xref>], some of which could directly bind small molecules and convert signals into transcriptional outputs.</p></sec><sec><title>Crenarchaeota as a distinct branch of the Archaea</title><p>In spite of the rather unexpected clustering with the pyrococci seen in the co-occurrence-based classification (Figure <xref ref-type="fig" rid="F3">3</xref>), the COG analysis provides evidence for the distinctness of Crenarchaeota as represented by <italic>A. pernix.</italic> At a quantitative level, this can be illustrated by comparing the number of COGs in which each of the archaeal species is missing but the rest are represented. The number of such COGs is notably greater in the case of <italic>A. pernix</italic> than in each of the euryarchaeal species (Figure <xref ref-type="fig" rid="F4">4</xref>). Qualitatively, a considerable number of apparently essential genes conserved in the euryarchaea are missing in <italic>A. pernix</italic> (Table <xref ref-type="table" rid="T6">6</xref>). <italic>A. pernix</italic> is expected to encode unrelated or distantly related proteins performing these functions (non-orthologous gene displacement [<xref ref-type="bibr" rid="B35">35</xref>]). Notably, unlike the euryarchaea with completely sequenced genomes, <italic>A. pernix</italic> does not encode enzymes of the <italic>de novo</italic> purine biosynthesis pathway (Table <xref ref-type="table" rid="T6">6</xref>). The enzymes for interconversion of IMP, GMP and AMP are present, and it appears likely that <italic>A. pernix</italic> partly relies on salvage pathways for the formation of purine nucleotides, but also probably imports nucleosides and/or bases into the cell. In this respect, <italic>A. pernix</italic> is similar to such parasitic bacteria as <italic>H. pylori</italic>, <italic>Borrelia burgdorferi</italic>, and <italic>Chlamydia</italic> that do not possess purine biosynthesis capabilities either and import nucleosides or bases from the surrounding medium. In contrast, the pyrimidine biosynthesis pathway genes are present in <italic>A. pernix</italic>, as they are in other archaea.</p><fig position="float" id="F4"><label>Figure 4</label><caption><p>COGs not represented in each of the archaeal species while including members of the remaining five species. For <italic>P. horikoshii</italic> and <italic>P. abyssi</italic>, the absence of the respective second pyrococcal species was allowed. For abbreviations, see the Materials and methods section.</p></caption><graphic xlink:href="gb-2000-1-5-research0009-4"/></fig><table-wrap position="float" id="T6"><label>Table 6</label><caption><p>Examples of COGs conserved in euryarchaea but missing in <italic>A. pernix</italic></p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">COG number</td><td align="left">Phylogenetic pattern^*</td><td align="left">Function</td><td align="left">Comments</td></tr></thead><tbody><tr><td align="left">0101</td><td align="left">amtks-yqvdcebrhujwgpolinx</td><td align="left">Pseudouridylate synthase (tRNA psi55)</td><td align="left">This RNA modification enzyme hitherto has been considered ubiquitous and essential [54]</td></tr><tr><td align="left">1549</td><td align="left">amtks- - - - - - - - - - - - - - - - - - - -</td><td align="left">Archaeosine tRNA-ribosyltransferase, contains PUA domain</td><td align="left">See text</td></tr><tr><td align="left">2036</td><td align="left">amtks-y- - - - - - - - - - - - - - - - - -</td><td align="left">Histones H3 and H4</td><td align="left">In the crenarchaeon <italic>Sulfolobus solfataricus</italic>, the role of histones in maintaining DNA supercoiling appears to be relegated to the DNA-binding protein Sso7d [55]. However, <italic>A. pernix</italic> does not encode an ortholog of this protein, and the mechanism of chromatin packaging remains unclear</td></tr><tr><td align="left">1933</td><td align="left">amtks- - - - - - - - - - - - - - - - - - - -</td><td align="left">Unique archaeal DNA polymerase, large subunit</td><td align="left">This DNA polymerase is highly conserved among the euryarchaea but so far has not been seen in other taxa [15]</td></tr><tr><td align="left">1311</td><td align="left">amtks-y- - - - - - - - - - - - - - - - - -</td><td align="left">Small subunit of DNA polymerase, predicted phosphatase (calcineurin-like superfamily)</td><td align="left">The absence of this subunit, which is represented by (predicted) active phosphatases in archaea and inactivated forms in eukaryotes [37] correlates with the absence of the large subunit</td></tr><tr><td align="left">1111</td><td align="left">amtks-y- - - - - - - - - - - - - - - - -</td><td align="left">ERCC4-like helicase</td><td align="left">A typical archaeal-eukaryotic repair protein, a predicted active helicase in euryarchaea and an inactivated form in eukaryotes [39]</td></tr><tr><td align="left">1107</td><td align="left">amtks- - - - - - - - - - - - - - - - - - - -</td><td align="left">Archaea-specific RecJ-like exonuclease, ontains DnaJ-type Zn finger domain</td><td></td></tr><tr><td align="left">1243</td><td align="left">amtks-y-v- - - - - - - - - - - - - - - -</td><td align="left">Transcription elongation factor ELP3</td><td align="left">Consists of an amino-terminal biotin-synthase-like domain and a carboxy-terminal histone acetylase domain (only the biotin-synthase-like domain is represented in some bacteria)</td></tr><tr><td align="left">0206</td><td align="left">amtks- -qvdcebrhuj-gpol-nx</td><td align="left">Cell division GTPase FtsZ</td><td align="left">A central component of bacterial and archaeal cell-division machinery which is homologous (although only weakly similar) and functionally analogous to eukaryotic tubulins [56]. So far, among prokaryotes, FtsZ has been found missing only in <italic>Chlamydia</italic> [57].</td></tr><tr><td align="left">0455</td><td align="left">amtks- -qvdcebr-uj- - -ol-n-</td><td align="left">ATPases involved in chromosome partitioning</td><td align="left">There may be a correlation between the absence of FtsZ and the presence of only one chromosome-partitioning ATPase (the ortholog of bacterial Mrp) in <italic>A. pernix</italic></td></tr><tr><td align="left">1149</td><td align="left">amtks- - -v- - - - - - - - - - - - - - -</td><td align="left">MinD superfamily P-loop ATPases containing an inserted ferredoxin domain</td><td></td></tr><tr><td align="left">0065</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">3-Isopropylmalate dehydratase large subunit</td><td align="left"><italic>A. pernix</italic> does not encode committed enzymes of leucine biosynthesis, indicating that, unlike all other free-living organisms whose genomes have been sequenced so far, it cannot produce leucine. It might depend on other archaea and/or bacteria as the source of leucine, or perhaps synthesize leucine by a distinct, completely uncharacterized pathway</td></tr><tr><td align="left">0066</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">3-isopropylmalate dehydratase small subunit</td><td></td></tr><tr><td align="left">0473</td><td align="left">amtks-yqvdcebrh-j- - - - - -nx</td><td align="left">3-isopropylmalate dehydrogenase</td><td></td></tr><tr><td align="left">0119</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">Isopropylmalate/homocitrate/citramalate synthase</td><td></td></tr><tr><td align="left">0015</td><td align="left">amtks-yqvdcebrhuj- - - - - -n-</td><td align="left">Adenylosuccinate lyase</td><td align="left">Most of the enzyme of the <italic>de novo</italic> pathway of purine biosynthesis are missing in <italic>A. pernix.</italic> It appears likely that, in this organism, all purine nucleotides are produced via the salvage pathway.</td></tr><tr><td align="left">0104</td><td align="left">amtks-yqvdcebrhuj- - - - - -n-</td><td align="left">Adenylosuccinate synthase</td><td></td></tr><tr><td align="left">0034</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">Glutamine phosphoribosyl-pyrophosphate amidotransferase</td><td></td></tr><tr><td align="left">0151</td><td align="left">amtks-yqvdcebrhuj- - - - - -n-</td><td align="left">Phosphoribosylamine-glycine ligase</td><td></td></tr><tr><td align="left">0150</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">Phosphoribosylamino-imidazol (AIR) synthetase PurM</td><td></td></tr><tr><td align="left">0152</td><td align="left">amtks-yqvdcebrh-j- - - - - -nx</td><td align="left">Phosphoribosylamino-imidazolesuccino-carboxamide (SAICAR) synthase</td><td></td></tr><tr><td align="left">0041</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">Phosphoribosylcarboxy-aminoimidazole (NCAIR) mutase PurE</td><td></td></tr><tr><td align="left">0047</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">Phosphoribosylformyl-glycinamidine (FGAM) synthase, glutamine amidotransferase</td><td></td></tr><tr><td></td><td></td><td align="left">domain</td><td></td></tr><tr><td align="left">0046</td><td align="left">amtks-yqvdcebrh-j- - - - - -n-</td><td align="left">Phosphoribosylformyl-glycinamidine (FGAM) synthase, synthetase domain</td><td></td></tr><tr><td align="left">0340</td><td align="left">amtks-yqvdcebrhuj- - - -linx</td><td align="left">Biotin-(acetyl-CoA carboxylase) ligase</td><td align="left"><italic>A. pernix</italic> apparently does not encode any enzymes of biotin synthesis or biotin-utilizing enzymes</td></tr><tr><td align="left">0511</td><td align="left">amtks-yqvdcebrhuj- - - -linx</td><td align="left">Biotin carboxyl carrier protein of acetyl-CoA carboxylase</td><td></td></tr><tr><td align="left">0157</td><td align="left">amtks-yqv-cebrhu- - - - - - -n-</td><td align="left">Nicotinate-nucleotide pyrophosphorylase</td><td></td></tr></tbody></table><table-wrap-foot><p>^*In the phylogenetic patterns, each letter indicates that a particular genome is represented in the given COG, and a dash indicates the absence of a representative from the corresponding genome. The one-letter code for genomes is as follows: a, <italic>Archeoglobus fulgidus</italic>; m, <italic>Methanococcus jannaschii</italic>; t, <italic>Methanobacterium thermoautotrophicum</italic>; k, <italic>Pyrococcus horikoshii</italic>; s, <italic>Pyrococcus abyssi</italic>; z, <italic>Aeropyrum pernix;</italic> y, <italic>Saccharomyces cerevisiae</italic>; q, <italic>Aquifex aeolicus</italic>; v, <italic>Thermotoga maritima</italic>; d, <italic>Deinococcus radiodurans</italic>; c, <italic>Synechocystis</italic> sp; e, <italic>Escherichia coli</italic>; b, <italic>Bacillus subtilis</italic>; r, <italic>Mycobacterium tuberculosis</italic>; h, <italic>Haemophilus influenzae</italic>; u, <italic>Helicobacter pylori</italic>; j, <italic>Campylobacter jejuni</italic>; w, <italic>Ureaplasma urealyticum</italic>; g, <italic>Mycoplasma genitalium</italic>; p, <italic>Mycoplasma pneumoniae</italic>; o, <italic>Borrelia burgdorferi</italic>; l, <italic>Treponema pallidum</italic>; i, <italic>Chlamydia trachomatis</italic> and <italic>C. pneumoniae</italic>; n, <italic>Neisseria meningitidis</italic>; x, <italic>Rickettsia prowazekii</italic>.</p></table-wrap-foot></table-wrap><p><italic>A. pernix</italic> lacks certain conserved RNA-modifying enzymes such as the tRNA-specific pseudouridine synthase (COG0101), which has so far been considered ubiquitous, and the tRNA archaeosine transglycosylase (COG1549). The absence of the latter enzyme suggests that archaeosine could be a euryarchaea-specific RNA modification.</p><p>Interestingly, <italic>A. pernix</italic> lacks several conserved proteins and features of domain architecture that are specifically shared by euryarchaea and eukaryotes. A particularly notable absence is that of the two subunits of the euryarchaeal DNA polymerase (Table <xref ref-type="table" rid="T6">6</xref>). The large subunit is present only in the euryarchaea [<xref ref-type="bibr" rid="B36">36</xref>], whereas the small subunit, which belongs to the calcineurin superfamily of phosphoesterases and is predicted to possess phosphatase activity, is conserved in euryarchaea and eukaryotes [<xref ref-type="bibr" rid="B37">37</xref>]. Unlike the euryarchaea, the replicative DNA polymerases in <italic>A. pernix</italic> are represented only by three paralogous members of the B family (one of them possibly inactivated; COG0417), which is shared by archaea and eukaryotes [<xref ref-type="bibr" rid="B38">38</xref>]. The replication factor A (RPA) ortholog (COG1599) from <italic>A. pernix</italic> contains a single OB-fold domain whereas euryarchaea and eukaryotes encode forms with multiple tandem repeats of the OB-fold that are more similar to each other. The eukaryotic DNA repair component ERCC4 and its ortholog in euryarchaea contain fused super-family II helicase and nuclease domains (COG1111 and COG1948 [<xref ref-type="bibr" rid="B39">39</xref>]). In contrast, the crenarchaea <italic>A. pernix</italic> and <italic>Sulfolobus</italic> possess only the nuclease domain (also represented in <italic>A. fulgidus</italic>), with no counterpart to the helicase domain. Two predicted helicases with a potential function in DNA repair (COG1112, a superfamily I helicase, and COG1205, a superfamily II helicase fused to a predicted metal-dependent nuclease domain) are also shared by the euryarchaea and eukaryotes, to the exclusion of <italic>A. pernix.</italic> The transcription machinery of <italic>A. pernix</italic> also shows several deviations from the general euryarchaeal-eukaryotic pattern. The transcription factor TFIIB from both <italic>A. pernix</italic> and <italic>Sulfolobus</italic> (COG1405) contains a disrupted amino-terminal Zn ribbon domain. However, <italic>A. pernix</italic> encodes a stand-alone version of the TFIIB Zn ribbon (APE0508 in COG1405) that could substitute for the disrupted version <italic>in trans.</italic> Importantly, the histones that function in the chromosomal structure maintenance and possibly also in transcription in euryarchaea (COG2036) are lacking in <italic>A. pernix</italic> (see Table <xref ref-type="table" rid="T6">6</xref> for more details). The RNA polymerase elongation factor ELP3, which combines a biotin synthase domain with a histone acetylase domain and is shared by euryarchaea and eukaryotes (COG1243), is missing in <italic>A. pernix.</italic></p><p>The current COG collection was not well suited for detecting COGs that are represented exclusively in crenarchaea and in eukaryotes because only one species from each of these taxa is represented. Our additional analysis, however, revealed very few such genes in <italic>A. pernix</italic> and no trends supporting a possible ancestral relationship between Crenarchaeota and eukaryotes were detected (L.A. and E.V.K., unpublished observations). The current status of archaeal genome analysis, with more 'eukaryotic' features seen in euryarchaea than in crenarchaea, offers no support for the 'eocyte hypothesis', which postulates origin of eukaryotes from crenarchaea [<xref ref-type="bibr" rid="B28">28</xref>]. For a more definitive interpretation of the evolutionary relationships between archaea and eukaryotes, however, additional genome sequences, particularly those from other crenarchaea, are required.</p></sec><sec><title><italic>Aeropyrum</italic> genes without homologs in Euryarchaeota and acquisition of bacterial genes by horizontal transfer</title><p>The 48 COGs that include a representative from <italic>A. pernix</italic> but not from euryarchaea are likely to reflect acquisition of bacterial genes by crenarchaea or loss of ancestral genes early in the evolution of euryarchaea (Table <xref ref-type="table" rid="T7">7</xref>). Rigorously distinguishing between these two possibilities may be difficult, but in particular cases there are indications in favor of one of them when there is a distinct similarity between an <italic>A. pernix</italic> protein and orthologs from a particular bacterial lineage (Table <xref ref-type="table" rid="T7">7</xref>). Generally, horizontal gene transfer appears to be the most plausible scenario for the origin of these genes in <italic>A. pernix</italic> because none of the genes in these 48 COGs is found in all bacteria and eukaryotes, and none shows specific affinity with eukaryotic orthologs. The converse would have suggested ancestral provenance.</p><table-wrap position="float" id="T7"><label>Table 7</label><caption><p><italic>A. pernix</italic> proteins conserved in a wide range of organisms but missing in euryarchaea (examples)</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left"><italic>A. pernix</italic> gene/COG number</td><td align="left">Phylogenetic pattern^*</td><td align="left">Function</td><td align="left">Comments</td></tr></thead><tbody><tr><td align="left">APE1618/1048</td><td align="left">- - - - -zyq-d-ebr- - - - - - - - - -x</td><td align="left">Aconitase A</td><td align="left">Unlike other archaea with sequenced genomes, <italic>A. pernix</italic> is an aerobe and possesses the complete TCA cycle. APE1618 belongs to a distinct family of (otherwise) bacterial aconitases (e.g <italic>E. coli</italic> AcnA and <italic>B. subtilis</italic> CitB). APE1816 also belongs to a specific family of bacterial fumarases.</td></tr><tr><td align="left">APE1816/0114</td><td align="left">- - - - -zy- -dcebrhu- - - - - -i-x</td><td align="left">Fumarase</td><td></td></tr><tr><td align="left">APE1677/1071</td><td align="left">- - - - -zy- -dc-br- - - -gp- -i-x</td><td align="left">pyruvate dehydrogenase E1 component, &#x003b1;-subunit</td><td></td></tr><tr><td align="left">APE1674/0022</td><td align="left">- - - - -zy- -dc-br- - - -gp- -i-x</td><td align="left">pyruvate dehydrogenase E1 component, &#x003b2;-subunit</td><td></td></tr><tr><td align="left">APE1671/0508</td><td align="left">- - - - -zy- -dcebrh- - -gp- -i-x</td><td align="left">Dihydrolipoamide acyltransferase</td><td></td></tr><tr><td align="left">APE1725/1290</td><td align="left">- - - - -z-q-dc-br-uj- - - - - -nx</td><td align="left">Cytochrome b</td><td align="left">As an aerobe, <italic>A. pernix</italic> encodes specific electron-transport chain components. APE1623 is closely related to the ortholog from <italic>Aquifex aeolicus</italic>.</td></tr><tr><td align="left">APE1623, APE0793_1/0843</td><td align="left">- - - - -z-q-dcebr-uj- - - - - -nx</td><td align="left">Cytochrome c oxidase, heme b and copper-binding subunit</td><td align="left"></td><td></td></tr><tr><td align="left">APE0793_2/</td><td align="left">- - - - -z-q-dcebr- - - - - - - - - -x</td><td align="left">Cytochrome oxidase, subunit 3</td><td></td></tr><tr><td align="left">1845</td><td></td><td></td><td></td></tr><tr><td align="left">APE1498/1171</td><td align="left">- - - - -zy-vdcebrh- - - - - - - - -x</td><td align="left">Threonine dehydratase</td><td align="left">Specifically related to a subfamily of bacterial catabolic threonine dehydratases (e.g. <italic>E. coli</italic> TdcB)</td></tr><tr><td align="left">APE1038/0295</td><td align="left">- - - - -zy-vd-ebrh- -wgpo- - - -</td><td align="left">Cytidine deaminase</td><td></td></tr><tr><td align="left">APE1353/0514</td><td align="left">- - - - -zy-dceb-h- - - - - -l- - -</td><td align="left">DNA helicase (RecQ family)</td><td align="left">APE1353 differs from other members of the RecQ family by the presence of long amino-terminal extension that probably form a non-globular domain. APE1353 shows no specific affinity to any of the bacterial orthologs.</td></tr><tr><td align="left">APE2450/0260</td><td align="left">- - - - -z-q-dcebrhu-wgp- -i-x</td><td align="left">Leucyl aminopeptidase</td><td></td></tr><tr><td align="left">APE0137/0405</td><td align="left">- - - - -zy- -dcebr-u- - - - - - - - -</td><td align="left">Gamma-glutamyltranspeptidase</td><td></td></tr><tr><td align="left">APE2464/</td><td align="left">- - - - -z-qvdcebr- - - - - - - - - - -</td><td align="left">Phosphate starvation-inducible protein</td><td></td></tr><tr><td align="left">COG1702</td><td></td><td align="left">PhoH, Predicted ATPase</td><td></td></tr><tr><td align="left">APE0993/&#x0003e;0813</td><td align="left">- - - - -z- - -d-eb-hu-wgp-l- - -</td><td align="left">Purine-nucleoside phosphorylase</td><td align="left">Correlates with the absence of <italic>de novo</italic> purine biosynthesis and the probable importance of salvage pathways</td></tr><tr><td align="left">APE2105/0813a</td><td align="left">- - - - -z- - - - - -e- -h- - - - -l- - -</td><td align="left">Uridine phosphorylase</td><td></td></tr><tr><td align="left">APE0033/1866</td><td align="left">- - - - -zy- -d-eb-h- - - - - - - - - -</td><td align="left">Phosphoenolpyruvate carboxykinase</td><td></td></tr></tbody></table><table-wrap-foot><p>^*The designations are as in Table <xref ref-type="table" rid="T6">6</xref>.</p></table-wrap-foot></table-wrap><p>The most notable group of genes that are found in <italic>A. pernix</italic>, but not in euryarchaea, reflects its aerobic metabolism. As an aerobe, <italic>A. pernix</italic> encodes the complete set of the tricarboxylic acid cycle (TCA) cycle enzymes, in contrast to the anaerobic euryarchaea which possess a truncated version of this pathway (Table <xref ref-type="table" rid="T7">7</xref>). In addition, <italic>A. pernix</italic> encodes accessory enzymes that are required for the formation of the pyruvate dehydrogenase complex, such as lipoate synthase (COG0320) and lipoate-protein ligase (COG0095). The set of <italic>A. pernix</italic> proteins that are related to respiration and not seen in euryarchaea additionally includes specific electron-transfer chain components such as cytochrome <italic>b</italic>, cytochrome oxidase, nitrate reductase, NADH-ubiquinone oxidoreductase, NADPH:quinone reductase and Rieske Fe-S protein.</p><p>A set of 23 COGs is shared by <italic>A. pernix</italic>, yeast and a subset of bacteria, to the exclusion of the euryarchaea. These cases seem to be readily explained by lateral acquisition of genes from a bacterial source in both eukaryotes (largely from mitochondria) and crenarchaea. Respiration-related enzymes mentioned above are an obvious case in point, but this explanation could also apply to at least some of the remaining few COGs in this set, for example, a RecQ-family helicase (COG0514).</p></sec><sec><title>Detecting interspecies differences in <italic>Pyrococcus abyssi</italic> and <italic>P. horikoshii</italic></title><p>In the case of two closely related genomes, such as those of <italic>P. abyssi</italic> and <italic>P. horikoshii</italic>, the COG analysis provides for straightforward genome subtraction (Table <xref ref-type="table" rid="T8">8</xref>). Given the high level of sequence conservation between orthologous proteins from these two species (typically over 70% identity), it seems somewhat unexpected that 80 COGs include proteins from <italic>P. abyssi</italic>, but not <italic>P. horikoshii</italic>, whereas the inverse is seen in 46 COGs. Many of these differences are likely to reflect differential gene loss, whereas others are probably due to horizontal gene acquisition. The greater number of COGs that <italic>P. abyssi</italic> is represented in, to the exclusion of <italic>P. horikoshii</italic>, seems to reflect the greater metabolic endowment of the former. In particular, the entire aromatic amino acid and cysteine biosynthesis pathways are present in <italic>P. abyssi</italic> but not in <italic>P. horikoshii</italic> (Table <xref ref-type="table" rid="T8">8</xref>). In the case of the aromatic amino acid pathway, the direction of evolution seems to be clear - loss of the respective genes by <italic>P. horikoshii</italic> in the course of its adaptation to the heterotrophic lifestyle which seems to have gone further than in <italic>P. abyssi.</italic> The case of the cysteine pathway is, however, particularly interesting because, among all archaea whose genomes are currently available, it is shared only by <italic>A. pernix</italic> and <italic>P. abyssi</italic>; the mechanism of cysteine formation in other euryarchaea remains a mystery [<xref ref-type="bibr" rid="B19">19</xref>]. In this case, acquisition of the respective genes from bacteria via horizontal transfer seems to be the most likely possibility because a gene-loss scenario would require several independent events in euryarchaea. Interestingly, probable horizontal acquisition of bacterial genes encoding the cytosine biosynthesis pathway enzymes has been described also in the euryarchaeon <italic>Methanosarcina barkeri</italic> [<xref ref-type="bibr" rid="B40">40</xref>]. Other COGs with differential representation of the two <italic>Pyrococcus</italic> species tend to include genes that are inherently mobile such as restriction-modification systems (Table <xref ref-type="table" rid="T8">8</xref>).</p><table-wrap position="float" id="T8"><label>Table 8</label><caption><p>Examples of differential genome display of <italic>Pyrococcus abyssi</italic> and <italic>Pyrococcus horikoshii</italic> using the COG approach</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left"></td><td></td><td></td><td></td></tr><tr><td align="left">Gene/COG Number</td><td align="left">Phylogenetic pattern^*</td><td align="left">Function</td><td align="left">Comment</td></tr></thead><tbody><tr><td align="left" colspan="4">Present in <italic>P.abyssi</italic> but not <italic>P. horikoshii</italic></td></tr><tr><td align="left">&#x02003;PAB2044/0547</td><td align="left">amt-szyqvdcebrhuj- - - - - -n-</td><td align="left">Anthranilate phosphoribosyltransferase</td><td align="left">The entire branched pathway for aromatic amino acid biosynthesis appears to be present in <italic>P. abyssi</italic> but not in <italic>P. horikoshii</italic></td></tr><tr><td align="left">&#x02003;PAB2045/0147</td><td align="left">amt-szyqvdcebrhuj- - - - - -n-</td><td align="left">Anthranilate synthase component I</td><td></td></tr><tr><td align="left">&#x02003;PAB2046/0512</td><td align="left">amt-szyqvdcebrhuj- - - - - -n-</td><td align="left">Anthranilate synthase component II</td><td></td></tr><tr><td align="left">&#x02003;PAB0307/0082</td><td align="left">- - - -szyqvdceb-huj- - - - -inx</td><td align="left">DAHP synthase</td><td></td></tr><tr><td align="left">&#x02003;PAB2049/0159</td><td align="left">amt-szyqvdcebrhuj- - - - -in-</td><td align="left">Tryptophan synthase &#x003b1; subunit</td><td></td></tr><tr><td align="left">&#x02003;PAB2048/0133</td><td align="left">amt-szyqvdcebrhuj- - - - -in-</td><td align="left">Tryptophan synthase &#x003b2; subunit</td><td></td></tr><tr><td align="left">&#x02003;PAB0250/0031</td><td align="left">- - - -szyqvdcebrhuj- - - - - -n-</td><td align="left">Cysteine synthase</td><td align="left"><italic>P. abyssi</italic> appears to be the only euryarchaeon that encodes the typical cysteine biosynthesis pathway which it shares with <italic>A. pernix</italic></td></tr><tr><td align="left">&#x02003;PAB1595/2046</td><td align="left">a- - -szyq-dc-b- - -j- - - - - - - -</td><td align="left">ATP sulfurylase</td><td></td></tr><tr><td align="left">&#x02003;PAB0781/0529</td><td align="left">a- - -szyq-dcebr- -j- - - - - -n-</td><td align="left">Adenylylsulfate kinase</td><td></td></tr><tr><td align="left">&#x02003;PAB1839/0035</td><td align="left">- -t-szyqvdcebrh-jwgp-l-n-</td><td align="left">Uracil phosphoribosyltransferase</td><td></td></tr><tr><td align="left">&#x02003;PAB2246/0827</td><td align="left">-m- -s- - - - - - -brhu-w-p- - - - -</td><td align="left">Adenine-specific DNA methyltransferases</td><td></td></tr><tr><td align="left">&#x02003;PAB2154/0610</td><td align="left">amt-s- - - - - -e- -hujw-p- - -n-</td><td align="left">Restriction enzymes type I helicase subunit</td><td></td></tr><tr><td align="left" colspan="4">Present in <italic>P. horikoshii</italic> but not <italic>P. abyssi</italic></td></tr><tr><td align="left">&#x02003;PH0369/0153</td><td align="left">- - -k- -y-v- -ebrh- - - - - -I---</td><td align="left">Galactokinase</td><td></td></tr><tr><td align="left">&#x02003;PH1048, PH1046/2309</td><td align="left">- - -k-z-q-d- -b- - - - - - -o- - - -</td><td align="left">Leucyl aminopeptidase (aminopeptidase T)</td><td></td></tr><tr><td align="left">&#x02003;PH0365/1085</td><td align="left">a- -k- -y-v- -e-rh- - - - - - - - - -</td><td align="left">Galactose-1-phosphate uridylyltransferase</td><td></td></tr><tr><td align="left">&#x02003;PH0896/1230</td><td align="left">- - -k- -yqvd-eb- - -j- - - - - -n-</td><td align="left">Co/Zn/Cd efflux system component</td><td></td></tr><tr><td align="left">&#x02003;PH0162/1353</td><td align="left">amtk- - -qv- - - -r- - - - - - - - - - -</td><td align="left">Predicted hydrolase of the HD superfamily</td><td></td></tr><tr><td align="left">&#x02003;PH1032/0338</td><td align="left">-m-k- - - - - -ce- -hu- - - - -I- - -</td><td align="left">Site-specific DNA methylase dam</td><td></td></tr><tr><td align="left">&#x02003;PH0873/1401</td><td align="left">- -tk- - -q-dceb-uj- - - - - - - -</td><td align="left">GTPase subunit (McrB) of a restriction endonuclease</td><td></td></tr><tr><td align="left">&#x02003;</td><td></td><td></td><td></td></tr></tbody></table><table-wrap-foot><p>^*The designations are as in Table <xref ref-type="table" rid="T6">6</xref>.</p></table-wrap-foot></table-wrap></sec></sec><sec><title>Conclusions</title><p>The annotation of a new genome is likely to be as good as the database(s) to which it is compared. The COG database was constructed on the phylogenetic principle of protein classification, namely clustering by (probable) orthology. In addition, considerable effort has been invested in the functional characterization and classification of the COGs. As a result, using the COGs for annotating new genomes of organisms that do not belong to already well-characterized groups provides for numerous functional predictions that are not readily attained in more routine annotation protocols. Furthermore, taking advantage of the structure of the COG database, it is possible to reveal the main functional systems of an organism and its probable evolutionary affinities, and to systematically uncover sets of genes whose presence or absence in the given genome is unexpected and informative from an evolutionary standpoint.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Sequence data and databases</title><p>The genome sequences and the sets of annotated proteins from <italic>Aeropyrum pernix</italic> and <italic>Pyrococcus abyssi</italic> were retrieved from the Genomes division of the Entrez system [<xref ref-type="bibr" rid="B41">41</xref>]. The COG database, as of 1 June, 2000, consisted of 2,112 COGs that included 26,919 out of the total of 43,897 proteins from 21 completely sequenced bacterial, archaeal and eukaryotic genomes. This release of the COG database included the following genomes. Bacteria: <italic>Aquifex aeolicus</italic> (Aa), <italic>Bacillus subtilis</italic> (Bs), <italic>Borrelia burgdorferi</italic> (Bb), <italic>Chlamydia trachomatis</italic> (Ct), <italic>Chlamydia pneumoniae</italic> (Cp), <italic>Escherichia coli</italic> (Ec), <italic>Haemophilus influenzae</italic> (Hi), <italic>Helicobacter pylori</italic> (Hp), <italic>Mycoplasma genitalium</italic> (Mg), <italic>Mycoplasma pneumoniae</italic> (Mp), <italic>Mycobacterium tuberculosis</italic> (Mtu), <italic>Synechocystis</italic> PCC6803 (Ssp), <italic>Thermotoga maritima</italic> (Tm), <italic>Treponema pallidum</italic> (Tp). Archaea: <italic>Archaeoglobus fulgidus</italic> (Af), <italic>Methanobacterium thermoautotrophicum</italic> (Mth), <italic>Methanococcus jannaschii</italic> (Mj) and <italic>Pyrococcus horikoshii</italic> (Ph). Eukaryotes: yeast, <italic>Saccharomyces cerevisiae</italic> (Sc). The following recently sequenced genomes were included in the present analysis, to become available in the new release of the COGs. Bacteria: <italic>Campylobacter jejunii</italic> (Cj), <italic>Deinococcus radiodurans</italic> (Dr), <italic>Neisseria meningitidis</italic> (Nm) and <italic>Ureaplasma urealyticum</italic> (Uu). Archaea: <italic>Aeropyrum pernix</italic> (Ap) and <italic>Pyrococcus abyssi</italic> (Pa). In addition to the COG database, the nonredundant (NR) database of protein sequences at the National Center for Biotechnology Information (NIH, Bethesda) was used for sequence similarity searches.</p></sec><sec><title>Sequence analysis and assignment of proteins to COGs</title><p>Protein sequence similarity searches were performed using the gapped BLASTP program or, for detecting subtle similarities, the position-specific iterative BLAST (PSI-BLAST) program, with either an individual protein or a collection of pre-computed position-specific scoring matrices used as a query [<xref ref-type="bibr" rid="B42">42</xref>]. Searches of nucleotide sequences translated in six frames were performed using the gapped TBLASTN program [<xref ref-type="bibr" rid="B42">42</xref>]. Regions of low complexity in protein sequences were identified using the SEG program [<xref ref-type="bibr" rid="B43">43</xref>].</p><p>Proteins were assigned to COGs using the COGNITOR program essentially as previously described [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. Briefly, each protein sequence from the analyzed genomes was compared to the protein sequences that comprise the COGs using the gapped BLASTP program and genome-specific best hits (BeTs) were registered. A protein was included in a COG when at least three BeTs were to the members of the given COG. In cases when two or more COGs met this criterion, the COG with the greater alignment scores was given priority. If no COGs satisfied this condition, an ambiguous result was reported. The new version of the COGNITOR program used in this analysis detects position-specific BeTs so that distinct domains of multidomain proteins were assigned to different COGs when justified by the above criteria.</p><p>Proteins from the analyzed genomes that could not be included in the pre-existing COGs were subjected to the original COG construction procedure [<xref ref-type="bibr" rid="B12">12</xref>]. Specifically, a new COG was formed when a triangle of consistent reciprocal BeTs was identified, which occurs when one protein from each of three distinct species gives as a genome-specific best hit the other two proteins, and vice versa. In the present analysis, only such elementary new COGs could be identified because those COGs that would include a greater number of species have already been created.</p></sec><sec><title>Manual validation of preliminary COG assignments</title><p>Each preliminary assignment was checked for validity, and adjustments were made as necessary. An assignment of a particular protein to a given COG was considered correct if: (1) the genome-specific best hits for that protein were members of the COG; (2) the size of the query protein was similar to the sizes of the COG members; (3) the region of similarity between the query and the COG members was extensive; (4) the similarity scores indicated statistical significance.</p><p>For cases that did not fit some of the straightforward criteria given above, other criteria or methods were used to confirm an assignment. For example, PSI-BLAST would be used to determine if the query is a diverged member of the COG, or sequence alignments would be examined to determine if the query shares conserved motifs with the COG members.</p><p>False positives that required reassignment generally fulfilled the criteria above, but failed to meet either condition (1) or condition (2). In the former case, some of the genome-specific best hits were to a different COG. In the latter case, the query protein merely contained a single domain of a COG whose members were composed of multiple domains. These proteins were assigned to COGs that contain the single domain only.</p><p>Certain false positives required removal from COGs. Although fulfilling the three-BeT criterion for inclusion, these assignments probably arose by chance as indicated by several lines of evidence: first, the genome-specific hits that caused the apparently incorrect assignment had lower alignment scores than those for non-COG proteins; second, the statistical significance of the hits was very low; third, the hits were non-reciprocal; fourth, the alignments with the COG members did not include diagnostic motifs of the respective protein family; and fifth, the protein was assigned to a COG based on hits to low-complexity (typically, coiled-coil) regions.</p><p>Proteins with multiple domains that hit members from more than one COG were manually divided according to the COGNITOR results. The sub-sequences were given separate FASTA entries in the COG database, and the domains were listed as members of the appropriate COG.</p></sec><sec><title>Classification of genomes by co-occurrence in COGs</title><p>The table of co-occurrence of genomes in COGs is available on the COG website [<xref ref-type="bibr" rid="B44">44</xref>]. These data were used for classifying genomes by cluster analysis and phylogenetic tree construction. The distance between genomes was calculated as Dij = 1 - (Cij/Ni + Nj - Cij), where Cij is the number of COGs in which genomes <italic>i</italic> and <italic>j</italic> co-occur, and Ni, Nj are the numbers of COGs that include the genomes i and j, respectively. This formula employs the Jackard co-occurrence coefficient, which is widely used in biometric studies to obtain comparable results for samples of different size [<xref ref-type="bibr" rid="B45">45</xref>]. Cluster dendrograms were generated using the UPGMA option of the NEIGHBOR program, and distance-matrix trees were generated using the FITCH program [<xref ref-type="bibr" rid="B46">46</xref>] or the Neighbor-Joining [<xref ref-type="bibr" rid="B47">47</xref>] option of the NEIGHBOR program. All tree-building programs are parts of the PHYLIP package [<xref ref-type="bibr" rid="B48">48</xref>].</p></sec></sec><sec><title>Additional data</title><p>The results of the analysis of the <italic>A. pernix</italic> and <italic>P. abyssi</italic> genomes described here are included, along with those for several recently sequenced bacterial genomes, on the COGs website [<xref ref-type="bibr" rid="B44">44</xref>] and a file of a <xref ref-type="supplementary-material" rid="S1">list of COGs</xref> is included. A list of <italic>A. pernix</italic><xref ref-type="supplementary-material" rid="S2">genes</xref>, in which genes overlapping with COG members and thought to be spurious are flagged, and newly detected genes have been added, and the corresponding FASTA library of <xref ref-type="supplementary-material" rid="S3">protein sequences</xref> are available by anonymous ftp (<ext-link ext-link-type="ftp" xlink:href="ftp:// ncbi.nlm.nih.gov/pub/koonin/Apernix"/>) and are included as text files.</p></sec><sec sec-type="supplementary-material"><title>Supplementary Material</title><supplementary-material content-type="local-data" id="S1"><caption><title>Additional data file 1</title></caption><media xlink:href="gb-2000-1-5-research0009-S1.txt" mimetype="text" mime-subtype="plain"><caption><p>The list of COGs as of August 15, 2000</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="S2"><caption><title>Additional data file 2</title></caption><media xlink:href="gb-2000-1-5-research0009-S2.txt" mimetype="text" mime-subtype="plain"><caption><p>A list of <italic>A. pernix</italic> genes, in which genes overlapping with COG members are flagged</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="S3"><caption><title>Additional data file 3</title></caption><media xlink:href="gb-2000-1-5-research0009-S3.txt" mimetype="text" mime-subtype="plain"><caption><p>Protein sequences of <italic>A. pernix</italic> genes</p></caption></media></supplementary-material></sec>

Accessing and distributing EMBL data using CORBA (common object request broker architecture)

<sec><title>Background:</title><p>The EMBL Nucleotide Sequence Database is a comprehensive database of DNA and RNA sequences and related information traditionally made available in flat-file format. Queries through tools such as SRS (Sequence Retrieval System) also return data in flat-file format. Flat files have a number of shortcomings, however, and the resources therefore currently lack a flexible environment to meet individual researchers' needs. The Object Management Group's common object request broker architecture (CORBA) is an industry standard that provides platform-independent programming interfaces and models for portable distributed object-oriented computing applications. Its independence from programming languages, computing platforms and network protocols makes it attractive for developing new applications for querying and distributing biological data.</p></sec><sec><title>Results:</title><p>A CORBA infrastructure developed by EMBL-EBI provides an efficient means of accessing and distributing EMBL data. The EMBL object model is defined such that it provides a basis for specifying interfaces in interface definition language (IDL) and thus for developing the CORBA servers. The mapping from the object model to the relational schema in the underlying Oracle database uses the facilities provided by Persistence^TM, an object/relational tool. The techniques of developing loaders and 'live object caching' with persistent objects achieve a smart live object cache where objects are created on demand. The objects are managed by an evictor pattern mechanism.</p></sec><sec><title>Conclusions:</title><p>The CORBA interfaces to the EMBL database address some of the problems of traditional flat-file formats and provide an efficient means for accessing and distributing EMBL data. CORBA also provides a flexible environment for users to develop their applications by building clients to our CORBA servers, which can be integrated into existing systems.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Wang</surname><given-names>Lichun</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Rodriguez-Tom&#x000e9;</surname><given-names>Patricia</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Redaschi</surname><given-names>Nicole</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>McNeil</surname><given-names>Phil</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Robinson</surname><given-names>Alan</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Lijnzaad</surname><given-names>Philip</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>The EMBL (European Molecular Biology Laboratory) Nucleotide Sequence Database (often referred to as the EMBL database) [<xref ref-type="bibr" rid="B1">1</xref>] is hosted at the European Bioinformatics Institute (EBI). It is a comprehensive database of DNA and RNA sequences that are directly submitted from researchers and genome sequencing groups, and collected from the scientific literature and patent applications. It is produced in an international collaboration with GenBank (NCBI, Bethesda, USA) and DDBJ (the DNA Data Bank of Japan, CIB, Mishima, Japan). Each of the three collaborating groups collects a portion of the total sequence data reported worldwide, and all new and updated database entries are exchanged daily. The amount of sequence data is growing exponentially.</p><p>As our scientific understanding deepens, the complexity of the related information increases as well. As a result, the structure of the data also keeps changing. The EMBL database is managed and maintained using the relational database management system (DBMS) Oracle. It contains over 130 tables and 140 relationships, having around 80 Gigabytes (Gb) of data comprising nearly 10 million objects of primary data and millions of sub-objects called 'features'. Traditionally, the sequences and related information, which have been collected over a long period of time, are made available in flat-file format via ftp, CD-ROM, www tools, and so on. The queries through tools such as SRS (Sequence Retrieval System, a network browser for databanks in molecular biology) [<xref ref-type="bibr" rid="B2">2</xref>] also return data in flat-file format. However, flat files have a number of shortcomings: the format may not be described formally; it is difficult to represent complex data and relationships, the meaningful units of information ('objects') are not represented or handled well; it is hard to retrieve objects separately; assembly of objects into bigger aggregates is difficult; elaborate parsing is often required; and so on. In general, the current availability of the resources is not matched by a flexible environment to meet individual researchers' needs.</p><p>An industry standard, the Object Management Group's (OMG) common object request broker architecture (CORBA), provides platform-independent programming interfaces and models for portable distributed object-oriented computing applications [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. Its independence from programming languages, computing platforms and network protocols provides a solution for developing new applications for querying and distributing biological data [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>], which can also be integrated into existing systems. Here we present a CORBA infrastructure developed at EMBL-EBI and show that the CORBA interfaces to the EMBL database address some of the limitations of the flat-file format and provide an efficient means for accessing and distributing EMBL data. CORBA also provides a flexible environment for users to develop application programs (for example, for sequence analysis or data mining).</p></sec><sec><title>Results and discussion</title><sec><title>EMBL object data model</title><p>The diversity and structure of biological data complicate their use. To develop a CORBA server that provides access to our biological data, we need a well-defined object model to model the real-world biological entities, that is, to describe the structure and constraints present in the data, as well as how the data can be accessed and queried. It is a specification of the data in the problem domain, independent of how the actual database is implemented. This model can then be expressed in IDL (interface definition language) interfaces to the CORBA server at one end, and mapped to a database schema for the underlying data management and storage at the other end.</p><p>We use the unified modeling language (UML) notation [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>] for this model. According to the UML, a model is organized into packages. A package groups classes that are semantically related and a dependency indicates that a package uses classes from another package. Each class belongs to exactly one package. A class is a descriptor for a set of objects that have similar properties, behavior and relationships to other objects. An attribute is a named property of a class. An attribute can be derived, that is, its value is computed from the values of other attributes. An operation is a procedure attached to a class, describing the behavior of the class. A class can be created by inheriting all attributes and operations from one or more of its superclasses. An association is a description of links or a set of links that specify connections among objects. An association can be reflexive, connecting a class with itself. Multiplicity defines how many times one object may link to another through an association. An association class has both association and class properties. It can be seen as an association that also has class properties, or as a class that also has association properties. It holds data that are relevant for the association, but for neither of the associated classes alone. A structured (composite) data type is represented as a class (as usual in object-oriented (OO) modeling). A class is used as an attribute type mainly if its role is solely to bundle simple data into a composite type (for example, Date). Multi-valued attributes are represented by instances of a parameterized class Coll{Type} (for example, Coll{string}).</p><p>Our object model of the EMBL database is organised into five main packages, as shown in Figure <xref ref-type="fig" rid="F1">1</xref>, where each package holds a set of closely related classes with a common purpose. The packages are: Sequence Info, classes representing biological sequences, general information about these sequences and administrative data associated with database entries; Feature Info, classes representing detailed sequence annotation (known as sequence features); Reference Info, classes representing bibliographic references that hold information about the sequences; Taxonomy Info, classes representing the taxonomy of the organisms from which the sequences were obtained; Location Info, classes representing locations on sequences.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>The database partitioning. The database is divided into five main packages: <italic>Sequence Info</italic>, all general information about sequences; <italic>Feature Info</italic>, detailed sequence annotation; <italic>Reference Info</italic>, bibliographic references; <italic>Taxonomy Info</italic>, the taxonomy of the organisms from which the sequences were obtained; <italic>Location Info</italic>, representing locations on sequences.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-1"/></fig><p>There is one additional package, Types, which holds classes representing all the special data types used in various parts of the model. Each package contains a relatively isolated part of the entire object model, and is a clear candidate for re-use in models for other databases.</p><p>Figure <xref ref-type="fig" rid="F2">2</xref> gives the definition of Sequence Info only. A full definition of the EMBL Nucleotide Sequence Database object model [<xref ref-type="bibr" rid="B16">16</xref>] can be found in <xref ref-type="supplementary-material" rid="S1">Additional data file 1</xref> with the online version of this article. The package Sequence Info defines class BioSeq, which represents biological sequences, and class Seqlnfo, which describes general information about these sequences. The administrative data associated with database entries are defined in Entrylnfo. The biological classes of sequence NSDBSeq, for nucleotide sequences, and PIDSeq, for protein sequences, are subclasses of BioSeq. VirtualSeq and PhysicalSeq are storage classes of sequence, that is, virtual or literal.</p><fig position="float" id="F2"><label>Figure 2</label><caption><p>Sequence Info. This package defines class <italic>BioSeq</italic>, which represents biological sequences, and class <italic>SeqInfo</italic>, which describes general information about these sequences. The administrative data associated with database entries are defined in <italic>EntryInfo</italic>. The biological classes of sequence <italic>NSDBSeq</italic>, which is for nucleotide sequences, and <italic>PIDSeq</italic>, which is for protein sequences, are subclasses of <italic>BioSeq</italic>. <italic>VirtualSeq</italic> and <italic>PhysicalSeq</italic> are storage classes of sequence, that is, virtual or literal.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-2"/></fig><p>The definition of some biological entities is prone to change because of the rapid developments in molecular biology. Any change made to the structure of the model needs to be propagated to both the IDL specification that defines the CORBA server interfaces and underlying relational schema. To handle this problem, a strategy of using both explicit model and meta model is employed in defining Feature Info. The structure of the model is therefore not affected by changes to the feature definition, which makes it suitable for defining stable IDL interfaces.</p><p>The EMBL CORBA server mainly covers Sequence Info, Location Info and Feature Info, which are grouped into a big package that also includes Reference Info and Taxonomy Info. The reference and taxonomy servers are independent servers for Reference Info and Taxonomy Info. This paper is focused on the EMBL server.</p></sec><sec><title>System architecture and CORBA development</title><p>The system architecture is shown in Figure <xref ref-type="fig" rid="F3">3</xref>. On the server side, CORBA implementation objects access and query the relational database via Persistence^TM (Persistence Software) [<xref ref-type="bibr" rid="B17">17</xref>], which acts as a middleware between our CORBA implementation and the Oracle database. To allow invoking operations on the objects, the server provides its clients interfaces in OMG IDL, which is independent of the server implementation. An object's interface is composed of the operations and types of data that can be passed to and from those operations. Clients access the CORBA objects via operation calls through an object request broker (ORB), where the distribution details are handled by the ORB.</p><fig position="float" id="F3"><label>Figure 3</label><caption><p>System architecture. On the server side, CORBA implementation objects access and query the relational database via Persistence^TM, which is a middleware between our CORBA implementation and the Oracle database. To allow invoking operations on the objects, the server provides its clients interfaces in OMG IDL, which is independent of the server implementation. An object's interface is composed of the operations and types of data that can be passed to and from those operations. Clients access the CORBA objects via operation calls through an Object Request Broker (ORB) where the distribution details are handled by the ORB.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-3"/></fig><p>The CORBA development overview is shown in Figure <xref ref-type="fig" rid="F4">4</xref>. CORBA object interfaces together with their operations and type of data are defined in IDL. For the ORB, we have chosen IONA Inc's C++ ORB, Orbix^TM [<xref ref-type="bibr" rid="B18">18</xref>]. Its IDL compiler generates skeleton code and stub code in C++. We provide the server object implementation code and the Persistence application code. These codes are subsequently compiled and linked together to become executable. Clients can be written in any language for which an ORB and IDL compiler are available, including Ada, C, C++, COBOL, CommonLisp, Eiffel, Java, Python, Peri, SmallTalk, Tcl, and so on. Note that we do not use the new features in CORBA 2.3, as ORBs that implement CORBA 2.3 have only become available recently.</p><fig position="float" id="F4"><label>Figure 4</label><caption><p>CORBA development overview. CORBA object interfaces together with their operations and type of data are defined in IDL. For the ORB, we have chosen IONA Inc's C++ ORB, Orbix^TM. Its IDL compiler generates skeleton code and stub code in C++. We provide the server object implementation code and the Persistence application code. These codes are subsequently compiled and linked together to become executable. Clients can be written in any language for which an ORB and IDL compiler are available, including Ada, C, C++, COBOL, CommonLisp, Eiffel, Java, Python, Perl, SmallTalk, Tcl, and so on.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-4"/></fig><sec><title>IDL definition</title><p>The OMG IDL is CORBA's fundamental abstraction mechanism for separating object interfaces from their implementations [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. It allows object interfaces to be defined in a manner that is independent of any particular programming language. It establishes a contract between client and server that describes the types and object interfaces used by an application. IDL definitions focus on object interfaces, the operations supported by those interfaces, and exceptions that might be raised by the operations. As data can only be exchanged between client and server if their types are defined in IDL, typically a large part of an IDL is concerned with the definition of data types. An interface can inherit from one or more other interfaces.</p><p>Following the EMBL data model, the IDL definition for the EMBL server comprises three IDL files: nsdb.idl, seqdb.idl and types.idl. The nsdb.idl defines the EMBL-specific sequences and related information and includes seven interfaces in the module nsdb: Entrylnfo, Embl, EmblSeq, NucSeq, NucFeature, Location and FeatureLocation. The seqdb.idl defines the module seqdb that includes three interfaces: BioSeq, Seqlnfo and Feature, which contain more general biological sequence information. The nsdb.idl and seqdb.idl use basic types defined in types.idl.</p><p>To reflect the accessing and querying of data, operations are defined in such a way that the return values of the operations represent attributes in the EMBL object data model. This supports 'creating objects on demand'. These objects are instances of 'data classes', which are the results of queries. Figure <xref ref-type="fig" rid="F5">5</xref> gives the IDL specification of interfaces BioSeq and Seqlnfo in module seqdb, and Entrylnfo, NucSeq and EmblSeq in module nsdb (the full IDL definition can be found in the Additional data file with the online version of this article) [<xref ref-type="bibr" rid="B19">19</xref>]). NucSeq inherits from seqdb::BioSeq and EmblSeq inherits from NucSeq, seqdb::SeqInfo, and Entrylnfo.</p><fig position="float" id="F5"><label>Figure 5</label><caption><p>Part of module <italic>seqdb</italic> and <italic>nsdb</italic>, extracted from the EMBL IDL specification. The interfaces <italic>BioSeq</italic> and <italic>SeqInfo</italic> are defined in module <italic>seqdb</italic>; The interfaces <italic>EntryInfo, NucSeq</italic> and <italic>EmblSeq</italic> are defined in module <italic>nsdb. NucSeq</italic> inherits from <italic>seqdb::BioSeq</italic> and <italic>EmblSeq</italic> inherits from <italic>NucSeq, seqdb::SeqInfo,</italic> and <italic>EntryInfo.</italic></p></caption><graphic xlink:href="gb-2000-1-5-research0010-5"/></fig></sec><sec><title>Class relationships</title><p>Each IDL interface is mapped into a class in C++ by the orb's IDL compiler (in our case, the ORB is iona's Orbix), and operations are mapped to member functions of the class. For the above interfaces BioSeq and Seqlnfo, we have two mapped classes as shown in Figure <xref ref-type="fig" rid="F6">6</xref>.</p><fig position="float" id="F6"><label>Figure 6</label><caption><p>Classes of <italic>BioSeq</italic> and <italic>SeqInfo</italic>. The class BioSeq has four methods: getBioSeqId, getLength, getAnySeq and getBioSeqVersion. The returned values of methods represent attributes of class BioSeq defined in the object model, providing information on biological sequences. The class SeqInfo has methods: getDescription, getKeywords, getComments, getDbXrefs and getReferences, representing general information on the sequences.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-6"/></fig><p>The module itself is also mapped into a class here. Although it can normally be mapped to a namespace, our C++ compiler (Sun's SparcWorks 4.2) does not support namespaces. The relationship between classes is shown in Figure <xref ref-type="fig" rid="F7">7</xref> in UML notation. The seqdb consists of three classes: Seqlnfo, BioSeq, and Feature. The nsdb class comprises four classes: EmblSeq, NucFeature, FeatureLocation and Embl. EmblSeq inherits from the classes of Entrylnfo, Seqlnfo and NucSeq that in turn inherits from BioSeq. FeatureLocation inherits from Location, and NucFeature inherits from Feature.</p><fig position="float" id="F7"><label>Figure 7</label><caption><p>Class relationship in UML notation. The class <italic>seqdb</italic> consists of 3 classes: <italic>SeqInfo</italic>, <italic>BioSeq</italic>, and <italic>Feature</italic>. The class <italic>nsdb</italic> comprises 4 classes: <italic>EmblSeq, NucFeature, FeatureLocation</italic> and <italic>Embl. EmblSeq</italic> inherits from the classes of <italic>EntryInfo, SeqInfo</italic> and <italic>NucSeq</italic> that in turn inherits from <italic>BioSeq. FeatureLocation</italic> inherits from <italic>Location</italic>, and <italic>NucFeature</italic> inherits from <italic>Feature</italic>.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-7"/></fig></sec><sec><title>Object-relational mapping</title><p>The EMBL CORBA server provides its clients with an object-oriented interface to the EMBL database. To achieve this, the object model needs to be mapped to the schema of the underlying Oracle relational database.</p><p>Persistence^TM, an object/relational tool from Persistence Software [<xref ref-type="bibr" rid="B17">17</xref>], is a mediator for transforming object operations to relational database calls and vice versa. It maps objects to relational rows and manages the objects in a shared cache, called the live object cache. It uses a proprietary object model description that maps classes to tables, objects to rows, attributes to columns and associations to foreign keys.</p><p>For inheritance relationships (single inheritance only), Persistence insists on a so-called horizontal mapping for performance reasons; that is, in the class hierarchy, only leaf nodes are represented by real database tables (or views). Non-leaf class objects are obtained as projections of the leaf class tables. However, for maximum flexibility, our existing database schema, which is independent and developed prior to the CORBA development, uses the so-called vertical mapping. In this case, each node in the class hierarchy has its own table, with subclass tables having no superclass attributes; their primary keys are also foreign keys to the superclass table. As our objects provide read-only access, it is possible to set up relational views that transform our tables into the horizontal object-to-relational mapping that is required by Persistence. This allows developers to create hierarchies of related objects from 'flat' tables. For example, for the class NsdbSeq that is inherited from BioSeq, its so-called Persistence horizontal object-to-relational mapping (view) is shown in Figure <xref ref-type="fig" rid="F8">8</xref>. This view is built from a number of tables (or views) as shown in Figure <xref ref-type="fig" rid="F9">9</xref>. The CORBA class EmblSeq is mapped to the view of NsdbSeq.</p><fig position="float" id="F8"><label>Figure 8</label><caption><p>The NsdbSeq view in Persistence. As the class NsdnSeq inherits from the class BioSeq, the view of NsdbSeq in Persistence therefore has attributes defined in both NsdbSeq and BioSeq, representing information on nucleotide sequences.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-8"/></fig><fig position="float" id="F9"><label>Figure 9</label><caption><p>Tables for the NsdbSeq view in the actual database. The view is built from a number of tables (or views) of the database.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-9"/></fig><p>Using the data model and schema description, Persistence can also offer an automatic generation of an IDL specification as well as a complete CORBA server. We have not used this facility, however, as we want full control over the IDL specification. This approach has an advantage in using views even when a one-to-one mapping from IDL to EMBL tables remains. When the structure of data changes at the database side as a result of the increasing complexity of biological data or the availability of new modeling capability in the database, we need only a change on the underlying views. We much less frequently require a change on the Persistence mapping and code as these can still map the changed tables or data to the same objects at the CORBA side.</p></sec></sec><sec><title>Object management</title><p>A 'live object cache' is a notion used in Persistence [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. The basic model for managing live objects is to cache data instances read from the database, to register their primary key values, and to respond to queries based on the cached data. As tuples are retrieved from the database, they are converted to objects and 'knitted' together according to the object-model mapping to form a network of in-memory objects. A live object cache maps information from relational tables into objects. Accessing and manipulating these objects in the live object cache is faster than querying the relational database, speeding up application performance considerably. Persistence can also ensure data integrity with appropriate locking and transaction management.</p><p>There are roughly three types of objects involved here: persistent objects, live objects, and CORBA objects. Here a persistent object is referred as a 'data object' in the database. A live object is an in-memory object in the live object cache. A CORBA object is a CORBA implementation object defined in IDL. Creation of a CORBA object is called instantiation. When a persistent object is loaded into memory, it becomes a live object. A CORBA object owns one or more live objects. Note that the orb's object adapter, no matter whether a basic object adapter (BOA) or portable object adapter (POA), only serves as the glue between CORBA objects and the ORB. It is an object that adapts the interface of one object to a different interface expected by a caller and allows the caller to invoke requests on an object without knowing the object's true interface. Although the future CORBA may include garbage collection, the management of objects is currently at the application developer's discretion. This section discusses the management of objects.</p><sec><title>Creation of live objects</title><p>Data kept in the database are only loaded into the cache on an as-needed basis. We employ Orbix's loader techniques [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B21">21</xref>] together with Persistence's live object caching [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B20">20</xref>] to build our loaders to support the creation of objects in the live object cache. When an operation invocation arrives at the process, Orbix ORB searches for the target object in the process's object table. Loaders are called when an object reference enters an address space via a function findMe, and Persistence live objects are then loaded. If no live objects available in the live object cache respond to the call, Persistence will create a new live object via querying and accessing the relational database.</p></sec><sec><title>Eviction of CORBA objects</title><p>Objects are created at the clients' request. When the server has been running for some period of time, possibly weeks or months, it will have created a number of CORBA objects, which in turn contain a number of Persistence live objects, and will consume the memory space. Some of them will not be needed any more. The evictor pattern [<xref ref-type="bibr" rid="B3">3</xref>] describes a general strategy for limiting memory consumption. The basic idea is that we use an object manager to instantiate objects on demand. However, instead of blindly instantiating a new object every time, the object manager checks instantiated objects in the pool that it manages. If the called object is already in the pool, it can be used directly. If not, it will check if the number of objects has reached a specified limit. If so, the object manager will evict an older instantiated object and then instantiate a new one for the current request. Consequently, the older related Persistence live objects will also be deleted from the cache and new live objects will be loaded in if requested.</p><p>One more interesting issue of the evictor pattern is how to choose which object to evict. There are a number of possible strategies, such as least recently used (LRU), least frequently used (LFU), evicting the object with the highest memory consumption (HMC), or using a weighted function that chooses an object for evictor based on a combination of factors (WF). We use a simple LRU algorithm to implement the evictor and prove it is effective.</p></sec></sec><sec><title>Accessibility of the EMBL database</title><p>When the CORBA server is up and running, a client, which can be developed using any CORBA-compliant ORB on the user's preferred environment and language (for which the ORB is available) at any local or remote machine, can access EMBL data through these objects using an IOR (Interoperable Object Reference) or via a Naming Service. We have published our EMBL server IOR [<xref ref-type="bibr" rid="B22">22</xref>] and its naming as 'databases/EMBL/nsdb/Embl', which is registered with the Naming Service [<xref ref-type="bibr" rid="B23">23</xref>]. We have also provided a number of demonstration clients for the EMBL server [<xref ref-type="bibr" rid="B24">24</xref>].</p><p>The client submits its query with bio-seq-id to the Embl object, which is a factory object representing the whole database. It invokes the operation findMe provided by EmblSeq object, which in turn invokes the loader object. The EmblSeq reference is returned to the client. Once the client has the EmblSeq object reference, it can then invoke the methods provided by EmblSeq to get the sequence information defined in Seqlnfo. The object attributes are obtained through invoking the methods. Further queries can be made through the invocation to other methods. The access is shown in Figure <xref ref-type="fig" rid="F10">10</xref>.</p><fig position="float" id="F10"><label>Figure 10</label><caption><p>Access to the EMBL database via the CORBA server. The client submits its query with <italic>bio-seq-id</italic> to the <italic>Embl</italic> object, which is a factory object representing the whole database. It invokes the operation findMe provided by <italic>EmblSeq</italic> object, which in turn invokes the loader object. The <italic>EmblSeq</italic> reference is returned to the client. Once the client has the <italic>EmblSeq</italic> object reference, it can then invoke the methods provided by <italic>EmblSeq</italic> to get the sequence information defined in <italic>SeqInfo</italic>. The object attributes are obtained through invoking the methods. Further queries can be made through the invocation to other methods.</p></caption><graphic xlink:href="gb-2000-1-5-research0010-10"/></fig><p>From Figure <xref ref-type="fig" rid="F10">10</xref>, it can be seen that CORBA interfaces to the EMBL database provide: meaningful units of information: objects; encapsulated methods, for example getLength(); interoperation between objects; easy access and distribution of data; and easy to comply with the standard. It provides a basis for developing further biological research tools.</p><p>Accessibility to the EMBL database via the EMBL CORBA server can be as fine as any small attribute defined in the EMBL data model. The EMBL CORBA server can also provide a blob object that contains a number of objects. Currently the EMBL server is undergoing a trial with internal and external users. There are increasing numbers of users developing their applications using our CORBA server.</p></sec></sec><sec><title>Conclusions</title><p>This paper presents a CORBA infrastructure developed at EMBL-EBI. The EMBL object model provides a basis to develop the CORBA server. Employing Persistence^TM maps the object model to the relational schema in the underlying Oracle database. To present Persistence with the right relations, views have been used to transform the vertically mapped tables to horizontal ones. Properly built loaders make use of the technique of 'live object caching' and enhance the performance. The evictor pattern is used for memory management. It has been demonstrated that the CORBA server addresses some problems of the flat-file format and provides a solution to accessing and distributing EMBL sequence data. It also provides a flexible and scalable environment for users to develop their applications by building clients.</p><p>The future work will include migrating the implementation of the EMBL server to comply with the emerging standard - OMG standard for biosequences. By OMG rules, the EBI, as a co-submitter on the Biomolecular Sequence Analysis (BSA) standard, is obliged to implement the standard. As the BSA standard proposal is not fully compatible with the EMBL IDL specification currently used, care will have to be taken to make this transition as easy as possible for existing clients.</p></sec><sec><title>Additional data</title><p>The following additional data are included with the online version of this article: <xref ref-type="supplementary-material" rid="S1">The EMBL Nucleotide Sequence Database object model</xref> and <xref ref-type="supplementary-material" rid="S2">The EMBL IDL specification</xref>.</p></sec><sec sec-type="supplementary-material"><title>Supplementary Material</title><supplementary-material content-type="local-data" id="S1"><caption><title>EMBL Nucleotide Sequence Database object model</title></caption><media xlink:href="gb-2000-1-5-research0010-S1.html" mimetype="text" mime-subtype="html"><caption><p>EMBL Nucleotide Sequence Database object model</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="S2"><caption><title>The EMBL IDL specification</title><p>It contains the IDL belonging to the EMBL CORBA servers:</p><p>The home page of the CORBA work at EBI is <ext-link ext-link-type="uri" xlink:href="http://corba.ebi.ac.uk/"/>. For questions and suggestions, please feel free to contact corba@ebi.ac.uk.</p></caption><media xlink:href="gb-2000-1-5-research0010-S2.idl" mimetype="text" mime-subtype="plain"><caption><p>general types used throughout the EMBL CORBA project (save as types.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S3.idl" mimetype="text" mime-subtype="plain"><caption><p>general sequence interfaces (save as seqdb.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S4.idl" mimetype="text" mime-subtype="plain"><caption><p>sequence interfaces specialised for EMBL (needs seqdb.idl) (save as nsdb.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S5.idl" mimetype="text" mime-subtype="plain"><caption><p>genetic code used for translation of nucleotide sequences in EMBL (save as geneticcode.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S6.idl" mimetype="text" mime-subtype="plain"><caption><p>the IDL for the reference server: access to the bibliographic references of EMBL (save as publication.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S7.idl" mimetype="text" mime-subtype="plain"><caption><p>the IDL for the taxonomy server: access to the taxonomy information of EMBL (save as taxonomy.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S8.idl" mimetype="text" mime-subtype="plain"><caption><p>meta information of the EMBL servers (save as meta.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S9.idl" mimetype="text" mime-subtype="plain"><caption><p>a file with only '#include' statements for the above IDL files, for easy compilation (save as embl.idl)</p></caption></media><media xlink:href="gb-2000-1-5-research0010-S10.txt" mimetype="text" mime-subtype="plain"><caption><p>Makefile is provided; it can be used to compile the IDL using one of three commonly used ORBs, i.e., ORBacus, Visibroker or OrbixWeb (save as Makefile). </p></caption></media></supplementary-material></sec>

Evidence for symmetric chromosomal inversions around the replication origin in bacteria

<sec><title>Background:</title><p>Whole-genome comparisons can provide great insight into many aspects of biology. Until recently, however, comparisons were mainly possible only between distantly related species. Complete genome sequences are now becoming available from multiple sets of closely related strains or species.</p></sec><sec><title>Results:</title><p>By comparing the recently completed genome sequences of <italic>Vibrio cholerae</italic>, <italic>Streptococcus pneumoniae</italic> and <italic>Mycobacterium tuberculosis</italic> to those of closely related species - <italic>Escherichia coli, Streptococcus pyogenes</italic> and <italic>Mycobacterium leprae,</italic> respectively - we have identified an unusual and previously unobserved feature of bacterial genome structure. Scatterplots of the conserved sequences (both DNA and protein) between each pair of species produce a distinct X-shaped pattern, which we call an X-alignment. The key feature of these alignments is that they have symmetry around the replication origin and terminus; that is, the distance of a particular conserved feature (DNA or protein) from the replication origin (or terminus) is conserved between closely related pairs of species. Statistically significant X-alignments are also found within some genomes, indicating that there is symmetry about the replication origin for paralogous features as well.</p></sec><sec><title>Conclusions:</title><p>The most likely mechanism of generation of X-alignments involves large chromosomal inversions that reverse the genomic sequence symmetrically around the origin of replication. The finding of these X-alignments between many pairs of species suggests that chromosomal inversions around the origin are a common feature of bacterial genome evolution.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Eisen</surname><given-names>Jonathan A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Heidelberg</surname><given-names>John F</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>White</surname><given-names>Owen</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Salzberg</surname><given-names>Steven L</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>Large-scale genomic rearrangements and duplications are important in the evolution of species. Previously, these large-scale genome-changing events were studied through genetic or cytological studies. With the availability of many complete genome sequences it is now possible to study such events through comparative genomics. The publication of the yeast genome has led to much better insight into the duplication events that have occurred in fungal and eukaryotic evolution (for example, see [<xref ref-type="bibr" rid="B1">1</xref>]). Large chromosomal duplications have also been found from analysis of completed chromosomes of <italic>Arabidopsis thaliana</italic> [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. The ability to detect large-scale genomic changes is dependent in large part on which genomes are available. Such studies in bacteria, for example, have been limited by the availability of genomes only from distantly related sets of species. Recently, however, the genomes of sets of closely related bacterial species have become available. We have compared these closely related bacterial genomes and have discovered an unusual phenomenon - alignments of whole genomes that show an X-shaped pattern (which we refer to as X-alignments). Here we present the evidence for these X-alignments and discuss mechanisms that might have produced them.</p></sec><sec><title>Results and discussion</title><sec><title>Whole-genome X-alignments between species at the DNA level</title><p>We compared the DNA sequences of the two chromosomes of <italic>Vibrio cholerae</italic> [<xref ref-type="bibr" rid="B4">4</xref>] with the sequence of the <italic>Escherichia coli</italic> chromosome [<xref ref-type="bibr" rid="B5">5</xref>] using a suffix tree alignment algorithm [<xref ref-type="bibr" rid="B6">6</xref>]. The analysis revealed a significant alignment at the DNA level between the <italic>V. cholerae</italic> large chromosome (chrI) [<xref ref-type="bibr" rid="B4">4</xref>] and the <italic>E. coli</italic> chromosome [<xref ref-type="bibr" rid="B5">5</xref>] spanning the entire length of these chromosomes (Figure <xref ref-type="fig" rid="F1">1a</xref>). Analysis of the reverse complement of <italic>V. cholerae</italic> chrI with <italic>E. coli</italic> also produced a significant alignment (Figure <xref ref-type="fig" rid="F1">1b</xref>). When superimposed, the two alignments produce a clear 'X' shape (Figure <xref ref-type="fig" rid="F1">1c</xref>) that is symmetric about the origin of replication of both genomes. This symmetry indicates that matching sequences tend to occur at the same distance from the origin but not necessarily on the same side of the origin. The X-alignment between <italic>V. cholerae</italic> and <italic>E. coli</italic> was found to be statistically significant using a test based on the number of matches found in diagonal strips in the alignment (see the Materials and methods section). Specifically, when <italic>V. cholerae</italic> chrI is aligned in the forward direction against <italic>E. coli</italic>, there are 459 maximal unique matching subsequences (MUMs; see the Materials and methods section), of which 177 occurred in a diagonal strip covering 10% of the total area (compared to the expected value of 46). The probability of observing this high a number of MUMs by chance is 4.7 &#x000d7; 10^-59. The alignment of <italic>V. cholerae</italic> chrI in the reverse direction against <italic>E. coli</italic> (which corresponds to the MUMs on the anti-diagonal) has a probability of 1.8 &#x000d7; 10^-90. As a control, we compared the genomes of distantly related species, such as <italic>E. coli</italic> and <italic>Mycobacterium tuberculosis.</italic> These do not show a significant X-alignment (Table <xref ref-type="table" rid="T1">1</xref>).</p><p>We have found that X-alignments of whole genomes are not limited to the <italic>V. cholerae</italic> versus <italic>E. coli</italic> comparison. For example, a whole-genome comparison of two bacteria in the genus <italic>Streptococcus - S. pyogenes</italic> [<xref ref-type="bibr" rid="B7">7</xref>] and <italic>S. pneumoniae</italic> (H. Tettelin, personal communication) - reveals a global X-alignment similar to that of <italic>V. cholerae</italic> versus <italic>E. coli</italic> (Figure <xref ref-type="fig" rid="F1">1d</xref>) which is also statistically significant (Table <xref ref-type="table" rid="T1">1</xref>). In addition, an X-alignment is found between two species in the genus <italic>Mycobacterium - M. tuberculosis</italic> [<xref ref-type="bibr" rid="B8">8</xref>] and <italic>M. leprae</italic> [<xref ref-type="bibr" rid="B9">9</xref>] (Figure <xref ref-type="fig" rid="F1">1e</xref>) - as well as between two strains of <italic>Helicobacter pylori</italic> (data not shown). The X-alignments observed between any two pairs of genomes are not identical in every aspect. For example, in the alignment between the two <italic>Mycobacterium</italic> species, each conserved region is much longer than in the other genome pairs. We believe this is due to different numbers of evolutionary events between the species (see below). Whole-genome X-alignments were not found between any other pairs of species, although a related pattern was seen between some of the chlamydial species (see below).</p></sec><sec><title>Whole-genome X-alignments between species are also found at the proteome level</title><p>To test whether the X-alignments found in the DNA analysis could also be found at the level of whole proteomes, we conducted comparisons of homologous proteins between species (see the Materials and methods section). Figure <xref ref-type="fig" rid="F2">2a</xref> shows a scatterplot of chromosome positions of all proteins homologous between <italic>V. cholerae</italic> chrI and <italic>E. coli.</italic> The presence of many large gene families causes a great deal of noise in this comparison. This noise can be reduced by considering only the best matching homolog for each open reading frame (ORF), rather than all protein homologs (Figure <xref ref-type="fig" rid="F2">2b</xref>). This filtered protein comparison results in an X-alignment that is statistically significant (Table <xref ref-type="table" rid="T2">2</xref>).</p></sec><sec><title>Whole-genome X-alignments within species</title><p>The finding of the X-alignment pattern between species led us to search for similar patterns within species; that is, global alignments of a genome with its own reverse complement. Of the genomes for which we found between-species X-alignments (<italic>M. tuberculosis, M. leprae, S. pyogenes, S. pneumoniae, E. coli</italic> and <italic>V. cholerae),</italic> statistically significant self-alignments are detected for all except <italic>M. tuberculosis</italic> (Figure <xref ref-type="fig" rid="F3">3</xref>; probabilities shown in Table <xref ref-type="table" rid="T1">1</xref>). Interestingly, these self-alignments are not as strong as those between species. Proteome analysis also shows an X-alignment within species (shown for <italic>V. cholerae</italic> chrI in Figure <xref ref-type="fig" rid="F2">2d</xref>; probabilities shown in Table <xref ref-type="table" rid="T2">2</xref>). The X-alignment of proteins within <italic>V. cholerae</italic> chrI is statistically significant only for recently duplicated-genes, but disappears when all paralogs are included. The importance of filtering for recent duplications is discussed below.</p></sec><sec><title>Model I: whole-genome inverted duplications</title><p>One possible explanation for an X-alignment within and between species is an ancestral inverted duplication of the whole genome, as has been suggested for <italic>E. coli</italic> [<xref ref-type="bibr" rid="B10">10</xref>]. The weak or missing X-alignment within species could be explained by gene loss of one of the two duplicates of many of the pairs of genes in the different lineages. Gene loss has been found to follow large chromosomal or genome duplications [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. This gene loss is thought to stabilize large duplications by preventing recombination events between duplicate genes. If gene loss is responsible for the weak X-alignment within species, then to maintain the X-alignments between species, the member of the gene pair lost in a particular lineage should be essentially random. If an ancient inverted duplication followed by differential gene loss is the correct explanation for the observed X-alignments, one would expect the genes along one diagonal to be orthologous between species (related to each other by the speciation event), while the genes along the other diagonal should be paralogous (related to each other by the genome duplication event before the speciation of the two lineages). However, the evidence appears to contradict this model: likely orthologous gene pairs are equally distributed on each diagonal (data not shown).</p></sec><sec><title>Model II: chromosomal inversions about the origin and/or terminus</title><p>A second possible explanation for the X-alignments is that an underlying mechanism allows sections of DNA to move within the genome but maintains the distance of these sections from the origin and/or terminus. There are a variety of possible mechanisms for such movement, but we believe the most likely explanation is the occurrence of large chromosomal inversions that pivot around the replication origin and/or terminus. Large chromosomal inversions, including those that occur around the replication origin and terminus, have been shown to occur in <italic>E. coli</italic> and <italic>Salmonella typhimurium</italic> in the laboratory (see, for example, [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]). The occurrence of such inversions over evolutionary time scales was first suggested by comparative analysis of the complete genomes of four strains in the genus <italic>Chlamydia</italic> [<xref ref-type="bibr" rid="B19">19</xref>]. In that study, we found that the major chromosomal differences between <italic>C. pneumoniae</italic> and <italic>C. trachomatis</italic> (shown in Figure <xref ref-type="fig" rid="F2">2c</xref>) were consistent with the occurrence of large inversions that pivoted around the origin and terminus (including multiple inversions of different sizes). In Figure <xref ref-type="fig" rid="F4">4</xref> we present a hypothetical model showing how a small number of inversions centered around the origin or terminus could produce patterns very similar to those seen in the <italic>Chlamydia, Mycobacterium</italic> and <italic>Helicobacter</italic> comparisons. The continued occurrence of such inversion over longer time scales would result in an X-alignment similar to that seen in the <italic>V. cholerae</italic> versus <italic>E. coli</italic> and <italic>S. pneumoniae</italic> versus <italic>S. pyogenes</italic> comparisons. Thus the different between-species X-alignments could be the result of different numbers of inversions between particular pairs of species.</p><p>Inversions about the origin and terminus could also produce an X-alignment within species, through the splitting of tandemly duplicated sequence. Many sets of tandemly duplicated genes are found in most bacterial genomes [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>] (also see Figure <xref ref-type="fig" rid="F3">3a</xref>,<xref ref-type="fig" rid="F3">c</xref>). As tandem duplications are inherently unstable (one of the duplicates can be rapidly eliminated by slippage and/or recombination events [<xref ref-type="bibr" rid="B21">21</xref>]), the fact that many tandem pairs are present within each genome suggests that tandem duplications occur frequently. Thus, it is reasonable to assume that occasionally a large inversion will split a pair of tandemly duplicated genes. An inversion that pivots about the origin and also splits a tandem duplication will result in a pair of paralogous genes spaced symmetrically on opposite sides of the origin.</p><p>If our inversion model is correct, then the genes along both diagonals in the <italic>between</italic>-species alignments should be orthologous, which is the case (see above). In contrast, genes along the anti-diagonal in the <italic>within</italic>-species X-alignments should be recent tandem duplicates that have been separated by inversions. This also appears to be the case - in the within-species analysis of <italic>V. cholerae</italic> chrI ORFs, the X-alignment shows up best when only recent duplicates are analyzed (Figure <xref ref-type="fig" rid="F2">2d</xref>). The splitting of tandem duplicates by inversions may be a general mechanism to stabilize the coexistence of duplicated genes, as it will prevent their elimination by unequal crossing-over or replication slippage events.</p><p>What could cause inversions that pivot around the origin and terminus of the genome to occur more frequently than other inversions? One possibility is that many inversions occur, but there is selection against those that change the distance of a gene from the origin or terminus. Such a possibility has been suggested by experimental work in <italic>E. coli</italic> [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. Additional studies have, however, suggested that there is little selective difference between inversions and that instead there may be certain regions that are more prone to inversion than others [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>]. Alternatively, the inversion events could be linked to replication, as has been suggested for small local inversion events [<xref ref-type="bibr" rid="B24">24</xref>]. Whatever the mechanisms, the fact that we find evidence for such inversions between many pairs of species suggests that they are a common feature of bacterial evolution. Many aspects of the X-alignments require further exploration. For example, to split a tandem duplication, an inversion must fall precisely on the boundary between two duplicated genes. This would appear to be unlikely, requiring a large number of inversions in order to generate a sufficient number of split gene pairs. If the mechanisms of gene duplication are somehow related to the mechanisms of inversion, however, then this model is more plausible. The process of duplicating a gene, if it occurs during replication, might promote a recombination event within the bacterial chromosome that inverts the sequence from the origin up to that point. As with inversion events, recombination and replication have been found to be tightly coupled [<xref ref-type="bibr" rid="B25">25</xref>].</p></sec></sec><sec><title>Conclusions</title><p>We present here a novel observation regarding the conservation between bacterial species of the distance of particular genes from the replication origin or terminus. The initial observation was only possible due to the availability of complete genome sequences from pairs of moderately closely related species (for example, <italic>V. cholerae</italic> and <italic>E. coli</italic>). This shows the importance of having genome pairs from many levels of evolutionary relatedness. Comparisons of distantly related species enable the determination of universal features of life as well as of events that occur very rarely. Comparison of very closely related species allows the identification of frequent events such as transitional changes at third codon positions or tandem duplications. To elucidate all other events in the history of life, genome pairs covering all the intermediate levels of evolutionary relatedness will be needed.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Genomes analyzed</title><p>Complete published genome sequences were obtained from the National Center for Biotechnology Information website [<xref ref-type="bibr" rid="B26">26</xref>] or from the TIGR Comprehensive Microbial Resource [<xref ref-type="bibr" rid="B27">27</xref>]. These included <italic>Aeropyrum pernix</italic> [<xref ref-type="bibr" rid="B28">28</xref>], <italic>Aquifex aeolicus</italic> [<xref ref-type="bibr" rid="B29">29</xref>], <italic>Archaeoglobus fulgidus</italic> [<xref ref-type="bibr" rid="B30">30</xref>], <italic>Bacillus subtilis</italic> [<xref ref-type="bibr" rid="B31">31</xref>], <italic>Borrelia burgdorferi</italic> [<xref ref-type="bibr" rid="B32">32</xref>], <italic>Campylobacter jejuni</italic> [<xref ref-type="bibr" rid="B33">33</xref>], <italic>Chlamydia pneumoniae</italic> AR39 [<xref ref-type="bibr" rid="B19">19</xref>], <italic>Chlamydia pneumoniae</italic> CWL029 [<xref ref-type="bibr" rid="B34">34</xref>], <italic>Chlamydia trachomatis</italic> (D/UW-3/Cx) [<xref ref-type="bibr" rid="B35">35</xref>], <italic>Chlamydia trachomatis</italic> MoPn [<xref ref-type="bibr" rid="B19">19</xref>], <italic>Deinococcus radiodurans</italic> [<xref ref-type="bibr" rid="B36">36</xref>], <italic>Escherichia coli</italic> [<xref ref-type="bibr" rid="B5">5</xref>], <italic>Haemophilus influenzae</italic> [<xref ref-type="bibr" rid="B37">37</xref>], <italic>Helicobacter pylori</italic> [<xref ref-type="bibr" rid="B38">38</xref>], <italic>Helicobacter pylori</italic> J99 [<xref ref-type="bibr" rid="B39">39</xref>], <italic>Methanobacterium thermoautotrophicum</italic> [<xref ref-type="bibr" rid="B40">40</xref>], <italic>Methanococcus jannaschii</italic> [<xref ref-type="bibr" rid="B41">41</xref>], <italic>Mycobacterium tuberculosis</italic> [<xref ref-type="bibr" rid="B8">8</xref>], <italic>Mycoplasma genitalium</italic> [<xref ref-type="bibr" rid="B42">42</xref>], <italic>Mycoplasma pneumoniae</italic> [<xref ref-type="bibr" rid="B43">43</xref>], <italic>Neisseria meningitidis</italic> MC58 [<xref ref-type="bibr" rid="B20">20</xref>], <italic>Neisseria meningitidis</italic> serogroup A strain Z2491 [<xref ref-type="bibr" rid="B44">44</xref>], <italic>Pyrococcus horikoshii</italic> [<xref ref-type="bibr" rid="B45">45</xref>], <italic>Rickettsia prowazekii</italic> [<xref ref-type="bibr" rid="B46">46</xref>], <italic>Synechocystis</italic> sp. [<xref ref-type="bibr" rid="B47">47</xref>], <italic>Thermotoga maritima</italic> [<xref ref-type="bibr" rid="B48">48</xref>], <italic>Treponema pallidum</italic> [<xref ref-type="bibr" rid="B49">49</xref>], and <italic>Vibrio cholerae</italic> [<xref ref-type="bibr" rid="B4">4</xref>]. In addition, a few unpublished genomes were analyzed: <italic>Streptococcus pyogenes</italic> (obtained from the Oklahoma University Genome Center website [<xref ref-type="bibr" rid="B7">7</xref>]), <italic>Streptococcus pneumoniae</italic> (H. Tettelin, personal communication), and <italic>Mycobacterium leprae</italic> (obtained from the Sanger Centre Pathogen Sequencing Group website [<xref ref-type="bibr" rid="B9">9</xref>]).</p></sec><sec><title>Whole-genome DNA alignments</title><p>DNA alignments of the complete genomic sequences of all bacteria used in this study were accomplished with the MUMmer program [<xref ref-type="bibr" rid="B6">6</xref>]. This program uses an efficient suffix tree construction algorithm to rapidly compute alignments of entire genomes. The algorithm identifies all exact matches of nucleotide subsequences that are contained in both input sequences; these exact matches must be longer than a specified minimum length, which was set to 20 base pairs for this comparison. To search for genome-scale alignments within species, complete bacterial and archaeal genomes (25 in total including all published genomes) were aligned with their own reverse complements. To search for between-species alignments, all genomes were aligned against all others in both orientations.</p></sec><sec><title>Whole-genome protein comparisons</title><p>The predicted proteome of each complete genome sequence (all predicted proteins in the genome) was compared to the proteomes of all complete genome sequences (including itself) using the fasta3 program [<xref ref-type="bibr" rid="B50">50</xref>]. Matches with an expected score (e-value) of 10^-5 or less were considered significant.</p></sec><sec><title>Statistical significance of X-alignments</title><p>To calculate the statistical significance of the X-alignments, the maximal unique matching subsequences (MUMs) for unrelated genomes were examined and found to be uniformly distributed [<xref ref-type="bibr" rid="B6">6</xref>]. With a uniform background, the expected density of MUMs in any region of an alignment plot is a simple proportion of the area of that region to the entire plot. In particular, in an alignment with <italic>N</italic> total MUMs, the probability (Pr) of observing at least <italic>m</italic> matches in a region with area <italic>p</italic> can be computed using the binomial distribution in Equation 1:</p><graphic xlink:href="gb-2000-1-6-research0011-i1.gif"/><p>The alignment of <italic>V. cholerae</italic> chrI (both forward and reverse strands) versus <italic>E. coli</italic> contains 926 MUMs. The MUMs forming X-alignments appear along the diagonal (<italic>y</italic> = <italic>x</italic>) and the anti-diagonal (<italic>y</italic> = <italic>L -x,</italic> where <italic>L</italic> is the genome length). To estimate the significance of the alignments in both directions, diagonal strips were sampled along each of the diagonals. The width of each strip was set at 10% of the plot area and significance values were calculated (Table <xref ref-type="table" rid="T1">1</xref>).</p></sec><sec><title>Identification of origins of replication</title><p>The origins of replication for the bacterial genomes have been characterized by a variety of methods. For <italic>E. coli, M. tuberculosis</italic> and <italic>M. leprae,</italic> the origins have been well-characterized by laboratory studies [<xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref>]. The origins and termini of <italic>C. trachomatis, C. pneumoniae</italic> and <italic>V. cholerae</italic> were identified by GC-skew [<xref ref-type="bibr" rid="B53">53</xref>] and by characteristic genes in the region of the origin [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. GC-skew uses the function (G-C)/(G+C) computed on 2,000 bp windows across the genome, which exhibits a clear tendency in many bacterial genomes to be positive for the leading strand and negative for the lagging strand. The origin of <italic>H. pylori</italic> was determined by oligomer skew [<xref ref-type="bibr" rid="B54">54</xref>] and confirmed by GC-skew. The origins and termini of <italic>S. pneumoniae</italic> and <italic>S. pyogenes</italic> were determined by the authors of the present study using GC-skew analysis and the locations of characteristic genes, particularly the chromosome replication initiator gene <italic>dnaA.</italic></p></sec></sec>

A search for reverse transcriptase-coding sequences reveals new non-LTR retrotransposons in the genome of <italic>Drosophila melanogaster</italic>

<sec><title>Background:</title><p>Non-long terminal repeat (non-LTR) retrotransposons are eukaryotic mobile genetic elements that transpose by reverse transcription of an RNA intermediate. We have performed a systematic search for sequences matching the characteristic reverse transcriptase domain of non-LTR retrotransposons in the sequenced regions of the <italic>Drosophila melanogaster</italic> genome.</p></sec><sec><title>Results:</title><p>In addition to previously characterized BS, Doc, F, G, I and Jockey elements, we have identified new non-LTR retrotransposons: Waldo, You and JuanDm. Waldo elements are related to mosquito RTI elements. You to the <italic>Drosophila</italic> I factor, and JuanDm to mosquito Juan-A and Juan-C. Interestingly, all JuanDm elements are highly homogeneous in sequence, suggesting that they are recent components of the <italic>Drosophila</italic> genome.</p></sec><sec><title>Conclusions:</title><p>The genome of <italic>D. melanogaster</italic> contains at least ten families of non-site-specific non-LTR retrotransposons representing three distinct clades. Many of these families contain potentially active members. Fine evolutionary analyses must await the more accurate sequences that are expected in the next future.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Berezikov</surname><given-names>Eugene</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Bucheton</surname><given-names>Alain</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Busseau</surname><given-names>Isabelle</given-names></name><xref ref-type="aff" rid="I2">2</xref><email>busseau@igh.cnrs.fr</email></contrib>

Genome Biology

<sec><title>Background</title><p>Non-long terminal repeat (non-LTR) retrotransposons, also known as LINEs (long interspersed nuclear elements), make up a very large class of transposable elements that are present in most eukaryotes [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. They transpose by reverse transcription of an RNA intermediate. They possess an open reading frame (ORF) with coding capacities for a protein with endonuclease, reverse transcriptase and sometimes RNase H activities. Many of them contain an additional ORF (ORF1) encoding a protein of an unknown function. Some non-LTR retrotransposons, such as R1 and R2 in arthropods, insert at specific sites in the genome [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>], whereas the others do not show specific sites of insertion. The cleavage specificity of some elements such as R2 is the result of the activity of an endonuclease showing similarities to restriction enzymes [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. The other elements, whether they are site-specific (like R1) or not (like mammalian L1 elements), have an endonuclease structurally related to apurinic/apyrimidinic (AP) endonucleases [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. Non-LTR retrotransposons can be assigned to 12 clades on the basis of the sequences of their reverse transcriptase domains and appear to be as old as eukaryotes [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. The chronology of the acquisition of the various enzymatic domains suggests that the first non-LTR retrotransposons contained an endonuclease related to restriction enzymes, and that the non-LTR retrotransposons that have a nuclease structurally related to AP endonucleases derived from this ancestral group. The RNase H domain would have been acquired later [<xref ref-type="bibr" rid="B2">2</xref>].</p><p>Reverse transcription of non-LTR retrotransposons is thought to occur on chromosomal DNA using the cut resulting from their endonuclease activity as a primer. This process is called target-primed reverse transcription (TPRT). The retrotransposition machinery of non-LTR retrotransposons is thought to be used for transposition of SINEs (short interspersed nuclear elements) and might also be at the origin of the formation of processed pseudogenes [<xref ref-type="bibr" rid="B10">10</xref>]. The shape of eukaryotic genomes is therefore largely influenced by non-LTR retrotransposons. They are particularly abundant in mammals. Humans contain one major class of non-LTR retrotransposons, the L1 elements, which make up more than 15% of the genome [<xref ref-type="bibr" rid="B11">11</xref>]. In contrast, the <italic>Drosophila melanogaster</italic> genome has more than ten non-LTR retrotransposon families, representing the Jockey, I, R1 and R2 clades [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B9">9</xref>], indicating that several elements have had the chance to spread in this species. Early studies of several families of non-site-specific <italic>Drosophila</italic> non-LTR retrotransposons have shown that, in many cases, they comprise recently transposed euchromatic elements dispersed on all chromosomal arms, and variously defective elements accumulated in pericentromeric heterochromatin [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. Euchromatic elements can be full size or truncated at the 5' end, presumably as the result of early arrest of reverse transcription. Heterochromatic elements are retrotranspositionally inactive and represent old components of the genome.</p><p>Many non-LTR retrotransposons in <italic>D. melanogaster</italic> were discovered during analysis of spontaneous mutations, as a result of their insertion within a gene. The sequence of most of the euchromatic part of the <italic>D. melanogaster</italic> genome has recently been reported [<xref ref-type="bibr" rid="B14">14</xref>], giving an interesting opportunity for studying all the non-LTR retrotransposons present in this species.</p></sec><sec><title>Results</title><p>Our aim was to identify all families of non-LTR retrotransposons in the sequenced part of the <italic>D. melanogaster</italic> genome. For this, we used software based on profile hidden Markov models (HMMs) to find all sequences matching the full-length reverse transcriptase model and containing the conserved motif [FY]XDD (in single-letter amino acid code) [<xref ref-type="bibr" rid="B15">15</xref>]. This approach makes it possible to simultaneously identify all potential reverse transcriptase sequences, including LTR elements, non-LTR elements and retroviruses. To distinguish between these classes of retrotransposons, sequences resulting from HMM search were used in BLAST searches against all known retrotransposons, and were assigned by the best match to the three groups. In our analyses, only the non-LTR fraction of the HMM search results was extracted for detailed investigation. Release 1 of the <italic>D. melanogaster</italic> genome sequence [<xref ref-type="bibr" rid="B14">14</xref>] and the Berkeley/European Drosophila Genome Projects (BDGP/EDGP) sequences [<xref ref-type="bibr" rid="B16">16</xref>] were analyzed. In both datasets, reverse transcriptase sequences of many known <italic>D. melanogaster</italic> non-LTR retrotransposons were identified. In addition, some reverse transcriptase sequences with highest similarity to retrotransposons from non-<italic>Drosophila</italic> species were recognized, providing a basis for identification of new families (see below). The results of these analyses are represented in Figure <xref ref-type="fig" rid="F1">1</xref>.</p><p>Each family of non-LTR retrotransposons identified in this way was further analyzed at the nucleotide level: BLASTN searches were performed, both in the sequences from Release 1 [<xref ref-type="bibr" rid="B14">14</xref>] and from BDGP/EDGP [<xref ref-type="bibr" rid="B16">16</xref>], to identify all members of the family in the genome, and full-size copies were identified. The BDGP/EDGP sequences appear to be a subset of Release 1 sequences. There is, however, a high error rate in the sequences of Release 1 containing repetitive DNA: comparison of a fraction of both datasets revealed 0.42% point differences in repetitive sequences compared with only 0.0046% point differences in non-repetitive sequences (see [<xref ref-type="bibr" rid="B17">17</xref>] and the Celera website [<xref ref-type="bibr" rid="B18">18</xref>]). In our analysis, we estimate that non-LTR retrotransposon sequences that could be found in both BDGP/EDGP and Release 1 datasets contain an average of one difference in every 300 base pairs (bp), that is 0.33%. About 40% of these differences are due to insertions and deletions rather than base mismatches, and they are associated with resolution of repeated residues: for example, 5 G should be 6 G, 3 T should be 2 T, 3 C should be 2 C, 8 A should be 7 A, and so on. Consequently, many full-size elements from Release 1 appear not to have coding capacities for complete ORF1 and ORF2 products because of frameshifts. We therefore chose as a representative of each new family identified a full-length element extracted from the BDGP/EDGP database.</p><p>The results of our searches are summarized in Table <xref ref-type="table" rid="T1">1</xref>. Most of the non-LTR retrotransposon families that we found were already identified and described. These include BS, Doc, F, G, I and Jockey. We have identified three new families: Waldo, JuanDm, and You. They belong to the R1, Jockey and I clades [<xref ref-type="bibr" rid="B2">2</xref>], respectively (Figure <xref ref-type="fig" rid="F2">2</xref>). Not surprisingly, no full-size copy of elements from the R1Dm, R2Dm, TART and HeT-A elements were identified, neither in Release 1 nor in the BDGP/EDGP sequences. Presumably this is because rDNA (target of R1Dm and R2Dm) and telomeric DNA (target of HET-A and TART) have been excluded from the sequencing project. Finally, sequences related to Q, LOA, Bilbo and Helena were found, but no full-size copy was identified for these families.</p><sec><title>Elements of the Jockey clade</title><p>The Jockey clade is by far the most represented clade in <italic>D. melanogaster.</italic> It includes one site-specific element, TART, and a large variety of non-site-specific elements such as Jockey, F, Doc, BS, G and Helena, although for these last two no potentially active copy has been identified. Our analysis has revealed one additional member of the Jockey clade, the JuanDm element.</p><sec><title>Jockey, F and Doc elements</title><p>Jockey, F and Doc are very similar in sequence and probably descend from a common ancestor [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. Our analyses confirm previous studies. These families include both full-size and variously defective copies. The copy number of Jockey, F and Doc was estimated to be around 50-100 by early studies [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>]. About half of these copies are located near the chromocenter. This estimate has been refined to 31.60 &#x000b1; 7.51 sites on chromosomal arms for Jockey, 31.40 &#x000b1; 10 for F, and 26.20 &#x000b1; 4.74 for Doc [<xref ref-type="bibr" rid="B24">24</xref>]. Our analyses reveal that each of these families in fact contains a very large total number of copies: at least 86 for Jockey, 94 for F and 128 for Doc. These are underestimates, as a large fraction of the heterochromatic part of the genome is not represented in the databases, and these regions are well known to contain transposable elements. We have identified seven full-size copies of Jockey, all surrounded by target site duplications (TSD) and mapping on chromosomal arms. The F and Doc families contain as many as 14 and 22 full-size copies, respectively. Surprisingly many full-size copies of F do not appear to be flanked by TSD. Large subsets of the defective copies of these families are truncated at the 5' end.</p><p>The Jockey family was reported to contain a subset of 3 kb long internally deleted elements [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Our analyses provide evidence for only two internally deleted elements of 2.6 and 2.9 kb in the genome of the <italic>y; cn bw sp</italic> strain. A subfamily of internally deleted elements might be specific for some strains, or might be confined to heterochromatic regions that are not represented in the databases.</p></sec><sec><title>G elements</title><p>No potentially active G element was reported by earlier studies. Only one complete G element was previously described, and it did not code for full-length ORF1 and ORF2 products [<xref ref-type="bibr" rid="B26">26</xref>]. However, some characteristic domains were recognized, such as cysteine-rich motifs of ORF1 and the reverse transcriptase domain of ORF2. The chromosomal distribution of G elements appeared to be fairly stable between strains. They were found mostly in tandem arrays in the nontranscribed spacer sequence of rDNA units [<xref ref-type="bibr" rid="B26">26</xref>]. Our analyses reveal 37 copies of G elements, most of which were short and variously deleted. We identified only one full-size G element surrounded by TSD, but it could not encode the complete ORF1 and ORF2 proteins. However, as this full-size element is not present in the sequences released by BDGP/EDGP, we cannot decide whether the fact that it appears to be unable to code for these products results from sequencing errors or if it is an inactive element. It is located in region 60E12-60F2 of the right arm of the second chromosome and is surrounded by other defective G elements organized in tandem repeats. In fact, the sequences of this G element and surrounding DNA are very similar to those previously described [<xref ref-type="bibr" rid="B26">26</xref>], indicating that it is the same inactive element.</p></sec><sec><title>TART elements</title><p>TART elements insert preferentially at the tips of the chromosomes with HeT-A elements and constitute the telomeres of the <italic>Drosophila</italic> chromosomes [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>]. As the telomeric regions are not represented in the released sequences we did not expect to pull out TART or HeT-A elements. Not surprisingly, we have identified only 11 sequences matching TART reverse transcriptase, and none of them encodes a large protein. We found only some short regions with less than 60% identity to HeT-A. This emphasizes the idea that TART and HeT-A elements have a strong preference for telomeric regions.</p></sec></sec><sec><title>BS elements</title><p>The number of copies of BS elements was estimated to be around five to ten [<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B29">29</xref>]. Although the first two copies were identified as recent insertions within a gypsy element, they seem to transpose very infrequently because their distribution pattern is highly conserved between strains as judged by Southern blot analyses [<xref ref-type="bibr" rid="B25">25</xref>]. Our analyses reveal 19 copies of BS in the genome, only two of which are full size and surrounded by TSD.</p><sec><title>Helena elements</title><p>Helena elements are non-LTR retrotransposons related to BS elements. They exist in various <italic>Drosophila</italic> species [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>]. Hybridization studies have revealed sequences homologous to the <italic>D. virilis</italic> Helena element close to the chromocenter of <italic>D. melanogaster,</italic> and one copy was partly sequenced (GenBank accession code AF012030). We have found nine sequences matching Helena reverse transcriptase. No full-size copy is recognizable. They are unable to encode a large protein and the mean genetic distance between the different copies is very high. Presumably, these are the remnants of very ancient elements related to Helena, present in <italic>D. melanogaster</italic> or an ancestor species, and now extinct in <italic>D. melanogaster.</italic></p><p>Some of the sequences found in our HMM search are equally distant from Helena and BS (52% identity). They possibly represent a new family, denoted Helena/BS in Figure <xref ref-type="fig" rid="F1">1</xref>. We noticed after this work was completed that a new non-LTR retrotransposon has recently been identified. The X element (GenBank accession code AF237761) belongs to the Helena/BS family.</p></sec><sec><title>JuanDm elements</title><p>We have identified one new family of non-LTR retrotransposons belonging to the Jockey clade, which we have called JuanDm because of its close relationship with Juan-A in <italic>Aedes aegypti</italic> [<xref ref-type="bibr" rid="B33">33</xref>] and Juan-C in <italic>Culex pipiens</italic> [<xref ref-type="bibr" rid="B34">34</xref>]. There are only eight JuanDm elements in the sequenced genome of the strain <italic>y</italic>; <italic>cn bw sp.</italic> One is only partially sequenced. The other seven are all surrounded by TSD (Figure <xref ref-type="fig" rid="F3">3</xref>). Four are full size, two are 5' truncated, and one has a 1.2 kb internal deletion. One of the full-size and one of the 5'-truncated JuanDm elements are also found in the sequences of BDGP/EDGP. There are five nucleotide differences between the sequences of the full-size copy of JuanDm from Release 1 and from BDGP/EDGP, including one base insertion and one base deletion, presumably resulting from sequencing errors. The complete JuanDm from BDGP/EDGP contains two overlapping ORFs with the capacity for encoding proteins very similar to those potentially encoded by Juan-A and Juan-C ORF1 and ORF2 (Figure <xref ref-type="fig" rid="F4">4</xref>).</p><p>There is very low heterogeneity between the eight JuanDm elements, which are more than 98% identical to each other at the DNA level. The JuanDm family is therefore a very young family in the genome of <italic>D. melanogaster.</italic> It is also present in other species from the <italic>D. melanogaster</italic> subgroup [<xref ref-type="bibr" rid="B35">35</xref>]: Figure <xref ref-type="fig" rid="F5">5a</xref> shows the result of hybridization of a JuanDm-specific DNA probe (Figure <xref ref-type="fig" rid="F3">3</xref>) to a Southern blot of genomic DNA digested with <italic>Spe</italic>I and <italic>Cla</italic>I. The probe reveals a JuanDm internal <italic>Spe</italic>I-<italic>Cla</italic>I fragment of 3 kb. A 3 kb fragment is observed in all species studied from the <italic>D. melanogaster</italic> subgroup: <italic>D. melanogaster, D. simulans, D. mauritiana, D. teissieri</italic> and <italic>D. yakuba.</italic> The strong signal intensity indicates that this fragment is present in multiple copies in <italic>D. melanogaster, D. simulans, D. mauritiana</italic> and <italic>D. yakuba.</italic> In addition, a few (two or three) fragments of various sizes are detected in these species. They might correspond to deleted elements. This indicates that the JuanDm family in these species is composed of very homogeneous elements. By contrast, in <italic>D. teissieri</italic> a 3 kb fragment is detected with a much lower intensity than in the other species. At least seven fragments of different sizes are also observed in this species. This suggests that the 3 kb internal fragment of JuanDm is not conserved in <italic>D. teissieri,</italic> and/or that these elements are more heterogeneous in size in <italic>D. teissieri</italic> than in other species of the <italic>D. melanogaster</italic> subgroup. No signal is detected in the more distant species <italic>D. virilis</italic> [<xref ref-type="bibr" rid="B35">35</xref>].</p></sec></sec><sec><title>Elements of the I clade</title><p>Until now the I element was the only member of the I clade to have been identified in <italic>D. melanogaster.</italic> This element has been extensively studied because it is responsible for the IR system of hybrid dysgenesis [<xref ref-type="bibr" rid="B36">36</xref>]. We have identified You, a new element of the I clade.</p><sec><title>I elements</title><p>Studies of the I element family have been recently reviewed [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>]. It is known that active I elements, also called I factors, are present only in some strains that are called inducer strains. Inducer strains are expected to contain five to ten full-size I elements in euchromatic regions, and about 30 defective elements mostly localized within pericentromeric heterochromatin [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. The isogenic <italic>y; cn bw sp</italic> strain is inducer. Six full-size I elements can be identified in the sequences of Release 1. There is probably a seventh one for which sequences are available for both the 5' and 3' ends but not the middle. None of these elements has the coding capacities of an active I factor [<xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B42">42</xref>]. We assume that this is due to sequencing errors. All these copies are surrounded by TSD. They map on euchromatic sites. We have also identified six 5' truncated elements that are surrounded by TSD, and eight that are not. In addition, about 30 elements more divergent from active I elements are found. None of those is full size. Some are 5' truncated, some 3' truncated, some are truncated at both ends, and many have internal deletions. They are usually within unallocated scaffolds, strongly suggesting that they are in heterochromatic regions. Those that possess the 3' end of I terminate with a sequence related to TAA (TAAA)_n instead of the regular TAA repeats of the active I elements. This termination was shown to be characteristic of I elements localized in pericentromeric heterochromatin [<xref ref-type="bibr" rid="B39">39</xref>].</p></sec><sec><title>You elements</title><p>The You family was first identified in the sequences of BDGP/EDGP on the basis of similarity between the DNA sequence of a full-size copy and that of the I element (58% all along the elements). Three full-size copies and one 5'-truncated copy surrounded by TSD can be found in the sequences of Release 1 (Figure <xref ref-type="fig" rid="F6">6</xref>). None of the complete You elements can produce full products of ORF1 and ORF2. This is presumably because of sequencing errors as one of these copies was also sequenced by BDGP/EDGP, revealing full coding capacities. You and I are closely related: they both encode very similar ORF1 products containing three cysteine-rich motifs (Figure <xref ref-type="fig" rid="F7">7</xref>). The products of ORF2 are also very similar and contain endonuclease, reverse transcriptase and RNase H domains as well as one cysteine-rich motif. You elements terminate at the 3' end with A/T rich sequences.</p><p>Strikingly, we have noticed that the six You elements that could be localized, map in distal or proximal regions of chromosomal arms: 1F and 19E-F on chromosome X, 39E and 40A on chromosome II, and 60F and 79E-F on chromosome III. This is a very unusual distribution, and it would be interesting to determine whether You elements are also localized similarly in other <italic>D. melanogaster</italic> strains. Examining the sequences surrounding You elements did not reveal an obvious bias for insertion sites. Their unusual distribution is therefore unlikely to be the result of sequence preference for insertion. Alternatively, euchromatic copies of You could be the few survivors of a former broader You family that is in the process of being eliminated from the genome by stochastic loss, as in the case of <italic>Drosophila</italic> mariner elements [<xref ref-type="bibr" rid="B43">43</xref>]. Their confinement near the frontiers between euchromatin and heterochromatin could reflect a lower rate of loss of sequences in these regions.</p><p>The You family is also present in other species of the <italic>D. melanogaster</italic> subgroup [<xref ref-type="bibr" rid="B35">35</xref>]: Figure <xref ref-type="fig" rid="F5">5b</xref> shows the result of hybridization of a You-specific DNA probe (Figure <xref ref-type="fig" rid="F6">6</xref>) to a Southern blot of genomic DNA digested with <italic>Sma</italic>I and <italic>Sac</italic>I. The probe reveals an internal <italic>Sma</italic>I-<italic>Sac</italic>I fragment of 4.9 kb. A 4.9 kb band hybridizing strongly to the probe is observed in <italic>D. melanogaster</italic> and in sibling species <italic>D. simulans</italic> and <italic>D. mauritiana,</italic> indicating that these species contain several copies of potentially full-size You elements. The more distant species <italic>D. teissieri</italic> and <italic>D. yakuba</italic> show much weaker signals, and the presence of the 4.9 kb fragment is not certain. These species contain sequences related to You, but these sequences are divergent from those of the <italic>D. melanogaster</italic> You elements. Finally, no hybridization signal is detected in <italic>D. virilis,</italic> which is outside the <italic>D. melanogaster</italic> subgroup [<xref ref-type="bibr" rid="B35">35</xref>], even with longer exposures.</p></sec></sec><sec><title>Elements of the R1 clade</title><p>Most non-LTR retrotransposons of the R1 clade are site-specific: R1Dm inserts preferentially at one site into the rDNA units. Recently, we have described two new subfamilies of elements - Waldo-A and Waldo-B - that belong to the R1 clade and do not seem to have strong insertion site preference.</p><sec><title>Waldo-A and Waldo-B elements</title><p>The Waldo-A and Waldo-B families were first identified by analyzing the sequences released by BDGP/EDGP and are described elsewhere [<xref ref-type="bibr" rid="B44">44</xref>]. Members of both subfamilies are found in the present study. There are two full-size Waldo-A and five full-size Waldo-B elements. Most of them are surrounded by TSD. All 5'-truncated Waldo-A elements are surrounded by TSD and are very similar to complete Waldo-A elements (&#x0003e;98% identity at the DNA level). Only a minority of Waldo-B 5'-truncated elements is surrounded by TSD. In addition variously defective elements with less than 98% identity to Waldo-A or Waldo-B elements have been found.</p></sec><sec><title>R1Dm elements</title><p>Only fragments of elements with similarity to R1Dm have been found in our search, presumably because R1Dm inserts preferentially within rDNA repeats, which are not represented in the databases.</p></sec></sec><sec><title>Elements of other clades</title><p>Partial R2Dm elements were identified in the BDGP/EDGP sequences but not in Release 1, presumably for the same reasons as for R1Dm.</p><p>We have found sequences that match the reverse transcriptase of the non-LTR retrotransposons Q in <italic>Anopheles gambiae</italic> [<xref ref-type="bibr" rid="B45">45</xref>] and LOA in <italic>D. silvestris</italic> [<xref ref-type="bibr" rid="B46">46</xref>]. These non-LTR retrotransposons belong to the LOA and the CR1 clades, respectively, which were supposed to have no representatives in <italic>D. melanogaster.</italic> We have also found some short sequences showing weak similarities to the element Bilbo in <italic>D. subobscura</italic> [<xref ref-type="bibr" rid="B47">47</xref>], which also belongs to the LOA clade. As in the case of Helena, these sequences are presumably remnants of very ancient elements from the LOA and CR1 clades, once present in <italic>D. melanogaster</italic> or an ancestor and now extinct.</p></sec></sec><sec><title>Discussion</title><p>Put together, the sequences of BS, Doc, F, G, I, Jockey, JuanDm, Waldo-A, Waldo-B and You elements presented in this work make 1134 kb, which is almost 1% of the 120 Mb of Release 1 [<xref ref-type="bibr" rid="B14">14</xref>]. This percentage would certainly increase if our analysis could be extended to all heterochromatic portions of the genome. In particular pericentromeric heterochromatin regions are known to be enriched in defective transposable elements that are thought to be old components of the genome [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. Only a small subset of the sequences in Release 1 (3.8 Mb) presumably originate from these regions because they could not be localized on chromosomal arms [<xref ref-type="bibr" rid="B14">14</xref>]. Many of the defective copies of non-LTR retrotransposons were found in these unlocalized sequences.</p><p>The mean genetic distance between members of a given family reflects their degree of heterogeneity. Assuming that non-LTR retrotransposons of different families accumulate mutations at the same rate, elements belonging to old families are expected to be more heterogeneous than elements of recent families. This reasoning does not take into account, however, possible bursts of invasion of a genome by a particular element or subset of elements of one family. This situation is well documented in the case of the I element family [<xref ref-type="bibr" rid="B37">37</xref>]. Indeed, active I elements that are now present in <italic>D. melanogaster</italic> have invaded the species very recently, during the 20th century. However, all <italic>D. melanogaster</italic> strains contain defective heterochromatic I elements that display an average of 94% sequence identity with each other and with the I factor [<xref ref-type="bibr" rid="B40">40</xref>]. These are very old remnants of former active I elements that were present in the common ancestor of all species from the <italic>D. melanogaster</italic> subgroup and were lost in the ancestor of the <italic>D. melanogaster</italic> species lineage. It is therefore not inconceivable that this kind of event may also have occurred in the case of other non-LTR retrotransposon families. When possible, we have estimated the genetic distance between only the subset of longest available elements within a family. This was done for the Doc, F, G, I, Jockey, JuanDm, Waldo-A, Waldo-B and You families (data not shown). Not surprisingly, the subset of longest elements is constantly more homogeneous than the whole family, indicating that the longest elements are probably still active or result from recent inactivation. One should keep in mind, however, that the error rate in the repeated sequences in Release 1 is not negligible and therefore the degree of heterogeneity observed in a small number of copies is probably overestimated.</p><p>The high rate of errors in repetitive sequences of Release 1 renders thorough analysis of transposable element families uncertain. For example, on the basis of the recognition of reverse transcriptase coding capacities, You elements were not recognized in the sequences of Release 1 (Figure <xref ref-type="fig" rid="F1">1a</xref>) but were identified in the BDGP/EDGP sequences (Figure <xref ref-type="fig" rid="F1">1b</xref>). It is possible that some non-LTR retrotransposon families have been missed by our study. More accurate sequences of repetitive DNA regions are expected in the near future [<xref ref-type="bibr" rid="B17">17</xref>]. They will provide invaluable material for the thorough phylogenetic analyses of families of transposable elements. In particular, analysis of the sequences and organization of the different copies of a given family will provide information useful in understanding the processes by which these elements evolve in the genome.</p><p>The annotations of the sequences of Release 1 are in progress. Most copies of non-LTR retrotransposons that we have found in the present study are not annotated yet. The coordinates of all sequences identified in our studies are available as an additional data file with the online version of this paper.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Identification of reverse transcriptase sequences</title><p><italic>Drosophila melanogaster</italic> sequences produced by Celera [<xref ref-type="bibr" rid="B14">14</xref>] and the BDGP/EDGP [<xref ref-type="bibr" rid="B16">16</xref>] were used in the analysis. A six-frame translation of all the sequences was produced. To reduce the size of the data set to be analyzed by time-consuming HMM software, only amino-acid sequences containing a motif [FY]XDD, which is conserved among reverse transcriptases [<xref ref-type="bibr" rid="B15">15</xref>], were extracted for the analysis. The HMMER 2.1.1 software [<xref ref-type="bibr" rid="B48">48</xref>] was used to identify all sequences in this subset matching the full-length model of reverse transcriptase, which was built using a seed alignment of reverse transcriptase sequences (accession number PF00078) obtained from the Pfam database [<xref ref-type="bibr" rid="B49">49</xref>]. Only matches with scores above zero were considered in the analysis.</p><p>Handling of sequences was facilitated by scripts from the SEALS package [<xref ref-type="bibr" rid="B50">50</xref>]. The results of HMMER searches were analyzed using scripts that we designed specially, grouped into families and classified on the basis of their similarities to known retrotransposons. The relationships between the families of non-LTR retrotransposons were determined by making neighbor-joining trees using the CLUSTAL W software [<xref ref-type="bibr" rid="B51">51</xref>].</p></sec><sec><title>Analysis of non-LTR retrotransposon families</title><p>BLASTN searches were performed in sequences of Release 1 and of BDGP/EDGP using WU-BLAST 2.0 package [<xref ref-type="bibr" rid="B52">52</xref>] and full-length elements from different families as queries. The results were analyzed using scripts that were especially written and based on the BioPerl package [<xref ref-type="bibr" rid="B53">53</xref>]. All high-scoring pairs (HSPs) with percentage identities greater than 70% and lengths greater than 200 nucleotides were used to define distinct copies in the same family. These limiting values were arbitrarily chosen to take into account all diverged and truncated copies in a family while filtering out doubtful hits. HSPs in the same hit were checked manually for overlaps or internal deletions relative to a full-length element and joined together when necessary. Genomic coordinates for all copies were determined in this way and element sequences with flanking regions were extracted for further analysis. Target site duplications were searched in a semi-manual manner with the aid of bl2seq program from the NCBI BLAST package [<xref ref-type="bibr" rid="B54">54</xref>]. Divergence of elements within a family was determined in the following manner: the longest element sequence in a family was used in BLAST searches against all other sequences in the family and resulting BLAST output was converted into multiple alignment with the MView program [<xref ref-type="bibr" rid="B55">55</xref>]. The alignment obtained was used to calculate genetic distances between copies in a family by CLUSTAL W software. Divergence of a family was determined as a mean genetic distance between the longest element and all the other elements in a family. The error rate was estimated by comparing sequences of full-length Jockey and JuanDm elements which have the same flanking sequences (100-200 bp flanks were used) in both Release 1 and BDGP/EDGP databases.</p></sec><sec><title>Southern blots</title><p>Digestion of genomic DNA, gel electrophoresis, transfer on NytranN nylon membranes (Schleicher and Schuell) and hybridization with ³²P-labeled DNA probes were performed following standard procedures [<xref ref-type="bibr" rid="B56">56</xref>] and suppliers' specifications. Hybridizations were carried out overnight at 42&#x000b0;C in 50% formamide. Washes were in 2x standard sodium sulfate (SSC), 0.1% sodium dodecyl sulfate (SDS) followed by 0.1 &#x000d7; SSC, then 0.1% SDS at 42&#x000b0;C. The DNA fragments used as probes were obtained by PCR amplifications from the isogenic <italic>y; cn bw sp</italic> strain genomic DNA using standard conditions with Taq DNA polymerase (Promega). For the JuanDm probe oligonucleotides 5'-AAGGAATAAACCACTAGTGGAGCGCC-3' and 5'-CAGGGAGTGTAAGCTTGGAGTGATG-3' were used as primers. For the You probe oligonucleotides 5'-GATCTTCTTATCAACGCGTACGTGC-3' and 5'-CCCAGGAGTATTGTGGATCCGTTAAG-3' were used as primers.</p></sec></sec><sec><title>Additional data</title><p>The following additional data are included with the online version of this paper: the <xref ref-type="supplementary-material" rid="S1">coordinates</xref> of all sequences identified in this study.</p></sec><sec sec-type="supplementary-material"><title>Supplementary Material</title><supplementary-material content-type="local-data" id="S1"><caption><title>Additional data file 1</title></caption><media xlink:href="gb-2000-1-6-research0012-S1.pdf" mimetype="application" mime-subtype="pdf"><caption><p>The coordinates of all sequences identified in this study</p></caption></media></supplementary-material></sec>

Interkingdom gene fusions

<sec><title>Background:</title><p>Genome comparisons have revealed major lateral gene transfer between the three primary kingdoms of life - Bacteria, Archaea, and Eukarya. Another important evolutionary phenomenon involves the evolutionary mobility of protein domains that form versatile multidomain architectures. We were interested in investigating the possibility of a combination of these phenomena, with an invading gene merging with a pre-existing gene in the recipient genome.</p></sec><sec><title>Results:</title><p>Complete genomes of fifteen bacteria, four archaea and one eukaryote were searched for interkingdom gene fusions (IKFs); that is, genes coding for proteins that apparently consist of domains originating from different primary kingdoms. Phylogenetic analysis supported 37 cases of IKF, each of which includes a 'native' domain and a horizontally acquired 'alien' domain. IKFs could have evolved via lateral transfer of a gene coding for the alien domain (or a larger protein containing this domain) followed by recombination with a native gene. For several IKFs, this scenario is supported by the presence of a gene coding for a second, stand-alone version of the alien domain in the recipient genome. Among the genomes investigated, the greatest number of IKFs has been detected in <italic>Mycobacterium tuberculosis,</italic> where they are almost always accompanied by a stand-alone alien domain. For most of the IKF cases detected in other genomes, the stand-alone counterpart is missing.</p></sec><sec><title>Conclusions:</title><p>The results of comparative genome analysis show that IKF formation is a real, but relatively rare, evolutionary phenomenon. We hypothesize that IKFs are formed primarily via the proposed two-stage mechanism, but other than in the Actinomycetes, in which IKF generation seems to be an active, ongoing process, most of the stand-alone intermediates have been eliminated, perhaps because of functional redundancy.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Wolf</surname><given-names>Yuri I</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Kondrashov</surname><given-names>Alexey S</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" corresp="yes" contrib-type="author"><name><surname>Koonin</surname><given-names>Eugene V</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>Comparative genome analysis has revealed major lateral gene transfer between the three primary kingdoms of life, Bacteria, Archaea, and Eukarya [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. The best recognized form of lateral gene flux is the transfer of numerous genes from mitochondria and chloroplasts to eukaryotic nuclear genomes [<xref ref-type="bibr" rid="B5">5</xref>]. Far beyond that, however, the role of lateral gene exchange, along with lineage-specific gene loss, as one of the principal factors of evolution, at least among prokaryotes, is obvious from the fact that the great majority of conserved families of orthologous genes show a 'patchy' phyletic distribution [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. In many cases, such families are shared by phylogenetically distant species (for example, bacteria and archaea), while they are missing in some of the more closely related species (for example, bacteria from the same lineage). Correlations have been noticed between the preferred routes of gene transfer and the lifestyles of the organisms involved. Thus, massive gene exchange seems to have occurred between archaeal and bacterial hyperthermophiles [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>], whereas certain parasitic bacteria, for example, chlamydia and spirochetes, appear to have acquired significantly more eukaryotic genes than free-living bacteria [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>].</p><p>Another evolutionary trend that is predominant in eukaryotes, but is important also in bacteria and archaea, involves the evolutionary mobility of protein domains that combine to form variable multidomain architectures [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. Domain fusion is one of the foundations of most forms of regulation and signal transduction in the cell. Examples include prokaryotic transcriptional regulators, most of which consist of the DNA-binding helix-turn-helix domain fused to a variety of small-molecule-binding domains [<xref ref-type="bibr" rid="B19">19</xref>], the two-component signal transduction system that is based on fusions of histidine kinases with sensor domains and of receiver domains with DNA-binding domains [<xref ref-type="bibr" rid="B20">20</xref>], and the sugar phosphotransferase (PTS) systems that include complex fusions of several enzymes [<xref ref-type="bibr" rid="B21">21</xref>]. In the evolution of eukaryotes, domain fusion takes the form of domain accretion, whereby proteins from complex organisms (such as animals) that are involved in various forms of regulation and signal transduction tend to accrue multiple domains that facilitate the formation of complex networks of interactions [<xref ref-type="bibr" rid="B22">22</xref>].</p><p>We were interested in exploring the possibility of a meeting between these two major evolutionary phenomena - lateral gene exchange and gene fusion - which would result in the formation of multidomain proteins in which different domains display distinct evolutionary provenance. In particular, we sought to identify fusions between domains originating from different primary kingdoms - Bacteria, Archaea and Eukarya - which we term interkingdom gene (domain) fusions (IKFs), and obtain clues to the pathways of IKF origin through comparative genome analysis. We show that, although IKF in general is a rare phenomenon, one bacterial lineage, the Actinomycetes, displays a significantly increased frequency of such events; we also propose a probable mechanism for IKF formation.</p></sec><sec><title>Results and discussion</title><p>To identify IKFs, all protein sequences encoded in the analyzed genomes were compared to the non-redundant protein database, and those proteins in which distinct parts showed the greatest similarity to homologs from different primary kingdoms were identified (see the Materials and methods section). In most cases, the reported alignments were highly statistically significant, leaving no doubt that true homologs were detected (Table <xref ref-type="table" rid="T1">1</xref>). On the few occasions when the database search statistics in themselves were not fully convincing (for example, the OB-fold nucleic acid-binding domain in the <italic>Bacillus subtilis</italic> protein YhcN and the methyltransferase domain in the YabN protein, also from <italic>B. subtilis</italic>), the homologous relationship was validated by detection of the salient sequence motifs known to be involved in the corresponding protein functions (data not shown). Such motif analysis was performed for all analyzed domains in order not only to validate homology, but also to distinguish between active and inactivated forms of enzymes. Figure <xref ref-type="fig" rid="F1">1</xref> shows multiple alignments of two domains involved in an IKF, illustrating the conservation of the characteristic functional motifs and the specific similarity between each of the domains of the IKF protein (in this case from <italic>Aquifex aeolicus</italic>) and their archaeal and bacterial homologs, respectively.</p><p>In several cases, the chimeric origin of a gene was obvious at a qualitative level because no homolog of the 'alien' domain with comparable sequence similarity was detected in the recipient superkingdom (Table <xref ref-type="table" rid="T1">1</xref>, Figure <xref ref-type="fig" rid="F2">2a</xref>,<xref ref-type="fig" rid="F2">b</xref>). For the rest of the candidate IKFs, phylogenetic tree analysis was performed to corroborate the origin of the invading domain by horizontal transfer; statistically significant grouping of a candidate IKF domain with homologs from the donor superkingdom provides such evidence (Figure <xref ref-type="fig" rid="F2">2c</xref>,<xref ref-type="fig" rid="F2">d</xref>). The overall number of confirmed IKFs is relatively small - 37 in 21 compared genomes (about 0.1% of the genes) - compared to the total number of likely interkingdom gene transfers. For completely sequenced bacterial genomes this has been conservatively estimated as 1-2% of the genes, with a greater fraction (2-10%) detected in archaea and hyperthermophilic bacteria ([<xref ref-type="bibr" rid="B23">23</xref>], and K.S. Makarova, L. Aravind and E.V.K., unpublished observations). Examination of the clusters of orthologous groups (COGs) of proteins from complete genomes [<xref ref-type="bibr" rid="B6">6</xref>], in which multidomain proteins are split into the constituent domains if the orthologs of the latter are present as stand-alone forms in some of the genomes, shows that IKFs constitute only a small fraction of all fusions of evolutionarily mobile domains (Figure <xref ref-type="fig" rid="F3">3</xref>). Generally, the small number of identified IKFs compared to the total number of inferred horizontal transfer events and the total number of domain fusions could be compatible with a random model of domain fusion subsequent to lateral gene transfer.</p><p>However, the distribution of IKFs among genomes is distinctly non-random, suggesting that such a simple model may be incorrect. Specifically, 12 IKFs were detected in <italic>Mycobacterium tuberculosis</italic> and 10 were found in the yeast <italic>Saccharomyces cerevisiae,</italic> but only a small number or none was identified in each of the other bacterial and archaeal genomes (Figure <xref ref-type="fig" rid="F2">2</xref>, Table <xref ref-type="table" rid="T1">1</xref>). The excess of IKFs in <italic>Mycobacterium</italic> is particularly notable, given that the fraction of genes horizontally transferred from archaea and eukaryotes in the mycobacterial genome is only slightly greater than that in most of the other bacteria, and considerably lower than that in the hyperthermophilic bacteria <italic>Aquifex</italic> and <italic>Thermotoga</italic> (K.S. Makarova, L. Aravind and E.V.K., unpublished observations). Similarly, whereas the overall number of domain fusions in <italic>M. tuberculosis</italic> is greater than in most other bacteria, the difference is insufficient to account for the over-representation of IKFs; furthermore, the cyanobacterium <italic>Synechocystis</italic> sp. has an even greater overall number of fusions but does not have any detectable IKFs (Figure <xref ref-type="fig" rid="F3">3</xref>). At present, we cannot provide a defendable biological explanation for the comparatively high frequency of IKF in <italic>Mycobacterium.</italic> It is tempting to interpret this trend in terms of adaptation of this bacterium to its relatively recently occupied parasitic niche, but examination of the individual IKF cases does not offer immediate clues in mycobacterial biology. The yeast IKFs clearly represent relatively recent horizontal transfers distinct from the gene influx from the mitochondria following the establishment of endosymbiosis because, under the protocol of IKF detection used here, only those alien domains were identified that have no counterparts in other eukaryotes.</p><p>Most of the IKFs are unique, but <italic>B. subtilis, M. tuberculosis</italic> and yeast each also encode families of two to three paralogous IKFs, which apparently have evolved by duplication subsequent to the respective fusion events (Table <xref ref-type="table" rid="T1">1</xref>). Strikingly, the same IKF, the three-domain uridine kinase, is shared by <italic>Treponema pallidum</italic> and <italic>Thermotoga maritima</italic> (Table <xref ref-type="table" rid="T1">1</xref>). Given that these two bacteria are not specifically related and that <italic>Borrelia burgdorferi,</italic> the second spirochete whose genome has been sequenced, encodes a typical bacterial uridine kinase, the presence of a common IKF in <italic>Treponema</italic> and <italic>Thermotoga</italic> cannot be realistically attributed to vertical inheritance of this gene from a common ancestor. It thus probably reflects horizontal transfer of the gene encoding the three-domain protein subsequent to its emergence in either the spirochetes or the Thermotogales.</p><p>Two evolutionary issues pertaining to IKFs need to be addressed, namely the mechanism(s) of their origin and the selective forces responsible for their preservation. From general considerations, it seems likely that IKFs have evolved via a two-step process, which involves lateral transfer of the complete gene coding for the IKF's alien portion, followed by domain fusion. This scenario rests on the assumption that the acquired foreign gene is selectively advantageous, because otherwise it would have been inactivated by mutations before recombination could take place. Under this mechanism, the alien portion of an IKF is likely to be present in the recipient genome also as a stand-alone gene. A clear-cut case of such a duplication of a horizontally transferred domain has been noticed in <italic>Chlamydia,</italic> whose genomes encode the SWI domain, implicated in chromatin condensation, both as a stand-alone protein and as the carboxy-terminal portion of topoisomerase I [<xref ref-type="bibr" rid="B10">10</xref>]. Apart from this case, the IKFs fall into two readily discernible classes, namely those from <italic>Mycobacterium</italic> and all the rest. <italic>M. tuberculosis</italic> (the only complete genome of an actinomycete available) possesses considerably more IKFs than any other bacterial or archaeal species (see above), and typically, the alien portions of these proteins show high level of similarity to the homologs from the donor superkingdom (eukaryotes). Most significantly, there is also, with a single exception, a stand-alone counterpart in the mycobacterial genome; in some cases, such a counterpart is seen only in a closely related species, <italic>M. leprae,</italic> and in one case, it is found in <italic>Streptomyces,</italic> a distantly related actinomycete (Table <xref ref-type="table" rid="T1">1</xref>). In the other genomes, the IKFs are generally less similar to the apparent donor and, with a few exceptions, stand-alone versions of the alien domains are missing (Table <xref ref-type="table" rid="T1">1</xref>). The hypothesis that seems to be most compatible with these observations is that IKFs indeed evolve via a stand-alone, horizontally transferred intermediate, but in the case of ancient IKFs, these intermediates are typically eliminated during evolution, perhaps because their function becomes redundant with the formation of the IKF. The IKFs identified in actinomycetes appear to result from relatively recent gene fusion events so that the original, stand-alone transferred genes are still present in the genome.</p><p>The IKFs include a variety of protein functions. Only some of these are well understood such as, for example, those of the bifunctional nucleotide and coenzyme metabolism enzymes that are particularly abundant in yeast (Table <xref ref-type="table" rid="T1">1</xref>). In other cases, the function of an IKF-encoded protein could be predicted only tentatively on the basis of the functions of its constituent domains (Table <xref ref-type="table" rid="T1">1</xref>). The selective advantage of the formation of multidomain proteins, at least as far as enzymes are involved, lies in the possibility of effective coupling of the reactions catalyzed by the different domains [<xref ref-type="bibr" rid="B16">16</xref>]; this may be generalized also for functional coordination of non-enzymatic domains. Fusion may result in the addition of a regulatory function to an enzymatic one. For example, it appears most likely that the RNA-binding TGS domain [<xref ref-type="bibr" rid="B24">24</xref>] in the uridine kinases of <italic>Treponema pallidum</italic> and <italic>Thermotoga maritima</italic> is involved in autoregulation of translation. The unusual aspect of the IKFs appears to be the compatibility of evolutionarily distant domains.</p><p>Examination of the phyletic distribution of the multidomain architectures of IKFs may help in pinpointing the evolutionary stage at which the fusion (but not necessarily the preceding horizontal gene transfer) has occurred. For example, the fusion of the SWI domain with topoisomerase belongs after the radiation of Chlamydia from other bacterial lineages, but before the radiation of <italic>Chlamydia pneumoniae</italic> and <italic>Chlamydia trachomatis</italic> (Table <xref ref-type="table" rid="T1">1</xref>). The majority of IKFs detected in the yeast <italic>S. cerevisiae</italic> are also present in <italic>Schizosaccharomyces pombe</italic> and/or other ascomycetes (Table <xref ref-type="table" rid="T1">1</xref>, and data not shown), but not in any other eukaryotes, and accordingly, they should have evolved at a relatively early stage of fungal evolution, but not before the fungal clade diverged from the rest of the eukaryotic crown group.</p><p>Finally, it should be noted that formation of some of the IKFs might have required more complex rearrangements of the contributing proteins than simple domain fusion. Figure <xref ref-type="fig" rid="F4">4</xref> shows the domain architectures of proteins that contribute domains to two IKFs. In each case, a simple fusion between genes encoding the respective individual domains is insufficient to explain the emergence of the IKF. For example, the uridine kinase example mentioned above (Figure <xref ref-type="fig" rid="F4">4a</xref>) should have involved isolation of the TGS-HxxxH domains of threonyl-tRNA synthetase before or concomitantly with their fusion with the uridine kinase. The specific molecular mechanism could have involved selective duplication of the upstream portion of the threonyl-tRNA synthetase gene. Similarly, the sialic acid synthase homologous domain, which is fused to hydroxymethylpyrimidine phosphate kinase in <italic>A. pernix</italic> and pyrococci, appears to have been derived from two-domain proteins that additionally contain a helix-turn-helix DNA-binding domain (Figure <xref ref-type="fig" rid="F4">4b</xref>). These hypotheses of a complex mechanism of gene fusion involved in the emergence of IKFs are based on a limited sample of sequenced genomes. An alternative possibility is that, before the postulated horizontal transfer event, the recipient domain(s) has been encoded by a stand-alone gene; such genes that do not contain the fused alien domain may yet be discovered in newly sequenced genomes. In fact, a stand-alone version of the sialic acid synthase homologous domain is seen in <italic>Methanobacterium,</italic> although it is considerably less similar to the IKF than the version fused to the HTH domain (Figure <xref ref-type="fig" rid="F4">4b</xref>).</p><p>The identification of IKFs underscores the complexity of the evolutionary process as revealed by comparison of multiple genomes. In and by itself, this phenomenon may not have a unique biological significance, but it reveals the overlap between two major evolutionary trends, horizontal gene transfer and protein domain rearrangement, and shows that domains, rather then entire proteins (genes), should be considered fundamental units of genetic material exchange.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><p>Protein sequences encoded in 21 complete genomes of archaea, bacteria and the yeast <italic>Saccharomyces cerevisiae</italic> were extracted from the Genome division of the Entrez retrieval system [<xref ref-type="bibr" rid="B25">25</xref>]. Each protein encoded in these genomes was used as the query in a comparison against the non-redundant protein sequence database (National Center for Biotechnology Information, NIH, Bethesda, USA) using the BLASTP program [<xref ref-type="bibr" rid="B26">26</xref>]. For each query, the set of local similarities detected by BLASTP was automatically (using a Perl script written for this purpose) screened for putative IKFs, that is situations in which the query did not have full-size homologs outside its immediate taxonomic group (for example, the Proteobacteria for <italic>Escherichia coli</italic>) and in which different regions of the query showed the greatest similarity to proteins from different primary kingdoms. The pseudocode for the script follows:</p><graphic xlink:href="gb-2000-1-6-research0013-i1.gif"/><p>The script itself is available as an <xref ref-type="supplementary-material" rid="S1">additional data file</xref>. The candidate IKF cases were further examined to detect situations where one or more distinct regions of the query could be classified as 'native' or 'alien' either on the basis of the lack of close homologs from the respective primary kingdom or using phylogenetic analysis. Multiple sequence alignments were generated using the ClustalW program [<xref ref-type="bibr" rid="B27">27</xref>], and when necessary, manually corrected to ensure the proper alignment of conserved motifs typical of the respective domains. Phylogenetic trees were constructed using the PROTDIST and FITCH programs of the PHYLIP package [<xref ref-type="bibr" rid="B28">28</xref>]. Trees were made separately for each domain of a putative IKF, and its mixed ancestry was considered confirmed if the affinities of the domains with different primary kingdoms were supported by bootstrap values of at least 50%. Additional iterative database searches were performed using the PSI-BLAST program [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B29">29</xref>] in order to predict functions of the individual domains of the identified IKFs in cases when these were not immediately clear.</p></sec><sec><title>Additional data</title><p>The following additional data are included with the online version of this paper: the <xref ref-type="supplementary-material" rid="S1">Perl script</xref> used to screen local similarities for putative IKFs.</p></sec><sec sec-type="supplementary-material"><title>Supplementary Material</title><supplementary-material content-type="local-data" id="S1"><caption><title>Additional data file 1</title></caption><media xlink:href="gb-2000-1-6-research0013-s1.txt" mimetype="text" mime-subtype="plain"><caption><p>The Perl script used to screen local similarities for putative IKFs</p></caption></media></supplementary-material></sec>

Genetic snapshots of the <italic>Rhizobium</italic> species NGR234 genome

<sec><title>Background:</title><p>In nitrate-poor soils, many leguminous plants form nitrogen-fixing symbioses with members of the bacterial family Rhizobiaceae. We selected <italic>Rhizobium</italic> sp. NGR234 for its exceptionally broad host range, which includes more than I 12 genera of legumes. Unlike the genome of <italic>Bradyrhizobium japonicum,</italic> which is composed of a single 8.7 Mb chromosome, that of NGR234 is partitioned into three replicons: a chromosome of about 3.5 Mb, a megaplasmid of more than 2 Mb (pNGR234<italic>b</italic>) and pNGR234<italic>a</italic>, a 536,165 bp plasmid that carries most of the genes required for symbioses with legumes. Symbiotic loci represent only a small portion of all the genes coded by rhizobial genomes, however. To rapidly characterize the two largest replicons of NGR234, the genome of strain ANU265 (a derivative strain cured of pNGR234<italic>a</italic>) was analyzed by shotgun sequencing.</p></sec><sec><title>Results:</title><p>Homology searches of public databases with 2,275 random sequences of strain ANU265 resulted in the identification of 1,130 putative protein-coding sequences, of which 922 (41%) could be classified into functional groups. In contrast to the 18% of insertion-like sequences (ISs) found on the symbiotic plasmid pNGR234<italic>a</italic>, only 2.2% of the shotgun sequences represent known ISs, suggesting that pNGR234<italic>a</italic> is enriched in such elements. Hybridization data also indicate that the density of known transposable elements is higher in pNGR234<italic>b</italic> (the megaplasmid) than on the chromosome. <italic>Rhizobium</italic>-specific intergenic mosaic elements (RIMEs) were found in 35 shotgun sequences, 6 of which carry RIME2 repeats previously thought to be present only in <italic>Rhizobium meliloti</italic>. As non-overlapping shotgun sequences together represent approximately 10% of ANU265 genome, the chromosome and megaplasmid may carry a total of over 200 RIMEs.</p></sec><sec><title>Conclusions:</title><p>'Skimming' the genome of <italic>Rhizobium</italic> sp. NGR234 sheds new light on the fine structure and evolution of its replicons, as well as on the integration of symbiotic functions in the genome of a soil bacterium. Although most putative coding sequences could be distributed into functional classes similar to those in <italic>Bacillus subtilis,</italic> functions related to transposable elements were more abundant in NGR234. In contrast to ISs that accumulated in pNGR234<italic>a</italic> and pNGR234<italic>b</italic>, the hundreds of RIME elements seem mostly attributes of the chromosome.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Viprey</surname><given-names>Virginie</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Rosenthal</surname><given-names>Andr&#x000e9;</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" corresp="yes" contrib-type="author"><name><surname>Broughton</surname><given-names>William J</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Perret</surname><given-names>Xavier</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>Many different Gram-negative bacteria colonize the nutrient-rich rhizospheres of plant roots. Some bacteria are pathogenic, whereas others form beneficial associations. In nitrate-poor soils, strains of <italic>Azorhizobium, Bradyrhizobium, Mesorhizobium</italic> and <italic>Rhizobium</italic> (collectively known as rhizobia), form nitrogen-fixing symbioses with leguminous plants. In compatible interactions, invading rhizobia penetrate their hosts through infection threads, which develop centripetally. At the same time, new structures called nodules develop from meristems induced in the cortex of infected roots. When infection threads reach nodule cells, rhizobia are released as symbiosomes into the cytoplasm of infected cells where they eventually enlarge and differentiate into nitrogen-fixing bacteroids. Continuous exchange of chemical signals between the two symbionts coordinates expression of bacterial and plant genes required for a symbiotic development. Flavonoids released by legume roots are amongst the first signals exchanged in this molecular dialog. By interacting with rhizobial regulators of the NodD family, flavonoids trigger the expression of nodulation genes (<italic>nod</italic>, <italic>noe</italic> and <italic>nol).</italic> In turn, most nodulation genes participate in the synthesis and secretion of a family of lipochito-oligosaccharide molecules, the Nod factors that are required for bacterial entry into root hairs. Little is known about how rhizobia migrate inside the infection threads, although it seems likely that genetic determinants of both partners are again involved (see [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]). Once within the cortex, the rhizobia differentiate into bacteroids where low free-oxygen tensions help coordinate the expression of genes involved in nitrogen fixation (<italic>nif</italic> and <italic>fix</italic>) [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>Taxonomic proposals based on DNA sequences of highly conserved genes indicate that rhizobia are a group of genetically diverse soil bacteria [<xref ref-type="bibr" rid="B4">4</xref>]. Other data suggest that in populations of soil bacteria, natural genetic mechanisms exist which can transform isolates with widely different chromosomal backgrounds into nodulating bacteria (that is, rhizobia) (for review see [<xref ref-type="bibr" rid="B1">1</xref>]). Comparisons of genomes of soil bacteria will help define the pools of symbiotic genes. Unfortunately, genomic studies of this kind have been hindered by the relatively large size of rhizobial genomes (6.5 to 8.7 Mb for <italic>R. meliloti</italic> and <italic>B. japonicum,</italic> respectively). Instead, as many symbiotic loci are often clustered on large plasmids in <italic>Rhizobium</italic> strains, or in chromosomal 'symbiotic islands' as in <italic>B. japonicum</italic> [<xref ref-type="bibr" rid="B5">5</xref>] and <italic>M. loti</italic> [<xref ref-type="bibr" rid="B6">6</xref>], physical and genetic analyzes of symbiotic plasmids or 'islands' prevailed. <italic>Rhizobium</italic> sp. NGR234 was selected for its exceptionally broad host range, which includes more than 112 genera of legumes in addition to the non-legume <italic>Parasponia andersonii</italic> [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. As in <italic>R. meliloti,</italic> the genome of NGR234 is partitioned into three replicons, a chromosome of about 3.5 Mb, a megaplasmid of more than 2 Mb (pNGR234<italic>b</italic>) and pNGR234<italic>a</italic>, a 536 kb symbiotic plasmid [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>]. Although various experiments have shown that most symbiotic genes are amongst the 416 open reading frames (ORFs) identified in the complete sequence of pNGR234<italic>a</italic> [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>], others are carried by the chromosome and/or the mega-plasmid [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B14">14</xref>].</p><p>Many ways of finding genes exist, but with the rapid advances in genomics, among the most effective are those that involve sequencing parts of or entire genomes. Although contiguous sequences of several symbiotic islands/plasmids will be released in the near future, <italic>R. meliloti</italic> strain 1021 as well as the phytopathogens <italic>Ralstonia solanacearum</italic> and <italic>Xanthomonas citri</italic> are the only plant-interacting microbes currently being sequenced [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. The cost of sequencing a complete genome is still well beyond the capability of most laboratories, however. Nevertheless, extensive information on the structure and content of genomes can be gained by randomly sequencing libraries made from total DNA [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. Here, we have used this approach to analyze the megaplasmid and chromosome of NGR234. A total of 2,275 individual shotgun sequences of ANU265 (a derivative strain of NGR234 cured of its symbiotic plasmid [<xref ref-type="bibr" rid="B22">22</xref>]) were searched for protein and/or DNA homologies, and putative coding sequences were grouped into 28 classes according to their putative function. In addition, clones carrying various <italic>Rhizobium</italic>-specific repeated elements such as RIME1 and RIME2 were also analyzed.</p></sec><sec><title>Results and discussion</title><sec><title>Random sequencing of the ANU265 genome</title><p>Total genomic DNA of ANU265 was used to construct an M13 library with inserts ranging in size from 0.9 to 1.5 kb. Of the 2,856 random clones analyzed, 80% (2,275) produced high-quality DNA sequence with an average read length of 253 bp (Table <xref ref-type="table" rid="T1">1</xref>). In this way, more than 575 kb of total nucleotide sequence was collected, which corresponds to approximately 10% of the ANU265 genome [<xref ref-type="bibr" rid="B11">11</xref>]. At 61.2 mol%, the mean G+C content of these sequences is similar to that found for the entire genome [<xref ref-type="bibr" rid="B23">23</xref>], but is also significantly higher than the value of 58.5 mol% calculated for pNGR234<italic>a</italic> [<xref ref-type="bibr" rid="B9">9</xref>]. This pool of 2,275 sequences was then screened for redundancy. A total of 381 overlapping sequences were identified, and grouped into 195 contigs (sets of overlapping sequences) of two to four elements each: 154 contigs represent pairs of clones, whereas the remaining 73 sequences belong to 23 groups of three elements and one of four clones. Because of the many highly conserved sequences repeated throughout the NGR234 genome [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B24">24</xref>], it was not possible to determine if overlapping clones represent contiguous sequences or DNA fragments from distinct repeats. Nevertheless, truly unique sequences represent 92% of the total number of clones. With an average insert size of 1.2 kb, clones tagged with non-overlapping sequences represent more than 40% (2.5 Mbp) of the ANU265 genome.</p></sec><sec><title>RIME- and IS-like sequences</title><p>Homology searches against nucleotide databases (BLASTN [<xref ref-type="bibr" rid="B25">25</xref>]) showed that 35 ANU265 sequences carried <italic>Rhizobium-</italic>specific intergenic mosaic elements (RIMEs). First identified in <italic>R. meliloti, R. leguminosarum</italic> bv. <italic>viciae</italic> and NGR234, RIME1 elements are 108 bp repeats characterized by two large palindromes, whereas RIME2 sequences are 109 bp repeats thought to be present only in <italic>R. meliloti</italic> [<xref ref-type="bibr" rid="B26">26</xref>]. RIMEs have many features of the short interspersed repeated elements that are non-coding, intercistronic sequences of less than 200 bp found in many prokaryotic genomes [<xref ref-type="bibr" rid="B27">27</xref>]. Of the 2,275 shotgun sequences of ANU265 collected, 29 contained RIME1 elements and 6 carried RIME2 repeats. Although Southern hybridizations indicated that approximately 20 copies of RIME1 were present in the genomes of. <italic>R. meliloti</italic> and NGR234 [<xref ref-type="bibr" rid="B26">26</xref>], our data indicate that there are many more. Among the 29 clones with RIME1 sequences, most (23) carry repeats that are very similar to the consensus ([<xref ref-type="bibr" rid="B26">26</xref>] and Figure <xref ref-type="fig" rid="F1">1</xref>). In another six (Figure <xref ref-type="fig" rid="F1">1</xref>, clones 27d06, 29g08, 0lf01, 11b07, 25e07 and 13c06), only one of the two large palindromic structures is conserved, however. This suggests that, in some cases, individual palindromes constitute independent repeats, not necessarily associated to form RIME1 elements. In the eight clones that code for putative proteins (Figure <xref ref-type="fig" rid="F1">1</xref>), RIME1 sequences are found immediately downstream of predicted ORFs (data not shown), indicating that these elements are probably confined to intergenic regions. Surprisingly, no RIME2 and a single RIME1 repeat were found on pNGR234<italic>a</italic> [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B11">11</xref>]. If these elements were regularly distributed throughout the NGR234 genome, more than a single RIME1 would have been expected on the 536 kb of pNGR234<italic>a</italic>. Thus, current data suggest that RIMEs preferentially accumulate on specific replicons, and that NGR234 carries possibly as many as 200 RIME-like elements.</p><p>In contrast to pNGR234<italic>a</italic>, which carries many IS sequences, only 2.2% (51) of the 2,275 ANU265 sequences were predicted to encode transposon-related functions. Although several clones that did not match database homologs may also carry sequences of yet uncharacterized IS elements, these results suggest that in proportion to their size, chromosome and megaplasmid carry fewer transposable elements than pNGR234<italic>a</italic>. Nevertheless most of the 51 clones (70%) matched ISs that were first identified in pNGR234<italic>a</italic> [<xref ref-type="bibr" rid="B9">9</xref>]. For example, ten sequences highly homologous to NGRIS-4 were found. This 3,316 bp element is duplicated in pNGR234<italic>a</italic> [<xref ref-type="bibr" rid="B9">9</xref>], whereas chromosome and megaplasmid carry two and five copies of NGRIS-4 respectively [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B24">24</xref>].</p></sec><sec><title>Identification of putative genes</title><p>To assign putative functions to the cloned DNA fragments, sequences were compared to protein and nucleotide databases [<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B28">28</xref>]. BLAST analyses showed that about 50% (1,130) of the 2,275 sequences matched protein-coding ORFs, three were homologous to rDNA and four to tRNA loci (see Table <xref ref-type="table" rid="T1">1</xref>). Of the 1,130 putative protein-coding sequences, 208 (or 9% of the 2,275 sequences) were similar to hypothetical genes with no known function (pioneer sequences) of rhizobia and other organisms. Thus, together with the 1,109 clones which showed no significant similarity to entries in nucleotide and amino-acid databases (see Table <xref ref-type="table" rid="T1">1</xref>), functions could not be assigned to 58% of the shotgun sequences. To provide an overview of the genetic organization of the ANU265 genome, predicted protein-coding sequences were grouped into various classes according to their putative function (Table <xref ref-type="table" rid="T2">2</xref>).</p></sec><sec><title>A genetic snapshot of the ANU265 genome</title><p>In total, 922 of the 2,275 sequences were grouped into 28 functional categories (Table <xref ref-type="table" rid="T2">2</xref>). Interestingly, comparison of this data with that derived from the complete sequence of the <italic>Bacillus subtilis</italic> genome [<xref ref-type="bibr" rid="B29">29</xref>] showed a similar distribution of genes in both organisms. Although <italic>B. subtilis</italic> is a Gram-positive bacterium, it is commonly found in soil, water sources and in associations with plants. Thus, with the exception of one homolog of a sporulation gene (which was not expected in rhizobia), the comparative analysis presented in Table <xref ref-type="table" rid="T2">2</xref> suggests that the number of shotgun sequences is probably sufficiently large to form a representative selection of ANU265 loci. All 1,130 sequences for which significant matches were found in database searches are classified by function in Table <xref ref-type="table" rid="T3">3</xref>.</p><p>As in other bacterial genomes, such as that of <italic>Escherichia coli</italic> [<xref ref-type="bibr" rid="B30">30</xref>], the largest functional class represents transport and binding proteins (see Tables <xref ref-type="table" rid="T2">2</xref> and <xref ref-type="table" rid="T3">3</xref>). A number of essential genes, including those required for replication, transcription and translation as well as those linked to primary metabolism, were also found. As expected of a soil-borne prokaryote, many loci (18%) involved in carbon and nitrogen metabolism were identified (encoding enzymes for the assimilation of nitrate/ammonia, the tri-carboxylic acid cycle, or transporters of dicarboxylic acids, and so on). In <italic>B. subtilis,</italic> 19% of the protein-coding genes are devoted to the metabolism of carbohydrates, amino acids and related molecules (Table <xref ref-type="table" rid="T2">2</xref>). This is in contrast to microorganisms such as <italic>Haemophilus influenzae</italic> and <italic>M. genitalium</italic> that are not able to grow on many nitrogen and carbon sources (only 10% of their predicted genes code for such metabolic functions [<xref ref-type="bibr" rid="B31">31</xref>]). Interestingly, homologs of various chaperones such as GroES/GroEL, DnaJ, and other small heat-shock proteins (sHsps), were identified (Table <xref ref-type="table" rid="T3">3</xref>, clones 308 to 318). The presence of multiple sHsps is not common in prokaryotes, but was shown to be widespread in rhizobia [<xref ref-type="bibr" rid="B32">32</xref>].</p><p>Obviously, the ability of rhizobia to respond to plant compounds that stimulate their growth contributes to successful colonization of the root [<xref ref-type="bibr" rid="B33">33</xref>] and absence of vitamins often limits the growth or rhizobia. Furthermore, the ability to either take up or synthesize vitamins is thought to be an essential characteristic of rhizobia [<xref ref-type="bibr" rid="B33">33</xref>]. For these reasons, it is not surprizing that several ANU265 sequences matched genes for biotin and thiamine utilization, such as that coding for a homolog of <italic>bioS</italic> (clone 745), a biotin-regulated locus of <italic>R. meliloti</italic> [<xref ref-type="bibr" rid="B34">34</xref>]. In <italic>R. meliloti, bioS</italic> is part of an operon which includes the <italic>surE</italic> and <italic>IppB/nlpD</italic> genes that are also found in ANU265 (clones 744 and 183). Homologs of thiamine biosynthetic genes <italic>thiCG of R. etli</italic> (clones 512 and 513) were also found. Miranda-Rios <italic>et al.</italic> [<xref ref-type="bibr" rid="B35">35</xref>] reported a direct correlation between the expression of <italic>thiC</italic> and the production of the symbiotic terminal oxidase <italic>cbb</italic>3<italic>,</italic> which is required for bacteroid respiration under conditions of low oxygen.</p><p>Putative symbiotic genes include loci involved in exopolysaccharide (EPS) biosynthesis and/or export, which are encoded by pNGR234<italic>b</italic> [<xref ref-type="bibr" rid="B10">10</xref>], as well as genes involved in the elaboration of acidic capsular polysaccharides (K-anti-gens), lipopolysaccharides and cyclic &#x003b2;-glucans (Table <xref ref-type="table" rid="T3">3</xref>, clones 245 to 270). A sequence homologous to <italic>fixN</italic> of <italic>R. meliloti</italic> was also identified (clones 208 and 209). The chromosomal <italic>fixNOPQ</italic> locus encodes an oxidase complex that is probably active during nitrogen fixation. Although sequences of the regulatory <italic>fixK</italic> genes [<xref ref-type="bibr" rid="B3">3</xref>] were identified (clone 683), no significant match to the oxygen-responsive system encoded by <italic>fixLJ</italic> was found. Members of other symbiotic two-component regulatory systems were detected in ANU265, however, including homologs of the sensor histi-dine kinase <italic>exoS</italic> (clone 200) and the response regulator <italic>chvI</italic> (clone 717). Both are necessary for regulating production of succinoglycans that are important in <italic>R. meliloti-Medicago sativa</italic> symbioses [<xref ref-type="bibr" rid="B36">36</xref>]. Similarly, the <italic>nwsA</italic> locus (clone 202) encodes a putative sensor kinase that is involved in the expression of nodulation genes in <italic>Bradyrhizobium</italic> strains [<xref ref-type="bibr" rid="B37">37</xref>].</p><p>It has been postulated that genes responsible for the synthesis (<italic>mos</italic>) and catabolism (<italic>moc</italic>) of rhizopines confer a competitive advantage on their host rhizobia [<xref ref-type="bibr" rid="B38">38</xref>]. Rhizopines are synthesized in nodules of <italic>M. sativa</italic> inoculated with <italic>R. meliloti</italic> strain L5-30, and can be used as growth substrates by certain rhizobia. Although <italic>mos</italic> and <italic>moc</italic> genes were thought to be limited to <italic>R. meliloti</italic> strains [<xref ref-type="bibr" rid="B39">39</xref>], homologs of <italic>mocABC,</italic> and <italic>mosA</italic> genes were also found in ANU265 (clones 543 to 549). Propagation of rhizobia in the soil, and hence their symbiotic efficiency, probably also depends on their tolerance to osmotic changes. It is thus notable that homologs of the <italic>R. meliloti betABC</italic> genes, which are involved in the osmoregulatory choline-glycine betaine pathway [<xref ref-type="bibr" rid="B40">40</xref>], were also found (clones 726 to 730).</p><p>Other putative symbiotic loci include homologs of the <italic>phbC</italic> and <italic>prsDE</italic> genes of <italic>R. meliloti,</italic> which encode a poly-3-hydroxybutyrate synthase [<xref ref-type="bibr" rid="B41">41</xref>] and a type I secretion system [<xref ref-type="bibr" rid="B42">42</xref>] (clones 741 to 743, and 298 to 301, respectively). Interestingly, PrsD and PrsE of <italic>R. meliloti</italic> are involved in the secretion of enzymes that modify succinoglycans [<xref ref-type="bibr" rid="B43">43</xref>], whereas a similar type I secretion system seems to be responsible for the export of the nodulation-signaling protein NodO in <italic>R. leguminosarum</italic> bv. <italic>viciae</italic> [<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>]. Although the role of these <italic>prsDE</italic> homologs in NGR234 is not clear, it is possible that more than one type of protein secretion system has a symbiotic role in this bacterium [<xref ref-type="bibr" rid="B46">46</xref>].</p></sec></sec><sec><title>Conclusions</title><p>Random sequencing of ANU265 followed by homology searches of public databases resulted in the identification of 1,130 putative protein-coding sequences, of which 922 (41%) could be classified into functional groups. Comparison of these data with those derived from the complete sequence of the <italic>B. subtilis</italic> genome showed a similar distribution of putative coding sequences, except perhaps for functions related to transposable elements (Table <xref ref-type="table" rid="T2">2</xref>). In fact, the genome of ANU265 carries more putative transposases and other IS-related functions (5.5% of all identified genes, and 2.2% of all shotgun sequences) than that of <italic>B. subtilis.</italic> Nevertheless, in proportion to their size, the chromosome and megaplasmid of NGR234 carry fewer IS sequences than pNGR234<italic>a</italic>. Furthermore, hybridization data indicate that the density of known transposable elements is higher in pNGR234<italic>b</italic> than on the chromosome (order of IS accretion is: pNGR234<italic>a</italic> &#x0003e; pNGR234<italic>b</italic> &#x0003e; chromosome) [<xref ref-type="bibr" rid="B11">11</xref>]. This suggests that IS elements preferentially accumulate on plasmids, possibly because they are less likely to disrupt essential functions. In contrast, the many RIME elements present in NGR234 are clearly more abundant on the chromosome and megaplasmid than on pNGR234<italic>a</italic>. Together, the distinct G+C contents and structural features of the symbiotic plasmid, megaplasmid and chromosome suggest that different evolutionary constraints and histories contributed to shape these three replicons.</p><p>'Skimming' the genome of <italic>Rhizobium</italic> sp. NGR234 has given new insights into the evolution of its replicons and the integration of symbiotic functions in the genome of a soil bacterium. It also reinforced the assumption, which originated from host-range extension experiments [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B47">47</xref>], that pNGR234<italic>a</italic> carries most of the symbiotic genes. Although few <italic>nod, nif</italic> and <italic>fix</italic> homologs were found amongst the random clones, it is likely that additional chromosome- and megaplasmid-encoded functions contribute to successful symbioses between NGR234 and its many host plants. In this respect, transcriptional analyses using shotgun sequences as hybridization templates [<xref ref-type="bibr" rid="B11">11</xref>] will help identify such new symbiotic loci.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Microbiological techniques</title><p><italic>Rhizobium</italic> strain ANU265 [<xref ref-type="bibr" rid="B19">19</xref>], a strain of <italic>Rhizobium</italic> sp. NGR234 [<xref ref-type="bibr" rid="B7">7</xref>] cured of pNGR234<italic>a</italic>, was grown in Rhizobium minimal medium supplemented with succinate (RMM) [<xref ref-type="bibr" rid="B47">47</xref>]. <italic>Escherichia coli was</italic> grown on SOC or in TY [<xref ref-type="bibr" rid="B48">48</xref>]. Subclones in M13mp18 vectors [<xref ref-type="bibr" rid="B49">49</xref>] were grown in <italic>E. coli</italic> strain DH5&#x003b1; F'IQ [<xref ref-type="bibr" rid="B50">50</xref>].</p></sec><sec><title>Preparation of the random genomic library and M13 templates</title><p>Genomic DNA <italic>of Rhizobium</italic> strain ANU265 was prepared as in Perret and Broughton [<xref ref-type="bibr" rid="B51">51</xref>]. ANU265 genomic DNA (15 &#x003bc;g) was sheared by sonication and incubated for 10 min at 30&#x000b0;C with 30 units of mung bean nuclease. The resulting digest was extracted with phenol/chloroform (1:1) and precipitated with ethanol. Fragments ranging in size from 900 to 1,500 bp were purified from agarose gels and ligated into <italic>Sma</italic>I-digested M13mp18 vector DNA. Ligation mixtures were electroporated into <italic>E. coli</italic> strain DH5&#x003b1;F'IQ [<xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B52">52</xref>], and transformants were plated on 5-bromo-4-chloro-indoyl-&#x003b2;-D-galactoside (X-Gal) and isopropyl-&#x003b2;-thiogalactopyranoside (IPTG)-containing petri dishes [<xref ref-type="bibr" rid="B48">48</xref>]. Fresh 1 ml cultures of <italic>E. coli</italic> DH5&#x003b1;F'IQ were infected with phages from randomly selected white plaques, and grown for 6 h at 37&#x000b0;C in TY medium. Phages were precipitated from 600 &#x003bc;l of the culture supernatant by adding 150 &#x003bc;l 2.5 M NaCl/20% polyethylene glycol (PEG-8,000) (20 min at 25&#x000b0;C). Afterwards, they were centrifuged for 20 min at 3,000<italic>g</italic> at 25&#x000b0;C, and resuspended in 20 &#x003bc;l Triton-TE extraction buffer (0.5% Triton X-100; 10 mM Tris-HCl, 1 mM EDTA pH 8.0). Following 10 min incubation at 80&#x000b0;C and ethanol precipitation, single-stranded phage DNA was recovered in 50 &#x003bc;l H₂O.</p></sec><sec><title>Sequence analysis</title><p>Dye-terminator cycle sequencing of individual M13 sub-clones, gel electrophoresis and sequence editing was performed as described by Freiberg <italic>et al.</italic> [<xref ref-type="bibr" rid="B53">53</xref>]. Shotgun sequences were checked for redundancy using the XGAP program [<xref ref-type="bibr" rid="B54">54</xref>] and for significant homologies with BLASTX-BLASTN software [<xref ref-type="bibr" rid="B55">55</xref>] using nonredundant databases at NCBI [<xref ref-type="bibr" rid="B25">25</xref>].</p></sec></sec>

Expression profiles during honeybee caste determination

<sec><title>Background</title><p>Depending on their larval environment, female honeybees develop into either queens or workers. As in other polyphenisms, this developmental switch depends not on genomic differences between queens and workers but on the differential expression of entire suites of genes involved with larval fate. As such, this and other polyphenic systems can provide a novel tool for understanding how genomes and environmental conditions interact to produce different developmental trajectories. Here we use gene-expression profiles during honeybee caste determination to present the first genomic view of polyphenic development.</p></sec><sec><title>Results</title><p>Larvae raised as queens or workers differed greatly in their gene-expression patterns. Workers remained more faithful than queens to the expression profiles of younger, bipotential, larvae. Queens appeared to both downregulate many of the genes expressed by bipotential larvae and turn on a distinct set of caste-related genes. Queens overexpressed several metabolic enzymes, workers showed increased expression of a member of the cytochrome P450 family, hexameric storage proteins and dihydrodiol dehydrogenase, and young larvae overexpressed two putative heat-shock proteins (70 and 90 kDa), and several proteins related to RNA processing and translation.</p></sec><sec><title>Conclusions</title><p>Large differences in gene expression between queens and workers indicate that social insect castes have faced strong directional selection pressures. Overexpression of metabolic enzymes by queen-destined larvae appears to reflect the enhanced growth rate of queens during late larval development. Many of the differently expressed genes we identified have been tied to metabolic rates and cellular responses to hormones, a result consistent with known physiological differences between queen and worker larvae.</p></sec>

<contrib id="A1" corresp="yes" contrib-type="author"><name><surname>Evans</surname><given-names>Jay D</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Wheeler</surname><given-names>Diana E</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib>

Genome Biology

<sec><title>Background</title><p>Social insect colonies are defined in part by reproductive division of labor, whereby some colony members are considerably more fecund than others. The reproductive roles of colony members often are fixed by events that occur during larval development. An understanding of the mechanisms behind role, or caste, determination allows tests of several models for the evolution and maintenance of social life [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Female honeybees (<italic>Apis mellifera</italic>) are fated to become queens or workers during the first few days of larval development. Despite detailed knowledge of the physiological differences between workers and queens during development [<xref ref-type="bibr" rid="B4">4</xref>], relatively little is known of the underlying genetic machinery that drives these changes. Although differences in honeybee larval transcription had been inferred [<xref ref-type="bibr" rid="B5">5</xref>], caste-biased genes were identified only recently [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. On the basis of inferred homology with known genes [<xref ref-type="bibr" rid="B3">3</xref>], these genes appear to be involved with diverse metabolic, nutritional and signaling processes during development.</p><p>Gene-expression arrays provide a powerful new tool to clarify the timing and nature of genetic events during insect development [<xref ref-type="bibr" rid="B8">8</xref>]. Here we use this technique to explore the genetic mechanisms of divergence between social insect castes. We contrast the gene-expression patterns of worker-destined and queen-destined larvae and then compare both castes with younger, bipotential, individuals. These comparisons allow tests of hypotheses concerning the evolutionary relationship between queen- or worker-destined developmental pathways in social insects (see, for example, [<xref ref-type="bibr" rid="B2">2</xref>]). More generally, changes in gene expression during insect caste determination can shed light on diverse developmental events such as differences in hormonal titers, the development or apoptosis of reproductive tissues, and developmental rates and metabolism [<xref ref-type="bibr" rid="B9">9</xref>]. We present evidence for widespread divergence in gene expression between workers and queens, and identify particular genes that appear to be integral to the production of these distinct adult phenotypes.</p></sec><sec><title>Results</title><p>We assayed gene expression using an array of gene fragments derived from reciprocal queen and worker larval sub-tractive cDNA libraries. Of 158 unique, consistently amplified clones isolated from these libraries, 63 showed a significant match to genes with known function. The functions assigned to these genes span approximately half of the major functional groups described by Adams <italic>et al.</italic> [<xref ref-type="bibr" rid="B10">10</xref>] for <italic>Drosophila melanogaster</italic>, suggesting that we have captured a fairly broad cross-section of the expressed genes in larval bees. Nevertheless, the clones derived from our libraries were biased toward particular functional groups, in a fashion that suggests a relationship between gene function and caste (Table <xref ref-type="table" rid="T1">1</xref>). For example, we found an apparent over-representation of both ribosomal proteins and hexameric storage proteins in our libraries, relative to the expected frequencies of these gene families based on the <italic>Drosophila</italic> sequence. Of the 63 characterized clones, 15 (24%) showed significant similarity to known ribosomal protein genes. In contrast, 128 of the 6,537 annotated genes (2%) in the <italic>Drosophila</italic> melanogaster genome encode ribosomal proteins [<xref ref-type="bibr" rid="B10">10</xref>]. Gene knock-out experiments have implicated several of these as regulators of development rate and body size in <italic>Drosophila</italic> [<xref ref-type="bibr" rid="B11">11</xref>]. Assuming that ribosomal proteins help modulate ribosome activity, they may be indicators of the developmental tempo of larvae, and therefore could help pinpoint the initiation of caste determination or other major developmental events. We also cloned six apparently distinct hexameric storage proteins from the libraries, indicating a strong bias toward this family given the frequencies of hexamerins in the <italic>Drosophila</italic> genome of &#x0003c;&#x0003c;1% [<xref ref-type="bibr" rid="B10">10</xref>].</p><p>We raised female honeybee larvae using standard methods, then measured gene-expression patterns for larvae collected at six 24-hour intervals from the embryo stage to the middle of the fifth and final instar. A subset (<italic>n</italic> = 13) of these larvae was raised as queens, while the others (<italic>n</italic> = 14) were raised workers. Analyses of covariance in gene expression indicate three primary clusters of genes with shared expression patterns across larvae (Figure <xref ref-type="fig" rid="F1">1</xref>). These clusters define genes upregulated primarily in queen-destined larvae relative to the other larvae screened; genes downregulated in queens; and genes upregulated in young, bipotential larvae. Genes within these clusters are more similar to each other in function than expected by chance (likelihood-ratio analysis based on functional group, df = 14, <italic>X</italic> = 38.2, <italic>p</italic> &#x0003c; 0.001).</p><p>Genes related to RNA processing and translation showed higher rates of expression in young larvae (first and second instars) than in older larvae (Table <xref ref-type="table" rid="T1">1</xref>). RNA helicase was strongly expressed by young larvae, as was an RNA-binding protein and translation elongation factor 2. Interestingly, two heat-shock proteins also were overexpressed in the youngest larvae. One shows a significant match to the Hsp70 family, whereas the second matches Hsp90. Heat-shock proteins are named for their ubiquitous presence following hyperthermy and other stresses [<xref ref-type="bibr" rid="B12">12</xref>], but members of this group have a more general role as molecular chaperones, and are involved in the assembly of newly translated proteins [<xref ref-type="bibr" rid="B13">13</xref>]. Hsp90 appears to be involved in the assembly of steroid hormone receptors [<xref ref-type="bibr" rid="B13">13</xref>]. Expression analyses on a more precise time scale, and at the level of specific tissues, should help determine whether the covariance between the expression of Hsps and agents involved in RNA processing reflects a joint role in producing essential proteins during early larval development. Queen-destined and worker-destined larvae showed broadly similar expression profiles late in the second instar, hours after queens and workers receive differential treatment in colonies. This suggests that most gene-expression changes associated with the caste programs occur after this point. Nevertheless, several genes were expressed differently at this early stage, including dihydrodiol dehydrogenase, which was expressed in lower levels by second-instar queens (normalized <italic>x</italic> = -0.067, SE = 0.025) than by either bipotential larvae (<italic>x</italic> = -0.026, SE = 0.009) or second-instar workers (<italic>x</italic> = 0.001, SE = 0.007). Dihydrodiol dehydrogenase has a general role in deactivating steroid hormones in mammals [<xref ref-type="bibr" rid="B14">14</xref>]. Consequently, it is intriguing that higher levels of expression of this enzyme by workers coincide with a period of decreased steroid titers in workers relative to queens. A putative lipid-binding protein also showed higher expression in workers than in queens from the second instar onward.</p><p>By the third instar, queen and worker larvae showed many differentially expressed genes (Figures <xref ref-type="fig" rid="F1">1</xref>,<xref ref-type="fig" rid="F2">2</xref>) and, in each case, retained these different levels of expression through the remaining two instars. A cytochrome P450 (from the CYP4 subfamily) was more strongly expressed by workers than by queens. Members of the CYP4 subfamily have been implicated in the downregulation of steroid levels [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>] although this subfamily is involved in a diversity of additional functions related to the metabolism of endogenous and exogenous substrates [<xref ref-type="bibr" rid="B17">17</xref>]. As a group, hexameric storage proteins were expressed at significantly higher levels in workers than in queens (Table <xref ref-type="table" rid="T1">1</xref>), although a single hexamerin, from clone 1CB6 [<xref ref-type="bibr" rid="B3">3</xref>], showed nearly equal levels of expression in queens and workers. Hexamerins have a nutrient storage role during insect development, by bundling amino acids that accumulate during larval development for use during metamorphosis or by the adult insects. The accelerated growth rates of queens may limit their ability to allocate resources to hexamerin storage during the larval stages we assayed, as shown for <italic>Manduca</italic> sexta larvae [<xref ref-type="bibr" rid="B18">18</xref>]. Other genes consistently overexpressed by workers include an oxidoreductase, a fatty-acid-binding protein, aldehyde dehydrogenase and a second lipid-binding protein.</p><p>Queens overexpressed two genes directly linked to increased metabolic rates - the nuclear-encoded mitochondrial protein ATP synthase and the mitochondrial gene cytochrome oxidase I. Higher expression of two mitochondrial genes was reported previously [<xref ref-type="bibr" rid="B6">6</xref>], and was suggested to be an indicator of higher metabolic respiration in queen versus worker larvae. Interestingly, increased expression of these genes by queen larvae in our study was not apparent until the fifth instar in the bees we followed, a fact consistent with a role in caste-biased respiration. During the final instar, queens show the most extreme differences in growth rates, relative to workers. Queens also showed relatively high expression of a set of structural (cuticular) proteins, a histone acetyltransferase bromodomain, an ortholog to the <italic>Smcx/y</italic> mammalian sex-differentiation gene, and pyruvate dehydrogenase (Figures <xref ref-type="fig" rid="F1">1</xref>,<xref ref-type="fig" rid="F2">2</xref>).</p></sec><sec><title>Discussion</title><p>As well as providing expression information for specific genes, the analyses presented here provide the first look at genome-level processes during the divergence of two social insect castes. Caste- and age-based biases in gene expression were indicated by a strong tendency for independent samples to group together in the clustering analyses (Figure <xref ref-type="fig" rid="F1">1</xref>). The castes differed widely in gene-expression patterns, supporting the idea that social insect castes have faced strong directional selection pressures [<xref ref-type="bibr" rid="B2">2</xref>]. Workers remained more faithful than queens to the expression profiles of younger, bipotential, larvae. In contrast, queens appeared simultaneously to downregulate many of the genes expressed by bipotential larvae and turn on a distinct set of caste-related genes. This difference could reflect, in part, a common impact of regulatory hormones on gene expression in young larvae and in larvae destined to become workers. Juvenile hormone and ecdysteroids are both much reduced in these larvae, relative to queen-destined larvae [<xref ref-type="bibr" rid="B4">4</xref>], and ecdysteroids have been shown to downregulate several genes expressed in the ovaries of developing workers [<xref ref-type="bibr" rid="B7">7</xref>].</p><p>Honeybees and other social insects provide a novel opportunity to measure the impacts of insect hormones on gene expression [<xref ref-type="bibr" rid="B9">9</xref>]. Specifically, hormone levels in different social insect castes are partially decoupled from molting and metamorphosis [<xref ref-type="bibr" rid="B19">19</xref>]. This fact can be used to retrieve expression differences caused by hormones <italic>per se</italic> from differences caused by other developmental events such as metamorphosis [<xref ref-type="bibr" rid="B8">8</xref>]. The identification of common transcriptional regulators for these genes, as shown in <italic>Drosophila</italic> embryos [<xref ref-type="bibr" rid="B20">20</xref>], would point to hormonal control mechanisms and unite caste-related genes with genes involved in insect development more generally.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Array development</title><p>To generate the genetic arrays, cDNA clones were isolated from four subtractive libraries derived from worker- and queen-destined larvae collected in the third and fourth instars. We chose 144 clones from worker-biased libraries and 144 clones from queen-biased libraries. These clones were amplified by the polymerase chain reaction (PCR) using endogenous adaptor primers (primers Nest1 and Nest2r, described in [<xref ref-type="bibr" rid="B3">3</xref>]), then separated and verified by agarose gel electrophoresis. After being denatured for 5 min at 95&#x000b0;C in a final molarity 0.2 M NaOH, approximately 5 &#x003bc;g DNA from each amplified clone was fixed onto nylon membranes (Hybond N+, Amersham) using a 5 &#x003bc;l slot-pin library replicator (V and P Scientific) followed by UV irradiation. Each of 96 quadrants in a 384-cell printing contained three samples and a negative control. An aliquot of each cDNA clone was purified then sequenced using fluorescent-dye labeling and an ABI Prism 373 DNA analysis machine (Applied Biosystems).</p></sec><sec><title>Sample collection</title><p>Embryos, along with worker- and queen-destined larvae, were harvested from several colonies of <italic>A. mellifera ligustica</italic> of the same genetic stock maintained in the Bee Research Lab apiary. Two groups of 50 late-stage (2-day-old) embryos were harvested from standard (worker) cells and immediately frozen at -80&#x000b0;C. Four groups of 20 bipotential larvae aged 24 h post-hatching also were pooled and frozen. Fourteen worker-destined larvae aged 48, 72, 96, and 120 h post-hatching were also collected, along with 13 queen-destined larvae from the same age groups, raised in natural queen-rearing (swarm) cells. The ages of these queen larvae were determined by wet weight and body size comparisons with larvae of known age raised in artificial queen cells. For both worker- and queen-destined larvae, ages were likely to be accurate within 8-12 h, on the basis of wet-weight comparisons with larvae of known age [<xref ref-type="bibr" rid="B3">3</xref>]. Total RNA was extracted from these samples using the RNAqueous protocol (Ambion).</p></sec><sec><title>Labeling and hybridization</title><p>DNA probes were generated from 5 &#x003bc;g total RNA by reverse transcriptase synthesis (Superscript II enzyme, Life Technologies), using oligo-dT primers and biotin-labeled dUTP and dATP (in a 1:10 molar ratio to unmodified dTTP and DATP, respectively). Probes were cleaned by spin filtration, denatured, and then incubated in hybridization buffer (50% formamide, 6 &#x000d7; SSPE, 0.5% SDS, 5 &#x000d7; Denhardt's solution) with 0.1 pmol poly-(A)₂₅ for 2 h at 42&#x000b0;C. Hybridization was carried out at 42&#x000b0;C, followed by a series of stringency washes, as in [<xref ref-type="bibr" rid="B3">3</xref>]. Membranes were washed, bathed in the Avidx alkaline-phosphatase conjugate then exposed to the chemiluminescent substrate CDP-star, according to the manufacturer's instructions (Tropix, Applied Biosystems). Membranes were next exposed to autoradiographic film for between 10 min and 8 h. Each membrane was used only once, giving an independent hybridization for 2 embryo replicates, 4 bipotential larvae, 13 queen, and 13 worker larval samples.</p></sec><sec><title>Data analysis</title><p>Cloned sequences were aligned using the CLUSTALW algorithm (Omiga 2.0, Oxford Molecular) to assess sequence quality and identify multiple captures of the same sequence. Each unique clone (167 out of 288) was compared against the GenBank database using the BLAST-X and BLAST-N algorithms at the site maintained by the US National Center for Biotechnology Information [<xref ref-type="bibr" rid="B21">21</xref>]. Matches showing a probability score of &#x0003c; 1.0 &#x000d7; 10^-3 were treated as significant matches. Of the 167 unique clones, nine were dropped from the analyses as a result of evidence of multiple bands or inconsistent PCR amplification. Of the remaining clones, 64 showed a match to one or more sequences in GenBank and, of these, 63 could be placed into one of the gene function groups defined in [<xref ref-type="bibr" rid="B10">10</xref>] and the Gene Ontology database [<xref ref-type="bibr" rid="B22">22</xref>]. All gene fragment sequences used are available, along with their BLAST search results, at the Beenome Project website [<xref ref-type="bibr" rid="B23">23</xref>], and in the NCBI dbEST database [<xref ref-type="bibr" rid="B24">24</xref>].</p><p>Scanned images of exposed films were scored using the software program Zero Dscan (Scanalytics), giving a densitometry score for each cell in the membrane's 384-cell matrix. Spreadsheet macros (Microsoft Excel) were then used to generate an average intensity value for each of 158 unique clones that consistently showed a single band in the PCR amplifications. These values were normalized as a proportion of the strongest signal on each membrane, then were further normalized and centered on unity across genes and samples [<xref ref-type="bibr" rid="B18">18</xref>]. Centroid clustering was then carried out and visualized using the programs Cluster 1.0 and TreeView 1.0, respectively [<xref ref-type="bibr" rid="B25">25</xref>].</p></sec></sec><sec><title>Additional data</title><p>The following additional data are included as Excel files: <xref ref-type="supplementary-material" rid="S1">raw expression data</xref> and <xref ref-type="supplementary-material" rid="S2">normalized data</xref> used in cluster analyses.</p></sec><sec sec-type="supplementary-material"><title>Supplementary Material</title><supplementary-material content-type="local-data" id="S1"><caption><title>Additional data file 1</title><p>Raw expression data.</p></caption><media xlink:href="gb-2000-2-1-research0001-S1.xls" mimetype="application" mime-subtype="vnd.ms-excel"><caption><p>Click here for data file 1</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="S2"><caption><title>Additional data file 2</title><p>Normalized expression data.</p></caption><media xlink:href="gb-2000-2-1-research0001-S2.xls" mimetype="application" mime-subtype="vnd.ms-excel"><caption><p>Click here for data file 2</p></caption></media></supplementary-material></sec>

Effectiveness of specific RNA-mediated interference through ingested double-stranded RNA in <italic>Caenorhabditis elegans</italic>

<sec><title>Background</title><p>In <italic>Caenorhabditis elegans,</italic> injection of double-stranded RNA (dsRNA) results in the specific inactivation of genes containing homologous sequences, a technique termed RNA-mediated interference (RNAi). It has previously been shown that RNAi can also be achieved by feeding worms <italic>Escherichia coli</italic> expressing dsRNA corresponding to a specific gene; this mode of dsRNA introduction is conventionally considered to be less efficient than direct injection, however, and has therefore seen limited use, even though it is considerably less labor-intensive.</p></sec><sec><title>Results</title><p>Here we present an optimized feeding method that results in phenotypes at least as strong as those produced by direct injection of dsRNA for embryonic lethal genes, and stronger for genes with post-embryonic phenotypes. In addition, the interference effect generated by feeding can be titrated to uncover a series of hypomorphic phenotypes informative about the functions of a given gene. Using this method, we screened 86 random genes on consecutive cosmids and identified functions for 13 new genes. These included two genes producing an uncoordinated phenotype (a previously uncharacterized POU homeodomain gene, <italic>ceh-6,</italic> and a gene encoding a MADS-box protein) and one gene encoding a novel protein that results in a high-incidence-of-males phenotype.</p></sec><sec><title>Conclusions</title><p>RNAi by feeding can provide significant information about the functions of an individual gene beyond that provided by injection. Moreover, it can be used for special applications for which injection or the use of mutants is sometimes impracticable (for example, titration, biochemistry and large-scale screening). Thus, RNAi by feeding should make possible new experimental approaches for the use of genomic sequence information.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Kamath</surname><given-names>Ravi S</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Martinez-Campos</surname><given-names>Maruxa</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Zipperlen</surname><given-names>Peder</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Fraser</surname><given-names>Andrew G</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" corresp="yes" contrib-type="author"><name><surname>Ahringer</surname><given-names>Julie</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Genome Biology

<sec><title>Background</title><p>RNA-mediated interference (RNAi) is the phenomenon first described in the nematode <italic>Caenorhabditis elegans</italic> in which introduction of double-stranded RNA (dsRNA) results in potent and specific inactivation of the corresponding gene through the degradation of endogenous mRNA [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. This technique rapidly produces gene-specific loss-of-function or hypomorphic phenotypes, and potent interference is also observed in the progeny of the affected animal. Thus, because RNAi results in a robust, specific and durable interference effect, and also because RNAi is the simplest and most efficient method for inactivating genes in <italic>C. elegans,</italic> it has been rapidly embraced as a reverse-genetics tool for determining the functions of specific genes.</p><p>Studies involving RNAi have shown that this mode of interference can function across cell boundaries; that is, the site of injection is not critical for successful gene inactivation [<xref ref-type="bibr" rid="B1">1</xref>]. As a result, it is also possible to initiate RNAi either by soaking worms in a solution of dsRNA or by feeding worms with <italic>Escherichia coli</italic> expressing target gene dsRNA, as RNA can be absorbed through the gut and distributed to somatic tissues and the germ line [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. However, these other delivery systems have seen limited use in published studies as the observed efficiency of gene inhibition is significantly lower than with microinjection of adult hermaphrodite worms [<xref ref-type="bibr" rid="B5">5</xref>].</p><p>Nevertheless, RNAi by feeding has several distinct advantages over microinjection. First, because feeding is far less labor-intensive than injection, it is extremely convenient for performing RNAi on a large number of worms. In this regard, RNAi by feeding has proved particularly useful in genetic screens to identify <italic>C. elegans</italic> genes involved in the RNAi mechanism [<xref ref-type="bibr" rid="B6">6</xref>]. Second, for the same reason, feeding is useful for performing RNAi on large numbers of genes. And third, feeding is considerably less expensive than injection and results in a durable reagent (a bacterial strain expressing dsRNA corresponding to a gene of interest) which can be reused to reproduce an RNAi phenotype easily and inexpensively.</p><p>For these reasons, we explored feeding as a means of delivering dsRNA for RNAi. We have developed an optimized protocol for feeding which is of similar sensitivity to injection and results in phenotypes at least as strong as those produced by injection. Furthermore, the interference effect produced by feeding can be titrated, resulting in the ability to generate a range of strong and hypomorphic phenotypes analogous to an allelic series of mutants. Thus, this method establishes RNAi by feeding as a viable or even preferable alternative to RNAi by injection in <italic>C. elegans.</italic></p></sec><sec><title>Results</title><sec><title>Establishment of optimal RNAi feeding conditions</title><p>Timmons and Fire [<xref ref-type="bibr" rid="B4">4</xref>] first described a method for RNAi in which bacteria expressing dsRNA are fed to <italic>C. elegans.</italic> A fragment corresponding to the gene of interest is cloned into a feeding vector (L4440) between two T7 promoters in inverted orientation and is transformed into a bacterial strain carrying IPTG-inducible expression of T7 polymerase [<xref ref-type="bibr" rid="B4">4</xref>]. Recently, Timmons and Fire also showed that use of an <italic>E. coli</italic> strain (HT115(DE3)), which lacks double-strand-specific RNase III, improves the ability to produce RNAi phenotypes by feeding (L. Timmons and A. Fire, personal communication; and Figure <xref ref-type="fig" rid="F1">1</xref>).</p><p>To determine feeding conditions that maximize observable phenotypes, we started with the existing L4440 vector and strain HT115(DE3) and varied a number of parameters that could affect the efficiency of RNAi. We chose two initial test genes that were easy to assay: <italic>gpb-1,</italic> for which mutants are embryonic lethal, and <italic>unc-22,</italic> which results in a post-embryonic uncoordinated phenotype (Unc), as determined by deletion mutants and by RNAi [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. We first tested different methods of induction with isopropylthiogalactoside (IPTG) to see if this would affect the RNAi phenotypes observed. Uninduced bacteria produced no phenotypes, but, somewhat unexpectedly, each presumably stronger method of induction resulted in a lower penetrance of phenotypes, culminating in 0% phenotype from overnight induction in culture (Table <xref ref-type="table" rid="T1">1</xref>). The best induction method was to grow bacteria in culture without induction, to seed these bacteria onto plates containing IPTG, and then to incubate overnight at room temperature; with this method, <italic>gpb-1</italic> produced 100% dead embryos and <italic>unc-22</italic> produced 99% Uncs. To further test this new induction method, we fed two more genes, <italic>par-1</italic> and <italic>par-3,</italic> mutations in either gene result in embryonic lethality [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. Using our optimized induction conditions, feeding <italic>par-1</italic> and <italic>par-3</italic> resulted in 100% and 96% dead embryos, respectively, similar to the results obtained with null mutants (Table <xref ref-type="table" rid="T1">1</xref>).</p><p>To determine the optimal concentration of IPTG for producing RNAi phenotypes, we titrated the IPTG concentration from 10 mM to 1 pM plus no IPTG. We found that 1 mM IPTG gave us the highest penetrance of phenotypes (Table <xref ref-type="table" rid="T2">2</xref>). In addition, we tested the effect of seeded bacterial density and growth phase on the ability to generate phenotypes; for both <italic>par-1</italic> and <italic>unc-22,</italic> saturated cultures produced phenotypes as well as log-phase bacteria (see the Materials and methods section). We also compared feeding at 15&#x000b0;C versus 22&#x000b0;C and found that although there is some gene-specific variation in RNAi effectiveness between these two temperatures, there is no generalizable difference (data not shown).</p><p>Finally, we tested the effect of length of feeding time on the penetrance of RNAi phenotypes observed. Many genes were not effectively silenced after 24 hours at 22&#x000b0;C, but were after 48 hours (see the Materials and methods section and Figure <xref ref-type="fig" rid="F2">2</xref>). In general, it appears that allowing worms to ingest dsRNA-expressing bacteria for longer periods of time increases the efficiency of RNAi by feeding. However, feeding worms beginning earlier than the L4 stage did not improve the penetrance of RNAi in the progeny (data not shown). After 48 hours, L4-stage hermaphrodites become older adults, and many stop laying fertilized eggs; thus we have routinely used the longest feeding time possible (36-40 hours at 22&#x000b0;C), which ensures that these worms will continue to lay fertilized eggs during the time window from which progeny are assayed (the subsequent 24 hours). We found that 72 hours at 15&#x000b0;C gives a similar level of RNAi inhibition for most genes (data not shown).</p></sec><sec><title>Titration of RNAi phenotypes by feeding</title><p>Many genes have pleiotropic effects <italic>in vivo</italic> but have one dominant mutant phenotype that masks other informative phenotypes; thus, we decided to see if we could elicit such masked phenotypes by titrating the concentration of IPTG and, by extension, the degree of RNAi. We titrated the IPTG concentration from 1 pM to 1 mM in three log increments; for comparison, we also tested no IPTG (uninduced) and 10 mM IPTG. We analyzed four genes that could possibly be titrated to other distinct phenotypes: <italic>unc-37,</italic> which has a known allelic series with strong alleles resulting in embryonic lethality and weak alleles resulting in an uncoordinated (Unc) phenotype [<xref ref-type="bibr" rid="B10">10</xref>]; <italic>hlh-2,</italic> which is embryonic lethal by RNAi injection and is expressed in some neural precursors which eventually form the ventral nerve cord ([<xref ref-type="bibr" rid="B11">11</xref>] and M. Krause, personal communication); <italic>mei-1,</italic> which is required for meiotic spindle formation and is embryonic lethal by RNAi injection, but for which a weak mutant allele produces a high-incidence-of-males (Him) phenotype [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]; and <italic>rba-2,</italic> a very strong embryonic lethal gene by RNAi injection which is also involved in repression of vulval cell fates [<xref ref-type="bibr" rid="B14">14</xref>].</p><p>All four genes were 100% embryonic lethal at 1 mM IPTG (Table <xref ref-type="table" rid="T2">2</xref>). For the first three genes, decreasing the IPTG concentration to 1 &#x003bc;M reduced the embryonic lethality sufficiently to expose high levels of a secondary phenotype. For <italic>unc-37</italic> and <italic>hlh-2,</italic> 100% of the surviving worms were Unc. For <italic>mei-1,</italic> a Him phenotype was seen: 8% of the surviving worms were male, which is significantly greater than the 0.5% which normally arise by non-disjunction in a wild type hermaphrodite culture. As we further decreased the IPTG concentration to 1 pM, these levels of embryonic lethality tapered off, as did the penetrance of secondary phenotypes. With no IPTG, all the worms fed these genes had wild-type progeny. For <italic>rba-2,</italic> however, embryonic lethality was observed at all concentrations of IPTG, including 56% lethality without any IPTG. Therefore, these bacteria are likely to express a low level of T7 polymerase in the absence of induction, which is sufficient to produce an RNAi phenotype for some genes. From these data, we conclude that it is possible to titrate RNAi phenotypes by feeding bacteria induced with different concentrations of IPTG, which results in a series of hypomorphic alleles analogous to an allelic series of mutants.</p></sec><sec><title>RNAi by feeding multiple genes</title><p>We also tested RNAi by feeding for two genes simultaneously. When we fed one embryonic lethal gene and one non-lethal gene, we found a reduced penetrance of embryonic lethality compared to feeding the lethal gene alone: feeding <italic>gpb-1</italic> and <italic>unc-22</italic> together reduced the embryonic lethality to 50% (versus 100% for <italic>gpb-1</italic> alone) and feeding <italic>par-3</italic> and <italic>unc-22</italic> together reduced the lethality to 85% (versus 96% for <italic>par-3</italic> alone). Furthermore, in both cases the resulting progeny also failed to display the <italic>unc-22</italic> phenotype (data not shown). Thus, it appears that feeding two genes greatly reduces the strength of phenotype produced by either.</p><p>In a separate experiment, we diluted the <italic>unc-37</italic> HT115 strain 1:1 with <italic>E. coli</italic> that does not produce dsRNA (OP50, a strain commonly used for growing <italic>C. elegans</italic>). In this case, the phenotype changed dramatically from 100% to 0% embryonic lethality, with 100% of progeny displaying post-embryonic phenotypes (uncoordinated, rolling, and body morphology defects), which is similar to the phenotypes obtained by inducing the bacteria with 1 fM IPTG (data not shown). Thus, diluting feeding bacteria with bacteria not expressing the dsRNA can also be used to generate weak hypomorphic phenotypes.</p></sec><sec><title>Target RNA expression in RNAi-treated hermaphrodites</title><p>To determine which tissues are susceptible to RNAi by feeding, we fed bacteria expressing green fluorescent protein (GFP) dsRNA to hermaphrodite worms with a transgenic <italic>GFP</italic> reporter gene expressed in all somatic tissues (<italic>egl-27::gfp;</italic> Figure <xref ref-type="fig" rid="F2">2a</xref>) [<xref ref-type="bibr" rid="B15">15</xref>]. After being fed for 24 hours at 15&#x000b0;C, GFP expression was markedly reduced compared to similarly treated unfed worms (<italic>n</italic> = 28; compare Figure <xref ref-type="fig" rid="F2">2b</xref> and <xref ref-type="fig" rid="F2">a</xref>). After 48 hours, with the exception of the nervous system, GFP was not detectable in somatic tissues; furthermore, neural GFP was dramatically reduced in 91% of worms (<italic>n</italic> = 22; Figure <xref ref-type="fig" rid="F2">2c</xref>). A similar level of inhibition of GFP expression was observed after 72 and 96 hours of feeding (Figure <xref ref-type="fig" rid="F2">2d</xref>,<xref ref-type="fig" rid="F2">e</xref>). Although GFP fluorescence in fed worms was abolished or severely reduced, GFP was sometimes expressed at high levels in late-stage embryos derived from these worms (data not shown). This suggests that some zygotic embryonic transcripts may be difficult to silence, possibly because a continuous supply of dsRNA cannot be provided through the eggshell. In summary, RNAi by feeding efficiently silences genes in most <italic>C. elegans</italic> somatic tissues; however, the nervous system has a delayed and somewhat less robust response to RNAi compared to other tissues.</p></sec><sec><title>Comparison of RNAi by feeding and injection for maternal-effect lethal genes</title><p>To test the strength of RNAi by feeding using the above optimized protocol, we compared it to RNAi by injection. We first tested a set of 14 known maternal-effect embryonic lethal genes (<italic>gpb-1</italic>, <italic>par-1</italic>, <italic>par-2</italic>, <italic>par-3</italic>, <italic>par-6</italic>, <italic>cyk-1</italic>, <italic>skn-1</italic>, <italic>dnc-1</italic>, <italic>bir-1</italic>, <italic>pal-1</italic>, <italic>dif-1</italic>, <italic>plk-1</italic>, <italic>dhc-1,</italic> and <italic>mex-3</italic>) to compare the lethality obtained with both methods (see [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>] for a review of maternal-effect genes). All genes tested were 100% embryonic lethal by both feeding and injection except for <italic>par-3,</italic> which resulted in 97% dead embryos by feeding but 100% by injection, and <italic>cyk-1,</italic> which resulted in 55% dead embryos by feeding but 100% by injection (Table <xref ref-type="table" rid="T3">3</xref>). For some genes, for example <italic>par-1,</italic> RNAi by feeding (<italic>n</italic> = 32 fed worms) and by injection (<italic>n</italic> = 9) both resulted in 100% embryonic lethality of the progeny in all cases. Whereas RNAi of <italic>par-3</italic> resulted in 100% embryonic lethality for all injected worms (<italic>n</italic> = 9), worms subjected to RNAi of <italic>par-3</italic> by feeding resulted in 100% dead embryos in 16/24 cases but lower levels of lethality (on average 88%) in 8/24 cases (data not shown). The overall comparison shows that RNAi by feeding is of similar strength to RNAi by injection for maternal-effect embryonic lethal genes. For some genes, however, RNAi by feeding appears to be somewhat more variable than RNAi by injection.</p><p>To further confirm that RNAi by feeding is comparable in effectiveness to injection for analyzing maternal-effect genes, we made four-dimensional time-lapse video recordings of developing embryos whose mothers were fed with dsRNA corresponding to the 10 above genes with a mutant phenotype detectable by the third cell division (<italic>gpb-1</italic>, <italic>par-1</italic>, <italic>par-2</italic>, <italic>par-3</italic>, <italic>par-6</italic>, <italic>cyk-1</italic>, <italic>dnc-1</italic>, <italic>bir-1</italic>, <italic>plk-1</italic>, and <italic>dhc-1,</italic>). In each case, the known null phenotypes were obtained (Figure <xref ref-type="fig" rid="F3">3</xref>, and data not shown; for <italic>cyk-1,</italic> one of two embryos recorded had the known phenotype and the other survived). For example, <italic>par-2(RNAi)</italic> by either method yielded spindle orientation defects, and <italic>bir-1(RNAi)</italic> by either method yielded embryos with cytokinesis defects (Figure <xref ref-type="fig" rid="F3">3</xref>) [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>].</p></sec><sec><title>Quantitative analysis of RNAi by feeding versus injection</title><p>To compare the sensitivity of feeding versus injection in detecting phenotypes from randomly selected genes, we tested 86 genes from the middle of <italic>C. elegans</italic> chromosome I on consecutive cosmids from K04G2 to R05D11. Of these genes, 12 yielded a phenotype by injection and 13 by feeding (Table <xref ref-type="table" rid="T4">4</xref>). These data suggest that feeding and injection are similarly sensitive for detecting genes that give an RNAi phenotype. The number of phenotypes detected in this experiment is, however, too small to allow a fair comparison of the strengths of the two methods.</p><p>To obtain a quantitative measure of the efficiency of feeding versus injection for RNAi, we compared the two methods on a larger data set. Fraser <italic>et al.</italic> have constructed an RNAi feeding library for <italic>C. elegans</italic> chromosome I [<xref ref-type="bibr" rid="B21">21</xref>]. After performing RNAi by feeding on the first 1,200 predicted genes on chromosome I, phenotypes were identified for 168 predicted genes. We then performed RNAi by injection on these genes and compared the phenotypes to those obtained by feeding (Figure <xref ref-type="fig" rid="F4">4</xref>; see the Materials and methods section for scoring criteria) We reasoned that using a set of genes initially identified by feeding for this comparison would be valid, as in our previous comparison of feeding and injection, feeding successfully detected all those genes with RNAi phenotypes detected by injection.</p><p>Of the 168 genes, 123 were determined to be embryonic lethal by either method, of which feeding detected 97% and injection 91%. The embryonic lethality was of equal penetrance using either method for 77% of these genes, was stronger by feeding for 15% of the genes, and was stronger by injection for 8% of the genes. Of 52 genes giving a sterile phenotype by either method, 96% were identified by feeding and 48% by injection; 44% were detected by both methods, 52% only by feeding, and 4% only by injection. And, finally, of 154 genes giving a post-embryonic phenotype by either method, 96% were identified by feeding and 64% by injection; 60% were detected by both methods, 36% only by feeding, and 4% only by injection. In addition to the above, we also injected 30 random genes that gave no phenotype by feeding, and none gave a detectable phenotype by injection (data not shown). Thus, from this expanded data set, we conclude that RNAi by feeding is roughly equivalent to injection for detecting embryonic lethality, and is somewhat better for detecting genes causing sterility or other post-embryonic phenotypes.</p></sec></sec><sec><title>Discussion</title><p>We present a method by which RNAi by feeding is as strong and as sensitive as RNAi by injection for detecting embryonic phenotypes, and, furthermore, is more sensitive for detecting sterility and post-embryonic phenotypes. The fact that feeding is superior to injection for detecting sterile phenotypes is most likely due to the fact that with feeding the RNA interference effect is begun at L4 stage, giving it more time to affect the germline or gonad, whereas young adults are usually injected; nevertheless, this does not diminish the utility of feeding for detecting sterility as L4-stage hermaphrodites are considerably more difficult to inject. Similarly, the fact that feeding is better than injection for producing post-embryonic phenotypes could be due to the fact that both fed mothers and their progeny are constantly exposed to dsRNA, whereas by injection only the mothers are subjected to a single dose. Again, this confers an inherent advantage on feeding in producing post-embryonic phenotypes that can be exploited by those studying genes important later in development. It should be noted, however, that there are gene-specific differences between RNAi by feeding and injection. We found that some genes are more sensitive to RNAi by injection and others to RNAi by feeding (Figure <xref ref-type="fig" rid="F4">4</xref>). In addition, for some genes, RNAi by feeding is more variable than RNAi by injection. Thus feeding is a useful tool to complement, rather than replace, RNAi by injection.</p><p>We also showed that by titrating the IPTG concentration, RNAi by feeding can generate a range of strong and hypomorphic phenotypes. Because many embryonic lethal genes, for example, have informative post-embryonic RNAi phenotypes, it is extremely useful to be able to elicit such phenotypes for specific genes. Hypomorphic phenotypes were also seen by analyzing the progeny of fed or injected worms immediately following exposure to dsRNA (during the time when RNAi was taking effect); however, these phenotypes were not as reproducible or as robust as those generated by titrating the IPTG concentration. Furthermore, titration allows a low level of RNAi to be consistently applied to a large number of worms, increasing the possibility of detecting low-penetrance phenotypes.</p><p>We also tested other variables that might have affected the strength and utility of feeding. Interestingly, the current feeding vector does not have any transcriptional terminators, and thus we would expect transcripts to vary in size and also to contain sequences from the vector backbone. As a result, we tested a modified vector containing T7 terminators just outside the T7 promoters and discovered that inclusion of such terminators greatly diminished the effectiveness of RNAi (data not shown). Because it is also useful to concomitantly inhibit the functions of two genes, we tested the ability to feed two dsRNAs and learned that this greatly reduces the ability of each gene to independently produce a phenotype. Other methods could be tested for this purpose, however, including co-transforming two feeding vectors into the same bacteria or inserting two gene fragments into the same vector, either as single fragments or as inverted repeats.</p><p>Our analysis of 86 genes on chromosome I by feeding versus injection identified 13 with an RNAi phenotype. Among these 13 genes were two, <italic>apr-1</italic> and <italic>mei-1,</italic> previously reported to have loss-of-function phenotypes. <italic>apr-1</italic> encodes a protein similar to the APC (adenomatous polyposis coli) protein; it is involved in Wnt signaling and has previously been shown to control endoderm induction in the embryo [<xref ref-type="bibr" rid="B22">22</xref>]. Although we failed to obtain embryonic lethality for <italic>apr-1</italic> by feeding, we did identify reproducible and specific phenotypes - uncoordinated movement, body morphology defects, and larval lethality - consistent with the expression pattern and previously demonstrated roles for this gene [<xref ref-type="bibr" rid="B23">23</xref>]. Both feeding and injection produced strong embryonic lethality for <italic>mei-1,</italic> a regulator of meiosis [<xref ref-type="bibr" rid="B12">12</xref>]. Furthermore, by titrating the IPTG concentration, we were able to phenocopy the Him phenotype of a weak <italic>mei-1</italic> mutant, which is indicative of X-chromosomal non-disjunction during meiosis [<xref ref-type="bibr" rid="B13">13</xref>].</p><p>The remaining 11 genes for which we identified an RNAi phenotype have no previously reported function. RNAi of two of these genes caused an Unc phenotype, suggesting roles in the neuromuscular system: <italic>ceh-6</italic> is a POU homeodomain protein, and D1081.2 encodes a MADS-box transcription factor [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Another gene, K02B12.8, produced a Him RNAi phenotype, suggesting a possible function in meiotic chromosome segregation. RNAi of D1081.8, which encodes a novel protein with a Myb-like DNA-binding domain, resulted in 100% embryonic lethality, as did, unsurprisingly, F52B5.6, which is thought to encode a ribosomal protein [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>]. K02B12.3 resulted in sterility by RNAi; this gene encodes a WD-domain protein which is weakly similar to TUP1, a general transcriptional repressor in <italic>Saccharomyces cerevisiae</italic> [<xref ref-type="bibr" rid="B28">28</xref>]. Finally, T19A6.2A, which is similar to a human breast cancer autoantigen, resulted in slow growth [<xref ref-type="bibr" rid="B29">29</xref>]. The ability of RNAi by feeding to detect a wide range of phenotypes for these genes demonstrates its value for studying gene function in <italic>C. elegans.</italic></p><p>One major advantage of RNAi by feeding over injection is that it is considerably less labor-intensive. In practice, this means that RNAi can be performed on thousands of worms for little more effort than feeding a single worm. Thus, feeding affords the possibility of using RNAi in ways not practically possible by injection, such as doing large-scale biochemistry on worms with a mutant RNAi phenotype. A second advantage of RNAi by feeding is that once a bacterial strain expressing a specific dsRNA is created, it can be reused indefinitely to repeatedly perform RNAi on a given gene. Thus, feeding is extremely useful for large-scale experiments in which either many worms will be subjected to RNAi, many genes will be used for RNAi, or a few genes will be used for RNAi many times. Indeed, this method has been used by Fraser <italic>et al.</italic> [<xref ref-type="bibr" rid="B21">21</xref>] to efficiently screen roughly 90% (2,500 predicted genes) of <italic>C. elegans</italic> chromosome I by RNAi.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Strains and clones</title><p>Standard methods were used for culturing <italic>C. elegans</italic> on NGM (nematode growth medium) [<xref ref-type="bibr" rid="B30">30</xref>]. Fragments designated for RNAi were obtained by polymerase chain reaction (PCR) from genomic DNA and were cloned into the L4440 feeding vector (pPD129.36) [<xref ref-type="bibr" rid="B4">4</xref>]; all fragments were between 500 and 2,700 base pairs (bp) in length. The resulting plasmids were transformed into the HT115(DE3) RNase Ill-deficient <italic>E. coli</italic> strain, which was previously shown by Timmons and Fire to be beneficial for RNAi by feeding (L. Timmons and A. Fire, personal communication). The HT115 genotype is as follows: (<italic>F^-, mcrA, mcrB, IN(rrnD-rrnE)1, lambda^-, rnc14::Tn10(DE3 lysogen:lacUV5 promoter-T7 polymerase</italic>)). The RNase III gene is disrupted by a Tn10 transposon carrying a tetracycline-resistance marker. Inclusion of tetracycline in feeding plates or in bacterial cultures used for feeding in many cases resulted in a weaker RNAi effect (data not shown), perhaps because the cultures grew very poorly, and thus it was not included in our feeding experiments; however, bacteria were selected on tetracycline plates before feeding.</p><p>The following primer pairs were used for PCR amplification: <italic>gpb-1</italic> (5'-ATGAGCGAACTTGACCAAC-3' and 5'-TTAATTCCAGATCTTGAGG-3'), <italic>par-1</italic> (5'-CAAAGCACGTGATAACCGG-3' and 5'-TTGGTGGCTCAATAAATGGC-3'), <italic>par-3</italic> (5'-TTTGGCTTCACTGTGACCG-3' and 5'-TGATGTGCTGTGGATCAGC-3'), <italic>par-2</italic> (5'-GCCGTCGCCCACTGTCG-3' and 5'-CCGGCTCCAGAGTGTCC-3'), <italic>cyk-1</italic> (5'-GAAGAACAGCTGACCAGCG-3' and 5'-GACGATTCAATGCAATGATGG-3'), <italic>skn-1</italic> (5'-CTGCCGAAGAGAATGCTCG-3' and 5'-GTTTGGTACAACTTCTGTTGG-3'), <italic>dnc-1</italic> (5'-TCTCCACTTTCTACTACAGC-3' and 5'-TGTTCTTGGAAGCCAGCG-3'), <italic>bir-1</italic> (5'-ATGGCACCCGGGACCAA-3' and 5'-TTATTTGCCGCGGCGGC-3'), <italic>pal-1</italic> (5'-GGGGTACCCCAATGTCGGTCGATGTCAAGTCG-3' and 5'-CATGCCATGGCATGGTACTTATAGCCGAATCTTCTG-3'), <italic>dif-1</italic> (5'-ACGCATTGAAATGTCGGACG-3' and 5'-TTGCAGGGAAAGCACGGAG-3'), and <italic>plk-1</italic> (5'-GACAAGGATCGTGGGACC-3' and 5'-AGCACAGCAACTTGGTGG-3'). Primer pairs for <italic>mex-3, dhc-1, par-6, unc-37, hlh-2, mei-1</italic> and <italic>rba-2,</italic> the 88 genes on cosmids K04G2 to R05D11, and the 1,200 genes used for large-scale comparison of feeding and injection were obtained as part of the Research Genetics <italic>C. elegans</italic> GenePairs collection. Fragments corresponding to <italic>unc-22</italic> and GFP were generated from L4440-based vectors containing those inserts (pLT 61.1 and pPD128.110, respectively) [<xref ref-type="bibr" rid="B4">4</xref>].</p></sec><sec><title>Bacterial induction method tests</title><p>Single colonies of HT115 bacteria containing cloned L4440 plasmids were picked and grown in culture in LB with 50 &#x003bc;g/ml ampicillin (Amp), except where indicated. The following methods were used to induce expression of dsRNA:</p><sec><title>Non-induced</title><p>Bacteria were grown for 8 h, then seeded directly onto NGM plates with 50 &#x003bc;g/ml Amp and incubated at room temperature overnight.</p></sec><sec><title>Protocol I (optimal)</title><p>Bacteria were grown for 8 h, then seeded directly onto NGM plates with 1 mM IPTG and 50 &#x003bc;g/ml Amp. (Similar results were obtained from bacterial cultures grown for 8-18 h before seeding plates; RNAi results obtained after growth longer than 24 h were sometimes weaker.) Seeded plates were allowed to dry at room temperature and induction was continued at room temperature overnight.</p></sec><sec><title>Protocol 2</title><p>Bacteria were grown to OD₅₉₅ = 0.4, then IPTG was added to 0.4 mM and bacteria were induced shaking at 37&#x000b0;C for 2 h. After induction, additional IPTG and Amp were added to a total concentration of 0.8 mM and 100 &#x003bc;g/ml, respectively, before seeding onto NGM plates with 50 &#x003bc;g/ml Amp.</p></sec><sec><title>Protocol 3</title><p>Bacteria were treated as in Protocol 1, but induction was performed on seeded plates at 37&#x000b0;C overnight.</p></sec><sec><title>Protocol 4</title><p>Bacteria were treated as in Protocol 2, but induction was performed in culture shaking at 37&#x000b0;C overnight.</p></sec></sec><sec><title>RNAi by feeding</title><p>L4-stage hermaphrodite worms were placed onto NGM plates containing seeded bacteria expressing dsRNA for each gene and were incubated for 36-40 h at 22&#x000b0;C or for 72 h at 15&#x000b0;C. Then, three worms were independently replica plated onto plates seeded with the same bacteria and were allowed to lay eggs for 24 h at 22&#x000b0;C before being removed. Progeny were scored for embryonic lethality after a further 24 h at 22&#x000b0;C, and post-embryonic phenotypes were scored blindly by two independent observers at the end of four successive 12-h intervals. Progeny laid on the first plate were also scored for post-embryonic phenotypes. A gene was found to be positive for a given phenotype if it could be observed in at least two of three worms or their progeny in at least two independent feeding experiments. Feeding times shorter than 36-40 h at 22&#x000b0;C or 72 h at 15&#x000b0;C were not always sufficient to produce a strong RNAi effect; for example, <italic>par-2</italic> and <italic>par-3</italic> were 53% and 23% embryonic lethal, respectively, after feeding for 24 h but were both 100% embryonic lethal after feeding for at least 36 h at 22&#x000b0;C (our unpublished results).</p></sec><sec><title>Titration of hypomorphic phenotypes</title><p>Single colonies of HT115 bacteria containing cloned L4440 plasmids with fragments corresponding to <italic>unc-37, hlh-2, mei-1</italic> or <italic>rba-2</italic> were treated as described in the optimal induction method above, except that for each gene, NGM plates were used with the following IPTG concentrations: 0, 1 pM, 1 nM, 1 &#x003bc;M, 1 mM, and 10 mM. Worms were incubated for 72 h at 15&#x000b0;C before being replica plated and scored.</p></sec><sec><title>Tissue susceptibility to RNAi by feeding</title><p>Worms containing an extrachromosomal array expressing <italic>egl-27::gfp,</italic> a ubiquitous somatic GFP reporter, were fed bacteria expressing GFP dsRNA for either 24, 48, 72, or 96 h at 15&#x000b0;C. Those worms, as well as control worms fed bacteria not expressing GFP dsRNA, were photographed under identical conditions. Figure <xref ref-type="fig" rid="F2">2</xref> illustrates the GFP expression of a representative worm from each time point. Many worms from the 72 h and 96 h time points had almost no detectable GFP expression in any tissue.</p></sec><sec><title>RNA synthesis and microinjection</title><p>Injections were performed as in [<xref ref-type="bibr" rid="B1">1</xref>]. Templates for dsRNA synthesis were made by PCR on L4440-based feeding constructs using T7 primer (5'-CGTAATACGACTCACTATAG-3'). Sense and antisense RNAs were synthesized in a single reaction <italic>in vitro</italic> using a T7 polymerase-based kit (Promega). Double-stranding was achieved by incubation at 72&#x000b0;C for 10 min, and the sizes of dsRNA products were verified by electrophoresis. dsRNA was injected at a concentration of 0.5-1.0 mg/ml into one or both gonad arms (we and others have found that injection into one or both gonad arms produces equivalent effects). Injected worms were allowed to recover at 22&#x000b0;C for 24 h post-injection, then were replica plated and allowed to lay eggs for 24 h. Injected worms and their progeny were scored as previously described for RNAi by feeding.</p></sec><sec><title>Four-dimensional recordings of developing embryos</title><p>Mothers fed dsRNA were dissected in egg buffer (118 mM And, 40 mM KCl, 3 mM CaCl₂, 3 mM MgCl₂, 5 mM HEPES pH 7.2) on a coverslip to release young embryos. The cover-slip was inverted onto a 3% agar pad, which was then sealed with petroleum jelly. A series of 12 focal planes was recorded every 30 sec for 1 h using Openlab software (Improvision) controlling either a Zeiss Axioplan 2 or Leica DMBRE microscope.</p></sec><sec><title>Quantitative comparison of RNAi by feeding versus injection</title><p>RNAi was performed by feeding at 15&#x000b0;C using the optimized method described above or by injection. Embryonic lethality was scored by estimating the percentage of dead embryos to the nearest 10% among the offspring of the three worms replica plated for each gene. For embryonic lethality, feeding and injection were considered to be of equal strength if the percentages of dead embryos were within 10%, and one method was considered stronger than the other if the percentage of dead embryos was at least 20% greater than that obtained by the other method. Fed or injected worms were considered sterile if they had fewer than 10 progeny on average, as wild-type worms under similar conditions typically have more than 50 progeny. Post-embryonic phenotypes were scored if more than 10% of the progeny had a given phenotype. For sterility and post-embryonic phenotypes, feeding and injection were considered to be of equal strength if both resulted in that phenotype, and one method was considered stronger if it resulted in that phenotype but the other method did not.</p></sec></sec>

Supervised harvesting of expression trees

<sec><title>Background</title><p>We propose a new method for supervised learning from gene expression data. We call it 'tree harvesting'. This technique starts with a hierarchical clustering of genes, then models the outcome variable as a sum of the average expression profiles of chosen clusters and their products. It can be applied to many different kinds of outcome measures such as censored survival times, or a response falling in two or more classes (for example, cancer classes). The method can discover genes that have strong effects on their own, and genes that interact with other genes.</p></sec><sec><title>Results</title><p>We illustrate the method on data from a lymphoma study, and on a dataset containing samples from eight different cancers. It identified some potentially interesting gene clusters. In simulation studies we found that the procedure may require a large number of experimental samples to successfully discover interactions.</p></sec><sec><title>Conclusions</title><p>Tree harvesting is a potentially useful tool for exploration of gene expression data and identification of interesting clusters of genes worthy of further investigation.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Hastie</surname><given-names>Trevor</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A2" corresp="yes" contrib-type="author"><name><surname>Tibshirani</surname><given-names>Robert</given-names></name><xref ref-type="aff" rid="I2">2</xref><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" corresp="yes" contrib-type="author"><name><surname>Botstein</surname><given-names>David</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A4" corresp="yes" contrib-type="author"><name><surname>Brown</surname><given-names>Patrick</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib>

Genome Biology

<sec><title>Background</title><p>In this paper we introduce 'tree harvesting' - a general method for supervised learning from gene expression data. The scenario is as follows. We have real-valued expression measurements for thousands of genes, measured over a set of samples. The number of samples is typically 50 or 100, but will be larger in the future. An outcome measurement is available for each sample, such as a survival time or cancer class. Our objective is to understand how the genes relate to the outcome.</p><p>The generic problem of predicting an outcome measure from a set of features is called 'supervised learning'. If the outcome is quantitative, the term 'regression' is used; for a categorical outcome, 'classification'. There are many techniques available for supervised learning: for example, linear regression, discriminant analysis, neural networks, support vector machines, and boosting. However, these are not likely to work 'off the shelf', as expression data present special challenges. The difficulty is that the number of inputs (genes) is large compared with the number of samples, and they tend to be highly correlated. Hastie <italic>et al</italic>. [<xref ref-type="bibr" rid="B1">1</xref>] describe one simple approach to this problem. Here we build a more ambitious model that includes gene interactions.</p><p>Our strategy is first to cluster the genes via hierarchical clustering, and then to consider the average expression profiles from all of the clusters in the resulting dendrogram as potential inputs into our prediction model. This has two advantages. First, hierarchical clustering has become a standard descriptive tool for expression data (see, for example, [<xref ref-type="bibr" rid="B2">2</xref>]), so by 'harvesting' its clusters, the components of our prediction model will be convenient for interpretation. Second, by using clusters as inputs, we bias the inputs towards correlated sets of genes. This reduces the rate of overfitting of the model. In fact we go further, and give preference to larger clusters, as detailed below.</p><p>The basic method is described in the next section for a quantitative output and squared error. We then generalize it to cover other settings such as survival data and qualitative responses. Tree harvesting is illustrated in two real examples and a simulation study is described to investigate the performance of the method. Finally, we generalize tree harvesting further, allowing nonlinear expression effects.</p></sec><sec><title>Results</title><sec><title>Tree harvesting</title><p>As our starting point, we have gene expression data <italic>x_ij</italic> for genes <italic>i</italic> = 1,2, ...<italic>p</italic> and samples <italic>j</italic> = 1,2 ...<italic>n</italic>, and a response measure <italic>y</italic> = (<italic>y</italic>₁, <italic>y</italic>₂, ... <italic>y_n</italic>) for each sample (each <italic>y_j</italic> may be vector-valued). The response measure can take many forms: for example, a quantitative measure such as percentage response to a treatment, a censored survival time, or one of <italic>K</italic> cancer classes. The expression data <italic>x_ij</italic> may be from a cDNA microarray, in which case it represents the log red to green ratio of a target sample relative to a reference sample. Or <italic>x_ij</italic> might be the expression level from an oligonucleotide array. </p><p>The basic method has two components: a hierarchical clustering of the gene expression profiles, and a response model. The average expression profile for each cluster provides the potential features (inputs) for the response model.</p><p>We denote a cluster of genes by <italic>X_c</italic>, and the corresponding average expression profile by <inline-graphic xlink:href="gb-2001-2-1-research0003-i1.gif"/><italic>_c</italic> = (<inline-graphic xlink:href="gb-2001-2-1-research0003-i1.gif"/><italic>_c</italic>_,1, <inline-graphic xlink:href="gb-2001-2-1-research0003-i1.gif"/><italic>_c</italic>_,2, ... <inline-graphic xlink:href="gb-2001-2-1-research0003-i1.gif"/><italic>_c</italic>_,n). Starting with <italic>p</italic> genes, a hierarchical clustering agglomerates genes in <italic>p</italic> - 1 subsequent steps, until all genes fall into one big cluster. Hence it generates a total of <italic>p</italic> + (<italic>p</italic> - 1) = 2<italic>p</italic> - 1 clusters, which we denote by <italic>c</italic>₁, <italic>c</italic>₂, ... <italic>c</italic>_{2<italic>p</italic> - 1}.</p><p>The response model approximates the response measurement by some of the average gene expression profiles and their products, with the potential to capture additive and interaction effects. To facilitate construction of the interaction model, we translate each <italic>x_ij</italic> to have minimum value 0 over the samples:</p><graphic xlink:href="gb-2001-2-1-research0003-i20.gif"/><p>The notation <inline-graphic xlink:href="gb-2001-2-1-research0003-i2.gif"/> denotes the average expression profile for a cluster <italic>c</italic>, using these translated values. The translation is done solely to make interactions in the model more interpretable. Note that the untranslated values are used in the clustering.</p><p>For a quantitative response <italic>y_j</italic>, <italic>j</italic> = 1,2,...<italic>n</italic>, the model takes the form:</p><graphic xlink:href="gb-2001-2-1-research0003-i21.gif"/><p>where &#x003b2;<italic>_k</italic> and <italic>&#x003b2;_kk'</italic> are parameters that are estimated by minimizing the sum of squared errors &#x02211;<italic>_j</italic> (<italic>y_j</italic> - <inline-graphic xlink:href="gb-2001-2-1-research0003-i3.gif"/><italic>_j</italic>)². As each <inline-graphic xlink:href="gb-2001-2-1-research0003-i4.gif"/> has minimum value 0, the product terms represent positive or negative synergy between the genes involved.</p><p>Clearly it is not feasible, or even desirable, to include all clusters in the sums in Equation 2. Instead we build up the model in a forward stepwise manner as follows. Initially the only term in the model <inline-graphic xlink:href="gb-2001-2-1-research0003-i5.gif"/> is the constant function 1. The candidate terms <inline-graphic xlink:href="gb-2001-2-1-research0003-i6.gif"/> consists of all of the 2<italic>p</italic> - 1 average expression profiles <inline-graphic xlink:href="gb-2001-2-1-research0003-i2.gif"/>. At each stage we consider all products consisting of a term in <inline-graphic xlink:href="gb-2001-2-1-research0003-i5.gif"/> and a term in <inline-graphic xlink:href="gb-2001-2-1-research0003-i6.gif"/>, and add in the term that most improves the fit of the model in terms of a score statistic <italic>S</italic>. We continue until some maximum number of terms <inline-graphic xlink:href="gb-2001-2-1-research0003-i5.gif"/> have been added to the model.</p><p>For example, at the first stage we enter the best average expression profile <inline-graphic xlink:href="gb-2001-2-1-research0003-i2.gif"/>; this corresponds to the product of <inline-graphic xlink:href="gb-2001-2-1-research0003-i2.gif"/> and the constant function 1. The resulting model has the form <inline-graphic xlink:href="gb-2001-2-1-research0003-i3.gif"/><italic>_j</italic> = <inline-graphic xlink:href="gb-2001-2-1-research0003-i7.gif"/>₀ + <inline-graphic xlink:href="gb-2001-2-1-research0003-i7.gif"/>₁<inline-graphic xlink:href="gb-2001-2-1-research0003-i8.gif"/>,<italic>j</italic>, where <inline-graphic xlink:href="gb-2001-2-1-research0003-i7.gif"/>₀, <inline-graphic xlink:href="gb-2001-2-1-research0003-i7.gif"/>₁ are found by least squares. At the second stage, the possible additions to the model are <inline-graphic xlink:href="gb-2001-2-1-research0003-i7.gif"/>₂<inline-graphic xlink:href="gb-2001-2-1-research0003-i9.gif"/>,<italic>j</italic> or <inline-graphic xlink:href="gb-2001-2-1-research0003-i7.gif"/>₁₂<inline-graphic xlink:href="gb-2001-2-1-research0003-i8.gif"/><italic>,j</italic><inline-graphic xlink:href="gb-2001-2-1-research0003-i9.gif"/>,<italic>j</italic> for some cluster <italic>c</italic>₂.</p><p>In general, this algorithm can produce terms involving the products of three or more average expression profiles. However, the user can put an explicit limit on the order of interaction, <italic>I</italic>, allowed in the model. For simplicity of interpretation, in the examples here we set <italic>I</italic> = 2, meaning that products are limited to pairwise products. This is achieved by only considering single terms (non-products) in <inline-graphic xlink:href="gb-2001-2-1-research0003-i5.gif"/> as candidates in the second step.</p><p>Models with pairwise interactions as in Equation 2 are often used in statistical applications. The interactions are usually included in an ad hoc basis, after the important additive terms have been included. An exception is the MARS (multivariate additive regression spline) procedure of Friedman [<xref ref-type="bibr" rid="B3">3</xref>]. This is a general adaptive learning method, which builds up an interaction model as products of piecewise linear functions of the inputs. The model is built up in the same way as in the tree-harvest procedure. MARS is a very popular methodology and inspired some of the ideas in this paper.</p><p>There are crucial computational details that make this algorithm run fast enough for practical applications. First, before the forward stepwise process is started, we need the average expression profiles for all of the 2<italic>p</italic> - 1 clusters. This is achieved in a natural recursive fashion using the tree structure available after a hierarchical clustering: the average expression profile in a node is the weighted average of the two average profiles of the daughter nodes, where the weights are the sizes of the daughter nodes. Other node specific statistics, such as variances and within-variances can be computed in a similar way.</p><p>Second, in the second step of the algorithm we must search over all 2<italic>p</italic> - 1 clusters to find the term that most improves the fit of the model. This is achieved by orthogonalizing the candidate average expression profiles with respect to the terms already in the model, and then computing a score test for each candidate term. With a quantitative response and least squares, this process gives exactly the contribution of each candidate term to the model. For survival, classification, and other likelihood-based models, it is a widely used approximation.</p></sec><sec><title>Additional features and issues</title><sec><title>Data normalization</title><p>As with most sets of microarray experiments, the data for each experiment come from different chips and hence must first be normalized to account for chip variation. We assume that the values for each experiment <italic>j</italic> have been centered, that is <italic>x_ij</italic> &#x02192;<italic>x_ij</italic> - (1/<italic>p</italic>) &#x02211;_{<italic>i</italic>}<italic>x_ij</italic>.</p></sec><sec><title>Choice of clustering method and criterion</title><p>The tree-harvest procedure just starts with a set of clusters, and these can be provided by any clustering method. We have chosen to base the procedure on hierarchical clustering, because of its popularity and effectiveness for microarray data (see, for example, [<xref ref-type="bibr" rid="B2">2</xref>]). The sets of clusters are conveniently arranged in a hierachical manner, and are nested within one another. Specifically if the clustering tree is cut off at two different levels, producing say four and five clusters, respectively, then the four clusters are nested within the five. Hence one can look at clusterings of all different resolutions at the same time. This feature is convenient for interpretation of the tree-harvest results, and is not a property of most other clustering methods. Despite this, other clustering methods might prove to have advantages for use in the tree-harvest procedure, including K-means clustering, self-organizing maps [<xref ref-type="bibr" rid="B4">4</xref>], and procedures that allow overlapping clusters (for example, gene shaving [<xref ref-type="bibr" rid="B1">1</xref>]). The choice of clustering criterion will also effect the results. Again, we have followed Eisen <italic>et al</italic>. [<xref ref-type="bibr" rid="B2">2</xref>] and used average linkage clustering, applied to the correlation matrix of the genes. The use of correlation makes the clustering invariant to scaling of the individual genes. Expanding the final clusters (see below) alleviates some of the sensitivity of the results to the choice of clustering method and criterion.</p></sec><sec><title>Biasing towards larger clusters</title><p>Typically gene expression datasets have many highly correlated genes. In addition, most clusters considered in the harvest procedure are subsets of other clusters. Hence if an average expression profile <inline-graphic xlink:href="gb-2001-2-1-research0003-i2.gif"/> is found to most improve the fit of the model in step 2 of the procedure, it is likely that the average expression profile of some larger cluster, perhaps containing the chosen cluster, does nearly as well as <inline-graphic xlink:href="gb-2001-2-1-research0003-i2.gif"/>. Now all else being equal we prefer larger clusters, because they are more likely to be biologically meaningful. Large clusters can result from a pathway of genes involved in a biological process, or a heterogeneous experimental sample containing different cell types. In addition, the finding of a large cluster correlated with the outcome is less likely to be spurious than that of a small cluster, because there are many more smaller clusters than larger clusters. For these reasons, we bias the selection procedure towards larger clusters. Specifically, if the score for the cluster c is <italic>S_c</italic>, we chose the largest cluster <italic>c</italic>' whose score <italic>S_c</italic>_' is within a factor (1 - &#x003b1;) of the best, that is satisfying <italic>S_c</italic>_' &#x02265; (1 - &#x003b1;) <italic>S_c</italic>. The parameter &#x003b1; may be chosen by the user: we chose &#x003b1; = 0.10 in our examples. The cluster <italic>c</italic>' often contains some or all of the genes in <italic>c</italic>, but this is not a requirement. Although this biases the selection towards larger clusters, a single gene can still be chosen if its contribution is spectacular and unique.</p></sec><sec><title>Model size selection and cross-validation</title><p>Having built a harvest model with some large number of terms, <italic>M</italic>, we carry out a backward deletion, at each stage discarding the term that causes the smallest increase in sum of squares. We continue until the model contains only the constant term. This gives a sequence of models with numbers of terms 1,2, ... <italic>M</italic>, and we wish to select a model size, and hence one of these models. The model size is chosen by <italic>K</italic>-fold cross-validation. The data is split into <italic>K</italic> parts. For each <italic>k</italic> = 1,2, ... <italic>K</italic> the harvest procedure is trained on all of the data except the <italic>k</italic>th part, and then data in the <italic>k</italic>th part is predicted from the trained model. The results are averaged over <italic>k</italic> = 1,2, ... <italic>K</italic>. This is illustrated in the examples in the next two sections.</p></sec><sec><title>Expanding the clusters</title><p>Hierarchical clustering uses a sequence of discrete partitions of genes. Hence, for a given cluster, there may be genes not in that cluster that are more highly correlated with the cluster's average expression profile than some of the genes in the cluster. To account for this, we simply look for such genes in the final set of clusters and report them as 'extra genes' belonging to each cluster.</p><p>We summarize all of the steps in Algorithm 1 (Box <xref ref-type="fig" rid="F0">1</xref>)</p></sec></sec><sec><title>Tree harvesting for general response variables</title><p>The tree-harvest method can be applied to most commonly occurring types of response data. Given responses <italic>y</italic> =(<italic>y</italic>₁, <italic>y</italic>₂, ... <italic>y_n</italic>), we form a model-based approximation &#x003b7; =(&#x003b7;₁, &#x003b7;₂, ... &#x003b7;<italic>_n</italic>) to minimize a loss function:</p><graphic xlink:href="gb-2001-2-1-research0003-i22.gif"/><p>Each quantity &#x003b7;<italic>_j</italic> is a function of the average gene expression profiles, having the form given in Equation 2:</p><graphic xlink:href="gb-2001-2-1-research0003-i23.gif"/><p>Some common response types and loss functions are listed in Table <xref ref-type="table" rid="T1">1</xref>.</p><p>As outlined in the previous section, the model is built up in a forward stepwise manner. Considering <inline-graphic xlink:href="gb-2001-2-1-research0003-i10.gif"/> to be a function of the parameters &#x003b2; = {&#x003b2;<italic>_k</italic>, <italic>&#x003b2;_k,k'</italic>}, addition of each new term to the model is based on the size of the score statistic:</p><graphic xlink:href="gb-2001-2-1-research0003-i24.gif"/><p>and similarly for <italic>&#x003b2;_k,k'</italic>. The censored survival time and categorical response models are illustrated in the next two sections.</p></sec><sec><title>Survival of lymphoma patients</title><p>Figure <xref ref-type="fig" rid="F1">1</xref> shows the dataset used in this example consisting of 3,624 gene expression measurements on 36 patients with diffuse large cell lymphoma (DLCL). These data are described in Alizadeh <italic>et al</italic>. [<xref ref-type="bibr" rid="B5">5</xref>]. The column labels refer to different patients, and the row labels identify the genes. We have applied hierarchical clustering to the genes and a separate clustering to the samples. Each clustering produces a (non-unique) ordering, one that ensures that the branches of the corresponding dendrogram do not cross. Figure <xref ref-type="fig" rid="F1">1</xref> displays the original data, with rows and columns ordered accordingly. </p><p>For each of the 36 patients, a (possibly censored) survival time is available; these range from 1.3 to 102.4 months, and 19 of the 36 patients died in the study period. An appropriate response model is Cox's proportional hazards model [<xref ref-type="bibr" rid="B6">6</xref>]. This has the form:</p><graphic xlink:href="gb-2001-2-1-research0003-i25.gif"/><p>Here <italic>z_j</italic> = (<italic>z</italic>_{1<italic>j</italic>}, <italic>z</italic>_{2<italic>j</italic>}, ... <italic>z_mj</italic>) are <italic>m</italic> risk factors (features) for sample <italic>j</italic>, and <italic>h</italic>(<italic>t</italic>|<italic>z_j</italic>) denotes the hazard function for an individual with feature values <italic>z</italic>; <italic>h</italic>₀(<italic>t</italic>) is the baseline hazard function for an individual with risk factors <italic>z</italic> = 0. The unknown function <italic>r</italic>(<italic>z_j</italic>) represents the log-relative risk of dying at any time <italic>t</italic> for an individual with <italic>z</italic> = <italic>z_j</italic> versus an individual with <italic>z</italic> = 0. In the tree harvest model, the features (<italic>z</italic>_{1<italic>j</italic>}, <italic>z</italic>_{2<italic>j</italic>}, ... <italic>z_mj</italic>) are average expression profiles and we take <italic>r</italic>(<italic>z_j</italic>) to be of the form:</p><graphic xlink:href="gb-2001-2-1-research0003-i26.gif"/><p>as in Equation 2. The tree-harvest algorithm computes an approximate score test from the partial likelihood, to decide which term is entered at each stage.</p><p>We ran the harvest procedure allowing a maximum of six terms, and it produced the results shown in Table <xref ref-type="table" rid="T2">2</xref>.</p><p>Some explanation is needed. At each stage the 'Node' refers to the cluster whose average expression profile is chosen for addition to the model. 'Parent' is the number of the cluster, already in the model, that is to be multiplied by the Node average expression profile; Parent = 0 refers to the constant function 1. Nodes starting with 's' for Node or Parent indicate single genes. 'Score' is the score value achieved by addition of the term; it is roughly a Gaussian variate, so that values &#x02265; 2 are reasonably large. </p><p>Focusing just on the selection of the first cluster, Figure <xref ref-type="fig" rid="F2">2</xref> shows all of the cluster scores. The green horizontal line is drawn at (1 - &#x003b1;) times the maximum score (&#x003b1; = 0.1), and we chose the largest cluster (blue point) above this line. This cluster is the eight-gene cluster 3005, shown in Figure <xref ref-type="fig" rid="F3">3</xref>.</p><p>Overall the resulting model has the form:</p><graphic xlink:href="gb-2001-2-1-research0003-i27.gif"/><p>A positive coefficient indicates increased risk. The training set and cross-validation curves are shown in Figure <xref ref-type="fig" rid="F4">4</xref>. The minimum of the cross-validation (CV) curve occurs at one term, suggesting that the subsequent terms may not improve prediction. </p><p>The gene clusters are shown in Figure <xref ref-type="fig" rid="F3">3</xref> and listed in the Additional data file, available with the online version of this article. Focusing only on the first cluster (3005), we computed the average expression for each of the 36 patients. Then the patients were divided into two groups: those with average expression below the median (group 1), and those with average expression above the median (group 2). The Kaplan-Meier survival curves for these two groups are shown in Figure <xref ref-type="fig" rid="F5">5</xref> and are significantly different (<italic>p</italic> = 2.4 x 10^-5).</p><p>If each of the 3,624 genes is ranked from lowest (1) to highest (3,624) value of the Cox score statistic, the average rank of the eight genes in the cluster 3005 is 3,574.5. Hence these genes are among the strongest individually for predicting survival, but are not the eight strongest genes. Rather they are a set of genes with very similar expression profiles, highly correlated with survival.</p></sec><sec><title>Human tumor data</title><p>In this example, the response is a categorical variable designating a cancer class. We use a subset of 61 of the tumors described in Ross <italic>et al</italic>. [<xref ref-type="bibr" rid="B7">7</xref>] and Scherf <italic>et al</italic>. [<xref ref-type="bibr" rid="B8">8</xref>], omitting the two prostate tumors and the one unknown class. There are expression values for 6,830 genes for each of the tumors, with the distribution across cancer classes shown in Table <xref ref-type="table" rid="T3">3</xref>.</p><p>Here, the tree-harvest method builds a multiple logistic regression (MLR) model in a stepwise fashion, using similar steps to those used for the Cox model for survival data. The goal here is to model the probability of the tumor class, given the expression values. In general terms, if the class variable is denoted by <italic>y</italic> taking values in {1,2, ..., <italic>J</italic>} and the predictor variables by <italic>x</italic>₁,<italic>x</italic>₂, ..., <italic>x_p</italic> a linear MLR model has the form:</p><graphic xlink:href="gb-2001-2-1-research0003-i28.gif"/><p><inline-graphic xlink:href="gb-2001-2-1-research0003-i29.gif"/></p><p><inline-graphic xlink:href="gb-2001-2-1-research0003-i30.gif"/></p><p>As before, the <italic>x</italic>₁ will be cluster averages, possibly individual genes, or pairwise products of these. The logistic transform is a natural scale on which to model the <italic>K</italic> probabilities; the inverse transformation:</p><graphic xlink:href="gb-2001-2-1-research0003-i31.gif"/><p>guarantees that the probabilities sum to 1 and are positive. The model is usually fit by multinomial maximum likelihood. Because the response is really multidimensional, we do not expect a single <italic>x</italic> to be able to distinguish all the cancer classes; this would imply that a single gene average creates an ordering that separates the cancer classes. Typically several are required.</p><p>At each stage, the tree-harvest algorithm considers augmenting the current fitted MLR model with a new term, candidates being any of the node averages, individual genes, or products of these with terms already in the model. As before, a score statistic is used, appropriate for the multinomial model. </p><p>The results of a tree harvest fit allowing seven terms are shown in Table <xref ref-type="table" rid="T4">4</xref>. The deviance is a measure of lack-of-fit of the multinomial model, and we see that with seven terms in the model we have a saturated fit (the model produces probability estimates that are essentially 1 for each observation and the relevant class). This is almost certainly an overfit situation, since we are fitting 56 parameters to 61 observations.</p><p>Figure <xref ref-type="fig" rid="F6">6</xref> shows all of the genes in the seven terms found by the model; the column order is chosen arbitrarily to separate the cancer classes (and is randomly chosen within cancer class). We used ten-fold cross-validation to find a good number of terms for the model. Figure <xref ref-type="fig" rid="F7">7</xref> shows the results, in terms of the deviance statistic (-2 &#x000d7; log-likelihood). For these data, the two-term model minimizes the CV deviance curve and corresponds to the top two bands in Figure <xref ref-type="fig" rid="F6">6</xref>. </p><p>Figure <xref ref-type="fig" rid="F8">8</xref> shows a scatterplot of the average expression for each of the first two clusters, with samples identified by cancer class. Some clear separation in the cancer classes is apparent. </p></sec><sec><title>Simulations</title><p>We carried out a simulation experiment to assess how well tree harvesting discovers 'true' structure. To ensure that the gene expression measurements were realistic in magnitude and correlation, we used the matrix of 3624 &#x000d7; 36 lymphoma expression measurements for our study. Artificial survival and censoring times were then generated, to produce a simulated dataset for harvesting.</p><p>Two scenarios were considered, additive and interaction. For the additive scenario, we chose a cluster at random and generated the censored survival time with a relative risk of 2 as a function of its average expression profile. As indicated in Table <xref ref-type="table" rid="T5">5</xref>, the randomly chosen cluster was taken from either single genes, small clusters (&#x0003c; 10 genes) or larger clusters (between 10 and 300 genes). Tree harvesting was allowed to enter just one term.</p><p>For the interaction scenario, we randomly chose one cluster <italic>c</italic>₁ with between two and ten genes, and then chose the second cluster <italic>c</italic>₂ to be the cluster containing between two and ten genes whose average expression profile had the smallest correlation with that for <italic>c</italic>₁. This made the two clusters as independent as possible, giving the harvest procedure the most chance of discovering their interaction. The survival data were then generated with relative risk function 4<inline-graphic xlink:href="gb-2001-2-1-research0003-i11.gif"/> + 4<inline-graphic xlink:href="gb-2001-2-1-research0003-i12.gif"/> + 3[<inline-graphic xlink:href="gb-2001-2-1-research0003-i11.gif"/><inline-graphic xlink:href="gb-2001-2-1-research0003-i12.gif"/> - <italic>r</italic>] where <italic>r</italic> is the projection of <inline-graphic xlink:href="gb-2001-2-1-research0003-i11.gif"/><inline-graphic xlink:href="gb-2001-2-1-research0003-i12.gif"/> on <inline-graphic xlink:href="gb-2001-2-1-research0003-i11.gif"/> and <inline-graphic xlink:href="gb-2001-2-1-research0003-i12.gif"/>. Tree harvesting was allowed to enter three terms.</p><p>The results are shown in the top panel of Table <xref ref-type="table" rid="T5">5</xref>. The numbers are averages over five simulations. The columns show the average number of genes in the true cluster, average number of genes in the cluster found by tree harvesting, the proportion of the genes found by tree harvesting that are in the true cluster, and vice versa. The final column shows the average absolute correlation of the average expression profile of the true cluster with the estimated cluster. For the interaction scenario, these quantities refer to the pooled set of genes that make up the interaction. If more than one interaction was found, the one having greatest overlap with the true interacting clusters is reported. We see that tree harvesting returns clusters that are a little too large when the true cluster is a single gene, and too small when the true cluster is large. In the additive scenario, it does a fairly good job at discovering the true cluster or one similar to it. However, it correctly discovers interactions only about a quarter of the time. A greater number of samples are needed to accurately find interactions among such a large set of genes. On the other hand, the correlations in the rightmost column are all quite high, indicating that tree harvesting is able to find clusters that are nearly as good as the true ones.</p><p>The middle panel of Table <xref ref-type="table" rid="T5">5</xref> shows the results for the additive scenarios when the relative risk is lowered to 1.0. As expected, they are somewhat worse, although the average correlations are still around 0.60.</p><p>To investigate whether a greater number of samples would improve the detection of interactions, we applied the same methodology to a set of 129 samples and 1,622 genes, from an unpublished study of breast cancer (T. Sorlie, C. Perou, and collaborators, personal communication). As before, we used the expression values and simulated sets of synthetic survival times. The results are shown in the bottom panel of Table <xref ref-type="table" rid="T5">5</xref>. Now the tree-harvest procedure does a good job of recovering the interactions. The greater number of samples, together with the smaller number of genes, resulted in a significant improvement in performance.</p></sec><sec><title>Nonlinear tree-harvest models</title><p>In the harvest procedure described above, the effect of gene expression is modeled linearly. Thus, in modeling each term we assume that increasing or decreasing gene expression has a consistent effect on the outcome. However, it is biologically plausible for a gene to have a nonlinear effect: for example, increasing expression may correlate with longer survival, but only up to some level. Beyond that level, the same or worse survival might result. </p><p>To allow for nonlinear effects, flexible bases of functions could be used for each gene. However, with a large number of genes this would tend to overfit quickly. Hence we allow a simple quadratic function for each gene:</p><graphic xlink:href="gb-2001-2-1-research0003-i32.gif"/><p>We first orthogonalize <italic>b</italic>(<italic>x</italic>) with respect to the linear term for the same gene, and then allow the transformed expression <italic>b</italic>(<italic>x</italic>) in place of the expression <italic>x</italic> in the tree-harvest model. In detail, the model has the same form as Equation 2:</p><graphic xlink:href="gb-2001-2-1-research0003-i33.gif"/><p>where <inline-graphic xlink:href="gb-2001-2-1-research0003-i13.gif"/> equals either <inline-graphic xlink:href="gb-2001-2-1-research0003-i14.gif"/> or <inline-graphic xlink:href="gb-2001-2-1-research0003-i16.gif"/> and &#x003b3; is chosen to make <inline-graphic xlink:href="gb-2001-2-1-research0003-i15.gif"/> uncorrelated with <inline-graphic xlink:href="gb-2001-2-1-research0003-i14.gif"/> over the dataset.</p><p>If a quadratic term is multiplied by a positive coefficient, then the effect of a gene has a 'U' shape, decreasing and then increasing. For a negative coefficient, the effect is an inverted 'U'. A product interaction between two quadratic terms would indicate a strong synergistic effect between the two genes, with direction of expression (below or above average) ignored. When the nonlinear option is used in harvesting, the procedure tries both linear and nonlinear terms at each stage, and chooses the one with maximum score.</p></sec><sec><title>Lymphoma data continued</title><p>We tried tree harvesting with the nonlinear option for the lymphoma dataset, and it gave the first four terms shown in Table <xref ref-type="table" rid="T6">6</xref>. Quadratic terms were entered in terms 2-4; these gave a better fit up to term 3 than the linear model fit earlier, but didn&#x02019;t do as well after that. The clusters from this model are shown in Figure <xref ref-type="fig" rid="F9">9</xref>.</p><p>In the second cluster, for example (marked '2' in Figure <xref ref-type="fig" rid="F9">9</xref>), we see that survival time is greatest for moderate expression levels, and is worse for very low or very high levels.</p><p>Overall, the lack of significant improvement of the nonlinear model over the linear model gives greater confidence that the linear shape for each term is appropriate in this example. However, quadratic models may well be useful for other gene expression experiments.</p></sec></sec><sec><title>Conclusions</title><p>The tree harvest procedure is a promising, general method for supervised learning from gene expression data. It aims to find additive and interaction structure among clusters of genes, in their relation to an outcome measure. This procedure, and probably any procedure with similar aims, requires a large number of samples to uncover successfully such structure. In the real data examples, the method was somewhat hampered by the paucity of available samples. We plan to try tree harvesting on larger gene expression datasets, as they become available. We used a forward stepwise strategy involving sum and products of the average gene expression of chosen clusters. We chose this strategy because it produces interpretable, biologically plausible models. Other models could be built from the average gene expression of clusters, including tree-based models or boosting methods (see, for example, Friedman <italic>et al</italic>. [<xref ref-type="bibr" rid="B9">9</xref>]).</p></sec><sec><title>Additional data</title><p>Additional data available with the online version of this article include <xref ref-type="supplementary-material" rid="S1">clusters</xref> from the harvest model applied to lymphoma data.</p></sec><sec sec-type="supplementary-material"><title>Supplementary Material</title><supplementary-material content-type="local-data" id="S1"><caption><title>Additional data file 1</title><p>Clusters from the harvest model applied to lymphoma data.</p></caption><media xlink:href="gb-2001-2-1-research0003-S1.txt" mimetype="text" mime-subtype="plain"><caption><p>Click here for data file</p></caption></media></supplementary-material></sec>

The effects of 1&#x003b1;,25-dihydroxyvitamin D₃ on matrix metalloproteinase and prostaglandin E₂ production by cells of the rheumatoid lesion

<sec><title>Introduction:</title><p>1&#x003b1;,25-dihydroxyvitamin D₃ [1&#x003b1;,25(OH)₂D₃], the biologically active metabolite of vitamin D₃, acts through an intracellular vitamin D receptor (VDR) and has several immunostimulatory effects. Animal studies have shown that production of some matrix metalloproteinases (MMPs) may be upregulated in rat chondrocytes by administration of 1&#x003b1;,25(OH)₂D₃; and cell cultures have suggested that 1&#x003b1;,25(OH)₂D₃ may affect chondrocytic function. Discoordinate regulation by vitamin D of MMP-1 and MMP-9 in human mononuclear phagocytes has also been reported. These data suggest that vitamin D may regulate MMP expression in tissues where VDRs are expressed. Production of 1&#x003b1;,25(OH)₂D₃ within synovial fluids of arthritic joints has been shown and VDRs have been found in rheumatoid synovial tissues and at sites of cartilage erosion. The physiological function of 1&#x003b1;,25(OH)₂D₃ at these sites remains obscure. MMPs play a major role in cartilage breakdown in the rheumatoid joint and are produced locally by several cell types under strict control by regulatory factors. As 1&#x003b1;,25(OH)₂D₃ modulates the production of specific MMPs and is produced within the rheumatoid joint, the present study investigates its effects on MMP and prostaglandin E₂ (PGE₂) production in two cell types known to express chondrolytic enzymes.</p></sec><sec><title>Aims:</title><p>To investigate VDR expression in rheumatoid tissues and to examine the effects of 1&#x003b1;,25-dihydroxyvitamin D₃ on cultured rheumatoid synovial fibroblasts (RSFs) and human articular chondrocytes (HACs) with respect to MMP and PGE₂ production.</p></sec><sec><title>Methods:</title><p>Rheumatoid synovial tissues were obtained from arthroplasty procedures on patients with late-stage rheumatoid arthritis; normal articular cartilage was obtained from lower limb amputations. Samples were embedded in paraffin, and examined for presence of VDRs by immunolocalisation using a biotinylated antibody and alkaline-phosphatase-conjugated avidin-biotin complex system. Cultured synovial fibroblasts and chondrocytes were treated with either 1&#x003b1;,25(OH)₂D₃, or interleukin (IL)-1&#x003b2; or both. Conditioned medium was assayed for MMP and PGE₂ by enzyme-linked immunosorbent assay (ELISA), and the results were normalised relative to control values.</p></sec><sec><title>Results:</title><p>The rheumatoid synovial tissue specimens (<italic>n</italic> = 18) immunostained for VDRs showed positive staining but at variable distributions and in no observable pattern. VDR-positive cells were also observed in association with some cartilage-pannus junctions (the rheumatoid lesion). MMP production by RSFs in monolayer culture was not affected by treatment with 1&#x003b1;,25(OH)₂D₃ alone, but when added simultaneously with IL-1&#x003b2; the stimulation by IL-1&#x003b2; was reduced from expected levels by up to 50%. In contrast, 1&#x003b1;,25(OH)₂D₃ had a slight stimulatory effect on basal production of MMPs 1 and 3 by monolayer cultures of HACs, but stimulation of MMP-1 by IL-1&#x003b2; was not affected by the simultaneous addition of 1&#x003b1;,25(OH)₂D₃ whilst MMP-3 production was enhanced (Table <xref ref-type="table" rid="T1">1</xref>). The production of PGE₂ by RSFs was unaffected by 1&#x003b1;,25(OH)₂D₃ addition, but when added concomitantly with IL-1&#x003b2; the expected IL-1 &#x003b2;-stimulated increase was reduced to almost basal levels. In contrast, IL-1&#x003b2; stimulation of PGE₂ in HACs was not affected by the simultaneous addition of 1&#x003b1;,25(OH)₂D₃ (Table <xref ref-type="table" rid="T2">2</xref>). Pretreatment of RSFs with 1&#x003b1;,25(OH)₂D₃ for 1 h made no significant difference to IL-1&#x003b2;-induced stimulation of PGE₂, but incubation for 16 h suppressed the expected increase in PGE₂ to control values. This effect was also noted when 1&#x003b1;,25(OH)₂D₃ was removed after the 16h and the IL-1 added alone. Thus it appears that 1&#x003b1;,25(OH)₂D₃ does not interfere with the IL-1&#x003b2; receptor, but reduces the capacity of RSFs to elaborate PGE₂ after IL-1&#x003b2; induction.</p></sec><sec><title>Discussion:</title><p>Cells within the rheumatoid lesion which expressed VDR were fibroblasts, macrophages, lymphocytes and endothelial cells. These cells are thought to be involved in the degradative processes associated with rheumatoid arthritis (RA), thus providing evidence of a functional role of 1&#x003b1;,25(OH)₂D₃ in RA. MMPs may play important roles in the chondrolytic processes of the rheumatoid lesion and are known to be produced by both fibroblasts and chondrocytes. The 1&#x003b1;,25(OH)₂D₃ had little effect on basal MMP production by RSFs, although more pronounced differences were noted when IL-1&#x003b2;-stimulated cells were treated with 1&#x003b1;,25(OH)₂D₃, with the RSF and HAC showing quite disparate responses. These opposite effects may be relevant to the processes of joint destruction, especially cartilage loss, as the ability of 1&#x003b1;,25(OH)₂D₃ to potentiate MMP-1 and MMP-3 expression by 'activated' chondrocytes might facilitate intrinsic cartilage chondrolysis <italic>in vivo</italic>. By contrast, the MMP-suppressive effects observed for 1&#x003b1;,25(OH)₂D₃ treatment of 'activated' synovial fibroblasts might reduce extrinsic chondrolysis and also matrix degradation within the synovial tissue. Prostaglandins have a role in the immune response and inflammatory processes associated with RA. The 1&#x003b1;,25(OH)₂D₃ had little effect on basal PGE₂ production by RSF, but the enhanced PGE₂ production observed following IL-1&#x003b2; stimulation of these cells was markedly suppressed by the concomitant addition of 1&#x003b1;,25(OH)₂D₃. As with MMP production, there are disparate effects of 1&#x003b1;,25(OH)₂D₃ on IL-1&#x003b2; stimulated PGE₂ production by the two cell types; 1&#x003b1;,25(OH)₂D₃ added concomitantly with IL-1&#x003b2; had no effect on PGE₂ production by HACs. In summary, the presence of VDRs in the rheumatoid lesion demonstrates that 1&#x003b1;,25(OH)₂D₃ may have a functional role in the joint disease process. 1&#x003b1;,25(OH)₂D₃ does not appear to directly affect MMP or PGE₂ production but does modulate cytokine-induced production.</p><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Comparative effects of 1 &#x003b1;,25-dihydroxyvitamin D₃ (1 &#x003b1;,25D3) on interleukin (IL)-1-stimulated matrix metalloproteinase (MMP)-1 and MMP-3 production by rheumatoid synovial fibroblasts and human articular chondrocytes <italic>in vivo</italic></p></caption><table frame="hsides" rules="groups"><tbody><tr><td></td><td align="center" colspan="2">Fibroblasts</td><td align="center" colspan="2">Chondrocytes</td></tr><tr><td></td><td colspan="2"><hr></hr></td><td colspan="2"><hr></hr></td></tr><tr><td></td><td align="center">MMP-1</td><td align="center">MMP-3</td><td align="center">MMP-1</td><td align="center">MMP-3</td></tr><tr><td colspan="5"><hr></hr></td></tr><tr><td align="left">Control</td><td align="center">1</td><td align="center">1</td><td align="center">1</td><td align="center">1</td></tr><tr><td align="left">+ 1&#x003b1;,25D3</td><td align="center">&#x000a0;&#x000a0;1.03 &#x000b1; 0.27</td><td align="center">&#x000a0;&#x000a0;2.07 &#x000b1; 0.35</td><td align="center">1.38 &#x000b1; 0.19</td><td align="center">&#x000a0;&#x000a0;1.59 &#x000b1; 0.22</td></tr><tr><td align="left">+ IL-1</td><td align="center">31.09 &#x000b1; 4.97</td><td align="center">31.28 &#x000b1; 8.49</td><td align="center">3.45 &#x000b1; 0.49</td><td align="center">&#x000a0;&#x000a0;9.05 &#x000b1; 0.62</td></tr><tr><td align="left">+ IL-1 + 1&#x003b1;,25D3</td><td align="center">15.55 &#x000b1; 5.86</td><td align="center">11.84 &#x000b1; 2.82</td><td align="center">3.71 &#x000b1; 0.53</td><td align="center">11.11 &#x000b1; 0.31</td></tr></tbody></table><table-wrap-foot><p>Data given are normalized relative to control values and are expressed &#x000b1; SEM for three cultures of each cell type.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Comparative effects of 1&#x003b1;,25-dihydroxyvitamin D₃ (1&#x003b1;,25D3) on Interleukin (IL)-1-stimulated prostaglandin E₂ production by rheumatoid synovial fibroblasts and human articular chondrocyte <italic>in vivo</italic></p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center">Fibroblasts</td><td align="center">Chondrocytes</td></tr></thead><tbody><tr><td align="left">Control</td><td align="center">1</td><td align="center">1</td></tr><tr><td align="left">+ 1&#x003b1;,25D3</td><td align="center">&#x000a0;&#x000a0;1.23 &#x000b1; 0.16</td><td align="center">1.35 &#x000b1; 0.25</td></tr><tr><td align="left">+ IL-1</td><td align="center">&#x000a0;&#x000a0;7.07 &#x000b1; 1.09</td><td align="center">&#x000a0;&#x000a0;3.7 &#x000b1; 1.05</td></tr><tr><td align="left">+ IL-1 + 1&#x003b1;,25D3</td><td align="center">1.61 &#x000b1; 0.7</td><td align="center">4.23 &#x000b1; 1.10</td></tr></tbody></table><table-wrap-foot><p>Data given are normalized relative to control values and are expressed &#x000b1; SEM for three cultures of each cell type.</p></table-wrap-foot></table-wrap></sec>

<contrib id="A1" contrib-type="author"><name><surname>Tetlow</surname><given-names>Lynne C</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>lynne.c.tetlow@man.ac.uk</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Woolley</surname><given-names>David E</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>The biologically active metabolite of vitamin D₃, 1&#x003b1;,25-dihydroxyvitaminD₃ [1&#x003b1;,25(OH)₂D₃], acts through an intracellular receptor [vitamin D receptor (VDR)] and has a main role in the regulation of calcium and phosphorus metabolism [<xref ref-type="bibr" rid="B1">1</xref>]. It also has several immunomodulatory actions such as its effect on the differentiation and proliferation of T lymphocytes, and the regulation of immunoglobulin production by B lymphocytes [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>].1&#x003b1;,25(OH)₂D₃ may affect chondrocytic function, such as proteoglycan and collagen synthesis [<xref ref-type="bibr" rid="B5">5</xref>]; and animal studies have shown that the production of some matrix metalloproteinases (MMPs), namely interstitial collagenase (MMP-1), stromelysin (MMP-3) and 72-kDa gelatinase (MMP-2), may be upregulated in rat chondrocytes by administration of the metabolite [<xref ref-type="bibr" rid="B6">6</xref>]. Discoordinate regulation by vitamin D of MMP-1 and MMP-9 in human mononuclear phagocytes has also been reported [<xref ref-type="bibr" rid="B7">7</xref>]. Together these data have suggested that vitamin D can regulate MMP expression in tissues or pathologies where receptors for the hormone are expressed.</p><p>The kidney is recognized as the primary source of 1&#x003b1;,25(OH)₂D₃, producing the metabolite via 1-hydroxylation of 25-hydroxyvitamin D₃-[<xref ref-type="bibr" rid="B1">1</xref>]. However, the local production of 1&#x003b1;,25(OH)₂D₃ within synovial fluids of arthritic joints, especially the macrophage component, has recently been indicated [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]; and receptors for vitamin D have also been demonstrated in rheumatoid synovial tissues and at sites of cartilage erosion [<xref ref-type="bibr" rid="B10">10</xref>]. Such studies have demonstrated a local source of 1&#x003b1;,25(OH)₂D₃ within the rheumatoid joint, but its regulation and physiological functions at this site remain obscure.</p><p>MMPs are reputed to play a major role in cartilage breakdown in the rheumatoid joint and are produced locally by several cell types, but especially by synovial fibroblasts and articular chondrocytes [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. MMP production and release is microenvironmental in nature and is tightly regulated by several factors, including the proinflammatory cytokines tumour necrosis factor-&#x003b1; and interleukin (IL)-1&#x003b2; [<xref ref-type="bibr" rid="B17">17</xref>]. Because 1&#x003b1;,25(OH)₂D₃ has been shown to modulate the production of specific MMPs and is produced within the rheumatoid joint, the present study was designed to investigate the effects of 1&#x003b1;,25(OH)₂D₃ on MMP and prostaglandin E₂ (PGE₂) production by rheumatoid synovial fibroblasts (RSFs) and human articular chondrocytes (HACs), cell types known to express chondrolytic enzymes both <italic>in vitro</italic> and <italic>in vivo</italic>.</p></sec><sec sec-type="methods"><title>Methods</title><sec><title>Tissue samples</title><p>Samples of rheumatoid synovial tissue, cartilage and cartilage-pannus junction were obtained from arthroplasty procedures performed on patients with classic late-stage rheumatoid arthritis. Normal articular cartilage samples were obtained from lower limb amputations. Samples were fixed in Carnoy's fixative at 20&#x000b0;C for 2 h, embedded in paraffin wax and 5 &#x003bc;m sections cut. Tissue sections were dewaxed, rehydrated and examined for the presence of VDR. </p></sec><sec><title>Immunolocalization of vitamin D receptors</title><p>Tissue sections were treated with 2N HCl at 37&#x000b0;C for 30 min, this being the antigen retrieval procedure recommended by the supplier of the primary antibody. Nonimmune rabbit serum at 10% (vol : vol) in TRIS-buffered saline was applied to the sections for 20 min at 20&#x000b0;C before incubation with the primary antibody. Rat monoclonal antibody to chick VDR (Biogenex, San Remo, USA), which is known to cross-react with human VDR, was applied to the sections for 2 h at 20&#x000b0;C after dilution 1 : 40 in TRIS-buffered saline. After 3 &#x000d7; 10 min washing in TRIS-buffered saline, biotinylated rabbit anti-rat immunoglobulin G (DAKO, Glostrup, Denmark) diluted 1 : 200 in TRIS-buffered saline was applied to the sections for 45 min at 20&#x000b0;C. After further washing in TRIS-buffered saline, alkaline phosphatase-conjugated ABC (Avidin-biotin complex system; DAKO) was applied to the sections for 45min at 20&#x000b0;C, diluted as instructed by the supplier. After further washing the alkaline phosphatase was developed using new fuchsin substrate to give a red colour. Sections were lightly counterstained using Harris's haematoxylin or toluidine blue. Non-immune rat immunoglobulin G was substituted for the primary antibody at similar concentrations on control tissue sections [<xref ref-type="bibr" rid="B10">10</xref>].</p></sec><sec><title>Cell cultures</title><p>Rheumatoid synovial tissue and human articular cartilage were enzymically digested to provide synovial fibroblast and chondrocyte cultures as previously described [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Cells were grown in Dulbecco's Modified Eagle's Medium + 10% (vol : vol) foetal calf serum, harvested and seeded into 12-well culture dishes (Nunc, Gibco, UK) Triplicate wells of confluent cell cultures in Dulbecco's Modified Eagle's Medium + 2% foetal calf serum were treated with 1&#x003b1;,25(OH)₂D₃ (10^-8 mol/l), IL-1&#x003b2; (0.05 ng/ml), or IL-1&#x003b2; + 1&#x003b1;,25(OH)₂D₃ (0.05 ng/ml and 10^-8 mol/l, respectively) and incubated for 48 h at 37&#x000b0;C. The conditioned medium was collected and assayed for MMP-1, MMP-2, MMP-3 and MMP-9, and PGE₂ using enzyme-linked immunosorbent assay (ELISA) methodology. Cell numbers per well were counted at the end of each experiment after 70% ethanol fixation and toluidine blue staining.</p></sec><sec><title>Enzyme linked immunosorbent assays</title><p>ELISA methodology was used to determine protein levels of MMP-1 (collagenase 1), MMP-3 (stromelysin) and MMP-9 (gelatinase B) as previously described [<xref ref-type="bibr" rid="B20">20</xref>]. MMP-2 (Gelatinase A) was measured using ELISA kits purchased from The Binding Site (Birmingham, UK); and PGE₂ was measured using an ELISA assay kit purchased from R &#x00026; D Systems Europe, Ltd (Abingdon,UK). </p><p>All ELISA results were initially calculated in ng or pg protein/ml culture medium/10⁶ cells per 48 h. Three different cultures of both RSFs and HACs were examined, but the capacities of each cell type to produce the MMPs and PGE₂ varied between the individual cultures. Therefore, the data from each culture was 'normalized' relative to control values, and the data sets from the three cultures of each cell type were subsequently pooled. This provided an evaluation that showed qualitative similarities for the RSFs and HACs, but demonstrated differences in 1&#x003b1;,25(OH)₂D₃ responses by each of these two cell types.</p></sec></sec><sec><title>Results</title><sec><title>Demonstration of the vitamin D receptor in rheumatoid tissues <italic>in vivo</italic></title><p>Specimens of rheumatoid synovial tissue (<italic>n</italic> = 18) immunostained for VDR were shown to have variable distributions of the receptor. All specimens showed some positive staining, but this could be less than 5% or as much as 70% of the total cell population. Different cell types within the synovial specimens were shown to express the receptor, including macrophages, endothelial cells, lymphocytes and fibroblastic cells, but no regular pattern was observed. Cells with fibroblastic morphology immunostained for VDR are shown in Figure <xref ref-type="fig" rid="F1">1a</xref>. Chondrocytes within articular cartilage from rheumatoid joints also expressed the receptor in six out of 10 specimens (Fig. <xref ref-type="fig" rid="F1">1b</xref>), this being a much higher frequency compared with the one in 10 specimens of normal articular cartilage from nonarthritic joints (data not shown). VDR-positive cells were also observed in association with some cartilage&#x02013;pannus junctions, described here as the rheumatoid lesion (Fig. <xref ref-type="fig" rid="F1">1c</xref>).</p></sec><sec><title>Effects of 1&#x003b1;,25-dihydroxyvitamin D₃ on matrix metalloproteinase production by rheumatoid synovial fibroblasts</title><p>1&#x003b1;,25(OH)₂D₃ alone had no effect on basal MMP production by RSFs in monolayer culture, but the simultaneous addition of 1&#x003b1;,25(OH)₂D₃ with IL-1&#x003b2; reduced the expected stimulation of MMP-1, MMP-3 and MMP-9 by up to 50% (Fig. <xref ref-type="fig" rid="F2">2</xref>: <italic>P</italic> = 0.096, 0.009 and 0.01, for IL-1&#x003b2; versus IL-1&#x003b2; + 1&#x003b1;,25(OH)₂D₃ for MMP-1, MMP-3 and MMP-9, respectively, by Student's <italic>t</italic>-test). MMP-2 production was not affected by either IL-1&#x003b2; or IL-1&#x003b2; + 1&#x003b1;,25(OH)₂D₃ (data not shown), an observation that is in accord with the constitutive nature of MMP-2 expression [<xref ref-type="bibr" rid="B21">21</xref>].</p></sec><sec><title>Effects of 1&#x003b1;,25-dihydroxyvitamin D₃ on matrix metalloproteinase production by human articular chondrocytes</title><p>In contrast to the data for RSFs, 1&#x003b1;,25(OH)₂D₃ had a slight stimulatory effect on basal production of MMP-1 and MMP-3 by monolayer cultures of HAC (Fig. <xref ref-type="fig" rid="F3">3</xref>: <italic>P</italic> = 0.098 and 0.002, for control versus 1&#x003b1;,25(OH)₂D₃, for MMP-1 and MMP-3, respectively, by Student's <italic>t</italic>-test). When stimulated with IL-1&#x003b2; MMP-1 and MMP-3 production was increased, and although simultaneous addition of 1&#x003b1;,25(OH)₂D₃ had no effect on the stimulation of the MMP-1 enzyme, MMP-3 production was further enhanced (Fig <xref ref-type="fig" rid="F3">3b</xref>: <italic>P</italic> = 0.008, by Students <italic>t</italic>-test). MMP-9 and MMP-2 were not produced in measurable quantities by these HAC cultures, either with or without IL-1&#x003b2; stimulation.</p></sec><sec><title>Effects of 1&#x003b1;,25-dihydroxyvitamin D₃ on prostaglandin E₂ production by rheumatoid synovial fibroblasts and human articular chondrocytes</title><p>PGE₂ production by RSFs was unaffected by the addition of 1&#x003b1;,25(OH)₂D₃ alone. Treatment of RSFs with IL-1&#x003b2; upregulated the production of PGE₂, but the addition of 1&#x003b1;,25(OH)₂D₃ together with IL-1&#x003b2; reduced the expected stimulation of PGE₂ almost to control values (Fig. <xref ref-type="fig" rid="F1">4a</xref>:<italic>P</italic> = 0.014, for IL-1&#x003b2; versus IL-1&#x003b2; + 1&#x003b1;,25(OH)₂D₃, by Student's <italic>t</italic>-test).</p><p>Treatment of HACs with IL-1&#x003b2; also increased the production of PGE₂, but in contrast to the effects noted for RSFs this IL-1-stimulation of PGE₂ was not affected by the concomitant addition of 1&#x003b1;,25(OH)₂D₃ (Fig. <xref ref-type="fig" rid="F4">4b</xref>).</p><p>To examine the possibility that 1&#x003b1;,25(OH)₂D₃ might obscure or interact with the IL-1&#x003b2; receptor of RSFs the latter were pretreated with 1&#x003b1;,25(OH)₂D₃ before incubation with IL-1&#x003b2;. Figure <xref ref-type="fig" rid="F4">4c</xref> shows that a 1-h preincubation with 1&#x003b1;,25(OH)₂D₃followed by IL-1&#x003b2; was not significantly different from the two factors added together, but preincubation with 1&#x003b1;,25(OH)₂D₃ for 16 h suppressed the expected increase in PGE₂ production to control values. This effect was noted even when the 1&#x003b1;,25(OH)₂D₃ was removed after the 16 h and IL-1&#x003b2; then added alone (Fig.4c, data column F). Thus, rather than directly interfering with the IL-1&#x003b2; receptor, it appears that 1&#x003b1;,25(OH)₂D₃ reduces the capacity of the RSFs to elaborate PGE₂ (and probably the MMPs shown in Fig. <xref ref-type="fig" rid="F2">2</xref>) after IL-1&#x003b2; induction.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Immunolocalization of the vitamin D receptor (VDR) in rheumatoid tissues. <bold>(a)</bold> Immunolocalization of VDR in rheumatoid synovium. Note positive red immunostaining of fibroblastic cells. (Counterstain Harris's haematoxylin; bar = 25 &#x003bc;m.) <bold>(b)</bold> Demonstration of VDR in cartilage from a rheumatoid joint. Note both positive and negative chondrocytes. (Counterstain Harris's haematoxylin; bar = 20 &#x003bc;m.) <bold>(c)</bold> VDR immunolocalization at the cartilage-pannus junction; cells within both pannus tissue and cartilage can be seen to be expressing the receptor. (Counterstain toluidine blue; bar = 25 &#x003bc;m.)</p></caption><graphic xlink:href="ar-1-1-063-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>The effects of 1&#x003b1;,25-hydroxyvitamin D₃ (1,25) on matrix metalloproteinase (MMP)-1, MMP-3 and MMP-9 production by rheumatoid synovial fibroblasts (RSFs) after 48 h incubation. <bold>(a)</bold> MMP-1 production by RSFs (<italic>n</italic> = 3) showing normalized values for control; + 1,25 (10^-8mol/l); + interleukin (IL)-1&#x003b2; (0.05 ng/ml); and + IL-1&#x003b2; and 1,25 (0.05 ng/ml and 10^-8 mol/l, respectively). Before normalization, control values for MMP-1 were in the range 50&#x02013;200 ng/ml culture medium/10⁶ cells per 48 h. <bold>(b)</bold> MMP-3 production by RSF (<italic>n</italic> = 3) showing normalized values for control; + 1,25; + IL-1&#x003b2;; and + IL-1&#x003b2; and 1,25. Before normalization, control values for MMP-3 were in the range 10&#x02013;40 ng/ml culture medium/10⁶ cells per 48 h. <bold>(c)</bold> MMP-9 production by RSFs (<italic>n</italic> = 3) showing normalized values for control; + 1,25; + IL-1&#x003b2;; and +IL-1&#x003b2; and 1,25. Before normalization, control values for MMP-9 were in the range 10-50 ng/ml culture medium/10⁶ cells per 48 h. Values are shown as means &#x000b1; SEM.</p></caption><graphic xlink:href="ar-1-1-063-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p>The effects of 1&#x003b1;,25-hydroxyvitamin D³ (1,25) on matrix metalloproteinase (MMP)-1 and MMP-3 production by human articular chondrocytes (HACs) after 48 h incubation. <bold>(a)</bold> MMP-1 production by HAC (<italic>n</italic> = 3) showing normalized data for control; + 1,25 (10^-8mol/l); + interleukin (IL)-1&#x003b2; (0.05 ng/ml); and + IL-1&#x003b2; and 1,25 (0.05 ng/ml and 10^-8 mol/l, respectively). Before normalization, control values for MMP-1 were in the range 50&#x02013;150 ng/ml culture medium/10⁶ cells per 48 h. <bold>(b)</bold> MMP-3 production by HAC (<italic>n</italic> = 3) showing normalized data for control; +1,25; + IL-1&#x003b2;; and + IL-1&#x003b2; and 1,25. Before normalization, control values for MMP-3 were in the range 10&#x02013;40 ng/ml culture medium/10⁶ cells per 48 h. Values are shown as means &#x000b1; SEM.</p></caption><graphic xlink:href="ar-1-1-063-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>The effects of 1&#x003b1;,25-hydroxyvitamin D₃ (1,25) on prostaglandin E₂ (PGE₂) production by rheumatoid synovial fibroblasts (RSFs) and human articular chondrocytes (HACs) after 48 h incubation. <bold>(a)</bold>PGE₂production by RSFs (<italic>n</italic> = 3) showing normalized data for control; + 1,25 (10^-8 mol/l); + interleukin (IL)-1&#x003b2; (0.05 ng/ml); and + IL-1&#x003b2; and 1,25 (0.05 ng/ml and 10^-8 mol/l, respectively). Before normalization, control values for PGE₂ production by RSFs were in the range 500&#x02013;2000 pg/ml culture medium/10⁶ cells per 48 h. <bold>(b) </bold>PGE₂ production by HACs (<italic>n</italic> = 3) showing normalized data for control; + 1,25; + IL-1&#x003b2;; and + IL-1&#x003b2; and 1,25. Before normalization control values for PGE₂ production by HAC were in the range 100&#x02013;300 pg/ml culture medium/10⁶ cells per 48 h. <bold>(c)</bold> Normalized data for PGE₂production by IL-1&#x003b2;-stimulated RSFs after preincubation with 1&#x003b1;,25(OH)₂D₃, as follows: A, control; B, +IL-1&#x003b2;; C, 1,25 + IL-1&#x003b2;; D, 1 h preincubation with 1,25, then + IL-1&#x003b2; and 1,25; E, 16 h preincubation with 1,25, then IL-1&#x003b2; + 1,25; F, 16 h preincubation with 1,25, then IL-1&#x003b2; alone. Values are shown as means &#x000b1; SEM.</p></caption><graphic xlink:href="ar-1-1-063-4"/></fig></sec></sec><sec><title>Discussion</title><p>The cell types within the rheumatoid lesion which were observed to express VDR included chondrocytes, fibroblasts, macrophages, lymphocytes and endothelial cells. These cells are all purported to be involved either directly or indirectly in the degradative processes associated with rheumatoid arthritis, possibly via their MMP and prostanoid production, or via the production of mediators responsible for inflammation and induction of proteinase expression by other cell types. Thus, the demonstration of VDR within the rheumatoid lesion provides support for a functional role of 1&#x003b1;,25(OH)₂D₃ in rheumatoid arthritis.</p><p>MMPs are considered to play important roles in the chondrolytic processes of the rheumatoid lesion [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. These enzymes are known to be produced by both fibroblasts and chondrocytes, but little has been reported in the literature regarding a relationship between 1&#x003b1;,25(OH)₂D₃and MMP production or its regulation, and most of the data to date have been obtained from animal studies or immortalized cell lines [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. 1&#x003b1;,25(OH)₂D₃ had little effect on basal MMP production by RSFs and marginally increased the basal production of MMP-1 and MMP-3 by chondrocytes. More pronounced differences were noted when IL-1&#x003b2;-stimulated or activated cells were treated with 1&#x003b1;,25(OH)₂D₃, the RSFs and HACs showing quite disparate responses. These opposite effects may be of relevance to the processes of joint destruction, especially cartilage loss, because the ability of 1&#x003b1;,25(OH)₂D₃to potentiate MMP-1 and MMP-3 expression by 'activated' chondrocytes might facilitate intrinsic cartilage chondrolysis <italic>in vivo</italic>. By contrast, the MMP-suppressive effects observed for 1&#x003b1;,25(OH)₂D₃ treatment of 'activated' synovial fibroblasts might reduce extrinsic chondrolysis and also matrix degradation within the synovial tissue. We recognize that the present study is somewhat restricted to the 1&#x003b1;,25(OH)₂D₃ effects on MMP-1 and MMP-3 production. Although these are prominent and well characterized MMPs, there are many other enzymes in this family, together with plasminogen activators and other proteinases, which have not been examined. From the disparate effects of 1&#x003b1;,25(OH)₂D₃ on the RSFs and HACs it would seem that further studies on the 1&#x003b1;,25(OH)₂D₃-modified proteinase phenotypes of these cells are warranted.</p><p>Prostaglandins are primary mediators of inflammation and have important roles in the immune response and the inflammatory processes associated with rheumatoid arthritis, and PGE₂ has been implicated in the potentiation of MMP production by some cell cultures [<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>]. 1&#x003b1;,25(OH)₂D₃ had little effect on basal PGE₂ production by RSFs, but the enhanced PGE₂ production observed following IL-1&#x003b2; stimulation of these cells was markedly suppressed by the concomitant addition of 1&#x003b1;,25(OH)₂D₃. By contrast, the increased PGE₂ production observed for IL-1&#x003b2;-treated HACs was unaffected by the simultaneous addition of 1&#x003b1;,25(OH)₂D_3. Thus, as with MMP production, 1&#x003b1;,25(OH)₂D₃ has disparate effects on IL-1&#x003b2;-stimulated PGE₂ production by these two cell types. Different responses by RSFs and HACs to the same ligand have been noted before; for example, IL-1&#x003b2; treatment was shown to stimulate glycosaminoglycan synthesis by RSFs, but inhibited its production by chondrocytes [<xref ref-type="bibr" rid="B24">24</xref>].</p><p>In summary, the immunolocalization of VDR in the rheumatoid lesion has demonstrated that the metabolite 1&#x003b1;,25(OH)₂D₃might have a functional role in the degradative and inflammatory processes of joint disease. Whereas 1&#x003b1;,25(OH)₂D₃ does not appear directly to affect the MMP or prostanoid production by unstimulated RSFs or HACs <italic>in vitro</italic>, it was shown to modulate the cytokine-induced MMP and PGE₂ production by these two cell cultures. The recognized immunomodulatory properties of 1&#x003b1;,25(OH)₂D₃ could well be important in rheumatoid tissues, in which the inflammatory response is a characteristic feature. The transient, local manifestations of cartilage and matrix-degrading activity [<xref ref-type="bibr" rid="B25">25</xref>] could be modified by 1&#x003b1;,25(OH)₂D₃ if the cells present express VDR and the metabolite is produced locally. This study has demonstrated that most rheumatoid synovial specimens were expressing VDR at the time of surgery, and that IL-1&#x003b2;-'activated' synovial fibroblasts and chondrocytes <italic>in vitro</italic> showed significant and different responses to 1&#x003b1;,25(OH)₂D₃ exposure with regard to MMP and PGE₂ production. Such observations suggest that 1&#x003b1;,25(OH)₂D₃ contributes indirectly rather than directly to MMP regulation via its action on other mediators or their signalling pathways, in accord with its recognized multifunctional and immunomodulatory properties [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B7">7</xref>].</p></sec>

Kinesin-like protein CENP-E is upregulated in rheumatoid synovial fibroblasts

<sec><title>Introduction:</title><p>Articular destruction by invading synovial fibroblasts is a typical feature in rheumatoid arthritis (RA). Recent data support the hypothesis that key players in this scenario are transformed-appearing synovial fibroblasts at the site of invasion into articular cartilage and bone. They maintain their aggressive phenotype toward cartilage, even when first cultured and thereafter coimplanted together with normal human cartilage into severe combined immunodeficient mice for an extended period of time. However, little is known about the upregulation of genes that leads to this aggressive fibroblast phenotype. To inhibit this progressive growth without interfering with pathways of physiological matrix remodelling, identification of pathways that operate specifically in RA synovial fibroblasts is required. In order to achieve this goal, identification of genes showing upregulation restricted to RA synovial fibroblasts is essential.</p></sec><sec><title>Aims:</title><p>To identify specifically expressed genes using RNA arbitrarily primed (RAP)-polymerase chain reaction (PCR) for differential display in patients with RA.</p></sec><sec><title>Methods:</title><p>RNA was extracted from cultured synovial fibroblasts from 10 patients with RA, four patients with osteoarthritis (OA), and one patient with psoriatic arthritis. RAP-PCR was performed using different arbitrary primers for first-strand and second-strand synthesis. First-strand and second-strand synthesis were performed using arbitrary primers: US6 (5' -GTGGTGACAG-3') for first strand, and Nuclear 1+ (5' -ACGAAGAAGAG-3'), OPN28 (5' -GCACCAGGGG-3'), Kinase A2+ (5' -GGTGCCTTTGG-3')and OPN24 (5' -AGGGGCACCA-3') for second-strand synthesis. PCR reactions were loaded onto 8 mol/l urea/6% polyacrylamide-sequencing gels and electrophoresed.Gel slices carrying the target fragment were then excised with a razor blade, eluated and reamplified. After verifying their correct size and purity on 4% agarose gels, the reamplified products derived from the single-strand confirmation polymorphism gel were cloned, and five clones per transcript were sequenced. Thereafter, a GenBank^&#x000ae; analysis was performed. Quantitative reverse transcription PCR of the segments was performed using the PCR MIMIC^&#x000ae; technique.In-situ expression of centromere kinesin-like protein-E (CENP-E) messenger (m)RNA in RA synovium was assessed using digoxigenin-labelled riboprobes, and CENP-E protein expression in fibroblasts and synovium was performed by immunogold-silver immunohistochemistry and cytochemistry. Functional analysis of CENP-E was done using different approaches (eg glucocorticoid stimulation, serum starvation and growth rate analysis of synovial fibroblasts that expressed CENP-E).</p></sec><sec><title>Results:</title><p>In RA, amplification of a distinct PCR product suitable for sequencing could be observed. The indicated complementary DNA fragment of 434 base pairs from RA mRNA corresponded to nucleotides 6615-7048 in the human centromere kinesin-like protein CENP-E mRNA (GenBank^&#x000ae; accession No. emb/Z15005).The isolated sequence shared greater than 99% nucleic acid (<italic>P</italic> = 2.9e^-169) identity with the human centromere kinesin-like protein CENP-E. Two base changes at positions 6624 (A to C) and 6739 (A to G) did not result in alteration in the amino acid sequence, and therefore 100% amino acid identity could be confirmed. The amplification of 10 clones of the cloned RAP product revealed the presence of CENP-E mRNA in every fibroblast culture examined, showing from 50% (271.000 &#x000b1; 54.000 phosphor imager arbitrary units) up to fivefold (961.000 &#x000b1; 145.000 phosphor imager arbitrary units) upregulation when compared with OA fibroblasts. Neither therapy with disease-modifying antirheumatic drugs such as methotrexate, gold, resochine or cyclosporine A, nor therapy with oral steroids influenced CENP-E expression in the RA fibroblasts. Of the eight RA fibroblast populations from RA patients who were receiving disease-modifying antirheumatic drugs, five showed CENP-E upregulation; and of the eight fibroblast populations from RA patients receiving steroids, four showed CENP-E upregulation.</p><p>Numerous synovial cells of the patients with RA showed a positive <italic>in situ</italic> signal for the isolated CENP-E gene segment, confirming CENP-E mRNA production in rheumatoid synovium, whereas in OA synovial tissue CENP-E mRNA could not be detected. In addition, CENP-E expression was independent from medication. This was further confirmed by analysis of the effect of prednisolone on CENP-E expression, which revealed no alteration in CENP-E mRNA after exposure to different (physiological) concentrations of prednisolone. Serum starvation also could not suppress CENP-E mRNA completely. </p></sec><sec><title>Discussion:</title><p>Since its introduction in 1992, numerous variants of the differential display method and continuous improvements including RAP-PCR have proved to have both efficiency and reliability in examination of differentially regulated genes. The results of the present study reveal that RAP-PCR is a suitable method to identify differentially expressed genes in rheumatoid synovial fibroblasts.</p><p>The mRNA, which has been found to be upregulated in rheumatoid synovial fibroblasts, codes for a kinesin-like motor protein named CENP-E, which was first characterized in 1991. It is a member of a family of centromere-associated proteins, of which six (CENP-A to CENP-F) are currently known. CENP-E itself is a kinetochore motor, which accumulates transiently at kinetochores in the G₂ phase of the cell cycle before mitosis takes place, appears to modulate chromosome movement and spindle elongation,and is degraded at the end of mitosis. The presence or upregulation of CENP-E has never been associated with RA.</p><p>The three-dimensional structure of CENP-E includes a coiled-coil domain. This has important functions and shows links to known pathways in RA pathophysiology. Coiled-coil domains can also be found in <italic>jun</italic> and <italic>fos</italic> oncogene products, which are frequently upregulated in RA synovial fibroblasts. They are also involved in DNA binding and transactivation processes resembling the situation in AP-1 (Jun/Fos)-dependent DNA-binding in rheumatoid synovium. Most interestingly, these coiled-coil motifs are crucial for the assembly of viral proteins, and the upregulation of CENP-E might reflect the influence of infectious agents in RA synovium. We also performed experiments showing that serum starvation decreased, but did not completely inhibit CENP-E mRNA expression. This shows that CENP-E is related to, but does not completely depend on proliferation of these cells. In addition, we determined the growth rate of CENP-E high and low expressors, showing that it was independent from the amount of CENP-E expression. supporting the statement that upregulation of CENP-E reflects an activated RA fibroblast phenotype. In summary, the results of the present study support the hypothesis that CENP-E, presumably independently from medication, may not only be upregulated, but may also be involved in RA pathophysiology.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Kullmann </surname><given-names>Frank</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Judex </surname><given-names>Martin</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Ballhorn </surname><given-names>Wibke</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>J&#x000fc;sten </surname><given-names>Hans-Peter</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Wessinghage </surname><given-names>Dieter</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Welsh</surname><given-names>John</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Yen</surname><given-names>Tim J</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Lang </surname><given-names>Bernhard</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Hittle</surname><given-names>Jim C</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A10" contrib-type="author"><name><surname>McClelland</surname><given-names>Michael</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A11" contrib-type="author"><name><surname>Gay</surname><given-names>Steffen</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A12" contrib-type="author"><name><surname>Sch&#x000f6;lmerich</surname><given-names>J&#x000fc;rgen</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A13" contrib-type="author"><name><surname>M&#x000fc;ller-Ladner</surname><given-names>Ulf</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Inflammation, altered cellular and humoral immune response, and synovial hyperplasia are typical findings in rheumatoid synovium pathophysiology [<xref ref-type="bibr" rid="B1">1</xref>]. On the other hand, there is increasing evidence that T-cell independent pathways, such as upregulation of proto-oncogenes, production of growth factors and the release of matrix-degrading enzymes, lead to progressive destruction of the affected joints [<xref ref-type="bibr" rid="B2">2</xref>]. Recent data [<xref ref-type="bibr" rid="B3">3</xref>] support the hypothesis that key players in this scenario are transformed-appearing synovial fibroblasts at the site of invasion into articular cartilage and bone. They maintain their aggressive phenotype toward cartilage even when first cultured and thereafter coimplanted together with normal human cartilage into severe combined immunodeficiency mice for an extended period.</p><p>To inhibit this progressive growth without interfering with pathways of physiological matrix remodelling, identification of pathways that operate specifically in rheumatoid arthritis (RA) synovial fibroblasts, and not in synovial fibroblasts of other origin [eg those of osteoarthritis (OA)], is required. To achieve this goal, identification of genes that show upregulation that is restricted to RA synovial fibroblasts is essential. Various strategies have been developed to examine tissue-specific gene differences in gene expression. Among these, the differential display approach of RNA arbitrarily primed (RAP)-polymerase chain reaction (PCR) [<xref ref-type="bibr" rid="B4">4</xref>] has been proved to be both efficient and reliable in numerous experimental settings [<xref ref-type="bibr" rid="B5">5</xref>].</p><p>In this study, we analyzed the differential expression of RNAs of RA fibroblasts versus OA fibroblasts derived from synovium of patients undergoing reconstructive surgery or synovectomy.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Synovial tissue and cell culture</title><p>Synovial fibroblasts and tissue were obtained from synovial biopsies of a total of 14 patients (Table <xref ref-type="table" rid="T1">1</xref>). Ten patients who met the criteria of the American College of Rheuma-tology for RA [<xref ref-type="bibr" rid="B6">6</xref>] were undergoing joint surgery (synovec-tomy or joint replacement by prosthesis implantation), and four patients with long-term OA were undergoing joint surgery (prosthesis implantation) because of severe articular dysfunction. In two patients with RA, fibroblast cultures were obtained not only from the proliferating synovium, but also separately from synovium that was extracted from visible, deep intraosseus invasion areas.</p><p>Culture of synovial fibroblasts was performed as described recently [<xref ref-type="bibr" rid="B3">3</xref>]. In brief, after enzymatic digestion, fibroblasts were grown in culture flasks in Dulbeccos Modified Eagles Medium-Cellgro (Mediatech, Washing-ton, DC, USA) containing 10% fetal calf serum (Gibco Life Technologies, Grand Island, NY, USA) and cultured for four passages. To exclude contamination, synovial fibroblasts were stained for fibroblast markers by immunocytochemistry [&#x0003e; 95% could be stained positively for the fibroblast enzyme prolyl-4-hydroxylase, and none were positive for the macrophage marker CD68 or the neutrophil marker cathepsin G (data not shown)] and tested for mycoplasmas. At 70&#x02013;80% confluency, cells were harvested and DNA was extracted as outlined below and stored at -70&#x000b0;C. Synovial tissue used for immunohistochemical analysis was immediately snap frozen and stored at -70&#x000b0;C.</p></sec><sec><title>RNA extraction</title><p>Total cellular RNA was extracted using the RNeasy spin column purification kit (Qiagen, Hilden, Germany). To remove contaminating genomic DNA, the total RNA was treated with DNase I (0.2U/&#x003bc;l; Boehringer Mannheim, Hannheim, Germany). RNA concentrations were measured spectrophotometrically at 260 nm and adjusted, and equal aliquots were then electrophoresed on 1% agarose gels stained with ethidium bromide to compare large and small ribosomal RNAs qualitatively and to exclude degradation. When starting with fresh RNAs, we performed one RAP-PCR, leaving out the reverse transcriptase as a control for DNA contamination.</p></sec><sec><title>RNA arbitrarily primed polymerase chain reaction of total cellular RNA</title><p>RAP-PCR of total cellular RNA was performed as previously described [<xref ref-type="bibr" rid="B4">4</xref>]. In brief, three different concentrations of RNA (500, 250 and 100 ng) were used as templates for each experiment. First-strand synthesis was carried out using 2 &#x003bc;mol/l first-strand arbitrary primer (for sequence see below), and second-strand synthesis was performed using 4 &#x003bc;mol/l arbitrary second primer and subsequently cycled through 35 low-stringency cycles. Arbitrary primers used were US6 (5' -GTGGTGACAG-3') for first-strand synthesis, and Nuclear 1+(5' -ACGAAGAAGAG-3'), OPN28 (5' -GCACCAGGGG-3'), kinase A2+ (5' -GGTGCCTTTGG-3') and OPN24 (5' -AGGGGCACCA-3') for second-strand synthesis. PCR products were loaded onto 8 mol/l urea/6% polyacrylamide sequencing gels. Electrophoresis was performed for 4&#x02013;6 h at 50 W in 1 &#x000d7; Tris-borate EDTA buffer. Gels were then transferred to 3MM Whatman paper, dried under vacuum at 80&#x000b0;C, and directly exposed to Kodak BioMax^&#x02122; autoradiography film (Kodak, Stuttgart, Germany) at room temperature for 12&#x02013;72 h, depending on the intensity of radiation of the amplified fragments.Several luminescence labels (autoradiogram markers; Stratagene, San Diego, CA, USA) were attached to the gel to facilitate alignment of the autoradiograms with the gels for subsequent isolation of differentially displayed gene fragments.</p></sec><sec><title>Isolation and purification of differentially amplified polymerase chain reaction products</title><p>Gel slices carrying the target fragment were then excized with a razor blade and placed in 50 &#x003bc;l TE buffer (10 mmol/l Tris-HCl, 1 mmol/l EDTA, pH8.0). DNA was eluted by incubating at 65&#x000b0;C for 3 h. Eluates were taken for reamplification of the gene fragment using the primers of the original fingerprint and the conditions outlined above for 20 cycles. PCR products were routinely checked by denaturing polyacrylamide gel electrophoresis, loading the reamplified product next to the original fingerprint to verify its size and purity. To exclude contamination of nondifferentially regulated gene products of similar size to that of the regulated transcripts, thereafter we used native polyacrylamide gels to separate the sequences of the reamplified mixture based on single-strand confirmation polymorphism (SSCP) as described previously [<xref ref-type="bibr" rid="B7">7</xref>]. After second identification of the gene segment using this procedure, it was cut from the SSCP gel and reamplified a second time.</p></sec><sec><title>Cloning, Southern blot and sequencing</title><p>After verifying its correct size and purity on 4% agarose gels, the reamplified products derived from the SSCP gel were cloned into PCR^&#x000ae;-II Topo using the TOPO-TA-Cloning^&#x000ae; Kit DUALPromoter (Invitrogen, De Schelp, The Nether-lands). After blue-white screening of the clones, 10 white colonies and one blue colony were picked and suspended in 50 &#x003bc;l water. Aliquots of these bacterial suspensions were checked by high-stringency PCR for the presence and the correct length of inserts using the T7 and the M13 (or M20) reverse sequencing primers. Clones of the correct length were subsequently grown overnight in 5ml LB medium containing 50 &#x003bc;g/ml of ampicillin for plasmid isolation. Five clones per transcript were sequenced with the Applied Biosystems 373 automatic sequencer using the Perkin Elmer (Norwalk, CT, USA) DNA sequencing kit.The databases of the National Center for Biological Information were screened to align the obtained sequences with known complementary DNA clones, genomic clones and cloned expressed sequence tags.</p><p>If sequences were multiply represented and confirmed within the majority of the clones resulting from one RAP-PCR product, inserts were reamplified and used as probes against Southern blots of the original fingerprint gel. This procedure confirms that selected sequences were in fact differentially amplified in the original fingerprint gels. For this, DNA from RAP-PCR fingerprint gels was transferred to nylon membranes (Duralon-UV, Stratagene, San Diego, CA, USA) by capillary action in a 10 &#x000d7; saline sodium citrate (SSC) buffer overnight. After ultraviolet cross-linking, blots were prehybridized and hybridized using established protocols [<xref ref-type="bibr" rid="B8">8</xref>].</p></sec><sec><title>Confirmation of differential expression by quantitative reverse transcription polymerase chain reaction</title><p>Quantitative reverse transcription PCR was performed using the PCR MIMIC^&#x000ae;technique (Clontech, Palo Alto, CA, USA). In this competitive PCR method, one set of primers is used to amplify both the target complementary DNA and the nonhomologous internal standard. As the amplified PCR fragments are designed to be different in length, they can be distinguished by gel electrophoresis. Using this method, relative transcript abundances of the different fibroblast populations could be compared after standardization of the coamplified highly abundant internal standard DNA. Radioactive PCR products amplified using centromere kinesin-like protein-E (CENP-E)-specific primers were evaluated by phosphor imaging densitometry (Phosphor Imager; Molecular Dynamics, Sunnyvale, CA, USA) and subsequent data analysis was performed using the Ambis software (Imagequant; Molecular Dynamics). Each PCR amplification was performed in quadruplicate, and relative amount was calculated as phosphor imager arbitrary units (PAU) of CENP-E amplicons of the RA and the OA fibroblast cultures.</p><p>Mean &#x000b1; standard error of the mean was calculated for each of the individual RA fibroblast cultures and compared with the CENP-E production of all OA synovial fibroblast cultures. Statistical analysis was performed using the Mann-Whitney test for nonpaired parameters. <italic>P</italic> &#x0003c; 0.05 was regarded as statistically significant.</p></sec><sec><title>In-situ expression of CENP-E messenger RNA in rheumatoid arthritis synovium</title><p>Plasmids containing the isolated CENP-E gene segment were extracted and purified using Nucleobond-AX-Columns (Macherey-Nagel, D&#x000fc;ren, Germany) and linearized to permit generation of antisense and sense riboprobes. In-situ hybridization was performed as published recently [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. In brief, antisense and sense RNA probes were transcribed by T3 and T7 RNA polymerase (Stratagene). Probes were labelled with digoxigenin-uridine triphosphate (Boehringer Mannheim). Frozen sections (4&#x02013;6 &#x003bc;m thick) were cut and fixed in 3% buffered paraformaldehyde for 1h at room temperature(20&#x02013;22&#x000b0;C). The sections were rinsed in 2 &#x000d7; SSC/0.25% acetic anhydride (Fisher Scientific, Springfield, NJ, USA) for 15 min at room temperature. After a rinsing step with 0.1mol/l tri-ethanolamine-HCl (pH8.0), prehybridization was performed. After the prehybridization, digoxigenin-labelled antisense or sense probe (for control) was applied to the tissue specimens in 15 l prehybridization buffer. The slides were sealed with nail polish and hybridized for 12 h in a humid chamber at 50&#x000b0;C. Then slides were washed at room temperature with SSC and sodium chloride-Tris-EDTA (STE) buffer. After digestion for 1 h at 37&#x000b0;C with 20 g/ml RNAse A (Boehringer Mannheim), slides were rinsed with 2 &#x000d7; SSC, 50% formamide for 5 min, followed by 1 &#x000d7; SSC, 0.1% sodium dodecyl sulphate for 10 min, and 0.5 &#x000d7; SSC, 0.1% sodium dodecyl sulphate for 15 min at 50&#x000b0;C. The slides were washed and developed using antidigoxigenin-5 nm gold-labelled antibody complex (Goldmark Biologicals, Phillipsburg, NJ, USA) according to a modification to the protocol of Komminoth <italic>et al</italic> [<xref ref-type="bibr" rid="B9">9</xref>].</p></sec><sec><title>Immunogold-silver immunohistochemistry for CENP-E</title><p>RA and OA synovial fibroblasts cultured in chamber wells and snap frozen sections (4&#x02013;6 &#x003bc;m thick) were fixed for 5 min in acetone and covered with a 4% milk, 2% normal horse serum Tris buffer for 30 min at room temperature (20&#x02013;22&#x000b0;C) to block nonspecific binding. Then, slides were washed and incubated with monoclonal mouse anti-CENP-E-antibody [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>], or antifibroblast antibodies (Dianova, Hamburg, Germany). The slides were washed and incubated with a biotinylated goat antimouse antibody (Dianova) diluted 1 : 600 in Tris buffer. After washing, sections were incubated with peroxidase-conjugated streptavidin (Dianova), diluted 1 : 600 in Tris buffer. After a rinsing step, 6nm gold-labelled goat antihorseradish peroxidase (Dianove), diluted 1 : 40 in Tris buffer was applied. Signal detection by the immunogold-silver technique was performed as described above.</p></sec><sec><title>Immunohistochemical double labelling (alkaline phosphatase anti-alkaline phosphatase method)</title><p>Double labelling was performed using the alkaline phosphatase anti-alkaline phosphatase method with mono-clonal antibodies against human fibroblasts (Dianova). Sections were covered with a 4% milk, 2% normal horse serum buffer for 30 min at room temperature to block nonspecific binding, washed in Tris-NaCl (pH7.5) and incubated for 45 min at room temperature (20&#x02013;22&#x000b0;C) with the primary antibodies diluted 1 : 50 to 1 : 100 in Tris-NaCl (pH 7.5). Colour development was performed as previously published [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>Negative controls were performed in each of the techniques described above by omitting the primary antibody, incubation with isotype-matched controls or using the sense probe in the <italic>in situ</italic> hybridization assay.</p></sec><sec><title>Glucocorticoid stimulation</title><p>Three of the RA fibroblast populations showing upregulation of CENP-E (RA08, RA09 and RA21) and three populations showing no upregulation (RA10, RA17 and RA22) were adjusted to 0.5 &#x000d7; 10⁶ cells and incubated with physiological concentrations of 10^-7 and 10^-9 mmol/l prednisolone for 6 h. Thereafter, CENP-E messenger (m)RNA was measured using reverse transcription PCR as outlined above.</p></sec><sec><title>Serum starvation</title><p>Three of the RA fibroblast populations showing upregulation of CENP-E (RA08, RA09 and RA21) and three populations showing no upregulation (RA10, RA17 and RA22) were adjusted to 0.5 &#x000d7; 10⁶ cells and were cultured for 72 h without fetal calf serum. Thereafter, CENP-E mRNA was measured using reverse transcription PCR as outlined above.</p></sec><sec><title>Growth rate</title><p>Cells (10⁵) of RA fibroblast populations showing upregulation of CENP-E (RA08, RA09 and RA21) and showing no upregulation (RA10, RA17 and RA22) were cultured in quintuplicate in 24 well plates. To evaluate the growth rate, cell counts were performed on days 2, 3, 4 and 5, and increase in cell numbers from day to day was evaluated using a Neubauer counting chamber and measured as percentage per day.</p><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Clinical data of patients</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Patient</td><td align="center">Age</td><td align="center">Male/female</td><td align="center">ESR (mm/1st hour)</td><td align="center">WBC (leucocytes/nl)</td><td align="center">Medication</td></tr></thead><tbody><tr><td align="left">RA02</td><td align="center">58</td><td align="center">F</td><td align="center">15</td><td align="center">16</td><td align="center">M, O</td></tr><tr><td align="left">RA03</td><td align="center">59</td><td align="center">F</td><td align="center">85</td><td align="center">16.5</td><td align="center">C, O</td></tr><tr><td align="left">RA04</td><td align="center">37</td><td align="center">M</td><td align="center">26</td><td align="center">8.8</td><td align="center">G, O</td></tr><tr><td align="left">RA06</td><td align="center">55</td><td align="center">M</td><td align="center">37</td><td align="center">7.8</td><td align="center">M, O</td></tr><tr><td align="left">RA08</td><td align="center">66</td><td align="center">F</td><td align="center">45</td><td align="center">8.7</td><td align="center">-</td></tr><tr><td align="left">RA09</td><td align="center">55</td><td align="center">F</td><td align="center">29</td><td align="center">14.5</td><td align="center">-</td></tr><tr><td align="left">RA10</td><td align="center">46</td><td align="center">F</td><td align="center">25</td><td align="center">11.7</td><td align="center">G, O</td></tr><tr><td align="left">RA17</td><td align="center">68</td><td align="center">M</td><td align="center">13</td><td align="center">6.7</td><td align="center">O</td></tr><tr><td align="left">RA21</td><td align="center">31</td><td align="center">M</td><td align="center">26</td><td align="center">9.7</td><td align="center">G</td></tr><tr><td align="left">RA22</td><td align="center">70</td><td align="center">F</td><td align="center">20</td><td align="center">12.5</td><td align="center">G, O</td></tr><tr><td align="left">OA02</td><td align="center">76</td><td align="center">F</td><td align="center">9</td><td align="center">10.4</td><td align="center">-</td></tr><tr><td align="left">OA03</td><td align="center">74</td><td align="center">F</td><td align="center">20</td><td align="center">8.1</td><td align="center">-</td></tr><tr><td align="left">OA05</td><td align="center">65</td><td align="center">M</td><td align="center">13</td><td align="center">7.1</td><td align="center">-</td></tr><tr><td align="left">OA06</td><td align="center">74</td><td align="center">M</td><td align="center">14</td><td align="center">5.5</td><td align="center">-</td></tr></tbody></table><table-wrap-foot><p>Abbreviations of disease-modifying antirheumatic drugs: C, cyclosporin A; G, gold; M, methotrexate; O, oral steroids; R, resochine; S, sulphasalazine.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Growth rate of CENP-E-expressing rheumatoid synovial fibroblasts</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td align="center">Growth rate</td></tr><tr><td></td><td align="center">(increase factor of cells per 24 h,</td></tr><tr><td></td><td align="center">mean &#x000b1; standard derivation)</td></tr></thead><tbody><tr><td align="left">CENP-E high</td><td align="center">1.14 &#x000b1; 0.19</td></tr><tr><td align="left">CENP-E low</td><td align="center">1.03 &#x000b1; 0.15</td></tr></tbody></table><table-wrap-foot><p>Determination of the growth rate of CENP-E high and low expressors reveals that the growth rate is slightly higher, but not significantly different with regard to the amount of CENP-E expression.</p></table-wrap-foot></table-wrap></sec></sec><sec><title>Results</title><sec><title>RNA arbitrarily primed polymerase chain reaction</title><p>Total RNA, prepared from synovial fibroblasts from 10 patients with RA and four patients with OA, was used for RAP-PCR. In total, approximately 150 RNAs were amplified using RAP-PCR per primer pair, of which most were expressed both by RA and OA synovial fibroblasts. Approximately 2% of the RNAs showed a polymorphism within the RA group. Five RAP-products were differentially expressed between RA and nonRA synovial fibroblasts, and one PCR product was suitable for further analysis. The autoradiography of the original RAP-PCR gel is shown in Figure <xref ref-type="fig" rid="F1">1</xref>, and the arrowheads indicate the complementary DNA fragment that we isolated for further characterization.</p></sec><sec><title>Cloning and sequencing of differentially amplified RNA arbitrarily primed product</title><p>The indicated complementary DNA fragment of 434 base pairs from RA mRNA corresponded to nucleotides 6615-7048 in the human centromere kinesin-like protein CENP-E mRNA (GenBank^&#x000ae;accession No. emb/Z15005) The RAP-PCR product was located entirely within the open reading frame. The isolated sequence shared greater than 99% nucleic acid (<italic>P</italic> = 2.9e^-169) identity with the human centromere kinesin-like protein CENP-E. Two base changes at positions 6624 (A to C) and 6739 (A to G) did not result in alteration of the amino acid sequence, and therefore 100% amino acid identity could be confirmed. The amplification of 10 clones of the cloned RAP product revealed the presence of CENP-E mRNA in every fibroblast culture examined.</p><p>We confirmed the predicted upregulation of CENP-E in RA using quantitative reverse transcription PCR. When analyzed by phosphor imaging densitometry, significant upregulation of CENP-E mRNA production in RA synovial fibroblasts as compared with OA synovial fibroblasts (173.000&#x000b1; 46.000 PAU) was observed in eight out of 12 synovial fibroblasts populations of patients with RA. CENP-E upregulation among the fibroblast populations ranged from approximately 50% (271.000 PAU, culture RA04) to fivefold (961.000PAU, culture RA21) upregulation (Fig. <xref ref-type="fig" rid="F2">2a</xref>). In six additional synovial fibroblast populations, the RA fibroblasts also showed an upregulation of CENP-E expression as compared with psoriatic arthritis fibroblasts (Fig. <xref ref-type="fig" rid="F2">2b</xref>). Of interest, neither therapy with disease-modifying antirheumatic drugs such as methotrexate, gold, resochine or cyclosporine A, nor therapy with oral steroids showed an influence on CENP-E expression in the RA fibroblasts. Of the eight RA fibroblast populations from RA patients receiving disease-modifying antirheumatic drugs, five showed an CENP-E upregulation; and of the eight fibroblast populations from RA patients receiving steroids, four showed a CENP-E upregulation. Furthermore, no difference in upregulation was seen in the RA synovial fibroblasts obtained from the intraosseus invasion areas (RA3K and RA9K), indicating a general upregulation of CENP-E in all synovial tissue compartments.</p></sec><sec><title>Confirmation of CENP-E protein synthesis</title><p>Synthesis of CENP-E was not only confirmed on the mRNA level as outlined above, but also on the protein level by immunohistochemistry using monoclonal antibodies against human CENP-E. As illustrated in Figure <xref ref-type="fig" rid="F3">3</xref>, cultured rheumatoid synovial fibroblasts expressed intensive amounts of CENP-E (Fig. <xref ref-type="fig" rid="F3">3a</xref>), whereas in OA fibroblasts CENP-E production was below detection level (Fig. <xref ref-type="fig" rid="F3">3b</xref>). Figure <xref ref-type="fig" rid="F3">3c</xref> shows the positive control using antifibroblast antibodies.</p></sec><sec><title>Expression of CENP-E in rheumatoid synovium</title><p>Numerous synovial cells of the patients with RA showed a positive <italic>in situ</italic> signal for the isolated CENP-E gene segment, confirming CENP-E mRNA production in rheumatoid synovium, whereas in OA synovial tissue CENP-E mRNA could not be detected. CENP-E expressing cells were found throughout the synovium, both in the lining layer as well as in the sublining. Because numerous CENP-E expressing cells showed an atypical fibroblast phenotype, we performed double-labelling using monoclonal fibroblast antibodies. The majority of CENP-E protein-expressing cells could be double-labelled with fibroblast antibodies, indicating that the CENP-E production seen in the cultured cells is not an effect caused by culture conditions. Figures <xref ref-type="fig" rid="F4">4a</xref> and <xref ref-type="fig" rid="F4">4b</xref> show synovial fibroblasts intensively expressing mRNA for the CENP-E gene segment isolated by RAP-PCR.</p><p>To evaluate whether CENP-E mRNA production led to protein production detectable by immunohistochemistry, monoclonal antibodies directed against human CENP-E [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>] were used. The most intensive CENP-E protein expression was found within highly inflamed areas in the sublining, and, to a lesser extent, in the lining layer. Figure <xref ref-type="fig" rid="F4">4c</xref> shows a synovial section with numerous synovial fibroblasts intensively expressing CENP-E protein.</p></sec><sec><title>Functional assays</title><p>As outlined above, CENP-E expression was independent from medication. This was further confirmed by analysis of the effect of prednisolone on CENP-E expression, which revealed no alteration in CENP-E mRNA after exposure to different (physiological) concentrations of prednisolone (Fig. <xref ref-type="fig" rid="F5">5</xref>). Serum starvation was also not able to suppress CENP-E mRNA completely. As shown in Figure <xref ref-type="fig" rid="F5">5</xref>, although in lesser amounts, CENP-E mRNA can still be detected in all RA synovial fibroblast populations, even after numerous days of serum starvation. Determination of the growth of CENP-E high and low expressors showed that the growth rate was slightly increased in the CENP-E high expressors, but by statistical criteria independent from the amount of CENP-E expression (Table <xref ref-type="table" rid="T2">2</xref>). Also, the doubling time of CENP-E high and low expressors did not differ (data not shown).</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Original fingerprint [RNA arbitrarily primed (RAP)-polymerase chain reaction (PCR)] of RNA of synovial fibroblasts derived from 10 different patients with rheumatoid arthritis and six patients with osteoarthritis. Each RAP-PCR was performed in triplicate using three different RNA concentrations. Left two lanes: DNA marker. Triplicate lanes 1&#x02013;10 show the RAP-PCR of the synovial fibroblasts of the 10 rheumatoid arthritis patients, and lanes 11&#x02013;16 show the RAP-PCR of those of the osteoarthritis patients. The arrow indicates the differentially displayed gene fragment seen in the rheumatoid arthritis patients (434 base pairs), which was used for subsequent analysis.</p></caption><graphic xlink:href="ar-1-1-071-1"/></fig><fig position="float" id="F2"><label>Figure 2</label><caption><p>Quantitative reverse transcription polymerase chain reaction of the CENP-E gene segment followed by phosphor imaging densitometry evaluation. The polymerase chain reaction MIMIC^&#x000ae; technique was used. <bold>(a)</bold> Significant upregulation in CENP-E in eight out of 12 rheumatoid arthritis synovial fibroblast cultures (white bars) as compared with all osteoarthritis fibroblast cultures (dotted bar; total <italic>n</italic> = 4); whereas four rheumatoid arthritis fibroblast populations did not show an upregulation in CENP-E upregulation (black bars). Values are shown as phosphor imager arbitrary units (PAU, mean &#x000b1; standard error of the mean). Each MIMIC^&#x000ae; was performed in quadruplicate. <italic>P</italic> &#x0003c; 0.05 was considered statistically significant. <bold>(b)</bold> Significant upregulation in six additional rheumatoid arthritis synovial fibroblast cultures (white bars) as compared with psoriatic arthritis fibroblasts (generous gift from E M&#x000e4;rker-Herrmann, University of Mainz, Germany). Ratio of intensities are shown (when compared to standard 18S RNA).</p></caption><graphic xlink:href="ar-1-1-071-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p>Immunocytochemistry of cultured rheumatoid arthritis fibroblasts using monoclonal antibodies against human CENP-E protein [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>]. Note the strong expression of CENP-E protein rheumatoid fibroblasts <bold>(a)</bold> in comparison with the absence of CENP-E protein in osteoarthritis fibroblasts <bold>(b)</bold>. <bold>(c)</bold> The positive control was performed with antifibroblast antibodies. Immunogold-silver staining, original magnifications &#x000d7; 800 (a, b) and &#x000d7; 1000 (c).</p></caption><graphic xlink:href="ar-1-1-071-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>Demonstration of CENP-E messenger RNA and protein in rheumatoid synovium by <italic>in situ</italic> hybridization and immunohistochemistry. <bold>(a, b)</bold> A digoxigenin-labelled RNA probe transcribed from the amplified RNA arbitrarily primed-PCR gene product is used on rheumatoid arthritis snap-frozen sections. Double-labelling for CENP-E messenger RNA (black staining), and alkaline phosphatase antialkaline phosphatase counterstaining using anti-fibroblast antibodies (red staining) shows CENP-E expression in numerous fibroblasts througout the synovium (a, arrows). (b) The intensive CENP-E messenger RNA expression in two synovial fibroblasts. <bold>(c)</bold> Shows numerous fibroblasts expressing CENP-E protein [black staining (arrows), counterstaining with fast red]. Original magnifications &#x000d7; 300 (a, c) and &#x000d7; 600 (b).</p></caption><graphic xlink:href="ar-1-1-071-4"/></fig><fig position="float" id="F5"><label>Figure 5</label><caption><p>CENP-E regulation. The figure shows CENP-E messenger RNA in CENP-E high (right lanes 3/4, 7/8, 11/12 and 15/16) and CENP-E low (left lanes 1/2, 5/6, 9/10 and 13/14) expressing rheumatoid arthritis synovial fibroblast populations after different prednisolone exposure (lanes 1&#x02013;8) and serum starvation (lanes 9&#x02013;12). As compared with unstimulated fibroblasts (lanes 13&#x02013;16) no significant alteration in CENP-E messenger RNA expression, either in CENP-E high- or in CENP-E low-expressing fibroblast populations, can be observed after prednisolone exposure. Serum starvation was also not able to suppress CENP-E messenger RNA completely (lanes 9&#x02013;12).</p></caption><graphic xlink:href="ar-1-1-071-5"/></fig></sec></sec><sec><title>Discussion</title><p>Since its introduction in 1992 [<xref ref-type="bibr" rid="B12">12</xref>], numerous variants of the differential display method [<xref ref-type="bibr" rid="B13">13</xref>] and continuous improvements including RAP-PCR [<xref ref-type="bibr" rid="B14">14</xref>] have proved to have both efficiency and reliability in examination of differentially regulated genes. Although the majority of experimental approaches have addressed malignant diseases [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>], the results of this study reveal that, along with differential subtraction [<xref ref-type="bibr" rid="B17">17</xref>] and the original differential display [<xref ref-type="bibr" rid="B18">18</xref>], RAP-PCR is a particularly suitable method to identify differentially expressed genes in rheumatoid synovial fibroblasts.</p><p>The mRNA, which has been found to be upregulated in rheumatoid synovial fibroblasts, codes for a kinesin-like motor protein named CENP-E, which was first characterized in 1991 [<xref ref-type="bibr" rid="B10">10</xref>]. It is a member of a family of centromere-associated proteins, of which six (CENP-A to CENP-F) are currently known. The complete CENP-E gene spans a length of 8371 bases, the open reading frame encodes 2663 amino acids, and the respective protein has a molecular weight of 312 kDa and is a member of the human centromere-kinetochore complex [<xref ref-type="bibr" rid="B11">11</xref>]. CENP-E itself is a kinetochore motor, which accumulates transiently at kinetochores in the G₂ phase of the cell cycle before mitosis takes place, appears to modulate chromosome movement and spindle elongation, and is degraded at the end of mitosis [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. Consistently, inhibition of the CENP-E protein by specific antibodies causes cell cycle arrest at metaphase [<xref ref-type="bibr" rid="B11">11</xref>]. In addition, further research showed that kinesin-like proteins, including CENP-E, specifically regulate microtubule formation and the consecutive movement of chromosomes during mitosis, and therefore are crucial for cell division [<xref ref-type="bibr" rid="B23">23</xref>].</p><p>Interestingly, the presence or upregulation of CENP-E has never been associated with RA, although antibodies to the CENP family are frequently found in numerous rheumatic diseases [<xref ref-type="bibr" rid="B24">24</xref>], and CENP-E has recently been proposed to be an autoantigen in the limited form of systemic sclerosis [<xref ref-type="bibr" rid="B25">25</xref>].</p><p>The three-dimensional structure of CENP-E includes a coiled-coil domain. This has important functions that have links to known pathways in RA pathophysiology. Coiled-coil domains can also be found in <italic>jun</italic> and <italic>fos</italic> oncogene products [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>], which are frequently upregulated in RA synovial fibroblasts [<xref ref-type="bibr" rid="B28">28</xref>]. They are also involved in DNA binding and transactivation processes, resembling the situation in activating protein-1 (Jun/Fos)-dependent DNA-binding in rheumatoid synovium [<xref ref-type="bibr" rid="B29">29</xref>]. Most interestingly, these coiled-coil motifs are crucial for the assembly of viral proteins [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>], and the upregulation of CENP-E might reflect the influence of infectious agents in RA synovium [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B32">32</xref>]. In addition, CENP-E may also be involved in activation of rheumatoid synovial fibroblasts, because presence of CENP-E has been found to correlate with the active state of centomeres in translocations [<xref ref-type="bibr" rid="B33">33</xref>]. These pathways may include a distinct dysregulation of the cell cycle (eg rare activation steps in the formation of neocentromeres [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>]), because rheumatoid synovial fibroblasts, although they have a transformed appearance, do not reveal an increased rate of proliferation <italic>in vitro</italic> or <italic>in vivo</italic> [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B37">37</xref>].</p><p>We also performed experiments showing that serum starvation decreased, but not completely inhibited CENP-E mRNA expression, showing that CENP-E is related to but does not completely depend on proliferation of these cells. In addition, we determined the growth rate of CENP-E high and low expressors, showing that it was independent on the amount of CENP-E expression, supporting the statement that upregulation of CENP-E reflects an activated RA fibroblast phenotype.</p><p>In summary, the results of this study support the hypothesis that CENP-E, presumably independently from medication, may not only be upregulated, but also be involved in RA pathophysiology.</p></sec>

Protection against cartilage and bone destruction by systemic interleukin-4 treatment in established murine type II collagen-induced arthritis

<sec><title>Introduction:</title><p>Rheumatoid arthritis (RA) is associated with an increased production of a range of cytokines including tumour necrosis factor (TNF)-&#x003b1; and interleukin (IL)-1, which display potent proinflammatory actions that are thought to contribute to the pathogenesis of the disease. Although TNF-&#x003b1; seems to be the major cytokine in the inflammatory process, IL-1 is the key mediator with regard to cartilage and bone destruction. Apart from direct blockade of IL-1/TNF, regulation can be exerted at the level of modulatory cytokines such as IL-4 and IL-10. IL-4 is a pleiotropic T-cell derived cytokine that can exert either suppressive or stimulatory effects on different cell types, and was originally identified as a B-cell growth factor and regulator of humoral immune pathways. IL-4 is produced by activated CD4⁺ T cells and it promotes the maturation of Th2 cells. IL-4 stimulates proliferation, differentiation and activation of several cell types, including fibroblasts, endothelial cells and epithelial cells. IL-4 is also known to be a potent anti-inflammatory cytokine that acts by inhibiting the synthesis of proinflammatory cytokines such as IL-1, TNF-&#x003b1;, IL-6, IL-8 and IL-12 by macrophages and monocytes. Moreover, IL-4 stimulates the synthesis of several cytokine inhibitors such as interleukin-1 receptor antagonist (IL-1Ra), soluble IL-1-receptor type II and TNF receptors IL-4 suppresses metalloproteinase production and stimulates tissue inhibitor of metalloproteinase-1 production in human mononuclear phagocytes and cartilage explants, indicating a protective effect of IL-4 towards extracellular matrix degradation. Furthermore, IL-4 inhibits both osteoclast activity and survival, and thereby blocks bone resorption <italic>in vitro</italic>. Of great importance is that IL-4 could not be detected in synovial fluid or in tissues. This absence of IL-4 in the joint probably contributes to the disturbance in the Th1/Th2 balance in chronic RA.</p><p>Collagen-induced arthritis (CIA) is a widely used model of arthritis that displays several features of human RA. Recently it was demonstrated that the onset of CIA is under stringent control of IL-4 and IL-10. Furthermore, it was demonstrated that exposure to IL-4 during the immunization stage reduced onset and severity of CIA. However, after cessation of IL-4 treatment disease expression increased to control values.</p></sec><sec><title>Aims:</title><p>Because it was reported that IL-4 suppresses several proinflammatory cytokines and matrix degrading enzymes and upregulates inhibitors of both cytokines and catabolic enzymes, we investigated the tissue protective effect of systemic IL-4 treatment using established murine CIA as a model. Potential synergy of low dosages of anti-inflammatory glucocorticosteroids and IL-4 was also evaluated.</p></sec><sec><title>Methods:</title><p>DBA-1J/Bom mice were immunized with bovine type II collagen and boosted at day 21. Mice with established CIA were selected at day 28 after immunization and treated for days with IL-4, prednisolone, or combinations of prednisolone and IL-4. Arthritis score was monitored visually. Joint pathology was evaluated by histology, radiology and serum cartilage oligomeric matrix protein (COMP). In addition, serum levels of IL-1Ra and anticollagen antibodies were determined.</p></sec><sec><title>Results:</title><p>Treatment of established CIA with IL-4 (1 &#x003bc;g/day) resulted in suppression of disease activity as depicted in Figure <xref ref-type="fig" rid="F1">1</xref>. Of great interest is that, although 1 &#x003bc;g/day IL-4 had only a moderate effect on the inflammatory component of the disease activity, it strongly reduced cartilage pathology, as determined by histological examination (Fig. <xref ref-type="fig" rid="F1">1</xref>). Moreover, serum COMP levels were significantly reduced, confirming decreased cartilage involvement. In addition, both histological and radiological analysis showed that bone destruction was prevented (Fig. <xref ref-type="fig" rid="F1">1</xref>). Systemic IL-4 administration increased serum IL-1Ra levels and reduced anticollagen type II antibody levels. Treatment with low-dose IL-4 (0.1 &#x003bc;g/day) was ineffective in suppressing disease score, serum COMP or joint destruction. Synergistic suppression of both arthritis severity and COMP levels was noted when low-dose IL-4 was combined with prednisolone (0.05 mg/kg/day), however, which in itself was not effective.</p></sec><sec><title>Discussion:</title><p>In the present study, we demonstrate that systemic IL-4 treatment ameliorates disease progression of established CIA. Although clinical disease progression was only arrested and not reversed, clear protection against cartilage and bone destruction was noted. This is in accord with findings in both human RA and animal models of RA that show that inflammation and tissue destruction sometimes are uncoupled processes. Of great importance is that, although inflammation was still present, strong reduction in serum COMP was found after exposure to IL-4. This indicated that serum COMP levels reflected cartilage damage, although a limited contribution of the inflamed synovium cannot be excluded.</p><p>Increased serum IL-1Ra level (twofold) was found after systemic treatment with IL-4, but it is not likely that this could explain the suppression of CIA. We and others have reported that high dosages of IL-1Ra are needed for marked suppression of CIA. As reported previously, lower dosages of IL-4 did not reduce clinical disease severity of established CIA. Of importance is that combined treatment of low dosages of IL-4 and IL-10 appeared to have more potent anti-inflammatory effects, and markedly protected against cartilage destruction. Improved anti-inflammatory effect was achieved with IL-4/prednisolone treatment. In addition, synergistic effects were found for the reduction of cartilage and bone destruction. This indicates that systemic IL-4/prednisolone treatment may provide a cartilage and bone protective therapy for human RA.</p><fig position="float" id="F1"><label>Figure 1</label><caption><p>Effects in mice of treatment with interleukin-4 or control on disease activity, cartilage damage and bone destruction. Mice were treated intraperitoneally for 7 days with either vehicle (control) or 1 &#x003bc;g/day interleukin-4 (IL-4). CIA, collagen-induced arthritis. *<italic>P</italic> &#x0003c; 0.05, versus control, by Mann-Whitney U test.</p></caption><graphic xlink:href="ar-1-1-081-1"/></fig></sec>

<contrib id="A1" contrib-type="author"><name><surname>Joosten</surname><given-names>Leo AB</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>l.joosten@reuma.azn.nl</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Lubberts</surname><given-names>Erik</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Helsen</surname><given-names>Monique MA</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Saxne</surname><given-names>Tore</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Coenen-de Roo</surname><given-names>Christina JJ</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Heineg&#x000e5;rd</surname><given-names>Dick</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>van den Berg</surname><given-names>Wim B</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Interleukin (IL)-4 is a pleiotropic T-cell-derived cytokine that can exert either suppressive or stimulatory effects on different cell types. It was originally identified as a B-cell growth factor and regulator of humoral immune pathways [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. IL-4 is produced by activated CD4⁺ T cells and it promotes the maturation of Th2 cells. IL-4 inhibits the differentiation of na&#x000ef;ve T cells to Th1 and cytokine (ie IL-2 and interferon-&#x003b3;) production by Th1 cells [<xref ref-type="bibr" rid="B3">3</xref>]. IL-4 stimulates proliferation, differentiation or activation of several cell types, including fibroblasts, endothelium cells and epithelium cells [<xref ref-type="bibr" rid="B4">4</xref>]. IL-4 is also known to be a potent anti-inflammatory cytokine that acts by inhibiting the synthesis of proinflammatory cytokines such as IL-1, tumour necrosis factor (TNF)-&#x003b1;, IL-6, IL-8 and IL-12 by macrophages and monocytes [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. Moreover, IL-4 stimulates the synthesis of several cytokine inhibitors such as interleukin-1 receptor antagonist (IL-1Ra), IL-1-receptor type II and TNF receptors [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. IL-4 suppresses metalloproteinase production and stimulates tissue inhibitor of metalloproteinase-1 production in human mononuclear phagocytes and cartilage explants, indicating a protective effect of IL-4 towards extracellular matrix degradation [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. Furthermore, IL-4 inhibits both osteoclast activity and survival, and thereby blocks bone resorption <italic>in vitro</italic> [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>].</p><p>RA is associated with an increased production of a range of cytokines including TNF&#x003b1; and IL-1, which display potent proinflammatory actions that are thought to contribute to the pathogenesis of rheumatoid arthritis (RA) [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. Although TNF-&#x003b1; seems to be the major cytokine involved in the inflammatory process, IL-1 is the key mediator with regard to cartilage and bone destruction [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. Apart from direct blockade of IL-1/TNF, regulation can be exerted at the level of modulatory cytokines such as IL-4 and IL-10. Of great importance is that IL-4 could not be detected in synovial fluid and tissues [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>], and this lack of IL-4 is likely to contribute to the uneven Th1/Th2 balance in chronic RA.</p><p>Although having a number of side effects, including osteoporosis and reduced adrenal function, glucocorticoids are potent and commonly used anti-inflammatory agents in human RA. Glucocorticoids downregulate proinflammatory cytokine production, such as IL-1 and TNF-&#x003b1;, by macrophages and monocytes via several mechanisms. One mechanism is through enhanced I&#x003ba;B&#x003b1; protein synthesis. I&#x003ba;B&#x003b1; forms inactive cytoplasmic complexes with nuclear factor-&#x003ba;B, which itself activates many immunoregulatory genes in response to proinflammatory cytokines [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. Other mechanisms of action that have been reported recently [<xref ref-type="bibr" rid="B23">23</xref>] are downmodulation of histone acetyltransferase and upregulation of histone deacetyltransferase, which both affected messenger RNA transcription negatively.</p><p>Murine collagen-induced arthritis (CIA) is a widely used experimental model of arthritis. Neutralization of the monokines IL-1 and TNF-&#x003b1; before or during onset of arthritis arrested the development of CIA [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Expression of CIA is also under particularly stringent control by IL-4 and IL-10. Treatment with anti-IL-4/anti-IL-10 shortly before onset accelerated the disease expression [<xref ref-type="bibr" rid="B26">26</xref>]. Furthermore, it was demonstrated that IL-12 plays a crucial role in the development of CIA, because blockade of endogenous IL-12 completely prevented onset of the disease [<xref ref-type="bibr" rid="B27">27</xref>]. In accord with these findings, during onset of CIA predominantly Th1 responses towards collagen type II were found [<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>]. It has been claimed [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>] that IL-4 exposure could induce immune deviation by enhanced development of Th2-like primary CD4 effector cells. Several animal studies indicated that IL-4 administration, starting just after immunization with the disease-inducing agent, ameliorated Th1-mediated models of autoimmune diseases such as diabetes in nonobese diabetic mice and experimental arthritis [<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>,<xref ref-type="bibr" rid="B34">34</xref>].</p><p>In the present study the effects of systemic high dose IL-4 therapy in established CIA were investigated. Furthermore, the potential synergy of combined prednisolone and IL-4 treatment were examined. We investigated the protective effect of IL-4 alone or in combination with prednisolone on disease activity as well as cartilage and bone destruction as determined histologically, radiologically and by serum measurements of cartilage oligomeric matrix protein (COMP). Anticollagen type II specific antibodies and serum IL-1Ra levels were assessed, in order to obtain an insight into the mechanism of action. The findings suggest that IL-4 treatment protects against cartilage and bone destruction, and that combined IL-4/steroid treatment may provide a safe, anti-inflammatory and anti-destructive therapy in human RA.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Animals</title><p>Male DBA-1/Bom mice were purchased from Bomholdg&#x000e5;rd (Ry, Denmark). The mice were housed in filter top cages, and were given free access to water and food. The mice were immunized at the age of 10&#x02013;12 weeks.</p></sec><sec sec-type="materials"><title>Materials</title><p>Complete Freund's adjuvant and <italic>Mycobacterium tuberculosis</italic> (strain H37Ra) were obtained from Difco Laboratories (Detroit, MI, USA). Bovine serum albumin and prednisolone 21-sodium succinate (P-4153) were purchased from Sigma Chemicals (St Louis, MO, USA). Antimurine IL-1Ra antibodies (capture MAP-480, detection BAF-480) were obtained from R&#x00026;D Systems (Minneapolis, MN, USA). PolyHRP-streptavidine (M2032) and Caseine colloid buffer (M2052) was from CLB (Amsterdam, The Netherlands). Recombinant murine IL-1Ra was purchased from R&#x00026;D systems. Recombinant murine IL-4 (6.5 &#x000d7; 10⁷ U/mg) was kindly provided by Dr S Smith (Schering-Plough, Kenilworth, NJ, USA).</p></sec><sec><title>Collagen preparation</title><p>Articular cartilage was obtained from metacarpophalangeal joints of 1&#x02013;2 year old cows. Bovine type II collagen was prepared according to the method of Miller and Rhodes [<xref ref-type="bibr" rid="B35">35</xref>]. It was dissolved in 0.05 mol/l acetic acid (5 mg/ml) and stored at -70&#x000ba;C.</p></sec><sec><title>Immunization</title><p>Bovine type II collagen was diluted with 0.05 mol/l acetic acid to a concentration of 2 mg/ml and was emulsified in an equal volume of complete Freund's adjuvant (2 mg/ml MT H37Ra). The mice were immunized intradermally at the base of the tail with 100 &#x003bc;l emulsion (100 &#x003bc;g collagen). At day 21 the animals were boosted with an intra-peritoneal injection of 100 &#x003bc;g collagen type II, diluted in phosphate-buffered saline (pH 7.4).</p></sec><sec><title>Assessment of arthritis</title><p>Mice were examined for visual appearance of arthritis in peripheral joints, and scores for severity were given (arthritis score) as previously described [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>]. Mice were considered arthritic when significant changes in redness and/or swelling were noted in digits or in other parts of the paws. At later time points ankylosis was also included in the arthritis score. Clinical severity of arthritis was graded on a scale of 0&#x02013;2 for each paw, according to changes in redness and swelling: 0, no changes; 0.5, significant; 1.0, moderate; 1.5, marked; and 2.0, maximal swelling and redness, and later on ankylosis. Arthritis score (mean&#x000b1; stan-dard deviation) was expressed as cumulative value for all paws, with a maximum of eight and expressed as percentage of the initial score at the beginning of treatment.</p></sec><sec><title>Treatment of collagen-induced arthritis with interleukin-4, prednisolone or interleukin-4/prednisolone</title><p>To evaluate the effect of IL-4, prednisolone or the combination IL-4/prednisolone on established CIA, mice with CIA were selected at day 28 and divided into groups of at least 10 mice with similar arthritis scores. Thereafter, mice were treated twice a day intraperitoneally with IL-4 (0.1 or 1&#x003bc;g/day), prednisolone (0.05 mg/kg/day), or with IL-4 and prednisolone (at the same doses for the noncombined regimens) for each of several days as indicated in the results.</p></sec><sec><title>Determination of interleukin-1 receptor antagonist levels</title><p>IL-1Ra was measured using enzyme-linked immunosorbent assay (ELISA). Briefly, Nunc Maxisorb ELISA plates (Nunc, Rostilde, Denmark) were coated with capture antibodies (5 &#x003bc;g/ml, carbonate buffer, pH 9.6, 24 h at 4&#x000b0;C), and thereafter nonspecific binding sites were blocked with 1% bovine serum albumin/phosphate-buffered saline-Tween. Standards and unknown samples were diluted in normal DBA-1 serum and incubated for 3 h at room temperature. Biotinylated detection antibodies were added at concentrations of 0.2&#x02013;0.4 &#x003bc;g/ml in 0.5% bovine serum albumin in phosphate-buffered slaine-Tween for 1.5 h at room temperature. Thereafter plates were incubated with PolyHRP (0.1 &#x003bc;g/ml in 1% caseine colloid buffer) for 45 min and orthophenylenediamine (0.8 mg/ml) was used as substrate. Plates were read at 495 nm.</p></sec><sec><title>Measurement of cartilage oligomeric matrix protein</title><p>At the end of the experiments, serum samples were taken and murine cartilage oligomeric matrix protein (COMP) levels were determinated using ELISA under similar conditions as those described for the assay for human COMP [<xref ref-type="bibr" rid="B36">36</xref>]. The assay was modified by using rat COMP for coating the microtitre plates, the standard curve included in each plate and by using the polyclonal antiserum raised against rat COMP to detect the antibody [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>]. A high cross-reactivity was found to murine COMP [<xref ref-type="bibr" rid="B39">39</xref>]. This was shown by parallel dilution curves of murine sera to the standard curve prepared with rat COMP, as well as in experiments in which a dilution of murine serum was added to the standard curve.</p></sec><sec><title>Determination of anticollagen antibodies</title><p>Antibodies against bovine type II collagen were examined by using an ELISA. Titres of total IgG, IgG₁ and IgG_2a were measured. Briefly, plates were coated with 10 &#x003bc;g bovine type II, and thereafter nonspecific bindings sites were blocked with 0.1 mol/l ethanolamin (Sigma Chemicals). Serial 1 : 2 dilutions of the sera were added, followed by incubation with isotype-specific goat antimouse peroxidase (Southern Biotechnology Associates, Birmingham, AL, USA) and substrate (5-aminosalicyclic acid; Sigma Chemicals). Plates were read at 492 nm. Titres were expressed as means &#x000b1; standard deviation dilution, which gives the half maximal value.</p></sec><sec><title>Radiological analysis of bone destruction</title><p>At the end of the experiments, knee joints were removed and used for radiological analysis as a measure of bone destruction. Radiographs were carefully examined using a stereo microscope, and joint destruction was graded on a scale from 0 to 5, ranging from no damage, minor bone destruction observed as one enlightened spot, moderate changes, two to four spots observed in one area, marked changes, two to four spots observed in more areas, severe erosions afflicting the joint, complete destruction of joint and/or new bone formations. Bone destruction was scored on the femoral head, tibia and patella as described previously [<xref ref-type="bibr" rid="B17">17</xref>].</p></sec><sec><title>Histology</title><p>Mice were killed by ether anaesthesia. Knee joints were removed and fixed for 4 days in 4% formaldehyde. After decalcification in 5% formic acid, the specimens were processed for paraffin embedding [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>]. Tissue sections (7 &#x003bc;m thick) were stained with haematoxylin and eosin, or safranin O. Histopathological changes were scored using the following parameters.</p><p>Infiltration of cells was scored on a scale from 0 to 3, depending on the amount of inflammatory cells in the synovial tissues. Inflammatory cells in the joint cavity were graded on a scale from 0 to 3 and expressed as exudate. Cartilage proteoglycan depletion was determined using safranin O staining. The loss of proteoglycans was scored on a scale from 0 to 3, ranging from fully stained cartilage to destained cartilage or complete loss of articular cartilage. A characteristic parameter in CIA is the progressive loss of articular cartilage. This destruction was separately graded on a scale from 0 to 3, ranging from the appearance of dead chondrocytes (empty lacunae) to complete loss of the articular cartilage. Bone erosion was scored on a scale ranging from 0 to 3, ranging from no abnormalities to complete loss of cortical and trabecular bone of the femoral head and patella. Histopathological changes in the knee joints were scored in the patella/femur region on 5 semiserial sections of the joint, spaced 70 &#x003bc;m apart. Scoring was performed on decoded slides by two observers, as described earlier [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>].</p></sec><sec><title>Statistical analysis</title><p>Differences between experimental groups were tested using the Mann-Whitney U test, unless otherwise stated.</p></sec></sec><sec><title>Results</title><sec><title>Amelioration of arthritis score in collagen-induced arthritis by <italic>in vivo</italic> treatment of interleukin-4</title><p>To investigate effects of <italic>in vivo</italic> treatment of established CIA with IL-4, mice that expressed CIA at day 28 after immunization were injected intraperitoneally with vehicle, 0.1 or 1 &#x003bc;g IL-4 per day. Figure <xref ref-type="fig" rid="F2">2</xref> shows that administration of 1 &#x003bc;g/day IL-4 results in significant amelioration of the arthritis score, but a lower dosage of 0.1 &#x003bc;g/day IL-4 was without effect. The anti-inflammatory effect of 1 &#x003bc;g/day IL-4 was further illustrated in Figure <xref ref-type="fig" rid="F3">3</xref>, in which disease progression is expressed as change in (&#x00394;) disease activity of all individual mice. Increased severity of CIA score can be seen in animals treated either with vehicle or 0.1 &#x003bc;g/day IL-4, whereas significantly decreased disease activity was noted after treatment with 1 &#x003bc;g/day IL-4. Histology revealed that no effect was found on the influx of inflammatory cells in joint tissues of IL-4-treated animals when compared with the vehicle-treated animals (Table <xref ref-type="table" rid="T1">1</xref>).</p></sec><sec><title>Interleukin-4 protects against cartilage destruction</title><p>Systemic treatment with high-dose IL-4 (1&#x003bc;g/day) significantly decreased cartilage destruction, determined as chondrocyte death and cartilage erosions (Fig. <xref ref-type="fig" rid="F4">4</xref>,Table <xref ref-type="table" rid="T1">1</xref>). It did not result in a significantly reduced loss of matrix proteoglycans, as determined by safranin O staining (Fig. <xref ref-type="fig" rid="F5">5</xref>, Table <xref ref-type="table" rid="T1">1</xref>). It has been demonstrated (data not shown) that there is a strong correlation between severe cartilage damage and increased serum COMP levels during murine CIA. In na&#x000ef;ve DBA-1 mice, serum COMP levels are approximately 4.0 &#x003bc;g/ml and COMP levels increased up to 8&#x02013;12 &#x003bc;g/ml in mice with fully established CIA. Serum COMP levels were determined in the various groups to identify the protection against severe cartilage destruction by IL-4. Figure <xref ref-type="fig" rid="F6">6</xref> shows that elevated COMP in CIA were not reduced by treatment with low-dose IL-4. It is of particular interest, that treatment with high-dose IL-4 (1 &#x003bc;g/day) significantly reduced serum COMP levels to values found in nonarthritic control animals.</p></sec><sec><title>Interleukin-4 protects against bone destruction</title><p>Bone destruction, which is a common feature of murine collagen arthritis, was examined by radiological analysis. Radiographs of knee joints were taken at the end of the treatment period. Figure <xref ref-type="fig" rid="F7">7</xref> showed that treatment with 1 &#x003bc;g/day IL-4 was sufficient to prevent bone destruction, determined as bone erosions on the head of the femur, the patella and the tibia. Little or no effect was noted after treatment with low-dose IL-4. Histological analysis of knee joints corroborated the protective effect of IL-4 (Table <xref ref-type="table" rid="T1">1</xref>). Figure <xref ref-type="fig" rid="F8">8</xref> (a, c) depicts degradation of patellar and femural cortical bone by osteoclasts in the vehicle-treated group, whereas almost no osteoclasts were seen in the IL-4-treated group (Fig. <xref ref-type="fig" rid="F8">8d</xref>). </p></sec><sec><title>Combined interleukin-4/prednisolone treatment</title><p>We examined potential synergistic effects of IL-4 and prednisolone, using low-dose prednisolone (0.05 mg/kg/day) and 0.1 or 1&#x003bc;g/day IL-4. Treatment of CIA with IL-4/prednisolone completely arrested the development of inflammatory signs of CIA (Figs <xref ref-type="fig" rid="F2">2</xref> and <xref ref-type="fig" rid="F3">3</xref>). Both combinations tested revealed full suppression of disease progression. In accord with previous observations, mice treated with 0.05 mg/kg/day prednisolone alone did not show significant suppression of arthritis. Histology taken after 7 days of treatment showed enhanced safranin O staining only in animals treated with IL-4/prednisolone (1 &#x003bc;g per kg/0.05 kg daily), indicating reduced depletion of matrix proteoglycans (Table <xref ref-type="table" rid="T1">1</xref>, Fig. <xref ref-type="fig" rid="F5">5d</xref>). This was in accord with the marked reduction in joint inflammation, as can be seen in Figure <xref ref-type="fig" rid="F4">4d</xref>. Both combinations of IL-4 and prednisolone reduced serum COMP to values found in na&#x000ef;ve DBA-1 mice. Interestingly, synergistic suppression of serum COMP was noted after exposure to low-dose IL-4 and prednisolone (Fig. <xref ref-type="fig" rid="F6">6</xref>). In contrast to serum COMP levels, combined IL-4/prednisolone treatment did not result in synergistic protection against bone destruction. High-dose IL-4 alone was already highly effective, and the combination of IL-4 with prednisolone did not improve the effect further, or was there an adverse effect of prednisolone (Table <xref ref-type="table" rid="T1">1</xref>, Figs <xref ref-type="fig" rid="F7">7</xref> and <xref ref-type="fig" rid="F8">8b</xref>). Treatment of CIA with 1 &#x003bc;g/day IL-4 alone and in combination with prednisolone (0.05 mg/kg/day) for 7 days caused similar reduction in osteoclast numbers (data not shown).</p></sec><sec><title>Effect of interleukin-4, or interleukin-4/prednisolone treatment on interleukin-1 receptor antagonist and anticollagen antibody levels</title><p>Serum IL-1Ra levels were determined at the end of the experiments and Table <xref ref-type="table" rid="T2">2</xref> shows a twofold increase after IL-4 treatment (1&#x003bc;g/day dose). Treatment with 0.1&#x003bc;g/day IL-4 showed no significant effects on serum IL-1Ra levels. Prednisolone reduced IL-1Ra levels when compared with vehicle-treated animals. In accord with these findings, combined IL-4/prednisolone (1 &#x003bc;g per day/ 0.05 mg per kg per day) treatment resulted in lower IL-1Ra levels than found with IL-4 alone.</p><p>Anticollagen antibodies were assayed at the end of treatment period at day 35. The antibody levels increased rapidly after clinical expression of CIA around day 28 after immunization. After IL-4 (1&#x003bc;g/day) treatment for 7 days, total IgGs levels as well as IgG₁ and IgG₂a anticollagen type II antibody levels were lower compared with vehicle treated animals (Fig. <xref ref-type="fig" rid="F9">9</xref>). Although all anticollagen type II antibodies were reduced, IgG_2a levels showed the most prominent reduction, indicating an effect on the Th1 rather than on the Th2 immune response. No decreased anticollagen type II antibody levels were found after treatment with low-dose IL-4. The high-dose IL-4/prednisolone regimen reduced anticollagen type II antibodies to levels similar to those found after treatment with 1&#x003bc;g/day IL-4.</p><fig position="float" id="F2"><label>Figure 2</label><caption><p>Dose dependent suppression of disease activity of collagen-induced arthritis (CIA) by interleukin (IL)-4 and the combination of IL-4/prednisolone (Pred). Mice with established CIA were divided into separate groups of at least 10 mice. Groups were treated intraperitoneally twice a day with vehicle, IL-4, prednisolone, or combined IL-4/prednisolone for 8 consecutive days. The data represent the mean arthritis score, expressed as percentage of initial value at day 28. Experiments were repeated once with approximately the same outcome. *<italic>P</italic> &#x0003c; 0.05, versus vehicle, by Mann-Whitney U test.</p></caption><graphic xlink:href="ar-1-1-081-2"/></fig><fig position="float" id="F3"><label>Figure 3</label><caption><p>Dose-dependent arrest of disease activity by treatment with interleukin (IL)-4 and IL-4/prednisolone (Pred). The enhanced disease activity between days 28 and 35 of each individual mouse is expressed as change in (&#x00394;) disease activity. For treatment protocol, see Fig. <xref ref-type="fig" rid="F2">2</xref>. <italic>P</italic> &#x0003c; 0.05, versus vehicle, by Mann-Whitney U test.</p></caption><graphic xlink:href="ar-1-1-081-3"/></fig><fig position="float" id="F4"><label>Figure 4</label><caption><p>Interleukin (IL)-4 treatment reduced cartilage destruction, whereas IL-4/prednisolone treatment additionally decreased cell influx. <bold>(a)</bold> Knee joint from vehicle-treated mouse. Severe cartilage destruction can be seen. Empty lacunae reflects chondrocyte death as marker of cartilage destruction, indicated by arrows. <bold>(b)</bold> Knee joint of a mouse treated with IL-4 1 &#x003bc;g/kg/day for eight consecutive days. Note the reduced cartilage destruction and chondrocyte death. <bold>(c)</bold> Knee joint of vehicle-treated animal. Note the severe cell influx in synovial tissues and joint cavity. <bold>(d)</bold> Knee joint of a mouse treated with IL-4/prednisolone (1 &#x003bc;g per day/0.05 mg per kg). Note the marked reduction of cell influx. All specimens were sampled at day 35. P, patella; F, femur; JS, joint space; C, cartilage; S, synovium. Haematoxylin and eosin staining was used. Original magnifications: &#x000d7; 200 (a, b) and &#x000d7; 100 (c, d).</p></caption><graphic xlink:href="ar-1-1-081-4"/></fig><fig position="float" id="F5"><label>Figure 5</label><caption><p>Effect of IL-4 or IL-4/prednisolone treatment on matrix proteoglycan loss. <bold>(a)</bold> Knee joint of a control na&#x000ef;ve mouse. The fully stained cartilage layers indicate no loss of proteoglycans. <bold>(b)</bold> Knee joint of an arthritic mouse treated with vehicle. Note the severe joint inflammation and complete loss of safranin O staining of the cartilage layers (indicated by arrows). <bold>(c)</bold> Mouse treated with IL-4 (1 &#x003bc;g/day). Loss of matrix proteoglycan can still be seen. <bold>(d) </bold>Knee joint of a mouse treated with IL-4/prednisolone (1 &#x003bc;g per day/0.05 mg per kg). Marked reduction in matrix proteoglycan depletion after combined treatment. For details see Fig. <xref ref-type="fig" rid="F4">4</xref>. Safranin O staining, original magnification &#x000d7; 100.</p></caption><graphic xlink:href="ar-1-1-081-5"/></fig><fig position="float" id="F6"><label>Figure 6</label><caption><p>Serum cartilage oligomeric matrix protein (COMP) level as a marker of cartilage turnover. Suppression of serum COMP was found after treatment with interleukin (IL)-4 and IL-4/prednisolone (Pred). IL-4(1 &#x003bc;g/day) and both doses (0.1 &#x003bc;g per day/0.05 mg per kg per day; and 1 &#x003bc;g per day/0.05 mg per kg per day) of IL-4/prednisolone reduced serum COMP levels to basic levels as found in nonimmunized animals (4.2 &#x000b1; 0.2 &#x003bc;g/ml). The data represent the mean&#x000b1; standard deviation COMP level of at least six sera per group. *<italic>P</italic> &#x0003c; 0.01, versus vehicle, by Mann-Whitney U test.</p></caption><graphic xlink:href="ar-1-1-081-6"/></fig><fig position="float" id="F7"><label>Figure 7</label><caption><p>Protection of interleukin (IL)-4 and IL-4/prednisolone (Pred) treatment on bone destruction. Knee joints were isolated at day 35 and bone destruction was examined by radiographic analysis. For treatment scheme see Fig. <xref ref-type="fig" rid="F2">2</xref>. Erosions were scored on a scale ranging from 0 to 5 on the femur head, tibia and patella. Each group consists of at least nine knee joints per group. *<italic>P</italic> &#x0003c; 0.01, versus vehicle, by Mann-Whitney U test.</p></caption><graphic xlink:href="ar-1-1-081-7"/></fig><fig position="float" id="F8"><label>Figure 8</label><caption><p>Bone destruction is prevented by interleukin (IL)-4 and IL-4/prednisolone treatment. <bold>(a)</bold> Severe bone destruction in patella and femur in knee joint of vehicle-treated animal. <bold>(b)</bold> Almost no bone degradation was noted after treatment with IL-4/prednisolone (1 &#x003bc;g per day/0.05 mg per kg). <bold>(c)</bold> Bone destruction in femur of a vehicle-treated animal at higher magnification. Osteoclasts, large multinuclear cells, located at the site of bone destruction (arrows). <bold>(d)</bold> No osteoclast-like cells were found in IL-4 (1 &#x003bc;g/day) treated animals. For treatment details see Fig. <xref ref-type="fig" rid="F4">4</xref>. S, synovium; B, bone; BM, bone marrow. Original magnifications &#x000d7; 200 (a, b), &#x000d7; 400 (c, d).</p></caption><graphic xlink:href="ar-1-1-081-8"/></fig><fig position="float" id="F9"><label>Figure 9</label><caption><p>Interleukin (IL)-4 or IL-4/prednisolone (Pred) treatment is associated with reduced anticollagen type II (CII) antibody levels. Treatment with 1 &#x003bc;g/day IL-4 resulted in lower anticollagen type II antibodies. Total Immunoglobulins (Ig tot), IgG¹ and IgG^2a levels were reduced. Similar effects were found after treatment with IL-4/prednisolone (1 &#x003bc;g per day/0.05 mg per kg). Anticollagen type II levels were determined in at least six mice per group. Data are expressed as means &#x000b1; standard deviation dilution, which gives the half maximal value.</p></caption><graphic xlink:href="ar-1-1-081-9"/></fig><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Effect of prednisolone, interleukin (IL)-4 or IL-4/prednisolone treatment on the joint pathology of collagen-induced arthritis in Mice</p></caption><table frame="hsides" rules="groups"><thead><tr><td></td><td></td><td></td><td align="center">Cartilage</td><td align="center">Proteoglycan</td><td align="center">Bone</td><td></td></tr><tr><td align="left">Treatment</td><td align="center">Dose</td><td align="center">Infiltrate</td><td align="center">destruction</td><td align="center">loss</td><td align="center">erosin</td><td align="center"><italic>n</italic></td></tr></thead><tbody><tr><td align="left">Vehicle</td><td align="center">-</td><td align="center">2.3 &#x000b1; 0.9</td><td align="center">2.2 &#x000b1; 0.9</td><td align="center">2.7 &#x000b1; 1.0</td><td align="center">1.9 &#x000b1; 0.9</td><td align="center">20</td></tr><tr><td align="left">Prednisolone</td><td align="center">0.05</td><td align="center">2.1 &#x000b1; 0.8</td><td align="center">2.1 &#x000b1; 1.2</td><td align="center">2.6 &#x000b1; 0.6</td><td align="center">1.7 &#x000b1; 1.1</td><td align="center">20</td></tr><tr><td align="left">IL-4</td><td align="center">0.1</td><td align="center">2.5 &#x000b1; 0.7</td><td align="center">2.5 &#x000b1; 0.8</td><td align="center">2.9 &#x000b1; 0.3</td><td align="center">2.0 &#x000b1; 0.8</td><td align="center">10</td></tr><tr><td align="left">IL-4</td><td align="center">1</td><td align="center">2.0 &#x000b1; 1.0</td><td align="center">1.2 &#x000b1; 0.8*</td><td align="center">2.0 &#x000b1; 0.7</td><td align="center">0.6 &#x000b1; 0.6*</td><td align="center">20</td></tr><tr><td align="left">IL-4/prednisolone</td><td align="center">0.1/0.05</td><td align="center">2.1 &#x000b1; 0.4</td><td align="center">1.6 &#x000b1; 0.7</td><td align="center">2.5 &#x000b1; 0.8</td><td align="center">2.1 &#x000b1; 0.6</td><td align="center">10</td></tr><tr><td align="left">IL-4/prednisolone</td><td align="center">1/0.05</td><td align="center">1.2 &#x000b1; 0.5*</td><td align="center">1.1 &#x000b1; 0.9*</td><td align="center">1.4 &#x000b1; 0.7*</td><td align="center">0.4 &#x000b1; 0.5*</td><td align="center">10</td></tr></tbody></table><table-wrap-foot><p>Histopathology scores of arthritic knee joints after treatment with vehicle, IL-4, prednisolone or the combination of IL-4/prednisolone. Mice were sacrified and knee joints were used for histology. Histology was scored as indicated in the Materials and methods section. Mice were treated twice a day intraperitoneally with either prednisolone (0.05 mg/kg), or IL-4 (0.1 or 1 &#x003bc;g/day], or IL-4 (at both dosages) combined with prednisolone (0.05 mg/kg). *<italic>P</italic> &#x0003c;0.05, versus vehicle, by Mann-Whitney U test.</p></table-wrap-foot></table-wrap><table-wrap position="float" id="T2"><label>Table 2</label><caption><p>Serum interleukin-1 receptor antagonist (IL-1Ra levels) after treatment with either interleukin (IL)-4, prednisolone, or IL-4/prednisolone</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Treatment</td><td align="center">Dose</td><td align="center">IL-1Ra (pg/ml)</td></tr></thead><tbody><tr><td align="left">Vehicle</td><td align="center">-</td><td align="center">414 &#x000b1; 155</td></tr><tr><td align="left">IL-4</td><td align="center">0.1</td><td align="center">386 &#x000b1; 213</td></tr><tr><td align="left">IL-4</td><td align="center">1</td><td align="center">838 &#x000b1; 187*</td></tr><tr><td align="left">Prednisolone</td><td align="center">0.05</td><td align="center">326 &#x000b1; 165</td></tr><tr><td align="left">IL-4/prednisolone</td><td align="center">0.1/0.05</td><td align="center">422 &#x000b1; 129</td></tr><tr><td></td><td align="center">1/0.05</td><td align="center">499 &#x000b1; 187</td></tr></tbody></table><table-wrap-foot><p>Serum IL-1Ra was determined using enzyme-linked immunosorbent assay at day 35 after immunization. Mice were treated as indicated in Table <xref ref-type="table" rid="T1">1</xref>. The data represent the mean&#x000b1; standard deviation of at least eight mice per group. The sensitivity of the IL-1Ra assay was to within 160 pg/ml. *<italic>P</italic> &#x0003c; 0.05, versus vehicle, by Mann-Whitney U test.</p></table-wrap-foot></table-wrap></sec></sec><sec><title>Discussion</title><p>The present study demonstrates clear tissue-protective effects of IL-4, although IL-4 did not prove to be a very potent anti-inflammatory cytokine. Both cartilage and bone erosion were prevented by IL-4 treatment of established CIA. Combination with low-dose prednisolone enhanced the anti-inflammatory capacity of IL-4. This might offer an attractive alternative to the use of high-dose prednisolone, because it can circumvent the unwanted side effects of the drug, including steroid-induced osteoporosis.</p><p>In previous studies of murine collagen arthritis [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B25">25</xref>] it was shown that TNF-&#x003b1; is important at onset of the disease, whereas IL-1 is the dominant cytokine, not only at the onset, but also in the progression of the arthritis and the concomitant cartilage destruction. Further support for the critical role of IL-1 is provided by the absence of collagen arthritis in IL-1&#x003b2;-deficient mice, and the marked reduction of this arthritis in ICE-deficient mice as well as in normal mice treated with IL-1&#x003b2;-converting enzyme inhibitors [<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref>]. Moreover, reduced onset of arthritis was noted in TNF-receptor-deficient mice, but once a joint was afflicted the arthritis progressed to full-blown expression and cartilage destruction, again emphasizing that TNF is important in onset, but is not the dominant cytokine in progression and tissue destruction [<xref ref-type="bibr" rid="B42">42</xref>].</p><p>In recent studies, it was clearly demonstrated that onset of CIA is under stringent control of IL-4 and IL-10, because blockade of both IL-4 and IL-10 by the use of antibodies accelerated disease onset [<xref ref-type="bibr" rid="B26">26</xref>]. Furthermore, treatment of established murine CIA with low-dose IL-4 showed no suppressive effect on disease activity and joint pathology. Interestingly, combination of low-dose IL-4 and IL-10 appeared to have more potent anti-inflammatory effects, and resulted in protection against cartilage pathology [<xref ref-type="bibr" rid="B26">26</xref>]. Systemic treatment of murine CIA with high-dose IL-4 (3 &#x003bc;g/day) during the immunization stage delays onset as well as reduces severity. When IL-4 administration was terminated, however, disease expression and activity rapidly accelerated and was indistinguishable from that in the vehicle-treated control group [<xref ref-type="bibr" rid="B34">34</xref>]. Systemic IL-4 treatment of streptococcal cell wall arthritis in rats resulted in suppression of disease activity, and ameliorated the chronic destructive process leading to decreased lesions [<xref ref-type="bibr" rid="B33">33</xref>]. This was associated with enhanced levels of IL-1Ra, the natural inhibitor of IL-1, which is in accord with observations in the present study and with studies in humans systemically treated with IL-4 [<xref ref-type="bibr" rid="B43">43</xref>]. However, it is not likely that the twofold increment in serum IL-1Ra levels, found after IL-4 exposure, is sufficient to suppress CIA. As previously mentioned, blockade of IL-1 by anti-IL-1 antibodies or very high-dose IL-1Ra completely suppressed CIA and lead to full protection against joint pathology [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Whether IL-4 acts locally or systemically is at present unknown. Further experiments on biodistribution of IL-4 are needed to resolve this issue.</p><p>IL-4 levels are virtually undetectable in arthritic tissue of RA patients, suggesting that the disease is either a selective Th1 process or is not driven at all by T cells. An alternative explanation could be the fact that IL-1&#x003b1; and IL-1&#x003b2; specifically inhibit IL-4 synthesis by T cells [<xref ref-type="bibr" rid="B44">44</xref>]. Other proinflammatory cytokines, such as TNF-&#x003b1;, IL-6 and IL-12 did not decrease IL-4 production, indicating the pivotal role of IL-1 in RA. It is known that IL-4 has a suppressive effect on Th1 activity and is a crucial factor in differentiation of na&#x000ef;ve T cells into the Th2 phenotype. This suppression has been suggested to be due to the inhibitory effect of IL-4 on IL-12 generation by antigen-presenting cells and macrophages [<xref ref-type="bibr" rid="B7">7</xref>]. IL-12, on the other hand, is a potent stimulator of the generation of Th1 cells. Analysis of anticollagen type II antibodies revealed that systemic IL-4 treatment did not alter the balance of IgG_2a/IgG₁ antibodies, indicating no suppressive effect on the Th1 immune response. Total anticollagen type II antibody levels were lower in both IL-4 (1 &#x003bc;g/day) and IL-4/prednisolone treated animals when compared with the vehicle group. We have previously found that anticollagen type II antibody levels rapidly increased after onset of CIA and reached the highest levels after 7 days (Joosten LAB, unpublished data). IL-4 treatment arrested the development of high anticollagen type II antibody levels after onset and did not alter IgG_2a/IgG₁ balance.</p><p>Cartilage alterations were screened for by histology as well as COMP levels in sera of mice at the end of the experiments. COMP is a prominent component of articular cartilage. In a process affecting cartilage turnover, fragments are released and eventually reach the circulation. Thus, serum levels may be used as a marker of generalized cartilage turnover [<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>]. More recent studies [<xref ref-type="bibr" rid="B46">46</xref>,<xref ref-type="bibr" rid="B47">47</xref>] have demonstrated the production of COMP by activated synovial cells and synovial tissue of RA and osteoarthritis patients. Although the relative contribution to serum levels is not firmly established, important information has been obtained from studies of collagen arthritis in rats. Thus, increased serum COMP levels are seen at time points when erosive changes appear in cartilage, whereas in early stages with marked inflammation in the synovium no increased COMP levels are seen [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>] (Larsson E, Saxne T, unpublished data). Furthermore, serum COMP levels are reduced to normal in murine CIA after treatment with IL-1-blocking antibodies, in correspondence with a marked suppression of the cartilage lesion as viewed histologically [<xref ref-type="bibr" rid="B17">17</xref>]. Thus, evidence so far indicates that changes in serum COMP relate to changes in the cartilage turnover. In accord with these findings, low-dose IL-4/prednisolone treatment did not suppress disease activity, largely reflecting synovitis, but clearly reduced serum COMP levels. Histology interestingly revealed that serum COMP levels correlated more with cartilage erosions than with loss of matrix proteoglycans, which is a reversible process.</p><p>Recently, it was shown that expression of neo-epitope VDIPEN correlated with marked cartilage erosions during experimental arthritis. This neoepitope is formed by proteolytic cleavage of aggrecan by matrix metalloproteinases (MMPs). VDIPEN expression reflects MMP-3 (eg stromelysin) activity and it colocalized with collagen breakdown epitopes, indicating severe cartilage damage by MMPs [<xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B49">49</xref>]. It was demonstrated that IL-4 down-regulates both stromelysin and collagenase synthesis and thereby contributed to inhibition of cartilage destruction [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B50">50</xref>]. Thus, reduction of cartilage destruction found after IL-4 treatment may well be due to a lower production of MMPs and/or inhibition of their activity. The fact that IL-4 treatment did not protect against proteoglycan loss does suggest that IL-4 has no major suppressive effect on aggrecanase. In a previous study [<xref ref-type="bibr" rid="B51">51</xref>] we showed that early proteoglycan loss is mediated by aggrecanase, whereas erosive, late destruction is linked to stromelysin.</p><p>Control of bone destruction is a most challenging objective in treatment of RA. In areas of tumour-like synovial tissue, erosion of subchondral and cortical bone is common, leading to the characteristic erosions seen on radiography. Osteoclasts can be seen in the areas of bone destruction during CIA. It has been reported that IL-4 inhibits bone resorption by inhibition of osteoclast development and activity <italic>in vitro</italic> [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. Here, we report for the first time that systemic IL-4 treatment of established CIA markedly reduced bone erosions, examined by radiographic analysis and histopathology. Neither bone destruction nor osteoclasts were noted in arthritic knee joints of animals treated with high-dose IL-4, indicating decreased formation of these cells. IL-4 furthermore downregulates IL-1, IL-6, TNF-&#x003b1; and prostaglandin E₂ production in several cell types that play a role in the resorption process of the bone. Interestingly, blocking studies with neutralizing antibodies directed against IL-4 in CIA indicated that the endogenous cytokine inhibited bone destruction. In animals treated with anti-IL-4, bone destruction determined by radiographic analysis was aggravated compared with that in vehicle-treated animals (data not shown).</p><p>Glucocorticoids are potent and commonly accepted anti-inflammatory agents, but the major drawback on continued usage in arthritis is the severe negative effect on the bone. More recent studies on the mechanism of action revealed strong downregulation of macrophage production of the proinflammatory cytokines TNF-&#x003b1; and IL-1, related to enhanced I&#x003ba;B&#x003b1; synthesis. Intriguingly, over a large dose range steroids not only inhibit TNF and IL-1, but also reduce the production of IL-1Ra and regulatory cytokines such as IL-4 and IL-10 [<xref ref-type="bibr" rid="B52">52</xref>]. This suggests that the net effect in joint inflammation is impaired by the lack of the protective cytokines, which inhibit TNF/IL-1 production as well as induce potent upregulators of scavengers such as soluble receptors for TNF and IL-1, and IL-1Ra [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B53">53</xref>]. Moreover, IL-4 powerfully reduces inducible nitric oxide synthase expression, thereby counteracting the suppressive effect of IL-1 on chondrocyte proteoglycan synthesis, which is mainly nitric oxide mediated. Evidence for the latter was provided in <italic>in vitro</italic> studies with nitric oxide inhibitors. In further support of a role <italic>in vivo</italic>, we recently demonstrated that IL-1 failed to inhibit chondrocyte proteoglycan synthesis in inducible nitric oxide synthase deficient mice [<xref ref-type="bibr" rid="B54">54</xref>].</p><p>The present data clearly demonstrates the synergistic effect of combination therapy of low-dose prednisolone and IL-4. Low-dose IL-4 was without suppressive effect on clinical disease activity, which is in accord with previous studies [<xref ref-type="bibr" rid="B39">39</xref>]. However, when combined with prednisolone the progression of CIA was completely arrested. Furthermore, synergistic suppression of cartilage destruction was demonstrated by lowered serum COMP levels, which was also reflected by histology. Only combined therapy with high-dose IL-4 and prednisolone was able to suppress the influx of inflammatory cells in joint tissues and reduce the loss of matrix proteoglycans.</p><p>In conclusion, IL-4 might offer an alternative cartilage-and bone-protective therapy that is complementary to TNF/IL-1 inhibitors. Its limited effect on the inflammatory process warrants combination with other therapeutic modalities. The present data suggest that combination with prednisolone at low dosages provides an intriguing option. In accord with earlier observations of both IL-10/prednisolone and IL-4/IL-10 synergy [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B39">39</xref>], it must be considered that a cocktail of IL-4, IL-10 and low-dose glucocorticosteroids or glucocorticoids might be an even more efficacious therapy for human RA.</p></sec>

Clonal expansion is a characteristic feature of the B-cell repertoire of patients with rheumatoid arthritis

<p>The present study was designed to analyze the level of B-cell clonal diversity in patients with rheumatoid arthritis by using HCDR3 (third complementarity determining region of the rearranged heavy chain variable region gene) length as a marker. A modified immunoglobulin V_H gene fingerprinting method using either genomic DNA or complementary (c)DNA derived from B cells of the peripheral blood, synovial fluid, and tissues of several rheumatoid arthritis patients was employed. These assays permitted the detection and distinction of numerically expanded B-cell clones from activated but not numerically expanded B-cell clones. The present data suggest that B-cell clonal expansion is a common and characteristic feature of rheumatoid arthritis and that it occurs with increasing frequency from the blood to the synovial compartments, resulting in a narrowing of the clonal repertoire at the synovial level. These clonal expansions can involve resting, apparently memory B cells, as well as activated B cells. Furthermore, some of these individual expansions can persist over extended periods of time. These findings support the hypothesis that a chronic ongoing (auto)immune reaction is operative in rheumatoid arthritis and that this reaction, at least at the B-cell level, may be unique to each individual joint. A determination of the targets of these autoimmune reactions may provide valuable clues to help understand the immunopathogenesis of this disease.</p>

<contrib id="A1" contrib-type="author"><name><surname>Itoh</surname><given-names>Kenji</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Patki</surname><given-names>Varsha</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Furie</surname><given-names>Richard A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Chartash</surname><given-names>Elliot K</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Jain</surname><given-names>Rita I</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Lane</surname><given-names>Lewis</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Asnis</surname><given-names>Stanley E</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Chiorazzi</surname><given-names>Nicholas</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>nchizzi@nshs.edu</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Rheumatoid arthritis is a chronic debilitating autoimmune disease of unknown etiology. Although the disease is characterized by synovitis of the joints, tendon sheaths, and bursae, manifestations that do not involve the synovium are not infrequent [<xref ref-type="bibr" rid="B1">1</xref>]. These articular and systemic manifestations appear to be mediated by immunologic processes [<xref ref-type="bibr" rid="B2">2</xref>]. The hallmarks of the synovial abnormalities in rheumatoid arthritis are synovial lining cell proliferation, neoangiogenesis, and inflammatory cell infiltration involving the myeloid, macrophage, and lymphoid lineages [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. There has been considerable controversy regarding the relative importance of the types of cells and their products involved in the inflammatory processes of rheumatoid arthritis [<xref ref-type="bibr" rid="B3">3</xref>]. Nevertheless, it seems likely that all of these cell types participate to some degree in disease pathogenesis.</p><p>Evidence in support of T-cell involvement in rheumatoid arthritis involves the description of restricted subsets of T cells in the blood and synovial tissue that either express or lack certain surface membrane proteins or that express a limited set of antigen receptors. For example, clonal amplifications of CD8⁺ CD57⁺ T cells are frequently found in the T-cell repertoire of rheumatoid arthritis patients [<xref ref-type="bibr" rid="B4">4</xref>]. Furthermore, expanded clones of CD4⁺ CD28^- T cells exist in the blood and synovial compartments of such patients [<xref ref-type="bibr" rid="B5">5</xref>] and these T cells appear to be autoreactive [<xref ref-type="bibr" rid="B6">6</xref>]. Finally, the T-cell receptors for antigen expressed by these and other T-cell subsets frequently display a bias in favor of receptors utilizing certain V&#x003b2; genes [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>].</p><p>In contrast to the extensive studies of the clonal distribution of T cells in rheumatoid arthritis, much less is known about the level of B-cell diversity in this disease. Previous studies, however, are consistent with the interpretation that the B-cell repertoire is also restricted. For example, flow cytometric analyses of circulating B cells [<xref ref-type="bibr" rid="B13">13</xref>] suggested that oligoclonality exists, and cell culture experiments [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>] demonstrated that synovial tissue explants spontaneously secrete immunoglobulins of restricted heterogeneity as defined by immunoglobulin (Ig)G subclass, isoelectric focusing, and idiotype expression. More recent molecular analyses of the immunoglobulin genes expressed by B cells in the synovial tissue of rheumatoid arthritis patients support these notions [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>].</p><p>These findings are important because they suggest that, at the B- and T-cell levels, an ongoing immune reaction is occurring that is directed at restricted sets of (auto)antigens. The present study was designed to analyze further the level of clonal diversity in rheumatoid arthritis B cells by using the length of the third complementarity determining region (CDR3) of the rearranged heavy (H) chain variable region (V) gene as a marker (herein referred to as HCDR3). A modification of the immunoglobulin V_H gene fingerprinting method [<xref ref-type="bibr" rid="B23">23</xref>] that has been used to analyze the diversity of B cells and T cells in several clinical settings [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>] was used to address this issue. The present data suggest that B-cell clonal expansion is a common and characteristic feature of rheumatoid arthritis, that it involves both resting and activated cells, and that it can persist over extended periods of time. These findings support the idea that a chronic (auto)immune reaction is operative in rheumatoid arthritis.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patients and patient samples</title><p>Heparinized venous blood, synovial fluid, and synovial tissue were obtained from patients who fulfilled the American College of Rheumatology criteria for the diagnosis of rheumatoid arthritis [<xref ref-type="bibr" rid="B27">27</xref>]. Synovial tissue removed at the time of either joint replacement or therapeutic synovectomy was digested with collagenase, DNAse, and hyaluronidase to obtain single-cell suspensions. Mononuclear cells (MNCs) were isolated from cell suspensions from blood, synovial fluid, and synovial tissue by density gradient centrifugation (Ficoll-Paque; Pharmacia LKB Biotechnology, Piscataway, NJ, USA).</p></sec><sec><title>Isolation of DNA and RNA, and preparation of complementary DNA</title><p>Genomic DNA was isolated from MNCs using the Puregene DNA Isolation Kit (Gentra Systems, Mineapolis, MN, USA) and total RNA was isolated using Ultraspec RNA (Biotech Laboratories, Houston, TX, USA). Both of these reagents were used according to the manufacturer's instructions. One microgram of RNA was reverse transcribed to complementary (c)DNA using 200U Moloney murine leukemia virua (M-MLV) reverse transcriptase (GIBCO BRL Life Technologies, Grand Island, NY, USA), 1U RNAse inhibitor (5 Prime 3 Prime, Boulder, CO, USA) and 20 pmol oligo dT primer in a total volume of 20 &#x003bc;l. These ingredients were incubated at 42&#x000b0;C for 1 h, heated to 65&#x000b0;C for 10 min to stop the reactions, and then diluted to a final volume of 100 &#x003bc;l.</p></sec><sec><title>Polymerase chain reaction conditions for immunoglobulin V_H gene fingerprinting assay</title><p>The original immunoglobulin V_H gene fingerprinting assay [<xref ref-type="bibr" rid="B23">23</xref>] was modified into two stages, starting with either genomic DNA or cDNA as templates (Fig. <xref ref-type="fig" rid="F1">1</xref>). The sequences of the primers used in these reactions were published previously [<xref ref-type="bibr" rid="B28">28</xref>].</p><sec><title>Stage I</title><p>Genomic DNA (100ng) was amplified using a sense V_H family-specific framework region (FR)1 primer in conjunction with an antisense J_H consensus primer. These reactions were carried out in 50 &#x003bc; l using 5 pmol of each primer, and were cycled with a 9600 GeneAmp System (Perkin Elmer, Emeryville, CA, USA) as follows: denaturation at 94&#x000b0;C for 40 s; annealing at 65&#x000b0;C for 45 s; and extension at 72&#x000b0;C for 40 s. After 35 cycles, extension was continued at 72&#x000b0;C for an additional 10 min.</p><p>cDNA (2&#x003bc;l) was amplified using a sense V_H family-specific FR1 primer in conjunction with the appropriate antisense C_H primer. The reactions were carried out in 50 &#x003bc;l using 5 pmol of each primer and cycled as follows: denaturation at 94&#x000b0;C for 45 s; annealing at 65&#x000b0;C for 45 s; and extension at 72&#x000b0;C for 45 s. After 35 cycles, extension was continued at 72&#x000b0;C for an additional 10 min.</p></sec><sec><title>Stage II</title><p>Polymerase chain reaction (PCR) products (2 &#x003bc; l) generated from either genomic DNA or cDNA were amplified using 5 pmol of nested sense V_H family-specific FR3 primer and radiolabeled antisense nested J_H consensus primer that had been end-labeled with &#x003b3;³²-P (New England Nuclear, Beverly, MA, USA) using T4 polynucleotide kinase (Promega, Madison, WI, USA). The reactions were carried out in 25 &#x003bc; l and cycled as follows: denaturation at 94&#x000b0;C for 30 s; annealing at 52&#x000b0;C for 45 s; and extension at 72&#x000b0;C for 30 s. After 15 cycles, extension was continued at 72&#x000b0;C for an additional 10 min. The radiolabeled PCR products that reflected the HCDR3 lengths of various B-cell clones in the cell suspension were electrophoresed through a 6% denaturing acrylamide sequencing gel for approximately 1.5 h. The gel was then dried and exposed to film overnight.</p></sec></sec><sec><title>DNA cloning and sequencing</title><p>DNA sequences were determined by reamplifying the original genomic DNA using the appropriate family-specific V_H leader and J_H consensus primers under the following PCR conditions: denaturation at 94&#x000b0;C for 45s; annealing at 62&#x000b0;C for 30s; and extension at 72&#x000b0;C for 45s. After 35 cycles, extension was continued at 72&#x000b0;C for an additional 10min. PCR products were then cloned into TA vector (Invitrogen, San Diego, CA, USA), processed using Wizard minipreps (Promega), and sequenced using M13 forward and reverse primers, a DNA Sequencing Kit (Perkin Elmer) and an automated sequenator (Applied Biosystems, Foster City, CA, USA).</p></sec></sec><sec><title>Results and discussion</title><sec><title>Identification of B-cell clonal expansions using a modified immunoglobulin V_H gene fingerprinting assay</title><p>During normal B cell development, the processes of gene segment recombination and coding end processing yield nucleotide HCDR3 lengths that are characteristic and virtually invariant for an individual B-cell clone. Therefore, these lengths can be used as signatures to identify members of a B cell clone. The immunoglobulin V_H gene fingerprinting approach [<xref ref-type="bibr" rid="B23">23</xref>] takes advantage of the wide range of HCDR3 lengths that can occur in human B cells (approximately 5-35 amino acids) to provide an estimate of clonal diversity in polyclonal populations. When polyclonal B lymphocytes from adults are analyzed using this assay, they display a Gaussian HCDR3 length distribution around a mean of approximately 15 amino acids. The presence of an individual dominant length that differs from this Gaussian distribution can be used as an indication of a specific B-cell clonal expansion.</p><p>Figure <xref ref-type="fig" rid="F1">1</xref> illustrates schematically the two-stage immunoglobulin V_H gene fingerprinting approach that we utilized. Note that when cDNA prepared from normal peripheral blood B cells is used as a template for these V_H family-specific and C_H-specific assays, ladders of HCDR3 lengths that differ by three nucleotides are identified. These individual HCDR3 lengths are signatures of the various individual B-cell clones contained within the polyclonal population. The intensities of the bands in virtually all of the ladders illustrated in Figure <xref ref-type="fig" rid="F1">1</xref> are relatively uniformly distributed around the mean. This indicates that there are no dominant HCDR3 lengths that skew the Gaussian distribution, and therefore that there are no significant clonal expansions among the B cells that express most of these V_H&#x02013;C_H combinations. Similar results are obtained using the genomic DNA-based assay, although these results cannot be interpreted in a C_H-specific manner (data not shown).</p><p>In the V_H6-IgG combination (Fig. <xref ref-type="fig" rid="F1">1</xref>), however, a non-Gaussian distribution is noted, even in this normal individual. This could be a reflection of the numbers of V_H genes present in the V_H family being analyzed (it is more likely to see a non-Gaussian distribution in families with small numbers of individual genes) or of the state of activation of a specific clone (because activated B cells contain much higher levels of V gene messenger RNA than resting B cells). We believe that in the instance illustrated in Figure <xref ref-type="fig" rid="F1">1</xref> the latter possibility is more likely, because the small V_H2 and V_H5 families (only two gene members per family) do not exhibit the same degree of oligoclonality as that observed with the only somewhat smaller V_H6 family (one gene member).</p></sec><sec><title>Distinction between clonal expansion and clonal activation using the modified immunoglobulin V_H gene fingerprinting assay</title><p>Because B-cell activation and differentiation result in dramatic increases in immunoglobulin V gene messenger RNA, these fingerprinting assays cannot readily distinguish between clonal expansion and activation when cDNA is used as a starting template. Because DNA levels are not appreciably altered by cellular activation, however, the use of genomic DNA as well as cDNA from the same sample of B cells helps to distinguish these two processes.</p><p>Thus, in the setting of specific B-cell clonal expansion without concomitant cellular activation, the DNA-based fingerprinting assay will indicate a dominant HCDR3 length, whereas the cDNA-based assay may not (data not shown). Conversely, in the setting of specific B-cell clonal activation without concomitant clonal expansion, the cDNA-based assay will indicate a dominant HCDR3 length, whereas the DNA-based assay may not. Finally, in the setting of specific B-cell clonal activation with concomitant clonal expansion, both the cDNA- and the DNA-based assays will indicate a dominant HCDR3 length.</p><p>These distinctions were very reproducible in the following studies. There were no situations in which evidence for cellular activation (either selective or accompanied by clonal expansion) was present in one set of analyses and not in a subsequent set using the same starting materials.</p></sec><sec><title>B cells in the blood, synovial fluid, and synovial tissue of rheumatoid arthritis patients exhibit clonal expansions of activated and resting B cells</title><p>We analyzed the peripheral blood, synovial fluid, and synovial tissue B cells of rheumatoid arthritis patients (<italic>n</italic> = 20, 10, and 5, respectively) using the genomic DNA- and cDNA-based fingerprinting assays to develop an understanding of the diversity of the B cells in these compartments. Figures <xref ref-type="fig" rid="F2">2</xref> and <xref ref-type="fig" rid="F3">3</xref> are illustrations of representative patients for whom concomitant blood and synovial fluid or blood and synovial tissue samples were available. In order to simplify the Figures, only the results for two large V_H families (V_H1 and V_H3) and two small V_H families (V_H5 and V_H6) are provided, although assays for each V_H family and each major CH family (&#x003bc;,&#x003b3;, and &#x003b1;) were performed and revealed similar findings.</p><p>The genomic DNA-based assays in both patients indicated that clonal expansions are common in the blood of rheumatoid arthritis patients. This type of result was obtained with all individuals tested. It was most convincingly demonstrated by the results in Figure <xref ref-type="fig" rid="F3">3</xref> obtained from B cells expressing genes of the V_H1 and V_H3 families. Because these V_H families contain the largest numbers of V_H genes, they would be more likely to display a polyclonal pattern.</p><p>An even more striking level of B-cell clonal dominance and expansion was seen when the genomic DNA-based assay was used to analyze B cells from the synovial fluid or synovial tissue (Figs <xref ref-type="fig" rid="F2">2</xref> and <xref ref-type="fig" rid="F3">3</xref>). In these analyses, virtually all V_H families demonstrated extensive B-cell oligoclonality. It should be pointed out that when an individual HCDR3 length comprises more than 50% of the radioactive counts of a V_H&#x02013;C_H ladder, clonality, based on DNA sequencing, is very likely; when an individual length comprises more than 70% of the radioactivity, clonality is virtually assured (data not shown).</p><p>Collectively, these data indicate that the B-cell repertoire of rheumatoid arthritis patients is skewed away from the typical, apparently random representation of normal individuals. The reason for this discrepancy is not clear, although one possibility is that restricted antigenic exposure alters the composition of the repertoire in favor of B cells reactive with the putative antigen(s). If this is so, the progressive narrowing of the repertoire from the blood to the synovial tissue is consistent with the ideas that the antigenic exposures are originating at these sites and that the synovial compartment is supporting clonal amplification. Because these are true clonal expansions (ie increased numbers of B cells per specific clone), it is likely that the antigenic exposures are chronic and therefore are increasing the numbers of memory B cells reactive with these determinants.</p><p>In order to confirm that these clonally expanded B cells were receiving ongoing antigenic stimulation and not limited solely to the memory compartment, we employed the cDNA-based assay to distinguish clonal expansions of activated B cells from resting (memory) cells. As illustrated in Figure <xref ref-type="fig" rid="F2">2</xref>, activated B-cell clones (identified by the letter 'A' in Fig. <xref ref-type="fig" rid="F2">2</xref>) expressing each of the immunoglobulin heavy-chain isotypes were easily identified in all the V_H families studied. In some instances, these activated clones were also expanded numerically (as defined by the genomic DNA-based assays, and identified by the letter 'E' in Fig. <xref ref-type="fig" rid="F2">2</xref>). In other cases, these activated clones did not appear to be numerically expanded. Similar examples can be found in Figures <xref ref-type="fig" rid="F3">3</xref> and <xref ref-type="fig" rid="F4">4</xref>, but they are not identified by letters in order to simplify the Figures. </p><p>Thus, it appears that many discrete B-cell clones exist in the synovial compartment of rheumatoid arthritis patients, and that these are increased in number, consistent with a response to a restricted antigenic challenge(s). Furthermore, these B-cell clones appear to be of both the resting (memory) and the activated types, suggesting that these antigenic challenges are chronic and ongoing. Thus, these data support and extend previous findings [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>] by indicating that specific B-cell clonal amplifications can occur both in previously stimulated B cells and in currently activated B cells. The presence of activated B cells that are not increased in number is consistent either with recent <italic>in situ</italic> synovium-specific stimulation of these B-cell clones, or with the influx of activated B cells that were stimulated by antigens outside of and not necessarily relevant to the synovial compartment.</p></sec><sec><title>B-cell clonal expansions in the blood and synovial compartments can be restricted to one or another compartment, or be common to the two</title><p>Because the preceding data indicated that the B-cell repertoire of rheumatoid arthritis patients contains expanded clones of B cells that can be resting or activated, we investigated whether the same clones could be identified in both the blood and synovial compartments. Figures <xref ref-type="fig" rid="F2">2</xref> and <xref ref-type="fig" rid="F3">3</xref> illustrate data that suggest that there are expanded clones that are blood restricted, joint restricted, or are common to both compartments. Examples of blood-restricted clones are highlighted on Figures <xref ref-type="fig" rid="F2">2</xref> and <xref ref-type="fig" rid="F3">3</xref> with the &#x025b8; symbol, those that are joint-restricted with the &#x025b9; symbol, and those that are common to the two compartments with the &#x02666; symbol. These findings suggest that there may be a degree of cellular trafficking between the blood and the synovial tissues.</p><p>In order to address this issue, we studied the B cells of the blood and two synovial sites (right and left hip) that were obtained from the same patient within 3 h of each other (Fig. <xref ref-type="fig" rid="F4">4</xref>). In this patient, the DNA-based assay provided examples of clonal expansions that were present in only one joint (eg the V_H3&#x02013;J_H and V_H5&#x02013;J_H combinations in Figure <xref ref-type="fig" rid="F4">4</xref>), and the companion cDNA-based assays indicated that in some instances these expansions were either activated or resting. In only rare instances, however, did the data suggest that a similar clone was present in two different synovial tissues. DNA sequence analyses confirmed the rarity of this event (data not shown). Thus it appears that in most instances the clonal amplifications occur <italic>in situ</italic> and are not the result of trafficking from one anatomic site to another. If so, this would suggest that the antigenic challenges driving these clonal expansions may not be common to all synovial tissues, but may be generated independently at each site, possibly by ongoing tissue breakdown.</p></sec><sec><title>Clonal persistence in the synovial fluid compartment</title><p>If the clonal expansions identified in the joints of rheumatoid arthritis patients are due to an ongoing response to antigen, then one would predict that at least some of the clones would persist over time. To test this, we studied the synovial fluid B cells from the same joints of three patients on two occasions spanning several months. Most of the clonal expansions detected on the initial samples were not present in the subsequent samples. In a few instances, however, B-cell clonal persistence was found.</p><p>Figure <xref ref-type="fig" rid="F5">5</xref> illustrates the best example of this phenomenon in a patient who was studied over a 4-month interval. The DNA-based assay using V_H4 family-specific primers indicated the presence of two similar clones on days 0 and 120, whereas the other V_H4-expressing clones detected at the first analysis were no longer present at the time of the second analysis. DNA sequence analyses of one of these two clones confirmed their identity, because each displayed the same rearranged V_HDJ_H gene with identical V_H mutations and identical HCDR3 sequences (data not shown). Therefore, certain clones can persist locally over time, suggesting that a common and persistent antigenic stimulation was operable in the joint of this rheumatoid arthritis patient.</p><p>The lack of persistence of the other B-cell clones suggests two possibilities. First, the initial set of B-cell clones might have been replaced by others that recognized and responded to different antigenic epitopes on the same original immunogenic protein. This type of clonal evolution to the recognition of different epitopes on the same immunogenic moiety is common in experimental situations in which repetitive immunizations with a defined antigen are delivered [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>]. The other possibility is that the B cells that disappeared over time were not reactive with tissue antigens. These could have been stimulated by irrelevant antigens in the periphery and therefore, after entering the synovial compartment, could not be restimulated and hence could not enter the memory pool and take up residence in the synovial tissue.</p></sec></sec><sec><title>Conclusion</title><p>The present data indicate that clonal expansion is a common occurrence in the B-cell repertoire of rheumatoid arthritis patients. These expansions involve both resting memory B cells and activated B cells, some of which are derived from the memory B-cell compartment. Because the extent of these clonal expansions increases from the blood to the synovial compartment, this progressive narrowing in diversity implies that antigens located in the synovia are responsible for these antigen-receptor biases. In support of this hypothesis are the observations that some of these clonal expansions are joint specific. Because identical clones are rarely found in two different joints, however, these immune reactions are probably unique to each individual joint. Furthermore, because it is unlikely that each joint would harbor a different foreign antigen, these B cells are most likely reacting with autoantigens generated locally, possibly by local tissue breakdown. </p><p>Recent studies [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B32">32</xref>] have demonstrated that the synovial tissue of rheumatoid arthritis patients can develop lymphoid aggregates that have the cellular components of an ectopic germinal center and that can sustain B-cell clonal expansion and diversification. It is likely that the B cells that mature in these 'pseudogerminal centers' and those that we have identified in the present studies are responding to specific (auto)antigens. Therefore, the identification of the antigenic reactivities of these B cells, and in particular those B cells within the memory compartment that have presumably traversed the pseudogerminal centers and undergone (auto)antigen and T cell selection and rescue, may provide important clues to the role of B lymphocytes and their immunoglobulin molecules in the immunopathogenesis of rheumatoid arthritis.</p></sec>

Activation of synovial fibroblasts in rheumatoid arthritis: lack of expression of the tumour suppressor PTEN at sites of invasive growth and destruction

<sec><title>Aims:</title><p>PTEN is a novel tumour suppressor which exhibits tyrosine phosphatase activity as well as homology to the cytoskeletal proteins tensin and auxilin. Mutations of PTEN have been described in several human cancers and associated with their invasiveness and metastatic properties. Although not malignant, rheumatoid arthritis synovial fibroblasts (RA-SF) exhibit certain tumour-like features such as attachment to cartilage and invasive growth. In the present study, we analyzed whether mutant transcripts of PTEN were present in RA-SF. In addition, we used <italic>in situ</italic> hybridization to study the expression of PTEN messenger (m)RNA in tissue samples of RA and normal individuals as well as in cultured RA-SF and in the severe combined immunodeficiency (SCID) mouse model of RA.</p></sec><sec><title>Methods:</title><p>Synovial tissue specimens were obtained from seven patients with RA and from two nonarthritic individuals. Total RNA was isolated from synovial fibroblasts and after first strand complementary (c)DNA synthesis, polymerase chain reaction (PCR) was performed to amplify a 1063 base pair PTEN fragment that encompassed the coding sequence of PTEN including the phosphatase domain and all mutation sites described so far. The PCR products were subcloned in <italic>Escherichia coli</italic>, and up to four clones were picked from each plate for automated sequencing. For <italic>in situ</italic> hybridization, digoxigenin-labelled PTEN-specific RNA probes were generated by <italic>in vitro</italic> transcription. For control <italic>in situ</italic> hybridization, a matrix metalloproteinase (MMP)-2-specific probe was prepared. To investigate the expression of PTEN in the absence of human macrophage or lymphocyte derived factors, we implanted RA-SF from three patients together with normal human cartilage under the renal capsule of SCID mice. After 60 days, mice were sacrificed, the implants removed and embedded into paraffin.</p></sec><sec><title>Results:</title><p>PCR revealed the presence of the expected 1063 base pair PTEN fragment in all (9/9) cell cultures (Fig. <xref ref-type="fig" rid="F1">1</xref>). No additional bands that could account for mutant PTEN variants were detected. Sequence analysis revealed 100% homology of all RA-derived PTEN fragments to those from normal SF as well as to the published GenBank sequence (accession number U93051). However, <italic>in situ</italic> hybridization demonstrated considerable differences in the expression of PTEN mRNA within the lining and the sublining layers of RA synovial membranes. As shown in Figure <xref ref-type="fig" rid="F2">2a</xref>, no staining was observed within the lining layer which has been demonstrated to mediate degradation of cartilage and bone in RA. In contrast, abundant expression of PTEN mRNA was found in the sublining of all RA synovial tissues (Figs <xref ref-type="fig" rid="F2">2a</xref> and <xref ref-type="fig" rid="F2">b</xref>). Normal synovial specimens showed homogeneous staining for PTEN within the thin synovial membrane (Fig. <xref ref-type="fig" rid="F2">2c</xref>). <italic>In situ</italic> hybridization using the sense probe gave no specific staining (Fig. <xref ref-type="fig" rid="F2">2d</xref>). We also performed <italic>in situ</italic> hybridization on four of the seven cultured RA-SF and followed one cell line from the first to the sixth passage. Interestingly, only 40% of cultured RA-SF expressed PTEN mRNA (Fig. <xref ref-type="fig" rid="F3">3a</xref>), and the proportion of PTEN expressing cells did not change throughout the passages. In contrast, control experiments using a specific RNA probe for MMP-2 revealed mRNA expression by nearly all cultured cells (Fig. <xref ref-type="fig" rid="F3">3b</xref>). As seen before, implantation of RA-SF into the SCID mice showed considerable cartilage degradation. Interestingly, only negligible PTEN expression was found in those RA-SF aggressively invading the cartilage (Fig. <xref ref-type="fig" rid="F3">3c</xref>). <italic>In situ</italic> hybridization for MMP-2 showed abundant staining in these cells (Fig. <xref ref-type="fig" rid="F3">3d</xref>).</p></sec><sec><title>Discussion:</title><p>Although this study found no evidence for mutations of PTEN in RA synovium, the observation that PTEN expression is lacking in the lining layer of RA synovium as well as in more than half of cultured RA-SF is of interest. It suggests that loss of PTEN function may not exclusively be caused by genetic alterations, yet at the same time links the low expression of PTEN to a phenotype of cells that have been shown to invade cartilage aggressively.</p><p>It has been proposed that the tyrosine phosphatase activity of PTEN is responsible for its tumour suppressor activity by counteracting the actions of protein tyrosine kinases. As some studies have demonstrated an upregulation of tyrosine kinase activity in RA synovial cells, it might be speculated that the lack of PTEN expression in aggressive RA-SF contributes to the imbalance of tyrosine kinases and phosphatases in this disease. However, the extensive amino-terminal homology of the predicted protein to the cytoskeletal proteins tensin and auxilin suggests a complex regulatory function involving cellular adhesion molecules and phosphatase-mediated signalling. The tyrosine phosphatase TEP1 has been shown to be identical to the protein encoded by PTEN, and gene transcription of TEP1 has been demonstrated to be downregulated by transforming growth factor (TGF)-&#x003b2;. Therefore, it could be hypothesized that TGF-&#x003b2; might be responsible for the downregulation of PTEN. However, the expression of TGF-&#x003b2; is not restricted to the lining but found throughout the synovial tissue in RA. Moreover, in our study the percentage of PTEN expressing RA-SF remained stable for six passages in culture, whereas molecules that are cytokine-regulated <italic>in vivo</italic> frequently change their expression levels when cultured over several passages. Also, cultured RA-SF that were implanted into SCID mice and deeply invaded the cartilage did not show significant expression of PTEN after 60 days. The drop in the percentage of PTEN expressing cells from the original cell cultures to the SCID mouse implants is of interest as this observation goes along with data from previous studies that have shown the prominent expression of activation-related molecules in the SCID mice implants that <italic>in vivo</italic> are found predominantly in the lining layer. Therefore, our data point to endogenous mechanisms rather than to the influence of exogenous human cytokines or factors in the downregulation of PTEN. Low expression of PTEN may belong to the features that distinguish between the activated phenotype of RA-SF and the sublining, proliferating but nondestructive cells.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Pap</surname><given-names>Thomas</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Franz</surname><given-names>Juliane K</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Hummel</surname><given-names>Klaus M</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Jeisy</surname><given-names>Elvira</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Gay</surname><given-names>Renate</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Gay</surname><given-names>Steffen</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>ruzgay@ruz.unizh.ch</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>PTEN is a novel tumour suppressor that exhibits tyrosine phosphatase activity as well as homology to the cytoskeletal proteins tensin and auxilin [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Mutations in <italic>PTEN</italic> have been described in several human cancers, and have been associated with the invasiveness and metastatic properties of malignancies [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Although not malignant, rheumatoid arthritis (RA) synovial fibroblasts (SF) are imbued with certain tumour-like features such as attachment to cartilage and invasive growth [<xref ref-type="bibr" rid="B4">4</xref>]. Moreover, it has been suggested that tyrosine kinase activity, which counteracts the action of tyrosine phosphatases, is increased in RA [<xref ref-type="bibr" rid="B5">5</xref>] and may be involved in the activation of mitogen-activated protein kinase in human synovial cells [<xref ref-type="bibr" rid="B6">6</xref>].</p><p>In the present study, we analyzed whether mutant transcripts of <italic>PTEN</italic> were present in RA-SF. In addition, we studied the expression of PTEN messenger (m)RNA in tissue samples from seven RA patients and two normal individuals, as well as in cultured RA-SF and in the severe combined immunodeficiency (SCID) mouse co-implantation model of RA. Aggressively invading RA-SF expressed only low levels of PTEN, which showed no evidence for mutations. This lack of expression was maintained in cultured RA-SF over several passages, and when RA-SF were implanted into SCID mice together with normal human cartilage.</p></sec><sec sec-type="methods"><title>Methods</title><sec><title>Tissue preparation and cell cultures</title><p>Synovial tissue specimens were obtained from seven patients with RA undergoing synovectomy or joint replacement and from two nonarthritic individuals. Immediately after surgery, one part of the tissue was embedded in Tissue-Tek OCT medium (Miles, Elkhart, IN, USA), snap-frozen and stored at -80&#x000b0;C, and a second was fixed in 4% buffered formalin for 6 h before embedding in paraffin. Another portion was digested enzymatically and the released cells were grown in Dulbecco's modified Eagle medium with 10% foetal calf serum [<xref ref-type="bibr" rid="B7">7</xref>]. At confluence, cells were harvested and half of them were used for complementary (c)DNA preparation. The remaining cells were used to maintain the culture, as well as for growing cells on chamber slides (Lab-Tek II; Nalge Nunc Int, Naperville, IL, USA) 48 h before <italic>in situ</italic> hybridization.</p></sec><sec><title>RNA isolation and reverse transcription polymerase chain reaction</title><p>Total RNA was isolated from cultured SA applying the TRIzol RNA isolation kit (Life Technologies, Basel, Switzerland) according to the manufacturer's protocol. After first strand cDNA synthesis using oligo-d(T)_12-18 primers and Moloney murine leukemia virus (M-MuLV) reverse transcriptase (Boehringer-Mannheim, La Jolla, California, USA), a 1063 base pair <italic>PTEN</italic> cDNA fragment was amplified using polymerase chain reaction (PCR) with <italic>Pyrococcus furiosus</italic> (Pfu) DNA-polymerase (Stratagene, La Jolla, California, USA). This fragment encompassed the coding sequence of <italic>PTEN</italic> including the phosphatase domain and all mutation sites described so far [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. The primer sequences were as follows: upper primer 5';-GAC AGC CAT CAT CAA AGA GA-3'; and lower primer, 5'-TGA CGG CTC CTC TAC TGT T-3'. The amplification was carried out for 32 cycles under annealing-extension conditions of 52&#x000b0;C for 1 min and 72&#x000b0;C for 2 min using a Perkin-Elmer (Foster City, California, USA) DNA-Thermocycler 480. To look for additional, low copy transcripts, the cycle number was increased stepwise up to 42 cycles and the annealing temperature was decreased to 48&#x000b0;C.</p></sec><sec><title>Cloning and sequencing of the <italic>PTEN</italic> fragments</title><p>The PCR products were then ligated into the PCR-Script Amp SK+ vector (Stratagene), and transformation of the vector into Epicurian Coli XL1-Blue MRF' Kan supercompetent cells was performed. After selection, up to four clones were picked from each plate and plasmid preparation of the <italic>PTEN</italic>-insert containing plasmids was performed using the Qiagen MiniPrep Kit (Qiagen, Basel, Switzerland). The sequences of the inserts were determined using automated, dideoxy sequencing.</p></sec><sec><title>Riboprobe preparation</title><p>The Qiagen MaxiPrep Kit was used for large scale preparation of PTEN-insert containing plasmids from two successfully transfected clones, and templates were prepared by linearization with BamH I or Not I (Life Technologies). Again, plasmid sequence was checked by automated sequencing, which confirmed the 100% identity of the <italic>PTEN</italic> fragment to the published GenBank sequence (accession number 193051). Antisense and sense RNA probes were then obtained by <italic>in vitro</italic> transcription using T3 and T7 RNA polymerase (Boehringer-Mannheim) with a commercially available transcription kit (Stratagene). For <italic>in situ</italic> hybridization, probes were labelled with digoxigenin-UTP (Boehringer-Mannheim). The RNA probe for control <italic>in situ</italic> hybridization to detect matrix metalloproteinase (MMP)-2 was prepared accordingly using a plasmid obtained from the American Type Culture Collection (ATCC, Rockville, Maryland, USA; ATCC number 79066).</p></sec><sec><title><italic>In situ</italic> hybridization</title><p><italic>In situ</italic> hybridization was performed as described by Kriegsmann <italic>et al</italic> [<xref ref-type="bibr" rid="B11">11</xref>]. Briefly, after fixation, tissue sections were hybridized with the digoxigenin-labelled riboprobes (either antisense or sense) in hybridization buffer containing 50% formamide, 1&#x000d7; Denhardt's solution, 10% dextran sulphate, 25 &#x003bc; g/ml herring sperm DNA (Boehringer-Mannheim), 40 mg/ml yeast transfer RNA (Sigma Chemical Co, St Louis, Missouri, USA) for 16 h at 52&#x000b0;C. After hybridization, unbound probe was digested at 37&#x000b0;C for 45 min with 10 &#x003bc; g/ml RNase A (Boehringer-Mannheim), and consecutive washing steps were performed at 50&#x000b0;C at the following stringencies: 50% formamide/2 &#x000d7; SSC (5 min); 1 &#x000d7; SSC + 1% sodium dodecyl sulphate (SDS; 15 min); 0.25 &#x000d7; SSC + 1% SDS (15 min); and 0.1% SSC + 1% SDS (15 min). Immunological detection was performed after blocking nonspecific binding sites with 2% horse serum (30 min at room temperature) by incubation with alkaline phosphatase-conjugated antidigoxigenin Fab fragments (Boehringer-Mannheim) for 1 h at room temperature, diluted 1/500 in Tris-NaCl, pH 7.6, containing 1% normal horse serum. After washing with Tris-NaCl (pH 7.6) and Tris-NaCl/MgCl₂ (pH 9.5), the sections were incubated with 5-bromo-4-chloro-3-indolyl-phosphate/4-nitro blue tetrazolium chloride colour substrate solution (Boehringer Mannheim) containing 1 mmol/l levamisole (DAKO, Zug, Switzerland), and developed at room temperature in darkness. Colour development was stopped with Tris-NaCl (pH 7.6).</p></sec><sec><title>Severe combined immunodeficiency mouse co-implantation experiments</title><p>SCID mice were obtained from the Charles Rivers GmbH (Sulzfeld, Germany) and kept permanently in sterile conditions. Implantation of RA-SF together with normal human cartilage was performed as described previously [<xref ref-type="bibr" rid="B7">7</xref>]. RA-SF from three different patients were used for the SCID mouse experiments. Briefly, after trypsinization, washing and centrifugation, 10⁵ cells were resuspended in 100 &#x003bc;l sterile culture medium and inserted into the cavity of an inert sponge (Gelfoam, Pharmacia &#x00026; Upjohn, D&#x000fc;bendorf, Germany) together with an 1 mm³ piece of normal human articular cartilage. Mice were anaesthetized intraperitoneally with 0.014 mg/g Xylocain (Lidocain hydrochloride; Astra Pharmaceutica, Dieticon, Switzerland) and 0.09 mg/g Ketalar (Ketamin hydrochloride; Parke-Davis, Baar, Switzerland) in an isotonic solution, and a 1 cm incision was made on the left flank of the animals. The left kidney was exteriorized and, once a small incision was made, an implant was placed under the renal capsule. The peritoneal layer and the skin were closed using 5.0 prolene suture material. After 60 days, mice were sacrificed and the implants removed. Tissue preparation included fixation in 4% buffered formalin and paraffin embedding according to standard procedures.</p></sec></sec><sec><title>Results</title><p>Using the specific primers, the expected 1063 base pair <italic>PTEN</italic> fragment was amplified from the total cDNA of all (all of nine) cell cultures by PCR. Moreover, no additional bands that could account for mutant <italic>PTEN</italic> variants were detected, even when PCR conditions were changed towards lower specificity (Fig. <xref ref-type="fig" rid="F1">1</xref>). PCR products were then subcloned into <italic>Escherichia coli</italic>, and up to four successfully transformed clones were picked for plasmid preparation from each culture plate (total number of samples 21). Sequence analysis revealed 100% homology of all RA-derived <italic>PTEN</italic> fragments to those obtained from normal SF as well as to the published GenBank sequence (accession number U93051).</p><p><italic>In situ</italic> hybridization with digoxigenin-labelled RNA probes, however, demonstrated considerable differences in the expression of PTEN mRNA within the lining and the sublining layers of RA synovial membranes. As shown in Figure <xref ref-type="fig" rid="F2">2a</xref>, no staining was observed within the lining layer, which has been demonstrated to mediate degradation of cartilage and bone in RA [<xref ref-type="bibr" rid="B4">4</xref>]. In contrast, abundant expression of PTEN mRNA was found in the sublining layer of all RA synovial tissues (Figs <xref ref-type="fig" rid="F2">2a</xref> and <xref ref-type="fig" rid="F2">b</xref>). Normal synovial specimens showed homogeneous staining for PTEN within the thin synovial membrane (Fig. <xref ref-type="fig" rid="F2">2c</xref>). Expression of PTEN mRNA was seen in the most superficial layer of normal synovium as well as in deeper regions, with most cells being of fibroblast shape (Fig. <xref ref-type="fig" rid="F2">2c</xref>). <italic>In situ</italic> hybridization using the sense probe gave no specific staining (Fig. <xref ref-type="fig" rid="F2">2d</xref>). We also performed <italic>in situ</italic> hybridization on four of the seven cultured RA-SF and followed one cell line from the first to the sixth passage. Interestingly, only 40% of cultured RA-SF expressed PTEN mRNA (Fig. <xref ref-type="fig" rid="F3">3a</xref>), and the proportion of PTEN-expressing cells did not change significantly throughout the passages. In contrast, control experiments using a specific RNA probe for MMP-2 revealed mRNA expression by nearly all cultured cells (Fig. <xref ref-type="fig" rid="F3">3b</xref>) indicating constitutive expression of MMP-2 but not of PTEN in the absence of macrophages and lymphocytes, and their locally derived factors.</p><p>To test the hypothesis further that PTEN downregulation in RA-SF is not caused by such exogenous factors, we co-implanted RA-SF from three patients together with normal human cartilage under the renal capsule of SCID mice and maintained the implants for 60 days as described previously [<xref ref-type="bibr" rid="B7">7</xref>]. Before implantation, RA-SF showed the above described expression pattern for PTEN. Histological evaluation of the implants after the rats were killed revealed considerable cartilage degradation by the RA-SF [<xref ref-type="bibr" rid="B7">7</xref>]. Interestingly, by <italic>in situ</italic> hybridization with PTEN-specific RNA probes, only negligible PTEN expression was found in those RA-SF aggressively invading the cartilage (Fig. <xref ref-type="fig" rid="F3">3c</xref>). Again, control <italic>in situ</italic> hybridization with RNA probes for MMP-2 showed abundant staining in these cells (Fig. <xref ref-type="fig" rid="F3">3d</xref>).</p></sec><sec><title>Discussion</title><p>Although the present study found no evidence for mutations of <italic>PTEN</italic> in RA synovium, the observation that PTEN expression is lacking in the lining layer of RA synovium as well as in more than half of cultured RA-SF is of interest. This suggests that loss of PTEN function may not exclusively be caused by genetic alterations, but that it links the low expression of PTEN to a phenotype of cells that have been shown to invade cartilage aggressively. Moreover, the present results may also have an impact on further investigations of other tumour suppressors such as p53, which has been found to be genetically altered simultaneously with <italic>PTEN</italic> in several cancers [<xref ref-type="bibr" rid="B10">10</xref>], and somatic mutations of which have also been described in RA [<xref ref-type="bibr" rid="B12">12</xref>]. On the basis of the inconsistency and great variability of p53 mutations in RA, it has been proposed that these mutations, although contributing to the invasive behaviour of rheumatoid tissue, may occur secondary to other changes and may not represent the primary step in the activation of RA-SF [<xref ref-type="bibr" rid="B12">12</xref>]. These data, together with our observations of PTEN downregulation in nonmalignant but aggressively invading RA-SF, suggest that the lack of PTEN expression may be specifically associated with certain features of malignant cells.</p><p>It has been proposed that the tyrosine phosphatase activity of PTEN is responsible for its tumour suppressor activity [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B13">13</xref>] by counteracting the actions of protein tyrosine kinases. Because some studies have demonstrated an upregulation of tyrosine kinase activity in RA synovial cells, it might be speculated that the lack of PTEN expression in aggressive RA-SF contributes to the imbalance of tyrosine kinases and phosphatases in this disease [<xref ref-type="bibr" rid="B5">5</xref>]. The extensive amino-terminal homology of the predicted protein to the cytoskeletal proteins tensin and auxilin, however, suggests a complex regulatory function involving cellular adhesion molecules and phosphatase-mediated signalling [<xref ref-type="bibr" rid="B9">9</xref>].</p><p>Tamura <italic>et al</italic> [<xref ref-type="bibr" rid="B13">13</xref>] most recently demonstrated that PTEN interacts with the focal adhesion kinase, and negatively regulates cellular interactions with the extracellular matrix by inhibiting integrin-mediated cell spreading, as well as formation of focal adhesions. Also, the tyrosine phosphatase TEP1 has been shown to be identical to the protein encoded by <italic>PTEN</italic>, and gene transcription of TEP1 has been demonstrated to be downregulated by transforming growth factor (TGF)-&#x003b2; [<xref ref-type="bibr" rid="B14">14</xref>]. Therefore, it could be hypothesized that TGF-&#x003b2;, which is expressed abundantly in the RA synovial membrane [<xref ref-type="bibr" rid="B15">15</xref>], might be responsible for the downregulation of PTEN. The expression of TGF-&#x003b2; is not restricted to the lining, however, but is found throughout the synovial tissue in RA [<xref ref-type="bibr" rid="B15">15</xref>]. Moreover, in the present study the percentage of PTEN-expressing RA-SF remained stable for six passages in culture, whereas molecules that are cytokine-regulated <italic>in vivo</italic> frequently change their expression levels when cultured over several passages. Also, cultured RA-SF that were implanted into SCID mice and deeply invaded the cartilage did not show significant expression of PTEN after 60 days.</p><p>The drop in the percentage of PTEN-expressing cells from the original cell cultures to the SCID mouse implants is of interest, because this observation is in accord with data from previous studies [<xref ref-type="bibr" rid="B16">16</xref>] that showed the prominent expression of activation related molecules in the SCID mice implants that are found predominantly in the lining layer <italic>in vivo</italic>. It may be speculated that activated, aggressive RA-SF are selected positively during the implantation by apoptosis of the nonaggressive cells.</p><p>With regard to PTEN, the present data point to endogenous mechanisms rather than to the influence of exogenous human cytokines or factors in the downregulation of PTEN. In this context, the question of whether the PTEN low-expressing phenotype constitutes a subset of RA-SF that are identical to previously described activated RA-SF is of critical importance. Some recent studies [<xref ref-type="bibr" rid="B16">16</xref>] searching for apoptosis regulating molecules as well as adhesion molecules and signalling cascades in RA-SF have provided novel insights into the nature of these aggressive RA-SF and have helped to characterize them on a molecular level. Thus far, though, there is no specific marker for the activated phenotype of RA-SF found in the lining layer of RA patients. Low expression of PTEN may be among the features that distinguish between the activated phenotype of RA-SF and the sublining, proliferating but nondestructive cells. It needs to be stressed, however, that the association between the lack of PTEN expression and the aggressive phenotype of RA-SF is, at this point, only phenomenological. A comprehensive analysis of different markers and pathways including functional analysis will be needed to clearly identify and specifically distinguish the activated phenotype of RA-SF that aggressively invade the cartilage in RA on a molecular level.</p></sec>

Mast cell activation and its relation to proinflammatory cytokine production in the rheumatoid lesion

<sec><title>Introduction:</title><p>Increased numbers of mast cells (MCs) are found in the synovial tissues and fluids of patients with rheumatoid arthritis (RA), and at sites of cartilage erosion. MC activation has been reported for a significant proportion of rheumatoid specimens. Because the MC contains potent mediators, including histamine, heparin, proteinases, leukotrienes and multifunctional cytokines, its potential contributions to the processes of inflammation and matrix degradation have recently become evident.</p><p>Proinflammatory cytokines are important mediators of inflammation, immunity, proteolysis, cell recruitment and proliferation. Tumour necrosis factor (TNF) reportedly plays a pivotal role in the pathogenesis of RA, especially its ability to regulate interleukin (IL)-1&#x003b2; expression, this being important for the induction of prostanoid and matrix metalloproteinase production by synovial fibroblasts and chondrocytes. IL-15 has been assigned numerous biological effects and has been implicated as an important factor in TNF-&#x003b1; expression by monocyte/macrophages. Some <italic>in vitro</italic> studies have placed IL-15 upstream from TNF-&#x003b1; in the cytokine cascade, suggesting an interdependence between TNF, IL-1 and IL-15 for the promotion of proinflammatory cytokine expression in the rheumatoid joint.</p></sec><sec><title>Aims:</title><p>To examine the <italic>in situ</italic> relationships of TNF-&#x003b1;, IL-1&#x003b2; and IL-15 in relation to MC activation in rheumatoid tissues by use of immunolocalization techniques; and to compare quantitatively the proinflammatory cytokine production by specific cell cultures and rheumatoid synovial explants with and without exposure to a MC secretagogue.</p></sec><sec><title>Materials and methods:</title><p>Samples of rheumatoid synovial tissue and cartilage&#x02013;pannus junction were obtained from patients (<italic>n</italic> = 15) with classic late-stage RA. Tissue sections were immunostained for MC (tryptase) and the proinflammatory cytokines IL-1, TNF-&#x003b1; and IL-15. Rheumatoid synovial tissue explants were cultured in Dulbecco's modified Eagles medium (DMEM) containing either the MC secretagogue rabbit antihuman immunoglobulin (Ig)E, or control rabbit IgG. Primary rheumatoid synovial cell cultures, human articular chondrocytes, synovial fibroblasts and synovial macrophages were prepared as described in the full article. Conditioned culture media from these cultures were collected and assayed for IL-1&#x003b2;, TNF-&#x003b1; and IL-15 using enzyme-linked immunosorbent assay methodology.</p></sec><sec><title>Results:</title><p>Immunohistological studies of rheumatoid synovial tissues have demonstrated local concentrations of MCs in most specimens of the rheumatoid lesion. Sites of MC activation were associated with localized oedema, and TNF-&#x003b1;, IL-1&#x003b1; and IL-1&#x003b2; production by a proportion of mononuclear inflammatory cells. By contrast, no evidence was found for IL-15 production in tissue sites containing either intact or activated MCs, and IL-15 expression, when observed, bore no relation to tissue sites where TNF-&#x003b1; and IL-1&#x003b2; were evident. The immunodetection of IL-15 was restricted to microfocal sites and was not typical of most junctional specimens, but was associated with a proportion of articular chondrocytes in a minority of junctional specimens.</p><p>MC activation within synovial explant cultures was induced by the addition of polyclonal antibody to human IgE. MC activation significantly reduced the levels of TNF-&#x003b1; and IL1&#x003b2; released into the medium, this representing approximately 33% of control values. By contrast, MC activation had little effect on the levels of IL-15 released into the culture medium, the average value being very low in relation to the release of TNF-&#x003b1; and IL-1&#x003b2; . Thus, induced MC activation brings about changes in the amounts of released tryptase, TNF-&#x003b1; and IL-1&#x003b2; , but not of IL-15.</p><p>Four preparations of primary rheumatoid synovial cell cultures produced more IL-1&#x003b2; than TNF-&#x003b1;, with only modest values for IL-15 production, indicating that all three cytokines are produced and released as free ligands by these cultures. Of specific cell types that produced IL-15 <italic>in vitro</italic>, macrophages produced more than fibroblasts, which in turn produced more than chondrocytes. This demonstrates that all three cell types have the potential to produce IL-15 <italic>in situ</italic>.</p></sec><sec><title>Discussion:</title><p>The biological consequences of MC activation <italic>in vivo</italic> are extremely complex, and in all probability relate to the release of various combinations of soluble and granular factors, as well as to the expression of appropriate receptors by neighbouring cells. The subsequent synthesis and release of cytokines such as TNF-&#x003b1; and IL-1 may well follow at specific stages after activation, or may be an induced cytokine response by adjacent macrophagic or fibroblastic cells. However, because no IL-15 was detectable either in or around activated or intact MCs, and the induced MC activation explant study showed no change in IL-15 production, it seems unlikely that the expression of this cytokine is regulated by MCs. The immunohistochemistry (IHC) demonstration of IL-15 at sites of cartilage erosion, and especially by some chondrocytes of articular cartilage, showed no spatial relationship with either T cells or neutrophils, and suggests other functional properties in these locations. The lack of evidence for an <italic>in situ</italic> association of IL-15 with TNF and IL-1 does not support a role for IL-15 in a proinflammatory cytokine 'cascade', as proposed by other <italic>in vitro</italic> experiments. We believe that sufficient evidence is available, however, to suggest that MC activation makes a significant contribution to the pathophysiological processes of the rheumatoid lesion.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Woolley</surname><given-names>David E</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>david.woolley@mri.cmht.nwest.nhs.uk</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Tetlow</surname><given-names>Lynne C</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Rheumatoid arthritis (RA) is characterized by chronic inflammation, cartilage destruction and joint deformity. Histopathological observations of the 'rheumatoid lesion' &#x02014; a term used to describe cartilage-pannus junctions and cartilage erosion sites [<xref ref-type="bibr" rid="B1">1</xref>] &#x02014; have identified a number of cell types, each of which may contribute different mediators to the inflammatory and degradative processes; these usually being microenvironmental in nature [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Although macrophages, fibroblastic synoviocytes, lymphocytes and neutrophils are recognized as important contributors to RA joint pathology, the mast cell (MC) has generally been neglected. Increased numbers of MCs are found in the synovial tissue and fluid of patients with RA [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>], and especially at sites of cartilage erosion [<xref ref-type="bibr" rid="B5">5</xref>]. MC activation has been reported in these locations for a significant proportion of rheumatoid specimens [<xref ref-type="bibr" rid="B3">3</xref>]. Because the MC contains a variety of potent mediators, including histamine, heparin, proteinases, leukotrienes and multifunctional cytokines [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>], its potential contributions to the processes of inflammation and matrix degradation have recently become evident [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B8">8</xref>].</p><p>Proinflammatory cytokines are important mediators of inflammation, immunity, proteolysis, cell recruitment and proliferation. Tumour necrosis factor (TNF)-&#x003b1; and TNF-&#x003b2;, and interleukin (IL)-1&#x003b1; and IL-1&#x003b2; have received much attention over the past decade [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>], and both TNF and IL-1 have been demonstrated in the rheumatoid lesion by immunohistochemistry [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. TNF reportedly plays a pivotal role in the pathogenesis of RA [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>], especially its ability to regulate IL-1&#x003b2; expression, this being important for the induction of prostanoid and matrix metalloproteinase production by synovial fibroblasts and chondrocytes [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. Cellular interactions mediated by TNF and IL-1, cytokines that are mainly produced by activated macrophages, have become prominent factors in the numerous reviews that have proposed a sequence of events leading to cartilage damage in RA [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. The precise factors that induce monocyte/macrophages to produce TNF and IL-1 remain obscure, however, although recent studies [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>] have suggested a role for IL-15.</p><p>IL-15 is reportedly expressed by activated monocytes, epithelial cells and fibroblasts [<xref ref-type="bibr" rid="B17">17</xref>]. IL-15 has been assigned numerous biological effects: it is a potent T-cell chemokine; it induces proliferation of mitogen-activated T cells; it contributes to B-cell proliferation and immunoglobulin (Ig) synthesis; and it enhances neutrophil and natural killer cell responses [<xref ref-type="bibr" rid="B17">17</xref>]. It was recently implicated as an important factor in TNF-&#x003b1; expression by monocyte/ macrophages. IL-15 together with IL-6 and TNF-&#x003b1; was shown to induce monocyte TNF-&#x003b1; production, a response similarly induced by IL-15-stimulated T cells via contact with monocytes [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. Other studies have placed IL-15 upstream from TNF-&#x003b1; in the cytokine cascade, showing that IL-15 activation of T cells directly produces TNF-&#x003b1; as well as amplifying inflammatory responses [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. Such reports, derived from <italic>in vitro</italic> cell studies, have suggested an interdependence between TNF, IL-1 and IL-15, and between monocyte/macrophages and T cells, for the promotion of proinflammatory cytokine expression in the rheumatoid joint. The present study examines these possible relationships <italic>in situ</italic>, especially in relation to MC activation, by use of immunolocalization techniques on freshly fixed rheumatoid tissues, and by a comparative quantitative assessment of proinflammatory cytokine production by rheumatoid synovial explants and specific cell cultures. We report here that IL-15 production <italic>in situ</italic> does not appear to be related to sites of TNF or IL-1 production, or with MC activation, and that synovial explants, synoviocytes, macrophages and chondrocytes <italic>in vitro</italic> produce quite modest amounts of IL-15 compared with those of TNF-&#x003b1; and IL-1&#x003b2; .</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Tissue samples</title><p>Samples of rheumatoid synovial tissue and cartilage-pannus junction were obtained from arthroplasty specimens from patients (<italic>n</italic> = 15) with classic late-stage rheumatoid arthritis (RA). Each sample was fixed in Carnoy's fixative at 20&#x000b0;C for 1&#x02013;2 h and embedded in paraffin wax, and 5 &#x003bc; m sections were cut. Tissue sections were dewaxed, rehydrated and examined for the presence of MCs and the proinflammatory cytokines IL-1, TNF-&#x003b1; and IL-15 using immunohistochemical techniques.</p></sec><sec><title>Immunolocalization of mast cells</title><p>Tissue sections were dewaxed and pretreated for 30 min with 10% rabbit serum (vol/vol) in Tris buffered saline (TBS) at pH 7.6. Mouse monoclonal antibody to human MC tryptase (Biogenesis, Poole, Dorset, UK) was diluted 1:200 and applied to the sections for 2 h at 20&#x000b0;C. After three 10-min washes in TBS, alkaline phosphatase-conjugated rabbit antimouse IgG (Dako Ltd, Cambridge, UK) diluted 1:50 was applied for 1 h at 20&#x000b0;C. After further washing the alkaline phosphatase was developed using new fuchsin substrate. Tissue sections were lightly counterstained with Harris's haematoxylin or Toluidine blue as previously described [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>].</p></sec><sec><title>Dual immunolocalization of mast cell tryptase and cytokines IL-1&#x003b1;, IL-1&#x003b2; or TNF-&#x003b1; </title><p>Tissue sections were pretreated for 30 min with rabbit serum at 10% (vol/vol) final concentration in TBS. After draining, the sections were incubated at 20&#x000b0;C for 2 h with a combination of primary antibodies, namely mouse monoclonal antitryptase (diluted 1 : 200) with either a goat polyclonal antibody to IL-1&#x003b1; or IL-1&#x003b2; (diluted 1 : 50) or goat polyclonal antibody to TNF-&#x003b1; (diluted 1 : 50). Polyclonal antibodies were purchased from R&#x00026;D Systems (Abingdon, UK). After three 10 min washes in TBS, secondary antibodies alkaline phosphatase-conjugated rabbit antimouse and horseradish peroxidase-conjugated rabbit antigoat (both from Dako) diluted 1 : 50 in TBS were applied together for 1 h at 20&#x000b0;C followed by three 10-min washes in TBS. Peroxidase was developed first using diaminobenzidine (brown colour) and the sections were washed, followed by development of alkaline phosphatase using new fuchsin (red colour) as described previously [<xref ref-type="bibr" rid="B2">2</xref>]. The sections were lightly counterstained in Harris's haematoxylin to demonstrate nuclei, dehydrated, mounted in Histomount (Mensura, Wigan, UK) and photographed using a Zeiss (Oberkochen, Germany) photomicroscope III and Ektachrome (Rochester, New York, USA) 160 tungsten film<bold>.</bold></p></sec><sec><title>Immunolocalization of interleukin-15</title><p>IL-15 was immunolocalized using a goat polyclonal antibody from R&#x00026;D Systems.</p><p>Tissue sections were pretreated with 10% (vol/vol) rabbit serum in TBS for 30 min. The primary antibody, diluted 1 : 100 in TBS, was applied to the sections and incubated for 2 h at 20&#x000b0;C. After three 10-min washes, biotinylated rabbit antigoat IgG was applied to the sections for 45 min, followed by a further washing in TBS. Alkaline phosphatase conjugated avidinbiotin complex was diluted as recommended by the supplier (Dako) and applied to the sections for 45 min at 20&#x000b0;C. After further washing alkaline phosphatase was developed using new fuchsin substrate (red). Consecutive sections of cartilage-pannus junction and rheumatoid synovium were stained for MC tryptase, IL-15, TNF-&#x003b1; and IL-1&#x003b2;, respectively.</p></sec><sec><title>Control tissue sections</title><p>Normal IgG from the same species and in concentrations similar to those used for the primary antibodies was substituted for each primary antibody and consistently gave negative results. In addition, each cytokine antibody was preabsorbed with its relevant antigen and used in place of the primary antibody; this also produced negative data and confirmed specificity.</p></sec><sec><title>Explant cultures</title><p>Rheumatoid synovial tissue obtained shortly after surgery was cut into small pieces of approximately 3 mm³ in Hanks balanced salt solution (HBSS; Gibco, Paisley, Scotland). Randomized explants (8&#x02013;10) were placed in each of four or six preweighed sterile culture dishes. After reweighing to determine the wet weight of the tissue per dish, 2 ml Dulbecco's modified Eagles medium (DMEM, Gibco) containing either the MC secretagogue rabbit antihuman IgE or control rabbit IgG, each at a final concentration of 150 &#x003bc; g/ml, was added to the cultures in duplicate or triplicate [<xref ref-type="bibr" rid="B18">18</xref>]. The dishes were incubated at 37&#x000b0;C in a 5% air incubator with humidified atmosphere.CO₂/95% air incubator with humidified atmosphere.Culture medium was removed at 20 h, aliquoted and stored at -20&#x000b0;C until it was assayed for the cytokines TNF-&#x003b1;, IL-1&#x003b2; and IL-15.</p></sec><sec><title>Preparation of rheumatoid synovial cell cultures</title><p>Rheumatoid synovial tissue was washed with HBSS, chopped into fine pieces and enzymically digested by incubation with 10 ml DMEM containing 1 mg/ml bacterial collagenase (<italic>Clostridium hystoliticum</italic>; Boehringer, Mannheim, Germany) in a shaking water bath for 3 h at 37&#x000b0;C. The resultant cell suspension was filtered, washed with DMEM and the pellet of cells was resuspended in DMEM + 10% (vol/vol) foetal calf serum (FCS), as described previously [<xref ref-type="bibr" rid="B19">19</xref>]. An aliquot of 200 &#x003bc; l dissociated synovial cells were plated at a density of 2 &#x000d7;10⁵ cells/well in 12-well culture dishes and incubated at 37&#x000b0;C with a 5% CO₂ incubator. After 48 h the conditionedmedium was harvested, centrifuged to remove nonadherent cells and stored at -20&#x000b0;C until assaying for the cytokines TNF-&#x003b1;, IL-1&#x003b2; and IL-15.</p></sec><sec><title>Synovial macrophage-enriched cultures</title><p>Dissociated rheumatoid synovial cells prepared as described above were plated into 80 cm² culture flasks and, after 24 h, the medium containing the nonadherent cells was removed and discarded. The remaining adherent cells, mainly fibroblasts and macrophages, were first washed with HBSS and then harvested by trypsinization. The cells were resuspended in HBSS containing 2% FCS using 80 &#x003bc; l buffer per 10⁷ cells. Cells of anti-fibroblast microbeads 20 &#x003bc; l per 10⁷ (Miltenyi Biotec, Bisley, Surrey, UK) was added to the cell suspension and the mixture was incubated at 20&#x000b0;C for 30 min. The cells were washed in 20 times the volume of the HBSS and centrifuged at 2000 revolutions/min for 10 min. After resuspension in 1 ml HBSS + 2% FCS, the cells were separated using a MS + magnetic separation column (Miltenyi Biotec). The resultant eluant gave cell preparations that were depleted of fibroblasts and rich in macrophages. The cells were plated out in DMEM + 10% FCS into 12-well culture dishes overnight and then transferred to DMEM + 2% FCS for 48 h. Cultures were shown to contain more than 90% macrophages by immunostaining with CD68 macrophage marker (Dako; data not shown). The medium was harvested and stored at -20&#x000b0;C until assaying for the cytokines IL-1&#x003b2;, TNF-&#x003b1; and IL-15 by enzyme-linked immunosorbent assay (ELISA). The cells were fixed in the culture wells, stained and counted.</p></sec><sec><title>Rheumatoid synovial fibroblasts</title><p>Subcultures of adherent synovial fibroblasts were used at passages two to four. Cells were plated out in DMEM + 10% FCS into 12-well culture dishes and, when confluent, the medium was replaced with DMEM + 2% FCS for 48 h. Medium was collected, stored and assayed for cytokines. Cells were fixed, stained and counted.</p></sec><sec><title>Chondrocyte cultures</title><p>Macroscopically normal articular cartilage was enzymically digested as described previously [<xref ref-type="bibr" rid="B20">20</xref>]. Chondrocytes were grown to confluence in 12-well culture dishes in DMEM + 10% FCS and used as described above for the synovial macrophage and fibroblast cultures, after which cells were fixed, stained and counted.</p></sec><sec><title>Measurement of cytokines</title><p>IL-1&#x003b2;, TNF-&#x003b1; and IL-15 were measured using Quantikine ELISA kits from R&#x00026;D Systems, following the manufacturer's instructions. Detection limits of the ELISAs were &#x0003c;1pg/ml, 4pg/ml and 1pg/ml for IL-1&#x003b2;, TNF-&#x003b1; and IL-15, respectively.</p></sec></sec><sec><title>Results</title><p>Immunohistological studies of rheumatoid synovial tissues have demonstrated local concentrations of MCs in most specimens of the rheumatoid lesion. Previous studies have shown evidence of MC activation <italic>in situ</italic>, as judged by the release of the MC-specific enzyme tryptase. Figure <xref ref-type="fig" rid="F1">1a</xref> shows MC activation with evidence of local oedema, associated with sites of TNF-&#x003b1; production by a proportion of mononuclear inflammatory cells. Figure <xref ref-type="fig" rid="F1">1b</xref> also shows dual immunolocalization of MC tryptase together with IL-1&#x003b2; production by some neighbouring cells. Associations of IL-1&#x003b2; and sites of MC activation were similarly observed [<xref ref-type="bibr" rid="B2">2</xref>]. Although MCs are reported to have the potential to express TNF-&#x003b1; and IL-1, only occasionally has TNF-&#x003b1; production by MCs been demonstrated in our rheumatoid specimens.</p><p>The association of TNF-&#x003b1; and IL-1&#x003b1; or IL-1&#x003b2; production by cells in similar locations to activated MCs was a common observation. By contrast, no evidence was found for IL-15 production in tissue sites containing either intact or activated MCs. Figures <xref ref-type="fig" rid="F1">1c</xref> and <xref ref-type="fig" rid="F1">1d</xref> show consecutive tissue sections stained for tryptase and IL-15, respectively; the latter showing negligible staining despite the presence of numerous MCs, some of which show evidence of degranulation. Indeed, IL-15 expression, when observed, bore no relation to tissue sites where TNF-&#x003b1; and IL-1&#x003b2; were evident. The immunodetection of IL-15 was variable between the different rheumatoid tissue specimens; whereas some showed prominent staining of synovial lining cells, especially CD68⁺ macrophages, other specimens were devoid of IL-15. Figure <xref ref-type="fig" rid="F1">1e</xref> shows a cartilage-pannus junction with evidence of extracellular IL-15 staining at cartilage erosion sites, and with intracellular staining localized to a few chondrocytic and macrophagic cells. This distribution of IL-15 was restricted to microfocal sites; it was not typical of most junctional specimens, but was not associated with the local expression of TNF-&#x003b1; or IL-1. Similarly, Figure <xref ref-type="fig" rid="F1">1f</xref> shows a proportion of articular chondrocytes at a cartilage-pannus junction stained for IL-15. Again, such observations were restricted to discrete sites where only a proportion of the cells were positive for IL-15. Thus, it appears that the production of IL-15 in rheumatoid tissues is independent of MC activation, and the local production of TNF-&#x003b1; and IL-1&#x003b1; and IL-1&#x003b2;.</p><p>We recognize that all of these specimens are fixed at one window in time, but despite the reported interdependence of these proinflammatory cytokines we have not as yet observed any colocalization of IL-15 with TNF or IL-1 in any of the 15 rheumatoid specimens fixed within minutes of surgical excision. Table <xref ref-type="table" rid="T1">1</xref> is a general summary of the immunolocalization data, which provides information on the relative frequencies for cytokine production in specimens of rheumatoid tissues and at the cartilage-pannus interface. Of special note is that most positive observations were microfocal in nature, seldom involving large groups of cells, with a significant proportion of the CD68 ⁺ macrophagic cells showing no evidence of cytokine production. Such observations suggest that at least <italic>in situ</italic> cytokine expression by macrophages and chondrocytes is subject to regulation, but the induction factors remain uncertain. These findings have been further examined using <italic>in vitro</italic> studies with rheumatoid synovial explants and monolayer cultures of specific cell types.</p><p>MC activation within synovial explant cultures was induced by the addition of polyclonal antibody to human IgE, this being an effective MC secretagogue. Its effect was confirmed by the elevated values for MC tryptase in the culture medium, with most cultures showing more than a threefold increase over controls (Table <xref ref-type="table" rid="T2">2</xref>). Because the cellular composition of each synovial specimen is known to be variable in terms of both total cell numbers and specific cell types, quantitative differences for tryptase and cytokine values between cultures were to be expected. MC activation significantly reduced the levels of TNF-&#x003b1; released into the medium, this representing approximately 30% of control values for most explant cultures (average values being 177.2 and 53.2 pg/ml per 100 mg tissue per 20 h for control and anti-IgE treatments, respectively). Similarly MC activation reduced the levels of IL-1&#x003b2; released into the medium, this representing approximately 33% of control values (averaged values being 50.1 and 16.1 pg/100 mg tissue per 20 h for control and anti-IgE treatments, respectively). By contrast, MC activation apparently had little effect on the levels of IL-15 released into the culture medium, the average value of less than 1 pg/ml per 100 mg tissue per 20 h being very low in relation to the release of TNF-&#x003b1; and IL-1&#x003b2; (Table <xref ref-type="table" rid="T2">2</xref>). Thus, the explant data demonstrate that induced MC activation brings about changes in the amounts of released tryptase, TNF-&#x003b1; and IL-1-&#x003b2;, but not of IL-15.</p><p>It is possible that the measurements of tryptase and cytokines in the conditioned culture medium do not necessarily reflect total release or production by the tissue explants, because the accessibility of each protein into the culture medium will depend to some extent on the relative interactions and binding properties to the extracellular matrix, cells and receptors. To overcome this problem of possible retention of cytokines by the three-dimensional properties of the explant tissue, production of the three cytokines by primary cultures of dissociated rheumatoid synovial cells was examined. Table <xref ref-type="table" rid="T3">3</xref> shows that, for four preparations of primary rheumatoid synovial cell cultures, variations were observed for the relative values of the three cytokines. This in all probability reflects the different compositions of specific cell types for each preparation. In three of the cultures, however, the IL-1&#x003b2; values were greater than those for TNF-&#x003b1;, with only modest values for IL-15 production. This trend is borne out by the averaged values for each cytokine; IL-1&#x003b2; production was greater than that of TNF-&#x003b1;, which in turn was greater than that of IL-15, these being 190, 85 and 14 pg/10⁶ cells per 48 h, respectively (Table <xref ref-type="table" rid="T3">3</xref>). It is recognized that membrane- or surface-bound forms of these cytokines exist for specific cell types, and these will not have contributed to the values presented here because all cells were removed from the conditioned medium before assay was carried out. Similarly, cytokine receptor expression by the cell cultures will contribute to some cytokine depletion of the conditioned medium. Nevertheless, these experiments with primary cultures of synovial cells give some indication that all three cytokines are produced and released as free ligands, with IL-15 representing approximately 5% of the three cytokines present in the medium after 48 h of culture.</p><p>Several cell types have been shown to produce IL-15 <italic>in vitro</italic>, but few studies have compared production by different cell types. Table <xref ref-type="table" rid="T4">4</xref> shows the values for IL-15 production and release into the culture medium by synovial macrophages, synovial fibroblasts and human articular chondrocytes. The data were obtained from three different preparations of each cell type, maintained in medium containing 2% FCS supplement, the latter having no detectable IL-15. In relation to cell numbers the ability of these cells to produce IL-15 followed the order macrophages, followed by fibroblasts, followed by chondrocytes, thereby demonstrating that all three cell types have the potential to produce this cytokine <italic>in situ</italic>.</p></sec><sec><title>Discussion</title><p>Recent histological studies of rheumatoid synovial tissues have demonstrated localized accumulations of MCs and evidence of their activation/degranulation, especially at cartilage erosion sites [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Because MCs contain or express numerous potent mediators, including histamine, heparin, proteinases, prostanoids, cytokines and growth factors, it seems most likely that MC secretion, activation or degranulation <italic>in situ</italic> will bring about changes in the local environment, not least the phenotype of neighbouring cells. For example, histamine induces tissue oedema via its effects on endothelial cells, activates chondrocytes via H₁ and H₂ receptors, and stimulates synoviocytes through H₁ receptors [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. Heparin interacts with growth factors, contributes to angiogenesis and stimulates IL-1 production by monocyte/macrophages [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. Tryptase and chymase are MC-specific proteinases that degrade various matrix components, and are capable of activating the zymogen forms of the matrix metalloproteinases, prostromelysin and procollagenase [<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>]. These MC enzymes are used to identify two subsets of MCs in human tissues. One subsetcontains both tryptase and chymase, and is termed MC_TC; the other contains only tryptase, and is known as MC_T. Both subsets have been demonstrated in rheumatoid synovial tissues, where they are purported to exert distinct functions with regard to inflammatory and degradative processes [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>]. In addition to proteinase heterogeneity the MC subsets are reported to have different cytokine profiles; whereas MC_T preferentially produce IL-4, IL-5 and IL-6, the MC_TC phenotype contains predominantly IL-4 [<xref ref-type="bibr" rid="B29">29</xref>]. Indeed, the realization that MCs can express several multi-functional cytokines (including the proinflammatory mediators TNF-&#x003b1; and IL-1&#x003b2; [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]) and various profibrotic cytokines (such as basic fibroblast growth factor and IL-4), and have the ability to release membrane-bound TGF-&#x003b1; [<xref ref-type="bibr" rid="B28">28</xref>], suggests numerous functional roles for MCs in the rheumatoid lesion [<xref ref-type="bibr" rid="B30">30</xref>].</p><p>Our earlier immunohistochemistry (IHC) studies of rheumatoid synovial tissue presented evidence for MC activation, as judged by the extracellular release of tryptase, and its association with the microenvironmental expression of TNF-&#x003b1;, IL-1&#x003b1;, IL-1&#x003b2;, stromelysin and collagenase-1 [<xref ref-type="bibr" rid="B2">2</xref>]. These observations were in accord with previous reports that soluble MC products stimulated collagenase production by cultures of synoviocytes and chondrocytes [<xref ref-type="bibr" rid="B31">31</xref>], and induced monocyte/macrophages to increase IL-1 production [<xref ref-type="bibr" rid="B24">24</xref>]. Such studies suggested that MC activation was likely to bring about increased proinflammatory cytokine and metalloproteinase production <italic>in situ</italic>, but paradoxically in the present study the induced MC activation of synovial explants resulted in a significant reduction in TNF-&#x003b1; and IL-1&#x003b2; release into the culture medium. Explanations for this are currently unresolved, but may relate to an increase in the expression of the relevant receptors or to some binding or sequestration of these ligands by the tissue matrix or released heparin.</p><p>The biological consequences of MC activation <italic>in vivo</italic> are extremely complex, and in all probability relate to the release of various combinations of soluble and granular factors, as well as the expression of appropriate receptors by neighbouring cells. As yet we have a poor understanding of the hierarchy of the MC mediators released upon degranulation. Although histamine is released rapidly and induces tissue oedema, the solubilization and release of factors from exocytosed granules provides a temporally regulated supply of specific signals within the localized domain of the degranulated MC [<xref ref-type="bibr" rid="B23">23</xref>]. The subsequent synthesis and release of specific cytokines may well follow at specific stages after activation, or, as demonstrated here by IHC, may be an induced cytokine response by adjacent macrophagic or fibroblastic cells. It is therefore possible that the IHC observations presented here reflect transient changes in cytokine expression, the extracellular tissue distribution observed for the tryptase protein possibly taking longer to process or remove from the tissue than the three cytokines. However, because no IL-15 was detectable either in or around activated or intact MCs, and the induced MC activation explant study showed no change in the IL-15 production, it seems unlikely that the expression of this cytokine is regulated by MCs.</p><p>The proinflammatory cytokines IL-1 and TNF-&#x003b1; are reported to play important roles in cartilage and bone degradation, with TNF-&#x003b1; occupying a primary position in the cytokine cascade through its ability to upregulate production of other cytokines, including IL-1, granulocyte-macrophage colony-stimulating factor, IL-6, IL-8 and IL-10 [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>]. TNF-&#x003b1; expression by monocyte/macrophages has been demonstrated both in synovial tissue and at the cartilage-pannus junction [<xref ref-type="bibr" rid="B32">32</xref>]. Similarly, IL-1 and its receptor have been demonstrated at sites of cartilage erosion [<xref ref-type="bibr" rid="B12">12</xref>]. Although monocyte/macrophages are recognized sources of both TNF-&#x003b1; and IL-1, in many specimens it is only a proportion of the macrophage numbers that express these cytokines at the time of surgery and tissue fixation. Similarly with MCs, although reported to produce TNF and IL-1 only very few have been shown to do so in rheumatoid synovial tissue. Such observations indicate that both TNF and IL-1 expression is subject to regulation, but the nature of the stimulatory agents remains uncertain [<xref ref-type="bibr" rid="B10">10</xref>]. Endotoxin, immunoglobulins, heparin and rheumatoid factors have been invoked, but more recent studies have shown the importance of T cells in mediating TNF and IL-1 production by monocyte/macrophages, especially via cell surface signals from T cells after their priming with IL-15 [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>].</p><p>Several cell types are capable of producing IL-15. These include activated monocytes, macrophages, epithelial cells, fibroblasts [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B33">33</xref>] and endothelial cells [<xref ref-type="bibr" rid="B34">34</xref>], and now chondrocytes as described in the present study. Other immunohistochemical studies of IL-15 in rheumatoid synovial tissues have identified CD68⁺ lining cells [<xref ref-type="bibr" rid="B14">14</xref>], macrophages, T cells and natural killer cells as positively stained for IL-15 [<xref ref-type="bibr" rid="B35">35</xref>]. One of its many functions is the recruitment, migration and activation of T cells, subsequently contributing to monocyte-derived TNF-&#x003b1; production via cell surface contact [<xref ref-type="bibr" rid="B17">17</xref>]. IL-15 is also reported to stimulate monocytes to produce the chemokines IL-8 and monocyte chemotactic protein-1, which play important roles in the regulation of leucocyte infiltration during inflammation [<xref ref-type="bibr" rid="B33">33</xref>], a concept also proposed for the cellular composition of rheumatoid synovial tissue [<xref ref-type="bibr" rid="B35">35</xref>]. </p><p>The present observations are generally in accord with these reports, atleast with regard to IL-15 expression by some synovial lining cells. However, the demonstration of IL-15 at sites of cartilage erosion, and especially by some chondrocytes of articular cartilage, showed no spatial relationship with T cells or neutrophils, and therefore suggest other functional properties in these locations. This is supported to some extent by the lack of evidence for an <italic>in situ</italic> association of IL-15 with TNF and IL-1, observations that do not necessarily support a role for IL-15 in a proinflammatory cytokine 'cascade', as determined by <italic>in vitro</italic> experiments [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>].</p><p>The present study has shown that the production of IL-15 by synovial explants and by cells <italic>in vitro</italic> was rather modest when compared with that for TNF-&#x003b1; and IL-1&#x003b2;. Harada <italic>et al</italic> [<xref ref-type="bibr" rid="B36">36</xref>] reported that fibroblast-like synoviocytes in culture produced large amounts of IL-15 protein and messenger RNA, this being elevated when stimulated with IL-1 or TNF. Receptor expression for each of these cytokines is clearly an important aspect for the understanding of these cytokine interactions, and so too is the realization that some cell types manifest membrane-bound cytokines. Biologically active IL-15 was detected in a constitutively expressed, membrane-bound form on normal human monocytes and monocytic cell lines, prompting Musso <italic>et al</italic> [<xref ref-type="bibr" rid="B37">37</xref>] to speculate that most of the natural biological effects of IL-15 are exerted by the cell surface-bound form. Cell membrane-bound forms of IL-1, TNF-&#x003b1; and IL-10 have also been reported, but it is unclear at present how these relate to the corresponding free forms of TNF-&#x003b1;, IL-1&#x003b2; and IL-15 measured in the experiments described here. It is apparent that the cellular interactions and proinflammatory cytokine responses that promote the inflammatory processes of rheumatoid synovial tissue and joint destruction are far from being resolved. We believe, however, that sufficient evidence is available to suggest that MC activation makes a significant contribution to the pathophysiological processes of the rheumatoid lesion.</p></sec><sec><title>Acknowledgements</title><p>We thank consultant orthopaedic surgeons T Dunningham (Tameside Hospital, Manchester) and M Morris (Devonshire Royal Hospital, Buxton) for the supply of rheumatoid tissues. This work was supported by project grants for the Arthritis Research Campaign, UK.</p></sec>

IFN-&#x003b3; production in response to <italic>in vitro</italic> stimulation with collagen type II in rheumatoid arthritis is associated with HLA-DRB1^*0401 and HLA-DQ8

<sec><title>Introduction:</title><p>Despite much work over past decades, whether antigen-specific immune reactions occur in rheumatoid arthritis (RA) and to what extent such reactions are directed towards joint-specific autoantigens is still questionable. One strong indicator for antigenic involvement in RA is the fact that certain major histocompatibility complex (MHC) class II genotypes [human leucocyte antigen (HLA)-DR4 and HLA-DR1] predispose for the development of the disease [<xref ref-type="bibr" rid="B1">1</xref>]. In the present report, collagen type II (CII) was studied as a putative autoantigen on the basis of both clinical and experimental data that show an increased frequency of antibodies to CII in RA patients [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>] and that show that CII can induce experimental arthritis [<xref ref-type="bibr" rid="B5">5</xref>].</p><p>It is evident from the literature that RA peripheral blood mononuclear cells (PBMCs) respond poorly to antigenic stimulation [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>], and in particular evidence for a partial tolerization to CII has been presented [<xref ref-type="bibr" rid="B9">9</xref>]. The strategy of the present work has accordingly been to reinvestigate T-cell reactivity to CII in RA patients, to relate it to the response to commonly used recall antigens and to analyze IFN-&#x003b3; responses as an alternative to proliferative responses.</p></sec><sec><title>Aims:</title><p>To study cellular immune reactivity to CII in patients with RA and in healthy control individuals and to correlate this reactivity to HLA class II genotypes and to the presence of antibodies to CII in serum.</p></sec><sec><title>Methods:</title><p>Forty-five patients who met the 1987 American College of Rheumatology classification criteria for RA [<xref ref-type="bibr" rid="B10">10</xref>] and 25 healthy control individuals of similar age and sex were included. Twenty-six of these patients who had low levels of anti-CII in serum were randomly chosen, whereas 19 patients with high anti-CII levels were identified by enzyme-linked immunosorbent assay (ELISA)-screening of 400 RA sera.</p><p>Heparinized blood was density gradient separated and PBMCs were cultured at 1 &#x000d7; 10⁶/ml in RPMI-10% fetal calf serum with or without antigenic stimulation: native or denatured CII (100 &#x003bc; g/ml), killed influenza virus (Vaxigrip, Pasteur M&#x000e9;rieux, Lyon, France; diluted 1 : 1000) or purified protein derivative (PPD; 10 &#x003bc; g/ml). CII was heat-denatured in 56&#x000b0;C for 30 min.</p><p>Cell supernatants were collected after 7days and IFN-&#x003b3; contents were analyzed using ELISA. HLA-DR and HLA-DQ genotyping was performed utilizing a polymerase chain reaction-based technique with sequence-specific oligonucleotide probe hybridization. Nonparametric statistical analyses were utilized throughout the study.</p></sec><sec><title>Results:</title><p>PBMCs from both RA patients and healthy control individuals responded with inteferon-&#x003b3; production to the same degree to stimulation with native and denatured CII (Fig. <xref ref-type="fig" rid="F1">1a</xref>), giving median stimulation indexes with native CII of 4.6 for RA patients and 5.4 for healthy control individuals, and with denatured CII of 2.9 for RA patients and 2.6 for healthy control individuals. RA patients with elevated levels of anti-CII had a weaker IFN-&#x003b3; response to both native and denatured CII than did healthy control individuals (<italic>P</italic> = 0.02 and 0.04, respectively).</p><p>Stimulation with the standard recall antigens PPD and killed influenza virus yielded a median stimulation index with PPD of 10.0 for RA patients and 51.3 for healthy control individuals and with influenza of 12.3 for RA patients and 25.7 for healthy, control individuals. The RA patients displayed markedly lower responsiveness to both PPD and killed influenza virus than did healthy control individuals (Fig. <xref ref-type="fig" rid="F1">1b</xref>). IFN-&#x003b3; responses to all antigens were abrogated when coincubating with antibodies blocking MHC class II.</p><p>The low response to PPD and killed influenza virus in RA patients relative to that of healthy control individuals reflects a general downregulation of antigen-induced responsiveness of T cells from RA patients [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. That no difference between the RA group and the control group was recorded in CII-induced IFN-&#x003b3; production therefore indicates that there may be an underlying increased responsiveness to CII in RA patients, which is obscured by the general downregulation of T-cell responsiveness in these patients. In order to address this possibility, we calculated the fraction between individual values for the CII-induced IFN-&#x003b3; production and the PPD-induced and killed influenza virus-induced IFN-&#x003b3; production, and compared these fractions. A highly significant difference between the RA and healthy control groups was apparent after stimulation with both native CII and denatured CII when expressing the response as a fraction of that with PPD (Fig. <xref ref-type="fig" rid="F2">2a</xref>). Similar data were obtained using killed influenza virus-stimulated IFN-&#x003b3; values as the denominator (Fig. <xref ref-type="fig" rid="F2">2b</xref>).</p><p>When comparing the compensated IFN-&#x003b3; response to denatured CII stimulation between RA patients with different HLA genotypes, highly significant differences were evident, with HLA-DRB1*0401 patients having greater CII responsiveness than patients who lacked this genotype (Fig. <xref ref-type="fig" rid="F3">3a</xref>). HLA-DQ8 positive patients also displayed a high responsiveness to CII as compared with HLA-DQ8 negative RA patients (Fig. <xref ref-type="fig" rid="F3">3b</xref>). These associations between the relative T-cell reactivity to denatured CII and HLA class II genotypes were not seen in healthy control individuals. Similar results were achieved using influenza as denominator (<italic>P</italic> = 0.02 for HLA-DRB1*0401 and <italic>P</italic> = 0.01 for HLA-DQ8).</p></sec><sec><title>Discussion:</title><p>No reports have previously systematically taken the general T-cell hyporesponsiveness in RA into account when investigating specific T-cell responses in this disease. In order to address this issue we used the T-cell responses to PPD and killed influenza virus as reference antigens. This was made on the assumption that exposure to these antigens is similar in age-matched and sex-matched groups of RA patients and healthy control individuals. The concept of a general hyporesponsiveness in RA T cells has been documented in several previous reports, in which both nominal antigens [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>] and mitogens [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>] have been used. The fact that a similar functional downregulation in RA PBMCs was obtained with both PPD and killed influenza virus as reference antigens strengthens the validity of our approach.</p><p>We identified an association between the IFN-&#x003b3; response to CII and HLA-DRB1*0401 and HLA-DQ8 in the RA patient group, which is of obvious interest because both these MHC class II alleles have been associated with high responsiveness to CII in transgenic mice that express these human MHC class II molecules [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. There was no association between high anti-CII levels and shared epitope (HLA-DRB1*0401 or HLA-DRB1*0404).</p></sec><sec><title>Conclusion:</title><p>CII, a major autoantigen candidate in RA, can elicit an IFN-&#x003b3; response <italic>in vitro</italic> that is associated with HLA-DRB1*0401 and HLA-DQ8 in RA patients. This study, with a partly new methodological approach to a classical problem in RA, has provided some additional support to the notion that CII may be a target autoantigen of importance for a substantial group of RA patients. Continued efforts to identify mechanisms behind the general hyporesponsiveness to antigens in RA, as well as the mechanisms behind the potential partial anergy to CII, may provide us with better opportunities to study the specificity and pathophysiological relevance of anti-CII reactivity in RA.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Berg</surname><given-names>Louise</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>louise.berg@cmm.ki.se</email></contrib><contrib id="A2" contrib-type="author"><name><surname>R&#x000f6;nnelid</surname><given-names>Johan</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Sanjeevi</surname><given-names>Carani B</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Lampa</surname><given-names>Jon</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Klareskog</surname><given-names>Lars</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Despite much work over past decades, whether antigen-specific immune reactions occur in rheumatoid arthritis (RA) and to what extent such reactions are directed towards joint-specific autoantigens is still questionable. The fact that certain major histocompatibility complex (MHC) class II genotypes [human leucocyte antigen (HLA)-DR4 and HLA-DR1] predispose to the development of RA [<xref ref-type="bibr" rid="B1">1</xref>] points to the possibility of antigenic involvement. The presence of an increased frequency of autoantibodies to cartilage-specific molecules such as collagen type II (CII) in RA [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>] and the fact that CII and other joint-derived proteins can cause arthritis in experimental animals after immunization [<xref ref-type="bibr" rid="B5">5</xref>] indicate a possible pathogenic role of autoimmune reactions towards these molecules. It has been difficult, however, to convincingly demonstrate cellular reactivity against joint-derived autoantigens such as CII in RA. This has contributed to the widespread notion that specific T-cell reactivities may not be all that important in RA, and that autoimmunity to joint antigens such as CII may only be of importance in very limited subgroups of RA patients.</p><p>A few lines of partly new evidence have now led us to reinvestigate the issue of cellular reactivity to CII in RA. First, there is increasing evidence that immunoreactivity to CII can indeed be associated with the RA-associated HLA allele DRB1*0401 and the closely linked allele DQA1*0301-DQB1*0302 (HLA-DQ8). The best evidence comes from studies in DRB1*0401 [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B16">16</xref>] or HLA-DQ8 [<xref ref-type="bibr" rid="B15">15</xref>] transgenic mice, which have a high susceptibility to CII-induced arthritis. There are also studies of antibody production to CII in joints that indicate that this is a relatively common feature among RA patients [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>] and is associated with the presence of HLA-DR4 [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Second, at least two possible mechanisms by which an existing T-cell responsiveness to CII may have remained undetected using conventional methods (such as determination of proliferation in response to CII <italic>in vitro</italic>) have been suggested. One mechanism may reside in the low T-cell reactivity to recall antigens in RA patients as compared with healthy control individuals [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. Another mechanism may reside in the partial tolerization to CII that can occur <italic>in vivo</italic>, as reported in transgenic mice in which the DNA sequence of an immunodominant peptide from rat CII was expressed in the mouse CII gene [<xref ref-type="bibr" rid="B9">9</xref>]. In that study, T-cell responses to the rat CII peptide was measured after immunization of rat CII. Proliferative responses were not detected but peptide-induced IFN-&#x003b3; production was evident <italic>in vitro,</italic> indicating that the mouse was tolerized to the antigen but that this tolerization only affected the proliferative response.</p><p>The strategy behind the present work has accordingly been to reinvestigate T-cell reactivity to CII in RA patients, taking the potential partial tolerization of CII-specific T cells and the general hyporesponsiveness of RA T cells into consideration. We investigated IFN-&#x003b3; responses after CII stimulation of peripheral blood mononuclear cells (PBMCs) <italic>in vitro</italic> and compensated for the general T-cell hyporesponsiveness by expressing the CII response as a fraction of the response to standard recall antigens. We also investigated possible associations of the recorded cellular reactivity to CII with HLA genotypes. Finally, we compared CII-induced IFN-&#x003b3; responses in RA patients with and without antibodies to CII.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patients</title><p>Initially 400 RA patients, including both those with early arthritis and those with more advanced stages of disease, were screened for the occurrence of high levels of anti-CII antibodies. From this screening 23 patients were identified as having more than 15U/ml anti-CII antibodies [for definition of these units, see separate method description of CII enzyme-linked immunosorbent assay (ELISA)]. These patients were asked to return and participate in the present study. Of these 19 also had elevated anti-CII levels (&#x0003e; 9U/ml) at inclusion into the study and comprised the group of RA patients with high anti-CII levels (median age 54 years, range 31&#x02013;86 years). In addition, 26 RA patients (median age 55 years, range 23&#x02013;78 years) with low anti-CII levels were randomly chosen. All patients met the 1987 American College of Rheumatology classification criteria for RA [<xref ref-type="bibr" rid="B10">10</xref>]. Twenty-five healthy staff members of similar age and sex from the Department of Rheumatology, Karolinska Hospital, served as healthy control individuals. Informed consent was obtained from all patients and control individuals included in the study. Patient characteristics are given in Table <xref ref-type="table" rid="T1">1</xref>.</p></sec><sec><title>Cell separation and stimulation</title><p>Peripheral blood was collected into heparinized tubes and diluted 1 : 2 with phosphate buffered saline (PBS). Mononuclear cells were isolated by density gradient centrifugation and diluted to 1 &#x000d7; 10⁶/ml in RPMI-1640 (Flow Laboratories, Irvine, Scotland, UK) supplemented with glutamine, HEPES buffer (Life Technologies, Paisley, Scotland, UK), penicillin, streptomycin and 10% of a defined batch of fetal calf serum (Flow Laboratories). Chick CII (Sigma, St Louis, MO, USA) or human CII [a kind gift from Alvar Gr&#x000f6;nberg (formerly Pharmacia), Uppsala, Sweden] diluted in 0.1 mol/l acetic acid, or just 0.1 mol/l acetic acid as a buffer control, was added to 1 ml of the cell suspension, giving a final concentration of 100 &#x003bc; g/ml of CII and 2 mmol/l acetic acid. When stimulating with denatured CII, a stock solution of CII was incubated at 56&#x000b0;C for 30 min and then added to the cell suspension.</p><p>As standard recall antigens, purified protein derivative (PPD; Statens Smittskyddsinstitut, Solna, Sweden; final concentration 10 &#x003bc; g/ml) and killed influenza virus (Vaxigrip, Pasteur M&#x000e9;rieux, Lyon, France; diluted 1:1000) was added to 1 ml of the cell suspension.</p><p>For blocking of HLA class II antigens, three different sets of antibodies were used in parallel: a mixture of 1 &#x003bc; g/ml of each of three monoclonal antibodies: antibody 2.06 [anti-DR, mouse immunoglobulin (Ig)G1], IVA12 (anti-DR, DP, mouse IgG1) and 9.3F10 (anti-DR, DQ, mouse IgG2a; all from ATCC, Rockville, Maryland, USA). This mixture of anti-class II antibodies has earlier been described to specifically block antigen-induced cytokine responses [<xref ref-type="bibr" rid="B20">20</xref>]. As control, an isotype-weighted mixture of the antikeyhole limpet haemocyanin monoclonal antibodies HS (mouse IgG1 [<xref ref-type="bibr" rid="B21">21</xref>]) and 7B4 (mouse IgG2a [<xref ref-type="bibr" rid="B21">21</xref>]) were used. In a second set of experiments, 10 &#x003bc; g/ml of L243 (mouse IgG1; ATCC [<xref ref-type="bibr" rid="B22">22</xref>]) were used with HS as control antibody. In some confirmatory experiments 10 &#x003bc; g/ml of a F(ab)' 2 rabbit polyclonal antibody against human class II^ antigens were used with rabbit F(ab)' 2 antibody against human IgG as control [<xref ref-type="bibr" rid="B23">23</xref>].</p><p>Cells were incubated in round bottomed 96-well plates (NUNC A/S, Roskilde, Denmark) for 7 days using a protocol that has been optimized for antigen induced IFN-&#x003b3; detection in supernatants ([<xref ref-type="bibr" rid="B24">24</xref>], our unpublished data). Supernatants were collected and frozen at -20&#x000b0;C for later analyses of cytokine content.</p></sec><sec><title>IFN-&#x003b3; measurements</title><p>ELISA plates (Maxisorp; NUNC A/S) were coated with 50 &#x003bc; l of 2 &#x003bc; g/ml catcher antibody (1-D1K; MabTech, Stockholm, Sweden) in PBS overnight at 4&#x000b0;C. After blocking the plates with 100 &#x003bc; l PBS + 1% bovine serum albumin for 1 h at room temperature, the plates were washed with PBS + 0.05% Tween. Recombinant IFN-&#x003b3; (R&#x00026;D Systems, Minneapolis, MN, USA) was diluted in medium and 50 &#x003bc; l of samples and cytokine standards were added in duplicates. The plates were incubated for 4 h at room temperature, washed and 50 &#x003bc; l/well of a secondary biotinylated antibody (7-B6-1; MabTech) diluted to 1 &#x003bc; g/ml in PBS + 1% bovine serum albumin + 0.05% Tween was added. Incubation was done at 4&#x000b0;C overnight. After washing, the plates were incubated for 1 h at room temperature with 50 &#x003bc; l/well avidine-alkaline phosphatase (Dakopatts, Glostrup, Denmark) diluted 1 : 1000 in PBS + 0.05% Tween. After washing, 50 &#x003bc; l/well substrate [<italic>p</italic>-nitrophenyl-phosphate tablets (Sigma) 1 mg/ml in diethanolamine buffer, pH9.8] was added and the reaction was read at 405 nm in a spectrophotometer. To correct for the general low T-cell reactivity seen in RA PBMCs (Fig. <xref ref-type="fig" rid="F1">1a</xref>), the CII-induced IFN-&#x003b3; production was expressed as a fraction of the PPD or killed influenza virus-induced IFN-&#x003b3; production obtained in parallel PBMC cultures.</p></sec><sec><title>Serum anticollagen type II detection by enzyme-linked immunosorbent assay</title><p>ELISA plates (Maxisorp; NUNC S/A) were coated with 100 &#x003bc; l/well with native chick CII (Sigma) diluted to 5 &#x003bc; g/ml in ice-cold PBS overnight at 4&#x000b0;C. After blocking with PBS + 1% bovine serum albumin and washing with PBS + 0.05% Tween, 100 &#x003bc; l/well of a high-titre standard serum (added in a dilution series) and the samples, diluted at least 1 : 5 in PBS + 1% bovine serum albumin, were added in duplicate and incubated for 2 h at room temperature. After washing, 100 &#x003bc; l of a biotinylated goat antihu-man IgG antibody (Tago, Burlingame, CA, USA) diluted to 1 &#x003bc; g/ml in PBS + 1% bovine serum albumin was added and incubated overnight at 4&#x000b0;C. The ELISA was developed as above. Elevated levels of IgG anti-CII were defined as mean &#x000b1; two standard deviations of sera from 39 healthy controls (&#x0003e; 9U/ml). The standard serum was arbitrarily defined as having 1500 U/ml.</p><sec><title>Human leucocyte antigen genotyping</title><p>HLA-DR and HLA-DQ genotyping was performed utilizing a polymerase chain reaction-based technique with sequence-specific oligonucleotide probes hybridization [<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>]. DRB1*01 subtypes were identified by allele-specific polymerase chain reaction primers [<xref ref-type="bibr" rid="B27">27</xref>].</p></sec></sec><sec><title>Statistical analysis</title><p>Nonparametric methods were used throughout the study. Differences between groups were analyzed using the Mann&#x02013;Whitney U-test, and analyses for matched pairs were performed using Wilcoxon's signed-rank test. When analyzing correlations, Spearman rank correlation was used. <italic>P</italic> &#x0003c;0.05 was considered statistically significant.</p></sec></sec><sec><title>Results</title><p>PBMCs from RA patients and healthy control individuals were challenged <italic>in vitro</italic> with native and denatured CII, and supernatants from these cultures were investigated for IFN-&#x003b3; content. In both groups, an increased IFN-&#x003b3; production was recorded after <italic>in vitro</italic> culture with both native and denatured CII as compared with cultures with no added antigen, giving a median stimulation index with native CII of 4.6 for RA patients and 5.4 for healthy control individuals, and with denatured CII of 2.9 for RA patients and 3.0 for healthy control individuals. No difference in IFN-&#x003b3; production in response to CII could be demonstrated between RA patients and healthy control individuals (Fig. <xref ref-type="fig" rid="F1">1a</xref>). As control, PBMCs from 12 RA patients and seven healthy control individuals were stimulated with pepsin (10 &#x003bc;g/ml). No IFN-&#x003b3; induction was evident with this stimulation (data not shown).</p><p>Parallel cultures were also stimulated with the standard recall antigens PPD and killed influenza virus. These antigens also induced the production of IFN-&#x003b3; in PBMCs from both RA patients and healthy control individuals, giving median stimulation indexes with PPD of 10.0 (RA patients) and 51.3 (healthy control individuals), and with killed influenza virus of 12.3 (RA patients) and 25.7 (healthy control individuals). For both PPD and influenza stimulations there was a significant difference between the RA group and the healthy control group in that the RA patients displayed a markedly lower responsiveness to PPD and to influenza (Fig. <xref ref-type="fig" rid="F1">1b</xref>).</p><p>In order to investigate whether the observed production of IFN-&#x003b3; in response to stimulation with CII, PPD and influenza was mediated by MHC class II-dependent T-cell activation, blocking antibodies were added to the cell cultures. As shown in Table <xref ref-type="table" rid="T2">2</xref> these antibodies inhibited the CII as well as the PPD-induced and killed influenza virus-induced IFN-&#x003b3; production. No such inhibitory effects were observed after incubation with isotype-matched control monoclonal antibodies.</p><p>As stated in the introduction to this report, the lower responses to PPD and killed influenza virus in RA patients than in healthy control individuals reflect a general down-regulation of antigen-induced responsiveness of T cells from RA patients [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. That no difference between the RA group and the control group was recorded in CII-induced IFN-&#x003b3; production therefore indicates that there may be an underlying increased specific and MHC class II-dependent responsiveness to CII in the RA patients, which is obscured by the general downregulation of T-cell responsiveness. In order to address this possibility, we calculated the fraction between individual values for the CII-induced IFN-&#x003b3; production and the PPD-induced and killed influenza virus-induced IFN-&#x003b3; production, and compared these fractions.</p><p>When compensating for the general T-cell hyporesponsiveness in RA-derived T cells in this manner, using the PPD response as the denominator, a highly significant difference between the RA group and the healthy control individuals was apparent after stimulation with both native CII and denatured CII (Fig. <xref ref-type="fig" rid="F2">2a</xref>). Similar data were obtained using killed influenza virus-stimulated IFN-&#x003b3; values as the denominator, but no statistical difference between RA patients and healthy control individuals could be determined for native CII (Fig. <xref ref-type="fig" rid="F2">2b</xref>).</p><p>We next investigated whether there was a difference between patients with elevated serum levels of anti-CII antibodies as compared with other RA patients with regard to the T-cell response to CII. Analyzing uncompensated IFN-&#x003b3; production in response to CII stimulation revealed, contrary to our expectations, that RA patients with high levels of anti-CII had lower IFN-&#x003b3; responses to denatured CII compared with those of RA patients with low levels of anti-CII (Fig. <xref ref-type="fig" rid="F4">4a</xref>). A similar tendency, but not statistically significant, could also be demonstrated using native CII. There was no difference in the PPD-induced or killed influenza virus-induced IFN-&#x003b3; responses between RA patients with high and low anti-CII levels (data not shown).</p><p>The results obtained after compensating for the T-cell hyporesponsiveness using PPD as the denominator demonstrate that both the RA patients with low anti-CII antibody levels and those with high levels had significantly greater responses to native CII than did healthy control individuals (Fig. <xref ref-type="fig" rid="F4">4b</xref>). For the group with low anti-CII levels there was also a highly significant increased response to denatured CII as compared with healthy control individuals, which could not be observed for patients with high anti-CII levels. Similar results were observed using influenza-induced IFN-&#x003b3; as the denominator (data not shown).</p><p>We also correlated the responsiveness to CII to various clinical parameters (C-reactive protein levels, presence of erosions as measured by radiography, numbers of swollen and tender joints, and disease duration). No correlations between C-reactive protein, presence of erosions or number of swollen joints and the compensated IFN-&#x003b3; response to CII could be demonstrated; the only significant association noted was a weak negative relationship between number of tender joints and the PPD-compensated CII-reactivity (<italic>P</italic> = 0.009, <italic>r</italic>s = -0.46, data not shown). The RA group with high levels of anti-CII had a significantly higher number of swollen joints than the RA group with low levels of anti-CII (<italic>P</italic> = 0.029, data not shown). A positive correlation was noted between disease duration and responsiveness to native or denatured CII (compensated values; <italic>P</italic> = 0.036, <italic>r</italic>s = 0.33 for native CII and <italic>P</italic> = 0.002, <italic>r</italic>s = 0.49 for denatured CII).</p><p>Finally, we compared the compensated IFN-&#x003b3; response to denatured CII stimulation between RA patients with different HLA genotypes. Of the RA patients included in this study, 13 were DRB1*0401 positive and 15 were DQ8 positive. Out of these patients, 12 were DRB1*0401/DQ8 double positive. Of the healthy control individuals, nine were DRB1*0401 positive and five were DQ8 positive (four were not genotyped for HLA-DQ). Out of these healthy control individuals, four were DRB1*0401/DQ8 double positive. Highly significant differences were evident, with HLA-DRB1*0401 patients having higher CII responsiveness than patients lacking this genotype (Fig. <xref ref-type="fig" rid="F3">3a</xref>). HLA-DQ8-positive patients also displayed a high responsiveness to CII as compared with HLA-DQ8-negative RA patients (Fig. <xref ref-type="fig" rid="F3">3b</xref>). Similar results were achieved using killed influenza virus as the denominator (<italic>P</italic> = 0.02 for HLA-DRB1*0401 and <italic>P</italic> = 0.01 for HLA-DQ8). These associations between the relative T-cell reactivity to denatured CII and HLA class II genotypes were not seen in healthy control individuals. No significant relationship between the analyzed HLA-DR or DQ haplotypes and responsiveness to CII remained when the two different groups of RA patients were analyzed separately. No association was detected between serum levels of anti-CII and the shared epitope (DRB1*0401 or DRB1*0404).</p></sec><sec><title>Discussion</title><p>The major findings reported in this paper are as follows. First, an increased IFN-&#x003b3; production in response to <italic>in vitro</italic> stimulation with CII in RA patients as compared with age-matched and sex-matched healthy control individuals when compensation was made for the general hyporesponsiveness of RA patient PBMCs to nominal antigens. Second, the response to CII was dependent on MHC class II molecules, as demonstrated by the blockade of the response with anti-MHC class II antibodies. Finally, RA patients positive for the genotypes HLA-DRB1^*0401 or HLA-DQ8 exhibited a higher IFN-&#x003b3; response than did RA patients without these alleles.</p><p>The choice of IFN-&#x003b3; production as the parameter for registration of the cellular response to CII was made both from experience in mice immunized with CII and from multiple sclerosis. In mice immunized with autologous CII, an autoimmune arthritis-provoking response may occur that is associated with <italic>in vivo</italic> antibody production to CII and with CII-induced IFN-&#x003b3; production <italic>in vitro,</italic> but not with any proliferative response to CII [<xref ref-type="bibr" rid="B9">9</xref>]. In multiple sclerosis patients, <italic>in vitro</italic> production of T-cell cytokines such as IFN-&#x003b3; is increased after <italic>in vitro</italic> stimulation with myelin antigens such as myelin basic protein and myelin-oligodendrocyte glyco-protein, despite the difficulties in detecting proliferative responses towards these antigens [<xref ref-type="bibr" rid="B28">28</xref>].</p><p>Although there is much previous support for the choice of IFN-&#x003b3; as a readout for T-cell responsiveness to antigens, thus taking the potential partial anergy to CII into account, no reports have previously systematically taken the general T-cell hyporesponsiveness in RA into account when investigating specific T-cell responses in this disease. In the present study we used the T-cell responses to PPD and killed influenza virus as reference antigens for making this compensation. This was based on the assumption that exposure to these antigens is similar in age-matched and sex-matched groups of RA patients and healthy control individuals. The concept of a general hyporesponsiveness of RA T cells has been documented in several previous reports in which both nominal antigens [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>] and mitogens [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>] were used. Even though the mechanisms behind this defective T-cell response are not clear, it appears that the hyporesponsiveness is a feature of the disease itself because it has been observed also in patients who are not on disease-modifying antirheumatic drug therapy [<xref ref-type="bibr" rid="B6">6</xref>]. The fact that a similar functional down-regulation in RA PBMCs was obtained with both PPD and killed influenza virus as reference antigens strengthens the validity of our approach.</p><p>Both native and heat-denatured CII were used in the present study on the basis that native and denatured CII may be processed in different ways by antigen-presenting cells. As shown in Figures <xref ref-type="fig" rid="F1">1a</xref> and <xref ref-type="fig" rid="F4">4a</xref>, native CII induced more IFN-&#x003b3; than did denatured CII, indicating an important role for the structure of CII in cellular stimulation. The difference in the amount of IFN-&#x003b3; induced by native and denatured CII might depend on differences in uptake or processing of these structurally different CII molecules.</p><p>Proliferation in response to <italic>in vitro</italic> CII stimulation of cells from both immunized animals and in human studies have in some instances been found to be a response to contaminating pepsin in the CII preparation [<xref ref-type="bibr" rid="B29">29</xref>]. This possibility was excluded in the present study by control experiments that failed to demonstrate any IFN-&#x003b3; production in response to pepsin (data not shown).</p><p>RA patients as well as healthy control individuals responded with IFN-&#x003b3; production <italic>in vitro</italic> when PBMCs were stimulated with chick or human CII. It has not directly been demonstrated that the IFN-&#x003b3;-producing cells are T cells, but the fact that blocking of MHC class II blocks the IFN-&#x003b3; response by RA PBMCs as well as by control PBMCs indicates that this is indeed the case. We also attempted to identify induction of a type 2 cytokine, but failed to detect interleukin-4 in CII stimulated cultures using methods that are capable of detecting inter-leukin-4 in cultures stimulated with a recall antigen.</p><p>We identified an association between the magnitude of the IFN-&#x003b3; response to CII and HLA-DRB1*0401 and HLA-DQ8 in the RA group, which is of obvious interest because both these MHC class II alleles have been associated with high responsiveness to CII in transgenic mice that express these human MHC class II molecules [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. Thus far, susceptibility to RA has mainly been associated with HLA-DRB1*0401 and with other MHC class II molecules carrying the 'shared epitope' [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>], whereas it remains unclear whether HLA-DQ8 has an independent association with RA. In <italic>in vitro</italic> studies of cellular responses to CII in RA patients, some reports have claimed that cellular reactivity to CII (measured as CII-induced proliferation or production of factors capable of stimulation of leucocyte functions) is associated with HLA-DR4 [<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>,<xref ref-type="bibr" rid="B34">34</xref>]. Others, however, have not found this association [<xref ref-type="bibr" rid="B35">35</xref>,<xref ref-type="bibr" rid="B36">36</xref>]. Anti-CII-producing B cells have been reported to be present in the joint [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>] and have in some instances been associated with HLA-DR4 [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. It is difficult to separate the influences of HLA-DR4 and HLA-DQ8 in studies of the association between MHC class II alleles and RA or in <italic>in vitro</italic> associations between cellular reactivity to CII and MHC class II alleles, because of the strong linkage disequilibrium between DRB1*0401 and HLA-DQ8. Finally, it should be noted that the RA patients in the present study were recruited in two different ways; 19 patients were selected on the basis of their high anti-CII antibody levels and 26 patients with low anti-CII levels were randomly recruited from our outpatient clinic. The HLA associations to cellular CII reactivity were only seen when both groups were combined, probably because of the rather low numbers of patients in each group. It appears unlikely to us that the association between HLA haplotype and CII-induced T-cell responsiveness seen in the combined group should be due to a bias introduced by selection of patients with high anti-CII levels for the following reasons: the high anti-CII group had a lower responsiveness to denatured CII than the randomly chosen group of RA patients; and there was no association between anti-CII levels and shared epitope (HLA-DRB1*0401 or HLA-DRB1*0404).</p><p>The IFN-&#x003b3; reactivity to CII seen in most individuals implies that T-cell tolerance to CII is incomplete. The finding that high cellular response to CII associates with HLA-DRB1*0401 and HLA-DQ8 could mean that selection in thymus favours a T-cell repertoire containing T-cell receptors with high specificity for some CII epitope(s). Another explanation for the HLA-associated cellular response to CII is that these class II molecules could be able to present some immunodominant peptides more efficiently than other class II molecules.</p><p>One recent study [<xref ref-type="bibr" rid="B37">37</xref>] has investigated CII-specific proliferative responses and their relation to the presence of circulating anti-CII antibodies in RA patients. That study could not demonstrate a difference in proliferative response to CII between RA patients and healthy control individuals. In contrast to the present results, however, a lower frequency of responsive RA patients without anti-CII compared with that in RA patients with anti-CII was recorded, which could not be explained by a difference in disease-modifying antirheumatic drug therapy between the two patient groups. This contradiction could be explained by the definition of anergy, as suggested by Mueller <italic>et al</italic> [<xref ref-type="bibr" rid="B38">38</xref>], that proliferative responses are lost whereas effector functions are retained in anergized T cells. Hence, T cells in RA patients without anti-CII might be anergic in the proliferative aspect but are still able to respond by cytokine production.</p><p>It has been shown by Cook <italic>et al</italic> [<xref ref-type="bibr" rid="B39">39</xref>] that levels of anti-CII in serum of RA patients decreases over time. This tendency was also apparent in the present study, but it was not designed to address this question (data not shown). We demonstrated a positive correlation between disease duration and IFN-&#x003b3; response to CII, indicating that the T-cell response to CII (expressed as a fraction of the PPD-induced IFN-&#x003b3; response) increases with time. As the PPD-induced IFN-&#x003b3; production decreases with disease duration (data not shown), this implies that general T-cell reactivity decreases gradually during the course of RA. CII-induced IFN-&#x003b3; production remains stable, however, and it increases when expressing the CII cellular reactivity as a fraction of the PPD cellular reactivity. Consequently, although humoral CII reactivity decreases with time [<xref ref-type="bibr" rid="B39">39</xref>], cellular CII reactivity increases. It is thus plausible that serum antibody levels to CII constitute a relatively poor mirror of T-cell responsiveness to CII. Because anti-CII antibodies and T cells reactive with CII may synergize in causing arthritis [<xref ref-type="bibr" rid="B40">40</xref>] (at least in rodents), there is an obvious need for further detailed parallel studies of B-cell and T-cell reactivity to CII in RA, taking the fine specificity (not addressed in the present study) into account.</p><p>In conclusion, CII, a major autoantigen candidate in RA, can elicit an IFN-&#x003b3; response <italic>in vitro</italic> that is associated with HLA-DRB1*0401 and HLA-DQ8 in RA patients. The present study, with a partly new methodological approach to a classical problem in RA, has provided some additional support to the notion that CII may be a target autoantigen of importance for a substantial group of RA patients. Continued efforts to identify mechanisms behind the general hyporesponsiveness to antigens in RA, as well as the mechanisms behind the potential partial anergy to CII, may provide us with better opportunities to study the specificity and pathophysiological relevance of anti-CII reactivity in RA.</p></sec>

Acute-phase serum amyloid A production by rheumatoid arthritis synovial tissue

<sec><title>Introduction:</title><p>Serum amyloid A (SAA) is the circulating precursor of amyloid A protein, the fibrillar component of amyloid deposits. In humans, four SAA genes have been described. Two genes (<italic>SAA1</italic> and <italic>SAA2</italic>) encode A-SAA and are coordinately induced in response to inflammation. <italic>SAA1</italic> and <italic>SAA2</italic> are 95% homologous in both coding and noncoding regions. <italic>SAA3</italic> is a pseudogene. <italic>SAA4</italic> encodes constitutive SAA and is minimally inducible. A-SAA increases dramatically during acute inflammation and may reach levels that are 1000-fold greater than normal. A-SAA is mainly synthesized in the liver, but extrahepatic production has been demonstrated in many species, including humans. A-SAA mRNA is expressed in RA synoviocytes and in monocyte/macrophage cell lines such as THP-1 cells, in endothelial cells and in smooth muscle cells of atherosclerotic lesions. A-SAA has also been localized to a wide range of histologically normal tissues, including breast, stomach, intestine, pancreas, kidney, lung, tonsil, thyroid, pituitary, placenta, skin and brain.</p></sec><sec><title>Aims:</title><p>To identify the cell types that produce A-SAA mRNA and protein, and their location in RA synovium.</p></sec><sec><title>Materials and methods:</title><p>Rheumatoid synovial tissue was obtained from eight patients undergoing arthroscopic biopsy and at joint replacement surgery. Total RNA was analyzed by reverse transcription (RT) polymerase chain reaction (PCR) for A-SAA mRNA. PCR products generated were confirmed by Southern blot analysis using human A-SAA cDNA. Localization of A-SAA production was examined by immunohistochemistry using a rabbit antihuman A-SAA polyclonal antibody. PrimaryRA synoviocytes were cultured to examine endogenous A-SAA mRNA expression and protein production.</p></sec><sec><title>Results:</title><p>A-SAA mRNA expression was detected using RT-PCR in all eight synovial tissue samples studied. Figure <xref ref-type="fig" rid="F1">1</xref> demonstrates RT-PCR products generated using synovial tissue from three representative RA patients. Analysis of RA synovial tissue revealed differences in A-SAA mRNA levels between individual RA patients.</p><p>In order to identify the cells that expressed A-SAA mRNA in RA synovial tissue, we analyzed primary human synoviocytes (<italic>n</italic> = 2). RT-PCR analysis revealed A-SAA mRNA expression in primary RA synoviocytes (<italic>n</italic> = 2; Fig. <xref ref-type="fig" rid="F2">2</xref>). The endogenous A-SAA mRNA levels detected in individual primary RA synoviocytes varied between patients. These findings are consistent with A-SAA expression in RA synovial tissue (Fig. <xref ref-type="fig" rid="F2">1</xref>). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels were relatively similar in the RA synoviocytes examined (Fig. <xref ref-type="fig" rid="F2">2</xref>). A-SAA protein in the supernatants of primary synoviocyte cultures from four RA patients was measured using ELISA. Mean values of a control and four RA samples were 77.85, 162.5, 249.8, 321.5 and 339.04 &#x003bc;g/l A-SAA, respectively, confirming the production of A-SAA protein by the primary RA synoviocytes. Immunohistochemical analysis was performed to localize sites of A-SAA production in RA synovial tissue. Positive staining was present in both the lining and sublining layers of all eight RA tissues examined (Fig. <xref ref-type="fig" rid="F2">3a</xref>). Staining was intense and most prominent in the cells closest to the surface of the synovial lining layer. Positively stained cells were evident in the perivascular areas of the sublining layer. In serial sections stained with anti-CD68 monoclonal antibody, positive staining of macrophages appeared to colocalize with A-SAA-positive cells (Fig. <xref ref-type="fig" rid="F2">3b</xref>). Immunohistochemical studies of cultured primary RA synoviocytes confirmed specific cytoplasmic A-SAA expression in these cells. The specificity of the staining was confirmed by the absence of staining found on serial sections and synoviocyte cells treated with IgG (Fig. <xref ref-type="fig" rid="F2">3c</xref>). </p></sec><sec><title>Discussion:</title><p>This study demonstrates that A-SAA mRNA is expressed in several cell populations infiltrating RA synovial tissue. A-SAA mRNA expression was observed in all eight unseparated RA tissue samples studied. A-SAA mRNA expression and protein production was demonstrated in primary cultures of purified RA synoviocytes. Using immunohistochemical techniques, A-SAA protein appeared to colocalize with both lining layer and sublining layer synoviocytes, macrophages and some endothelial cells. The detection of A-SAA protein in culture media supernatants harvested from unstimulated synoviocytes confirms endogenous A-SAA production, and is consistent with A-SAA mRNA expression and translation by the same cells. Moreover, the demonstration of A-SAA protein in RA synovial tissue, RA cultured synoviocytes, macrophages and endothelial cells is consistent with previous studies that demonstrated A-SAA production by a variety of human cell populations.</p><p>The RA synovial lining layer is composed of activated macrophages and fibroblast-like synoviocytes. The macrophage is the predominant cell type and it has been shown to accumulate preferentially in the surface of the lining layer and in the perivascular areas of the sublining layer. Nevertheless, our observations strongly suggest that A-SAA is produced not only by synoviocytes, but also by synovial tissue macrophage populations. Local A-SAA protein production by vascular endothelial cells was detected in some, but not all, of the tissues examined. The reason for the variability in vascular A-SAA staining is unknown, but may be due to differences in endothelial cell activation, events related to angiogenesis or the intensity of local inflammation.</p><p>The value of measuring serum A-SAA levels as a reliable surrogate marker of inflammation has been demonstrated for several diseases including RA, juvenile chronic arthritis, psoriatic arthropathy, ankylosing spondylitis, Beh&#x000e7;et's disease, reactive arthritis and Crohn's disease. It has been suggested that serum A-SAA levels may represent the most sensitive measurement of the acute-phase reaction. In RA, A-SAA levels provide the strongest correlations with clinical measurements of disease activity, and changes in serum levels best reflect the clinical course.</p><p>A number of biologic activities have been described for A-SAA, including several that are relevant to the understanding of inflammatory and tissue-degrading mechanisms in human arthritis. A-SAA induces migration, adhesion and tissue infiltration of circulating monocytes and polymorphonuclear leukocytes. In addition, human A-SAA can induce interleukin-1&#x003b2;, interleukin-1 receptor antagonist and soluble type II tumour necrosis factor receptor production by a monocyte cell line. Moreover, A-SAA can stimulate the production of cartilage-degrading proteases by both human and rabbit synoviocytes. The effects of A-SAA on protease production are interesting, because in RA a sustained acute-phase reaction has been strongly associated with progressive joint damage. The known association between the acute-phase response and progressive joint damage may be the direct result of synovial A-SAA-induced effects on cartilage degradation.</p></sec><sec><title>Conclusion:</title><p>In contrast to noninflamed synovium, A-SAA mRNA expression was identified in all RA tissues examined. A-SAA appeared to be produced by synovial tissue synoviocytes, macrophages and endothelial cells. The observation of A-SAA mRNA expression in cultured RA synoviocytes and human RA synovial tissue confirms and extends recently published findings that demonstrated A-SAA mRNA expression in stimulated RA synoviocytes, but not in unstimulated RA synoviocytes.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>O'Hara</surname><given-names>Rosemary</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Murphy</surname><given-names>Evelyn P</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Whitehead</surname><given-names>Alexander S</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>FitzGerald</surname><given-names>Oliver</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Bresnihan</surname><given-names>Barry</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>b.bresnihan@svcpc.ie</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>SAA is the circulating precursor of amyloid A protein, the fibrillar component of amyloid deposits [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Four human SAA genes have been described. Two genes (<italic>SAA1</italic> and <italic>SAA2</italic>) encode A-SAA and are coordinately induced in response to inflammation. <italic>SAA1</italic> and <italic>SAA2</italic> are 95% homologous in both coding and noncoding regions. <italic>SAA3</italic> is a pseudogene. <italic>SAA4</italic> encodes constitutive SAA and is minimally inducible. A-SAA increases dramatically during acute inflammation and may reach levels that are 1000-fold greater than normal [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. A-SAA is mainly synthesized in the liver [<xref ref-type="bibr" rid="B5">5</xref>], but extrahepatic production has been demonstrated in many species, including in humans. A-SAA mRNA is expressed in RA synoviocytes [<xref ref-type="bibr" rid="B6">6</xref>] and in monocyte/macrophage cell lines such as THP-1 cells (Human monocyte cell line American Type Culture Collection Tumor Immunology bank 17 - ATCC TIB) [<xref ref-type="bibr" rid="B7">7</xref>], in endothelial cells and smooth muscle cells of atherosclerotic lesions [<xref ref-type="bibr" rid="B8">8</xref>]. A-SAA has also been localized to a wide range of histologically normal tissues, including breast, stomach, intestine, pancreas, kidney, lung, tonsil, thyroid, pituitary, placenta, skin and brain [<xref ref-type="bibr" rid="B9">9</xref>].</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patients</title><p>Synovial membrane samples were obtained following informed consent from patients with RA at arthroscopic biopsy (<italic>n</italic> = 7) or joint replacement surgery (<italic>n</italic> = 1). RA was diagnosed according to the criteria of the American College of Rheumatology [<xref ref-type="bibr" rid="B10">10</xref>]. Histologically normal synovium (<italic>n</italic> = 1) was obtained from the knee joint of a patient undergoing lower limb amputation.</p></sec><sec><title>Isolation and culture of synovial cells</title><p>Synovial cells were obtained by enzymatic digestion of synovial membrane with 1mg/ml collagenase type I (Worthington Biochemical, Freehold, NJ, USA) in RPMI (GibcoBRL, Paisley, UK) for 4 h at 37&#x000b0;C in 5% carbon dioxide. Dissociated cells were plated in RPMI supplemented with 10% foetal calf serum (GibcoBRL), 10 ml of 1 mmol/l HEPES (GibcoBRL), penicillin (100 units/ml), streptomycin (100 units/ml) and fungizone (0.25 &#x003bc;g/ml). The cells were grown to confluency (approximately 10days) at 37&#x000b0;C in a 5% carbon dioxide atmosphere, then harvested with trypsin and passaged. Synoviocytes were found to be morphologically homogenous fibroblast-like cells and were used between the third and seventh passage. To confirm synoviocyte cultures were not contaminated by monocytes, staining for the monocyte marker CD14 was carried out. Cells were placed in serum-free medium 24 h before total RNA extraction.</p></sec><sec><title>Reverse transcription-polymerase chain reaction</title><p>Total RNA, isolated from freshly obtained synovial biopsies or cultured primary synoviocytes, was converted by RT into cDNA. For each total RNA sample, the following materials were used for RT at 42&#x000b0;C for 50 min: 1 &#x003bc;g total RNA and 200units SuperscriptII (GibcoBRL); RT buffer as supplied; 100 mmol/l dithiothreitol, 40units RNasin Ribonucleic Inhibitor (Promega, Madison, WI, USA); 1.25 mmol/l each of dATP, dCTP, dGTP and dTTP; and 500ng of oligo dTs. PCR was performed with the following materials: 2.0 &#x003bc;l cDNA; 1.25 mmol/l of each of dATP, dCTP, dGTP and dTTP; 2.5 units AmpliTaq (Perkin Elmer, Brachburg, NJ,USA); 2.5 mmol/l MgCl₂ (GAPDH) and 2.0 mmol/l MgCl₂(A-SAA); 2.5 &#x003bc;l 10 &#x000d7;PCRII buffer (Perkin Elmer) and 20 ng of each specific PCR primer pair in a 25 &#x003bc;l total volume. Specific primers for human A-SAA were used to amplify a 335 base-pair (bp) A-SAA product: sense primer (5' -AAG CTT CTT TCC GTT CCT TGG-3') and antisense primer(5' -GAG AGC AGA GTG AAG AGG AAG C-3'). The A-SAA primers used span an intron. Thus, the PCR generates an unequivocally RNA-derived band, based on its size. GAPDH primers were designed to generate a 635-bp product: sense primer (5' -CCA CCC ATG GCA AAT TCC ATG GCA-3') and antisense primer (5' -TCT AGA CGG CAG GTC AGG TCC ACC-3'). After preincubation (94&#x000b0;C, 10 min) each PCR sample underwent a 35-cycle amplification regimen of denaturation (94&#x000b0;C, 1 min), primer annealing (60-56&#x000b0;C, 1 min) and extension (72&#x000b0;C, 1 min), with a final extension (72&#x000b0;C, 10 min) in a thermal cycler (MJ Research, Inc, Cambridge, MA, USA).</p></sec><sec><title>Northern blot analysis</title><p>Total RNA was isolated (RNeasy, Qiagen, Crawley, UK) from cultured primary synoviocytes and quantified by ultraviolet absorption. Of total RNA, 10 &#x003bc;g was electrophoresed on a standard northern gel and transferred onto a nylon membrane (BioRad, Richmond, CA, USA). The human A-SAA cDNA [<xref ref-type="bibr" rid="B11">11</xref>] was radiolabelled to a high specific activity using [&#x003b1;-³²P] dCTP and a random primer labelling system (Promega). All membranes were probed under high stringency conditions. Blots were exposed to film at -80&#x000b0;C using intensifying screens and autoradiographic intensity was quantified using an imaging densitometer.</p></sec><sec><title>Southern blot analysis</title><p>PCR products generated were run on a 2% agarose gel and transferred onto a nylon membrane (BioRad) using standard procedures. Human A-SAA and GAPDH cDNA probes were radiolabelled to a high specific activity using [&#x003b1;-³²P] dCTP and a random labelling system (Promega). All membranes were probed under high stringency conditions. Blots were exposed to film at -80&#x000b0;C using intensifying screens.</p></sec><sec><title>Measurement of acute-phase serum amyloid A by ELISA</title><p>A-SAA protein levels were measured using a sandwich enzyme immunoassay (Biotrin International, Dublin, Ireland). Supernatants derived from primary RA synoviocyte cultures were harvested. Samples were added to a microtitre plate, which was precoated with IgG (anti-A-SAA) and incubated for 1 h. The microtitre plate was washed and IgG (anti-A-SAA)-horseradish peroxidase conjugate added. After 1 h incubation and plate washing, substrate was added. The absorbance was measured at 450nm, using 650nm as a reference. Colour intensity is directly proportional to the amount of A-SAA in the sample. The detection limit of the assay was determined as 2.25 &#x003bc;g/l [<xref ref-type="bibr" rid="B11">11</xref>].</p></sec><sec><title>Immunohistology</title><p>Synovial tissue was placed in the cryopreservative embedding media OCT compound (Tissue Tek, Sakura, Finetek, Europe BV, Zoeterwoude, The Netherlands) and immediately frozen in liquid nitrogen. Sections (7 &#x003bc; m) were cut on a microtome (Microm HM 505N, GmbH 69190 Walldorf, Germany), placed on glass slides coated with 2% 3-amino-propyl-triethoxy-silane (Sigma-Aldrich Ireland Ltd, Dublin, Ireland) in acetone and dried overnight at room temperature. Isolated RA synoviocytes were trypsinized and placed into a six-well plate with apyrogenic cell culture coverslips. Once grown to confluency, the medium was removed and the cells were treated with methanol for 15 min. Synoviocytes were stained essentially as described for the tissue sections. Tissue sections were allowed to reach room temperature, fixed in acetone, air-dried and incubated for 1 h at room temperature with blocking serum (Vectastain Rabbit Elite Kit, Vector Laboratories Ltd, Peterborough, UK). The slides were incubated with avidin for 15 min, rinsed and then incubated with biotin for 15 min. The primary polyclonal antibody for A-SAA (1:1200-1:1600; rabbit antihuman) was incubated for 1 h at room temperature. Secondary antibodies (antirabbit and antimouse; Vectastain) were prepared and added to the relevant sections and incubated for 30 min. The secondary antibody was washed off and the slides were incubated with Avidin:Biotinylated enzyme complex solution for 30 min and incubated for 6 min with 3,3' -diaminobenzidine and counterstained in haemotoxylin stain for 1 min.</p></sec></sec><sec><title>Results</title><sec><title>Acute-phase serum amyloid A mRNA in inflamed human synovial tissue</title><p>RT-PCR analysis was employed to examine peripheral A-SAA mRNA expression in RA synovial membrane. Endogenous A-SAA mRNA expression was detected in all eight RA synovial tissue samples studied. Figure <xref ref-type="fig" rid="F1">1</xref> demonstrates RT-PCR products generated using synovial tissue from three representative RA patients. The specificity of the cDNAs generated were confirmed by Southern blot analysis using cDNA probes for human A-SAA and GAPDH (Fig. <xref ref-type="fig" rid="F2">4</xref>). RA synovial tissue samples examined showed increased levels of A-SAA mRNA when compared with normal synovium (Fig. <xref ref-type="fig" rid="F2">4</xref>; lane 1). Analysis of RA synovial tissue revealed differences in A-SAA mRNA levels between individual RA patients. Expression levels of the house-keeping gene (GAPDH) were similar in all patients.</p></sec><sec><title>Acute-phase serum amyloid A mRNA expression in cultured human synovial cells</title><p>In order to identify the cells that express A-SAA mRNA in RA synovial tissue, we analyzed primary human synoviocytes. Northern blot analysis did not detect endogenous A-SAA mRNA in primary unstimulated RA synoviocytes (Fig. <xref ref-type="fig" rid="F2">5a</xref>; lane 1). In contrast, abundant A-SAA mRNA levels were observed in KB oral epidermal cells [Human Epidermoid, carcinoma; American Type Culture Collection Certified cell lines (ATCC CCL) 17] stimulated with interleukin-1 (10 ng/ml) and interleukin-6 (10 ng/ml) and dexamethasone (10^-6 mol/l) for either 24 or 48 h (Fig. <xref ref-type="fig" rid="F2">5a</xref>; lanes 2 and 3). RT-PCR analysis revealed A-SAA mRNA expression in primary RA synoviocytes (<italic>n</italic> = 2; Fig. <xref ref-type="fig" rid="F2">2</xref>). The sensitivity of the PCR technique was greatly improved by the utilization of the ultrapure AmpliTaq gold polymerase (Perkin Elmer). The endogenous A-SAA mRNA levels detected in individual primary RA synoviocytes varied between patients. These findings are consistent with A-SAA expression in RA synovial tissue (Fig. <xref ref-type="fig" rid="F2">1</xref>). GAPDH levels were relatively similar in the RA synoviocyte cells examined (Fig. <xref ref-type="fig" rid="F2">2</xref>).</p><p>Acute-phase serum amyloid A protein in primary synoviocyte culture supernatants</p><p>A-SAA protein was measured by ELISA in the supernatants of primary synoviocyte cultures from four RA patients. Supernatants were separated from cell cultures between the third and seventh passages. Mean values of control and four RA samples of 77.85, 162.5, 249.8, 321.5 and 339.04 &#x003bc;g/l A-SAA, respectively, were obtained, confirming the production of A-SAA protein by the primary RA synoviocytes (Fig. <xref ref-type="fig" rid="F2">6</xref>).</p></sec><sec><title>Immunohistochemical localization of acute-phase serum amyloid A in human synovial tissue and cultured synoviocytes</title><p>Immunohistochemical analysis was performed in order to localize sites of A-SAA production in RA synovial tissue.Positive A-SAA staining was observed in all eight RA tissues examined. Positive staining was present in both the lining and sublining layers of the tissue (Fig. <xref ref-type="fig" rid="F2">3a</xref>). In the synovial lining layer, staining was usually intense and most prominent in the cells closest to the surface. In the sublining layer positively stained cells were also prominent in the perivascular areas. In serial sections stained with anti-CD68 monoclonal antibody, positive staining of some macrophages appeared to colocalize with A-SAA-positive cells, particularly those in the lining layer surface and the perivascular areas (Fig. <xref ref-type="fig" rid="F2">3b</xref>). In some sections the vascular endothelium also demonstrated positive A-SAA staining, but this finding was not consistent. Immunohistochemical studies of cultured primary RA synoviocytes confirmed specific cytoplasmic A-SAA expression in these cells (Fig. <xref ref-type="fig" rid="F2">7a</xref>). The specificity of the staining was confirmed by the absence of staining found on serial sections (Fig. <xref ref-type="fig" rid="F2">3c</xref>) and synoviocytes (Fig. <xref ref-type="fig" rid="F2">7b</xref>) treated with IgG.</p></sec></sec><sec><title>Discussion</title><p>The aim of the present study was to examine synovial A-SAA production in RA and to identify the cell populations expressing A-SAA in inflamed tissue. In contrast to non-inflamed synovium, A-SAA mRNA expression was identified in all RA tissues examined. A-SAA appeared to be produced by synovial tissue synoviocytes, macrophages and endothelial cells.</p><p>This study demonstrates that A-SAA mRNA is expressed in several cell populations infiltrating RA synovial tissue, but not in normal synovial tissue. First, by employing RT-PCR, A-SAA mRNA expression was observed in all eight unseparated RA tissue samples studied. Second, A-SAA mRNA expression and protein production was demonstrated in primary cultures of purified RA synoviocytes. Finally, using immunohistochemical techniques, A-SAA protein appeared to colocalize with both lining layer and sublining layer synoviocytes, macrophages and some endothelial cells. The observation of A-SAA mRNA expression in cultured RA synoviocytes confirms and extends the recent findings reported by Kumon <italic>et al</italic> [<xref ref-type="bibr" rid="B6">6</xref>], which demonstrated A-SAA mRNA expression in stimulated RA synoviocytes, but not in unstimulated RA synoviocytes. The demonstration of A-SAA production in unstimulated RA synoviocyte cultures in that study is probably due to the increased sensitivity of the analytical methods employed. Additionally, the detection of A-SAA protein in culture media supernatants harvested from unstimulated synoviocytes confirms endogenous A-SAA production and is consistent with A-SAA mRNA expression and translation by the same cells. Moreover, the demonstration of A-SAA protein in RA synovial tissue, RA cultured synoviocytes, macrophages and endothelial cells is consistent with previous studies [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B12">12</xref>] demonstrating A-SAA production by a variety of human cell populations.</p><p>The RA synovial lining layer is composed of activated macrophages and fibroblast-like synoviocytes [<xref ref-type="bibr" rid="B13">13</xref>]. The macrophage is the predominant cell type and it has been shown [<xref ref-type="bibr" rid="B14">14</xref>] to accumulate preferentially in the surface of the lining layer. Macrophages also accumulate in the perivascular areas of the sublining layer [<xref ref-type="bibr" rid="B15">15</xref>]. As demonstrated in the present study, not all lining layer and sublining layer macrophages appeared to produce A-SAA. Futher studies of isolated synovial tissue macrophages, and immunohistological studies employing double-labelling techniques, will elucidate this observation. Nevertheless, the observations reported in this study strongly suggest that A-SAA is produced not only by synoviocytes, but also by synovial tissue macrophage populations. Local A-SAA protein production by vascular endothelial cells was detected in some, but not all, of the tissues examined. The reason for the variability in vascular A-SAA staining is unknown, but this variability may be due to differences in endothelial cell activation, events relating to angiogenesis or the intensity of local inflammation.</p><p>The value of measuring serum A-SAA levels as a reliable surrogate marker of inflammation has been demonstrated [<xref ref-type="bibr" rid="B1">1</xref>] in several diseases including RA, juvenile chronic arthritis, psoriatic arthropathy, ankylosing spondylitis, Beh&#x000e7;et's disease, reactive arthritis and Crohn's disease. It has been suggested [<xref ref-type="bibr" rid="B16">16</xref>] that serum A-SAA levels may represent the most sensitive measurement of the acute-phase reaction. Cunnane <italic>et al</italic> [<xref ref-type="bibr" rid="B16">16</xref>] quantified serum A-SAA levels in 140 patients with various inflammatory joint diseases with duration of less than 2 years, and demonstrated significant correlations with other acute-phase measurements such as C-reactive protein and the erythrocyte sedimentation rate. The magnitude of the A-SAA response was greatest, and the highest levels occurred in RA. In RA, A-SAA levels provided the strongest correlations with clinical measurements of disease activity, and changes in serum levels best reflected the clinical course.</p><p>The principal biologic function of A-SAA is not known [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. A number of biologic activities have been described, however, including several that are relevant to the understanding of inflammatory and tissue degrading mechanisms in human arthritis. A-SAA induces migration, adhesion and tissue infiltration of circulating monocytes and polymorphonuclear leucocytes [<xref ref-type="bibr" rid="B17">17</xref>]. In addition, human A-SAA can induce interleukin-1&#x003b2;, interleukin-1 receptor antagonist and soluble type II tumour necrosis factor receptor production by a monocyte cell line [<xref ref-type="bibr" rid="B18">18</xref>]. Moreover, A-SAA can stimulate the production of cartilage-degrading proteases by both human and rabbit synoviocytes [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. The effects of A-SAA on protease production are particularly interesting because, in RA, a sustained acute-phase reaction has been strongly associated with progressive joint damage [<xref ref-type="bibr" rid="B21">21</xref>]. The known association between the acute-phase response and progressive joint damage may be the direct result of synovial A-SAA-induced effects on cartilage degradation.</p></sec>

Active synovial matrix metalloproteinase-2 is associated with radiographic erosions in patients with early synovitis

<sec><title>Introduction:</title><p>In cancer the gelatinases [matrix metalloproteinase (MMP)-2 and MMP-9] have been shown to be associated with tissue invasion and metastatic disease. In patients with inflammatory arthritis the gelatinases are expressed in the synovial membrane, and have been implicated in synovial tissue invasion into adjacent cartilage and bone. It is hypothesized that an imbalance between the activators and inhibitors of the gelatinases results in higher levels of activity, enhanced local proteolysis, and bone erosion.</p></sec><sec><title>Objectives:</title><p>To determine whether the expression and activity levels of MMP-2 and MMP-9, and their regulators MMP-14 and tissue inhibitor of metalloproteinase (TIMP), are associated with early erosion formation in patients with synovitis of recent onset.</p></sec><sec><title>Patients and method:</title><p>A subset of 66 patients was selected from a larger early synovitis cohort on the basis of tissue availability for the study of synovial tissue and serum gelatinase expression. Patients with peripheral joint synovitis of less than 1 years' duration were evaluated clinically and serologically on four visits over a period of 12 months. At the initial visit, patients underwent a synovial tissue biopsy of one swollen joint, and patients had radiographic evaluation of hands and feet initially and at 1year. Serum MMP-1, MMP-2, MMP-9, MMP-14, and TIMP-1 and TIMP-2 levels were determined, and synovial tissue was examined by immunohistology for the expression of MMP-2 and MMP-9, and their molecular regulators. Gelatinolytic activity for MMP-2 and MMP-9 was quantified using a sensitive, tissue-based gel zymography technique. Four healthy individuals underwent closed synovial biopsy and their synovial tissues were similarly analyzed.</p></sec><sec><title>Results:</title><p>Of the 66 patients studied, 45 fulfilled American College of Rheumatology criteria for rheumatoid arthritis (RA), with 32 (71%) being rheumatoid factor positive. Of the 21 non-RA patients, seven had a spondylarthropathy and 14 had undifferentiated arthritis. Radiographically, 12 of the RA patients had erosions at multiple sites by 1 year, whereas none of the non-RA patients had developed erosive disease of this extent. In the tissue, latent MMP-2 was widely expressed in the synovial lining layer and in areas of stromal proliferation in the sublining layer and stroma, whereas MMP-9 was expressed more sparsely and focally. MMP-14, TIMP-2, and MMP-2 were all detected in similar areas of the lining layer on consecutive histologic sections. Tissue expression of MMP-14, the activator for pro-MMP-2, was significantly higher in RA than in non-RA patients (8.4 &#x000b1; 5 versus 3.7 &#x000b1; 4 cells/high-power field; <italic>P</italic> = 0.009). In contrast, the expression of TIMP-2, an inhibitor of MMP-2, was lower in the RA than in the non-RA samples (25 &#x000b1; 12 versus 39 &#x000b1; 9 cells/high-power field; <italic>P</italic> = 0.01). Synovial tissue expressions of MMP-2, MMP-14, and TIMP-2 were virtually undetectable in normal synovial tissue samples. The synovial tissue samples of patients with erosive disease had significantly higher levels of active MMP-2 than did those of patients without erosions (Fig. <xref ref-type="fig" rid="F1">1</xref>). Tissue expression of MMP-2 and MMP-9, however, did not correlate with the serum levels of these enzymes.</p><p>With the exception of serum MMP-2, which was not elevated over normal, serum levels of all of the other MMPs and TIMPs were elevated to varying degrees, and were not predictive of erosive disease. Interestingly, MMP-1 and C-reactive protein, both of which were associated with the presence of erosions, were positively correlated with each other (<italic>r</italic> = 0.42; <italic>P</italic> &#x0003c; 0.001).</p></sec><sec><title>Discussion:</title><p>MMP-2 and MMP-9 are thought to play an important role in the evolution of joint erosions in patients with an inflammatory arthritis. Most studies have concentrated on the contribution of MMP-9 to the synovitis, because synovial fluid and serum MMP-9 levels are markedly increased in inflammatory arthropathies. Previously reported serum levels of MMP-9 have varied widely. In the present sample of patients with synovitis of recent onset, serum MMP-9 levels were elevated in only 21%. Moreover, these elevations were not specific for RA, the tissue expression of MMP-9 was focal, and the levels of MMP-9 activity were not well correlated with early erosions. Although serum MMP-2 levels were not of prognostic value, high synovial tissue levels of MMP-2 activity were significantly correlated with the presence of early erosions. This may reflect augmented activation of MMP-2 by the relatively high levels of MMP-14 and low levels of TIMP-2 seen in these tissues. We were able to localize the components of this trimolecular complex to the synovial lining layer in consecutive tissue sections, a finding that is consistent with their colocalization.</p><p>In conclusion, we have provided evidence that active MMP-2 complexes are detectable in the inflamed RA synovium and may be involved in the development of early bony erosions. These results suggest that strategies to inhibit the activation of MMP-2 may have the potential for retarding or preventing early erosions in patients with inflammatory arthritis.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Goldbach-Mansky</surname><given-names>Raphaela</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Lee</surname><given-names>Jennifer M</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Hoxworth</surname><given-names>Joseph M</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Smith II</surname><given-names>David</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Duray</surname><given-names>Paul</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Schumacher</surname><given-names>Ralph H</given-names><suffix>Jr</suffix></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Yarboro</surname><given-names>Cheryl H</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Klippel</surname><given-names>John</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Kleiner</surname><given-names>David</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A10" contrib-type="author"><name><surname>El-Gabalawy</surname><given-names>Hani S</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>elgabala@exchange.nih.gov</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Radiographic erosions are characteristic of a number of chronic inflammatory arthropathies, and are associated with articular destruction leading to functional loss and disability [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. The prevention or retardation of erosions is a major objective for early therapeutic interventions in patients with new onset arthritis [<xref ref-type="bibr" rid="B3">3</xref>]. Although the pathologic processes that underlie the development of erosions are incompletely understood, a considerable body of evidence has suggested that a pannus of cells that originate in chronically inflamed, proliferative synovial tissue becomes locally invasive, and enzymatically degrades the matrix of the articular cartilage and periarticular bone [<xref ref-type="bibr" rid="B4">4</xref>].</p><p>Although a number of enzymes have been shown to be involved in erosion formation, several members of the matrix metalloproteinase (MMP) family are thought to play a central role in the degradation of the extracellular matrix of articular bone and cartilage. Of the MMPs, the collagenases [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>] and stromelysins [<xref ref-type="bibr" rid="B7">7</xref>] have been the best studied, and marked elevations in serum and synovial levels have been found in rheumatoid arthritis (RA) [<xref ref-type="bibr" rid="B8">8</xref>]. A number of studies have shown that the serum and synovial levels of both gelatinases MMP-2 [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>] and MMP-9 [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>] are similarly elevated in RA. Although both of these MMPs are expressed at the junction between synovial pannus and cartilage or bone [<xref ref-type="bibr" rid="B14">14</xref>], further evidence of their direct role in erosion formation is lacking. In addition members of this subfamily of MMPs are involved in the invasiveness of cancers [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>], and thus may also play a similar role in the invasion of synovial pannus.</p><p>The regulation of MMP activity is complex, and occurs at multiple levels [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. In most cases, MMPs are secreted as proenzymes and then undergo extracellular activation. Few studies have evaluated the levels of active MMPs <italic>in situ</italic>. Gel zymography has proved to be a sensitive and effective method of detecting the activity of gelatinases in small tissue samples [<xref ref-type="bibr" rid="B20">20</xref>]. In the present study we evaluated the levels of active MMP-2 and MMP-9 in small synovial biopsy samples obtained from patients with synovitis of recent onset. We also examined the synovial expression of the gelatinases by immunohistology, as well as their serum levels in order to determine whether these indices may predict erosions in RA patients.</p></sec><sec><title>Patients and method</title><sec><title>Patient population</title><p>A total of 66 patients that formed a subset of a larger early synovitis cohort, recruited into an ongoing US National Institutes of Health protocol (94-AR-0194), were selected for analysis on the basis of having stored optimum cooling temperature (OCT) compound samples available for analysis. The patients had peripheral joint synovitis with duration longer than 6 weeks but less than 12 months. After the initial evaluation they were followed prospectively for 12 months. Of those patients, 45 fulfilled the American College of Rheumatology criteria for RA [<xref ref-type="bibr" rid="B21">21</xref>], seven patients fulfilled the European Spondylarthropathy Study Group criteria for spondylarthropathies [<xref ref-type="bibr" rid="B22">22</xref>], and 14 could not be classified and were diagnosed with undifferentiated arthritis.</p><p>Serum aliquots obtained at the initial visit were used for the analysis of serum MMP and tissue inhibitor of metalloproteinase (TIMP) levels. All patients underwent a closed needle biopsy of synovial tissue [<xref ref-type="bibr" rid="B23">23</xref>] from an inflamed joint, typically a knee. Tender and swollen joint counts were obtained from 66 joint areas; hips were excluded. Anterioposterior and lateral radiographs of the hands and feet were obtained at the initial visit and at 1-year follow-up visit. Radiographs were read by experienced radiologists, who were unaware of the patient's diagnosis or test results, and were scored for erosions. Erosions were defined as unequivocal loss of cortical and subcortical bone in at least two different joints of either the hands or feet. Of the 66 patients who underwent a clinical evaluation and analysis of serum MMP levels, 28 had synovial tissue biopsies with detectable lining layer and large enough samples to determine MMP and TIMP expression by gel zymography and immunohistochemistry.</p><p>Healthy volunteers (<italic>n</italic> = 4) were recruited under the same protocol and a synovial knee biopsy was performed once signed informed consent was obtained. These tissues were also assayed by gel zymography and stained immunohistochemically.</p></sec><sec><title>Antibodies and gels</title><p>Antibodies to MMP-2 (immunoglobulin G₁), MMP-9 (immunoglobulin G₁) and MMP-14 (immunoglobulin G₁) were purchased from Research Diagnostics Inc (Flanders, NJ, USA). The TIMP-2 (immunoglobulin G₁) antibody was obtained from Calbiochem (La Jolla, CA, USA). Polyclonalmouse immunoglobulin G₁ was used as negative controlin the staining experiments. Enzyme-linked immunosorbent assay (ELISA) kits for serum MMP-1, TIMP-1 and TIMP-2, and the activity assays for MMP-2 and MMP-9 were purchased from Biotrak (Amersham Pharmacia Biotech, Piscataway, NJ, USA). Precast zymography gels containing 10% Tris glycine with 0.1% gelatin as substrate were purchased from Novex (San Diego, CA, USA), and protein quantification kits were purchased from Bio-Rad Laboratories (Hercules, CA, USA).</p></sec><sec><title>Matrix metalloproteinase and tissue inhibitor of metalloproteinase enzyme-linked immunosorbent assay</title><p>ELISA assays for MMP-1, TIMP-1, and TIMP-2 were performed according to the manufacturer's instructions. Briefly, 100 &#x003bc; l diluted serum samples were pipetted in duplicates into the appropriate ELISA microtiter wells and incubated for 2 h at 20-25 &#x000b0;C. Wells were washed and incubated with 100 &#x003bc;l of the antiserum for 2 h at 20-25&#x000b0;C, and incubated with 100 &#x003bc;l of the peroxidase conjugate for 1h. After washing 100 &#x003bc;l of 3,3',5,5' -tetramethylbenzi-dine/hydrogen peroxide substrate were pipetted into the wells and incubated for 30 min at room temperature. Absorbance at 630 nm was measured spectrophotometrically in an automated plate reader.</p></sec><sec><title>Serum matrix metalloproteinase activity assays</title><p>Active and total MMP-2 and MMP-9 were measured by activity assays. Assays were performed according to the manufacturer's instructions. Briefly, 100 &#x003bc;l diluted serum samples were plated in quadruplicate and refrigerated at 4 &#x000b0;C overnight. Plates were washed in an automated plate washer and 50 &#x003bc;l of a 1mol/l p-aminophenylmercuric acetate, which activates pro-MMP, was added only into duplicate wells in which total MMP activity was to be measured. After 2 h of incubation at 37 &#x000b0;C, 50 &#x003bc;l detection reagent containing a modified urokinase and S-2444 peptide substrate was added to all wells. Plates were read spectrophotometrically at time 0 and after a 2-h incubation at 37 &#x000ba;C for 2 h at an absorbance of 410 nm. MMP activity was represented by the change of absorbance over time. These values were compared with a standard curve of serial dilutions of a known concentration of activated enzyme. Normal range of values was provided by the manufacturer.</p></sec><sec><title>Gel zymography</title><p>Gel zymography on synovial tissue samples was performed according to a protocol developed by Kleiner <italic>et al</italic> [<xref ref-type="bibr" rid="B20">20</xref>]. Unfixed cryosections of synovial tissue, obtained from closed needle biopsies as described above, were scraped off the slide with a 22-gauge needle, transferred into a microfuge tube, and vortexed. Media from a HT-1080 fibrosarcoma cell line was used as a positive control for MMP-2 and MMP-9 activity. Of each sample suspension, 30 &#x003bc;l were loaded onto a precast sodium dodecylsulfate- polyacrylamide gel electrophoresis (SDS-PAGE) containing 0.1% gelatin. After electrophoresis, gels were washed in 2.5% Triton X for 3 h and incubated at 37 &#x000b0;C in low salt collagenase buffer containing 50 mmol/l Tris, 0.2mol/l NaCl, 5mmol/l anhydrous CaCl₂, and 0.02% Brijdetergent. The gels were subsequently stained with 0.5% Comassie blue (G-250), destained with 30% methanol and 10% acetic acid, and incubated in 30% methanol and 5% glycerol. They were dried between cellophane sheets. Areas of gelatinase activity appeared as nonstaining bands on the gels.</p><p>The zymography gels were scanned and analyzed using US National Institutes of Health Image 1.6 software. Integrated pixel density for each gelatinolytic band was recorded and reported in volume units of pixel intensity &#x000d7; mm². Each value was normalized against the protein concentration, determined by a protein quantitation kit according to the manufacturer's instructions, and reported as gelatinase activity in nanograms per milligrams of protein. Gelatinolytic bands at 92, 72, and 62 kDA represent latent MMP-2, latent MMP-9, and active MMP-2 activity, respectively.</p></sec><sec><title>Immunohistochemistry</title><p>Immunohistochemistry was performed according to standard techniques [<xref ref-type="bibr" rid="B24">24</xref>]. OCT-embedded synovial tissue was cryosectioned into 8-&#x003bc;m thick sections and acetone fixed on glass slides. After hydration and blocking of endogenous peroxidase activity, tissues were blocked with a dilution of human and goat serum. Sections were incubated with primary antibody at the following concentrations: MMP-2 (1 : 75), MMP-9 (1 : 75), TIMP-2 (1 : 75), and MMP-14 (1 : 30). Negative controls using either mouse immunoglobulin G₁antibodies or no primary antibody were performed with each experiment. Slides were washed and incubated with goat antimouse biotinylated antibody at a 1 : 750 dilution followed by incubation with streptavidine horseradish peroxidase. Slides were developed with a substrate chromogen solution until a brown reaction product was observed, then counterstained with hematoxylin and coverslipped.</p></sec><sec><title>Histologic evaluation of the synovial tissue</title><p>Synovial tissues suitable for analysis were histologically evaluated for the presence and degree of inflammation. All tissues were analyzed in a blinded manner by two independent observers. Proliferation of the lining layer, inflammatory cellular infiltration of the sublining layer, lymphocytic aggregates, and stromal proliferation were scored semiquantitatively from 0 to 3 in five high-power fields (HPFs; 400 &#x000d7;) for each category. Scores were added and divided by four to derive a tissue composite index. Tissue scores for MMP expression were obtained by counting the number of positive cells in the lining layer and sublining layer (one HPF underneath the lining layer) in five representative HPFs (400&#x000d7;) and expressed as positive cells/HPF.</p></sec><sec><title>Statistical methods</title><p>Differences between patient groups were analyzed using &#x003c7;², analysis of variance, and Kruskal-Wallis tests.</p></sec></sec><sec><title>Results</title><sec><title>Patient clinical features</title><p>The clinical features of the 66 patients studied are shown in Table <xref ref-type="table" rid="T1">1</xref>. Patients with RA were older, had more tender and swollen joints, were rheumatoid factor positive, and were more likely to be on disease-modifying antirheumatic drug therapy than were non-RA patients. Of the patients with RA, 12 had erosions at multiple sites by 1 year, whereas none of the non-RA patients had developed definite erosions. Patient characteristics in the subset of 28 patients who underwent synovial tissue immunohistology and gel zymography did not differ significantly from the larger sample (data not shown).</p></sec><sec><title>Serum levels of the metalloproteinases and their inhibitors in patients with early synovitis</title><p>MMP-1 serum concentrations in all 66 (100%) patients examined were elevated, as were total MMP-9 levels in 14 (21%), TIMP-1 levels in 65 (98%), and TIMP-2 levels in 14 (21%) patients (Table <xref ref-type="table" rid="T2">2</xref>). Serum levels were not elevated over the normal range provided by the manufacturer. Patients with erosive disease had higher serum MMP-1 and mean C-reactive protein (CRP) levels than patients without erosion (<italic>P</italic> &#x0003c; 0.01). levels of mmp-1, but not of other mmps, significantly correlated with the <italic>crp</italic> levels (&#x0003e; r = 0.42; <italic>P</italic> &#x0003c; 0.001). overall serum mmp-2 and mmp-9 levels did not correlate with tissue expression of these enzymes (data not shown).</p></sec><sec><title>Localization of MMP-2, MMP-9, MMP-14, and TIMP-2 expression in synovial tissue</title><p>Expressions of MMP-2, MMP-9, MMP-14, and TIMP-2 in representative locations in the synovial tissue specimens are shown in Figure <xref ref-type="fig" rid="F2">2</xref>. Of note, the antibodies used to detect MMP-2 recognized only the latent zymogen form. MMP-2 was widely expressed in the synovial lining layer and in areas of stromal proliferation in the sublining and stromal layer. MMP-9 expression was more focal and was observed sparsely in the lining layer and in the endothelium of single vessels in both RA and non-RA tissues. MMP-14 and TIMP-2 were detected primarily in the lining layer on consecutive tissue sections.</p></sec><sec><title>Comparisons of tissue levels of MMP-2, MMP-9, MMP-14, and TIMP-2</title><p>MMP expression was assayed in 21 RA tissues and seven non-RA tissues, and results were compared among groups and with tissues from four healthy volunteers without evidence of an arthropathy or traumatic injury (Table <xref ref-type="table" rid="T3">3</xref>). Tissue composite indices of inflammation were similar in the RA and non-RA tissues. Five of the 21 RA patients had erosions at multiple sites, as compared with none of the non-RA patients. Synovial tissue expressions of MMP-2, MMP-14, and TIMP-2 were virtually undetectable in all normal volunteers; MMP-9 expression was observed in an isolated area of one normal synovial tissue sample. Tissue expression for MMP-14, the activator for pro-MMP-2, was significantly higher in RA than in non-RA patients (8.4 &#x000b1; 5 versus 3.7 &#x000b1; 4 cells/HPF; <italic>P</italic> = 0.009). In contrast, the expression of TIMP-2, an inhibitor of MMP-2, was lower in the RA than in the non-RA samples (25 &#x000b1; 12 versus 39 &#x000b1; 9 cells/HPF; <italic>P</italic> = 0.01). Activated and latent MMP-2 activities, as measured by gel zymography, tended to be higher overall in the RA patients than in the non-RA patients, but results were not statistically different.</p></sec><sec><title>Radiographic erosions are associated with higher MMP-2 activity in the synovial tissue</title><p>The results of synovial tissue MMP-2 and MMP-9 expressions were compared among patients with erosive disease, patients without erosions, and normal volunteers (Fig. <xref ref-type="fig" rid="F1">1</xref>). Patients with erosive disease had significantly higher synovial levels of total MMP-2 (10.9 &#x000b1; 8.1 versus 18.9 &#x000b1; 9.4 ng/mg; <italic>P</italic> = 0.04) and active MMP-2 (3.5 &#x000b1; 2.1 versus 1.9 &#x000b1; 2.5 ng/mg; <italic>P</italic> = 0.04) than patients without erosions. Similarly, mean MMP-9 activity was higher in tissue samples from patients with erosive disease, but wide variations in tissue expression resulted in poor discrimination between groups. As expected, synovium of normal volunteers had significantly lower MMP-2 activity than that of the patients (5.7 &#x000b1; 0.7 versus 12.3 &#x000b1; 8.7 ng/mg; <italic>P</italic> &#x0003c; 0.05). this trend was also observed for mmp-9, but did not reach statistical significance, probably due to the very wide standard deviation (4.0 &#x000b1; 1.3 versus 28.6 &#x000b1; 71.2 ng/mg in normal individuals and in the patients, respectively; &#x0003e; P = 0.07).</p></sec></sec><sec><title>Discussion</title><p>Joint destruction is one of the hallmarks of erosive arthropathies and leads to significant morbidity and disability. Newer imaging studies have concluded that erosion formation can occur very early in the disease [<xref ref-type="bibr" rid="B25">25</xref>]. Detailed pathologic studies, although generally performed on destroyed joints obtained at the time of arthroplasty, have suggested that these erosions result from direct invasion of synovial tissue into the adjacent articular bone and cartilage. It has been postulated that the locally invasive cells that form the aggressive pannus have acquired a transformed phenotype, and analogies to invasive cancer cells have been made [<xref ref-type="bibr" rid="B26">26</xref>]. Several metalloproteinases that are capable of the proteolytic degradation of extracellular matrix are thought to play a major role in the dissolution of the bone and cartilage matrix, allowing the invasive cells to migrate into the stroma across cell boundaries. Although the relative contribution of each MMP to this process remains unclear, the gelatinases, particularly MMP-2, appear to play a prominent role in tissue invasion [<xref ref-type="bibr" rid="B16">16</xref>]. In patients with arthritis, most studies have concentrated on the contribution of MMP-9 to the underlying mechanism of joint destruction. In contrast to MMP-2, MMP-9 levels tend to be higher in inflammatory arthropathies than in traumatic, osteoarthritic, and tissue repair processes [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B27">27</xref>]. Furthermore, these two MMPs differ in tissue distribution and transcriptional regulation. These observations suggest that MMP-2 and MMP-9 may contribute differentially to the pathophysiologic processes that lead to joint destruction.</p><p>In the joint, MMP-9 is expressed in macrophages in the synovial lining layer, in vascular endothelial cells, leukocytes, chondrocytes, and osteoclasts. It is also expressed on cultured fibroblasts and on activated fibroblasts <italic>in vivo</italic> [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B28">28</xref>]. MMP-9 is inducible by inflammatory cytokines such as tumor necrosis factor-&#x003b1; and interleukin-1 [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B29">29</xref>], and the post-transcriptional regulation <italic>in vivo</italic> remains incompletely understood. MMP-9 is secreted in an inactive zymogen form and can be activated <italic>in vitro</italic> by the plasmin cascade and through other MMPs [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>]. In its inactive form MMP-9 is complexed to TIMP-1 [<xref ref-type="bibr" rid="B34">34</xref>]. The regulation of MMPs by inflammatory cytokines, and the expression of MMP-9 in macrophages and vascular structures have previously suggested that MMP-9 may be associated with the synovial inflammation and neovascularization [<xref ref-type="bibr" rid="B35">35</xref>,<xref ref-type="bibr" rid="B36">36</xref>].</p><p>In contrast to MMP-9, MMP-2, in the inactive zymogen form, is constitutively expressed by many cell types, including fibroblasts and endothelial cells, and is involved in many aspects of normal tissue remodeling and angiogenesis [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>]. Its unique regulatory requirements at the cell surface have suggested a special role for MMP-2 in tissue invasion. The transcriptional and post-translational activation of this enzyme is tightly regulated [<xref ref-type="bibr" rid="B16">16</xref>]. MMP-2 differs from several other members of the metalloproteinase family in that its transcription is not induced in response to the proinflammatory cytokines tumor necrosis factor-&#x003b1; and interleukin-1, and seems primarily regulated post-translationally. As is characteristic of most of the MMPs, MMP-2 is produced as a proenzyme, and requires activation by enzymatic cleavage. Evidence from several investigators has suggested that MMP-2 in its active form is bound in a trimolecular complex at the cell surface [<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B42">42</xref>]. Pro-MMP-2 complexed with TIMP-2 binds to a TIMP-2 binding site on the membrane-bound MMP-14 receptor, and through a conformational change the catalytic domain becomes exposed forming the active enzyme complex. Upon dissociation from the MMP-14 receptor, TIMP-2 binds the catalytic site and suppresses the activity. Thus, the activity of MMP-2 is regulated locally by a delicate balance between MMP-14 activation and TIMP-2 inhibition. Studies have localized this trimolecular complex to the invadopodia of malignant cells at sites of tissue invasion into stroma [<xref ref-type="bibr" rid="B43">43</xref>,<xref ref-type="bibr" rid="B44">44</xref>] and serum levels and high tissue expression levels of active MMP-2 are predictors of disease severity and shortened survival in several malignancies [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>].</p><p>In view of the analogies between erosion formation in arthritis and tumor invasion in cancer, we hypothesized that activated MMP-2 and possibly MMP-9 might play an important role in the early stages of invasion of synovial pannus into bone and cartilage. Synovial tissue specimens obtained from a cohort of patients with early synovitis were examined for the presence of MMP-9 and components of the MMP-2-associated trimolecular complex. A highly sensitive zymographic technique was used to evaluate tissue levels of active and latent MMP-2. Our immunohistologic data suggest that MMP-2, TIMP-2, and MMP-14 are all localized to the synovial lining layer. Previous reports have found colocalization of these molecules in areas of tissue remodeling, loosening of prosthetic joints [<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>,<xref ref-type="bibr" rid="B47">47</xref>], and in the cartilage-pannus junction [<xref ref-type="bibr" rid="B14">14</xref>]. Although the results of these experiments do not prove that the expression of these molecules occurs on the same cell, they are consistent with the view that the synovial lining layer is a site of MMP-2 activation.</p><p>Because of the blind nature of the biopsy technique used in the present study, we did not specifically sample areas of synovium directly adjacent to cartilage and bone. We did, however, ensure that all of the needle biopsy samples examined immunohistologically and zymographically had clear evidence of a well-defined synovial lining layer, were of adequate size, and were thus appropriate for comparative evaluation. The selection of biopsy material with detectable lining layer and sufficiently large size for analysis may have biased our results toward patients with more proliferative synovial lesions, and possibly more severe disease. This selection bias would have occurred in the RA and non-RA patients, however, and is therefore unlikely to account for the differences seen in MMP-2 activity between the patient groups. We compared the synovium of patients with early RA with that of patients with other forms of early synovitis, and with that of normal volunteers. The RA samples tended to have the highest expression levels of MMP-2 and MMP-14, but they exhibited low levels of TIMP-2 expression. More importantly, we showed that the patients with radiographic erosions, all of whom had RA, had the highest levels of active MMP-2 by gel zymography. We therefore propose that the augmented levels of activated MMP-2 detected in the synovia of patients with early erosive RA may relate, at least in part, to an imbalance between activation by MMP-14 and inhibition by TIMP-2.</p><p>Immunohistologically, MMP-9 expression was clearly higher in inflamed synovium than in normal synovium, where it was virtually undetectable. Furthermore, MMP-9 activity levels tended to be higher in the synovial samples of patients with radiographic erosions than those of patients without erosions, but the measured values varied widely in the specimens examined. The focal expression of MMP-9 in these synovial tissues, combined with the small number of samples examined, might have contributed to the wide tissue variations observed.</p><p>Overall, the present observations suggest an association between the presence of active synovial gelatinases and the early development of erosive articular damage. Although the evidence is indirect, these data are consistent with the hypothesis that high levels of activated synovial gelatinases reflect augmented synovial invasiveness; these observations are similar to those made in the context of invasive cancers [<xref ref-type="bibr" rid="B16">16</xref>]. Although there continues to be controversy regarding the origins of the invasive cells at the junction of cartilage/bone and pannus, considerable evidence has accumulated that the lining cells in RA synovium exhibit a 'transformed' phenotype that is typical of invasive cells [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B48">48</xref>].</p><p>We attempted to evaluate the more readily available serum levels of active MMPs as predictors of joint erosions. Overall, only the serum levels of MMP-1 were significantly associated with erosions and with CRP levels. Persistent CRP elevation in RA patients has been shown to be associated with the progression of radiographic erosions [<xref ref-type="bibr" rid="B49">49</xref>], but a previous RA study failed to associate MMP-1 levels with erosions [<xref ref-type="bibr" rid="B50">50</xref>]. In general, serum levels of MMPs have tended to vary widely in the reported studies [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B51">51</xref>], and appear not to be of prognostic value. Furthermore, degranulation of platelets and neutrophils may contribute to the elevation of serum levels of MMP-2 and MMP-9 seen in the patients, and may be an explanation for the high baseline level in the serum [<xref ref-type="bibr" rid="B52">52</xref>,<xref ref-type="bibr" rid="B53">53</xref>]. In contrast, this is not the case for MMP-1, because no blood cell carries a preformed secretory granule of this enzyme and this may explain why MMP-1 levels relate more closely to radiographic erosions. The measurement of serum MMP levels in the context of controlled therapy trials has not been well explored.</p><p>In the present study we provided evidence that active MMP-2 complexes are expressed in the synovium and may be involved in the development of bony erosions. Therapeutic strategies to inhibit a broad spectrum of MMPs, including MMP-2 and possibly MMP-9, may therefore present a more powerful approach to retard or prevent early erosions in patients with an inflammatory arthropathy.</p></sec>

Specificity of T cells in synovial fluid: high frequencies of CD8⁺ T cells that are specific for certain viral epitopes

<sec><title>Introduction:</title><p>Epstein-Barr virus (EBV) is transmitted orally, replicates in the oropharynx and establishes life-long latency in human B lymphocytes. T-cell responses to latent and lytic/replicative cycle proteins are readily detectable in peripheral blood from healthy EBV-seropositive individuals. EBV has also been detected within synovial tissue, and T-cell responses to EBV lytic proteins have been reported in synovial fluid from a patient with rheumatoid arthritis (RA). This raises the question regarding whether T cells specific for certain viruses might be present at high frequencies within synovial fluid and whether such T cells might be activated or able to secrete cytokines. If so, they might play a 'bystander' role in the pathogenesis of inflammatory joint disease.</p></sec><sec><title>Objectives:</title><p>To quantify and characterize T cells that are specific for epitopes from EBV, cytomegalovirus (CMV) and influenza in peripheral blood and synovial fluid from patients with arthritis.</p></sec><sec><title>Methods:</title><p>Peripheral blood mononuclear cells (PBMCs) and synovial fluid mononuclear cells (SFMCs) were obtained from patients with inflammatory arthritis (including those with RA, osteoarthritis, psoriatic arthritis and reactive arthritis). Samples from human leucocyte antigen (HLA)-A2-positive donors were stained with fluorescent-labelled tetramers of HLA-A2 complexed with the GLCTLVAML peptide epitope from the EBV lytic cycle protein BMLF1, the GILGFVFTL peptide epitope from the influenza A matrix protein, or the NLVPMVATV epitope from the CMV pp65 protein. Samples from HLA-B8-positive donors were stained with fluorescent-labelled tetramers of HLA-B8 complexed with the RAKFKQLL peptide epitope from the EBV lytic protein BZLF1 or the FLRGRAYGL peptide epitope from the EBV latent protein EBNA3A. All samples were costained with an antibody specific for CD8. CD4⁺ T cells were not analyzed. Selected samples were costained with antibodies specific for cell-surface glycoproteins, in order to determine the phenotype of the T cells within the joint and the periphery. Functional assays to detect release of IFN-&#x003b3; or tumour necrosis factor (TNF)-&#x003b1; were also performed on some samples.</p></sec><sec><title>Results:</title><p>The first group of 15 patients included 10 patients with RA, one patient with reactive arthritis, one patient with psoriatic arthritis and three patients with osteoarthritis. Of these, 11 were HLA-A2 positive and five were HLA-B8 positive. We used HLA-peptide tetrameric complexes to analyze the frequency of EBV-specific T cells in PBMCs and SFMCs (Figs <xref ref-type="fig" rid="F1">1</xref> and <xref ref-type="fig" rid="F2">2</xref>). Clear enrichment of CD8⁺ T cells specific for epitopes from the EBV lytic cycle proteins was seen within synovial fluid from almost all donors studied, including patients with psoriatic arthritis and osteoarthritis and those with RA. In donor RhA6, 9.5% of CD8⁺ SFMCs were specific for the HLA-A2 restricted GLCTLVAML epitope, compared with 0.5% of CD8⁺ PBMCs. Likewise in a donor with osteoarthritis (NR4), 15.5% of CD8⁺ SFMCs were specific for the HLA-B8-restricted RAKFKQLL epitope, compared with 0.4% of CD8⁺ PBMCs. In contrast, we did not find enrichment of T cells specific for the HLA-B8-restricted FLRGRAYGL epitope (from the latent protein EBNA3A) within SFMCs compared with PBMCs in any donors. In selected individuals we performed ELISpot assays to detect IFN-&#x003b3; secreted by SFMCs and PBMCs after a short incubation <italic>in vitro</italic> with peptide epitopes from EBV lytic proteins. These assays confirmed enrichment of T cells specific for epitopes from EBV lytic proteins within synovial fluid and showed that subpopulations of these cells were able to secrete proinflammatory cytokines after short-term stimulation.</p><p>We used a HLA-A2/GILGFVFTL tetramer to stain PBMCs and SFMCs from six HLA-A2-positive patients. The proportion of T cells specific for this influenza epitope was low (&#x0003c;0.2%) in all donors studied, and we did not find any enrichment within SFMCs.</p><p>We had access to SFMCs only from a second group of four HLA-A2-positive patients with RA. A tetramer of HLA-A2 complexed to the NLVPMVATV epitope from the CMV pp65 protein reacted with subpopulations of CD8⁺ SFMCs in all four donors, with frequencies of 0.2, 0.5, 2.3 and 13.9%. SFMCs from all four donors secreted TNF after short-term incubation with COS cells transfected with HLA-A2 and pp65 complementary DNA. We analyzed the phenotype of virus-specific cells within PBMCs and SFMCs in three donors. The SFMC virus-specific T cells were more highly activated than those in PBMCs, as evidenced by expression of high levels of CD69 and HLA-DR. A greater proportion of SFMCs were CD38⁺, CD62L low, CD45RO bright, CD45RA dim, CD57⁺ and CD28^- when compared with PBMCs.</p></sec><sec><title>Discussion:</title><p>This work shows that T cells specific for certain epitopes from viral proteins are present at very high frequencies (up to 15.5% of CD8⁺ T cells) within SFMCs taken from patients with inflammatory joint disease. This enrichment does not reflect a generalized enrichment for the 'memory pool' of T cells; we did not find enrichment of T cells specific for the GILGFVFTL epitope from influenza A or for the FLRGRAYGL epitope from the EBV latent protein EBNA3A, whereas we found clear enrichment of T cells specific for the GLCTLVAML epitope from the EBV lytic protein BMLF1 and for the RAKFKQLL epitope from the EBV lytic protein BZLF1.</p><p>The enrichment might reflect preferential recruitment of subpopulations of virus-specific T cells, perhaps based on expression of selectins, chemokine receptors or integrins. Alternatively, T cells specific for certain viral epitopes may be stimulated to proliferate within the joint, by viral antigens themselves or by cross-reactive self-antigens. Finally, it is theoretically possible that subpopulations of T cells within the joint are preferentially protected from apoptotic cell death. Whatever the explanation, the virus-specific T cells are present at high frequency, are activated and are able to secrete proinflammatory cytokines. They could potentially interact with synoviocytes and contribute to the maintenance of inflammation within joints in many different forms of inflammatory arthritis.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Tan</surname><given-names>Linda C</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Mowat</surname><given-names>Alastair G</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Fazou </surname><given-names>Chrysoula</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Rostron </surname><given-names>Tim</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Roskell</surname><given-names>Helen</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Dunbar </surname><given-names>P Rod</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Tournay</surname><given-names>Claire</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Romagn&#x000e9;</surname><given-names>Fran&#x000e7;ois</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Peyrat</surname><given-names>Marie-Alix</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A10" contrib-type="author"><name><surname>Houssaint</surname><given-names>Elisabeth</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A11" contrib-type="author"><name><surname>Bonneville</surname><given-names>Marc</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A12" contrib-type="author"><name><surname>Rickinson</surname><given-names>Alan B</given-names></name><xref ref-type="aff" rid="I5">5</xref></contrib><contrib id="A13" contrib-type="author"><name><surname>McMichael </surname><given-names>Andrew J</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A14" contrib-type="author"><name><surname>Callan</surname><given-names>Margaret FC</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>mcallan@molbiol.ox.ac.uk</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Epstein-Barr virus (EBV) is a common gammaherpesvirus that infects over 90% of individuals worldwide [<xref ref-type="bibr" rid="B1">1</xref>]. The virus is transmitted orally, replicates in the oropharynx and subsequently establishes life-long latent infection of B lymphocytes [<xref ref-type="bibr" rid="B2">2</xref>]. Replicative infection is associated with the expression of approximately 70EBV 'lytic' proteins, whereas latent infection is associated with the expression of up to nine 'latent' proteins [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. The cellular immune response plays a crucial role in controlling EBV infection. T-cell responses to EBV latent proteins have been well studied and are easily detectable in healthy EBV-seropositive individuals [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. More recently, responses to some of the EBV lytic cycle proteins have also been described [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]; reactivities to these antigens are frequently found to be more abundant than those directed against the EBV latent antigens in healthy EBV carriers [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>].</p><p>An association between EBV and rheumatoid arthritis (RA) was first proposed on the basis of high titres of EBV-specific antibodies found in some patients with RA [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. The observation that some of the anti-EBV antibodies were cross-reactive with autoantigens such as collagen was put forward as a further argument in support of the link [<xref ref-type="bibr" rid="B13">13</xref>]. The importance of these findings in the pathophysiology of rheumatoid arthritis has never been established, however. More recently the cloning of T cells that are specific for EBV antigens from synovial fluid taken from a patient with RA [<xref ref-type="bibr" rid="B14">14</xref>] has reopened the debate regarding the importance of this virus in the aetiology of arthritis. David-Ameline <italic>et al</italic> [<xref ref-type="bibr" rid="B15">15</xref>] showed that a large proportion of T-cell clones derived from synovial fluid taken from one individual with RA, under polyclonal activation conditions, recognized EBV-transformed lymphoblastoid cell lines in an human leucocyte antigen (HLA)-restricted manner. Subsequent work revealed that these T-cell clones recognized epitopes from EBV lytic cycle proteins [<xref ref-type="bibr" rid="B14">14</xref>]. Analysis of the T-cell receptor use of the EBV-specific T cell clones and of the T-cell receptor repertoire of synovial fluid lymphocytes [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>] suggested that the EBV-specific T cells were clonally expanded within the synovial fluid from this donor. In some other donors, short-term T-cell lines derived from synovial fluid lymphocytes, but not those derived from peripheral blood lymphocytes, secreted low levels of tumour necrosis factor (TNF) in response to stimulation with an EBV antigen [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. These results suggested that EBV-specific T cells might form a component of joint infiltrating lymphocytes in patients with RA.</p><p>In the present study we used tetrameric HLA-peptide complexes [<xref ref-type="bibr" rid="B17">17</xref>] to investigate the T-cell response to EBV, cytomegalovirus (CMV) and influenza in individuals with inflammatory arthritis. We made tetramers of HLA molecules complexed to peptide epitopes from EBV latent and lytic proteins [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B10">10</xref>], to an epitope from a CMV structural protein [<xref ref-type="bibr" rid="B18">18</xref>] and to an epitope from influenza A matrix protein [<xref ref-type="bibr" rid="B19">19</xref>]. We used these tetramers to quantify and characterize virus-specific CD8⁺ T cells within samples of peripheral blood and synovial fluid taken from patients with inflammatory arthritis. The experiments show that CD8⁺ T cells specific for certain viral antigens are enriched within synovial fluid. The enrichment could reflect recruitment into, stimulation within or preferential survival within inflamed joints. The virus-specific T cells are activated and subpopulations are able to secrete proinflammatory cytokines. These T cells could therefore interact with synoviocytes and play a role in the maintenance of inflammation in chronic arthritis.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patient samples</title><p>Samples of synovial fluid and peripheral venous blood were initially obtained from a group of 15 EBV-seropositive patients, 13 of whom were HLA-A2 positive and/or HLA-B8 positive. Ten patients (RhA1-RhA10) suffered from RA, one from reactive arthritis (NR1), one from psoriatic arthritis (NR2) and three from osteoarthritis (NR3-NR5). Details of these patients are given in Table <xref ref-type="table" rid="T1">1</xref>. Peripheral blood mononuclear cells (PBMCs) and synovial fluid mononuclear cells (SFMCs) were isolated using Lymphoprep (Nycomed Pharma, Oslo, Norway) gradient centrifugation and cryopreserved in 10% dimethyl sulphoxide, 90% fetal calf syndrome at -70&#x000b0;C. Only synovial lymphocytes were available for analysis from a smaller second group of HLA-A2-positive RA patients (RA3, RA11, RA14 and RA15), whose clinical status has been described elsewhere [<xref ref-type="bibr" rid="B16">16</xref>].</p></sec><sec><title>Human leucocyte antigen typing</title><p>Genomic DNA was extracted from whole blood using a Puregene kit (Gentra systems, Minneapolis, MN, USA) according to the manufacturer's instructions. Each patient was tissue typed using polymerase chain reaction, as previously described [<xref ref-type="bibr" rid="B20">20</xref>]. HLA-A and HLA-B types are summarized in Table <xref ref-type="table" rid="T1">1</xref>.</p></sec><sec><title>Tetrameric class I human leucocyte antigen-peptide complexes.</title><p>HLA-peptide tetramers were produced as previously described [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Briefly, recombinant class I heavy chain or &#x003b2;₂-microglobulin was produced in <italic>Escherichia coli</italic> cells transformed with the relevant expression plasmids. Expression of the heavy chain was limited to the extracellular domain and the sequence of this domain was modified by the addition of a substrate sequence for BirA biotinylation at the carboxyl terminus. The HLA-peptide complexes were folded <italic>in vitro</italic> using 30mg heavy chain, 25mg &#x003b2;₂-microglobulin and 10mg peptide, then biotinylated using purified BirA enzyme (Avidity, Denvery, CO, USA). The biotinylated complexes were recovered by fast performance liquid chromatography (FPLC) purification and ion exchange chromatography. Tetramers were made by mixing biotinylated protein complex with streptavidin-phycoerythrin (PE) at a molar ratio of 4:1. The three HLA-A2 tetrameric complexes synthesized for this study contained either the influenza A virus matrix peptide GILGFVFTL [<xref ref-type="bibr" rid="B19">19</xref>] (A2/FluM tetramer), the EBV BMLF1 peptide GLCTLVAML [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B14">14</xref>] (A2/GLC tetramer), or the CMV pp65 peptide NLVPMVATV [<xref ref-type="bibr" rid="B18">18</xref>] (A2/NLV tetramer). Two HLA-B8 tetramers were also made, with the EBV EBNA3A peptide FLRGRAYGL [<xref ref-type="bibr" rid="B21">21</xref>] (B8/FLR tetramer), or the EBV BZLF1 peptide RAKFKQLL [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>] (B8/RAK tetramer).</p></sec><sec><title>ELISpot assay for IFN-&#x003b3; release by stimulated T cells</title><p>This assay was used to detect IFN-&#x003b3; production by T cells in fresh samples of PBMCs and SFMCs on antigen stimulation. Ninety-six-well polyvinylidene difluoride backed plates (Millipore, Watford, UK) were precoated with 15 &#x003bc; g/ml of anti-IFN-&#x003b3; monoclonal antibody 1-DIK(Mabtech, Stockholm, Sweden). PBMCs were added in duplicate wells at 2.5&#x000d7;10⁵, 1.25&#x000d7;10⁵ and 6.25&#x000d7;10⁴ cells/well and incubated overnight at 37&#x000b0;C, 5% carbon dioxide with the exact 8-mer or 9-mer peptide antigen(2 &#x003bc; mol/l final concentration). The cells were discarded the following day and the second biotinylated anti-IFN-&#x003b3; mono-clonal antibody, 7-B6-1 biotin (Mabtech, Stockholm, Sweden), was added at 1 &#x003bc; g/ml and left for 3h at room temperature, followed by streptavidin-conjugated alkaline phosphatase (Mabtech, Stockholm, Sweden) for a further 2h. Individual cytokine-producing cells were detected as dark spots after a 30-min reaction with 5-bromo-4-chloro-3-indolyl phosphate and nitroblue tetrazolium using a pre-mixed alkaline phosphatase conjugate substrate kit (Bio-Rad, Richmond, CA, USA). The spots were counted under a dissection microscope. The number of specific responders was calculated after subtracting negative control values.</p></sec><sec><title>COS transfections and T cell stimulation assay</title><p>COS cells were transfected with HLA-A2 and pp65 complementary DNA as previously described [<xref ref-type="bibr" rid="B16">16</xref>]. CD8⁺ SFMCs (10⁴ and 10⁵) were incubated with the transfected COS cells for 6h. The culture supernatant was harvested and tested for TNF-&#x003b1; content by measuring culture supernatant cytotoxicity against Wehi 164 clone 13 in a colorimetric assay as previously described [<xref ref-type="bibr" rid="B16">16</xref>].</p></sec><sec><title>Cell staining and flow cytometry</title><p>PBMCs were incubated for 30min in phosphate-buffered saline with 0.16% bovine serum albumin and 0.1% sodium azide, containing 0.5mg/ml phycoerythrin-labelled tetrameric complex, washed and then stained on ice with saturating amounts of an anti-CD8 monoclonal antibody directly conjugated to Tricolor (Caltag Laboratories, San Francisco, CA, USA). For phenotypic analysis, selected samples were also incubated with one of a panel of mono-clonal antibodies directed against cell-surface markers. This panel consisted of anti-CD28 fluorescein isothiocyanate (FITC) (Immunotech, Marseilles, France), anti-CD45RA FITC (Immunotech, Marseilles, France), anti-CD45RO FITC (DAKO, Glostrup, Denmark), anti-CD57 FITC (Becton-Dickinson, Mountain View, CA, USA), anti-CD62L FITC (Pharmingen, San Diego, CA, USA), anti-CD69 FITC (DAKO) and anti-HLA DR FITC (DAKO).</p><p>All samples were fixed in 2% formaldehyde and analyzed using a fluorescent antibody cell sorter using CELLQuest software (Becton-Dickinson). For two-colour analysis 50000 live cells were analyzed. For three-colour analysis 200000 live cells were analyzed. In each experiment the lymphocyte pool was identified using forward and side scatter analysis and markers were set to analyze the CD8^hi subset of CD3⁺ T cells. CD4⁺ T-cell responses were not analyzed in the present study.</p></sec></sec><sec><title>Results</title><sec><title>Presence of T cells specific for Epstein-Barr virus in synovial fluid</title><p>In the first group of patients we used HLA-peptide tetrameric complexes to analyze the frequency of CD8⁺ T cells specific for two EBV lytic protein epitopes, the HLA-A2 restricted epitope (GLCTLVAML) from BMLF1 [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B14">14</xref>] and the HLA-B8 restricted epitope (RAKFKQLL) from BZLF1 [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>], and for an EBV latent protein epitope, the HLA-B8 restricted epitope (FLRGRAYGL) from EBNA3A [<xref ref-type="bibr" rid="B21">21</xref>].</p><p>We initially studied 11 HLA-A2⁺ patients with inflammatory arthritis. In patient RhA6, 9.5% of CD8⁺ T cells within synovial fluid were specific for the GLCTLVAML epitope from BMLF1, whereas only 0.5% CD8⁺ T cells within peripheral blood were specific for this epitope (Figs <xref ref-type="fig" rid="F1">1a</xref> and <xref ref-type="fig" rid="F2">2a</xref>). Likewise in patient RhA5, T cells that were specific for the GLCTLVAML epitope were clearly enriched within synovial fluid (4.5% of CD8⁺ T cells) when compared with peripheral blood (1.0%; Figs <xref ref-type="fig" rid="F1">1b</xref> and <xref ref-type="fig" rid="F2">2a</xref>). In four of the six other HLA-A2⁺ individuals with RA (RhA2, RhA3, RhA8, RhA9), we also found higher frequencies of T cells specific for GLCTLVAML in synovial fluid than in peripheral blood (Fig. <xref ref-type="fig" rid="F2">2a</xref>). This finding was not restricted to patients with RA; we found enrichment of GLCTLVAML-specific T cells in synovial fluid taken from patients with psoriatic arthritis (NR2) and osteoarthritis (NR3; Fig. <xref ref-type="fig" rid="F2">2a</xref>).</p><p>We identified five HLA-B8⁺ patients, three of whom suffered from RA (RhA4, RhA5, RhA7), one of whom had psoriatic arthritis (NR2) and one of whom had osteoarthritis (NR5). In patient RhA5, 13.1% CD8⁺ T cells within synovial fluid reacted with the B8-RAK tetrameric complex compared with 4.3% within peripheral blood (Figs <xref ref-type="fig" rid="F1">1c</xref> and <xref ref-type="fig" rid="F2">2b</xref>). Similarly in patient RhA7, we found that 7.3% of CD8⁺ T cells were specific for the RAKFKQLL epitope in synovial fluid compared with 0.9% CD8⁺ T cells within the periphery (Figs <xref ref-type="fig" rid="F1">1d</xref> and <xref ref-type="fig" rid="F2">2b</xref>). In the third RA patient, as well as in patients with psoriatic arthritis and osteoarthritis, the frequencies of T cells specific for the RAKFKQLL epitope were also much higher in synovial fluid than in peripheral blood (Fig. <xref ref-type="fig" rid="F2">2b</xref>).</p><p>T cells that were specific for the HLA-B8 restricted epitope (FLRGRAYGL) from the EBV latent protein EBNA3A were either undetectable or present at only very low frequencies (&#x0003c;0.1% of CD8⁺ T cells) in three of the five HLA-B8⁺ patients studied. In donor RhA7 we found FLRGRAYGL-specific T cells at a frequency of 0.5% CD8⁺ T cells in peripheral blood and 0.4% CD8⁺ T cells in synovial fluid. In a patient with osteoarthritis (NR4) we found FLRGRAYGL-specific T cells at a frequency of 0.4% in peripheral blood and 0.2% in synovial fluid. Thus, using this technique we did not find evidence for enrichment of T cells specific for this EBV latent epitope within synovial fluid of patients with inflammatory arthritis.</p><p>In general the frequencies of EBV-reactive T cells within peripheral blood of the patients studied were similar to those previously reported in healthy individuals [<xref ref-type="bibr" rid="B7">7</xref>] using this technique, despite the use of immunosuppressive drugs.</p><p>In selected individuals we performed ELISpot assays to detect IFN-&#x003b3; secreted by SFMCs and PBMCs after a short incubation <italic>in vitro</italic> with peptide epitopes from EBV lytic proteins (Fig. <xref ref-type="fig" rid="F3">3</xref>). We calculated that the estimates of frequency of virus-specific T cells obtained with an ELISpot assay for IFN-&#x003b3; secretion were a mean of 19.9-fold and 20.1-fold lower than those obtained with tetramer staining in peripheral blood and synovial fluid, respectively. We have previously shown that in healthy control individuals estimates of frequency of EBV-specific T cells obtained with the ELISpot assay are a mean of 4.4-fold lower than those obtained by staining with tetrameric complexes [<xref ref-type="bibr" rid="B7">7</xref>]. Thus, the ability of the virus-specific T cells to secrete IFN-&#x003b3; appeared to be impaired both in peripheral blood and synovial fluid in the present patient group. The correlation between the numbers of T cells that were detected using an ELISpot assay versus tetramer staining was relatively poor in this group of patients, suggesting interindividual and intersite differences in the functional capacity of the antigen-specific T cells. Despite these limitations, the results of the assays showed enrichment for T cells that were specific for epitopes from EBV lytic proteins within synovial fluid, and showed that subpopulations of these cells are able to secrete proinflammatory cytokines after short-term stimulation.</p></sec><sec><title>T cells specific for cytomegalovirus may also be present at high frequencies within synovial fluid</title><p>In light of recent results that suggest increased responses of synovial lymphocytes from some RA patients to CMV antigens derived from pp65 and IE1 proteins [<xref ref-type="bibr" rid="B16">16</xref>], we also assessed the number of SFMCs specific for a CMV epitope in a second group of four patients. HLA-A2 tetramers carrying a CMV-derived epitope (pp65 NLVPMVATV) were made and used to stain CD8⁺ synovial fluid lymphocytes from four HLA-A2⁺ RA patients (RA3, RA11, RA14 and RA15). Paired samples of peripheral blood from these donors were not available for comparative analysis. The frequency of A2/NLV-reactive cells varied greatly from one patient to another, but reached up to 13.9% in one patient (RA15; Table <xref ref-type="table" rid="T2">2</xref>). Consistent with these findings, CD8⁺ SFMCs from these donors secreted TNF-&#x003b1; after short-term incubation with COS cells transfected with HLA-A2 and pp65 complementary DNA (Table <xref ref-type="table" rid="T2">2</xref>) [<xref ref-type="bibr" rid="B16">16</xref>].</p></sec><sec><title>T cells specific for influenza A are not present at high frequencies within synovial fluid</title><p>We used a tetramer of HLA-A2 complexed with the GILGFVFTL epitope from influenza A matrix protein to stain paired samples of peripheral blood and synovial fluid from six HLA-A2 patients, four of whom had RA, one of whom had osteoarthritis and one of whom had reactive arthritis. Influenza-specific CD8⁺ T cells were undetectable in both PBMC and SFMC preparations in four donors. The frequencies of the influenza-specific CD8⁺ T cells in the other two donors were very low (&#x0003c;0.2%) in peripheral blood and showed no enrichment within synovial fluid.</p></sec><sec><title>Phenotype of Epstein-Barr virus-specific T lymphocytes within peripheral blood and synovial fluid</title><p>We analyzed expression of cell surface markers of activation and differentiation by EBV antigen-specific cells in paired samples of peripheral blood and synovial fluid in donor RhA7 (Figs <xref ref-type="fig" rid="F4">4</xref> and <xref ref-type="fig" rid="F5">5</xref>). CD69 and HLA-DR are often upregulated by activated CD8⁺ T cells; these molecules were found to be expressed at higher levels on RAKFKQLL-specific T cells within synovial fluid than in those found in peripheral blood (Fig. <xref ref-type="fig" rid="F4">4</xref>). Coexpression of CD69 and HLA-DR represents an unusual phenotype, but one that has been described previously within the joint [<xref ref-type="bibr" rid="B22">22</xref>]. CD38 expression has also been used to identify activated cells; this molecule was expressed on 56% of RAKFKQLL-specific T cells within the joint, compared with on 15% within the periphery (Fig. <xref ref-type="fig" rid="F5">5a</xref>).</p><p>L-selectin (CD62L) is downregulated after antigen stimulation, with expression often being regained in the stable memory state [<xref ref-type="bibr" rid="B23">23</xref>]. This molecule was expressed on only 5% of the RAKFKQLL-specific synovial fluid lymphocytes, compared with on 23% of the cells within peripheral blood (Fig. <xref ref-type="fig" rid="F5">5b</xref>).</p><p>CD45RA was not expressed at high levels by significant numbers of RAKFKQLL-specific cells within synovial fluid, but was expressed by 25% of those within peripheral blood (Fig. <xref ref-type="fig" rid="F5">5c</xref>). The vast majority of RAKFKQLL-specific cells within synovial fluid were CD45RO bright; 69% of these antigen-specific cells within the periphery expressed CD45RO [<xref ref-type="bibr" rid="B24">24</xref>] (Fig. <xref ref-type="fig" rid="F5">5d</xref>).</p><p>Expression of CD57, a glycoprotein of unknown function, is thought to occur on CD8⁺ T cells in a late differentiation compartment [<xref ref-type="bibr" rid="B25">25</xref>]. This molecule was expressed on 9% of RAKFKQLL-specific T cells within peripheral blood and on 55% of these cells within synovial fluid (Fig. <xref ref-type="fig" rid="F5">5e</xref>).</p><p>CD28 is a costimulatory molecule that binds B7; down-regulation of CD28 expression is associated with a diminished proliferative capacity and may reflect a state of late-terminal differentiation [<xref ref-type="bibr" rid="B26">26</xref>]. This molecule was expressed on 55% of RAKFKQLL-specific T cells within peripheral blood and on 39% of T cells in synovial fluid (Fig. <xref ref-type="fig" rid="F5">5f</xref>).</p><p>Analysis of GLCTLVAML-specific T cells within peripheral blood and synovial fluid from two further donors, with RA and osteoarthritis, respectively, revealed the same pattern, with EBV-specific T cells within the synovial fluid showing an increase in markers of activation and late differentiation as compared with those in peripheral blood (data not shown).</p></sec></sec><sec><title>Discussion</title><p>The present study shows that virus-specific CD8⁺ T cells are enriched in synovial fluid from individuals with inflammatory arthritis. In particular, CD8⁺ T cells specific for two epitopes from EBV lytic cycle antigens (GLCTLVAML from BMLF1 and RAKFKQLL from BZLF1) may be present at very high frequency within the joints of EBV-seropositive patients. In donor RhA5, staining with tetrameric HLA-peptide complexes showed that T cells specific for these two epitopes accounted for 17.6% of all CD8⁺ T cells within synovial fluid (over 10⁶ cells in a single joint aspirate). The findings described are not specific for RA and we obtained similar results in a patient with psoriatic arthritis (NR2) and in patients with osteoarthritis (NR3 and NR4). We also found very large numbers of T cells specific for an epitope (NLVPMVATV) from the CMV tegument protein pp65 in one patient with RA, and clearly detectable populations of T cells specific for this epitope in three other patients with RA. We did not find high frequencies of T cells specific for an epitope (FLRGRAYGL) from the EBV latent protein EBNA3A or of T cells specific for an epitope (GILGFVFTL) from the influenza A matrix protein within SFMCs in the patients studied. Thus, the synovial T-cell population is not simply a 'concentrated' pool of peripheral memory T cells. Many factors may contribute to the enrichment of T cells specific for certain viral antigens within the joint; these include preferential migration of subsets of T cells, local stimulation within the joint and protection from apoptotic cell death within the joint.</p><p>T-cell migration to sites of inflammation is a carefully controlled process [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>] and T cells with an activated and memory phenotype may be preferentially recruited to a site of inflammation. The experiments described here show that the antigen-specific T cells within PBMCs are relatively more common in the CD62L^lo, CD45RO⁺ and CD45RA^- compartments than in the CD62L^hi, CD45RO^- and CD45RA⁺ compartments. Thus, selection for activated/memory T cells is likely to account at least partly for the observed enrichment of virus-specific T cells within the joints. Current experiments are aimed at investigating the importance of expression of integrins and chemokine receptors by T cells in the recruitment of virus-specific T cells into joints [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>].</p><p>It is also possible that T cells are stimulated to proliferate within the joints. Preliminary experiments (unpublished data) suggest that a small proportion of CD8⁺ T cells (usually &#x0003c;5%), including some EBV-specific T cells, are in cell cycle within synovial fluid. the stimulus to proliferation might be the relevant viral antigen itself. CMV dna has been detected in rheumatoid synovium [<xref ref-type="bibr" rid="B32">32</xref>]. Bcells as well as T cells are recruited into inflamed joints and, in EBV-seropositive individuals, a subpopulation of these Bcells will be latently infected with EBV. One might therefore expect to find this virus within inflamed joints. Although some early studies [<xref ref-type="bibr" rid="B33">33</xref>] found no evidence of EBV infection within joints, other reports [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>] described detection of EBV DNA within the joints of patients with RA and one study [<xref ref-type="bibr" rid="B36">36</xref>] described the use of <italic>in situ</italic> hybridization to detect EBV-encoded small RNA1 and LMP1 transcripts in synovial lining cells from RA patients. Furthermore Koide <italic>et al</italic> [<xref ref-type="bibr" rid="B37">37</xref>] derived a fibroblastoid cell line that expressed EBV proteins from the synovium of a patient with RA. Very recently, Edinger <italic>et al</italic> [<xref ref-type="bibr" rid="B38">38</xref>] reported detection of EBV DNA within synovia from 10 out of 11 patients with RA. That study also provides evidence of transcription of EBV EBER1 and BZLF1 in samples of synovia from patients with RA and osteoarthritis. Thus, expression of BZLF1 within the joint may be stimulating T cells specific for the HLA-B8 restricted RAKFKQLL epitope from BZLF1. Transcription of BMLF1 has, however, not been detected within synovial tissue, suggesting that alternative mechanisms may be responsible for driving the proliferation of CD8⁺ T cells specific for the HLA-A2 restricted epitope (GLCTLVAML) from BMLF1. One possible alternative is that antigen-presenting cells such as dendritic cells may take up EBV antigens and subsequently be recruited to joints where they present epitopes from the EBV antigens by 'cross-presentation' [<xref ref-type="bibr" rid="B39">39</xref>]. A second possibility is that the virus-specific T cells are being stimulated by cross-reactive self-antigens expressed within the joint. In favour of this is the fact that a subpopulation of T cells specific for the HLA-B8 restricted RAKFKQLL peptide epitope is able to cross-react with a self peptide from a serine-threonine kinase [<xref ref-type="bibr" rid="B40">40</xref>].</p><p>Relative resistance to apoptotic cell death is a further theoretical factor that could influence the frequency of the virus-specific T cells within the joint. Although there is evidence that T cells within the joint are protected from apoptosis by type 1 IFN [<xref ref-type="bibr" rid="B41">41</xref>] there is no obvious reason to believe that the T cells specific for the RAKFKQLL and GLCTLVAML epitopes should survive more efficiently than other CD8⁺ T cells.</p><p>The results of the ELISpot assays for IFN-&#x003b3; release after incubation of SFMCs with peptide epitopes from EBV and of the assays for TNF-&#x003b1; release after incubation of SFMCs with COS cells transfected with HLA-A2 and CMV pp65 suggest that the T cells within the synovial fluid retain their capacity to secrete proinflammatory cytokines. Within the joint, the secretion of such cytokines could lead to activation of synoviocytes and hence to the maintenance of inflammation [<xref ref-type="bibr" rid="B42">42</xref>]. Cell-cell contact between the activated virus-specific T cells within the joint and the synoviocyte population is a second mechanism whereby the virus-specific T cells might interact with the indigenous cells within the joint and contribute to the pathogenesis of inflammatory joint disease [<xref ref-type="bibr" rid="B43">43</xref>].</p><p>Importantly, these experiments show that large numbers of T cells within the joint are specific for epitopes from certain viral proteins. Many previous studies analyzed T-cell receptor use of T cells within the joint and found evidence of clonality, and concluded that the T cells are being driven by a specific self-antigen. Paliard <italic>et al</italic> [<xref ref-type="bibr" rid="B44">44</xref>] found that clonality was particularly marked within V&#x003b2; 14⁺ T cells within synovial fluid and that V&#x003b2; 14⁺ T cells were not well represented within peripheral blood of patients with RA. Those authors suggested that a superantigen might have caused activation of V&#x003b2; 14⁺ T cells, with recruitment of selected V&#x003b2; 14⁺ T cell clones into the joint followed by deletion of V&#x003b2; 14⁺ T cells in the periphery. </p><p>It will be interesting to analyze V&#x003b2; usage of our HLA-viral peptide tetramer-reactive CD8⁺ T cells. From the work we have described, it seems likely that at least some of the clonally expanded populations of CD8⁺ T cells found in synovial fluid are specific for viral antigens. The reasons for the presence of large numbers of synovial T cells specific for certain viral epitopes and not others remain unclear, and the role that these virus-specific 'bystander' T cells may play in the maintenance of inflammation needs further investigation.</p></sec>

Apoptosis and p53 expression in rat adjuvant arthritis

<sec><title>Introduction:</title><p>RA is a chronic inflammatory disorder that is characterized by inflammation and proliferation of synovial tissue. The amount of DNA fragmentation is significantly increased in rheumatoid synovium. Only low numbers of apoptotic cells are present in rheumatoid synovial tissue, however. The proportion of cells with DNA strand breaks is so great that this disparity suggests impaired apoptosis. Therefore, the development of novel therapeutic strategies that are aimed at inducing apoptosis in rheumatoid synovial tissue is an attractive goal.</p><p>Although animal models for arthritis only approximate RA, they provide a useful test system for the evaluation of apoptosis-inducing therapies. AA in rats is among the most commonly used animal models for RA. For the interpretation of such studies, it is essential to characterize the extent to which apoptosis occurs during the natural course of the disease. Therefore, we evaluated the number of apoptotic cells and the expression of p53 in various phases of AA.</p></sec><sec><title>Materials and methods:</title><p>In order to generate the AA rat model, Lewis rats were immunized with <italic>Mycobacterium tuberculosis</italic> in mineral oil on day 0. Paw swelling usually started around day 10. For the temporal analysis rats were sacrificed on days 0, 5 (prearthritis), 11 (onset of arthritis), 17 (accelerating arthritis), or 23 (chronic arthritis).</p><p>For the detection of apoptotic cells, the hind paws were harvested on days 0(<italic>n</italic>=6),5 (<italic>n</italic>=6), 11 (<italic>n</italic>=6), 17 (<italic>n</italic>=6), or 23 (<italic>n</italic>=4). The right ankle joints were fixed in formalin, decalcified in ethylenediaminetetra-acetic acid, embedded in paraffin, and sectioned. The TUNEL method was applied. The percentage of TUNEL-positive cells of the total inflammatory cell infiltrate was noted.</p><p>For Western blot analysis, hind paws were harvested on days 0 (<italic>n</italic>=2), 5 (<italic>n</italic>=3), 11 (<italic>n</italic>=4), 17 (<italic>n</italic>=4), or 23 (<italic>n</italic>=4). In addition, hind paws of normal rats (<italic>n</italic>=2) were studied. The right ankle joints were snap frozen and pulverized. Synovial tissue was also obtained by arthroscopy of three patients with longstanding (&#x0003e;5 years) RA. After protein extraction in lysis buffer, equal amounts of protein samples from lysates were pooled and examined by Western bolt analysis using anti-p53 monoclonal antibody D07, which recognizes wild-type and mutant p53 from rodents and humans.</p><p>For immunohistochemical analysis, six rats were sacrificed on day 23 after immunization and synovial tissue of the right ankle joints was snap frozen and evaluated by immunohistochemistry using anti-p53-pan. The sections were evaluated semi-quantitatively using a 0-4 scale.</p><p>The kruskal-Wallis test for several group means was used to compare the percentage of TUNEL-positive cells at different time points. </p></sec><sec><title>Results:</title><p>The percentages of TUNEL-positive cells were strongly dependent on the stage of the disease. Very few TUNEL-positive cells were detected in normal rats or in the early phases of AA; the number of TUNEL-positive cells was 1% or less of the total cell infiltrate, including neutrophils, from days 0-17 (Table <xref ref-type="table" rid="T1">1</xref>). On day 23, however, the percentage of TUNEL-positive cells was significantly increased [15.8&#x000b1;5.1% (mean &#x000b1; standard error of the mean); <italic>P</italic>=0.01]. TUNEL-positive cells were observed in the intimal lining layer and synovial sublining of the invasive front, as well as in the articular cartilage (Fig. <xref ref-type="fig" rid="F1">1</xref>).</p><p>Subsequently, we examined expression of the tumor suppressor gene <italic>p53</italic>, because this is a key regulator of apoptosis. Expression of p53 in pooled rat AA joint extracts gradually increased from day 0 (6 arbitrary units) to day 23 (173 arbitrary units), which was markedly higher than p53 levels in RA synovium (32 arbitrary units; Table <xref ref-type="table" rid="T1">1</xref>). Overexpression of p53 protein on day 23 was confirmed by immunohistochemistry in a separate experiment in six rats with AA. Overexpression of p53 was observed in the intimal lining layer and synovial sublining in all rats on day 23. In all cases a semiquantitative score of 4 was assigned, indicating that 51% or more of the cells were positive, whereas control sections were negative.</p></sec><sec><title>Discussion:</title><p>The results presented here reveal that the number of TUNEL-positive cells remained very low until chronic arthritis developed. This indicates that, although there was sufficient DNA damage to cause an increment in p53 expression in the early phases, DNA strand breaks that can be detected by TUNEL assays only occurred in chronic AA. The observation that TUNEL-positive cells were nearly absent in early AA clearly indicates that only very few cells were undergoing programmed cell death. This is an important observation, which makes it possible to study the effects of apoptosis-inducing therapies <italic>in situ</italic> in early and accelerating AA. An effective therapy would obviously increase the number of TUNEL-positive cells.</p><p>There is already some overexpression of p53 in the preclinical phase and during the onset of the arthritis, with an additional increment in p53 expression during accelerating and chronic arthritis. Presumably, this is wild-type p53, because the disease duration is likely too short to allow for the development of <italic>p53</italic> mutations. Transcription of p53 is probably increased in response to the toxic environment of the inflamed joint. The increased expression of p53 in the joints of rats with chronic AA was even greater than that observed in synovial tissue of RA patients with long-standing disease. </p><p>Overexpression of p53 and increased numbers of apoptotic cells did not occur simultaneously in this model; rather p53 overexpression preceded increased apoptosis. Activation of <italic>p53</italic> leads to induction of cell growth arrest, allowing time for DNA repair. It appears that DNA damage is only extensive enough to induce apoptosis in the latter stages of AA. Factors other than <italic>p53</italic> may also play an important role in the actual induction of apoptosis</p><p>Taken together, significant apoptosis only occurs late in AA and it follows marked p53 overexpression, making it a useful model for testing proapoptotic therapies. AA is not the best model for <italic>p53</italic> gene therapy, however, because dramatic p53 overexpression occurs in the latter stages of the disease.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Tak</surname><given-names>Paul P</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref><email>P.P.Tak@amc.uva.nl</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Klapwijk</surname><given-names>Maartje S</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Broersen</surname><given-names>Sophie FM</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>van de Geest</surname><given-names>Deliana A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Overbeek</surname><given-names>Marieke</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Firestein</surname><given-names>Gary S</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>RA is a chronic inflammatory disorder that is characterized by inflammation and proliferation of synovial tissue. The disease is still associated with long-term morbidity and early mortality, despite treatment with antirheumatic drugs. Inadequate apoptosis appears to contribute toward prolonged survival and constitutive activation of specialized cells in rheumatoid synovium [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. The amount of DNA fragmentation is significantly increased in rheumatoid synovium [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>], which is presumably due to the toxic environment of the chronically inflamed joint [<xref ref-type="bibr" rid="B5">5</xref>]. Only low numbers of apoptotic cells are present in rheumatoid synovial tissue, however [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. The proportion of cells with DNA strand breaks is so great that this disparity suggests impaired apoptosis. The observation that mice with the lymphoproliferative or generalized lymphoproliferative disorder, which have mutations that inactivate Fas and Fas ligand, respectively, develop pathology similar to that observed in immune-mediated diseases [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>] illustrates that reduced apoptosis may play an important role in the pathogenesis of synovitis.</p><p>The <italic>p53</italic> tumor suppressor is a key regulator of DNA repair and cell replication [<xref ref-type="bibr" rid="B11">11</xref>]. DNA damage activates <italic>p53</italic>, thereby inducing cell growth arrest to allow time for DNA repair. When DNA damage is extensive the cells may undergo apoptosis. Inactivation of the <italic>p53</italic> gene renders cells less susceptible to undergo apoptosis [<xref ref-type="bibr" rid="B12">12</xref>]. The p53 system ensures that cells with damaged DNA either die or are repaired. We have previously proposed that impaired apoptosis in rheumatoid synovial tissue might be explained in part by the development of permanent genetic changes in the <italic>p53</italic> tumor suppressor gene [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. In addition, other factors may be involved, such as protection against apoptosis by nuclear factor-&#x003ba;B activation [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>], a relative deficiency of functional Fas ligand in the RA joint [<xref ref-type="bibr" rid="B17">17</xref>], and expression of antiapoptotic molecules, such as bcl-2 [<xref ref-type="bibr" rid="B3">3</xref>] and sentrin [<xref ref-type="bibr" rid="B18">18</xref>]. Therefore, the development of novel therapeutic strategies aimed at inducing apoptosis in rheumatoid synovial tissue is an attractive goal.</p><p>Although animal models of arthritis only approximate RA, they provide a useful test system for the evaluation of apoptosis-inducing therapies. AA in rats is among the most commonly used animal models for RA [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. This model has recently been used to investigate the effects of bisindolylmaleimide, a compound that facilitates Fas-mediated apoptosis [<xref ref-type="bibr" rid="B21">21</xref>]. Rat AA might also provide a useful screening model for the evaluation of gene therapies that are aimed at induction of apoptosis, because the size of the joints permits relatively easy intra-articular injection [<xref ref-type="bibr" rid="B22">22</xref>]. For the interpretation of such studies, however, it is essential to characterize the extent to which apoptosis occurs during the natural course of the disease. Therefore, we evaluated the number of apoptotic cells and the expression of p53 in various phases of AA.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Adjuvant arthritis model</title><p>Male Lewis rats (150-200 g) were immunized at the base of the tail with 1 mg <italic>Mycobacterium tuberculosis</italic> H37RA (Difco, Detroit, MI, USA) in 0.1 ml mineral oil on day 0 [<xref ref-type="bibr" rid="B23">23</xref>]. Paw swelling usually started around day 10. For the temporal analyses, rats were killed on days 0, 5 (prearthritis), 11 (onset of arthritis), 17 (accelarating arthritis), or 23 (chronic arthritis) by carbon dioxide narcosis. All animals were handled in accordance with University of California San Diego Animal Subjects Committee and United States Department of Agriculture guidelines.</p></sec><sec><title>Detection of apoptotic cells</title><p>The hind paws were harvested on days 0 (<italic>n</italic>=6), 5 (<italic>n</italic>=6), 11 (<italic>n</italic>=6), 17 (<italic>n</italic>=6), or 23 (<italic>n</italic>=4). The right ankle joints were fixed in formalin, decalcified for 4 weeks in 15% ethylenediaminetetra-acetic acid in phosphate-buffered saline, embedded in paraffin, and sectioned. For detection of apoptotic cells the TUNEL method was applied, based on terminal deoxynucleotidyl transferase-mediated labeling of free 3'-hydroxy termini exposed in cells that exhibit DNA strand breaks. An <italic>in situ</italic> cell death detection alkaline phosphatase kit from Boehringer Mannheim (Indianapolis, IN, USA) was used according to the manufacturer's instructions. For detection of alkaline phosphatase activity we used the alkaline phosphatase substrate kit I (Fast Red) from Vector Laboratories (Burlingame, CA, USA). The percentage of TUNEL-positive cells of the total inflammatory cell infiltrate was noted.</p></sec><sec><title>Western blot analysis</title><p>Hind paws were harvested on days 0 (<italic>n</italic>=2), 5 (<italic>n</italic>=3), 11 (<italic>n</italic>=4), 17(<italic>n</italic>=4), or 23 (<italic>n</italic>=4). In addition, hind paws of normal rats (<italic>n</italic>=2) were studied. After removal of skin and muscle tissue, the right ankle joints were snap frozen in liquid nitrogen and pulverized. Synovial tissue was also obtained by anthroscopy of three patients with longstanding (&#x0003e;5 years) rheumatoid factor-positive, erosive RA; these patients have been described previously [<xref ref-type="bibr" rid="B24">24</xref>]. All patients had active arthritis in a knee joint and elevated serum levels of C-reactive protein. The patients were treated with nonsteroidal anti-inflammatory drugs. None were treated with corticosteroids or immunosuppressive drugs, such as azathioprine, methotrexate, or cyclophophamide, within 3 months before study entry [<xref ref-type="bibr" rid="B24">24</xref>]. After protein extraction in lysis buffer, equal amounts of protein samples (in total 20 &#x003bc;g/lane) from lysates were pooled and run on a gel in order to normalize for differences in synovial cellularity [<xref ref-type="bibr" rid="B24">24</xref>]. The pooled samples were then transferred onto a nitrocellulose membrane, and p53 protein detected with 0.25 &#x003bc;g/ml of the Immunoglobulin G_2b mouse anti-p53 monoclonal anti-body DO7 (Novocastra Laboratories Ltd, Newcastle, UK), which recognizes wild-type and mutant p53 from rodents and humans. After incubation with horseradish peroxidase-conjugated goat-antimouse antibody, horseradish peroxidase activity was detected using hydrogen peroxide as the substrate and visualized by chemiluminescence. Densitometry was performed with Image software version 1.57 (National Institutes of Health, Bethesda, MD, USA). Results are expressed as arbitrary densitometry units.</p></sec><sec><title>Immunohistochemistry</title><p>Six rats were sacrificed on day 23 immunization, and synovial tissue of the right ankle joints was snap frozen in Tissue-Tek OCT (Miles Diagnostics, Elkhart, IN, USA) by immersion in methylbutane (-70&#x000b0;C). All slides were stained in one procedure. Endogenous peroxidase activity was inhibited using 0.1% sodium azide and 0.3% hydrogen peroxide in phosphate-buffered saline for 30 min. The biotinylated anti-p53-pan (Boehringer Mannheim) was diluted to a final concentration of 2&#x003bc;g/ml and incubated for 60 min. In negative control sections the primary anti-body was omitted or irrelevant antibody was applied at the same concentration as the primary antibody. This was followed by incubation with avidin-biotin-peroxidase complex (Vectastain ABC Kit; Vector Laboratories), biotinylated tyramine, and horseradish peroxidase-conjugated streptavidin, as previously described [<xref ref-type="bibr" rid="B24">24</xref>]. Horseradish peroxidase activity was detected using hydrogen peroxide as substrate and 3,3'-diaminobenzidine (DAB; Vector Laboratories) as dye. Sections were coded and randomly analyzed [<xref ref-type="bibr" rid="B24">24</xref>]. The sections were evaluated semiquantitatively using a 0-4 scale as follows: 0, no staining; 1, rare positive staining or trace staining (1-5%); 2, scattered clusters of positive cells (6-15%); 3, moderate staining in a specific region (16-50%) and 4, extensive staining throughout a region (51-100%) [<xref ref-type="bibr" rid="B24">24</xref>].</p></sec><sec><title>Statistical analysis</title><p>The Kruskal-Wallis test for several group means was used to compare the percentage of TUNEL-positive cells at different time points.</p></sec></sec><sec><title>Results</title><sec><title>Apoptosis in different phases of adjuvant arthritis</title><p>Detection of apoptotic cells was performed on the basis of <italic>in situ</italic> labeling of DNA strand breaks. Representative examples of the TUNEL stainings in relation to the paw volumes in various phases of the disease are shown in Figure <xref ref-type="fig" rid="F1">1</xref>. The percentages of TUNEL-positive cells were strongly dependent on the stage of the disease. Very few TUNEL-positive cells were detected in normal rats or during the early phases of AA; the number of TUNEL-positive cells was 1% or less of the total cell infiltrate, including neutrophils, from days 0-17 (Table <xref ref-type="table" rid="T1">1</xref>). On day 23, however, the percentage of TUNEL-positive cells was significantly increased (15.8&#x000b1;5.1% [mean &#x000b1; standard error of the mean]; <italic>P</italic>=0.01). TUNEL-positive cells were observed in the intimal lining layer and synovial sublining of the invasive front as well as in the articular cartilage (Fig. <xref ref-type="fig" rid="F1">1</xref>).</p></sec><sec><title>Expression of p53 in different phases of adjuvant arthritis</title><p>Subsequently, we examined expression of the tumor suppressor gene <italic>p 53</italic>, because this is a key regulator of apoptosis. Expression of p53 in pooled rat AA joint extracts gradually increased from day 0 (6 arbitrary units) to day 23 (173 arbitrary units), which was markedly higher than <italic>p53</italic> levels in RA synovium (32 arbitrary units; Table <xref ref-type="table" rid="T1">1</xref> and Fig. <xref ref-type="fig" rid="F2">2</xref>). Overexpression of p53 protein on day 23 was confirmed by immunohistochemistry in a separate experiment in six rats with AA. Overexpression of p53 was observed in the intimal lining layer and synovial sublining in all rats on day 23 (Fig. <xref ref-type="fig" rid="F3">3</xref>). In all cases a semiquantitative score of 4 was assigned, indicating that 51% or more of the cells were positive, whereas control sections were negative.</p></sec></sec><sec><title>Discussion</title><p>Rat AA is a T-cell dependent disease, which is characterized by paw swelling, joint erosions and ankylosis, as well as systemic manifestations. Infiltration of the synovium by leukocytes precedes the development of clinical signs and symptoms of arthritis [<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>]. There is also an increase in the numbers of CD8⁺ T cells and B cells in the regional lymph nodes in the preclinical phase [<xref ref-type="bibr" rid="B27">27</xref>]. As shown in the present study, clinical signs of arthritis usually appear by days 10-12. Subsequently, paw volume markedly increases as a result of cellular infiltration and edema of synovial tissue [<xref ref-type="bibr" rid="B28">28</xref>].</p><p>The results presented here reveal that the number of TUNEL-positive cells remained very low until chronic arthritis developed. Severe disease and marked paw swelling characterize this phase. The results indicate that, although there was sufficient DNA damage to cause an increment in p53 expression in the early phase, DNA strand breaks that can be detected by TUNEL assays only occurred in chronic AA. In general, the results may be false positive, because TUNEL-positive cells are not necessarily apoptotic [<xref ref-type="bibr" rid="B2">2</xref>]. The observation that TUNEL-positive cells were nearly absent in early AA, however, clearly indicates that only very few cells were undergoing programmed cell death. We have recently shown that there is increased oxidative stress in chronic arthritis [<xref ref-type="bibr" rid="B29">29</xref>], which may play a pivotal role in the induction of DNA strand breaks [<xref ref-type="bibr" rid="B5">5</xref>]. The results in the AA model suggest that the production of reactive oxygen and nitrogen species is not sufficiently increased in the earliest phases of arthritis to lead to induction of apoptosis. This is an important observation, which makes it possible to study the effects of apoptosis-inducing therapies <italic>in situ</italic> in early and accelerating AA. An effective therapy would obviously increase the number of TUNEL-positive cells. </p><p>There is already some overexpression of p53 in the preclinical phase and during the onset of the arthritis, with an additional increment in p53 expression during accelerating and chronic arthritis. Presumably, this is wild-type p53, because the disease duration is probably too short to allow for the development of <italic>p53</italic> mutations. Transcription of <italic>p53</italic> is probably increased in response to the toxic environment of the inflamed joint, with local production of oxygen radicals, nitric oxide [<xref ref-type="bibr" rid="B30">30</xref>], and cytokines. Similarly, exposure of human fibroblasts to nitric oxide generated from a nitric oxide donor or from overexpression of inducible nitric oxide synthase may result in accumulation of wild-type p53 [<xref ref-type="bibr" rid="B31">31</xref>]. There are also several examples of overexpression of wild-type p53 in human inflammatory diseases, which include the following: inflammation in atherosclerotic plaques [<xref ref-type="bibr" rid="B32">32</xref>], idiopathic pulmonary fibrosis [<xref ref-type="bibr" rid="B33">33</xref>], <italic>Helicobacter pylori</italic>-associated gastritis [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>], ulcerative colitis [<xref ref-type="bibr" rid="B36">36</xref>], Crohn's disease [<xref ref-type="bibr" rid="B36">36</xref>], chronic pancreatitis [<xref ref-type="bibr" rid="B37">37</xref>], infectious colitis [<xref ref-type="bibr" rid="B38">38</xref>], lymphocytic thyroiditis [<xref ref-type="bibr" rid="B39">39</xref>], and RA [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref>]. The increased expression of p53 in the joints of rats with chronic AA was even greater than that observed in synovial tissue of RA patients with long-standing disease. Because there is already dramatic p53 overexpression in the latter stages of the disease, AA is probably not the best model to evaluate <italic>p53</italic> gene therapy. Indeed, we could achieve only a marginal additional increase in p53 expression <italic>in vivo</italic> gene transfer (unpublished data).</p><p>Overexpression of p53 and increased numbers of apoptotic cells did not occur simultaneously in this model; rather p53 overexpression preceded increased apoptosis. Activation of p53 leads to induction of cell growth arrest, allowing time for DNA repair. It appears that DNA damage is only extensive enough to induce apoptosis in the latter stages of AA. Factors other than <italic>p53</italic> may also play an important role in the actual induction of apoptosis, such as tumor necrosis factor-&#x003b1;, interactions between Fas and Fas ligand, and degranulation of granules that contain granzymes and perforin.</p><p>Taken together, significant apoptosis only occurs late in AA and it follows marked p53 overexpression, making it is useful model for testing proapoptotic therapies. AA is not the best model for <italic>p53</italic> gene therapy, however, because dramatic p53 overexpression occurs in the latter stages of the disease.</p></sec>

Rheumatoid arthritis associated autoantibodies in patients with synovitis of recent onset

<sec><title>Introduction:</title><p>A spectrum of autoantibodies is now known to be specifically associated with RA. There continues to be uncertainty as to what stage of the disease each of these autoantibodies develop, and whether they are associated with unique clinical features.</p></sec><sec><title>Aims:</title><p>To help address these questions, a spectrum of autoantibodies known to be associated with RA in a cohort of patients with early synovitis was evaluated.</p></sec><sec><title>Methods:</title><p>An inception cohort of 238 patients having peripheral joint synovitis of less than 12 months duration was evaluated clinicially then followed prospectively for 1 year. Patients were classified as having RA on the basis of fulfilling the 1987 criteria. Serum samples collected at the time of the initial evaluation were tested for anti-Sa and anti-RA-33 using immunoblotting, and to (pro)filaggrin (AFA), anti-CCP, and calpastatin (anti-RA-1) using enzyme-linked immunosorbent assay techniques. AKA were detected using immunoflurescence on human epidermal tissue. RF was tested by nephelometry. HLA-DRB1 alleles were determined using sequence specific primers. Initial and 1 year radiographs were evaluated for the presence of erosions.</p></sec><sec><title>Results:</title><p>Of the 238 patients with synovitis of recent onset in the cohort, 106 (45%) met RA criteria, 102 (96%) of whom met the criteria on their initial visit. Diagnoses in the remaining patients included 22 (9%) with reactive arthritis, 14 (6%) with psoriatic arthritis or another form of spondylarthropathy, 11 (5%) with another well-defined rheumatic diagnosis, and 85 (36%) with undifferentiated arthritis. The RA patients were significantly older than the nonRA patients (46 &#x000b1; 13 versus 39 &#x000b1; 13; <italic>P</italic> &#x0003c; 0.001), had higher mean swollen joint count (13.8 &#x000b1; 9.7 versus 2.3 &#x000b1; 2.3; <italic>P</italic> &#x0003c; 0.001), and higher C-reactive protein (CRP) level (1.9 &#x000b1; 1.9 versus 1.6 &#x000b1; 2.4; <italic>P</italic> &#x0003c; 0.01). Table <xref ref-type="table" rid="T1">1</xref> summarizes the prevalence of the various RA associated antibodies in patients diagnosed as having RF-positive (RF+) RA, RF-negative (RF-) RA, and nonRA. Regarding the characteristics of these tests, RF had the highest sensitivity at 66%, and all the other antibodies individually were less than 50% sensitive. AFA, anti-Sa, anti-CCP were greater than 90% specific for RA, while RF and AKA were 80-90% specific, and anti-RA-33 and anti-RA-1 was not specific for this diagnosis. The data further indicate that adding any one of AFA, AKA, anti-Sa, or anti-CCP to RF increases the specificity for RA from 80 to 90%. In the absence of RF, the presence of one or more of these antibodies carried a sensitivity of only 31% for RF- RA, with anti-Sa being the most specific at 98%. Overall, there was a high degree of correlation between AFA, AKA, anti-Sa or anti-CCP, this being highest between anti-Sa and anti-CCP (odds ratio, 13.3; <italic>P</italic> &#x0003c; 0.001). Despite this high level of correlation, of the 101 patients who were positive for at least one of these four autoantibodies, 57% were positive for only one, suggesting considerable variability in individual reactivity patterns.</p><p>RA has been shown in multiple populations to be associated with HLA-DRB1 alleles encoding for the shared epitope (SE). In this study, as illustrated in Table <xref ref-type="table" rid="T2">2</xref>, the presence of each of these autoantibodies was significantly associated with having two shared epitope alleles, even when only the RA patients were considered.</p><p>Patients with anti-Sa antibodies were predominantly male (61% versus 28%; <italic>P</italic>&#x0003c;0.01), had significantly higher swollen joint counts (18 &#x000b1; 12 versus 13 &#x000b1; 9; <italic>P</italic>=0.02), and higher CRP levels (2.6 &#x000b1; 3 mg/dl versus 1.6 &#x000b1; 1.4 mg/dl; <italic>P</italic>=0.03) at the initial visit. Despite subsequently begin treated with significantly higher doses of prednisone (4.8 &#x000b1; 6.0 mg/day versus 1.8 &#x000b1; 3.3 mg/day; <italic>P</italic>&#x0003c;0.01), and more disease modifying antirheumatic drug therapy (1.4 &#x000b1; 0.8 versus 0.9 &#x000b1; 0.7 disease modifying antirheumatic drugs; <italic>P</italic>&#x0003c;0.01), the anti-Sa-positive RA patients had a higher frequency of erosions than the rest of the RA patients (60% versus 33%; <italic>P</italic>=0.03). Neither RF nor SE were associated with the disease severity measures, and analyses evaluating all the other autoantibodies failed to reveal a similar trend.</p></sec><sec><title>Discussion:</title><p>Despite a well-documented lack of specificity, RF continues to be a central part of the definition of RA, primarily because of its favourable sensitivity profile. In our cohort, RF had a sensitivity of 66%, a specificity of 87%, and an overall accuracy of 78% for the diagnosis of RA. AFA, anti-Sa, anti-CCP were all highly specific for this diagnosis, and when any of them were present in conjunction with RF, the specificity for RA approached 100%. Potentially of more importance to the clinician is the diagnostic value of these antibodies when RF is not detectable. Our data indicate that only 31% of RF- RA patients had any of AKA, AFA, anti-Sa or anti-CCP, and that anti-Sa was the most specific for this diagnosis. This modest level of sensitivity suggests that testing for this spectrum of autoantibodies carries little advantage over RF alone in diagnosing early RA.</p><p>AFA, AKA, and antiperinuclear factor (APF) have all been proposed to identify a common antigen present in the skin protein (pro)filaggrin. It has continued to be puzzling why a skin antigen would be targeted relatively specifically in a disorder that is primarily articular. A potential explanation for this may relate to the demonstration that citrulline appears to be an essential constituent of the antigenic determinants recognized by AKA, APF, and AFA. The citrulline rich (pro)filaggrin molecule makes an ideal substrate for detecting this reactivity. Moreover, the SA antigen, which, unlike (pro)filaggrin, is detectable in rheumatoid synovium, has recently been shown to also be citrullinated. It is thus possible that AKA, AFA, APE, and anti-Sa all recognize one or more citrullinated antigens. Despite this possibility, the modest degree of concordance between them in individual patient sera suggests that it is unlikely that a single antigen is involved in generating these responses.</p><p>This study provides evidence suggesting that anti-Sa antibodies appear to be a marker for a subset of early RA patients whose disease may be more severe and erosive. Moreover, it was determined that anti-Sa, AFA, and anti-CCP were all highly associated with SE, particularly two copies. We examined a spectrum of potential RA severity indicators including the number of swollen joints, CRP level, and presence of early radiographic erosions. Our data indicate that anti-Sa was more highly associated with these measures of RA severity than any other parameter, including the most accepted prognostic indicators, RF and SE.</p><p>In conclusion, it is demonstrated that antibodies directed against putatively citrullinated antigens including SA, filaggrin, keratin, and CCP are the most specific for RA, and are detectable early in the disease course. It will be of interest to find out whether the cumulative prevalence of specific autoantibody subsets tends to increase over time, as this would suggest that the mechanisms underlying the development of these reactivities continue to evolve over the course of the arthropathy.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Goldbach-Mansky</surname><given-names>Raphaela</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Lee</surname><given-names>Jennifer</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>McCoy</surname><given-names>Angela</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Hoxworth</surname><given-names>Joseph</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Yarboro</surname><given-names>Cheryl</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Smolen</surname><given-names>Josef S</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Steiner</surname><given-names>G&#x000fc;nter</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Rosen</surname><given-names>Antony</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Zhang</surname><given-names>Cindy</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A10" contrib-type="author"><name><surname>M&#x000e9;nard</surname><given-names>Henri A</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A11" contrib-type="author"><name><surname>Zhou</surname><given-names>Zhi Jie</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A12" contrib-type="author"><name><surname>Palosuo</surname><given-names>Timo</given-names></name><xref ref-type="aff" rid="I5">5</xref></contrib><contrib id="A13" contrib-type="author"><name><surname>Van Venrooij</surname><given-names>Walther J</given-names></name><xref ref-type="aff" rid="I6">6</xref></contrib><contrib id="A14" contrib-type="author"><name><surname>Wilder</surname><given-names>Ronald L</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A15" contrib-type="author"><name><surname>Klippel</surname><given-names>John H</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A16" contrib-type="author"><name><surname>Schumacher</surname><given-names>H Ralph</given-names><suffix>Jr</suffix></name><xref ref-type="aff" rid="I7">7</xref></contrib><contrib id="A17" contrib-type="author"><name><surname>EI-Gabalawy</surname><given-names>Hani S</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>elgabala@exchange.nih.gov</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Autoantibodies can be detected in a spectrum of rheumatic diseases where they may be highly associated with distinct clinical syndromes. These are often helpful for diagnosis, and to some extent, prognosis. In RA, RF is detected in 70-80% of patients with established disease, and is an integral part of the definition of this disorder. AKA, APF, AFA, and anti-Sa have all been shown to be associated with RA, and appear to be more specific than RF for this disease [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. Anti-RA-33, [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>] and anti-RA-1 [<xref ref-type="bibr" rid="B13">13</xref>] have also been shown to be prevalent in RA patient populations, but not specific for this disease.</p><p>There is an evolving understanding of the antigens to which these RA associated antibodies are directed, although their ultimate pathogenic role, if any, continues to be unclear. AKA, APF, and AFA identify epitopes carried by keratin, profilaggrin and filaggrin, respectively [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. These proteins follow a common post-translational pathway, which involves a peptidyl-arginine to citrulline deimination. This modification appears to be central to specific antigenicity, giving rise to epitopes recognized by sera from RA patients [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. Anti-Sa antibodies are directed towards an unknown antigen that is abundant in placental tissue and rheumatoid synovium [<xref ref-type="bibr" rid="B7">7</xref>]. Interestingly, the Sa antigen also appears to be citrullinated [<xref ref-type="bibr" rid="B18">18</xref>]. Anti-RA-33 antibodies recognize the A2 protein, an antigen found in the heterogeneous ribonucleoprotein of the splicosome [<xref ref-type="bibr" rid="B19">19</xref>]. Anti-RA-1 antibodies are directed towards domains 3 and 4 of calpastatin, the natural inhibitor of calpains, which are members of the cysteine proteinases that have been implicated in articular damage [<xref ref-type="bibr" rid="B13">13</xref>].</p><p>This spectrum of autoantibodies has been evaluated primarily in cohorts of patients with well-established RA. A number of important questions are unresolved regarding the stage of the disease at which each of these antibodies arise, and whether indeed some may antedate the clinical expression of this disorder. The fact that RF, AKA or APF could sometimes antedate the appearance of clinical RA by several years has been reported [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. Anti-Sa, AFA and anti-RA-33 have been reported to be present in early RA patients, although the cohorts studied have generally been small, and the clinical characterization of the patients not detailed [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. Anticyclic citrullinated peptide (anti-CCP) has recently been evaluated in a cohort of patients with early synovitis, and was found to be more specific than RF for early RA, while having comparable sensitivity [<xref ref-type="bibr" rid="B25">25</xref>].</p><p>Recently, the clinical features and human leukocyte antigen (HLA) associations of a well-characterized cohort of patients with synovitis of recent onset were presented [<xref ref-type="bibr" rid="B26">26</xref>]. In the current study, we sought to determine the prevalence and diagnostic value of the RA associated autoantibodies in this heterogeneous cohort of patients with early synovitis. Sera obtained within 12 months of symptom onset were tested for these antibodies, and their presence was related to disease features and subsequent clinical course over a 1 year period. Our data indicate that AFA, anti-Sa, and anti-CCP are all highly specific for early RA, but demonstrate only a modest degree of concordance in the sera of individual patients. Anti-Sa identifies a subset of male RA patients with severe disease.</p></sec><sec sec-type="methods"><title>Methods</title><sec><title>Patients</title><p>Two hundred and thirty-eight patients were recruited to an early synovitis study at the National Institutes of Health (protocol 94-AR-194). The study patients had persistent synovitis (&#x0003e;6 weeks) of at least one peripheral joint, which had been present for less than 1 year. Patients with traumatic, septic, and crystal induced arthritis were specifically excluded.</p><p>Patients underwent a complete clinical evaluation. The number of swollen joints was determined by evaluating for the presence of effusion and/or synovial thickening in 66 peripheral joints (hips were excluded). Patients were then followed clinically over a 1 year period. Routine laboratory data were obtained on each visit, including an evaluation of acute phase reactants. Anteroposterior and lateral radiographs of the hands and feet were either available for evaluation or obtained at the time of assessment. A second set of radiographs was obtained at the 1 year visit. All radiographs were evaluated for the presence of erosions. Although a proportion of the radiographs were felt to have 'possible' or 'questionable' erosions, only patients with definite erosions on either set of radiographs were included in the definition of erosions. At the completion of the year of observation, patients who were diagnosed as having RA had all met the 1987 American College of Rheumatology (ACR) criteria [<xref ref-type="bibr" rid="B27">27</xref>] on at least one visit. If a patient had met ACR criteria, but with follow up was unequivocally diagnosed as having another well-characterized rheumatic disease, this individual was not included in the definition of RA.</p></sec><sec><title>Detection of autoantibodies</title><p>IgM RF was measured by nephelometry, and a level &#x0003e;20 IU/ml was considered positive. Anti-Sa antibodies were detected by immunoblot, as previously described [<xref ref-type="bibr" rid="B7">7</xref>]. A partially purified preparation of placental Sa antigen was provided by one of the authors (HAM), and the western blot assay performed blindly by another (GS). RA-33 was detected using immunoblot, as described previously [<xref ref-type="bibr" rid="B19">19</xref>]. Filaggrin was obtained from human epidermis and purified using reversed phase high-performance liquid chromatography. After covalently coupling to 96-well plates, AFA were detected by enzyme-linked immunosorbent assay (ELISA) as previously described [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B5">5</xref>]. The cutoff level for positivity (OD, 0.15) was determined on the basis of results of sera from 100 middle-aged (40-65 years) blood donors. Antibodies to CCP were evaluated after generating a cyclic peptide from linear citrulline containing peptides, by substituting serine residues by cysteine. Anti-CCP were detected using ELISA as previously described with the linear peptides [<xref ref-type="bibr" rid="B17">17</xref>]. Control peptides had an unmodified arginine rather than citrulline. The cutoff level for positivity (OD, 0.3) was determined on the basis of generating 100% specificity for RA in previous assays using local normal controls. AKA were detected using immunoflurescence. Sections of human breast epidermis were used, and the slides were read independently by two observers (AR and CZ). Any slides in which the observers did not agree were reread, and a consensus agreed. In all cases, the detected autoantibodies were of the IgG class. The results were not corrected for total IgG levels.</p></sec><sec><title>HLA typing</title><p>HLA-DR typing was done by the molecular polymerase chain reaction-sequence specific primers methods using specific oligonucleotide sequences as primers as previously described [<xref ref-type="bibr" rid="B28">28</xref>]. Shared epitope (SE) alleles included DRB1^*0101, ^*0102, ^*0401, ^*0404, ^*0405, ^*0408, ^*1001, ^*1402.</p></sec><sec><title>Statistical analysis</title><p>Patient groups were compared using analysis of variance of the Kruskal-Wallis test for continuous variables, and using the Chi-squared test for proportions. The significance level for associations between individual autoantibodies and HLA-DRB1 alleles was adjusted for multiple comparisons using the Bonferroni method. Statistical analysis was performed using Epi Info statistical software (Center for Disease Control, Atlanta, GA, USA: http://www.cdc.gov/epo/epi/epiinfo.htm).</p></sec></sec><sec><title>Results</title><p>Of the 238 patients with synovitis of recent onset in the cohort, 106 (45%) met RA criteria, 102 (96%) of whom met the criteria on their initial visit. Four patients initially met RA criteria, but were given nonRA diagnoses on follow up. Two of these patients had SLE, and two had psoriatic arthritis. All four presented with RF- polyarthritis. Diagnoses in the remaining nonRA patients included 22 (9%) with reactive arthritis, 14 (6%) with psoriatic arthritis or another form of spondylarthropathy, 11 (5%) with another well-defined rheumatic diagnosis, and 85 (36%) with undifferentiated arthritis. The clinical characteristics of the RF-positive (RF+) RA, RF-negative (RF-) RA and nonRA patients are summarized in Table <xref ref-type="table" rid="T1">1</xref>.</p><sec><title>Prevalence of the autoantibodies in the patient cohort</title><p>Table <xref ref-type="table" rid="T1">1</xref> indicates the prevalence of a spectrum of known autoantibodies and of the RA associated antibodies in patients diagnosed as having RF+ RA, RF- RA, and nonRA. In total, 87 patients were RF-positive, 17 of whom did not meet RA criteria any point during the study period, and were classified as having undifferentiated arthritis. Antinuclear antibodies were detected in 65/238 (27%) patients, and were of comparable prevalence in the patient groups. Less than 10% of the patients in all groups demonstrated antibodies to SSA, SSB, dsDNA, ribonucleoprotein (RNP), and Sm. Of the RA associated antibodies tested, AFA, anti-Sa, anti-CCP, and AKA were all significantly more prevalent in the RF+ RA group compared with either the RF- RA group or the nonRA group. Anti-RA-1 antibodies were detected in 35% of the cohort with no significant difference between the groups. Anti-RA-33 antibodies were detected in three patients in the cohort. Please note, the prevalence of anti-Sa antibodies was significantly higher in the RF- RA group compared with the nonRA group (14% versus 2%; <italic>P</italic>&#x0003c;0.01), but there were no differences in the prevalence of the other autoantibodies between these two patient groups. In total, 90/106 (85%) of the RA patients and 75/132 (57%) of the nonRA patients had at least one of the RA associated antibodies.</p></sec><sec><title>Diagnostic value of antibodies for RA</title><p>The diagnostic value of the individual antibodies for RA is shown in Table <xref ref-type="table" rid="T3">3</xref>. RF had the highest sensitivity at 66%, and all the other antibodies were individually less than 50% sensitive. AFA, anti-Sa and anti-CCP were greater than 90% specific for RA, while RF and AKA were 80-90% specific, and anti-RA-1 was not specific for this diagnosis. The data further indicate that adding any one of AFA, AKA, anti-Sa, or anti-CCP to RF increases the specificity for RA from 80 to 90%. In the absence of RF, the presence of one or more of these antibodies carried a sensitivity of only 31% for RF- RA, with anti-Sa being the most specific at 98%. At the completion of the study, of the 37/132 (28%) nonRA patients positive for one of AFA, AKA, anti-Sa, or anti-CCP, 12 had some form of spondylarthropathy, 4 had a connective tissue disease, and 21 continued with a diagnosis of 'undifferentiated arthritis'.</p></sec><sec><title>Correlation between antibodies in patient sera</title><p>Previous studies have generally suggested a high degree of correlation in RA patient sera between AFA, AKA, anti-CCP, and anti-Sa. We tested the correlation between these antibodies in the patient sera. The correlation was highest between anti-CCP and anti-Sa [odds ratio (OR), 13.3; <italic>P</italic>&#x0003c;0.001), and lowest between anti-Sa and AKA (OR, 2.4; <italic>P</italic>=0.09). In total, 101 patients had at least one of the four antibodies, 64% of whom had RA. Of these 101 patients, 57 were in fact positive for only one antibody (AKA = 20, AFA = 12, anti-Sa = 6, anti-CCP =19). Only seven patients were positive for all four antibodies.</p><p>Seropositivity for anti-CCP and AFA was determined on the basis of an ELISA cut-off level as described in the Methods section. The relationship between the titers of these two antibodies, and between each of them and RF, was examined. There was a significant correlation in the titers of anti-CCP and AFA in individual sera (<italic>r</italic>=0.46; <italic>P</italic>&#x0003c;0.001). Titers of both anti-CCP and AFA were not correlated with RF titers. When only anti-CCP-positive patients were examined, the mean CCP titer was significantly higher in the RF-positive patients (<italic>n</italic>=44) compared with the RF-negative patients (<italic>n</italic>=10) (1.07 &#x000b1; 0.55 versus 0.6 &#x000b1; 0.34; <italic>P</italic>=0.01). A similar tread was seen with AFA, although this did not reach statistical significance. Of note also, the mean RF titer for RF-positive patients who did not meet RA criteria did not differ from the RF-positive patients who were diagnosed as having RA (202 &#x000b1; 172 versus 207 &#x000b1; 169; <italic>P</italic>= not significant).</p></sec><sec><title>RF, Sa, AFA, and CCP are highly associated with shared epitope in early RA</title><p>RA has been shown in multiple populations to be associated with HLA-DR alleles encoding for the SE. We sought to determine if specific autoantibodies were particularly associated with SE alleles in our cohort. These data are presented in Table <xref ref-type="table" rid="T2">2</xref>. Overall, the SE alleles were highly associated with the diagnosis of RA (OR 3.0; <italic>P</italic>&#x0003c;0.0001). RF, AFA, anti-Sa, and anti-CCP were all associated with SE alleles, particularly two copies. Interestingly, 6/6 patients with ^*0101/^*0401 were positive for at least one of these autoantibodies.</p></sec><sec><title>Anti-Sa antibodies identify a male predominant subset of RA patients with severe disease</title><p>Table <xref ref-type="table" rid="T4">4</xref> demonstrates that, in comparison with all other RA patients, RA patients with anti-Sa antibodies were predominantly male (61% versus 28%; <italic>P</italic>&#x0003c;0.01), had significantly higher swollen joint counts (18 &#x000b1; 12 versus 13 &#x000b1; 9; <italic>P</italic>=0.02), and higher CRP levels (2.6 &#x000b1; 3 mg/dl versus 1.6 &#x000b1; 1.4 mg/dl; <italic>P</italic>=0.03) at the initial visit. Despite subsequently being treated with significantly higher doses of prednisone (4.8 &#x000b1; 6.0 mg/day versus 1.8 &#x000b1; 3.3 mg/day; <italic>P</italic>&#x0003c;0.01), and more disease modifying antirheumatic drug therapy (1.4 &#x000b1; 0.8 versus 0.9 &#x000b1; 0.7 disease modifying antirheumatic drugs; <italic>P</italic>&#x0003c;0.01), the anti-Sa-positive RA patients had a higher frequency of erosions that the rest of the RA patients (60% versus 33%; <italic>P</italic>=0.03). Table <xref ref-type="table" rid="T4">4</xref> further indicates that neither RF nor SE was associated with the disease severity measures, and analyses evaluating all the other autoantibodies failed to reveal a similar trend (data not shown). Of interest, only six RA patients demonstrated the presence of nodules, and this feature was more associated with RF and SE than with anti-Sa. Analysis of the nonRA patients did not reveal any clinically meaningful associations with any of the autoantibodies.</p></sec></sec><sec><title>Discussion</title><p>A cohort of patients with synovitis of recent onset was evaluated and we sought to determine the prevalence and potential clinical utility of a spectrum of autoantibodies that has been shown to be associated with RA. To date, most of the studies evaluating these antibodies have been in patients with established, well-characterized disease, and their prevalence and diagnostic value in patients with early inflammatory arthritis has not been defined. Our study evaluated patients within a few months of the onset of synovitis, and then followed them for a 1 year period to determine the best clinical diagnosis. In particular, we sought to determine how sensitive and specific each of the antibodies, alone and in combination, were for early RA. We utilized the accepted definition of RA using the 1987 ACR criteria [<xref ref-type="bibr" rid="B27">27</xref>]. Of note, a large epidemiologic study of patients with early inflammatory arthritis found that the proportion of patients who met the ACR criteria increased if the criteria were applied cumulatively over a 5 year period [<xref ref-type="bibr" rid="B29">29</xref>]. Although the duration of follow up in this study was shorter, we found that almost all of the patients who had met the RA criteria at the completion of the study period had done so on their initial visit.</p><p>Despite a well-documented lack of specificity, RF continues to be a central part of the definition of RA, primarily because of its favorable sensitivity profile. In our cohort, RF had a sensitivity of 66%, a specificity of 87%, and an overall accuracy of 78% for the diagnosis of RA. AFA, anti-Sa and anti-CCP were all highly specific for this diagnosis, and when any of them were present in conjunction with RF, the specificity for RA approached 100%. Potentially of more importance to the clinician is the diagnostic value of these antibodies when RF is not detectable. Our data indicate that only 31% of RF- RA patients had any of AKA, AFA, anti-Sa or anti-CCP, and that anti-Sa was the most specific for this diagnosis. This modest level of sensitivity suggests that testing for this spectrum of autoantibodies carries little advantages over RF alone in diagnosing early RA.</p><p>This study provides evidence suggesting that anti-Sa anti-bodies appear to be a marker for a subset of early RA patients whose disease may be more severe and erosive. These data are consistent with observations in a French RA cohort, where anti-Sa and SE were associated with severe radiographic erosions [<xref ref-type="bibr" rid="B9">9</xref>]. Indeed, it was determined that anti-Sa, AFA, and anti-CCP were all highly associated with SE, particularly two copies. We examined a spectrum of potential RA severity indicators including the number of swollen joints, CRP level, and presence of early radiographic erosions. Our data indicate that anti-Sa was more highly associated with these measures of RA severity than any other parameter, including the most accepted prognostic indicators, RF and SE. In particular, there was a significantly higher prevalence of radiographic erosions in the anti-Sa-positive RA patients compared with the rest of the RA population, despite the fact that the anti-Sa-positive patients had received significantly higher doses of prednisone and more disease modifying antirheumatic drug therapy. Interestingly, the anti-Sa-positive subset in this cohort was strikingly predominated by males (61% of Sa-positive versus 28% of Sa-negative RA patients; <italic>P</italic>&#x0003c;0.001). It has recently been shown that, in comparison with female RA patients, male RA patients tended to develop erosions earlier in the disease course [<xref ref-type="bibr" rid="B30">30</xref>]. The finding that anti-Sa antibodies are associated both with male gender and with severe early RA further emphasizes the importance of gender differences in the clinical expression of this heterogeneous disease.</p><p>AFA, AKA, and APF have been proposed to all identify a common antigen present in the skin protein (pro)filaggrin [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. It has continued to be puzzling why a skin antigen would be targeted relatively specifically in a disorder that is primarily articular. A potential explanation for this may relate to the demonstration that citrulline appears to be an essential constituent of the antigenic determinants recognized by AKA, APF, and AFA [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. The citrulline rich (pro)filaggrin molecule makes an ideal substrate for detecting this reactivity. Moreover, the Sa antigen, which, unlike (pro)filaggrin, is detectable in rheumatoid synovium, has recently also been shown to be citrullinated [<xref ref-type="bibr" rid="B18">18</xref>]. It is thus possible that AKA, AFA, APF, and anti-Sa all recognize one or more citrullinated antigens. As with previous studies, the current study documents a good overall correlation between these antibodies. Nevertheless, the data clearly indicate that reactivity to CCP does not capture that complete spectrum of AFA, AKA, and anti-Sa reactivity. Indeed, it has been shown that 56% of patients who were positive for one or more of these antibodies were positive for only one, and that only 7% of these patients were positive for all of them. Similarly, previous studies evaluating AKA, APF, and AFA have shown varying degrees of discordance in the seropositivity of individual patient sera [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. Moreover, there was a spectrum of reactivity patterns seen when RA sera were tested against various citrulline containing peptides [<xref ref-type="bibr" rid="B17">17</xref>]. Together, these data are most consistent with the hypothesis that individual RA patients respond to unique antigenic determinants, that may preferentially be citrullinated, but that are likely distinct from those of targeted by other RA patients.</p><p>It was surprising to discover that anti-RA-33 antibodies were all but absent in this cohort of patients, particularly in view of the previously reported prevalence of 20-40% in other RA cohorts [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. This antibody was reported to have a prevalence of only 6% in a Finnish cohort of early RA patients [<xref ref-type="bibr" rid="B24">24</xref>]. It should be pointed out that the testing for anti-RA-33, including that performed in the current study, was all performed in the same laboratory (GS). The reasons for this discrepancy are unclear, and may represent inherent differences in the populations studied. Alternatively, it is possible that the development of anti-RA-33 reactivity increases as the diseases progresses. Longitudinal follow up of early RA cohorts, such as the present one, will help to further clarify this issue.</p><p>In the current study, it was demonstrated that antibodies directed against putatively citrullinated antigens including Sa, filaggrin, keratin, and CCP are the most specific for RA, and are detectable early in the disease course. It will be of interest to find out whether the cumulative prevalence of specific autoantibody subsets tends to increase over time, as this would suggest that the mechanisms underlying the development of these reactivities continue to evolve over the course of the arthropathy.</p></sec>

Gene therapy for established murine collagen-induced arthritis by local and systemic adenovirus-mediated delivery of interleukin-4

<sec><title>Introduction:</title><p>Rheumatoid arthritis (RA) is a chronic systemic autoimmune disease that is characterized by joint inflammation, and progressive cartilage and bone erosion. Recent research has identified certain biologic agents that appear more able than conventional therapies to halt effectively the progression of disease, as well as ameliorate disease symptoms. One potential problem with the use of biologic agents for arthritis therapy is the need for daily or weekly repeat dosing. The transfer of genes directly to the synovial lining can theoretically circumvent the need for repeat dosing and reduce potential systemic side effects [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. However, although many genes have been effective in treating murine CIA if administrated at a time before disease onset, local intra-articular or periarticular gene transfer has not been highly effective in halting the progression of established disease. IL-4, similar to tumor necrosis factor (TNF)-&#x003b1; and IL-1 inhibitors, has been shown be therapeutic for the treatment of murine CIA when administered intravenously as a recombinant protein, either alone or in combination with IL-10. IL-4 can downregulate the production of proinflammatory and T-helper (Th)1-type cytokines by inducing mRNA degradation and upregulating the expression of inhibitors of proinflammatory cytokines such as IL-1 receptor antagonist (IL-1Ra) [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. IL-4 is able to inhibit IL-2 and IFN-&#x003b3; production by Th1 cells, resulting in suppression of macrophage activation and the production of the proinflammatory cytokines IL-1, IL-6, IL-8, and TNF-&#x003b1; by monocytes and macrophages [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>].</p></sec><sec><title>Objective:</title><p>In order to examine the therapeutic effects of local and systemic IL-4 expression in established CIA, an adenoviral vector carrying the gene for murine IL-4 (Ad-mIL-4) was generated. The ability of Ad-mIL-4 to treat established CIA was evaluated by local periarticular and systemic intravenous injection of Ad-mIL-4 into mice at various times after disease onset.</p></sec><sec><title>Materials and methods:</title><p>Male DBA/1 lacJ (H-2^{<italic>q</italic>}) mice, aged 7-8 weeks, were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). The mice were immunized intradermally at the base of tail with 100 &#x003bc; g bovine type II collagen. On day 21 after priming, mice received a boost injection (intradermally) with 100 &#x003bc; g type II collagen in incomplete adjuvant. For the synchronous onset of arthritis, 40 &#x003bc; g lipopolysaccharide (Sigma, St Louis, MO, USA) was injected intraperitoneally on day 28. Ad-mIL-4 was injected periarticularly into the hind ankle joints of mice on day 32 or intravenously by tail vein injection on day 29. Disease severity was monitored every other day using an established macroscopic scoring system ranging from 0 to 4: 0, normal; 1, detectable arthritis with erythma; 2, significant swelling and redness; 3, severe swelling and redness from joint to digit; and 4, maximal swelling with ankylosis. The average of macroscopic score was expressed as a cumulative value for all paws, with a maximum possible score of 16 per mouse. Cytokine production by joint tissue or serum were assessed using enzyme-linked immunosorbent assay (ELISA; R&#x00026;D Systems, Minneapolis, MN, USA).</p></sec><sec><title>Results:</title><p>To examine the therapeutic effects of IL-4 gene transfer in a murine model of arthritis, 5&#x000d7;10⁸ particles of Ad-mIL-4 and enhanced green fluorescent protein (Ad-eGFP) were administered by periarticular injection into the ankle joints of mice with established disease 4 days after lipopolysaccharide injection. All mice had established disease at time of injection. As shown in Figure <xref ref-type="fig" rid="F1">1</xref>, the severity of arthritis (Fig. <xref ref-type="fig" rid="F1">1a</xref>), paw thickness (Fig. <xref ref-type="fig" rid="F1">1b</xref>), and the number of arthritic paws (Fig. <xref ref-type="fig" rid="F1">1c</xref>) were all significantly reduced in the Ad-mIL-4 group, compared with the saline- and Ad-eGFP-treated groups. Analysis of the bones in the ankle joints of control arthritic mice showed evidence of erosion with an associated monocytic infiltrate around the joint space compared with the Ad-mIL-4-treated and nonarthritic control joints. In addition, injection of the ankle joints in the hind legs resulted in a therapeutic effect in the front paws. A similar contralateral effect has been observed with adenoviral-mediated delivery of viral (v)-IL-10. Interestingly, a high level of murine IL-10 also was detected from the joint lysates of Ad-mIL-4-treated na&#x000ef;ve and arthritic mice, with the production of endogenous IL-10 correlating with the dose of Ad-mIL-4. The administration of recombinant IL-4 protein systemically has been shown to be therapeutic in murine CIA models if given before disease onset. To examine the effect of systemic IL-4 delivered by gene transfer, 1&#x000d7;10⁹ particles of Ad-mIL-4 were injected via the tail vein of collagen-immunized mice the day after lipopolysaccharide injection. Whereas the immunized control mice, injected with Ad-eGFP, showed disease onset on day 3 after lipopolysaccharide injection, Ad-mIL-4-treated mice showed a delay in disease onset and as a reduction in the total number of arthritic paws. Also, systemic injection of Ad-mIL-4 suppressed the severity of arthritis in CIA mice according to arthritis index.</p></sec><sec><title>Discussion:</title><p>Gene therapy represents a novel approach for delivery of therapeutic agents to joints in order to treat the pathologies associated with RA and osteoarthritis, as well as other disorders of the joints. In the present study we examined the ability of local periarticular and systemic gene transfer of IL-4 to treat established and early-stage murine CIA, respectively. We have demonstrated that both local and systemic administration of Ad-mIL-4 resulted in a reduction in the severity of arthritis, as well as in the number of arthritic paws. In addition, the local gene transfer of IL-4 reduced histologic signs of inflammation and of bone erosion. Interestingly, local delivery of Ad-mIL-4 was able to confer a therapeutic effect to the untreated, front paws through a currently unknown mechanism. In addition, both local and systemic expression of IL-4 resulted in an increase in the level of endogenous IL-10, as well as of IL-1Ra (data not shown). Previous experiments have shown that gene transfer of IL-10 and IL-1 and TNF inhibitors at the time of disease initiation (day 28) is therapeutic. However, delivery of these agents after disease onset appeared to have only limited therapeutic effect. In contrast, the present results demonstrate that IL-4, resulting from local periarticular and systemic injection of Ad-mIL-4, was able partially to reverse progression of established and early-stage disease, respectively. These results, as well as those of others, support the potential application of IL-4 gene therapy for the clinical treatment of RA.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Kim</surname><given-names>Seon Hee</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Evans</surname><given-names>Christopher H</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Kim</surname><given-names>Sunyoung</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Oligino</surname><given-names>Thomas</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Ghivizzani</surname><given-names>Steven C</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Robbins</surname><given-names>Paul D</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>probb@pitt.edu</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>RA is a chronic systemic autoimmune disease that is characterized by joint inflammation, and progressive cartilage and bone erosion. Currently the symptoms of arthritis are managed using pharmacologic agents, including both steroidal and nonsteroidal drugs, and disease-modifying drugs such as methotrexate. No pharmacologic agents have yet proven effective in halting the progression of disease, however. Recent research has identified certain biologic agents that appear more able than conventional therapies to halt effectively the progression of disease, as well as ameliorate disease symptoms. In particular, inhibitors of TNF-&#x003b1; and IL-1 have proven effective in clinical trials, and the US Food and Drug Administration has approved the use of soluble TNF-&#x003b1; receptor for treatment of human RA. One potential problem with the use of biologic agents for arthritis therapy is the need for daily or weekly repeat dosing. The transfer of genes directly to the synovial lining can theoretically circumvent the need for repeat dosing and reduce potential systemic side effects [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Ex-vivo and in-vivo methods have been used to deliver therapeutic genes such as those that encode IL-10, v-IL-10, soluble TNF and IL-1 receptors, and IL-1Ra to arthritic mouse [<xref ref-type="bibr" rid="B10">10</xref>], rat [<xref ref-type="bibr" rid="B11">11</xref>], dog [<xref ref-type="bibr" rid="B12">12</xref>], and rabbit joints [<xref ref-type="bibr" rid="B13">13</xref>]. Although many genes have been effective in treating murine CIA if they are administered before disease onset, local periarticular gene transfer has not been highly effective in halting the progression of established disease.</p><p>IL-4, similar to TNF-&#x003b1; and IL-1 inhibitors, has been shown be therapeutic for the treatment of murine CIA when administered intravenously as a recombinant protein, either alone or in combination with IL-10. IL-4 can down-regulate the production of proinflammatory and Th1-type cytokines by inducing mRNA degradation and upregulating the expression of inhibitors of proinflammatory cytokines such as IL-1Ra [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. IL-4 is able to inhibit IL-2 and IFN-&#x003b3; production by Th1 cells, resulting in suppression of macrophage activation and the production of the proinflammatory cytokines IL-1, IL-6, IL-8, and TNF-&#x003b1; by monocytes and macrophages [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. In addition, IL-4 inhibits growth factor-induced RA synoviocyte proliferation, and expression of prostaglandin E and matrix metalloproteinase-3 in RA synovial fibroblast [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>], and reduces bone resorption. Given that IL-4 is able to inhibit the production of both IL-1 and TNF-&#x003b1;, block synovial cell proliferation and bone loss, and promote the formation of Th2 lymphocytes, it represents an attractive cytokine for treating arthritis by gene transfer.</p><p>To examine the therapeutic effects of local and systemic IL-4 expression in established CIA, we have generated an adenoviral vector carrying the gene for murine IL-4 (Ad-mIL-4). The ability of Ad-mIL-4 to treat established CIA was evaluated by local periarticular and systemic intravenous injections of Ad-mIL-4 into mice at various times after disease onset. Local injection of Ad-mIL-4 resulted in a reduction in the severity of arthritis and joint swelling, and reduced macroscopic signs of joint inflammation and bone erosion. Injection of the ankle joints in the hind legs also resulted in a therapeutic effect in the untreated, front paws. A high level of endogenous murine IL-10 was detected in the joint tissues of mice injected locally with Ad-mIL-4. Systemic delivery of murine IL-4 by intravenous injection of Ad-mIL-4 also resulted in a slight delay in the onset of disease, with a significant reduction in the number of arthritic paws. These results demonstrate that local and systemic gene transfer of IL-4 is able to treat established and early-stage disease, respectively, in a mouse model of arthritis, and thus may be useful for clinical applications for the treatment of RA.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Mice</title><p>Male DBA/1 lacJ (H-2^{<italic>q</italic>}) mice, aged 7-8 weeks, were purchased from The Jackson Laboratory. They were maintained in a specific pathogen-free animal facility of the University of Pittsburgh Biotechnology Center. For local periarticular injection, mice were anesthetized with inhalation of 2.5% isoflurane gas. A maximum of 5 &#x003bc; l diluted adenovirus was injected into joint articular space under the ankle joint bone in each hind paw with a 26-gauge Hamilton syringe. For systemic delivery of adenovirus, 500&#x003bc; l diluted adenovirus was injected intravenously via the tail vein.</p></sec><sec><title>Vector construction and adenovirus generation</title><p>Ad-mIL-4 was constructed and propagated according to standard protocols, as previously described [<xref ref-type="bibr" rid="B17">17</xref>]. Briefly, recombinant adenoviruses were generated by homologous recombination in CRE8 cells, a 293 cell line that expresses Cre recombinase, between psi-5, an Ad5-derived, E1- and E3-deleted adenovirus, and pAd-lox, the adenoviral shuttle vector that expresses IL-4. The inserted cDNA sequences are expressed under the transcriptional control of human cytomegalovirus promoter.</p><p>Vectors were purified using two consecutive CsCl gradient ultracentrifugation, dialyzed at 4&#x000b0;C against sterile virus buffer, aliquoted, and stored at -80&#x000b0;C until use. The titers of the adenoviral stocks were determined by incubating serial dilutions of the stocks at 37&#x000b0;C for 60 min with sub-confluent 293 cells. Viral titers were calculated by determining the highest dilution that produced 100% viral cytopathic effect, which was then multiplied by the number of cells per well and by the dilution factor. CRE8 and 293 cell lines were grown in Dulbecco's modified eagle medium (GIBCO-BRL, Rockville, MD, USA) supplemented with 10% fetal bovine serum.</p></sec><sec><title>Induction of arthritis</title><p>Bovine typeII collagen (Chondrex, Seattle, WA, USA) was dissolved in 0.05mol/l acetic acid at a concentration of 2mg/ml by stirring overnight at 4&#x000b0;C and emulsified in equal volumes of FCA containing 2mg/ml heat-killed <italic>Mycobacterium tuberculosis</italic> strain H37Ra. The mice were immunized intradermally at the base of tail with 100 &#x003bc; g collagen. On day 21 after priming, the mice were boosted with 100&#x003bc; g type II collagen in incomplete adjuvant. In order to synchronize the onset of arthritis, 40 &#x003bc; g lipopolysaccharide (Sigma) was injected intraperitoneally on day 28.</p></sec><sec><title>Disease evaluation</title><p>Disease severity was assessed every other day using an established macroscopic scoring system ranging from 0 to 4: 0, normal; 1, detectable arthritis with erythma; 2, significant swelling and redness; 3, severe swelling and redness from joint to digit; and 4, maximal swelling with ankylosis. The macroscopic score (mean &#x000b1; standard deviation) was expressed as a cumulative value for all paws, with a maximum possible score of 16. The thickness of each paw was also measured using a spring-load caliper. The paw swelling for each mouse was calculated by adding the four thicknesses of the individual paws. In addition, the number of arthritic paws of individual mice were counted and added to represent the number of arthritic paws in a group. The <italic>in vivo</italic> experiments were performed with 10 mice/group and repeated three times to ensure reproducibility.</p></sec><sec><title>Histologic examination</title><p>Joint tissues from freshly dissected mice were immersion-fixed in 10% neutral buffered formalin and decalcified in 15% ethylene diamine tetra-acetic acid/30% glycerol for 2 weeks. Tissues were then dehydrated in graded alcohols, embedded in paraplast, sectioned at 5&#x003bc; m on a microtome, and stained with hematoxylin and eosin. Sections were evaluated in a blinded manner for histologic signs of arthritis and scored as follows: 1, synovial cell proliferation, synovial hypertrophy with villus formation and/or fibrin deposition; 2, inflammation, synovitis and/or generalized inflammation; 3, cartilage disruption, chondrocyte degeneration and/or ruffling of cartilage surface and/or dystrophic cartilage; and 4, joint destruction, cartilage erosion with abundant inflammation and pannus formation with bone erosion.</p></sec><sec><title>Type II collagen antibody titration</title><p>Serum level of antibody against type II collagen was measured using a standard ELISA assay. Briefly, a 96-well Immuno-Maxisorp Plate (Nunc, Naperville, IL, USA) was coated with bovine type II collagen (10 &#x003bc; g/ml) overnight at 4&#x000b0;C and blocked with 10% fetal bovine serum in phosphate-buffered saline. Sample sera were diluted to 1:100000 (vol:vol) and incubated for 2h at 37&#x000b0;C. After washing, bound antibody isotope was detected with biotin-conjugated antimouse whole IgG (heavy and light chain) antibody (Pharmingen, San Diego, CA, USA). Thereafter, plates were washed, incubated with 100 &#x003bc; l 2,2-azino-di-(-3-ethyl-benzthiazolinesulfonate) substrate (ABTS; Boeringer Mannheim, Indianapolis, IN, USA) at 1mg/ml and read at 405nm.</p></sec><sec><title>Cytokine production</title><p>Cytokine production in the joint tissue or serum was assessed by ELISA (R&#x00026;D systems). For measuring the cytokine production, peeled joint tissues from the upper portion of ankle to the middle of the paw were ground by homogenizer in the equal volume of the lysis buffer (100mmol/l potassuim phosphate, pH 7.8 and 0.2% Triton-X 100). Cytokine production was standardized as amount of cytokine per gram of tissue.</p></sec><sec><title>Statistical analysis</title><p>Results were compared using the Student's <italic>t</italic>-test and by analysis of variance. <italic>P</italic> &#x0003c;0.05 was considered statistically significant.</p></sec></sec><sec><title>Results</title><sec><title>Local delivery of Ad-mIL-4 in established CIA mouse model</title><p>To establish CIA, 8-week-old aged DBA1/lacJ male mice were immunized with 100 &#x003bc; g bovine type II collagen emulsified in complete adjuvant. After 21 days, the mice were boosted with the same amount of collagen in incomplete adjuvant, with disease pathology observed starting 28 days after immunization. In order to synchronize the onset of arthritis, lipopolysaccharide was injected intraperitoneally on day 28, excluding the mice that had already showed signs of disease pathology. All of the collagen-immunized mice had swollen and red paws and ankle joints within 3 days after lipopolysaccharide injection.</p><p>In order to examine the therapeutic effects of IL-4 gene transfer in a murine model of arthritis, increasing doses of an Ad-mIL-4 recombinant virus were administered by periarticular injection into the hind ankle joints of mice with established disease 4 days after lipopolysaccharide injection.</p><p>As shown in Figure <xref ref-type="fig" rid="F1">1</xref>, all mice had established disease at time of injection. However, the severity of arthritis (Fig. <xref ref-type="fig" rid="F1">1a</xref>), paw thickness (Fig. <xref ref-type="fig" rid="F1">1b</xref>), and the number of arthritic paws (Fig. <xref ref-type="fig" rid="F1">1c</xref>) were all significantly reduced in the Ad-mIL-4 group, compared with the saline- and Ad-eGFP-treated groups.</p><p>Analysis of the bones in the ankle joints of control arthritic mice showed evidence of erosion with an associated monocytic infiltrate around the joint space (Fig. <xref ref-type="fig" rid="F2">2a</xref>) compared with the nonarthritic control joints. In contrast, the joints from Ad-mIL-4-treated mice showed less inflammatory joint tissue, reduction in bone erosion, and reduction in the number of inflammatory cells. The changes in histology of the ankle joints from sections from five mice per group were also scored in a blinded manner. As shown in Figure <xref ref-type="fig" rid="F2">2b</xref>, significant inhibition of disease progression as assessed by joint histology was observed in the Ad-mIL-4-treated group.</p><p>These results suggest that local expression of IL-4 after gene transfer to joints with established disease could effectively protect tissue from inflammation as well as block bone erosion. It is important to note that the possible inflammatory responses to adenoviral injection were examined by injection of the same number of particles of Ad-mIL-4 and Ad-eGFP into na&#x000ef;ve joints of DBA mice. At the doses of virus used, however, no inflammation was observed (data not shown).</p></sec><sec><title>Reduction in disease severity in noninjected front paws by local injection of IL-4</title><p>Previously, we have noted a contralateral effect, in which treatment of a diseased joint by adenovirus-mediated transfer of the gene that encodes v-IL-10 resulted in a therapeutic effect in noninjected joints. To determine whether Ad-mIL-4 was able to confer a contralateral effect, the severity of disease in the hind and front paws of the mice injected only in the hind ankle joints was evaluated. As shown in Figure <xref ref-type="fig" rid="F3">3a</xref>, a reduction in the arthritis index was observed in the Ad-mIL-4 injected hind leg ankle joints of the CIA mice. Interestingly, the noninjected front paws also showed a statistically significant reduction in the severity of arthritis (Fig. <xref ref-type="fig" rid="F3">3b</xref>). Taken together, these results suggest that local periarticular injection of Ad-mIL-4 resulted in a therapeutic effect in both injected and noninjected joints. However, unlike v-IL-10, IL-4 is able to confer a contralateral therapeutic effect in mice with established disease.</p></sec><sec><title>Stimulation of endogenous IL-10 expression by local IL-4 gene delivery</title><p>To test the level and duration of gene expression, 5&#x000d7;10⁸ particles of Ad-mIL-4, Ad-mIL-10, and Ad-eGFP were injected periarticularly into the ankle joints of na&#x000ef;ve DBA1 mice. Joint tissues were isolated at indicated time points, homogenized in lysis buffer, and the lysates analyzed for cytokine production by ELISA. Maximal IL-4 expression was detected on day 7, with the level gradually decreasing over time (Fig. <xref ref-type="fig" rid="F4">4a</xref>). However, it is important to note that it is unclear whether the level of IL-4 detected in the joints at the different time points is partially due to induction of endogenous IL-4. Interestingly, a high level of murine IL-10 also was detected from the joint lysates of Ad-mIL-4-injected animals (Fig. <xref ref-type="fig" rid="F4">4b</xref>) that directly correlated with the dose of Ad-mIL-4 (data not shown). An increase in the level of IL-10 expression could even be detected 3 weeks after injection of a high dose of Ad-mIL-4.</p><p>The levels of IL-4 and IL-10 expression also were examined in arthritic mice treated with Ad-mIL-4 at day 35 after injection. As shown in Figure <xref ref-type="fig" rid="F5">5a</xref>, a slight increase in IL-4 expression was observed in that correlated with the dose of Ad-mIL-4 (Fig. <xref ref-type="fig" rid="F5">5a</xref>). A similar increase at day 35 was also observed in na&#x000ef;ve mice (data not shown). In addition, a significant dose-dependent increase in the level of endogenous IL-10 35 days after Ad-mIL-4 administration was observed in the CIA-treated animals (Fig. <xref ref-type="fig" rid="F5">5b</xref>). These results demonstrated that IL-4 transgene expression was detectable for up to 5 weeks after Ad-mIL-4 delivery and that IL-4 was able to induce the production of endogenous IL-10 in the joints of both na&#x000ef;ve and immunized mice. In addition, expression of IL-4 resulted in a twofold increase in the level of endogenous IL-1Ra (data not shown). In contrast, periarticular injection of Ad-mIL-10 did not affect the level of expression of endogenous levels of IL-4 (Fig. <xref ref-type="fig" rid="F4">4a</xref>) and IL-1Ra (data not shown). Taken together, these results suggest that the therapeutic effects of exogenous IL-4 could be mediated in part through the induction of endogenous IL-4 and IL-1Ra.</p></sec><sec><title>Delay of disease onset by systemic Ad-mIL-4 administration</title><p>The administration of recombinant IL-4 protein systemically has been shown to be therapeutic in murine CIA models if given before onset of disease onset. To examine the effect of systemic IL-4 delivered by gene transfer, 10⁹ particles of Ad-mIL-4 were administered intravenously by tail vein injection of the collagen-immunized mice on the day after lipopolysaccharide injection. Whereas the immunized mice injected with Ad-eGFP showed disease onset on day 3 after lipopolysaccharide injection, IL-4-treated mice showed a delay in disease onset (Fig. <xref ref-type="fig" rid="F6">6a</xref>) as well as a reduction in the total number of arthritic paws (Fig. <xref ref-type="fig" rid="F6">6c</xref>). In addition, seven out of the 15 Ad-mIL-4-treated mice were disease-free for up to 10 days after lipopolysaccharide injection, and two mice were disease free at the end of the experiment on day 65. In addition, systemic injection of Ad-mIL-4 suppressed the severity of arthritis in the CIA mice according to arthritis index (Fig. <xref ref-type="fig" rid="F6">6b</xref>). Thus, systemic delivery of IL-4 by adenovirus-mediated gene transfer is able to reduce the onset and severity of early-stage disease.</p></sec><sec><title>IL-4 and IL-10 expression in sera of CIA mice administered Ad-mIL-4 systemically</title><p>To examine the duration of IL-4 expression as well as induction of endogenous IL-10 expression, serum was collected from the mice and tested for levels of IL-4 and IL-10 (Fig. <xref ref-type="fig" rid="F7">7</xref>). An elevated level of IL-4 was detected in on day 7 (Fig. <xref ref-type="fig" rid="F7">7a</xref>) and on day 30 (Fig. <xref ref-type="fig" rid="F7">7c</xref>) after IL-4 injection. In addition, although there was no observed increase in the expression of IL-10 in the sera at day 7 after systemic delivery of IL-4 (Fig. <xref ref-type="fig" rid="F7">7b</xref>), IL-10 was elevated in the sera at day 35 (Fig. <xref ref-type="fig" rid="F7">7d</xref>). In contrast, the levels of IL-4 and IL-10 in the saline or Ad-eGFP-treated control groups were not significantly above background after either or local Ad-mIL-4 delivery. Thus, the early therapeutic effects of IL-4 after systemic delivery most likely are not mediated by IL-10. In contrast, it is possible that the therapeutic effects of IL-4 after local injection are conferred in part by IL-10.</p></sec></sec><sec><title>Discussion</title><p>Gene therapy represents a novel approach for delivery of therapeutic agents to joints in order to treat the pathologies associated with RA and osteoarthritis, as well as other disorders of the joints. Previously, we and others have shown that local <italic>ex vivo</italic> and <italic>in vivo</italic> gene transfer of v-IL-10, IL-1Ra, IFN-&#x003b3;, soluble IL-1 receptor, and soluble TNF receptor are able to block certain intra-articular pathologies in rabbit, rat, and murine models of arthritis. IL-4, like IL-10, is a Th2 cytokine that has been demonstrated to be therapeutic for the treatment of arthritis after systemic administration of recombinant protein.</p><p>In the present study we examined the ability of local periarticular gene transfer of IL-4 to treat established murine CIA. We demonstrated that both local and systemic administration of Ad-mIL-4 resulted in a reduction in the severity of established and early-stage arthritis, respectively, as well as in the number of arthritic paws. In addition, the local gene transfer of Ad-mIL-4 reduced histologic signs of inflammation as well as bone erosion. Previous experiments have shown that gene transfer of IL-10 and IL-1 and TNF inhibitors at the time of disease initiation (day 28) was therapeutic. However, delivery of these agents after disease onset appeared to have only limited therapeutic effect. In contrast, the present results demonstrate that local IL-4 delivery was able partially to reverse progression of established disease after local periarticular injection.</p><p>Interestingly, local and systemic expression of IL-4 resulted in an increase in the level of endogenous IL-10 as well as IL-1Ra. Previous reports have shown that an additive or possible synergistic therapeutic effect can be achieved in animal models of arthritis with combined treatment with recombinant IL-4 and IL-10. Treatment of mice with IL-4 and IL-10 results in reduction in TNF-&#x003b1; and IL-1&#x003b2; production with a concomitant shift in the IL-1Ra:IL-1 ratio [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. The fact that the therapeutic effects of IL-4 and IL-10 are potentially synergistic is possibly due to the fact that they suppress proinflammatory mediator production through different mechanisms. IL-4 is able to block transcription of TNF and IL-1 genes, whereas IL-10 stimulates degradation of TNF and IL-1&#x003b2; mRNAs [<xref ref-type="bibr" rid="B20">20</xref>]. Thus, it is likely that the therapeutic effects we observed after local injection of Ad-mIL-4 are due to both exogenous IL-4 and endogenous IL-10 production. However, the delayed induction of IL-10 after systemic administration of Ad-mIL-4 suggests that the initial therapeutic effects are not conferred by IL-10.</p><p>A mechanism whereby IL-4 may alter IL-10 and IL-1Ra levels could involve the regulation of transcription factors that regulate the respective genes for these cytokines. In particular, nuclear factor-&#x003ba; B and signal transducer and activator of transcription-6 are important for conferring transcriptional regulation by IL-4. Signal transducer and activator of transcription-6, after tyrosine phosphorylation, can bind directly to nuclear factor-&#x003ba; B [<xref ref-type="bibr" rid="B21">21</xref>], resulting in synergic activation in certain cell types.</p><p>The present results suggest that gene transfer of IL-4 can stimulate expression of endogenous cytokines, such as IL-10 and IL-1Ra, as well as possibly endogenous IL-4 expression. Endogenous IL-10, which can be induced by IL-4, is a natural suppressor of a number of inflammatory responses. Inhibition of endogenous IL-10 with neutralizing antibodies enhanced endotoxic shock, IgG immune complex-induced lung injury, and the severity of CIA [<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Periarticular delivery of murine or viral IL-10 by gene transfer resulted in the inhibition of CIA in mouse models [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B26">26</xref>] if delivered before or at the time of disease onset. The present results are also consistent with the observation that IL-4, but not IL-10, has been shown to enhance the production of IL-1Ra by RA synovial cells [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B27">27</xref>].</p><p>IL-4 is a potent mediator in shifting the balance of Th1/Th2 cells and skewing the production of antibody subtypes [<xref ref-type="bibr" rid="B28">28</xref>]. Th2 effector cell differentiation also is dependent on the presence of IL-4 during priming [<xref ref-type="bibr" rid="B29">29</xref>]. Blocking endogenous IL-4 using neutralizing antibodies has been shown [<xref ref-type="bibr" rid="B30">30</xref>] to result not only in the absence of T cells with a Th2-like phenotype, but also in the appearance of T cells producing IL-2 and IFN-&#x003b3; after restimulation. IL-4 is also able to skew the production of antibody subtype [<xref ref-type="bibr" rid="B28">28</xref>]. However, we have shown that anticollagen antibody levels are already very high at day 30 after immunization and that IL-4 treatment did not significantly change these elevated levels (data not shown). In addition, we did not observe significant differences between IgG isotypes in the control and Ad-mIL-4 treated mice (data not shown). Thus, the mechanisms through which local and systemic IL-4 administration are able to suppress CIA are still unclear.</p><p>We have demonstrated previously that periarticular injection of adenovirus-mediated gene transfer of v-IL-10 into the hind paws of mice with early-stage arthritis was able to confer a therapeutic effect in the untreated front paws. Similarly, administration of v-IL-10 into one knee of rabbits with antigen-induced arthritis was able to confer a therapeutic effect in the contralateral untreated knee. This observed contralateral effect was not limited to v-IL-10, in that coadministration of adenoviral vectors expressing soluble IL-1 and soluble TNF receptors also conferred a similar protective effect to untreated knees. Although the mechanism of the contralateral effect is unclear, we have demonstrated that adoptive transfer of dendritic cells from animals treated with adenovirus-mediated gene transfer of v-IL-10 to untreated immunized animals is able to confer a therapeutic effect (unpublished data). Thus, it is possible that local expression of v-IL-10 or IL-1 and TNF inhibitors modulates the activity of dendritic cells. However, although v-IL-10 is able to block early-stage disease, it is ineffective in reversing established disease. In contrast, periarticular injection of Ad-mIL-4 was able to reverse pathology in established disease not only in the treated hind paws, but also in the untreated front paws. Interestingly, we have observed that intravenous injection of na&#x000ef;ve dendritic cells, genetically modified to express IL-4, is able to effectively treat established arthritis by inhibiting the Th1 response (unpublished data).</p><p>During the preparation of this manuscript, a similar study using a recombinant Ad-mIL-4 vector for treatment of murine CIA was reported [<xref ref-type="bibr" rid="B31">31</xref>]. However, the vector was administered before onset of disease into knee joints, whereas the vector in the present study was injected after disease onset into ankle joints. Similar to the present results, in that study local injection of the recombinant Ad-mIL-4 vector resulted in a reduction in bone erosion and a reduction in cartilage degradation. However, unlike the present results, no effect on joint inflammation was observed. Moreover, in the present report we demonstrated that both local and systemic injection resulted in induction of endogenous IL-10. Given the ability of IL-4 to reverse established disease partially, our studies, as well as those of others, support the potential application of IL-4 gene therapy for the clinical treatment of RA.</p></sec>

IgVH genes from different anatomical regions, with different histopathological patterns, of a rheumatoid arthritis patient suggest cyclic re-entry of mature synovial B-cells in the hypermutation process

<sec><title>Introduction:</title><p>Although IgV genes in rheumatoid B cells have been intensively analyzed, many questions concerning antigen driven B-cell maturation and recirculation remain unanswered. It would be interesting to know whether B-cell maturation in rheumatoid tissue is different from that in secondary lymphatic organs. Moreover, it would be interesting to know whether there exists a restricted number of antigens that act on the lesions of different anatomical sites of the RA patient, and whether B cells recirculate between the different joints.</p></sec><sec><title>Methods:</title><p>RNA and genomic DNA were prepared from tissue sections from three different anatomical sites, with different histopathologies and different onsets (left and right peroneal tendons and cubita synovial membrane), of a RA patient. Genomic DNA was amplified by seminested polymerase chain reaction (PCR), and the cDNA corresponding to the RNA was amplified by PCR using primers specific for each IgVH family. The obtained sequences were compared with their germline counterparts on the V-Base data Bank [<xref ref-type="bibr" rid="B1">1</xref>]. An immunohisto-chemical analysis of the infiltrate and the clinical data of local disease activity were also included.</p></sec><sec><title>Results:</title><p>In the locations with longer disease duration (right peroneal tendon 5 months, left peroneal tendon 2 months) a very intense inflammatory infiltrate with germinal centers containing Ki-M4-positive follicular dendritic cells (FDC) was observed. In the location with shorter disease duration (right cubita 2 weeks) a low, diffuse and nonfollicular infiltration with marked oedema was detected. From the 55 analyzed clones seven expressed nonfunctional rearrangements (pseudogenes) with stop codons, and 48 were found to express functional genes. Among the 48 clones that expressed functional genes, there were two that had amino acid deletions on their complementarity determining region (CDR)2 - clones <italic>K194/1</italic> and <italic>K194/111</italic> - similar to the ones described by Wilson <italic>et al</italic> [<xref ref-type="bibr" rid="B2">2</xref>] and Goossens <italic>et al</italic> [<xref ref-type="bibr" rid="B3">3</xref>]. Two types of mixed molecules were found. Mixed molecules of the first type (<italic>k194/57</italic>, <italic>k194/67</italic> and <italic>k194/109</italic>) are composed of rearrangements of two different IgV genes. Mixed molecules of the second type (<italic>k194/126</italic>, <italic>k194/119</italic>, <italic>k194/30</italic> and <italic>k194/99</italic>) are composed of a IgV gene rearrangement that is fragmented by insertions of small random sequences. These insertions are different from the ones described by Wilson <italic>et al</italic> [<xref ref-type="bibr" rid="B2">2</xref>] since they are not duplicates or parts of IgV genes. The ratio of replacement mutations to silent mutations (R/S ratio) increased with disease duration. There was strong heterogeneity among the CDR3 segments.</p><p>The amino acid sequences that belonged to the VH1-family obtained from the three anatomical regions were primarily compared with the amino acid sequences of their closest germline counterparts (Fig. <xref ref-type="fig" rid="F1">1a</xref>). One result from this comparison was the heterogeneity in the CDR3 rearrangements. Moreover, sequences <italic>k194/58</italic> and <italic>k194/82</italic> are clonally related (confirmed at nucleotide level). Then, the 21 amino acid sequences were compared with the most widely used germline counterpart <italic>IgHV1-18^*01</italic> (Fig. <xref ref-type="fig" rid="F1">1b</xref>). All of these VH1 sequences had mainly conservative mutations in the framework region (FR) and nonconservative mutations in the CDR. Also, there was an almost overall conservation of the mutational cold spots and 'structural cold spots' [<xref ref-type="bibr" rid="B4">4</xref>] among the 19 VH1 segments. The replacement (11 from 19 replacements resulted in a proline residue) in position 34 of CDR2 could be interpreted as an antigen-selected mutational hotspot.</p><p>The comparison of the five sequences belonging to <italic>IgHV4-30-1/4-31^*02</italic> resulted in two types of clonal relation (Fig. <xref ref-type="fig" rid="F2">2a</xref>). The first type of clonal relation, between sequences <italic>k194/100</italic> and <italic>k194/101</italic> (Fig. <xref ref-type="fig" rid="F2">2b</xref>), suggests that both sequences are derived from a single progenitor cell. The second type of clonal relation is between sequences <italic>k194/23</italic>, <italic>k194/102</italic> and <italic>k194/103</italic> (Fig. <xref ref-type="fig" rid="F2">2c</xref>). It suggests that an unmutated progenitor cell gave rise to <italic>k194/23</italic> (left peroneal tendon), from which <italic>k194/103</italic> (right cubita) derived and later generated <italic>k194/102</italic> (right cubita).</p></sec><sec><title>Discussion:</title><p>The analysis of the 55IgVH sequences corroborates the findings of other groups that studied a singlelocation and RA B-cell hybridomas [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>] and adds further information on B-cell distribution and activation in RA. First, amino acid deletions and mixed molecules could be interpreted as novel pathways to generate antibody specificities, leading, for instance, to autoreactive antibodies that could contribute to the local and systemic tissue destruction. Second, an apparently conserved mutational pattern among the 19 amino acid VH1 segments suggests that in all three RA lesions of this patient the synovial B cells are dealing with a restricted number of antigens. Third, the existence of clonally related B cells in the cubita and left peroneal tendon leaves no doubt that in this patient there is a cyclic re-entry of mutated B cells in the hypermutation process [<xref ref-type="bibr" rid="B11">11</xref>]. The already mutated B cells from the early RA lesions sequentially colonize new germinal centers in secondary lymphatic organs as proposed by Kepler <italic>et al</italic> [<xref ref-type="bibr" rid="B12">12</xref>]. These reactivated B-cells then invade new anatomical regions, leading to the perpetuation of the chronic inflammation in RA.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Souto-Carneiro</surname><given-names>Maria M</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>path119@mail.uni-wuerzburg.de</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Krenn</surname><given-names>Veit</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Hermann</surname><given-names>Ralph</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>K&#x000f6;nig</surname><given-names>Achim</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>M&#x000fc;ller-Hermelink</surname><given-names>Hans-Konrad</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Molecular analysis of synovial tissue and B-cell hybrido-mas [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B13">13</xref>] has demonstrated that synovial B cells, which are a characteristic feature of RA [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>], are expanded in an antigen-dependent manner [<xref ref-type="bibr" rid="B16">16</xref>]. Because germinal centers may be detected primarily in synovial tissue of severely affected joints of RA patients [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B17">17</xref>], this is very suggestive that antigens that drive local B-cell expansion are directly involved in the pathogenesis of RA.</p><p>Although IgV genes in rheumatoid B cells have been intensively analyzed, many questions concerning the antigen driven B-cell maturation and recirculation remain unanswered. It would be interesting to know whether B-cell maturation in rheumatoid tissue is different from that in secondary lymphatic organs. Moreover, it would be interesting to know whether there exists a restricted number of antigens that act on the lesions of different anatomical sites of the RA patient, and whether B cells recirculate between the different joints.</p><p>Therefore, in the present study IgVH genes from synovial tissue B cells of different anatomical regions (with different times of disease onset) from a RA patient were analyzed. Furthermore, we included a histopathological analysis and clinical data of local disease activity, which can give a more complete picture of the role of B cells in the pathogenesis of RA.</p></sec><sec><title>Patient and methods</title><sec><title>Patient, disease activity and tissue samples</title><p>Tissue samples (one from the right and another from the left tendon of musculus peroneus longus and one from the cubita synovial tissue) from a 48-year-old female patient with confirmed seropositive RA [<xref ref-type="bibr" rid="B18">18</xref>] were obtained at synovectomy and were snap frozen. The patient was receiving antirheumatic medication (gold, methotrexate and sulphasalazine). In the present investigation, the degree of local disease activity was scored according to the method of Fuchs <italic>et al</italic> [<xref ref-type="bibr" rid="B19">19</xref>] and Krenn <italic>et al</italic> [<xref ref-type="bibr" rid="B20">20</xref>], on the basis of warmth, effusion and swelling. The patient was seropositive for rheumatoid factors. All tissues (approximately 50%) were fixed in formalin and embedded in paraffin (Giemsa, haematoxylin and eosin staining) to be used for diagnosis and scoring of the degree of inflammation. The only material used for immunohistochemical analyses was tissue that exhibited macroscopic signs of inflammation, taken from at least three different regions of the resected synovial membrane.</p></sec><sec><title>Immunohistochemistry</title><p>For immunohistochemical staining, 7-&#x003bc; m cryosections (mounted on poly-L-lysine-coated slides) were used. Immediately before staining, the cryosections were treated with acetone for 10min, air dried at room temperature (10-20min) and double immunohistochemical staining was performed, as described by Krenn <italic>et al</italic> [<xref ref-type="bibr" rid="B7">7</xref>]. Briefly, the indirect immunoperoxidase technique (Ki-M4; DAKO, Hamburg, Germany) was combined with the alkaline phosphatase/antialkaline phosphatase technique (CD20; DAKO). No counterstaining was performed. In all cases, control staining was performed and single stainings were compared with double stainings in order to ensure that the pattern of immunohistochemical reaction remained unaltered.</p></sec><sec><title>Histopathological score of inflammatory infiltration</title><p>A portion of tissue (approximately 50%) was fixed in formalin and paraffin-embedded (Giemsa, haematoxylin and eosin staining) for use in diagnosis and scoring of the degree of the inflammatory infiltration, which in the present study was performed according to the method of Krenn <italic>et al</italic> [<xref ref-type="bibr" rid="B7">7</xref>] on a semiquantitative 1-5 scale. Very low inflammatory infiltration was indicated by 1 on the scale: the synovial intima is slightly enlarged (two to three cell layers thick); the degree of lymphocytic infiltration is very low, showing a diffuse pattern; and the subsynovial region exhibits chronic tissue granulation with slight fibrosis. Low inflammatory infiltration was indicated by 2 on the scale: the synovial intima is slightly enlarged (two to three cell layers thick), and the degree of inflammatory infiltration is low, with a diffuse perivascular lymphocytic and plasma cell infiltration; and the subsynovial region shows chronic tissue granulation with moderate fibrosis. Moderate inflammatory infiltration was indicated by 3 on the scale: the synovial intima is moderately enlarged (three to five cell layers thick), and the degree of lymphocytic infiltration is moderate, with small follicle-like aggregates near small blood vessels; and there is moderate cellularity of the subsynovial region, which exhibits slight fibrosis. Strong inflammatory infiltration was indicated by 4 on the scale: the synovial intima is extensively enlarged (five to 10 cell layers), and lymphocytes exhibit a dense follicle-like pattern; and the 'interfollicular' area exhibits very high cellularity without fibrosis. Very strong inflammatory infiltration was indicated by 5 on the scale: the synovial intima is extensively enlarged, and the distribution of lymphocytes exhibits a dense follicle-like pattern with formation of germinal centers; and granulomas and hemigranulomas can be seen in the subsynovial region. In each histopathological analysis, 10 fields were examined, and the most prominent finding in a given field determined the score.</p></sec><sec><title>cDNA synthesis and polymerase chain reaction amplification</title><p>Total RNA from about 50 tissue sections of 20 &#x003bc; m was prepared using the method of Chomczynski and Sacchi [<xref ref-type="bibr" rid="B21">21</xref>]. cDNA synthesis was performed with 5 &#x003bc; g RNA using Gibco BRL (Karlsruhe, Germany) M-MLV reverse transcriptase according to the supplier's manual. The amplification of the VH genes was carried out in a 25 &#x003bc; l volume containing 1.75mmol/l MgCl₂, 0.4pmol/l primer,1U Taq polymerase (MBI Fermentas, St Leon-Rot, Germany) and 200 &#x003bc; mol/l of each dNTP. The cycle profile for amplification consisted of DNA denaturation at 95&#x000b0;C for 2min, followed by 45 cycles of 94&#x000b0;C for 30s, primer annealing at 65&#x000b0;C for 30s for VH3 and VH4 primers (60&#x000b0;C for VH1, VH2 and VH5), and extension at 72&#x000b0;C for 80s. Primer sequences were described previously [<xref ref-type="bibr" rid="B22">22</xref>], and are located from codons 17 to 24 (according to V-Base [<xref ref-type="bibr" rid="B1">1</xref>] sequence comparison). In brief, the following primers, given in the 5' -3' direction, were used:</p><p>V_H1 5' CCT CAG TGA AGT YTC CTG CAA GGC 3' </p><p>V_H2 5' GTC CTG CGC TGG TGA AAS CCA CAC A 3' </p><p>V_H3 5' GGG GTC CCT GAG ACT CTC CTG TGC AG 3' </p><p>V_H4 5' GAC CCT GTC CCT CAC CTG CRC TGT C 3' </p><p>V_H5 5' AAA AAG CCC GGG GAG TCT CTG ARG A 3' </p><p>V_H6 5' ACC TGT GCC ATC TCC GGG GAC AGT G 3' </p><p>J_H1-5 5' GGT GAC CAG GGT BCC YTG GCC CCA G 3' </p><p>J_H6 5' GGT GAC CGT GGT CCC TTG CCC CCA G 3' </p></sec><sec><title>DNA extraction and amplification of IgVH genes by nested polymerase chain reaction</title><p>DNA extraction and amplification of IgVH genes was performed according to the method of Kim <italic>et al</italic> [<xref ref-type="bibr" rid="B16">16</xref>] with minor modifications. In short, DNA was prepared by incubating 10&#x000d7;5&#x003bc; m tissue sections at 50&#x000b0;C for 1h with proteinase K (Boehringer Mannheim, Mannheim, Germany), which was inactivated by heating at 95&#x000b0;C. To improve the specificity of the PCR amplification, seminested PCR reactions were carried out as follows. In the first step, amplification with Taq polymerase was performed with VH 5' primers and external J_H region-specific 3' primers [<xref ref-type="bibr" rid="B16">16</xref>]. In the second round,aliquots were specifically amplified for the heavy-chain genes using the same 5' V_H region primers, but internal J_H region primers (seminested PCR). In brief, the following primers, given in the 5' -3' direction, were used:</p><p>V_H1 5' CCA TGG ACT GGA CCT GGA 3' </p><p>V_H2 5' ATG GAC ATA CTT TGT TCC AC 3' </p><p>V_H3 5' CCA TGG AGT TTG GGC TGA GC 3' </p><p>V_H4 5' ATG AAA CAC CTG TGG TTC TT 3' </p><p>V_H5 5' ATG GGG TCA ACC GCC ATC CT 3' </p><p>V_H6 5' ATG TCT GTC TCC TTC CTG AT 3' </p><p>J_Hexternal 5' CTC ACC TGA GGA GAC GGT GAC C 3' </p><p>J_Hinternal</p><p>5' TGA (AG)GA GAC GGT GAC C(AG)(GT) GT(GCT) CC 3' </p><p>The final concentrations of the reagents were 0.1mmol/l MgCl₂, 200 &#x003bc; mol/l of each dNTP, 10pmol/l of each primer and 2U Taq DNA polymerase. The cycle program consisted of a denaturation step at 95&#x000b0;C for 5min followed by five cycles at 95&#x000b0;C for 40s, 65&#x000b0;C for 40s and 72&#x000b0;C for 1min and 50s; five cycles at 95&#x000b0;C for 40s, 60&#x000b0;C for 40s and 72&#x000b0;C for 1min and 50s; and 25 cycles at 95&#x000b0;C for 40s, 55&#x000b0;C for 40s and 72&#x000b0;C for 1min and 50s. The cycles were followed by a final 10-min incubation at 72&#x000b0;C.</p></sec><sec><title>Sequence analysis</title><p>Aliquots of the final PCR products were separated by electrophoresis using a 2% low melting agarose gel (Roth, Karlsruhe, Germany), and DNA bands in the range of 350bp were purified from the agarose gel using High-Pure DNA gel extraction kit (Boehringer Mannheim). Cloning of PCR fragments was performed using the pCR-Script Amp SK(+) cloning kit (Stratagene, Heidelberg, Germany). Positive clones were sequenced using the DyeDeoxy Termination Cycle Sequencing Kit (Applied BioSystems Inc, Weiterstadt, Germany), and analyzed using an automated DNA sequencer ABIPrism373 (Applied BioSystems Inc). Both strands were sequenced using T3 and T7 primers. The sequences were analyzed using DNAman for Windows software (Lynon BioSoft, Vaudreuil, Quebec, Canada), Genebank and v-Base databases [<xref ref-type="bibr" rid="B1">1</xref>].</p></sec></sec><sec><title>Results</title><sec><title>Local disease activity and duration of local disease</title><p>The female patient, who suffered from a confirmed seropositive RA with involvement of tendon sheaths, exhibited severe signs of local disease activity, with tenosynovitis of the right distal peroneus longus tendon and left distal peroneus longus tendon, and synovialis of the right cubita (Table <xref ref-type="table" rid="T1">1</xref>). The durations of local disease were 5 months, 2 months and 2 weeks, respectively.</p></sec><sec><title>Histopathology and immunohistochemistry of synovial tissue</title><p>A heterogeneous inflammatory infiltrate could be observed in the different locations. In both locations with longer disease duration (right peroneal tendon [5 months] and left peroneal tendon [2 months]) a very intense inflammatory infiltrate, with Ki-M4-positive FDC-containing germinal centers (Fig. <xref ref-type="fig" rid="F3">3a</xref>,<xref ref-type="fig" rid="F3">3b</xref> and insert), could be observed (inflammatory score 5). However, in the right cubita (disease duration 2 weeks) a low, diffuse and nonfollicular infiltration with marked oedema (inflammatory score 2) was detected (Fig. <xref ref-type="fig" rid="F3">3c</xref>). The latter synovialitis histopathologically showed a more acute inflammatory reaction, whereas in the right and left peroneal tendons the morphological pattern of a typical chronic tendosynovitis was present. Immunohistochemically, the right and left peroneal tendons exhibited a dense follicular-like infiltration, with Ki-M4-positive FDC and peripherally located CD20-positive B lympocytes representing germinal centers (insert in Fig. <xref ref-type="fig" rid="F3">3a</xref> and <xref ref-type="fig" rid="F3">3b</xref>). In the right cubita only a very low and diffuse nonfollicular distribution of lymphocytes without Ki-M4-positive FDC could be recognized.</p></sec><sec><title>Comparison of the mutated VH segments with the germline genes</title><sec><title><italic>Presence of pseudogenes</italic></title><p>From the 55 analyzed clones (Table <xref ref-type="table" rid="T2">2</xref>), seven expressed nonfunctional rearrangements (pseudogenes) with stop codons, and 48 were found to express functional genes. The existence of pseudogenes has been largely described in IgV genes amplified from genomic DNA of healthy [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B23">23</xref>] and diseased [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B24">24</xref>] individuals. Furthermore, the presence of pseudogenes in diseased individuals is largely related to specific mutations on the RYGW motifs [<xref ref-type="bibr" rid="B4">4</xref>]. Based on these findings, it is not surprising that we found pseudogenes in the genomic DNA amplificates (<italic>k194/81</italic>, <italic>k194/120</italic> and <italic>k194/126</italic>). However, we also identified pseudogenes in the cDNA amplificates (<italic>k194/30</italic>, <italic>k194/33</italic>, <italic>k194/130</italic> and <italic>k194/135</italic>), which have not yet been described in the literature. These pseudogenes could be the product of a PCR artifact (maybe due to an elevated number of cycles) that introduced STOP-codons in the IgV-gene sequence. However, because the sequences were read in both directions (5' -3' and 3' -5'), and both readings yielded the same confirmatory results, we do not consider the pseudogenes in the cDNA amplificates to be PCR artifacts. The existence of such pseudogenes could be explained by the findings of Drapkin <italic>et al</italic> [<xref ref-type="bibr" rid="B25">25</xref>] that DNA repair enzymes are part of the RNA polymerase II transcription initiation process. Hence there could be a defective DNA repair mechanism that, in the case presented here, could lead to the introduction of STOP codons in the RNA molecule.</p></sec><sec><title><italic>Deletions and mixed molecules</italic></title><p>Among the 48 clones that express functional genes, there were two presenting amino acid deletions on their CDR2: clones <italic>K194/1</italic> and <italic>K194/111</italic>. These detected deletion events should be regarded as the result of somatic hypermutation and not as a PCR artifact, because they were found in the intrinsic somatic hypermutation hotspots [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>], and also involved triplets from CDR2, which leaves the transcripts functionally in frame without profoundly altering the backbone structure of the molecule, as defined by Wilson <italic>et al</italic> [<xref ref-type="bibr" rid="B2">2</xref>].</p><p>Two types of mixed molecules were found. Mixed molecules of the first type (<italic>k194/57</italic>, <italic>k194/67</italic> and <italic>k194/109</italic>) are composed of rearrangements of two different IgV genes. These two mixed molecules could be considered PCR artifacts, like the ones described by Bridges <italic>et al</italic> [<xref ref-type="bibr" rid="B30">30</xref>] for amplified V&#x003ba; gene segments in RA synovium. This could be due to the fact that RNA is very unstable and could have fragmented while the samples had not been snap frozen (for reasons of hygiene it is not allowed to take snap-freezing apparatus into the operating theatre). However, two facts counter the PCR artifact hypothesis. First, the B cells of the RA synovial samples are used in our laboratory not only for IgV analysis, but also for hybridoma production, so their RNA must be intact to allow successful cell fusion [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. Second, the sequences were read in both directions (5' -3' and 3' -5'), and both readings yielded the same confirmatory results of functional mixed molecules.</p><p>Mixed molecules of the second type (<italic>k194/126</italic>, <italic>k194/119</italic>, <italic>k194/30</italic> and <italic>k194/99</italic>) are composed of a IgV gene rearrangement fragmented by insertions of small random sequences. These insertions are different from the ones described by Wilson <italic>et al</italic> [<xref ref-type="bibr" rid="B2">2</xref>] since they are not duplicates or parts of IgV genes. On one hand, this could happen because of the insertion of incorrectly amplified fragments, thus resulting in a PCR hybrid artifact. On the other hand, however, the sequences were read in both directions (5' -3' and 3' -5'), and both readings yielded the same confirmatory results. Also the use of nested PCR to amplify the genomic DNA strongly reduces the possibility of amplification and insertion of incorrect fragments [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B31">31</xref>], thus rendering improbable the hypothesis of PCR artifact.</p><p>Although not considering the deletions and the mixed molecules as PCR artifacts, we made the decision not to consider them for further mutational analyses, because a rational comparison with these sequences is not possible.</p></sec><sec><title><italic>Local overall R/S ratios increase with disease duration</italic></title><p>The 41 in-frame functional clones accumulated between 4 and 46 replacements on their amino acid sequence. The R/S ratios in the CDR of all clones from each anatomical region were all higher than 3. There was a direct correlation between the R/S values and the time of local disease duration. Locations with longer disease activity (right and left peroneus longus tendons) also had higher R/S values in the CDR than the location with a later onset (right cubita; Fig. <xref ref-type="fig" rid="F4">4</xref>).</p></sec><sec><title><italic>Heterogenity among the CDR3s</italic></title><p>Even though a relatively reduced number of different VH germline gene segments was used, the CDR3s were encoded by D-gene segments that differed in both amino acid sequence and length, and all of the six known human JH gene segments were found. As expected for the normal adult Ig repertoire [<xref ref-type="bibr" rid="B32">32</xref>], the JH4 and JH6 segments were the most commonly used.</p></sec></sec><sec><title>Comparison of the sequences from the same VH family amplified from each location</title><p>The comparison was restricted to the VH1 and VH4 families, because they yielded the more relevant results.</p><sec><title><italic>VH1 family</italic></title><p>The mutational patterns of immunoglobulin VH1 genes was studied by Borretzen <italic>et al</italic> [<xref ref-type="bibr" rid="B33">33</xref>] in peripheral blood monoclonal IgM rheumatoid factors of healthy individuals and RA patients. However, this kind of mutational pattern comparison has not been extended to B cells from RA synovial tissue. As is widely known, the primary structure of an antibody, formed by the amino acid sequence, determines all of its chemical and biological properties. Thus, the amino acid sequences that belong to the VH1 family obtained from the three anatomical regions were compared with the amino acid sequences of their closest germline counterparts (Fig. <xref ref-type="fig" rid="F1">1a</xref>). One result from this comparison is the heterogenity in the CDR3 rearrangements. Moreover, sequences k194/58 and k194/82 are clonally related (confirmed at nucleotide level, data not shown).</p><p>Then, the 21 amino acid sequences were compared with the most widely used germline counterpart <italic>IgHV1-18^*01</italic> to determine whether a common motif could be discerned (Fig. <xref ref-type="fig" rid="F1">1b</xref>). As postulated, the conservation of the amino acid sequence of all three FRs is crucial for the interaction with the antigen [<xref ref-type="bibr" rid="B34">34</xref>]. In fact, for FR 1+2 we observed highly conserved regions (Fig. <xref ref-type="fig" rid="F5">5</xref>).</p><p>The residues at positions number 6, 10 and 11 from FR1 and the complete FR2 (except positions 22 and 26) showed a high conservation of the amino acid residues. In FR3, we found high sequence diversity (Figs <xref ref-type="fig" rid="F1">1b</xref> and <xref ref-type="fig" rid="F5">5</xref>), even though there was still conservation within residues 48-49, 51-52, 54, 56, 58-60, 62-64, 66-67, 69 and 71-75. All the other positions had a total of 30 nonconservative substitutions, and therefore they probably do not play a determining role in the antigen-mediated activation process. As expected, the number of nonconservative substitutions in both CDR1 and CDR2 was highly elevated. Nevertheless, the 3' end of CDR2, covering residues 41-47, contained a total of 25 substitutions, but only five were nonconservative (Fig. <xref ref-type="fig" rid="F5">5</xref>). More striking evidence was that residues in position 34 of CDR2 in 11 out of 19 replacements resulted in a proline residue. It could even be speculated that position 34 of CDR2 may be an antigen-selected mutational hotspot, because it does not belong to the defined somatic hypermutation hotspots [<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>]. In position 45 there are 13 substitutions from Leu to Phe, but 12 of them cannot be considered real substitutions because, as shown in Figure <xref ref-type="fig" rid="F1">1a</xref>, the germline <italic>IgHV1-18^*01</italic> is the only one to have a Leu in that position instead of the more frequent Phe. Therefore, in this position we only considered Leu-Ile and one Leu-Phe (for <italic>k194/62</italic>) as real substitutions.</p><p>As stated by Chang and Casali [<xref ref-type="bibr" rid="B35">35</xref>], the CDR1 is the IgVH gene region with higher susceptibility to amino acid replacement, and this was in fact the case for all the obtained sequences. The CDR1 had the highest number of nonconservative replacement mutations, which makes it very unlikely to be primarily involved in the antigen-mediated activation.</p></sec><sec><title><italic>VH4 family</italic></title><p>From the seven amplified sequences of the VH4 family (four from the cubita and three from the left peroneal tendon) five had <italic>IgHV4-30-1/4-31^*02</italic> as their closest germline counterpart, and the other two had <italic>IgHV4-59^*01</italic> and <italic>IgHV4-30-4^*06</italic> (Table <xref ref-type="table" rid="T2">2</xref>). When comparing the five sequences that belong to <italic>IgHV4-30-1/4-31^*02</italic> (Fig. <xref ref-type="fig" rid="F2">2a</xref>) with each other, there was evidence of two different clonal relations. The first clonal relation was between sequences <italic>k194/100</italic> and <italic>k194/101</italic> (Fig. <xref ref-type="fig" rid="F2">2b</xref>), suggesting that both sequences are derived from a single progenitor cell with the rearrangement <italic>IgHV4-30-1/4-31^*02-IgHD4-17^*01-IgHJ5^*02</italic>. The second clonal relation was between sequences <italic>k194/23</italic>, <italic>k194/102</italic> and <italic>k194/103</italic> (Fig. <xref ref-type="fig" rid="F2">2c</xref>), suggesting that sequence <italic>k194/102</italic> was derived from <italic>k194/103</italic>, which in turn was derived from sequence <italic>k194/23</italic> that had a progenitor cell with the rearrangement <italic>IgHV4-30-1/4-31^*02-IgHD2-2^*02inv-IgHJ4^*01</italic>. Furthermore, the small number of mutations of all of these five sequences when compared with the germline could be taken as indirect evidence that the germline already encodes a high-affinity antibody, as suggested by Williams and Taylor [<xref ref-type="bibr" rid="B8">8</xref>].</p></sec></sec></sec><sec><title>Discussion</title><p>Various studies have demonstrated that somatically mutated B-cells are present in RA synovial tissue [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>] and in human RA hybridomas [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B36">36</xref>]. However, we studied the IgV genes of synovial B cells taken from different anatomical regions, with distinct histopathology and local disease duration, of the same RA patient. The analysis of the 55IgVH sequences corroborates the findings of other groups that studied a single location, and adds further information on B-cell distribution and activation in RA.</p><sec><title>Amino acid deletions and mixed molecules: novel pathways to generate antibody specificities?</title><p>Recently, the introduction of deletions and duplications, in addition to nucleotide exchanges, has been described as a feature of the somatic hypermutation process [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Amino acid deletions in the IgV genes have only been described in lymphomas and healthy secondary lymphatic tissue, however [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. In the present study we report the existence of such amino acid deletions in the IgV genes from synovial B cells of an autoimmune disease. For the first time, amino acid deletions in the VH genes were found in B cells of an autoimmune disease. The detection of these deletion events in RA synovialitis stresses the functional homology of the synovial membrane to secondary lymphatic tissue.</p><p>Also for the first time, IgV gene mixed molecules were found that are formed either by two segments of different IgV genes or by a IgV gene rearrangement that is fragmented by random insertions. The mechanisms that underlie the formation of these mixed molecules could be modified pathways to the unified model for somatic hypermutation, as proposed by Maizels [<xref ref-type="bibr" rid="B37">37</xref>] (Fig. <xref ref-type="fig" rid="F6">6</xref>); an initiating lesion could lead to a hypermutation either templated (using another germline gene as template) or untemplated (by inserting small random sequences). In some cases the process of untemplated hypermutation could lead to the insertion of STOP codons, rendering the gene nonfunctional.</p><p>Hence, the production of mixed molecules and the introduction of deletions could represent novel pathways for RA synovial B cells to generate new specificities that lead, for instance, to autoreactive antibodies that could contribute to the local and systemic tissue destruction.</p></sec><sec><title>Apparent mutational pattern among the 19 amino acid VH1 segments</title><p>The comparison of the amino acid sequences of the 19 VH1 segments from the different locations provided some valuable data on the interaction of the RA synovial B cells and their target antigen(s).</p><p>All of the VH1 sequences had mainly conservative mutations in the FR and nonconservative in the CDR, thus, agreeing with the results from Wedemayer <italic>et al</italic> [<xref ref-type="bibr" rid="B38">38</xref>]. When resolving at 2.1&#x000c5;, the crystal structure of a germline antibody Fab fragment and its complex with hapten, they observed an expansion of the binding potential of the primary antibody repertoire. This expansion derived from configurational stability due to antigen binding and somatic mutations, nonconservative mutations in the CDR that raised the affinity for the hapten, and conservative mutations in the FR.</p><p>Another important finding was the almost overall conservation of the mutational cold spots and 'structural cold spots' [<xref ref-type="bibr" rid="B4">4</xref>] among the 19VH1 segments. The interesting absence of mutations in positions 17, 23, 28, 30, 48, 56 and 71 suggests the existence of more 'structural cold spots' in the VH1 family than those described.</p><p>During the germinal center reaction (for instance in the follicles of the two earlier lesions of this patient), rearranged B cells with low-affinity receptors improve their affinity by somatic hypermutation [<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. Nevertheless, these mutations can also decrease the affinity instead of expanding it. Therefore, as reported by Meffre <italic>et al</italic> [<xref ref-type="bibr" rid="B41">41</xref>], under appropriate regulation VDJ rearrangements take place in mature B cells of human tonsil. Hence, the heterogenity observed on the CDR3 of the 19 VH1 sequences could be due to a reactivation of the rearrangement process in order to rescue these RA synovial B cells from deleterious somatic mutations, or to further increase their binding affinity.</p><p>Based on the above findings, there appears to be a conserved mutational pattern among all 19 VH1 segments, hence suggesting that in all three RA lesions of this patient the synovial B cells were activated by a restricted number of antigens. This is strengthened by the replacement in position 34 of CDR2, which could be interpreted as an antigen-selected mutational hotspot.</p></sec><sec><title>Cyclic re-entry of mutated rheumatoid arthritis synovial B-cells in the hypermutation process</title><p>The increment of mutations with antigen dose [<xref ref-type="bibr" rid="B42">42</xref>] possibly indicates that the maturation of the immune response is a continuous process with the production of an increasing number of hypermutated memory B cells with time. In the special case of RA, the local joint destruction may release antigens that lead to the hypermutation process. Characteristic for B-cell hypermutation are the elevated R/S ratios in the CDR. In the present study there was a direct association of the overall R/S ratios with the duration of local disease. Synovial B cells were shown to undergo a germinal center-like reaction in RA [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. Therefore, we suggest that the activated B cells from this patient have undergone a local maturation in the germinal center-like structures detected in the peroneal tendons (left and right). On the other hand, the fact that one lesion was free from FDCs and exhibited only an acute inflammatory infiltrate could support the model proposed by Oprea and Perelson [<xref ref-type="bibr" rid="B43">43</xref>]; the already mutated germinal center B cells from the peroneal tendons might have migrated into the cubita synovial tissue, as has been shown for closely located finger joints [<xref ref-type="bibr" rid="B44">44</xref>], and re-entered in a cyclic hypermutation process. The apparent existence of a mutational pattern on amino acid level of clones obtained from the different regions could support this hypothesis. However, the existence of clonally related B-cells in the cubita and left peroneal tendon leaves no doubts that, in the patient studied, there is cyclic re-entry of the mutated B cells from the early RA lesions in the hypermutation process [<xref ref-type="bibr" rid="B11">11</xref>] that sequentially colonize new germinal centers, as proposed by Kepler and Perelson [<xref ref-type="bibr" rid="B12">12</xref>]. These reactivated B cells then invade new anatomical regions, leading to the perpetuation of the chronic inflammation in RA.</p></sec></sec>

T cells that are naturally tolerant to cartilage-derived type II collagen are involved in the development of collagen-induced arthritis

<sec><title>Introduction:</title><p>A discussion is ongoing regarding the possible role of cartilage-directed autoimmunity as a part of the pathogenesis of rheumatoid arthritis (RA). One possibility is that the association of RA with shared epitope-expressing DR molecules reflects a role for major histocompatibility complex (MHC) class II molecules as peptide receptors, and that the predilection of the inflammatory attack for the joint indicates a role for cartilage as a source of the antigenic peptides. A direct role for CII in the development of arthritis is apparent in the CIA model, in which a definite role for MHC class II molecules and a role for CII-derived peptides have been demonstrated [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Remarkably, it was found that the identified MHC class II molecule in the CIA model A^q has a structurally similar peptide binding pocket to that of the shared epitope, expressing DR4 molecules [<xref ref-type="bibr" rid="B4">4</xref>]. In fact, DR4 (DRB1^*0401) and DR1 (DRB1^*0101) transgenic mice are susceptible to CIA because of an immune response to a peptide that is almost identical to that which is involved in A^q-expressing mice [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. They are both derived from position 260-273 of the CII molecule; the peptide binds to the A^qmolecule with isoleucine 260 in the P1 pocket, but with phenylalanine 263 in the P1 pocket of the DR4 and DR1 molecules.</p><p>Although these findings do not prove a role for CII in RA, they show that such recognition is possible and that there are structural similarities when comparing mouse with human. However, there are also strong arguments against such a possibility. First, arthritis can evolve without evidence for a cartilage-specific autoimmunity, as seen with various adjuvant-induced arthritis models [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>] and in several observations using transgenic animals with aberrant immunity to ubiquitously expressed proteins [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>]. Moreover, the MHC association in the adjuvant arthritis models correlates with severity of the disease rather than susceptibility [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>], as has also been observed in RA [<xref ref-type="bibr" rid="B12">12</xref>]. Second, it has not been possible to identify the CII-reactive T cells from RA joints, or to achieve a strong and significant CII proliferative response from T cells derived from RA joints. Most recently these negative observations were corroborated using DR4+CII peptide tetramer reagents [<xref ref-type="bibr" rid="B13">13</xref>]. On the other hand, it has also been difficult to isolate autoreactive CII-specific T cells from CIA, and it can be anticipated that, even in the CIA model, T cells that are specific for CII will be hard to find in the joints [<xref ref-type="bibr" rid="B4">4</xref>].</p><p>We believe that the explanations for these observations in both experimental animals and humans are related to tolerance. The CIA model in the mouse is usually induced with heterologous CII, and is critically dependent on an immune response to the glycosylated CII peptide 256-270, which is bound to the MHC class II A^q molecule. In CII transgenic mice, expressing the heterologous (rat) form of the immunodominant CII 256-270 epitope in cartilage, we observed partial T-cell tolerance. This tolerance is characterized by a low proliferative activity, but with maintained effector functions such as production of IFN-&#x003b3; and the ability to give help to B cells to produce anti-CII IgG antibodies [<xref ref-type="bibr" rid="B14">14</xref>]. Interestingly, these mice were susceptible to arthritis. However, a possibility was that T cells that had newly emerged from the thymus and that were not yet tolerized when the mice were immunized with CII led to the induction of arthritis. We have now addressed this possibility and found that induction of tolerance occurs within a few days, and that mice lacking recent thymic emigrants (ie thymectomized mice) display partially tolerant T cells and susceptibility to arthritis to the same extent as nonthymectomized mice. In addition we found that T cells that are reactive with the nonmodified peptides are relatively more affected by tolerance than T cells that are reactive with the more immunodominant glycosylated variants.</p></sec><sec><title>Objectives:</title><p>To investigate the possibility that T cells that are naturally tolerant to the cartilage protein CII are involved in the development of arthritis, and to exclude a role for nontolerized recent thymic T-cell emigrants in the development of arthritis.</p></sec><sec><title>Materials and methods:</title><p>A mutated mouse CII, expressing glutamic acid instead of aspartic acid at position 266, was expressed in a transgenic mouse called MMC (mutated mouse collagen) that has been described earlier [<xref ref-type="bibr" rid="B14">14</xref>]. The mice were thymectomized, or sham-operated, at 7 weeks of age and allowed to recover for 4 weeks before being immunized with rat CII in complete Freund's adjuvant. Arthritis development was recorded and sera analyzed for anti-CII IgG, IgG₁ and IgG_{2<italic>a</italic>} levels. To assay T-cell effector functions, other MMC and control mice were immunized in the hind footpads with rat CII in complete Freund's adjuvant, and the draining popliteal lymph nodes were taken 10 days later. The lymph node cells (LNCs) were used for proliferation assay, IFN-&#x003b3; enzyme-linked immunosorbent assay (ELISA) and B-cell enzyme-linked immunospot (ELISPOT). For the proliferation assay, 10⁶ cells were put in triplicate cultures in microtitre wells together with antigen and incubated for 72h before thymidine-labelling and harvesting 15-18h later. For IFN-&#x003b3; ELISA analysis, supernatant from the proliferation plates was removed before harvesting and used in an ELISA to quantify the amount of IFN-&#x003b3; produced [<xref ref-type="bibr" rid="B15">15</xref>]. B-cell ELISPOT was performed to enumerate the number of cells producing anti-CII IgG [<xref ref-type="bibr" rid="B16">16</xref>].</p><p>T-cell lines that were reactive towards rat CII were established by immunization with rat CII. An established T-cell line that was reactive with CII and specific for the CII 256-270 peptide was restimulated with freshly collected, irradiated, syngenic spleen cells and rat CII for 3 days followed by 2 weeks of IL-2 containing medium. Immediately before transfer, the cells were labelled with the cytoplasmic dye 5 (and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE) [<xref ref-type="bibr" rid="B17">17</xref>]. Labelled cells (10⁷) were injected intravenously into transgenic MMC mice and nontransgenic littermates. The mice were killed 4 days after cell transfer, and the concentration of CFSE-labelled cells was determined by flow cytometry.</p></sec><sec><title>Results and discussion:</title><p>To investigate whether and how quickly CII-reactive T cells will encounter CII <italic>in vivo</italic>, an established T-cell line that is reactive towards rat CII was labelled with the cytoplasmic dye CFSE and transferred into MMC-QD and control mice. Four days later the mice were killed, and it was found that MMC-transgenic mice had dramatically fewer CFSE-labelled cells in the spleen than did nontransgenic littermates (0.11% compared with 0.57%). Similarly, reduced numbers of CFSE-positive cells were observed in blood. This indicates that the T cells encountered the mutated CII that was present in the cartilage of MMC mice, but not in the nontransgenic littermates. Presumably, CII from cartilage is spread by antigen-presenting cells (APCs) to peripheral lymphoid organs. This observation also suggests that newly exported T cells from the thymus will be tolerized to CII in the periphery within less than 4 days.</p><p>To further investigate whether the MMC mice harbours na&#x000ef;ve or tolerized T cells, the mice were immunized with CII at different time points after thymectomy that were well in excess of the times required for their encounter with CII. After 10 days, the response was analyzed <italic>in vitro</italic> towards both the nonglycosylated and the glycosylated CII 256-270 peptides as well as towards purified protein derivative. The galactosylated form of the peptide (Fig. <xref ref-type="fig" rid="F1">1</xref>) was used because this is the most immunodominant modification [<xref ref-type="bibr" rid="B18">18</xref>]. In contrast to control mice, LNCs from transgenic mice did not proliferate significantly towards the nonglycosylated peptide, indicating that these cells have been specifically tolerized, which is in accordance with earlier observations [<xref ref-type="bibr" rid="B14">14</xref>]. A reduced, but still significant proliferation was also observed toward the immunodominant glycosylated CII peptide. Most important, however, was that the proliferative response in the MMC mice did not decrease after thymectomy. Similarly, a significant IFN-&#x003b3; production towards the glycosylated CII peptide was observed in the MMC mice. The response was somewhat reduced compared with that observed in nontransgenic littermates, and this was especially true for the response toward the nonglycosylated peptide. Again, no decrease in the MMC response by thymectomy was observed. Taken together, the T-cell response in transgenic mice was reduced in comparison with that in the nontransgenic littermates. Furthermore, the response in transgenic animals did not decrease by thymectomy (4 or 8 weeks before immunization), showing that autoreactive T cells are still maintained (and partially tolerized) with significant effector functions at least up to 8 weeks after thymectomy, excluding a exclusive role for recent thymic emigrants in the autoimmune response towards CII. To investigate whether thymectomized mice, lacking recent CII-specific thymic emigrants, were susceptible to CIA, mice were immunized with CII 4 weeks after thymectomy and were observed for arthritis development during the following 10 weeks. Clearly, the thymectomized MMC mice were susceptible to arthritis (five out of 18 developed arthritis; Fig. <xref ref-type="fig" rid="F2">2</xref>), and no significant differences in susceptibility between thymectomized and sham-operated mice, or between males and females, were seen. In accordance with earlier results [<xref ref-type="bibr" rid="B14">14</xref>], MMC transgenic mice had a significantly reduced susceptibility to arthritis as compared with the nontransgenic littermates (<italic>P</italic> &#x0003c; 0.0001 for arthritic scores, disease onset and incidence). All mice were bled at 35 days after immunization, and the total levels of anti-CII IgG were determined. Transgenic mice developed levels of anti-CII IgG significantly above background, but the antibody titres were lower than in nontransgenic littermates (<italic>P</italic> &#x0003c; 0.0001). No effect on the antibody levels by thymectomy was observed, nor did thethymectomy affect the distribution of IgG₁ versus IgG_{2<italic>a</italic>} titres,indicating that the observed tolerance is not associated with a shift from a T-helper-1- to a T-helper-2-like immune response. These findings show that T cells that are specific for a tissue-specific matrix protein, CII, are partially tolerized within a few days after thymus export and that these tolerized cells are maintained after thymectomy. Most important, mice that lack newly exported CII reactive T cells are still susceptible to CIA, suggesting that the partially tolerant T cells are involved in development of arthritis.</p><p>In the light of these data it is possible to explain some of the findings in RA. T-cell reactivity to CII has been shown in RA patients, but with a very weak proliferative activity [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. This is fully compatible with observations in mouse and rat CIA when autologous CII, and not heterologous CII, are used for immunization. This is particularly true if the responses are recorded during the chronic phase of disease, in which the antigen-specific T-cell responses seem to be suppressed in both humans and experimental animals. These observations were confirmed in a recent report [<xref ref-type="bibr" rid="B21">21</xref>] in which it was shown that CII-reactive T-cell activity could be detected in RA patients if IFN-&#x003b3; production but not proliferation was measured. In the present studies in mice the strongest response is seen towards post-translational modifications of the peptide. Because the T-cell contact points are the same whether the peptide is bound to DR4 or to A^q, it is fully possible that post-translational modifications of the peptide also plays a significant role in humans [<xref ref-type="bibr" rid="B22">22</xref>]. The fact that IgG antibodies specific for CII are found in many RA patients could be explained by maintained B-cell helper functions of CII-reactive T cells. In fact, it has been reported [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>] that the occurrence of IgG antibodies to CII is associated with shared epitope DR4 molecules. These observations are thus compatible with a role for CII reactivity in RA. To avoid any confusion, it needs to be stressed that RA is a heterogeneous syndrome in which not only CII, but also other cartilage proteins and other mechanisms are of importance. Such a pathogenic heterogeneity is reflected by the multitude of experimental animal models that have demonstrated how many different pathways may lead to arthritis [<xref ref-type="bibr" rid="B25">25</xref>].</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Malmstr&#x000f6;m</surname><given-names>Vivianne</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>rikard.holmdahi@inflam.lu.se</email></contrib><contrib id="A2" contrib-type="author"><name><surname>B&#x000e4;cklund</surname><given-names>Johan</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Jansson</surname><given-names>Liselotte</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Kihlberg</surname><given-names>Jan</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Holmdahl</surname><given-names>Rikard</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>T cells originate from multipotent cells in the bone marrow, but mature in the thymus. They are educated in the thymus both by positive selection (ie in which only the thymocytes that are capable of interacting with self-MHC survive [<xref ref-type="bibr" rid="B26">26</xref>]) and by negative selection (ie in which recognition of self-antigens that are present in the thymus leads to apoptosis [<xref ref-type="bibr" rid="B27">27</xref>]). After approximately 2 weeks the T cells leave the thymus as mature but na&#x000ef;ve cells [<xref ref-type="bibr" rid="B28">28</xref>]. Outside the thymus T cells circulate in search for a peptide-MHC complex to recognize. If a na&#x000ef;ve T-cell fails to find such an antigenic complex its lifespan will be limited [<xref ref-type="bibr" rid="B29">29</xref>]. Instead, new T cells will continuously emerge from the thymus and replace the ones leaving the pool of na&#x000ef;ve T cells.</p><p>The extent to which the lifespan is prolonged if such T cells meet their antigen is currently under debate. One possibility is that na&#x000ef;ve T cells may only respond to antigens that are presented by professional APCs and in the proper microenvironment; potential antigens that are located in other cells and tissues are ignored [<xref ref-type="bibr" rid="B30">30</xref>]. Alternatively, some T cells may interact with antigens outside of the context of an immune response (eg in the physiological state by self-antigens presented for the T cells in draining lymph nodes). Recent data [<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>] suggest that a low-affinity interaction, for example by the appropriate MHC but without the relevant peptide, will lead to the prolongation of the lifespan of newly exported na&#x000ef;ve T cells. Other findings suggest that the transgenic expression of foreign antigen in peripheral tissues will lead to maintained survival of transgenic T cells with a tolerized phenotype [<xref ref-type="bibr" rid="B33">33</xref>]. The importance of such T cells in autoimmune disease is of great significance, but their existence and function are still obscure.</p><p>Target antigens, which are of importance for development of autoimmune disease, may evoke either ignorance or tolerance. For example, in mice that are transgenic for a T-cell receptor (TCR) that is specific for myelin basic protein (MBP), which is a central nervous system-specific protein, the T cells ignore the existence of the antigen unless they are not activated by other means [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>]. In other autoimmune diseases, however, the relevant self-tissue is more readily exposed to the immune system and consequently a pool of T cells that recognize the self-antigen, but which fail to be deleted, might exist. We believe that RA, as well as CIA, which is an experimental model of RA, are such diseases.</p><p>Both RA and CIA show MHC class II association, whereby CIA in mice is associated with the A^q molecule expressed by the H-2^q haplotypes [<xref ref-type="bibr" rid="B36">36</xref>] and RA in humans with the DRB1^*0401/DRA molecule in the DR4 haplotype [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>]. Interestingly, the MHC association indicates structural relevance, because the peptide binding cleft of the A^q and the DRB1^*0401/DRA molecule are very similar and bind peptides from the same CII region [<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. Consequently, mice that express the DRB1^*0401/DRA molecule are susceptible to CIA [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B41">41</xref>]. The MHC class II association provides a strong indication that the model is T-cell dependent [<xref ref-type="bibr" rid="B42">42</xref>,<xref ref-type="bibr" rid="B43">43</xref>,<xref ref-type="bibr" rid="B44">44</xref>].</p><p>CIA is not only a useful model for RA, but is also a valuable tool for studying the interactions between the immune system and cartilage. CII is the major protein constituent of cartilage, and is highly conserved, with only minor amino acid substitution between species. When comparing the triple helical region of mouse [<xref ref-type="bibr" rid="B45">45</xref>] and rat [<xref ref-type="bibr" rid="B46">46</xref>] CII, only 13 out of 1015 amino acids were substituted. CII is further subjected to extensive post-translational modifications, including hydroxylation of prolines and lysines and glycosylation of hydroxylysines.</p><p>There are only a few reports in the literature in which glycopeptides were involved in immune responses [<xref ref-type="bibr" rid="B47">47</xref>,<xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B49">49</xref>,<xref ref-type="bibr" rid="B50">50</xref>,<xref ref-type="bibr" rid="B51">51</xref>], and the present CII 256-270 peptide is the only one that is naturally selected [<xref ref-type="bibr" rid="B52">52</xref>,<xref ref-type="bibr" rid="B53">53</xref>]. When H-2^q mice are immunized with rat CII, a T-cell response toward CII 256-270 is elicited. This T-cell epitope contains three sites for hydroxylations and two sites for glycosylations [<xref ref-type="bibr" rid="B2">2</xref>], and these modifications participate in the immune response. To study self-tolerance of CII 256-270 reactive T cells, we previously generated mice with transgenic, cartilage-specific expression of the rat CII epitope [<xref ref-type="bibr" rid="B14">14</xref>]. These MMC transgenic mice have a reduced susceptibility to rat CIA; only 45% of the mice develop a severe polyarthritis after immunization, as compared with 95% of the non-transgenic littermates. The decreased incidence of CIA can be explained by T-cell tolerance, because the immunodominant T-cell epitope from rat CII is now expressed as self in cartilage. The existence of immune tolerance towards glycopeptides has not been addressed before, and is an important issue because glycosylated CII is more arthritogenic than nonglycosylated CII [<xref ref-type="bibr" rid="B52">52</xref>].</p><p>The present study was designed to investigate whether tolerized T cells are capable of mediating disease or whether na&#x000ef;ve T cells, that are not yet tolerized, make a necessary contribution to pathology. Although the thymus shrinks with age, it continues to produce fresh thymus emigrants (ie na&#x000ef;ve T cells). If these are specific for a distant tissue-specific self-antigen we postulated that it would take time before the T cells acquired tolerance. Possibly, the arthritis in the transgenic mice could occur as a result of the encounter of such nontolerized T cells with the immunogenic CII, rather than first being tolerized by the CII derived from cartilage. To address this possibility we thymectomized adult mice and immunized them with CII after 4 weeks, thereby allowing the T cells to become tolerized. The results indicate that the tolerized T cells themselves could be involved in the development of arthritis. We also studied immune responses in lymph node cultures derived from thymectomized mice, and showed T-cell tolerance toward the glycosylated version of the CII 256-270 epitope. In spite of peripheral tolerance this peptide elicited IFN-&#x003b3; production, which we believe is important for disease development.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Mice</title><p>C3H.Q mice (H-2^q) were originally a gift from Dr DC Shreffler, St Louis, USA; DBA/1 mice originated from the Jackson Lab (Bar Harbor, Maine, USA); and B10.Q mice (H-2^q) were from Dr Jan Klein, T&#x000fc;bingen, Germany. All arthritis and <italic>in vitro</italic> experiments were performed on F1 animals (B10.Q&#x000d7;C3H.Q). The transgenic MMC-1 mice (in this work referred to as MMC mice) have previously been described [<xref ref-type="bibr" rid="B14">14</xref>]. Briefly, the MMC mouse, which was originally from the C3H.Q background, was back-crossed eight generations to the B10.Q mice (MMC-BQ). The MMC transgene is a mutated mouse CII gene, in which position 266 has been changed from a D into an E, thereby containing the rat CII 256-270 epitope and showing cartilage-restricted expression. Nontransgenic QD mice (B10.Q&#x000d7;DBA/1) F1 were used for the establishment of a CII-specific T-cell line. MMC-QD mice were used in transfer experiments as recipient mice for migration studies. All animals were bred and used in our animal facilities.</p></sec><sec><title>Antigens</title><p>Rat CII was prepared from the SWARM chondrosarcoma by pepsin digestion [<xref ref-type="bibr" rid="B54">54</xref>], and further purified as previously described [<xref ref-type="bibr" rid="B55">55</xref>]. The peptides were synthesized as previously described [<xref ref-type="bibr" rid="B2">2</xref>], the glycosylated CII 256-270 peptide with a &#x003b2;-D-galactopyranose residue on L-hydroxylysine at position 264 (Fig. <xref ref-type="fig" rid="F1">1</xref>) [<xref ref-type="bibr" rid="B56">56</xref>]. Both collagen and collagen peptides were dissolved and stored in 0.1mol/l acetic acid.</p></sec><sec><title>Thymectomy, immunization and arthritis evaluation</title><p>Adult mice (7 weeks old) were thymectomized under anaesthesia. In parallel, mice were also sham-operated under the same conditions. The mice were allowed to recover for 4 weeks before CII immunization. The mice were immunized intradermally in the base of the tail with 100 &#x003bc; g rat CII emulsified 1:1 in complete Freund's adjuvant (IFA; Difco, Detroit, MI, USA). They were also boosted with 50 &#x003bc; g rat CII emulsified 1:1 in IFA (Difco), 5weeks later. At that time blood samples were also taken for analysis of anti-CII antibody responses. The amounts of total anti-CII IgG as well as the IgG₁ and IgG_{2<italic>a</italic>} isotypes were determined through quantitative ELISA [<xref ref-type="bibr" rid="B57">57</xref>]. Using this assay no CII antibody reactivities may be found in either wild-type or MMC transgenic mice before immunization. Development of clinical arthritis was followed through visual scoring of the mice, starting 2 weeks after immunization and continuing until the end of the experiment. The arthritis was scored using a scale of 1-3 (1, one affected joint; 2, two or more arthritic joints; and 3, severe arthritis involving the entire paw), giving a maximum score of total 12 per individual.</p></sec><sec><title>Lymphocyteassays</title><p>To assay T-cell effector functions, mice were immunized in the hind footpads with 50 &#x003bc; g rat CII emulsified 1:1 in complete Freund's adjuvant and the draining popliteal lymph nodes were taken 10 days later. The LNCs were used for proliferation assay, IFN-&#x003b3; ELISA and B-cell ELISPOT. For the proliferation assay, 10⁶ cells were put in triplicate cultures in microtitre wells together with antigen and incubated for 72h before thymidine-labelling and harvesting 15-18h later. For IFN-&#x003b3; ELISA, supernatant from the proliferation plates was removed before harvesting and used in an ELISA to quantify the amount of IFN-&#x003b3; produced, as previously described [<xref ref-type="bibr" rid="B15">15</xref>]. B-cell ELISPOT analyses were performed as previously decribed (but with minor modifications) [<xref ref-type="bibr" rid="B16">16</xref>] in order to enumerate the number of cells that produce anti-CII IgG. In all experiments the LNCs were assayed from individual mice and statistics were calculated from the biological variation.</p></sec><sec><title>T-cell lines and CFSE assay</title><p>A T-cell line that is reactive toward rat CII was established by immunization of QD mice with rat CII in the hind footpads. Eight days later draining lymph nodes were collected and restimulated <italic>in vitro</italic> with rat CII for 4 days. After 4 days of antigen stimulation, cells were collected, washed and allowed to rest in the absence of APCs and in the presence of IL-2 for 1 week. Thereafter the T cells were restimulated with freshly collected, irradiated (3000 Rad), syngenic spleen cells and rat CII for 3 days (5&#x000d7;10⁵ T cells/ml, 5&#x000d7;10⁶ APCs/ml, 10 &#x003bc; g rat CII/ml) followed by 2 weeks of resting in IL-2-containing medium. At the time of restimulation, an aliquot of the cell line was also tested for antigen specificity. The cell line responded toward denatured CII, the nonmodified CII 256-270 peptide and the glycosylated CII 256-270 peptide with proliferation and IFN-&#x003b3; production (data not shown). This scheme of 2-weekly restimulation cycle was repeated a further two times before transferring the cells to recipient mice.</p><p>Because our purpose was to follow the migration pattern of CII-specific T cells <italic>in vivo</italic>, and not transfer of disease, the cells were allowed to rest without APCs or antigen <italic>in vitro</italic> for 7 days after restimulation in the presence of IL-2 before transfer. The cells were phenotypically characterized at this stage, and were found to be CD4-positive with high and intermediate expression of CD54 and CD69, respectively, and low expression of CD25 and CD95 ligand. Immediately before transfer, the cells were labelled with the cytoplasmic dye CFSE (Molecular Probes, Leiden, The Netherlands) [<xref ref-type="bibr" rid="B17">17</xref>] by a 10-min incubation of the cells with 5 &#x003bc; mol/l CFSE at 37&#x000b0;C. Labelled cells (10⁷) were then injected intravenously into transgenic MMC mice and nontransgenic littermates. The mice were sacrificed 4 days after cell transfer and the frequency of CFSE-labelled cells were determined by flow cytometry with a fluorescence-activated cell sorter and using CellQuest Software (Becton Dickinson, Mountain View, CA, USA).</p></sec><sec><title>Statistical analysis</title><p>Incidence of arthritis was analyzed using the &#x003c7;² test, and antibody levels and proliferative responses were analyzed using the Mann-Whitney U-test.</p></sec></sec><sec><title>Results</title><sec><title>Thymectomized mice develop CIA with the same features as nonthymectomized animals</title><p>Seven-week-old mice were thymectomized or sham-operated. Four weeks later the mice were immunized to induce CIA and disease was scored for the following 10 weeks. Clearly, the thymectomized MMC mice were susceptible to arthritis (five out of 18 developed arthritis). No significant differences in susceptibility between thymectomized and sham-operated mice, or between males and females, were seen (Fig. <xref ref-type="fig" rid="F2">2</xref>, Table <xref ref-type="table" rid="T1">1</xref>). In concordance with earlier results [<xref ref-type="bibr" rid="B14">14</xref>], MMC transgenic mice had a significantly reduced susceptibility to arthritis as compared with the nontransgenic littermates (<italic>P</italic> &#x0003c; 0.0001 for arthritic scores, disease onset and incidence). Disease incidence in the transgenic mice was even lower than has previously been found, which is most likely due to the change of gene background from C3H.Q to (B10.Q&#x000d7;C3H.Q) F1. A similar reduction in incidence was also observed in (B10.RIII&#x000d7;C3H.Q) F1 MMC transgenic mice [<xref ref-type="bibr" rid="B36">36</xref>].</p></sec><sec><title>Antibody responses</title><p>All mice were bled at 35 days after immunization, and the total levels of anti-CII IgG determined. As shown in Table <xref ref-type="table" rid="T1">1</xref>, transgenic mice developed significant levels of anti-CII IgG, but the antibody titres were lower than in nontransgenic littermates (<italic>P</italic> &#x0003c; 0.0001). An effect on the antibody levels by thymectomy could not be observed, and neither did the thymectomy affect the distribution of IgG₁ titres versus those of IgG_{2<italic>a</italic>} (Fig. <xref ref-type="fig" rid="F3">3</xref>), indicating that the observed tolerance is not associated with a shift from a T-helper-1- to a T-helper-1-like immune response. Only four animals had more IgG₁ than IgG_{2<italic>a</italic>} antibodies, two with arthritis (one transgenic and one littermate) and two without clinical disease (one transgenic and one littermate).</p></sec><sec><title>Lymph node cultures</title><p>Antigen-specific T-cell activation was investigated at different time points after thymectomy to determine whether the mice harboured tolerized or na&#x000ef;ve T cells. The mice were immunized with rat CII, and 10 days later the recall response was determined toward the nonglycosylated CII 256-270 peptide, the glycosylated CII 256-270 peptide and purified protein derivative (Fig. <xref ref-type="fig" rid="F4">4</xref>). The galactosylated form of the peptide (Fig. <xref ref-type="fig" rid="F1">1</xref>) was used because this is the most immunodominant modification [<xref ref-type="bibr" rid="B18">18</xref>].</p><p>LNCs from transgenic mice proliferated poorly in response to CII peptides, indicating that these cells had been specifically tolerized, which is in accordance with earlier observations (Fig. <xref ref-type="fig" rid="F4">4</xref>) [<xref ref-type="bibr" rid="B14">14</xref>]. Nevertheless, the proliferative response to the glycosylated peptide was still significant (<italic>P</italic> =0.002 versus background values, and <italic>P</italic> =0.003 versus the response toward the nonglycosylated peptide as analyzed in all mice), suggesting that the remaining response is directed toward this peptide. Most important, however, was that the proliferative response in the transgenic mice did not decrease after thymectomy.</p><p>The supernatants of the proliferation cultures were also analyzed for IFN-&#x003b3; content (Fig. <xref ref-type="fig" rid="F4">4</xref>). The response was somewhat reduced compared with that observed in non-transgenic littermates, notably when comparing the response towards the nonglycosylated peptide. Again, no reduction in the IFN-&#x003b3; response was observed after thymectomy. In fact, a significant response towards the nonglycosylated peptide was observed in mice immunized 8 weeks after thymectomy. Still, this response was very much reduced in comparison with that induced with the glycosylated peptide or with that observed in non-transgenic littermates.</p><p>In summary, the autoimmune response in transgenic animals did not decrease after thymectomy, showing that these T cells were not newly exported, but were maintained after being partially tolerized. LNCs from transgenic mice produced less IFN-&#x003b3; than the nontransgenic litter-mates in response to the nonglycosylated CII 256-270 peptide, as previously reported for MMC mice with intact thymus. The time after thymectomy (4 or 8 weeks) did not attenuate the autoimmune response to the nonglycosylated or the glycosylated peptide. Thus, the responses were still significant 8 weeks after thymectomy.</p><p>In similarity with the observed T-cell responses, significant numbers of B cells producing anti-CII IgG were seen in CII-immunized MMC mice using quantitative ELISPOT assays (Fig. <xref ref-type="fig" rid="F5">5</xref>). The response did not decrease after thymectomy. Together with the data shown in Figure <xref ref-type="fig" rid="F3">3</xref>, these data show that the significant, but reduced antibody levels to CII in the MMC mouse is reflected by numbers of antibody-secreting cells. In comparison with earlier data using the MMC transgenic mice, this was a more pronounced reduction in the B-cell response. The different genetic background of the mice used in the present study may explain this, in which an F1 with the relatively low responder B10 background was used. Again the same phenomenon was observed in (B10.RIII&#x000d7;C3H.Q) F1 transgenic mice immunized with rat CII [<xref ref-type="bibr" rid="B46">46</xref>].</p></sec><sec><title><italic>In vivo</italic> transfer of rat collagen type II-reactive T cells</title><p>An activated CFSE-labelled T-cell line, which was reactive towards rat CII, was transferred into mice to follow T-cell migration and survival. Four days after cell transfer the transgenic mice had dramatically fewer CFSE-labelled cells in the spleen than did nontransgenic littermates (Fig. <xref ref-type="fig" rid="F6">6a</xref>); similar reduced numbers of CFSE-labelled cells were also seen in blood (data not shown). Only occasionally could cells be detected in the lymph nodes or in the thymus. These had only one fluorescence peak, indicating that they had not proliferated <italic>in vivo</italic>. Extracted spleen cells were analyzed for phenotypic markers, such as CD71, CD69 and CD43. CFSE-labelled cells stained positive for CD43 and negative for CD71, and approximately 50-70% were CD69 positive (Fig. <xref ref-type="fig" rid="F6">6b</xref>, and data not shown). However, there was no difference in the expression levels of these activation markers between cells transferred into MMC or nontransgenic mice.</p><p>The dramatic differences in detectable events in the spleen and blood between MMC and littermate recipients suggest that the T cells interacted with the antigens that were present in MMC mice, but that this did not occur in their nontransgenic littermates.</p></sec></sec><sec><title>Discussion</title><p>In the present study we showed that T cells that are specific for a tissue-specific matrix protein, CII, are partially tolerized and that these cells are maintained after thymectomy. By adult thymectomy, the continuous outflow of na&#x000ef;ve T cells in mice with an established, normal T-cell repertoire was terminated. Still, the partial tolerant state, defined by lack of proliferative capacity but with significant IFN-&#x003b3; response and B-cell helping capacity, was maintained up to 8 weeks after thymectomy. In addition, thymectomy did not affect the susceptibility to CIA, suggesting that the partially tolerant T cells are involved in development of arthritis.</p><p>The loss of detectable CFSE-labelled rat CII reactive T cells in MMC mice suggests that the T cells had encountered their antigen. However, where this interaction takes place and the outcome of it was not deduced in the present study. It may be speculated that the interaction takes place in the circulation due to the inability of T cells to reach lymph nodes (the T-cell line is CD62 ligand negative; data not shown). Furthermore, the rapid decrease in cell numbers implies that this interaction occurs shortly after cells enter the periphery. However, it could also be caused by a migration of T cells into tissue, even in the absence of local inflammation, which would promote such traffic. If the T cells do migrate out into the tissue the outcome could possibly be retention and even expansion of the T-cell population, and the latter may even lead to outbreak of arthritis. Importantly, we showed that rat CII-reactive T cells interact with the antigen present in MMC mice, because fewer cells are found 4 days after cell transfer in transgenic mice compared with in control animals.</p><p>These findings suggest that the immune system normally interacts with cartilage and that the T cells that participate in such an interaction survive and have maintained function. The questions that arise are when, how and where the T cells and the CII antigen normally meet.</p><p>Presumably the interaction takes place in the lymph nodes that drain the joints, and depends on the amount of CII that is degraded and transported to these lymph nodes, the nature of the transporting APCs and the MHC-binding efficiency after processing. There is no lymphatic drainage of cartilage, but the surrounding synovial tissue contains numerous macrophages that have the capacity to engulf proteins scavenged from cartilage. It is likely, but not proven, that some of these macrophages transport CII to lymph nodes and present them to the immune system. It is not likely that synovial dendritic cells fulfil these functions, because dendritic cells do not present collagen [<xref ref-type="bibr" rid="B58">58</xref>,<xref ref-type="bibr" rid="B59">59</xref>], in contrast to the crucial role of dendritic cells to prime immune responses to other antigens [<xref ref-type="bibr" rid="B60">60</xref>]. Thus, the presentation of CII in the periphery is limited but must occur.</p><p>The circulation of newly exported T cells is most likely very rapid, but the possibility of T cells encountering specific antigens is of course also limited by the low numbers of CII-reactive T cells. If tolerance is an important mechanism for protection from immune-mediated inflammatory attacks on self-tissues, it should be important to allow the interactions between self-reactive T cells and the respective self-antigens to occur as quickly as possible. In this context, 4 weeks is a long time for a circulating T cell. Studies on class I restricted CD8⁺ T cells [<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B61">61</xref>] have shown that na&#x000ef;ve T cells exported from thymus will not expand unless they meet their specific antigen. However, they may survive for at least 2 weeks but will subsequently disappear. Memory cells on the other hand are more easily maintained, even in the absence of antigen. The lifespan of CD4⁺ T cells is more controversial, but their survival is probably more limited than for CD8⁺ T cells. The maintenance of memory CD4⁺ T cells has been suggested [<xref ref-type="bibr" rid="B62">62</xref>] to be dependent on the continuous presence of the specific antigen. In the case of the CII-reactive T cells, the specific antigen is in the cartilage and encounters with this antigen may lead to maintenance, which most likely is synonymous with 'partial' tolerance induction.</p><p>We describe the observed tolerance as 'partial' because we have no better word for the observed state. The description is made on the population level, and not on the cellular level. Thus, the observed partial tolerance could have different explanations. One possibility is tolerization on the cellular level, for instance, by induction of a different mode of signalling through TCR or through costimulatory molecules [<xref ref-type="bibr" rid="B63">63</xref>]. Alternatively, the tolerance could be achieved by deletion of T cells that express TCRs with high avidity for the CII peptide, leaving only the low-avidity T cells, which might reach the critical threshold for 'partial' activation (IFN-&#x003b3; secretion and B-cell help, but no IL-2 secretion). Nevertheless, if this should be the case the triggered low avidity CII-reactive cells are maintained and play a role in arthritis.</p><p>Interestingly, the glycosylated peptide elicited a stronger recall response than the synthetic peptide, which is in accordance with the finding that glycosylated CII is more arthritogenic than nonglycosylated CII [<xref ref-type="bibr" rid="B52">52</xref>]. How can such a finding be explained? One possibility is that the stronger response to the glycosylated peptide reflects a greater number of various TCRs (and thus T cells) that are capable of interacting with the glycosylated peptide, possibly because of the sugar being quite flexible in its position. Alternatively, the expression level of the glycosylated epitope will also influence tolerance induction. If there is low expression of this variant, less glycopeptide-specific T cells will become tolerized <italic>in vivo</italic>. It should also be emphasized that the glycopeptide-specific T cells, which may recognize a structure much larger than any amino acid side chain, are apparently efficiently selected in the thymus, despite the absence of the relevant peptides. The structural basis of the positive selection of these T cells remains to be clarified.</p><p>The role of long-lived self-reactive T cells in mediating effector functions is of course a central issue in the understanding of chronic autoimmune disease. The location and the physical state of the relevant self-antigen will, to a large extent, determine the activity of such T cells. In experimental allergic encephalomyelitis (EAE), an animal model for multiple sclerosis, an interesting study [<xref ref-type="bibr" rid="B64">64</xref>] has been reported on the breakage of tolerance. Mice were administered the MBP peptide Ac1-11, and developed a severe and acute encephalomyelitis. After recovering from disease the mice were given a superantigen that was reactive to the same subtype of T cells that initially mediated disease. These T cells were now tolerant, but the superantigen was able to break this tolerance and the mice suffered a second exacerbation of disease. Thus, T cells that are specific for EAE seem normally to ignore the existence of the self-antigen unless they are activated, which leads to trafficking into the central nervous system and subsequently to induction of disease. Consequently, the recall T-cell response to MBP is as strong to self-MBP as to foreign MBP. In contrast, the recall response to CII differs dramatically between self-CII and foreign CII due to tolerance induction to the CII 256-270 peptide, which is derived from cartilage.</p><p>From the above-stated reasons it is likely that tolerance to cartilage-derived collagen peptides expressed on APCs in peripheral lymphoid organs, but not in the thymus, occurs. When tolerance is broken, for example after the introduction of adjuvant, which will activate APCs, autoimmune disease can ensue. At present we know very little regarding what can cause breakage of tolerance. In mice we often immunize with tissue-specific proteins in the presence of adjuvant or with adjuvant only [<xref ref-type="bibr" rid="B65">65</xref>], but what happens in humans with autoimmune diseases that are triggered by unknown causes? Peripheral tolerance towards tissue-restricted antigens in mice may include different routes of tolerance induction and maintenance [<xref ref-type="bibr" rid="B33">33</xref>]. Knowledge about such mechanisms would certainly help in understanding the basic mechanisms that cause autoimmune disease and will provide possibilities to control and limit the chronic development of pathologic inflammation.</p></sec>

Autologous stem-cell transplantation in refractory autoimmune diseases after <italic>in vivo</italic> immunoablation and <italic>ex vivo</italic> depletion of mononuclear cells

<sec><title>Introduction:</title><p>Patients with persistently active autoimmune diseases are considered to be candidates for autologous SCT. We performed a phase 1/2 study in a limited number of patients who were refractory to conventional immunosuppressive treatment. Following a period of uncontrolled disease activity for at least 6 months, autologous SCT was performed, after <italic>in vivo</italic> immunoablation and <italic>ex vivo</italic> depletion of mononuclear cells.</p></sec><sec><title>Aims:</title><p>To investigate feasibility, toxicity and efficacy of the treatment, and the incidence of emergent infections.</p></sec><sec><title>Methods:</title><p>Seven patients (aged between 23 and 48 years) were included in the single-centre trial: one had relapsing polychondritis, three had treatment-refractory SLE and three patients had SSc. Stem-cell mobilization was achieved by treatment with moderate-dose cyclophosphamide (2 g/m²; in terms of myelotoxic side effects or myelosuppression) and granulocyte colony-stimulating factor (G-CSF). CD34^- cells of the leukapheresis products were removed by high-gradient magnetic cell sorting. After stem-cell collection, immunoablation was performed with high-dose cyclophosphamide (200 mg/kg body weight) and antithymocyte globulin (ATG; 90 mg/kg body weight). Autologous SCT was followed by reconstitution of the immune system, which was monitored by six-parameter flow cytometry and standard serology. The trial fulfilled the European League Against Rheumatism (EULAR) and the European Group for Blood and Marrow Transplantation (EBMT) guidelines for blood and bone marrow stem-cell transplants in autoimmune disease.</p></sec><sec><title>Results:</title><p>Among the seven patients studied, the patient with relapsing polychondritis and the patients with SLE were successfully treated and remained in complete remission during a follow up of 10-21 months. Remission persisted despite reconstitution of the immune system, resulting in high numbers of effector-/memory-type T-helper lymphocytes and increasing populations in the na&#x000ef;ve T-cell compartment. Before autologous SCT, one of the patients with SLE had a long-lasting secondary antiphospholipid syndrome, with high anticardiolipin antibodies and thromboembolic events. After autologous SCT the antiphospholipid antibodies became negative, and no thrombosis occurred during follow up. Two of the patients with SSc were unaffected by treatment with autologous SCT for 6 or 13 months. The other patient with SSc died 2 days after autologous SCT because of cardiac failure.</p><p>During stem-cell mobilization with G-CSF, flares of autoimmune disease were seen in the patient with polychondritis and in one patient with SLE. The strategy utilized for depletion of CD34^- cells led to a reduction by 4.5-5 log of contaminating CD3⁺ cells in the transplant. T-cell add-back was required in the patient with polychondritis and in one patient with SLE to provide a dose of 1&#x000d7;10⁴ CD3⁺ cells/kg body weight for the transplant.</p></sec><sec><title>Discussion:</title><p><italic>In vivo</italic> immunoablation in combination with autologous SCT after <italic>ex vivo</italic> depletion of CD34^- cells can block the autoimmune process in relapsing polychondritis or SLE without incidence of severe infections. The remissions were achieved in patients with advanced disease that was refractory to previous intensive immunosuppressive therapy. The present results do not indicate that large-scale contamination of the stem-cell transplant with autoreactive cells after selection for CD34⁺cells occurred. After the preparative regimen, the application of G-CSF was avoided, because induction of flares of the autoimmune disease were noticed during the mobilization of stem cells. In SSc patients, distinct remissions were not observable after autologous SCT; the serological and clinical status did not improve. Follow-up periods of more than 12 months may be required to identify successful treatment with autologous SCT in SSc patients. Among the various autoimmune diseases the efficacy of autologous SCT appears to be dependent on the underlying pathophysiology. The results of the present phase 1/2 study suggest that patients with advanced stage SSc should not be treated with autologous SCT, until the reasons for the lack of response and the possible mortality due to cardiac complications are identified. The observation of flares of autoimmune disease after application of G-CSF emphasizes the need for critical evaluation of the role of G-CSF in immunoablative regimens.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Rosen</surname><given-names>Oliver</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>oliver.rosen@rz.hu-berlin.de</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Thiel</surname><given-names>Andreas</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Massenkeil</surname><given-names>Gero</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Hiepe</surname><given-names>Falk</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>H&#x000e4;upl</surname><given-names>Thomas</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Radtke</surname><given-names>Hartmut</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Burmester</surname><given-names>Gerd R</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Gromnica-Ihle</surname><given-names>Erika</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Radbruch</surname><given-names>Andreas</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="B0" contrib-type="author"><name><surname>Arnold</surname><given-names>Renate</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Refractory autoimmune diseases cause a high degree of morbidity and even mortality, although they are not considered to be malignant diseases. During treatment with conventional and experimental immunosuppression, patients can experience treatment-related morbidity without significant gain in quality of life. Autologous SCT is a novel experimental approach for treating patients with refractory autoimmune diseases [<xref ref-type="bibr" rid="B1">1</xref>]. Worldwide, 74 patients with severe autoimmune disease have thus far been treated in 22 centres [<xref ref-type="bibr" rid="B2">2</xref>]. Of these 74 patients, 38 received autologous SCT for treatment of rheumatic autoimmune diseases.</p><p>In the present study one patient with therapy-resistent polychondritis, three patients with advanced SLE and three patients with SSc qualified for an aggressive experimental therapy. After stem-cell mobilization all patients were treated with a rigorous immunosuppressive regimen including cyclophosphamide and ATG to achieve <italic>in vivo</italic> depletion of T cells and other mononuclear cells. The preparative regimen was followed by autologous SCT of CD34⁺ cells after an effective <italic>ex vivo</italic> depletion of mononuclear cells by high-gradient magnetic cell sorting in order to exclude contamination of the transplant with CD34^- cells. The present phase 1/2 trial was aimed at investigating the toxicity of this protocol and the incidence of infections. In addition, the efficacy of autologous SCT with respect to clinical and serological remissions and their duration was evaluated.</p></sec><sec sec-type="methods"><title>Patients and methods</title><sec><title>Patients</title><p>All patients had long-lasting histories of severe and progressive disease without any signs of improvement under conventional immunosuppressive treatment. Inclusion criteria were defined as persistently active disease with poor prognosis and inadequate response to standard protocols (glucocorticoids and at least two different regimens of immunosuppressive drugs, such as intravenous cyclophosphamide 800-1000 mg/application). Furthermore, the patients needed to have adequate function of all major organs in order to tolerate conditioning and transplantation. The exclusion criteria were infections and uncontrolled arrhythmia or congestive heart failure. Further exclusion criteria were as follows: ejection fraction below 50% determined by echocardiogram; lung function test (LFT; transfer factor for carbon monoxide [TLCO] &#x0003c;45%); glomerular filtration rate below 40ml/min or serum creatinine greater than 2.0 mg/dl; hyperalimentation; and age greater than 59 years. The patients were included in the trial only after written consent had been obtained. The present study on autologous SCT for refractory autoimmune diseases was approved by the state ethics committee.</p><p>Patient 1</p><p>A 41-year-old female was admitted with relapsing polychondritis, which was first diagnosed in 1985. The disease was manifested by severe arthralgias, costosternal pain, vasculitis, scleritis, saddle nose and tracheal involvement; the patient had also sustained a life-threatening episode of pyoderma gangrenosum. Despite continuous and intensive conventional therapy for several years, no remission was achieved. During disease progression there was a risk of developing a tracheo-oesophageal fistula. The previous therapy regimens had included intravenous Ig, high-dose methylprednisolone, methotrexate, anti-CD4 antibody and intravenous cyclophosphamide (cumulative dose 6.0 g/m² per month) with concomitant application of steroids. At admission, the daily dose of methylprednisolone was 30 mg. Her Karnofsky score was 60%.</p><p>Patient 2</p><p>A 27-year-old female was diagnosed as having severe SLE at the age of 16 years. During the course of disease, erythema, arthralgia, myalgia, abdominal vasculitis, polyserositis, nephrotic syndrome and pericardial effusions had been observed. Despite consecutive treatments with high-dose methylprednisolone, hydroxychloroquine, azathioprine, intravenous cyclophosphamide (cumulative dose 2.8 g/m² per month), cyclosporine A, mycophenolate mofetile and daily doses of prednisolone of at least 30mg, the disease activity remained uncontrolled for 1.5 years before stem-cell therapy. The patient had been hospitalized for the 15 months before autologous SCT. Her Karnofsky score was 40% and her European Consensus on Lupus Activity measurement (ECLAM) score was 6.5. This patient had serum antibodies against double-stranded DNA (Table <xref ref-type="table" rid="T1">1</xref>); she fulfilled the classification criteria of the American College of Rheumatology [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>Patient 3</p><p>A 48-year-old female had had severe SLE since 1993. The disease manifested as polyserositis, arthralgias, peripheral neuropathy, nephrotic syndrome, pericardial effusions and ventricular tachycardia (the latter was treated with propanolol). Treatment had included high-dose methylprednisolone, hydroxychloroquine, azathioprine, methotrexate, intravenous Ig, monthly intravenous cyclophosphamide (cumulative dose 2.7 g/m² per month) and mycophenolate mofetile. At admission, the patient was under treatment with prednisolone (20 mg/day) and oral morphium sulphate (120 mg/day). Her Karnofsky score was 60% and her ECLAM score was 6. This patient had serum antibodies against double-stranded DNA (Table <xref ref-type="table" rid="T1">1</xref>); she fulfilled the classification criteria of the American College of Rheumatology [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>Patient 4</p><p>A 37-year-old male had been diagnosed with SLE in 1989, with a nephrotic syndrome and oral lesions, erythema, arthralgias, and cardiac and pulmonary involvement. Despite treatment with prednisolone, azathioprine, intravenous cyclophosphamide (cumulative dose 7.3 g/m²) and high-dose methylprednisolone, the nephrotic syndrome (histology indicated lupus nephritis of World Health Organization grade IV) and other manifestations had not improved, and the ventricular arrhythmia (multiple couplets, one triplet, multiple bigemini) remained uncontrolled. At admission, the dose of prednisolone was 100 mg/day. His Karnofsky score was 70% and his ECLAM score was 10. This patient had serum antibodies against double-stranded DNA (Table <xref ref-type="table" rid="T1">1</xref>); he fulfilled the classification criteria of the American College of Rheumatology [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>Patient 5</p><p>A 23-year-old female was first diagnosed as having diffuse SSc at age 12 years. During the course of disease, microstomia, xerostomia, arthralgias, dysphagia, cutaneous necrosis with Raynaud's phenomenon and the onset of lung fibrosis (by high-resolution computed tomography [HRCT] scan; LFTs - total lung capacity [TLC] 72.6%, residual volume [as percentage of TLC] 127%, single breath (SB) TLCO 61.8%) had occurred. Progression of disease was observed under consecutive treatment periods with D-penicillamine, prednisolone, azathioprine, cyclosporine A, oral cyclophosphamide for 12 months (cumulative dose 3.8 g/m²) and dapsone. Treatment at admission was only symptomatic and without steroids. Her Karnofsky score was 60% and her skin score was 19.</p><p>Patient 6</p><p>A 25-year-old male was diagnosed with diffuse SSc in 1995 with microstomia, arthralgias, dysphagia, cutaneous necrosis with Raynaud's phenomenon, and onset of lung fibrosis (by HRCT scan; LFTs - TLC 93.2%, residual volume [in percentage of TLC] 159%, TLCO-SB 86.2 %). His finger mobility was severely limited, and he had lost 10 kg in weight since 1997. Treatment had included prednisolone, azathioprine and symptomatic therapy. At admission, the daily dose of prednisolone was 5 mg. His Karnofsky score was 60%, and his skin score was 30. In this patient steroids were applied due to the rapid progression of the disease.</p><p>Patient 7</p><p>A 45-year-old female had diffuse SSc that was first diagnosed in 1996. During the preceding 6 months she had lost 13 kg in weight, presumably due to oesophageal involement. Further manifestations were microstomia, xerostomia, arthralgias, Raynaud's phenomenon, cutaneous necrosis, intermittent tachyarrhythmia and the onset of lung fibrosis (by HRCT scan; LFT - TLC 71.1%, residual volume [in percentage of TLC] 182%, steady-state (SS) TLCO 47.3%), but she had normal echocardiography (ejection fraction 60%). Despite pretreatment with prednisolone, methotrexate, mycophenolate mofetile, azathioprine and one course of intravenous cyclophosphamide (cumulative dose 0.7 g/m²), progression of disease continued. During hospitalization before autologous SCT, the patient was treated with prednisolone 30 mg/day. Her Karnofsky score was 40% and her skin score was 32. In this patient steroids were applied due to the rapid progression of the disease.</p></sec><sec><title>Specific antibodies and cytometry</title><p>Disease-related autoantibodies in SLE and SSc were analyzed at admission and regularly during follow up. Monolayers of Hep-2 cells (Bios GmbH, Gr&#x000e4;felfing/Munich, Germany) were used to detect antinuclear antibodies (ANAs) by indirect immunofluorescence. Anti-double-stranded DNA antibodies were identified by indirect immunofluorescence on <italic>Crithidia luciliae</italic> and by enzyme-linked immunosorbent assay as previously described [<xref ref-type="bibr" rid="B4">4</xref>]. Autoantibodies against extractable nuclear antigens (Sm, U1RNP, Ro/SS-A, La/SS-B, Scl-70, Jo-1, centromere) and anticardiolipin antibodies were analyzed using enzyme-linked immunosorbent assay (IMTEC Immundiagnostika GmbH, Zepernick, Germany).</p><p>Antibodies conjugated to phycoerythrin, fluorescein or biotin, and conjugated to peridinin-chlorophyll protein were obtained from Becton Dickinson (Heidelberg, Germany) and Pharmingen (Hamburg, Germany). For cytometry, anti-CD45RO (clone UCHL-1) was coupled to Cy5 (Amersham, Braunschweig, Germany), according to the manufacturer's instructions. Cell staining and flow cytometry were performed using standard protocols on freshly prepared peripheral blood mononuclear cells. The cells were analyzed using a dual-laser, six-parameter FACSCalibur flow cytometer (Becton Dickinson, Heidelberg, Germany); the data were evaluated using commercial software (Becton Dickinson). UCHL-1 (anti-CD45RO) was a generous gift from Imperial Cancer Research Technology (London, UK).</p></sec><sec><title>Stem-cell mobilization and collection</title><p>In all patients mobilization of stem cells was achieved with cyclophosphamide at 2 g/m². After 5 days, G-CSF (10 &#x003bc;g/kg body weight) was administered daily, until harvest of CD34⁺ cells. Leukapheresis was performed when the leucocyte numbers had reached 4.0&#x000d7;10⁹/l. If required, leukapheresis (Cobe Spectra; Cobe BCT, Lake-wood, CO, USA) was repeated until a minimum number of 4&#x000d7;10⁶CD34⁺ cells/kg body weight had been collected for the transplants.</p></sec><sec><title>Engineering of transplants</title><p>Removal of CD34^- leucocytes from the stem-cell transplant was performed by selection for CD34⁺ cells through high-gradient magnetic cell sorting, using a CliniMacs^&#x02122; device (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. If required, CD3⁺ cells from the CD34^-fraction were additionally supplied to the purified CD34⁺cells to transplant a minimum of 1.0&#x000d7;10⁴/kg body weight CD3⁺cells. Until transplantation the CD34⁺ cell suspensions were cryopreserved with 5vol% dimethyl sulphoxide.</p></sec><sec><title>Preparative regimen and autologous stem-cell transplantation</title><p>The preparative regimen consisted of 200 mg cyclophosphamide/kg body weight (days -5 to -2) and ATG (rabbit; obtained from Fresenius, Bad Homburg, Germany) 90 mg/kg body weight (days -4 to -2) [<xref ref-type="bibr" rid="B8">8</xref>]. During ATG treatment 500mg methylprednisone was administered twice a day. The median time interval between cyclophosphamide for mobilization of stem cells and autologous SCT was 38 days (range 29-61 days). Supportive care was provided, according to standard protocols for allogeneic bone marrow transplantation, including isolation of the patient and prophylaxis against infection. Substitution of Igs (10 g every other week) was applied to avoid hypoimmunoglobinaemia, and was ended in all patients after 6 months.</p></sec><sec><title>Evaluation of response</title><p>The function of the organs involved was monitored by technical examinations. Apart from the the clinical course, serological parameters were evaluated (ie ANAs, anti-double-stranded DNA, Scl-70 and other extractable nuclear antigens). Activity of SLE was determined by the ECLAM score [<xref ref-type="bibr" rid="B9">9</xref>]. For SSc the skin score was used [<xref ref-type="bibr" rid="B10">10</xref>]. Therapeutic response was defined as 50% improvement in clinical and serological parameters. Complete remission was defined as normalization without clinical symptoms of disease. The trial fulfilled the EBMT/EULAR guidelines for blood and bone marrow stem-cell transplants in auto-immune diseases [<xref ref-type="bibr" rid="B11">11</xref>].</p></sec></sec><sec><title>Results</title><sec><title>Preparation of stem cells</title><p>For mobilization of stem cells one leukapheresis was sufficient in five out of seven patients; only patients 2 and 3, both with SLE, needed two procedures to collect the number of CD34⁺ cells required for transplant engineering. The median number of CD34⁺ cells collected was 14.1&#x000d7;10⁶CD34⁺ cells/kg body weight (range 4.7-70.0&#x000d7;10⁶ CD34⁺ cells/kg body weight). After enrichment for CD34⁺ cells by using the CliniMacs^&#x02122; device, the transplants contained a median of 6.1&#x000d7;10⁶CD34⁺ cells/kg body weight (range 2.4-7.3&#x000d7;10⁶ CD34⁺ cells/kg body weight). The <italic>ex vivo</italic> purging procedure led to a reduction in contaminating mononuclear cells by 4.5-5 log, resulting in 1.0&#x000d7;10⁴ CD3⁺ cells/kg body weight (range 0.3-1.6&#x000d7;10⁴ CD3⁺cells/kg body weight). Due to this effective strategy, patients 1 and 4 needed a T-cell add-back to provide a dose of 1&#x000d7;10⁴ CD3⁺ cells/kg body weight in the transplant.</p><p>One day after the first application of G-CSF for stem-cell mobilization, severe arthralgias were observed in patient 1 (with polychondritis), suggesting a flare of the autoimmune disease. Patient 3 with SLE showed Raynaud's phenomenon and athralgias after the first day of G-CSF treatment. On day 7, she developed pericardial and pleural effusions, followed by ventricular arrhythmia with a trial fibrillation on day 10 of G-CSF application, presumably related to activation of the autoimmune process. The cardiac condition disappeared under digoxin. Patient 7 with SSc had a reactivation of haemorrhagic oesophagitis, which vanished upon specific treatment. Febrile periods of unknown origin were observed in patients 1 and 7. The symptoms disappeared after the application of high-dose steroids, which could be reduced gradually.</p></sec><sec><title>Immunoablation and reconstitution of the immune system</title><p>During the immunoablative regimen, autologous SCT and haematological reconstitution the median period of hospitalization of the patients was 34 days (range 30-71 days). Reconstitution of granulocytes and platelets occurred rapidly within 2 weeks in all patients. After autologous SCT, the absolute number of nucleated cells reached 1.0&#x000d7;10⁹/l on median day +14 (range day +12 to day +16). The platelet count was 50&#x000d7;10⁹/l on median day +12 (range day +9 to day +15). At day +20 after autologous SCT, patients 5 and 6 (with SSc) showed rapid recovery of up to approximately 0.8&#x000d7;10⁹/l Tlymphocytes or natural killer cells in the peripheral blood. The median number of platelet transfusions applied during bone marrow aplasia was 5 units (range 2-10). A median of 10 units (range 6-17) of red blood cells were administered.</p><p>In the immunoablative phase, systemic inflammatory response syndrome occurred in patients 1-6 during the first infusion of ATG. Severe athralgias and pleural effusion were observed in patient 2; during septicaemia, although she was still in aplasia after autologous SCT, arthritis and abdominal vasculitis were noted. The side effects were interpreted as flares of the autoimmune disease (Table <xref ref-type="table" rid="T1">1</xref>). Remission was achieved by high-dose steroids. For up to 2 months after autologous SCT, the absolute counts of CD4⁺ cells remained below the limit of detection in all patients, and reached pretransplantation levels 4-6 months later (data not shown). Almost all CD4⁺ cells detectable during the second phase of reconstitution (2-5 months after autologous SCT) were CD45RA^-/CD45RO⁺memory/effector cells (Table <xref ref-type="table" rid="T2">2</xref>). The activation marker HLA-DR was expressed on up to 50% of these cells in patient 1, and approximately 20% in patient 2 (data not shown). The transient appearence of activated memory/ effector cells was in concurrence with viral or bacterial infections (interstitial pneumonia, localized infections of the perianal region and of the urinary tract in patient 1, and a <italic>Streptococcus mitis</italic> pneumonia in patient2). Na&#x000ef;ve CD4⁺/CD45RA⁺/CD45RO^- cells were nearly undetectable in patients 1-4 until 6 months after autologous SCT.</p><p>For CD8⁺ lymphocytes, an early transient rapid expansion was observed in patients 2 and 3 within 2 months after autologous SCT. During the follow up of 10 and 6 months, respectively, the absolute numbers of CD8⁺ cells declined in these patients, but recovered later and remained at levels fourfold to 10-fold higher than before autologous SCT (Table <xref ref-type="table" rid="T2">2</xref>). In patient 1 the absolute number of CD8⁺ cells was low during the first 7 months, but had increased fivefold at 1 year after autologous SCT in comparison with the status at admission.</p></sec><sec><title>Clinical outcome after autologous stem-cell transplantation</title><p>With a follow up between 6 and 21 months, the four patients with polychondritis or SLE were in remission, as defined by the disappearance of any clinical symptoms of disease. The physical ability of these patients had improved continuously, as shown by the Karnofsky index and the ECLAM scores (Table <xref ref-type="table" rid="T1">1</xref>). In the SLE patients, the disease-related autoantibody titres (ANAs, anti-double-stranded DNA, cardiolipin) declined to within the normal range. In patient 3 the increased cardiolipin antibody titre was reduced to normal for the first time since August 1996 and remained low to the last date of follow up (February 2000). Despite the withdrawal of propanolol and the reduction in prednisolone dose, ventricular arrythmias were no longer observed. In patient 4, who was suffering from nephrotic syndrome when admitted, the proteinuria improved dramatically after autologous SCT. In addition, patient 4 had a fresh deep venous thrombosis of the leg combined with symptomatic pulmonary embolism at admission, which improved after autologous SCT; anticoagulation therapy has been continued for safety reasons. It was possible to reduce the application of corticosteroids gradually in all patients.</p><p>Patients 5 and 6 (with SSc) had neither clinical nor serological improvement (Table <xref ref-type="table" rid="T3">3</xref>), although progression of the disease was not observable. In the LFT no major alterations were detectable, and the skin score remained stable. In both patients, Raynaud's phenomenon improved after autologous SCT only in warm climates. Against medical advice patient 5 became pregnant during reconstitution of the immune system and gave birth to a healthy child 14 months after autologous SCT. In patient 7 (with SSc), clinical and serological progression (weight loss of 2 kg or 5% of body weight and a fourfold increase in ANA titres) was observed after moderate-dose cyclophosphamide and G-CSF for stem-cell mobilization. During immunoablation, fluid retention led to a weight gain of 7 kg within 1 week and to the occurrence of plural effusions. Transiently, she was stablilized at a significantly reduced level of performance status before the onset of ventricular tachycardia and subsequent electromechanical uncoupling. She died on day +2 after autologous SCT due to cardiac failure, although no signs of cyclophosphamide-induced cardiotoxicity were observed during autopsy. The postmortem histology revealed an advanced stage of pulmonary fibrosis, with all the signs of cor pulmonale.</p></sec></sec><sec><title>Discussion</title><p>The present study was conducted to evaluate the efficacy of autologous SCT in treatment-refractory autoimmune disease. After a median follow up of 14 months (range10-21 months) four patients with polychondritis or SLE are in clinical remission. Following the immunoablative regimen, including cyclophosphamide, ATG and steroids, the disease-specific titres of autoantibodies became negative in patients 2-4, who had SLE. No symptoms of recurrent autoimmune reactivity have been detected in these patients. Obviously, the rigorous regimen of immunoablation, consisting of high-dose cyclophosphamide in combination with ATG, was successful in achieving rather complete aplasia. This combination increased the efficacy of immunoablation in comparison with that achieved by other studies [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>], by grossly reducing the number of autoreactive immune cells. Plasma cells can be found resting in bone marrow for more than 90 days [<xref ref-type="bibr" rid="B14">14</xref>], and apparently were eliminated in the patients with SLE in the present study under treatment with ATG, presumably by the recognition of specific surface antigens by immunoglobulins of the ATG.</p><p>Side effects during immunoablation may be severe, and flares of the autoimmune disease are particularly detrimental to the patient's condition. Comparable with a previously reported observation of the induction of flares by G-CSF in a patient with rheumatoid arthritis [<xref ref-type="bibr" rid="B15">15</xref>], in the present trial flares of SLE were diagnosed in patient 2 who was in aplasia after the immunoablative regimen (Table <xref ref-type="table" rid="T1">1</xref>). Flares may be attributed to `cytokine-primed' clinical situations in which G-CSF is released from macrophages and residual lymphocytes during bone marrow aplasia. Flares can be controlled by high-dose steroids. After gradual reduction in the dose of steroids and without further treatment of patients 1-4, recurrence of the flares was not observed during the follow up of more than 12 months. The flares that occurred in patient 2, in aplasia and during septicaemia, may represent a `burn-out' of autoreactivity in terms of an exhaustion of mediators participating in autoimmunity. In order to avoid the induction of flares, G-CSF was not applied after termination of the immunoablative regimen and during the phase of haematological reconstitution.</p><p>In a previous study in five patients with haematological malignancies or solid tumours and concomitant refractory autoimmune disease, who were treated by autologous SCT [<xref ref-type="bibr" rid="B13">13</xref>], the disease persisted or relapsed within 3 months. In that investigation the transplantations were performed without <italic>in vivo</italic>/<italic>ex vivo</italic> depletion of T cells, emphasizing the importance of effective immunoablation. Successful immunoablation <italic>in vivo</italic>, combined with less intensive purging of the transplant (2-3 log depletion of CD3⁺ cells) was reported to halt disease progression in patients with multiple sclerosis, rheumatoid arthritis or SLE [<xref ref-type="bibr" rid="B16">16</xref>]. The further reduction in CD3⁺ cells, as applied in the present study, but not in the previous investigation [<xref ref-type="bibr" rid="B16">16</xref>], did not lead to a higher incidence of severe infections. Thus, it is tempting to speculate that in future studies even lower amounts than 1&#x000d7;10⁴ CD3⁺ cells/kg body weight, as applied in the present study, may be tolerated.</p><p>In the present trial, preparations with greatly reduced numbers of CD3⁺cells were used for autologous SCT. By application of the recently developed CliniMACS^&#x02122; device [<xref ref-type="bibr" rid="B7">7</xref>], a highly efficient technology was introduced for the selection of haematopoietic stem cells. High-gradient magnetic cell sorting was able to purify CD34⁺ cells effectively from G-CSF-mobilized peripheral blood, resulting in 4.5-5 log depletions of CD3⁺ T cells, thus minimizing the risk of retransplantation of autoreactive T cells. In fact, after autologous SCT no relapse of autoreactivity was observed in the patients with polychondritis or those with SLE, even though the absolute counts of CD4⁺, CD8⁺ and all other leucocytes had reached pretreatment levels. On the other hand, the large-scale depletion of T cells from the transplants did not provoke life-threatening infections before and during immunological reconstitution. The relevance of the <italic>in vivo</italic>/<italic>ex vivo</italic> depletion procedure used in our investigation may be confirmed by controlled studies.</p><p>Stem-cell support is essential in shortening the duration of aplasia and in the reconstitution of haemopoiesis after immunoablation by a regimen of cyclophosphamide and ATG. Any immunoablative treatment without subsequent autologous SCT is associated with the risk of severe infections during neutropenia and thrombopenic bleeding. This appears to be in contrast to the results of a recent approach with high-dose cyclophosphamide followed by G-CSF but without autologous SCT [<xref ref-type="bibr" rid="B12">12</xref>], in which two patients with SLE had follow-up periods of 12 or 14 months with complete or partial remission. The results appear to be due to the G-CSF-induced priming after high-dose cyclophosphamide. In patients with aplastic anaemia treated with allogeneic bone marrow transplantation the advantage of a combined immunosuppression with cyclophosphamide and ATG was emphasized by a significant reduction in graft rejections [<xref ref-type="bibr" rid="B8">8</xref>]. The present data on patients with polychondritis or SLE are in accord with the recent results achieved by high-stringency immunoablation followed by autologous SCT in a panel of 10 patients suffering from multiple sclerosis, rheumatoid arthritis or SLE [<xref ref-type="bibr" rid="B16">16</xref>]. With follow-up periods of 6 and 12 months in that study the two patients with SLE were in remission at the time of publication.</p><p>In the present investigation, a rapid decrease in levels of pathological autoantibodies to normal values was observed in the SLE patients responding to autologous SCT. The early phase in the reconstitution of the immune system was marked by rapid recovery of granulocytes and platelets in all patients. During the second phase of reconstitution CD4⁺ cells, exclusively of the antigen-experienced memory/effector type, were observed 2-5 months after autologous SCT (CD45RO⁺, CD45RA^-; Table <xref ref-type="table" rid="T2">2</xref>). Similar kinetics of reconstitution were described in patients after allogeneic bone marrow transplantation for haematological malignancies [<xref ref-type="bibr" rid="B17">17</xref>]. In the patients we studied, the activated T-helper cells may reflect the clonal expansion of persisting cells after the preparative regimen, which might have been stimulated by minor infections during reconstitution [<xref ref-type="bibr" rid="B18">18</xref>].</p><p>During the second phase of reconstitution, patient 2 in the present study suffered an episode of varizella-zoster infection. After <italic>in vitro</italic> incubation of mononuclear cells of that patient with varizella-zoster virus antigen, the secretion of IFN-&#x003b3; by a subpopulation of T cells was observed (data not shown). This may exclude a persisting general deficiency of the immune system due to the aggressive immunoablation. Deficiency of the immune system was considered as a basic reason for self-tolerance in autoimmune disease after immunosuppression followed by autologous SCT [<xref ref-type="bibr" rid="B16">16</xref>].</p><p>The two patients with SSc who were evaluable for follow up showed no clinical and serological responses at 6 or 13 months after autologous SCT. We presume that insufficient immunoablation may have a role in the treatment failures. This is supported by the persistance of Scl-70 autoantibody, and by the early recovery of lymphocytes (patient 5) and natural killer cell reconstitution (patient 6). In SSc no correlation between the activity of the disease and the presence of autoantibodies has been shown, although serum Scl-70 is associated with poor prognosis with regard to pulmonary or cardiac involvement [<xref ref-type="bibr" rid="B19">19</xref>]. Patients 5 and 6 were still positive for Scl-70 after autologous SCT, suggesting a need for intensification of treatment. However, the reason for insufficient immunoablation in the SSc patients is not clear, and may be related to the underlying pathophysiology that is different from that in SLE. SSc appears to be less responsive, at least at the advanced stage of tissue destruction, due to fibrotic processes that are not present in SLE.</p><p>The present results are in contrast to those of a previous report of a significant decline in ANAs in a SSc patient within 6 months after autologous SCT [<xref ref-type="bibr" rid="B20">20</xref>]. The stable disease in patient 5 in the present study was accompanied by a pregnancy during the follow-up period. Pregnancy in patients with autoimmune disease has been postulated to be a reason for stability in SSc and multiple sclerosis [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. Patient 7 of the present trial, who had a brief history of SSc, died 2 days after autologous SCT from cardiac failure due to massive pulmonary fibrosis. After 2 g/m² cyclophosphamide was administered for stem-cell mobilization, she developed clinical and serological progress until autologous SCT. In hindsight, the advanced stage of pulmonary fibrosis was not foreseeable, and is to be considered the cause for the treatment-associated mortality.</p><p>In conclusion, the present study demonstrates that the induction of immune tolerance for disease-related antigens is feasible and can be achieved with immunoablation and subsequent autologous SCT in the case of refractory polychondritis and SLE. Effective <italic>ex vivo</italic> depletion of CD34^-cells can be achieved with state-of-the-art technologies, and appears to be essential for sustained tolerance after immunological reconstitution. These results show unambiguously that a `reset' of the immune system was brought about, which was able to deal successfully with pathogens. The treatment consisted of one admission into hospital for stem-cell mobilization, and another one for immunoablation and autologous SCT, with median durations of hospitalization of 20 and 34 days, respectively. The high costs of the complex and intensive therapy performed in the present study may be acceptable when compared with those of disease-related long-term hospitalization and invalidity.</p></sec>

The role of X-chromosome inactivation in female predisposition to autoimmunity

<sec><title>Introduction:</title><p>A reduction in the sex ratio (male : female) is characteristic of most autoimmune disorders. The increased prevalence in females ranges from a modest 2:1 for multiple sclerosis [<xref ref-type="bibr" rid="B1">1</xref>], to approximately 10:1 for systemic lupus erythematosus [<xref ref-type="bibr" rid="B2">2</xref>]. This tendency toward autoimmunity in females is often ascribed to hormonal differences, because in a number of experimental disease models estrogens exacerbated disease, and androgens can inhibit disease activity [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. However, human studies have failed to demonstrate a clear-cut influence of hormonal environment on disease susceptibility to lupus or other autoimmune disorders. In addition, many childhood forms of autoimmunity, such as juvenile rheumatoid arthritis, exhibit female predominance [<xref ref-type="bibr" rid="B5">5</xref>]. Interestingly, juvenile (type 1) diabetes is an exception to this general trend, with a sex ratio close to 1 in most studies [<xref ref-type="bibr" rid="B6">6</xref>]. Therefore, it is reasonable to consider alternative explanations for the increased prevalence of autoimmune diseases in human females.</p><p>A unifying feature of autoimmune disorders appears to be the loss of immunologic tolerance to self-antigens, and in many of these diseases there is evidence that T-cell tolerance has been broken. The most profound form of T-cell tolerance involves deletion of potentially self-reactive T cells during thymic selection. Thus, lack of exposure to a self-antigen in the thymus may lead to the presence of autoreactive T cells and may increase the risk of autoimmunity. An elegant example of this has recently been reported [<xref ref-type="bibr" rid="B7">7</xref>].</p><p>The existence of X-chromosome inactivation in females offers a potential mechanism whereby X-linked self-antigens may escape presentation in the thymus or in other peripheral sites that are involved in tolerance induction. Early in female development, one of the two X chromosomes in each cell undergoes an ordered process of inactivation, with subsequent silencing of most genes on the inactive X chromosome [<xref ref-type="bibr" rid="B8">8</xref>]. This phenomenon occurs at a very early embryonic stage [<xref ref-type="bibr" rid="B9">9</xref>], and thus all females are mosaic and may occasionally exhibit extreme skewing towards one or the other parental X chromosome. In theory, this may result in a situation in which polymorphic self-antigens on one X chromosome may fail to be expressed at sufficiently high levels in a tolerizing compartment, such as the thymus, and yet may be expressed at a considerable frequency in the peripheral soma. Thus, females may be predisposed to a situation in which they can occasionally express X-linked autoantigens in the periphery to which they have been inefficiently tolerized. Stewart [<xref ref-type="bibr" rid="B10">10</xref>] has recently speculated that such a mechanism may play a role in the predisposition to systemic lupus.</p><p>This hypothesis predicts that females with autoimmunity may be particularly prone to this mechanism of `inadequate tolerization' by virtue of extremely skewed X-chromosome inactivation. We therefore performed a comprehensive analysis of X-chromosome inactivation patterns in populations of females with multiple sclerosis, systemic lupus erythematosus, juvenile rheumatoid arthritis, and type 1 (insulin-dependent) diabetes mellitus, and in female control individuals. The results do not provide support for a major role for skewed X-chromosome inactivation in female predisposition to autoimmunity; however, neither is the underlying hypothesis disproved by the present data.</p></sec><sec><title>Materials and method:</title><p>DNA was obtained from female patients from the following sources: 45 persons with juvenile diabetes seen at the Virginia Mason Research Center in Seattle, Washington; 58 multiple sclerosis patients seen at the New York Hospital Multiple Sclerosis Center; 46 patients with systemic lupus erythematosus seen at the Hospital for Special Surgery (New York); 18 patients with juvenile rheumatoid arthritis seen at the Children's Hospital Medical Center in Cleveland. In addition, 30 healthy age-matched females were studied as normal controls.</p><p>Employing a modification of previously described methods [<xref ref-type="bibr" rid="B11">11</xref>], we utilized a fluorescent Hpa II/PCR assay of the androgen receptor (AR) locus to assess X-chromosome inactivation patterns. The AR gene contains a polymorphic CAG repeat, which is flanked by Hpa II sites. These Hpa II sites are methylated on the inactive X chromosome, and are unmethylated on the active X chromosome. By performing PCR amplification across this region after cutting with the methylation-sensitive enzyme Hpa II, the relative amounts of the methylated AR alleles can be quantitatively determined with a high degree of accuracy; variance on repeated assays is approximately 4% [<xref ref-type="bibr" rid="B12">12</xref>].</p><p>Skewing of X-chromosome inactivation is expressed as percentage deviation from equal (50:50) inactivation of the upper and lower AR alleles. Therefore, the maximal possible deviation is 50%, in which case all of the X chromosomes bearing one of the AR alleles are inactivated.</p></sec><sec><title>Results:</title><p>We examined X-chromosome inactivation patterns in several different populations. The results are summarized in Fig. <xref ref-type="fig" rid="F1">1</xref>. A wide range of X-inactivation skewing was observed in all five groups. Approximately 5% (nine out of 197) of individuals exhibited extreme skewing (greater than 40% deviation from a 50:50 distribution). However, there was no difference between the groups, either in the overall mean skewing, or in the fraction of individuals with extreme skewing (&#x0003e;40%).</p><p>Although the present study was not initiated in order to examine allelic variation in the AR gene <italic>per se</italic>, the data provide an opportunity to address this question. Excessively long CAG repeats in the AR are a rare cause of spinal-bulbar muscular atrophy [<xref ref-type="bibr" rid="B13">13</xref>], and AR repeat length appears to have an influence on the biology of certain tumors [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. In this context, it has been shown that transcription of AR correlates inversely with repeat length [<xref ref-type="bibr" rid="B16">16</xref>]. We therefore compared AR repeat length in control individuals and patients with autoimmunity. No differences were observed for mean repeat length, or for maximum and minimum repeat length, among the five groups. </p></sec><sec><title>Discussion:</title><p>The reason for the female predominance in most autoimmune diseases remains obscure. The present study was initiated in order to address the hypothesis that a nonhormonal mechanism related to X inactivation might be involved. The hypothesis rests on the idea that skewing of X inactivation might lead to a deficiency of tolerance induction in the thymus, particularly with respect to polymorphic X-linked autoantigens. The hypothesis predicts that skewed X inactivation would be more prevalent in females with autoimmune diseases than in female control individuals. This was not observed.</p><p>Nevertheless, these negative data do not rule out a role for X inactivation in female predisposition to loss of tolerance. A general model for how this mechanism might operate is shown in Fig. <xref ref-type="fig" rid="F2">2</xref>. Thymocytes undergo selection in the thymic parenchyma and, in the case of negative selection, the selecting elements appear to be derived from the bone marrow and consist mainly of thymic dendritic cells. If the thymic dendritic cell population exhibits random X inactivation, it is highly likely that differentiating thymocytes will contact dendritic cells that express self-antigens on both X chromosomes. This situation is outlined schematically on the left side of Fig. <xref ref-type="fig" rid="F2">2</xref>. However, if there is extremely skewed X inactivation in the thymic dendritic cell population, a particular thymocyte might not come into contact with dendritic cells that express one of the two X chormosomes. This would lead to a situation where T cells may undergo thymic maturation without having been negatively selected for antigens that are expressed on the predominantly inactive X chromosome. This situation is shown on the right side of Fig. <xref ref-type="fig" rid="F2">2</xref>.</p><p>In order for this mechanism to be physiologically relevant, some assumptions must be made. First, defective tolerance from skewed X inactivation should only be directed at X-linked antigens that are polymorphic, and for which the individual is heterozygous. Thus, this mechanism would not be expected to lead to lack of tolerance commonly, unless there are at least several highly polymorphic X-linked autoantigens in the population that are involved in thymic deletion events. Second, if this actually leads to autoimmunity, it also predicts that the initial break in tolerance that leads to disease should involve an X-linked autoantigen that is expressed in a peripheral nontolerizing site or circumstance.</p><p>A recent report [<xref ref-type="bibr" rid="B7">7</xref>] has elegantly demonstrated the importance of thymic deletion events in predisposition to autoimmune disease. The proteolipid protein (PLP) autoantigen is expressed in alternatively spliced forms, which exhibit tissue specific expression. A nonspliced variant is expressed in peripheral neural tissue. However, in the thymus a splice variant results in the lack of thymic expression of an immunodominant peptide. This results in loss of tolerace of T cells to this peptide, presumably on the basis of lack of thymic deletion of thymocytes that are reactive with this antigen. Interestingly, PLP is encoded on the X chromsome. However, there is no evidence that genetic polymorphisms control the level splicing of PLP within the thymus. Nevertheless, these data illustrate the potential importance of deficiencies in thymic deletion for autoimmune T-cell reactivity.</p><p>The present results suggest that if skewed X inactivation is relevant to thymic tolerance induction, then the effect does not depend on global skewing of X-chromosome inactivation, at least in the hematopoietic compartment. In this study we examined X-inactivation patterns in peripheral blood mononuclear cells, and the results should reflect the state of X inactivation in all mesenchymal tissues, including dendritic cells. X inactivation occurs at a very early time point in development, and thus the results in one tissue should reflect the general situation in the rest of the body. However, there may be exceptions to this. We have occasionally observed differences in X-inactivation patterns between buccal mucosa (an ectodermally derived tissue) and peripheral blood in the same individiual (unpublished observations). This could be a chance event, or it may result from selection for certain X-linked alleles during embryonic development, as has been described in carriers of X-linked immunodeficiencies [<xref ref-type="bibr" rid="B17">17</xref>].</p><p>Another consideration is that certain tissue microenvironments may be derived from very small numbers of founder cells, and thus may exhibit skewed utilization of one or the other X chromosome, even if the tissue as a whole is not skewed. This situation could vary over time. Thus, there may be time points at which certain thymic microenvironments are populated by dendritic cells that, for stochastic reasons, all utilize the same X chromosome. This would create a `window of opportunity' in which a given thymocyte, in a given selecting location, could escape negative selection by antigens on the inactive X chromosome. The likelihood of this happening would obviously depend on the number of dendritic cells that are usually contacted by a thymocyte during thymic selection. There is limited information on this point, although Stewart [<xref ref-type="bibr" rid="B10">10</xref>] has theorized that this number may be as low as 15. If this is the case, then escape from thymic deletion may still occur in females who are heterozygous for a relevant X-linked antigen, even if the hematopoietic cells in general do not exhibit extreme skewing.</p><p>In conclusion, we suggest that X-chromosome inactivation needs to be considered as a potential factor in the predominance of females in most autoimmune diseases. Our inability to show an increase in X-chromosome skewing in females with autoimmunity does not eliminate this as an etiologic contributor to loss of immunologic tolerance. Future experiments must be directed at a detailed analysis of tissue patterns of X inactivation, as well as at a search for potential X-linked autoantigens.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Chitnis</surname><given-names>Smita</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>peterg@nshs.edu</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Monteiro</surname><given-names>Joanita</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Glass</surname><given-names>David</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Apatoff</surname><given-names>Brian</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Salmon</surname><given-names>Jane</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Concannon</surname><given-names>Patrick</given-names></name><xref ref-type="aff" rid="I5">5</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Gregersen</surname><given-names>Peter K</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>A reduction in the sex ratio (male : female) is characteristic of most autoimmune disorders. The increased prevalence in females ranges from a modest 2:1 for multiple sclerosis [<xref ref-type="bibr" rid="B1">1</xref>], to approximately 10:1 for systemic lupus erythematosus [<xref ref-type="bibr" rid="B2">2</xref>]. This tendency towards autoimmunity in females is often ascribed to hormonal differences because, in a number of experimental disease models, estrogens exacerbate disease and androgens can inhibit disease activity [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>]. However, human studies have failed to demonstrate a clear-cut influence of hormonal environment on disease susceptibility to lupus or other autoimmune disorders. In addition, many childhood forms of autoimmunity, such as juvenile rheumatoid arthritis, exhibit female predominance [<xref ref-type="bibr" rid="B5">5</xref>]. Interestingly, juvenile (type 1) diabetes is an exception to this general trend, with a sex ratio close to 1 in most studies [<xref ref-type="bibr" rid="B6">6</xref>]. Therefore, it is reasonable to consider alternative explanations for the increased prevalence of autoimmune diseases in human females.</p><p>A unifying feature of autoimmune disorders appears to be the loss of immunologic tolerance to self-antigens, and in many of these diseases there is evidence that T-cell tolerance has been broken. The most profound form of T-cell tolerance involves deletion of potentially self-reactive T cells during thymic selection. Thus, lack of exposure to a self-antigen in the thymus may lead to the presence of autoreactive T cells and increase the risk of autoimmunity. An elegant example of this has recently been reported [<xref ref-type="bibr" rid="B7">7</xref>].</p><p>The existence of X-chromosome inactivation in females offers a potential mechanism whereby X-linked self-antigens may escape presentation in the thymus or in other peripheral sites that are involved in tolerance induction. Early in female development, one of the two X chromosomes in each cell undergoes an ordered process of inactivation, with subsequent silencing of most genes on the inactive X chromosome [<xref ref-type="bibr" rid="B8">8</xref>]. This phenomenon occurs at a very early embryonic stage [<xref ref-type="bibr" rid="B9">9</xref>], and thus all females are mosaic and may occasionally exhibit extreme skewing towards one or the other parental X chromosome. In theory, this may result in a situation in which polymorphic self-antigens on one X chromosome may fail to be expressed at sufficiently high levels in a tolerizing compartment, such as the thymus, and yet may be expressed at considerable frequency in the peripheral soma. Thus, females may be predisposed to a situation in which they can occasionally express X-linked autoantigens in the periphery to which they have been inefficiently tolerized. Stewart [<xref ref-type="bibr" rid="B10">10</xref>] has recently speculated that such a mechanism may play a role in the predisposition to systemic lupus.</p><p>This hypothesis predicts that females with autoimmunity may be particularly prone to this mechanism of `inadequate tolerization' by virtue of extremely skewed X-chromosome inactivation. We therefore performed a comprehensive analysis of X-chromosome inactivation patterns in populations of females with multiple sclerosis, systemic lupus, juvenile rheumatoid arthritis, and type 1 (insulin-dependent) diabetes mellitus, and in female control individuals. The results do not provide support for a major role for skewed X-chromosome inactivation in female predisposition to autoimmunity; however, neither is the underlying hypothesis disproved by our data.</p></sec><sec><title>Materials and method</title><sec><title>Subjects</title><p>DNA was obtained from female patients from the following sources: 45 persons with juvenile diabetes seen at the Virginia Mason Research Center in Seattle, Washington; 58 multiple sclerosis patients seen at the New York Hospital Multiple Sclerosis Center; 46 patients with systemic lupus erythematosus seen at the Hospital for Special Surgery (New York); 18 patients with juvenile rheumatoid arthritis seen at the Children's Hospital Medical Center in Cleveland. In addition, 30 healthy age-matched females were studied as normal controls.</p></sec><sec><title>Hpa II/polymerase chain reaction assay for X-chromosome inactivation</title><p>A modification to a previously described assay for AR methylation [<xref ref-type="bibr" rid="B11">11</xref>] was used for these studies. In the normal females the AR gene is methylated on the inactive X chromosome and is undermethylated on the active X chromosome. Furthermore, the presence of a highly polymorphic triplet repeat within the AR gene allows for the discrimination of each X chromosome in most female subjects (Fig. <xref ref-type="fig" rid="F3">3</xref>). Thus, allele-specific methylation patterns can be distinguished using this gene.</p><p>Genomic DNA (approximately 200 ng) was digested overnight using a methylation-sensitive restriction enzyme HpaII, as per the manufacturer's instructions (Gibco BRL, Rockville, MD, USA). A mock sample as control was prepared simultaneously without the HpaII enzyme. After the digestion the samples were boiled for 10 min to inactivate the enzyme. PCR amplification was performed, with final concentration of the digest equivalent to approximately 100 ng of the DNA, 1 &#x000d7; PCR bufferII (Perkin-Elmer, Foster City, CA, USA), 2.5 mmol/l MgCl<sub>2</sub>, and amplitaq gold 1.25 units.</p><p>Primers used were AR1 (5'-TCCAGAATCTGTTCCA-GAGCGTGC-3') and AR2 (5'-GCTGTGAAGGTTGCT-GTTCCTCAT-3'). These primers flank the triple repeat CAG and the HpaII sites in the first exon of the AR gene (Fig. <xref ref-type="fig" rid="F3">3</xref>). The AR1 primer was fluorescence labeled with 6-Fam and TET. The 6-Fam-AR1 and the TET-AR1 were used for mock and HpaII-digested samples, respectively. This combination allowed simultaneous analyses of the PCR products of the HpaII- and mock-treated samples using the ABI310 Genetic Analyzer (Perkin-Elmer). The final concentration of the AR1 primer was 3 pmol 6-Fam-AR1+7 pmol unlabelled AR1 and 10 pmol of unlabeled AR2 in a 50 &#x003bc;l reaction.</p><p>The cycling conditions were as follows: denaturation at 95&#x000b0;C/12 min to activate the amplitaq gold, followed by 35 cycles of denaturation at 95&#x000b0;C/45 s, annealing at 60&#x000b0;C/30 s and extension at 72&#x000b0;C/30 s. After 35 cycles, the final extension was done at 72&#x000b0;C/10 min. The PCR products of each sample (mock and HpaII) were diluted 1:10 in the same vial. One microliter of the diluted product was then added to a mixture containing 12 &#x003bc;l deionized formamide and 0.5 &#x003bc;l molecular weight standard GS 350 TAMRA.The vials were denatured at 95&#x000b0;C/5 min, cooled on ice, and resolved using a 310 Genetic Analyzer (Perkin-Elmer).</p></sec><sec><title>Calculations for percentage of skewing of X inactivation</title><p>Two peaks corresponding to two alleles were obtained from the 310 Genetic Analyzer for each mock- and HpaII-treated sample. The relative intensity of the larger AR allele (higher molecular weight, peak 2) with respect to the smaller AR allele (low molecular weight, peak 1) was calculated and expressed as the ratio R (peak 2 area/peak 1 area). The ratios in the mock-digested (R<sub>M</sub>) and HpaII-digested (R<sub>H</sub>) samples were calculated separately, and were then averaged for each of the two sets of duplicate samples. For each individual, a normalized ratio (R<sub>N</sub> = R<sub>H</sub>/R<sub>M</sub>) was calculated to correct for occasional minor variation in efficiency of amplification of the two AR alleles. This normalized ratio was used to determine the percentage of inactivation of the X chromosome bearing the larger AR allele: percentage inactivation = [R<sub>N</sub>/(R<sub>N</sub> + 1)] &#x000d7; 100. The degree of skewing was calculated by subtracting 50 from the observed degree of inactivation.</p></sec></sec><sec><title>Results</title><sec><title>Fluorescent Hpa II/polymerase chain reaction assay for X-chromosome inactivation</title><p>Employing a modification of previously described methods [<xref ref-type="bibr" rid="B11">11</xref>], we utilized a fluorescent Hpa II/PCR assay of the AR locus to assess X-chromosome inactivation patterns. As shown in Fig. <xref ref-type="fig" rid="F3">3</xref>, the AR gene contains a polymorphic CAG repeat that is flanked by Hpa II sites. These Hpa II sites are methylated on the inactive X chromosome, and are unmethylated on the active X chromosome. By performing PCR amplification across this region after cutting with the methylation-sensitive enzyme Hpa II, the relative amounts of the methylated AR alleles can be quantitatively determined with a high degree of accuracy; variance on repeated assays is approximately 4% [<xref ref-type="bibr" rid="B12">12</xref>].</p><p>Two examples of this assay are shown in Fig. <xref ref-type="fig" rid="F4">4</xref>. In Fig. <xref ref-type="fig" rid="F4">4a</xref>, an individual with equivalent methylation of both AR alleles is shown. In this case, the relative peak intensity is equivalent in the mock-digested and Hpa II-digested samples. In Fig. <xref ref-type="fig" rid="F4">4b</xref>, an individual with extremely skewed X inactivation is shown, in whom the lower allele (271 bp) is relatively over-methylated compared with the upper (284 bp) allele. Thus, the X chromosome bearing the lower allele is preferentially inactivated. As described in the Materials and method section, all assays were normalized to the mock results and performed in duplicate. Skewing of X-chromosome inactivation is expressed as a percentage deviation from equal (50:50) inactivation of the upper and lower AR alleles. Therefore, the maximal possible deviation is 50%, in which case all of the X chromosome bearing one of the AR alleles are inactivated.</p></sec><sec><title>Comparison of X-chromosome inactivation patterns in normal females and females with autoimmune diseases</title><p>We examined X-chromosome inactivation patterns in several different populations. The results are summarized in Fig. <xref ref-type="fig" rid="F1">1</xref>. A wide range of X-inactivation skewing was observed in all five groups. Approximately 5% (nine out of 197) of individuals exhibited extreme skewing (greater than 40% deviation from a 50:50 distribution). However, there was no difference between the groups, either in the overall mean skewing, or in the fraction of individuals with extreme skewing (&#x0003e;40%).</p></sec><sec><title>Androgen receptor gene allele size comparisons among patients and controls</title><p>Although the present study was not initiated to examine allelic variation in the AR gene <italic>per se</italic>, the data provide an opportunity to address this question. Excessively long CAG repeats in the AR gene are a rare cause of spinal-bulbar muscular atrophy [<xref ref-type="bibr" rid="B13">13</xref>], and AR repeat length appears to have an influence on the biology of certain tumors [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. In this context, it has been shown [<xref ref-type="bibr" rid="B16">16</xref>] that transcription of the AR gene correlates inversely with repeat length.We therefore compared AR repeat length in control individuals and patients with autoimmunity. No differences were observed for mean repeat length, or for maximum and minimum repeat length, among the five groups.</p></sec></sec><sec><title>Discussion</title><p>The reason for the female predominance in most autoimmune diseases remains obscure. The present study was initiated in order to address the hypothesis that a nonhormonal mechanism related to X inactivation might be involved. The hypothesis rests on the idea that skewing of X inactivation might lead to a deficiency of tolerance induction in the thymus, particularly with respect to polymorphic X-linked autoantigens. The hypothesis predicts that skewed X inactivation would be more prevalent in females with autoimmune diseases than in female control individuals. This was not observed.</p><p>Nevertheless, these negative data do not rule out a role for X inactivation in female predisposition to loss of tolerance. A general model for how this mechanism might operate is shown in Fig. <xref ref-type="fig" rid="F1">1</xref>. Thymocytes undergo selection in the thymic parenchyma, and in the case of negative selection, the selecting elements appear to be derived from the bone marrow and consist mainly of thymic dendritic cells. If the thymic dendritic cell population exhibits random X inactivation, it is highly likely that differentiating thymocytes will contact dendritic cells that express self-antigens on both X chromosomes. This situation is outlined schematically on the left side of Fig. <xref ref-type="fig" rid="F2">2</xref>. However, if there is extremely skewed X inactivation in the thymic dendritic cell population, a particular thymocyte may not come into contact with dendritic cells that express one of the two X chromosomes. This would lead to a situation where T cells may undergo thymic maturation without having been negatively selected for antigens that are expressed on the predominantly inactive X chromosome. This situation is shown on the right side of Fig. <xref ref-type="fig" rid="F2">2</xref>.</p><p>In order for this mechanism to be physiologically relevant, several assumptions must be made. First, defective tolerance from skewed X inactivation should only be directed at X-linked antigens that are polymorphic, and for which the individual is heterozygous. Thus, this mechanism would not be expected to commonly lead to lack of tolerance unless there are at least several highly polymorphic X-linked autoantigens in the population that are involved in thymic deletion events. Second, if this actually leads to autoimmunity, it also predicts that the initial break in tolerance that leads to disease should involve an X-linked autoantigen that is expressed in a peripheral nontolerizing site or circumstance.</p><p>A recent report [<xref ref-type="bibr" rid="B7">7</xref>] elegantly demonstrated the importance of thymic deletion events in predisposing to autoimmune disease. The PLP autoantigen is expressed in alternatively spliced forms, which exhibit tissue-specific expression. A nonspliced variant is expressed in periperal neural tissue. However, in the thymus a splice variant results in the lack of thymic expression of an immunodominant peptide. This results in loss of tolerace of T cells to this peptide, presumably on the basis of lack of thymic deletion of thymocytes that are reactive with this antigen. Interestingly, PLP is encoded on the X chromsome. However, there is no evidence that genetic polymorphisms control the level splicing of PLP within the thymus. Nevertheless, these data illustrate the potential importance of deficiencies in thymic deletion for autoimmune T-cell reactivity.</p><p>The present results suggest that if skewed X inactivation is relevant to induction of thymic tolerance, then the effect does not depend on global skewing of X-chromosome inactivation, at least in the hematopoietic compartment. In this study we examined X-inactivation patterns in peripheral blood mononuclear cells, and the results should reflect the state of X inactivation in all mesenchymal tissues, including dendritic cells. X inactivation occurs at a very early time point in development, and thus the results in one tissue should reflect the general situation in the rest of the body. However, there may be exceptions to this. We have occasionally observed differences in X inactivation patterns between buccal mucosa (an ectodermally derived tissue) and peripheral blood (mesoderm) in the same individual (unpublished observations). This could be a chance event, or may result from selection for certain X-linked alleles during embryonic development, as has been described in carriers of X-linked immunodeficiencies [<xref ref-type="bibr" rid="B17">17</xref>]. However, in general, subsets of cells within the hematopoietic compartment do not display differences in X-inactivation patterns (unpublished observations).</p><p>Another consideration is that certain tissue microenvironments may be derived from very small numbers of founder cells, and thus may exhibit skewed utilization of one or the other X chromosome, even if the tissue as a whole is not skewed. This situation could vary over time. Thus, there may be time points at which certain thymic microenvironments are populated by dendritic cells that, for stochastic reasons, all utilize the same X chromosome. This would create a `window of opportunity' in which a given thymocyte, in a given selecting location, could escape negative selection by antigens on the inactive X chromosome. The likelihood of this happening would obviously depend on the number of dendritic cells that are usually contacted by a thymocyte during thymic selection. There is limited information on this point, although Stewart [<xref ref-type="bibr" rid="B10">10</xref>] has theorized that this number may be as low as 15. If this is the case, escape from thymic deletion may still occur in females who are heterozygous for a relevant X-linked antigen, even if the hematopoietic cells in general do not exhibit extreme skewing.</p><p>If the latter scenario is operative, it will be extremely difficult to document it by studying X-inactivation patterns. One might examine X-inactivation skewing specifically in thymic dendritic populations, but if the effect is at the level of microenvironment and varies over time, then it will not be possible to detect this using methods directed at large cell populations, as we have done here. Conceivably, examination of thymic tissue sections could provide support for the hypothesis. Another aspect of this is that some thymic deletion events appear to be mediated by thymic epithelial cells at the cortico-medullary junction. Interestingly, thymic epithelial cells appear to be derived from very few founder cells [<xref ref-type="bibr" rid="B18">18</xref>], and thus should exhibit a rather large degree of `patchiness' with respect to X inactivation. This again might lead to local epithelial cell microenvironments that fail to delete for X-linked autoantigens.</p><p>It has recently become apparent that mechanisms of peripheral tolerance also exist, and we have considered the possibility that skewed X inactivation in a peripheral tolerizing compartment might also lead to inefficient tolerance. A major mechanism of peripheral tolerance induction appears to involve the recognition of tolerizing antigens by T cells in the absence of costimulation [<xref ref-type="bibr" rid="B19">19</xref>]. This may specifically occur in the paracortical regions of lymph nodes, without further progression of the tolerized T cells into lymph node follicles. Because this mechanism presumably involves circulating antigen, it is difficult to invoke a role for X-inactivation skewing in altering this process, unless the skewing were virtually complete. Tolerance induction by parenchymal tissue has also been described [<xref ref-type="bibr" rid="B20">20</xref>], and may be a multistage process that is still not entirely understood. Conceivably, skewed expression of X-linked autoantigens could play a role here, but this would require invoking a peripheral tolerizing compartment that is limiting with respect to the dosage of tolerizing cells to which peripheral T cells can be exposed. No such compartment has yet been defined.</p><p>In conclusion, we suggest that X-chromosome inactivation needs to be considered as a potential factor in the predominance of females in most autoimmune diseases. Our inability to show an increase in X-chromosome skewing in females with autoimmunity does not eliminate this as an etiologic contributor to loss of immunologic tolerance. Future experiments must be directed at detailed analysis of tissue patterns of X inactivation, as well as at a search for potential X-linked autoantigens.</p></sec>

Heterogeneous nuclear ribonucleoproteins C1/C2 identified as autoantigens by biochemical and mass spectrometric methods

<sec><title>Introduction:</title><p>The classification of antinuclear antibodies (ANAs) is important for diagnosis and prognosis and for understanding the molecular pathology of autoimmune disease. Many of the proteins that associate with RNA in the ribonucleoprotein (RNP) complexes of the spliceosome have been found to react with some types of ANA [<xref ref-type="bibr" rid="B1">1</xref>], including proteins of the heterogeneous nuclear RNP (hnRNP) complex that associate with newly transcribed pre-mRNA. Autoantibodies to the A2, B1, and B2 proteins of hnRNP found in some patients may be markers of several overlap syndromes [<xref ref-type="bibr" rid="B2">2</xref>]. However, ANAs with specificity for these proteins as well as for the D protein also appear to occur in many distinct connective-tissue diseases, although epitope specificities may differ [<xref ref-type="bibr" rid="B3">3</xref>]. ANAs with specificity for the C component of hnRNP (consisting of the C1 and C2 proteins) have to our knowledge so far been described in only one case [<xref ref-type="bibr" rid="B4">4</xref>]. We here describe the approach taken to unambiguously identify the C1/C2 proteins as ANA targets in the sera of some patients.</p></sec><sec><title>Aims:</title><p>To determine the fine specificity of sera containing an unusual speckled ANA-staining pattern using a combination of 2D gel electrophoresis and MS.</p></sec><sec><title>Methods:</title><p>Patient sera were screened for ANAs by indirect immunofluorescence microscopy on HEp-2 cells (cultured carcinoma cells). Sera with an unusual, very regular, speckled ANA pattern were tested for reactivity with components of nuclear extracts of HeLa cells that were separated by one-dimensional (1D) or 2D gel electrophoresis or by reversed-phase high-performance liquid chromatography (HPLC). IgG reactivity was assessed by immunoblotting. Reactive protein spots from 2D separations were excised from the gels and subjected to in-gel digestion with trypsin for subsequent peptide mapping, partial peptide sequencing, and protein identification by MS and tandem MS on a hybrid electrospray ionization/quadrupole/time-of-flight (ESI-Q-TOF) mass spectrometer [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>].</p></sec><sec><title>Results:</title><p>We observed a strong nuclear staining pattern (titer &#x0003e;1280) with the characteristic even-sized coarse speckles and no staining of nucleoli in sera from three patients. On immunoblots of nuclear extracts from HeLa cells, these sera stained two distinct bands, at <italic>M</italic>_r 42 000 and 41 000. There activity strongly resembled that of the patient originally described by Stanek <italic>et al</italic> [<xref ref-type="bibr" rid="B4">4</xref>]. The antigens were enriched by fractionating the extract using reversed-phase HPLC on a C4 column, and the two reactive spots on 2D separations were excised for identification. The two components appeared to be of approximately the same isoelectric points, although their molecular masses differed by approximately 2000. Peptide-mass mapping was performed by matrix-assisted laser desorption-ionization time-of-flight (MALDI-TOF) MS on the tryptic peptide mixture generated by digestion of the two excised proteins. The database search suggested that the two proteins were C1/C2 hnRNPs (Swissprot accession number P07910). The identity of the proteins was further confirmed by tandem MS using an ESI-Q-TOF instrument. One peptide carrying two positive charges (m/z 580.32 Da), corresponding to a peptide mass of 1158.7 Da, was selected as a precursor ion and partially sequenced by collisional fragmentation. The fragmented peptide was found to represent the tryptic fragment VDSLLENLEK, ie amino acids 207-216 (C2 protein numbering). Four other peptides were partially sequenced and all of them matched the human C1/C2 hnRNP sequence. The theoretical masses of C1 and C2 are 32.0 and 33.3 kDa, respectively. The difference between the two sequences is a 13-amino-acid insert in C2 between positions 107 and 108 of C1. The presence of a specific tryptic fragment in the MALDI-TOF peptide-mass map from the higher-molecular-mass spot containing a 13-amino-acid insert that was not present in the lower-molecular-mass spot, further demonstrated that the two components represented the two isoforms of the C class of hnRNPs.</p><p>The patient whose case prompted us to investigate the specificities of these antibodies was a 72-year-old man who had arthralgias and oligoarthritis but did not fulfill the criteria for rheumatoid arthritis and did not have dermatological complaints. The reactivity of various patient groups to the C1/C2 hnRNP autoantigens was subsequently tested by immunoblotting of HeLa-cell nuclear extracts. Of 59 patients with rheumatoid arthritis, 19 with polymyositis, 33 with scleroderma, and 10 with psoriatic arthritis, none had IgG antibodies reacting with the two bands. Of sera from 139 consecutive patients who had moderately to strongly positive speckled ANA patterns shown by indirect immunofluorescence on HEp-2 cells, only two reacted with the C1/C2 hnRNP bands in immunoblotting. One of these was from a young woman (22 years old) whose complaints of muscle tenderness were not explained by objective findings or abnormal laboratory test results. The third patient that we identified through ANA screening followed by immunoblotting was a 54-year-old male who was being treated with methotrexate for long-standing polymyositis in addition to psoriasis and possible osteoporosis. </p></sec><sec><title>Discussion:</title><p>The results confirm the existence of anti-C1/C2 antibodies in some patients with speckled ANAs. The antigens were identified through the use of biochemical methods using high-resolution separation techniques combined with mass-spectrometry peptide mapping and database searches. As a general approach, this is a powerful way to identify new antigens using small amounts of material without the need for conventional protein sequencing. The approach does require, however, that the proteins can be found in databases, that they are not extensively post-translationally modified, that they can be digested enzymatically, and that they can be isolated in appropriately pure form by the separation technique used.</p><p>It is not known at present if the C1/C2 antibodies may have pathogenic relevance and/or relate to specific diagnoses or subsets within the group of connective-tissue diseases. It does appear that the reactivity is quite rare among ANA-positive patients, and therefore many patients will have to be examined to determine these issues. The fact that the antibodies to the C1/C2 hnRNPs are revealed by indirect immunofluorescence would indicate that the epitopes are accessible in intact, fixed HEp-2 cells and thus probably reside outside the nucleic-acid-binding domains that would be expected to be covered by RNA.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>HH Heegaard</surname><given-names>Niels</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>R Larsen</surname><given-names>Martin</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Muncrief</surname><given-names>Terri</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Wiik </surname><given-names>Allan</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Roepstorff</surname><given-names>Peter</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>The determination of the molecular and submolecular specificity of ANAs is important for diagnosis and prognosis in clinical medicine and for understanding the molecular pathology of autoimmune disease. Different antigenic specificities of ANA are associated with different diseases and different patterns of disease manifestations. The various protein antigens associated with ANAs characteristically represent molecules that bind nucleic acids. Thus, many of the proteins that associate with RNA in the RNP complexes of the spliceosome have been found to react with some types of ANA [<xref ref-type="bibr" rid="B1">1</xref>]. Also, the hnRNP complex that associates with newly transcribed pre-mRNA contains proteins that have been shown to be targets of ANAs in some rheumatic diseases. The protein components of hnRNP are the most abundant nonhistone proteins of the cell nucleus and are classified roughly on the basis of their molecular masses into at least 21 protein groups (A through U). Autoantibodies to the A2, B1, and B2 proteins of hnRNP found in some patients may be markers of several overlap syndromes [<xref ref-type="bibr" rid="B2">2</xref>]. However, ANAs with specificity for these proteins as well as for the D protein and the I protein also appear to occur in many distinct connective-tissue diseases, although epitope specificities may differ [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. Despite the discovery of the A, B, and R hnRNP ANA specificities in patients [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>], antibodies against the C component of hnRNP (consisting of the C1 and C2 proteins) have to our knowledge so far been described in only one case [<xref ref-type="bibr" rid="B4">4</xref>]. The identification in this study was based on the colocalization of patient immunoreactivity on immunoblots of nuclear extracts with the band developed with a monoclonal anti-C1/C2 antibody. In the present study, we identify the C1/C2 specificity of circulating ANAs in three patients by high-resolution separation techniques (2D gel electrophoresis) combined with the assignment of protein identities by mass-spectrometric (MS and tandem MS) peptide mapping. This protein was also found to be the target of the ANA found in the patient described earlier [<xref ref-type="bibr" rid="B4">4</xref>].</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Reagents</title><p>Trifluoroacetic acid was from Fluka (Buchs, Switzerland), HPLC-grade water and acetonitrile from Merck (Darmstadt, Germany), sequencing-grade trypsin from Promega (Madison, WI, USA). The bicinchonic acid assay for protein was from Pierce (Rockford, IL, USA). All other chemicals were of analytical grade from Sigma Chemical Co (St Louis, MO, USA).</p></sec><sec><title>Cells and antibodies</title><p>HeLa-cell nuclei were obtained from 4C (Mons, Belgium). Cultured carcinoma (HEp-2)-cell slides were from Immunoconcepts, Inc (Sacramento, CA, USA). Fluorescein- and alkaline-phosphatase-labeled rabbit anti-human IgG antibodies (F0202 and D0336, respectively) were from Dako (Glostrup, Denmark). A patient serum sample previously described as containing antibodies reacting with C1/C2 hnRNP [<xref ref-type="bibr" rid="B4">4</xref>] was generously provided by Dr J Vencovsk&#x000fd;, Prague.</p></sec><sec><title>Indirect immunofluorescence</title><p>Patient sera were screened for ANAs by indirect immunofluorescence microscopy on slides of acetone-fixed HEp-2 cells using a cutoff dilution of 1:160 for the patient sera.</p></sec><sec><title>Preparation of HeLa-cell nuclear extracts</title><p>To establish the best conditions for extracting the antigens, the nuclear material was subjected to five different procedures: high-salt extraction using either 0.5 M NaCl or 6 M guanidinium hydrochloride in 125 mM Tris/HCl, pH 6.8 (sonicated for 3&#x000d7;5 s at 20 000 Hz on ice); detergent extraction in 125 mM Tris/HCl, pH 6.8, with 5% &#x003b2;-octyl glucoside or with 2.5% Triton X-100, using sonication as above; or extraction with 2.5% SDS for 10 min on ice after mixing. All samples were subsequently centrifuged at 16 000&#x000d7; <bold><italic>g</italic></bold> for 5 min and the supernatants were then analyzed by gel electrophoresis and immunoblotting (high-salt samples after being diluted 10 times in water). Extracts were either stored at -80&#x000b0;C or processed immediately for 1D or 2D gel electrophoresis or for reversed-phase HPLC.</p></sec><sec><title>Reversed-phase HPLC</title><p>HeLa-cell nuclear extracts were separated by reversed-phase HPLC on an analytical C4 column (4.6&#x000d7;250 mm, 300 &#x000c5;, 5 &#x003bc;m particle size) from Vydac (Hesperia, CA, USA) using an &#x000c4;kta chromatography system (Pharmacia, Lund, Sweden). The column was eluted with a 50-min gradient of 0-70% acetonitrile in 0.1% aqueous trifluoroacetic acid at 1 ml/min with monitoring at 210 and 280 nm. Fractions of 1 ml were collected. Prior to testing for reactivity with patient sera, the collected fractions were dried down in a Savant SpeedVac (Farmingdale, NY, USA) and resolubilized in 100 &#x003bc;l water. Dots of 2 &#x003bc;l of these fractions were subsequently placed on nitrocellulose, dried, and developed with patient sera as detailed below.</p></sec><sec><title>One- and 2D gel electrophoresis and immunoblotting</title><p>SDS-PAGE was performed in 10% precast gels from Novex (San Diego, CA, USA). Samples were solubilized at a protein concentration of 1-2 mg/ml by boiling 20-&#x003bc;l aliquots for 3 min in the presence of 2.5% (w/v) SDS and 25 mM dithiothreitol in 10 mM Tris/HCl, pH8, 10% (v/v) glycerol. The sample buffer also contained 50 &#x003bc;g/ml Pyronin G as a marker. The electrophoresis running buffer was 25 mM Tris, 192 mM glycine, 0.1% SDS, pH8.3 (Novex), and separation took place at 120 V for 1-2 h until the marker reached 1 cm from the bottom edge of the gel.</p><p>Two-dimensional electrophoresis was used to analyze crude HeLa-cell nuclear extracts and HPLC-purified sub-fractions of the extracts. Prior to isoelectric focusing, samples were solubilized at a protein concentration of approximately 1 mg/ml in 1.5% (v/v) Ampholine 3.5-9.5 (Pharmacia), 0.75% (w/v) SDS, 12 mM dithiothreitol, 2 M urea, 1% CHAPS (3- [{3-cholamidopropyl}-dimethylammonio]-1-propane sulfonate), and an aliquot of bromphenol blue. Isoelectric focusing of 17-ml samples took place in 1-mm&#x000d7;130-mm tube gels (4% acrylamide [C=2.7%] containing 9 M urea, 1% [w/v] CHAPS, and 0.5% [v/v] Nonidet P40). The first-dimension isoelectric focusing was performed at 400 V for 17-18 h and then at 800 V for 1 h. After the isoelectric focusing, the tube gels were extruded and placed on top of a large (13&#x000d7;13 cm) 11% polyacrylamide gel (C=2.7%) prepared in a Protean Cell (BioRad, Hercules, CA, USA). After equilibration for about 1 min in the presence of 500 &#x003bc;l transfer buffer (6.25 mM Tris/HCl, pH6.8, 10% [v/v] glycerol, 2% [w/v] SDS, 200 mM dithiothreitol, and an aliquot of bromophenol blue), the second-dimension SDS-PAGE took place at a constant 15 W/gel for 4-5 h. One- and 2D gels were either silver-stained using a modified protocol of Heukeshoven and Dernick [<xref ref-type="bibr" rid="B11">11</xref>], in which glutaraldehyde had been left out, or they were electroblotted for 1 h at 1 mA/cm² onto nitrocellulose of pore size 0.2 &#x003bc;m in a semidry electroblotting cell, according to standard procedures. Blots were blocked with non-ionic detergent (2% Tween 20) for 5 min and developed with patient sera (diluted 1:100 unless otherwise noted) and alkaline-phosphatase-labeled anti-human IgG antibodies at 1:2000 using Nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate for the color reaction as described elsewhere [<xref ref-type="bibr" rid="B12">12</xref>].</p></sec><sec><title>In-gel digestion of proteins separated by gel electrophoresis</title><p>In-gel digestion was performed as described previously [<xref ref-type="bibr" rid="B6">6</xref>]. The excised gel plugs were washed in 50 mM NH₄HCO₃/acetonitrile (60/40) and dried by vacuum centrifugation. The trypsin (sequencing grade), dissolved in 50 mM NH₄HCO₃ at 12 ng/&#x003bc;l, was added to the dry gel pieces and incubated on ice for 1 h for reswelling. The supernatant was removed, 60 &#x003bc;l digestion buffer was added, and the digestion was continued at 37&#x000b0;C for 4-18 h. The peptide mixture was analyzed directly by MALDI-MS without extraction by depositing a 0.8-&#x003bc;l 1:1 mixture of analyte solution in 2% trifluoroacetic acid and matrix solution (15-20 g/l &#x003b1;-cyano-4-hydroxycinnamic acid in 70% acetonitrile) directly on the MALDI target.</p><p>In other cases, the samples were desalted and concentrated as described elsewhere [<xref ref-type="bibr" rid="B13">13</xref>], and for analysis by MALDI-MS, the peptides were eluted with 0.5 &#x003bc;l matrix solution (15-20 g/l of &#x003b1;-cyano-4-hydroxycinnamic acid in 70% acetonitrile) and deposited directly on the MALDI target. For tandem MS, the peptides on the column were eluted with 1.5 &#x003bc;l 50% methanol/49% H₂O/1% formic acid directly into the nanoelectrospray needle.</p></sec><sec><title>Peptide-mass mapping by MALDI-MS</title><p>A Bruker REFLEX model MALDI-TOF mass spectrometer (Bruker-Franzen Analytik GmbH, Bremen, Germany) equipped with the Scout source and variable detector bias gating was used in positive-ion reflector mode for mass analysis of peptide mixtures (peptide-mass mapping). The ion acceleration voltage was 20 kV.</p></sec><sec><title>ESI-Q-TOF MS</title><p>Tandem MS of peptides generated by in-gel digestion was performed on a nano-ESI-Q-TOF mass spectrometer (Micromass, Manchester, UK). Ions were produced in an atmospheric pressure ionization (API)/ESI ion source at 40&#x000b0;C. The flow rate of the drying gas was 50 l/h. A potential of 1 kV was applied to the precoated borosilicate nanoelectrospray needles (Protana A/S, Odense, Denmark), and a stable flow rate (10-30 nl/min) was produced with a nitrogen back-pressure of 0-5 psi (0-34.5 kPa). The cone voltage was 50 V. A quadrupole analyzer was used to select the precursor ion for fragmentation in the hexapole collision cell. The collision gas was argon at a pressure of 6&#x000d7;10^-5 mbar (6&#x000d7;10^-3 Pa) and the collision voltage was 20-25 V. Product ions were analyzed using an orthogonal TOF analyzer. Data were processed with a Mass Lynx Windows NT-based program.</p></sec><sec><title>Protein identification</title><p>Protein identification by peptide-mass fingerprinting was performed by searching the peptide masses in a comprehensive non-redundant protein sequence database (NRDB; European Bioinformatics Institute, Hinxton, UK) containing approximately 420 000 protein sequences, using the database search program PepSea (Protana, Odense, Denmark). The peptide-mass maps and the protein identifications were evaluated as described in [<xref ref-type="bibr" rid="B6">6</xref>]. The partial amino acid sequences of the peptides generated by tandem MS together with the associated masses were used to construct the peptide sequence tags [<xref ref-type="bibr" rid="B7">7</xref>] that were used to identify the proteins in the NRDB.</p></sec></sec><sec><title>Results</title><sec><title>Indirect immunofluorescence on HEp-2 cells</title><p>The starting point for the investigation into the specificity of the antibodies in the present study was the observation of an unusual staining pattern in a patient serum being tested for IgG reactivity on HEp-2 cells. A strong nuclear staining pattern (titer &#x0003e;1280) with very even-sized, coarse speckles and no staining of nucleoli was observed (Fig. <xref ref-type="fig" rid="F1">1</xref>). This staining pattern did not resemble the well-known variably sized spliceosomal speckled pattern seen with antibodiesto Sm/U₁RNP or the fine speckled pattern characteristic of SSA/SSB-antibody-positive sera. In accordance with this, the patient had none of these antibody specificities when tested by specific assays (data not shown).</p></sec><sec><title>Determination of antigen molecular mass in HeLa-cell nuclear extracts</title><p>The IgG reactivity of the serum shown in Fig. <xref ref-type="fig" rid="F1">1</xref> was tested on an SDS-PAGE-separated nuclear extract from HeLa cells as shown in Fig. <xref ref-type="fig" rid="F2">2</xref>. A double band in the <italic>M</italic>_r 40 000-42 000 range and a high-molecular-mass band (around 200 000) were observed as the sole reactivities of this particular patient (Fig. <xref ref-type="fig" rid="F2">2</xref>, lane 2). Sera from two other patients, with a similar atypical ANA-staining pattern, also stained components of this doublet in immunoblotting (Fig. <xref ref-type="fig" rid="F2">2</xref>, lanes 3 and 4). The reactivity strongly resembled that of the patient originally described by Stanek <italic>et al</italic> [<xref ref-type="bibr" rid="B4">4</xref>] (Fig. <xref ref-type="fig" rid="F2">2</xref>, lane 5). The two different, more weakly reacting, sera appear to react preferentially with either the lower (lane 3) or the higher (lane 4) band, respectively. There was no reactivity with a normal serum (lane 1, Fig. <xref ref-type="fig" rid="F2">2</xref>). The antigens were present both in high-salt, SDS, and non-ionic detergent extracts of nuclei but appeared to be extracted to the highest degree in the &#x003b2;-octyl glucoside extracts (data not shown).</p></sec><sec><title>Enrichment of antigen preparation</title><p>After verification by immunoblotting that antibody reactivity was present in crude HeLa-cell nuclear extracts obtained by treatment with &#x003b2;-octyl glucoside (Fig. <xref ref-type="fig" rid="F2">2</xref>), the antigen was enriched by fractionating the extract by reversed-phase HPLC on a C4 column. Two of the collected fractions contained material with a molecular mass around 40 000 that was reactive with the patient antibody as judged by SDS-PAGE immunoblotting (data not shown).</p></sec><sec><title>Micropreparative separation of antigens</title><p>The contents of the crude nuclear extract obtained after &#x003b2;-octyl glucoside treatment were separated by 2D gel electrophoresis into several hundred components (Fig. <xref ref-type="fig" rid="F3">3a</xref>). However, the patient antibodies only reacted with two components in the mixture on immunoblots of the 2D separation (Fig. <xref ref-type="fig" rid="F3">3b</xref>). The two components appeared to be of approximately the same isoelectric points, but they differed in molecular mass by approximately 2000. The purified fractions from the C4 reversed-phase HPLC separation gave a less complicated 2D separation pattern (Fig. <xref ref-type="fig" rid="F3">3c</xref>), in which the two reactive spots (encircled) could be clearly identified and were positive on immunoblotting with patient serum (Fig. <xref ref-type="fig" rid="F3">3d</xref>). These two spots were excised for individual characterization.</p></sec><sec><title>Protein identification</title><p>MALDI-TOF peptide-mass mapping was performed on the tryptic peptide mixture generated by in-gel digestion of the two components after excision from 2D gels in two independent experiments. The database search compared the detected peptide masses with the theoretical tryptic peptide masses of the protein sequences present in the database. This approach in both sets of experiments suggested that the two proteins were hnRNPs C1/C2 (Swissprot accession number P07910), a conclusion suggested by the matching of 8 of 27 peptide masses entered (data not shown). The low number of matched peptides relative to the number of peptides included in the search, together with the small difference in the number of matching peptides to the next hit in the database (three peptides), make this identification only tentative. Therefore, the identity of the proteins was further confirmed by tandem MS using an ESI-Q-TOF instrument. One peptide carrying two positive charges (m/z 580.32 Da) corresponding to a peptide mass of 1158.7 Da was selected as a precursor ion and fragmented in the collision cell. The fragment ion spectrum is illustrated in Fig. <xref ref-type="fig" rid="F4">4a</xref>. A peptide sequence tag [<xref ref-type="bibr" rid="B7">7</xref>] consisting of a partial sequence and its associated masses was constructed -(632.36)LLSD(1060.70) - and was used to search the NRDB. The database search subsequently unambiguously identified the protein as human C1/C2 hnRNP (data not shown). The fragmented peptide was found to represent the tryptic fragment VDSLLENLEK, ie amino acids 207-216 (C2 protein numbering). Four other peptides were partially sequenced, and all were found to match the human C1/C2 hnRNP sequence (data not shown). Several of the peptide masses that were not matched to the sequence could be matched to peptides derived from keratin and from autoproteolysis of trypsin (data not shown).</p><p>The theoretical masses of C1 and C2 are 32.0 and 33.3 kDa, respectively. The difference between the two sequences is a 13-amino-acid insert in C2 between positions 107 and 108 of C1. The presence of a specific tryptic fragment in the MALDI-TOF peptide-mass map from the upper spot but not from the lower spot containing a 13-amino-acid insert in the protein sequence (SAAEMYGSVT-EHPSPSPLLSSSFDLDYDFQR) clearly demonstrates that the two components in the gel represent the two isoforms of the C-class of hnRNPS (Fig. <xref ref-type="fig" rid="F4">4b</xref>). The peptide-mass map of C1 contains a peptide signal at m/z 2101.9 Da, corresponding to the tryptic peptide without the 13-amino-acid insert - SAAEMYGSSFDLDYDFQR.</p></sec><sec><title>Patients and clinical data</title><p>The patient whose case prompted us to investigate the specificities of these antibodies was a 72-year-old man with a speckled ANA titer of &#x0003e;1280, who had arthralgias and oligoarthritis but did not fulfill the criteria for rheumatoid arthritis and did not have dermatological complaints. The reactivity of sera from various patient groups towards the C1/C2 hnRNP autoantigens was subsequently tested by immunoblotting of HeLa-cell nuclear extracts. None of 59 patients with rheumatoid arthritis, 19 with polymyositis, 33 with scleroderma, and 10 with psoriatic arthritis had IgG antibodies reacting with the two bands. Among 139 consecutive patients who had moderately to strongly positive speckled ANA patterns by indirect immunofluorescence on HEp-2 cells, only two additional patients had sera that reacted with the C1/C2 hnRNP bands in immunoblotting (Fig. <xref ref-type="fig" rid="F2">2</xref>, lanes 3 and 4). The serum of one of these patients reacted preferentially with the C1 hnRNP, and that of the other reacted with the C2 hnRNP and thus may be specific for the C2 insert. One patient was a young woman (22 years old) with complaints of muscle tenderness but without objective findings or other abnormal laboratory test findings, whose sera showed speckled ANAs, with a titer of &#x0003e;1280. A third patient whose case we identified through ANA screening followed by immunoblotting was a 54-year-old man, whose sera also showed speckled ANAs, with a titer of &#x0003e;1280. This man was receiving methotrexate and had long-standing polymyositis in addition to psoriasis and possible osteoporosis.</p></sec></sec><sec><title>Discussion</title><p>Although the prototype serum found in our study strongly matched the reactivity of the serum from the report by Stanek <italic>et al</italic> [<xref ref-type="bibr" rid="B4">4</xref>] with respect to the singular staining pattern shown in Fig. <xref ref-type="fig" rid="F1">1</xref>, it probably would not be recognized as such in routine diagnostic laboratory tests.</p><p>The present study confirms the existence of anti-C1/C2 antibodies in some patients with speckled ANAs. The antigens were identified through the use of biochemical methods using high-resolution separation techniques combined with mass-spectrometry peptide mapping and database searches. As a general approach, this is a powerful way to identify new antigens using small amounts of material without the need for conventional protein sequencing and consequently with no problems related to blocked N-termini. The approach does require, however, that the proteins can be found in databases, that they are not extensively post-translationally modified, that they can be digested enzymatically, and that they can be isolated in appropriately pure form by the separation technique used.</p><p>The C1 and C2 proteins are identical except for an insert of 13 amino acids after serine 107 in the 303-amino-acid-containing C2 protein [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. The presence of this insert accounts for the separation of the C1 and C2 proteins by SDS-PAGE. Like other hnRNPs, they contain consensus sequence RNA-binding domains and nuclear localization signals [<xref ref-type="bibr" rid="B15">15</xref>]. The C1/C2 proteins form anisotropic([C1]₃ [C2]₁) tetramers that bind strongly but apparentlynonspecifically to RNA and probably form a basic structural unit in pre-mRNA-processing and RNA-packaging complexes [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. Together with the A and B groups of hnRNPs, they constitute the core proteins of the 40S hnRNP particles [<xref ref-type="bibr" rid="B19">19</xref>]. C1 and C2 are phosphorylated, contain unusually acidic C-terminal domains, and, accordingly, are the most acidic of the hnRNPs.</p><p>It is not known at present if the C1/C2 antibodies are pathogenically relevant and/or relate to specific diagnoses or subsets within the group of connective-tissue diseases. Although ELISA with native antigen may increase the number of positive findings, reactivity does seem to be quite rare among ANA-positive patients. Therefore, clarification of these issues would require the study of many patients.</p><p>Autoantibodies to many of the core proteins of the hnRNP complex have now been described, including the A₁, A₂/RA33, A₃, B₁, and B₂ proteins as well as D/AUF-1 and, recently, also the hnRNP R protein [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. These proteins are primarily involved in packaging and processing of pre-mRNA and in the regulation of alternative splicing. They may also be involved in RNA transport between the nucleus and the cytoplasm. ANAs directed against these proteins do not give clear-cut ANA reactivity on commonly used ANA substrates such as HEp-2 cells [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. This is probably because the antibodies specifically target RNA-binding domains that are shielded by bound RNA in intact cells. The fact that these antibodies often appear together with other ANAs in systemic lupus erythematosus, in mixed connective-tissue diseases, and in adult and juvenile rheumatoid arthritis has made it difficult to study their presence except by immunoblotting of extracts treated with DNase and RNase to remove nucleic acids. The fact that antibodies to the C1/C2 hnRNPs are revealed by indirect immunofluorescence indicates that the epitopes are accessible in intact, fixed HEp-2 cells and thus are probably outside nucleic-acid-binding domains that would be expected to be covered by RNA.</p></sec>

Cytokine, activation marker, and chemokine receptor expression by individual CD4⁺ memory T cells in rheumatoid arthritis synovium

<sec><title>Introduction:</title><p>In RA large numbers of CD4⁺ memory T cells infiltrate the inflamed synovium [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. The accumulated CD4⁺ memory T cells in the RA synovium appear to be activated, because they express cytokines and activation markers [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. Expressed cytokines and activation markers should play important roles in the pathogenesis of RA. However, the frequency of cytokine expression by RA synovial CD4⁺ T cells has not been analyzed accurately. Recently, the roles of chemokine and chemokine receptor interactions in T-cell migration have been intensively examined. Interactions of chemokine and chemokine receptors might therefore be important in the accumulation of the CD4⁺ T cells in the RA synovium. Accordingly, correlation of cytokine and chemokine receptor expression might be important in delineating the function and potential means of accumulation of individual CD4⁺ memory T cells in the RA synovium.</p><p>In the present study we analyzed cytokine (IL-2, IL-4, IL-6, IL-10, IL-13, IFN-&#x003b3; , TNF-&#x003b1; , and LT-&#x003b1; ), activation marker (CD154 [CD40 ligand] and TRANCE - also called receptor activator of nuclear factor &#x003ba; B ligand [RANKL] or osteoclast differentiation factor [ODF]), and chemokine receptor expression by individual CD4⁺ memory T cells isolated from rheumatoid synovium and blood. To achieve this we employed a single-cell reverse transcription (RT) polymerase chain reaction (PCR) technique. This technique made it possible to correlate mRNAs expressed by individual CD4⁺ memory T cells in the synovium and blood.</p></sec><sec><title>Materials and method:</title><p>Synovial tissues from three RA patients and peripheral blood mononuclear cells from two RA patients and a normal donor were analyzed.</p><p>Cytokine (IL-2, IL-4, IL-6, IL-10, IL-13, IFN-&#x003b3;, TNF-&#x003b1;, and LT-&#x003b1; ) and activation marker (CD154 and TRANCE) expression by individual CD4⁺CD45RO⁺ T cells from RA synovium or blood were analyzed using a single-cell RT-PCR. In brief, single CD4⁺CD45RO⁺T cells was sorted into each well of a 96-well PCR plate using a flow cytometer. cDNA from individual cells was prepared, and then the cDNA was nonspecifically amplified. The product was then amplified by PCR using gene-specific primers to analyze cytokine and activation marker expression.</p></sec><sec><title>Results:</title><p>Cytokine and activation marker expression by individual CD4⁺CD45RO⁺T cells from RA synovial tissues was analyzed using a single-cell RT-PCR method. Expression of mRNAs was analyzed in 152 individual synovial tissue CD4⁺CD45RO⁺ T cells sorted from three RA patients in which T-cell receptor (TCR) C&#x003b2; mRNA was detected. Frequencies of CD4⁺ memory T cells expressing cytokine and activation marker mRNA in RA synovium are shown in Table <xref ref-type="table" rid="T1">1</xref>. IL-2, IL-4, and IL-6 were not expressed by the synovial tissue CD4⁺CD45RO⁺ T cells, whereas 2-20% of cells expressed the other cytokine mRNAs.</p><p>Few correlations between cytokine and activation marker mRNAs were observed. Notably, no cells contained both IFN-&#x003b3; and LT-&#x003b1; mRNAs, cytokines that are thought to define the T-helper (Th)1 phenotype [<xref ref-type="bibr" rid="B9">9</xref>]. However, the frequency of TRANCE-positive cells in IL-10-positive cells was significantly higher than that in IL-10-negative cells (Table <xref ref-type="table" rid="T2">2</xref>). Moreover, the frequency of TRANCE-positive cells in TNF-&#x003b1;-positive cells was also significantly higher than that in TNF-&#x003b1;-negative cells. </p><p>Varying percentages of CD4⁺ memory T cells expressed CC and CXC chemokine receptors. The frequency of CCR5-positive cells in IFN-&#x003b3;-positive cells was significantly higher than that in IFN-&#x003b3;-negative cells, whereas the frequency of CCR6-positive cells in LT-&#x003b1;-positive cells was significantly higher than that in LT-&#x003b1;-negative cells, and the frequency of CCR7-positive cells in IL-10-positive cells was significantly higher than that in IL-10-negative cells. Furthermore, the frequency of CXCR4-positive cells in TRANCE-positive cells was significantly higher than that in TRANCE-negative cells.</p><p>Expression of cytokine and activation marker mRNAs was also analyzed in 48 individual peripheral blood CD4⁺CD45RO⁺ T cells from two RA patients. IL-2, IL-4, IL-6, and LT-&#x003b1; were not expressed by the peripheral CD4⁺CD45RO⁺ T cells, whereas 4-17% of cells expressed the other markers. The most striking difference between synovial tissue and peripheral blood CD4⁺ memory T cells was the presence of LT-&#x003b1; expression in the former, but not in the latter. IFN-&#x003b3; and TNF-&#x003b1; were not expressed by normal peripheral blood CD4⁺ memory T cells, although they were expressed by RA peripheral blood CD4⁺ memory T cells. </p></sec><sec><title>Discussion:</title><p>The present study employed a single-cell PCR technology to analyze cytokine expression by unstimulated RA synovial tissue CD4⁺ memory T cells immediately after isolation, without <italic>in vitro</italic> manipulation. The results confirm the Th1 nature of rheumatoid inflammation. It is noteworthy that no individual synovial CD4⁺ memory T cells expressed both IFN-&#x003b3; and LT-&#x003b1; mRNAs, even though these are the prototypic Th1 cytokines [<xref ref-type="bibr" rid="B9">9</xref>]. These results imply that, in the synovium, regulation of IFN-&#x003b3; and LT-&#x003b1; must vary in individual cells, even though both Th1 cytokines can be produced.</p><p>The present data showed that CCR5 expression correlated with IFN-&#x003b3; but not with LT-&#x003b1; expression by synovial CD4⁺ memory T cells. It has been reported that CCR5 expression is upregulated in RA synovial fluid and synovial tissue T cells [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>] and that CCR5 &#x00394; 32 deletion may have an influence on clinical manifestations of RA [<xref ref-type="bibr" rid="B13">13</xref>], suggesting that CCR5 might play an important role in RA. Recently, it has been claimed that CCR5 was preferentially expressed by Th1 cell lines [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. However, in the present study CCR5 was not expressed by all IFN-&#x003b3;-expressing cells. Moreover, CCR5 expression did not correlate with expression of LT-&#x003b1; by RA synovial CD4⁺ memory T cells. Therefore, it is unclear whether CCR5 is a marker of Th1 cells in RA synovium.</p><p>IL-10 expression correlated with CCR7 expression by RA synovial CD4⁺ memory T cells. Recently, it was reported [<xref ref-type="bibr" rid="B16">16</xref>] that in the blood CCR7⁺CD4⁺ memory T cells express lymph-node homing receptors and lack immediate effector function, but efficiently stimulate dendritic cells. These cells may play a unique role in the synovium as opposed to in the blood. By producing IL-10, they might have an immunoregulatory function. In addition, IL-10 expression also correlated with expression of TRANCE. Although it is possible that IL-10 produced by these cells inhibited T-cell activation in the synovium, TRANCE expressed by these same cells might function to activate dendritic cells and indirectly stimulate T cells, mediating inflammation in the synovium. These results imply that individual T cells in the synovium might have different, and sometimes opposite functional activities.</p><p>LT-&#x003b1; expression correlated with CCR6 expression by synovial CD4⁺ memory T cells. It has been reported that CCR6 is expressed by resting peripheral memory T cells [<xref ref-type="bibr" rid="B17">17</xref>], whereas LT-&#x003b1; expression is associated with the presence of lymphocytic aggregates in synovial tissue [<xref ref-type="bibr" rid="B7">7</xref>]. The correlation between the expression of these two markers therefore suggests the possibility that CCR6 may play a role in the development of aggregates of CD4⁺ T cells that are characteristically found in rheumatoid synovium.</p><p>TRANCE is known to be expressed by activated T cells, and can stimulate dendritic cells and osteoclasts [<xref ref-type="bibr" rid="B18">18</xref>]. Of note, TRANCE-mediated activation of osteoclasts has recently been shown [<xref ref-type="bibr" rid="B19">19</xref>] to play an important role in the damage to bone that is found in experimental models of inflammatory arthritis. It is therefore of interest that TRANCE was expressed by 3-16% of the RA synovial CD4⁺ memory T cells. Of note, 67% of TNF-&#x003b1;-positive cells expressed TRANCE. In concert, TNF-&#x003b1; and TRANCE expressed by this subset of CD4⁺ memory T cells might make them particularly important in mediating the bony erosions that are characteristic of RA.</p><p>Interestingly, there was a correlation between expression of IFN-&#x003b3; and IL-10 in RA peripheral blood CD4⁺ memory T cells. In RA peripheral blood, CD154 expression correlated with that of CXCR3 by CD4⁺ memory T cells. It has been claimed [<xref ref-type="bibr" rid="B15">15</xref>] that CXCR3 is preferentially expressed by <italic>in vitro</italic> generated Th1 cells. However, in the present study CXCR3 did not correlate with IFN-&#x003b3; expression. Although IFN-&#x003b3; and TNF-&#x003b1; mRNAs were expressed <italic>in vivo</italic> by peripheral blood CD4⁺ T cells from RA patients, LT-&#x003b1; mRNA was not detected, whereas IFN-&#x003b3; , TNF-&#x003b1; , and LT-&#x003b1; were not detected in samples from healthy donors. These findings indicate that RA peripheral blood CD4⁺ memory T cells are stimulated <italic>in vivo</italic>, although they do not express LT-&#x003b1; mRNA. The present studies indicate that the frequencies of CD4⁺ memory T cells that expressed IFN-&#x003b3; in the blood and in the synovium are comparable. These results imply that activated CD4⁺ memory T cells migrate between blood and synovium, although the direction of the trafficking is unknown. The presence of LT-&#x003b1; mRNA in synovium, but not in blood, indicates that CD4⁺ memory cells are further activated in the synovium, and that these activated CD4⁺ memory T cells are retained in the synovium until LT-&#x003b1; mRNA decreases.</p><p>In conclusion, CD4⁺ memory T cells are biased toward Th1 cells in RA synovium and peripheral blood. In the synovium, IFN-&#x003b3; and LT-&#x003b1; were produced by individual cells, whereas in the rheumatoid blood no LT-&#x003b1;-producing cells were detected. Furthermore, there were modest correlations between individual cells that expressed particular cytokines, such as IL-10, and certain chemokine receptor mRNAs.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Nanki</surname><given-names>Toshihiro</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>lipskyp@mail.nih.gov</email></contrib><contrib id="A2" contrib-type="author"><name><surname>E Lipsky</surname><given-names>Peter</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>In RA large numbers of CD4⁺ memory T cells infiltrate the inflamed synovium [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. The accumulated CD4⁺ memory T cells in the RA synovium appear to be activated, because they express cytokines and activation markers [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. Expressed cytokines and activation markers should play important roles in the pathogenesis of RA. However, the frequency of cytokine expression by RA synovial CD4⁺ T cells has not been analyzed accurately. Recently, the roles of chemokine and chemokine receptor interactions in T-cell migration have been intensively examined. Interactions of chemokine and chemokine receptors might therefore be important for the accumulation of the T cells in the RA synovium. Accordingly, correlation of cytokine and chemokine receptor expression might be important in delineating the function and potential means of accumulation of individual CD4⁺ memory T cells in the RA synovium.</p><p>In order to analyze cytokine, activation marker, and chemokine receptor expression by individual CD4⁺ memory T cells, we used a single-cell RT-PCR technique. The method made it possible to analyze cytokine and chemokine receptor expression by individual RA synovial CD4⁺ memory T cells without <italic>in vitro</italic> stimulation, and to correlate cytokine and chemokine receptor expression.</p><p>Using this technique, we analyzed cytokine (IL-2, IL-4, IL-6, IL-10, IL-13, IFN-&#x003b3;, TNF-&#x003b1;, and LT-&#x003b1;), activation marker (CD154 [CD40 ligand] and TRANCE - also called receptor activator of nuclear factor-&#x003ba; B ligand [RANKL] or osteoclast differentiation factor [ODF]), and chemokine receptor expression by individual CD4⁺ memory T cells isolated from rheumatoid synovium and blood. The results indicate that CD4⁺ memory T cells are biased toward Th1 cells in RA synovium, although individual cells produced IFN-&#x003b3; or LT-&#x003b1;, but not both. A similar pattern of cytokine production was observed with CD4⁺ memory T cells from RA blood, with the exception that no cells expressing LT-&#x003b1; were detected. There were modest correlations between individual cells that expressed particular cytokine and chemokine receptor mRNAs.</p></sec><sec><title>Materials and method</title><sec><title>Specimens</title><p>Synovial tissues were obtained at surgery from three RA patients. The synovial tissue was minced and incubated with 0.3 mg/ml collagenase (Sigma, St Louis, MO, USA) for 1h at 37&#x000b0;C in RPMI 1640 medium (Life Technologies, Gaithersburg, MD, USA). Partially digested pieces of the tissue were pressed through a metal screen to obtain single-cell suspensions. Mononuclear cells were then isolated by ficoll-hypaque (Pharmacia Biotech, Piscataway, NJ, USA) gradient centrifugation. RA was diagnosed according to the American College of Rheumatology criteria [<xref ref-type="bibr" rid="B20">20</xref>].</p><p>Also, peripheral blood mononuclear cells were separated by ficoll-hypaque gradient centrifugation from two RA patients and a normal donor.</p></sec><sec><title>Single cell sorting and reverse transcription polymerase chain reaction</title><p>The method for construction of cDNA libraries from single cells was similar to previously reported techniques [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. RA synovial tissue T cells were analyzed without <italic>in vitro</italic> culture or stimulation by staining synovium mononuclear cells with FITC-conjugated anti-CD4 mono-clonal antibody (Q4120; Sigma) and PE-conjugated anti-CD45RO monoclonal antibody (UCHL-1), after which individual CD4⁺CD45RO⁺ T cells were sorted in a 96-well PCR plate (Robbins Scientific, Sunnyvale, CA, USA) using the FACStar^Plus (Becton Dickinson, San Jose, CA, USA). Peripheral CD4⁺CD45RO⁺ single cells were also sorted into wells of 96-well PCR plates using the FACStar^Plus flow cytometer.</p><p>Each well contained 4 &#x003bc;l lysis buffer (50 mmol/l Tris-HCl [pH8.3], 75 mmol/l KCl, 3 mmol/l MgCl₂, 1 mmol/l DTT,10 &#x003bc;mol/l dNTP [Sigma], 5 U/ml PRIME RNase Inhibitor [5 Prime ?? 3 Prime Incorporated, Boulder, CO, USA], 300 U/ml RNAguard [Pharmacia Biotech], 200 ng/ml oligo [dT]₂₄ [Integrated DNA Technologies Incorporated, Coralville, IA, USA], and 0.5% NP-40). The samples were heated to 65&#x000b0;C for 1 min, cooled to 20&#x000b0;C for 3 min, and maintained on ice. Two units of AMV Reverse Transcriptase (Promega, Madison, WI, USA) and 50 U of M-MLV Reverse Transcriptase (Life Technologies) was added, and the samples were incubated at 37&#x000b0;C for 15 min before heat inactivation at 65&#x000b0;C for 10 min. For polyadenylate tailing at the 3' end of the cDNA, 5 &#x003bc;l tailing buffer (200 mmol/l potassium cacodylate [pH 7.2], 4 mmol/l CoCl₂, 0.4 mmol/l DTT), 2 mmol/l dATP (Roche, Indianapolis, IN, USA), and 10 U terminal transferase (Roche) were added, and incubated at 37&#x000b0;C for 20 min, followed by heat inactivation at 65&#x000b0;C for 10 min. To amplify the cDNA non-specifically, PCR was performed with 100 &#x003bc;l of 10 mmol/l Tris-HCl (pH 9.0), 50 mmol/l KCl, 2.5 mmol/l MgCl₂, 0.01% Triton-X, 1 mmol/l dNTP, 10 U Taq DNA polymerase (Promega), and 2 &#x003bc;mol/l X-(dT)₂₄ primer (ATG TCG TCC AGG CCG CTC TGG ACA AAA TAT GAA TTC-dT₂₄; Integrated DNA Technologies Incorporated). Twenty-five cycles of amplification were performed with 1 min at 94&#x000b0;C, 2 min at 42&#x000b0;C, and 6 min at 72&#x000b0;C, plus 10 s extension per cycle. Afterward, 5U Taq DNA polymerase was added, followed by an additional 25 cycles of PCR.</p><p>For gene-specific amplification, 1 &#x003bc;l of nonspecifically amplified cDNA was amplified by PCR in 25 &#x003bc; l 10 mmol/l Tris-HCl (pH 9.0), 50 mmol/l KCl, 1.5 mmol/l MgCl₂, 0.01% Triton-X, 200 &#x003bc;mol/l dNTP and 0.625 U Taq DNA polymerase. The cycling program was: 94&#x000b0;C for 1 min, 60&#x000b0;C for 1 min (58&#x000b0;C: IL-2, IL-4, IFN-&#x003b3;), and 72&#x000b0;C for 1 min for 35 cycles, followed by a final extension for 7 min. For nested amplification, 1 &#x003bc;l of amplified PCR reaction mixture was added to a second PCR reaction mixture (50 &#x003bc;l of 10 mmol/l Tris-HCl [pH 9.0], 50 mmol/l KCl, 1.5 mmol/l MgCl₂, 0.01% Triton-X, 200 &#x003bc;mol/l dNTP, and 1.25 U Taq DNA polymerase). The cycling program was: 94&#x000b0;C for 1 min, 60&#x000b0;C for 1 min, and 72&#x000b0;C for 1 min for 35 cycles, followed by a final extension for 7 min. The PCR products were then separated by electrophoresis through 2.0% agarose. The primers were designed to be within 600 bp of the 3' end of each mRNA. The primers used were as shown in Table <xref ref-type="table" rid="T3">3</xref>.</p><p>To confirm that the PCR products were amplified from the corresponding genes, the nucleotide sequences of the PCR products were analyzed. More than five PCR products of each cytokine from a total of two or three donors were sequenced. All the sequences of the PCR products were identical to the previously published sequences (data not shown).</p><p>To confirm that each well contained only one cell after sorting, TCR V&#x003b2; mRNA was analyzed by single-cell RT-PCR using V&#x003b2; family-specific primers. In the wells analyzed, only one TCR V&#x003b2; was detected (data not shown).</p></sec><sec><title>Statistical analyses</title><p>To analyze correlations between cytokines and chemokine receptor expressions, and to compare frequencies of chemokine receptor-expressing cells between different T-cell subsets, Fisher's exact probability test was used.</p></sec></sec><sec><title>Results</title><sec><title>Cytokine and activation marker expression by individual rheumatoid synovial tissue CD4⁺ memory T cells</title><p>Cytokine and activation marker expressions by individual CD4⁺CD45RO⁺T cells from RA synovial tissues were analyzed by employing a single-cell RT-PCR method. Expressions of mRNAs were analyzed in 152 individual synovial tissue CD4⁺CD45RO⁺ T cells sorted from three RA patients in which TCR C&#x003b2; mRNA was detected. Cytokine and activation marker expressions by 50 synovial tissue CD4⁺CD45RO⁺ T cells (RA1) are shown in Fig. <xref ref-type="fig" rid="F1">1</xref>.</p><p>Frequencies of CD4⁺ memory T cells that expressed cytokine and activation marker mRNA in RA synovium are shown in Table <xref ref-type="table" rid="T1">1</xref>. IL-2, IL-4, and IL-6 were not expressed by the synovial tissue CD4⁺CD45RO⁺ T cells, whereas 2-20% of cells expressed the other mRNAs.</p></sec><sec><title>Correlation of the expression of cytokines by rheumatoid synovial tissue CD4⁺ memory T cells</title><p>Few correlations between cytokine and activation marker mRNAs were observed. Notably, no cells contained mRNAs for both IFN-&#x003b3; and LT-&#x003b1;, cytokines that are thought to define the Th1 phenotype [<xref ref-type="bibr" rid="B9">9</xref>]. However, the frequency of TRANCE-positive cells in IL-10-positive cells was significantly higher than that in IL-10-negative cells (Table <xref ref-type="table" rid="T2">2</xref>). Moreover, the frequency of TRANCE-positive cells in TNF-&#x003b1;-positive cells was also significantly higher than that in TNF-&#x003b1;-negative cells.</p></sec><sec><title>Correlation of cytokine and chemokine receptor expression by RA synovial tissue CD4⁺ memory T cells</title><p>Varying percentages of CD4⁺ memory T cells expressed CC and CXC chemokine receptors as shown in Tables <xref ref-type="table" rid="T4">4</xref> and <xref ref-type="table" rid="T5">5</xref>. Thus, for example, 21% of RA synovial CD4⁺ memory T cells expressed CCR5, 39% CCRb, and 19% CCR7 mRNAs, whereas 16% expressed CXCR3 and 76% expressed CXCR4 mRNAs. The frequency of CCR5-positive cells in IFN-&#x003b3;-positive cells was significantly higher than that in IFN-&#x003b3;-negative cells (Table <xref ref-type="table" rid="T4">4</xref>), whereas the frequency of CCR6-positive cells in LT-&#x003b1;-positive cells was significantly higher than that in LT-&#x003b1;-negative cells, and the frequency of CCR7-positive cells in IL-10-positive cells was significantly higher than that in IL-10-negative cells. Furthermore, the frequency of CXCR4-positive cells in TRANCE-positive cells was significantly higher than that in TRANCE-negative cells (Table <xref ref-type="table" rid="T5">5</xref>).</p></sec><sec><title>Cytokine and activation marker expression by peripheral blood CD4⁺ memory T cells from RA patients and a normal donor</title><p>Expressions of cytokine and activation marker mRNAs was also analyzed in 48 individual peripheral blood CD4⁺CD45RO⁺ T cells sorted from two RA patients and in 33 individual peripheral blood CD4⁺CD45RO⁺ T cells sorted from a normal donor. Frequencies of CD4⁺ memory T cells that expressed cytokine and activation marker mRNA are shown in Table <xref ref-type="table" rid="T6">6</xref>. IL-2, IL-4, IL-6, and LT-&#x003b1; were not expressed by the RA peripheral blood CD4⁺CD45RO⁺ T cells, whereas 4-17% of cells expressed the other markers. The most striking difference between RA synovial tissue and peripheral blood CD4⁺ memory T cells was the presence of LT-&#x003b1; expression in the former, but not in the latter. IFN-&#x003b3; and TNF-&#x003b1; were not expressed by normal peripheral blood CD4⁺ memory T cells, although they were expressed by RA peripheral blood CD4⁺ memory T cells.</p><p>Variable frequencies of RA peripheral CD4⁺ memory T cells expressed chemokine receptor mRNAs. Except for significantly decreased expressions of CCR5 and CXCR4, there were no differences between chemokine receptor expressions by synovial tissue and peripheral blood CD4⁺ memory T cells. In peripheral blood CD4⁺ memory T cells, there was a significant correlation between IFN-&#x003b3; and IL-10 expressions, and IFN-&#x003b3; and CCR6 expressions (<italic>P</italic> &#x0003c;0.05; Table <xref ref-type="table" rid="T7">7</xref>). In addition, CD154 and CXCR3 expressions correlated (<italic>P</italic> &#x0003c;0.005). No other correlations were detected (data not shown).</p></sec></sec><sec><title>Discussion</title><p>The present study employed single-cell PCR technology to analyze cytokine mRNA expressions by unstimulated RA synovial tissue CD4⁺ memory T cells immediately after isolation, without <italic>in vitro</italic> manipulation. The results are consistent with the Th1 nature of rheumatoid inflammation. These data showed that 6-22% of RA synovial CD4⁺ memory T cells produced IFN-&#x003b3; mRNA. Previous studies [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>] reported that 1-10% of RA synovial T cells expressed IFN-&#x003b3; protein by immunohistologic analysis. Although there is some variation in the results obtained with the different methodologies, both results are consistent with the conclusion that there is a Th1 bias in RA [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B23">23</xref>]. It is noteworthy that no individual synovial CD4⁺ memory T cells expressed both IFN-&#x003b3; and LT-&#x003b1; mRNA, even though these are the prototypic Th1 cytokines [<xref ref-type="bibr" rid="B9">9</xref>]. These results imply that, in the synovium, regulation of IFN-&#x003b3; and LT-&#x003b1; must vary in individual cells, even though both Th1 cytokine mRNAs can be expressed.</p><p>The present data showed that CCR5 expression correlated with IFN-&#x003b3; but not with LT-&#x003b1; expression by synovial CD4⁺ memory T cells. It has been reported that CCR5 expression is upregulated in RA synovial fluid and synovial tissue T cells [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>], and that CCR5 &#x00394; 32 deletion may have an influence on clinical manifestations of RA [<xref ref-type="bibr" rid="B13">13</xref>], suggesting that CCR5 might play an important role in RA. Recently, it has been claimed [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>] that CCR5 was preferentially expressed by Th1 cell lines. In the present study, however, CCR5 was not expressed by all IFN-&#x003b3; expressing cells. Moreover, CCR5 expression did not correlate with expression of LT-&#x003b1; by RA synovial CD4⁺ memory T cells, although it correlated with IFN-&#x003b3; . Therefore, it is unclear whether CCR5 is a marker of Th1 cells in RA synovium.</p><p>Of RA synovial CD4⁺ T cells 6-14% expressed IL-10, and the expression correlated with CCR7 expression. It has been reported that approximately 1.5% of synovial T cells express IL-10 by immunohistochemistry [<xref ref-type="bibr" rid="B5">5</xref>], and that 4% of synovial CD4⁺ T cells have the potential to express IL-10 [<xref ref-type="bibr" rid="B3">3</xref>]. Recently, it was reported [<xref ref-type="bibr" rid="B16">16</xref>] that, in the blood, CCR7⁺CD4⁺ memory T cells express lymph-node homing receptors and lack immediate effector function, but efficiently stimulate dendritic cells. However, because 19% of RA tissue CD4⁺ memory T cells expressed CCR7 and there was a correlation between IL-10 production and CCR7 expression, these cells may play a unique role in the synovium as opposed to in the blood. By producing IL-10, they may exert an immunoregulatory function. In addition, it is interesting to note that IL-10 expression also correlated with expression of TRANCE. Although it is possible that IL-10 produced by these cells inhibited T-cell activation in the synovium, TRANCE expressed by these same cells might function to activate dendritic cells and indirectly stimulate T cells, mediating inflammation in the synovium. These results imply that individual T cells in the synovium might have different, and sometimes opposite functional activities.</p><p>LT-&#x003b1; was expressed by 3-12% of the synovial CD4⁺ memory T cells, and the expression correlated with CCR6 expression, which is expressed by 39% of the synovial CD4⁺ memory T cells. However, there were no LT-&#x003b1;-expressing CD4⁺ T cells that also produced IFN-&#x003b3;, although synovial CD4⁺ memory T cells that produced each cytokine were found in abundance. It has been reported that CCR6 is expressed by resting peripheral memory T cells [<xref ref-type="bibr" rid="B17">17</xref>], whereas LT-&#x003b1; expression is associated with the presence of lymphocytic aggregates in synovial tissue [<xref ref-type="bibr" rid="B7">7</xref>]. The correlation between the expression of these two markers therefore suggests the possibility that CCR6 might play a role in the development of aggregates of CD4⁺ T cells that are characteristically found in rheumatoid synovium.</p><p>TRANCE is known to be expressed by activated T cells, and can stimulate dendritic cells and osteoclasts [<xref ref-type="bibr" rid="B18">18</xref>]. Of note, TRANCE-mediated activation of osteoclasts has recently been shown [<xref ref-type="bibr" rid="B19">19</xref>] to play an important role in the damage to bone found in experimental models of inflammatory arthritis. Recently, the presence of TRANCE in rheumatoid synovium was reported [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. It is therefore of interest that TRANCE was expressed by 3-16% of the RA synovial CD4⁺ memory T cells. Of note, TRANCE expression correlated with IL-10, TNF-&#x003b1;, and CXCR4 expressions. Especially noteworthy was that 67% of TNF-&#x003b1;-positive cells expressed TRANCE. In concert, TNF-&#x003b1; and TRANCE expressed by this subset of CD4⁺ memory T cells might make them particularly important in mediating the bony erosions that are characteristic of RA.</p><p>Interestingly, there was a correlation between expression of IFN-&#x003b3; and IL-10 in RA peripheral blood CD4⁺ memory T cells. Production of IL-10 in humans differs from that in the mouse, in that IL-10 production does not appear to be restricted to Th2 cells [<xref ref-type="bibr" rid="B9">9</xref>]. As noted here, RA peripheral CD4⁺ T cells could express both IFN-&#x003b3; and IL-10. In RA peripheral blood, CD154 expression correlated with CXCR3 by CD4⁺ memory T cells. It has been claimed [<xref ref-type="bibr" rid="B15">15</xref>] that CXCR3 was preferentially expressed by <italic>in vitro</italic> generated Th1 cells. However, in the present study CXCR3 did not correlate with IFN-&#x003b3; expression. Although IFN-&#x003b3; and TNF-&#x003b1; mRNAs were expressed <italic>in vivo</italic> by peripheral blood CD4⁺ T cells from RA patients, LT-&#x003b1; mRNA was not detected, whereas IFN-&#x003b3;, TNF-&#x003b1;, and LT-&#x003b1; were not detected from healthy donors. These findings indicate that RA peripheral blood CD4⁺ memory T cells are stimulated <italic>in vivo</italic>, but that they do not express LT-&#x003b1; mRNA. Previous studies have documented the presence of IFN-&#x003b3; and IL-10 mRNA in circulating T cells of RA patients [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>]. The present studies indicate that the frequency of CD4⁺ memory T cells that express IFN-&#x003b3; in the blood and synovium is comparable, although the percentages that secrete IFN-&#x003b3; are not known. These results imply that activated CD4⁺ memory T cells migrate between blood and synovium, although the direction of the trafficking is unknown. The presence of LT-&#x003b1; mRNA in synovium but not in blood indicates that CD4⁺ memory cells are further activated in the synovium, and that these activated CD4⁺ memory T cells are retained in the synovium until LT-&#x003b1; mRNA decreases.</p><p>In conclusion, CD4⁺ memory T cells are biased toward Th1 cells in RA synovium and peripheral blood. In the synovium, IFN-&#x003b3; and LT-&#x003b1; were produced by individual cells, whereas in the rheumatoid blood no LT-&#x003b1;-producing cells were detected. Furthermore, there were modest correlations between individual cells that expressed particular cytokines and certain chemokine receptor mRNAs.</p></sec>

Monoarticular antigen-induced arthritis leads to pronounced bilateral upregulation of the expression of neurokinin 1 and bradykinin 2 receptors in dorsal root ganglion neurons of rats

<sec><title>Introduction:</title><p>Ongoing pain and hyperalgesia (enhanced pain response to stimulation of the tissue) are major symptoms of arthritis. Arthritic pain results from the activation and sensitization of primary afferent nociceptive nerve fibres ('pain fibres') supplying the tissue (peripheral sensitization) and from the activation and sensitization of nociceptive neurons in the central nervous system (central sensitization). After sensitization, nociceptive neurons respond more strongly to mechanical and thermal stimulation of the tissue, and their activation threshold is lowered. The activation and sensitization of primary afferent fibres results from the action of inflammatory mediators such as bradykinin (BK), prostaglandins and others on membrane receptors located on these neurons. BK is a potent pain-producing substance that is contained in inflammatory exudates. Up to 50% of the primary afferent nerve fibres have receptors for BK. When primary afferent nerve fibres are activated they can release neuropeptides such as substance P (SP) and calcitonin gene-related peptide from their sensory endings in the tissue. SP contributes to the inflammatory changes in the innervated tissue (neurogenic inflammation), and it might also support the sensitization of nociceptive nerve fibres by binding to neurokinin 1 (NK1) receptors. NK1 receptors are normally expressed on a small proportion of the primary afferent nerve fibres.</p></sec><sec><title>Aims:</title><p>Because the expression of receptors on the primary afferent neurons is essential for the pain-producing action of inflammatory mediators and neuropeptides, we investigated in the present study whether the expression of BK and NK1 receptors on primary afferent neurons is altered during the acute and chronic phases of an antigen-induced arthritis (AIA). AIA resembles in many aspects the inflammatory process of human rheumatoid arthritis. Because peptide receptors are expressed not only in the terminals of the primary afferent units but also in the cell bodies, we removed dorsal root ganglia (DRGs) of both sides from control rats and from rats with the acute or chronic phase of AIA and determined, after short-term culture of the neurons, the proportion of DRG neurons that expressed the receptors in the different phases of AIA. We also characterized the inflammatory process and the nociceptive behaviour of the rats in the course of AIA.</p></sec><sec><title>Materials and methods:</title><p>In 33 female Lewis rats 10 weeks old, AIA was induced in the right knee joint. First the rats were immunized in two steps with methylated bovine serum albumin (m-BSA) emulsified with Freund's complete adjuvant, and heat-inactivated <italic>Bordetella pertussis</italic>. After immunization, m-BSA was injected into the right knee joint cavity to induce arthritis. The joint swelling was measured at regular intervals. Nociceptive (pain) responses to mechanical stimulation of the injected and the contralateral knee were monitored in the course of AIA. Groups of rats were killed at different time points after the induction of AIA, and inflammation and destruction in the knee joint were graded by histological examination. The DRGs of both sides were dissected from segments L1&#x02013;L5 and C1&#x02013;C7 from arthritic rats, from eight immunized rats without arthritis and from ten normal control rats. Excised DRGs were dissociated into single cells which were cultured for 18 h.</p><p>The expression of the receptors was determined by assessment of the binding of SP-gold or BK-gold to the cultured neurons. For this purpose the cells were slightly fixed. Binding of SP-gold or BK-gold was detected by using enhancement with silver and subsequent densitometric analysis of the relative grey values of the neurons. Displacement controls were performed with SP, the specific NK1 receptor agonist [Sar⁹, Met(O₂)¹¹]-SP, BK, the specific BK 1 (B1) receptor agonist _D-Arg (Hyp³-Thi^5,8-_D-Phe⁷)-BK and the specific BK 2 (B2) receptor agonist (Des-Arg¹⁰)-Lys-BK.</p></sec><sec><title>Results:</title><p>The inflammatory process in the injected right knee joint started on the first day after induction of AIA and persisted throughout the observation period of 84 days (Fig. <xref ref-type="fig" rid="F1">1</xref>). The initial phase of AIA was characterized by strong joint swelling and a predominantly granulocytic infiltration of the synovial membrane and the joint cavity (acute inflammatory changes). In the later phases of AIA (10&#x02013;84 days after induction of AIA) the joint showed persistent swelling, and signs of chronic arthritic alterations such as infiltration of mononuclear leucocytes, hyperplasia of synovial lining layer (pannus formation) and erosions of cartilage and bone were predominant. The contralateral knee joints appeared normal at all time points. Destruction was observed only in the injected knee but some proteoglycan loss was also noted in the non-injected, contralateral knee. In the acute and initial chronic phases of AIA (1&#x02013;29 days) the rats showed mechanical hyperalgesia in the inflamed knee (limping, withdrawal response to gentle pressure onto the knee). In the acute phase (up to 9 days) a pain response was also seen when gentle pressure was applied to the contralateral knee.</p><p>Figure <xref ref-type="fig" rid="F2">2</xref> displays the changes in the receptor expression in the DRG neurons during AIA. The expression of SP&#x02013;gold-binding sites in lumbar DRG neurons (Fig. <xref ref-type="fig" rid="F2">2a</xref>) was substantially increased in the acute phase of arthritis. In untreated control rats (<italic>n</italic> = 5), 7.7 &#x000b1; 3.8% of the DRG neurons from the right side and 10.0 &#x000b1; 1.7% of the DRG neurons from the left side showed labelling with SP&#x02013;gold. The proportion of SP&#x02013;gold-labelled neurons in immunized animals without knee injection (<italic>n</italic> = 3) was similar. By contrast, at days 1 (<italic>n</italic> = 2 rats) and 3 (<italic>n</italic> = 5 rats) of AIA in the right knee, approximately 50% of the DRG neurons exhibited labelling with SP&#x02013;gold, and this was seen both on the side of the injected knee and on the opposite side. At day 10 of AIA (<italic>n</italic> = 3 rats), 26.3 &#x000b1; 6.1% of the ipsilateral DRG neurons but only 15.7 &#x000b1; 0.6% of the contralateral neurons exhibited binding of SP&#x02013;gold. At days 21 (<italic>n</italic> = 5 rats), 42 (<italic>n</italic> = 3 rats) and 84 (<italic>n</italic> = 5 rats) of AIA, the proportion of SP&#x02013;gold-positive neurons had returned to the control values, although the arthritis, now with signs of chronic inflammation, was still present. Compared with the DRG neurons of the untreated control rats, the increase in the proportion of labelled neurons was significant on both sides in the acute phase (days 1 and 3) and the intermediate phase (day 10) of AIA (Mann&#x02013;Whitney <italic>U</italic>-test). The size distribution of the neurons was similar in the DRG neurons of all experimental groups. Under all conditions and at all time points, SP&#x02013;gold binding was found mainly in small and medium-sized (less than 700 &#x003bc;m²) neurons. In the cervical DRGs the expression of NK1 receptors did not change in the course of AIA. The binding of SP&#x02013;gold to the neurons was suppressed by the coadministration of the specific NK1 receptor agonist [Sar⁹, Met(O₂)¹¹]&#x02013;SP in three experiments, showing that SP&#x02013;gold was bound to NK1 receptors.</p><p>The expression of BK&#x02013;gold-binding sites in the lumbar DRG neurons showed also changes in the course of AIA, but the pattern was different (Fig. <xref ref-type="fig" rid="F2">2b</xref>). In untreated control rats (<italic>n</italic> = 5), 42.3 &#x000b1; 3.1% of the DRG neurons of the right side and 39.6 &#x000b1; 2.6% of the DRG neurons of the left side showed binding of BK&#x02013;gold. At days 1 (<italic>n</italic> = 2 rats) and 3 (<italic>n</italic> = 5 rats) of AIA, approximately 80% of the DRG neurons on the side of the knee injection (ipsilateral) and approximately 70% on the opposite side were labelled. In comparison with the untreated control group, the increase in the proportion of labelled neurons was significant on both sides. The proportion of labelled neurons in the ipsilateral DRGs remained significantly increased in both the intermediate phase (day 10, <italic>n</italic> = 3 rats) and chronic phase (days 21, <italic>n</italic> = 5 rats, and 42, <italic>n</italic> = 3 rats) of inflammation. At 84 days after the induction of AIA (<italic>n</italic> = 5 rats), 51.0 &#x000b1; 12.7% of the neurons showed an expression of BK&#x02013;gold-binding sites and this was close to the prearthritic values. However, in the contralateral DRG of the same animals the proportion of BK&#x02013;gold-labelled neurons declined in the intermediate phase (day 10) and chronic phase (days 21&#x02013;84) of AIA and was not significantly different from the control value. Thus the increase in BK&#x02013;gold-labelled neurons was persistent on the side where the inflammation had been induced, and transient on the opposite side. The size distribution of the DRG neurons of the different experimental groups was similar. In the cervical DRGs the expression of BK receptors did not change in the course of AIA. In another series of experiments, we determined the subtype(s) of BK receptor(s) that were expressed in DRGs L1&#x02013;L5 in different experimental groups. In neither untreated control animals (<italic>n</italic> = 5) nor immunized rats without knee injection (<italic>n</italic> = 5) nor in rats at 3 days (<italic>n</italic> = 5) and 42 days (<italic>n</italic> = 5) of AIA was the binding of BK&#x02013;gold decreased by the coadministration of BK&#x02013;gold and the B1 agonist. By contrast, in these experimental groups the binding of BK&#x02013;gold was suppressed by the coadministration of the B2 agonist. These results show that B2 receptors, but not B1 receptors, were expressed in both normal animals and in animals with AIA.</p></sec><sec><title>Discussion:</title><p>These results show that in AIA in the rat the expression of SP-binding and BK-binding sites in the perikarya of DRGs L1&#x02013;L5 is markedly upregulated in the course of knee inflammation. Although the inflammation was induced on one side only, the initial changes in the binding sites were found in the lumbar DRGs of both sides. No upregulation of SP-binding or BK-binding sites was observed in the cervical DRGs. The expression of SP-binding sites was upregulated only in the first days of AIA, that is, in the acute phase, in which the pain responses to mechanical stimulation were most pronounced. By contrast, the upregulation of BK-binding sites on the side of AIA persisted for up to 42 days, that is, in the acute and chronic phase of AIA. Only the B2 receptor, not the B1 receptor, was upregulated. The coincidence of the enhanced expression of NK1 and BK receptors on sensory neurons and the pain behaviour suggests that the upregulation of these receptors is relevant for the generation and maintenance of arthritic pain.</p><p>In the acute phase of AIA, approximately 50% of the lumbar DRG neurons showed an expression of SP-binding sites. Because peptide receptors are transported to the periphery, the marked upregulation of SP-binding receptors probably leads to an enhanced density of receptors in the sensory endings of the primary afferent units. This will permit SP to sensitize more neurons under inflammatory conditions than under normal conditions. However, the expression of NK1 receptors was upregulated only in the acute phase of inflammation, suggesting that SP and NK1 receptors are less important for the generation of hyperalgesia in the chronic phase of AIA.</p><p>Because BK is one of the most potent algesic compounds, the functional consequence of the upregulation of BK receptors is likely to be of immediate importance for the generation and maintenance of inflammatory pain. The persistence of the upregulation of BK receptors on the side of inflammation suggests that BK receptors should be an interesting target for pain treatment in the acute and chronic phases. Only B2 receptors were identified in normal animals and in rats with AIA. This is surprising because previous pharmacological studies have provided evidence that, during inflammation, B1 receptors can be newly expressed.</p><p>Receptor upregulation in the acute phase of AIA was bilateral and almost symmetrical. However, hyperalgesia was much more pronounced on the inflamed side. It is most likely that receptors on the contralateral side were not readily activated because in the absence of gross inflammation the local concentration of the ligands BK and SP was probably quite low. We hypothesize that the bilateral changes in receptor expression are generated at least in part by mechanisms involving the nervous system. Symmetrical segmental changes can be produced only by the symmetrical innervation, involving either the sympathetic nervous system or the primary afferent fibres. Under inflammatory conditions, primary afferent fibres can be antidromically activated bilaterally in the entry zone of afferent fibres in the spinal cord, and it was proposed that this antidromic activation might release neuropeptides and thus contribute to neurogenic inflammation. Because both sympathetic efferent fibres and primary afferent nerve fibres can aggravate inflammatory symptoms, it is also conceivable that they are involved in the regulation of receptor expression in primary afferent neurons. A neurogenic mechanism might also have been responsible for the bilateral degradation of articular cartilage in the present study.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Segond von Banchet</surname><given-names>Gisela</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>schailbe@mti-n.uni-jena.de</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Petrow</surname><given-names>Peter K</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Br&#x000e4;uer</surname><given-names>Rolf</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Schaible</surname><given-names>Hans-Georg</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Persistent pain is a major symptom of arthritis such as human rheumatoid arthritis. Inflammatory pain is caused by the activation and sensitization of primary afferent nociceptive neurons ('pain fibres') supplying the tissue (peripheral sensitization), and from the activation and sensitization of nociceptive neurons in the central nervous system (central sensitization). After sensitization, nociceptive neurons respond more strongly to mechanical and thermal stimulation of the tissue, and their activation threshold is lowered. These neuronal changes cause persistent pain as well as hyperalgesia, an enhanced pain sensitivity to mechanical and thermal stimuli [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>]. The peripheral sensitization is produced by the action of inflammatory mediators such as bradykinin (BK) and prostaglandins on the primary afferent neurons that express receptors for these compounds in their sensory endings [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. Subgroups of primary afferent neurons also express receptors for neuropeptides such as substance P (SP) that are released from primary afferent fibres.</p><p>Because the activation of primary afferent neurons by mediators is dependent on the receptors located on the neurons, we investigated in the present experiments the expression of BK and SP [neurokinin 1 (NK1)] receptors in primary afferent neurons in the acute and chronic phases of antigen-induced arthritis (AIA), which resembles in many aspects the inflammatory process of human rheumatoid arthritis [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. BK is a potent pain-producing substance in animals and in humans. When it is applied to the tissue it causes many neurons to fire action potentials, and it sensitizes the neurons so that they respond more strongly to mechanical and thermal stimuli [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. BK is produced under inflammatory conditions from a precursor and is released in the plasma; it is contained in inflammatory exudates, for example in the joint [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>]. It acts on BK 2 (B2) receptors that are expressed in sensory neurons under normal conditions. Under inflammatory conditions an additional expression of a BK 1 (B1) receptor has been reported [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>].</p><p>SP is a peptide that is produced in and released from the primary afferent fibres themselves. Indeed, SP is synthesized in 10&#x02013;20% of the primary afferent neurons [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>], most of which seem to be nociceptive neurons with high thresholds [<xref ref-type="bibr" rid="B22">22</xref>]. SP is transported from the cell bodies in the dorsal root ganglia (DRGs) to the sensory endings in the tissue, and the release from the sensory endings causes a neurogenic inflammation [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. SP is also transported to the central terminals of the primary afferent neurons. At this site it is involved in the activation and in the generation of inflammation-evoked hyperexcitability of spinal cord neurons [<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>]. SP acts on NK1 receptors [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>]. The application of SP to the tissue can sensitize a proportion of the nociceptive primary afferent neurons, suggesting that SP contributes to the generation of hyperalgesia [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>]. However, other studies have failed to show effects of SP on primary afferent neurons [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B37">37</xref>]. Thus, under normal conditions, the role of SP is questionable.</p><p>Because it is very difficult to study the expression of peptide receptors <italic>in situ</italic>, we took advantage of the fact that peptide receptors are expressed not only in the terminals of the primary afferent units but also in the cell bodies [<xref ref-type="bibr" rid="B33">33</xref>,<xref ref-type="bibr" rid="B38">38</xref>,<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. We therefore removed DRGs of both sides from control rats and from rats with the acute or chronic phase of AIA and determined, after short-term culture of the neurons, the proportion of DRG neurons that expressed the receptors under the different experimental conditions. We also characterized the inflammatory process and the nociceptive behaviour of the rats in the course of AIA.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Induction of joint inflammation</title><p>In 33 female Lewis rats 10 weeks old (Charles River, Sulzfeld, Germany), an inflammation was induced in the right knee joint. In the first step the rats received a subcutaneous injection of 500 &#x003bc;g of antigen [methylated bovine serum albumin (m-BSA); Sigma, Deisenhofen, Germany] in 500 &#x003bc;l of saline and emulsified with 500 &#x003bc;l of Freund's complete adjuvant [supplemented with 2 mg/ml <italic>Mycobacterium tuberculosis</italic> strain H37RA (Difco, Detroit, MI, USA] and an intraperitoneal injection of 2 &#x000d7; 10⁹ heat-inactivated <italic>Bordetella pertussis</italic> (Pertussis-Referenzlabor, Krankenhaus Berlin-Friedrichshain, Berlin, Germany). The same immunization procedure was repeated seven days later. After a further 14 days a sterile solution of antigen (m-BSA) (500 &#x003bc;g in 50 &#x003bc;l of saline) was injected into the cavity of the right knee joint (day 0). Eight animals received the same immunization procedure excluding the injection of antigen into the knee joint. The mediolateral diameters of the knee joints were measured at regular intervals by means of a vernier caliper (see Fig. <xref ref-type="fig" rid="F1">1a</xref>).</p><p>At 1, 3, 10, 21, 42 or 84 days after the induction of inflammation in the knee joint, the rats were killed by cervical dislocation during anaesthesia with ether. A total of 10 untreated rats of the same age and sex were used as normal control animals. The immunized rats without arthritis induction were killed 14 days after the second immunization (day 0).</p><p>Two series of experiments were performed. In the first series (<italic>n</italic> = 31 rats), we determined the expression of SP-binding and BK-binding sites in the DRGs of lumbar segments L1&#x02013;L5. The nerves innervating the knee joint are contained in these segments. In the second series of experiments (<italic>n</italic> = 20 rats) we again took DRGs from segments L1&#x02013;L5 to determine the BK receptor subtypes, and we removed DRGs from cervical segments C1&#x02013;C7 to determine the expression of NK1 and BK receptors in the primary afferent neurons at the cervical level. In addition, a detailed documentation of the behaviour was performed in the second series. All procedures complied with the regulations of the Thuringian Commission for Animal Protection.</p></sec><sec><title>Histology and grading of arthritis</title><p>After the rats had been killed, both knee joints were removed, skinned, fixed in 4% buffered formalin, decalcified in EDTA, embedded in paraffin, cut into 5 &#x003bc;m frontal sections and stained with haematoxylin&#x02013;eosin for microscopic examination. Three sections per knee joint were examined in a blind fashion by two independent observers (PKP and RB) with the use of a semiquantitative score (0 = no, 1 = mild, 2 = moderate, 3 = severe alterations). The acute inflammatory reaction was assessed by evaluating the quantity of fibrin exudation and the relative number and density of granulocytes in the synovial membrane and in the joint space. The chronic inflammatory reaction was quantified on the basis of the relative number and density of infiltrating mononuclear leucocytes in the synovial membrane, the degree of synovial hyperplasia, and the extent of fibrosis in the synovial tissues. The histological score is the sum of the three parameters evaluated. A score from 0 to 4 was used to assess the degree of cartilage destruction: 0, no destruction; 1, unequivocal erosions of less than 10% of cartilage and bone cross sections; 2, erosion of 10&#x02013;25%; 3, erosion of 25&#x02013;50%; 4, erosion of more than 50% of cartilage and bone cross sections. Additional sections were stained with safranin O to determine the loss of proteoglycan in the cartilage matrix, with the use of the same score as for the evaluation of inflammatory reactions.</p></sec><sec><title>Testing of nociceptive behaviour</title><p>At several time points after the induction of AIA, the behaviour of rats (at least <italic>n</italic> = 5 rats for each time point) was assessed by using a score. First each rat was placed in a box in which it could move freely. The severity of disturbances of walking was graded: 4, no walking; 3, walking on three legs; 2, limping with the leg with inflammation; 1, limping with the leg with inflamed knee only after pressure on the knee; 0, normal walking. Other parameters (exploratory behaviour, standing on hindlimbs) were also checked and documented. Thereafter the rats were held in the hand by one experimenter and the following tests were performed by the other experimenter: flexion/extension of the left and right knee joint as well as an application of moderate non-painful and strong, lightly painful pressure onto the ankle joints and the knee joints. In all these cases we determined whether the rat showed a nociceptive reaction, namely a withdrawal of the stimulated leg. By using a mechanical device that applied pressure to a small area, we administered pressure to the lateral side of the knee joint and determined the pressure range at which the rat withdrew the leg. Responses to pressure in the range 0&#x02013;100 g were scored 3 (this stimulus evokes a touch sensation in humans), responses to pressure in the range 100&#x02013;200 g were scored 2 (this stimulus evokes a pressure sensation), responses to 200&#x02013;250 g were scored 1 and a lack of response to 250 g was scored 0 (the application of 250 g evokes a weak pain sensation). For analysis, the scores of all animals were added and divided by the number of animals tested in the group.</p></sec><sec><title>Preparation of the DRGs</title><p>Because the fixation of the animals by perfusion impairs the binding of gold-labelled peptides and because the structure of DRGs in cryostat sections is not well maintained, we removed the DRGs after killing the animals and put the neurons into short-term culture. The DRGs of both sides were dissected quickly from segments L1&#x02013;L5 or from segments C1&#x02013;C7. The ganglia were incubated at 37&#x000b0;C in 0.28 U/ml collagenase type A (Boehringer, Mannheim, Germany) dissolved in Dulbecco's modified Eagle's medium (DMEM; Gibco BRL, Eggenstein, Germany) for 100 min. After being washed with phosphate-buffered saline (PBS, 20 mmol, pH 7.4), the ganglia were incubated in PBS containing 25,000 U/ml trypsin (Sigma) for 11 min at 37&#x000b0;C. Then the ganglia were dissociated into single cells by gentle agitation and by trituration through a fire-polished Pasteur pipette. The cells were washed three times in DMEM by centrifugation (500 <italic>g</italic>, 5 min). The final cell pellet was suspended in Ham's F-12 medium (Gibco BRL) containing 10% heat-inactivated horse serum (Gibco BRL), 100 U/ml penicillin (Gibco BRL), 100 &#x003bc;g/ml streptomycin (Gibco BRL) and 100 ng/ml nerve growth factor (7S, recombinant human; Boehringer). Cells were plated on 13 mm glass coverslips coated with poly-(_L-lysine) (200 &#x003bc;g/ml) and kept for 18 h at 37&#x000b0;C in a humidified incubator gassed with 3.5% CO₂in air. Cells were fed after 14&#x02013;16h with supplemented Ham's F-12 medium (see above).</p></sec><sec><title>Preparation of the SP&#x02013;gold and BK&#x02013;gold conjugates</title><p>SP (Sigma) (1 &#x003bc;mol) or 1 &#x003bc;mol of BK (Bachem, Heidelberg, Germany) was dissolved in 500 &#x003bc;l of HEPES (20 mmol, pH 7.5). This solution was added to 6 nmol of sulpho-<italic>N</italic>-hydroxysuccinimido Nanogold reagent (BioTrend, K&#x000f6;ln, Germany), dissolved in 500 &#x003bc;l of doubly distilled water and incubated for 1 h at room temperature. To separate SP&#x02013;gold or BK&#x02013;gold conjugates from unbound SP or BK, membrane centrifugation (Amicon Microcon-10 system) was used. The SP&#x02013;gold and BK&#x02013;gold conjugates were dissolved in PBS containing 0.1% bovine serum albumin (BSA), 0.2 mol of sucrose, 4 &#x003bc;g/ml leupeptin and 10 mmol of sodium azide. This solution was aliquoted and stored at &#x02013;20&#x000b0;C for a maximum of 3 months.</p></sec><sec><title>SP&#x02013;gold and BK&#x02013;gold binding to cultured neurons</title><p>The cells were pre-fixed with 2% paraformaldehyde and 0.05% glutaraldehyde in 0.1 mol of phosphate buffer (pH 7.2) for 30 min. After being washed with PBS (20 mmol, pH 7.4), the cells were pretreated with 50 mmol of glycine in PBS and thereafter with 5% BSA and 0.1% gelatine in PBS for 30min. The cells were then washed with 0.1% acetylated BSA (BSA-C) and incubated overnight with 0.3 nmol/ml SP&#x02013;gold in PBS or with 0.3 nmol/ml BK&#x02013;gold containing 0.1% BSA-C, bacitracin (40 &#x003bc;g/ml), leupeptin(4 &#x003bc;g/ml) and chymostatin (2 &#x003bc;g/ml) at 4&#x000b0;C in a moist chamber. After being washed with PBS plus 0.1% BSA-C and thereafter with PBS to remove unbound SP&#x02013;gold or BK&#x02013;gold, cells were postfixed with 2% glutaraldehyde in PBS for 10 min. After extensive washing with PBS and doubly distilled water, the gold particles were intensified with silver enhancer (R-Gent, pH 5.5; BioTrend) for 15 min at 22&#x000b0;C. The reaction was stopped by washing in doubly distilled water. The preparations were dehydrated and embedded in DePeX (Fluka, Neu-Ulm, Germany).</p></sec><sec><title>Control incubations</title><p>To determine the specificity of the SP&#x02013;gold or BK&#x02013;gold complex used in the binding studies, a displacement control was performed. Neurons were incubated in 1 &#x003bc;mol/ml unlabelled SP or 1 &#x003bc;mol/ml unlabelled BK together with 0.3 nmol/ml peptide&#x02013;gold. The unlabelled peptides should decrease or prevent binding of the gold-labelled peptides. Furthermore, to test whether SP&#x02013;gold is bound specifically to NK1 receptors, 0.3 nmol/ml SP&#x02013;gold was incubated in the presence of 1 &#x003bc;mol/ml [Sar⁹, Met(O₂)¹¹]-SP (Bachem), a specific agonist at the NK1 receptor. To examine whether the binding of BK&#x02013;gold is related to BK receptors, 0.3 nmol/ml BK&#x02013;gold was incubated in the presence of 1 &#x003bc;mol/ml of _D-Arg (Hyp³-Thi^5,8-D-Phe⁷)-BK, a BK analogue specific for the B1 receptor (Bachem), and/or 1 &#x003bc;mol/ml of (Des-Arg¹⁰)-Lys-BK, a B2 receptor agonist (Bachem).</p></sec><sec><title>Data analysis</title><p>At different time points after induction of inflammation, DRG preparations of two to five rats were used to determine SP&#x02013;gold-binding and BK&#x02013;gold-binding sites. The DRGs of untreated rats served as the control group. DRGs were also removed from immunized but non-arthritic animals (day 0 group). To analyse the data, from every coverslip 50 or 100 structurally intact neurons were examined with a light microscope (Zeiss; Axiophot, Jena, Germany) coupled to a CCD colour video camera [AVT-BC6(0)] and an image-analysing system (Kontron, Eching, Germany). On each coverslip, neurons were randomly selected; neurons obstructed by other neurons or by tissue were not included. In total 30 500 neurons were analysed. The relative grey value (grey value of the soma/grey value of the substrate background) was determined for each soma. From each individual binding experiment one coverslip was used to perform a displacement control incubation (see above). This allowed us to determine the grey value range of neurons with no SP&#x02013;gold or no BK&#x02013;gold binding in each experiment (see Fig. <xref ref-type="fig" rid="F5">5</xref>, white bars). In all other coverslips of the particular experiment, neurons were considered as positive for peptide-binding sites if they had a relative grey value above that of neurons from the control incubation. For the final analysis, data from the experiments were pooled. Group results are expressed as means and standard deviations. To evaluate the cell size, the cross-sectional area was taken from each selected neuron. For statistical evaluation, four groups of DRGs were formed, namely DRGs of untreated control rats, of the acute phase (1 and 3 days), of the chronic phase (21 and 42 days) and of the intermittent phase (10 days) of AIA. For comparison of the AIA groups with the control group the Mann&#x02013;Whitney <italic>U</italic>-test was used, with adjustment of significance values for multiple comparisons; significance was accepted at <italic>P</italic> &#x0003c;0.05 [<xref ref-type="bibr" rid="B41">41</xref>].</p></sec></sec><sec><title>Results</title><sec><title>Joint inflammation</title><p>Figure <xref ref-type="fig" rid="F1">1</xref> shows the time course of the inflammatory process. The swelling of the injected knee is shown in Fig. <xref ref-type="fig" rid="F1">1a</xref> (dots). The diameter of the contralateral knees showed a slight increase over time owing to growth. The time course of acute inflammatory changes (a predominantly granulocytic infiltration of the synovial membrane and the joint cavity) is shown in Fig. <xref ref-type="fig" rid="F1">1b</xref>; the time course of chronic alterations (infiltration of mononuclear leucocytes, hyperplasia of synovial lining layer, extent of fibrosis) is displayed in Fig. <xref ref-type="fig" rid="F1">1c</xref>. The contralateral knee joints seemed normal at all time points. Figure <xref ref-type="fig" rid="F1">1d</xref> summarizes the destruction of the joints, taking into account the erosion of cartilage and bone and the formation of synovial pannus. Figure <xref ref-type="fig" rid="F1">1e</xref> depicts the depletion of cartilage proteoglycan (evidenced by the loss of staining with safranin O). Although destruction was observed only in the injected knee, some proteoglycan loss was also noted in the non-injected, contralateral knee.</p></sec><sec><title>Nociceptive behaviour (pain response)</title><p>At 1&#x02013;29 days after the induction of AIA, the rats showed signs of mechanical hyperalgesia in the leg with inflammation. Most striking were the disturbances of gait (Fig. <xref ref-type="fig" rid="F3">3a</xref>). At 1&#x02013;3 days most rats did not use the leg with inflammation for walking. At 9&#x02013;21 days of AIA, the rats showed marked limping of the leg with the injected knee. At 29 days of AIA, limping was seen only after the knee had been pressed. Normal gait was observed after 37 days of AIA. Furthermore, at 1&#x02013;9 days of AIA, rats did not stand on the hindfeet or visibly loaded only the foot of the leg with the non-injected knee. Exploring behaviour fell from day 2 to day 9 of AIA. When the inflamed knee was manually flexed and extended, the rats showed an immediate defence reaction at 1&#x02013;9 days of AIA. Movements of the contralateral knee did not elicit defence behaviour. Local pressure onto the injected and non-injected knee revealed that the local threshold in the injected knee was markedly reduced at 1&#x02013;9 days of AIA but a reduction was also seen on the contralateral side (Fig. <xref ref-type="fig" rid="F3">3b</xref>). Between 14 and 29 days of AIA, the rats showed a withdrawal response when the injected knee was compressed between two fingers with moderate intensity, whereas strong and painful compression was necessary to evoke a withdrawal response to pressure on the contralateral knee.</p></sec><sec><title>Expression of SP&#x02013;gold- and BK&#x02013;gold-binding sites in DRGs from control and arthritic rats</title><p>Figure <xref ref-type="fig" rid="F4">4</xref> shows neurons that were isolated from DRGs L1&#x02013;L5 3 days after the induction of AIA and then cultured. Neurons labelled with SP&#x02013;gold (Fig. <xref ref-type="fig" rid="F4">4a</xref>) or with BK&#x02013;gold (Fig. <xref ref-type="fig" rid="F4">4b</xref>) appear dark as a result of the silver staining of the gold particles (arrows). The next sections will illustrate the data analysis and summarize the data on the receptor expression.</p></sec><sec><title>SP&#x02013;gold binding sites</title><sec><title>Lumbar DRGs</title><p>Figure <xref ref-type="fig" rid="F5">5</xref> exemplifies the data analysis. In the DRGs of both sides of all animals, control incubations were performed in which SP was administered in excess together with SP&#x02013;gold, to suppress the binding of SP&#x02013;gold. Figure <xref ref-type="fig" rid="F5">5a</xref> displays the distribution of the grey values of neurons of the DRGs of both sides from these control incubations. The grey values were in the range 0.0&#x02013;0.16 (white bars), and thus this range of grey values does not indicate labelling with SP&#x02013;gold. In all other incubations only SP&#x02013;gold was used. In the DRGs of healthy untreated control animals (<italic>n</italic> = 5) approximately 9% of the DRG neurons of both sides exhibited grey values of more than 0.16 (black bars), indicating binding of SP&#x02013;gold (Fig. <xref ref-type="fig" rid="F5">5b</xref>). Similar findings were obtained in DRG neurons from immunized non-arthritic rats (<italic>n</italic> = 3; Fig. <xref ref-type="fig" rid="F5">5c</xref>). By contrast, in DRG from rats (<italic>n</italic> = 5) in which the inflammation had been induced on the ipsilateral side three days previously, a much higher proportion of DRG neurons of both sides exhibited grey values of more than 0.16, indicating binding of SP&#x02013;gold (Fig. <xref ref-type="fig" rid="F5">5d</xref>). These inflammation-induced changes were reversible (Fig. <xref ref-type="fig" rid="F5">5e</xref>).</p><p>Figure <xref ref-type="fig" rid="F2">2a</xref> displays the proportions of DRG neurons that showed binding of SP&#x02013;gold in the different groups of rats. In the untreated control rats (<italic>n</italic> = 5), 7.7&#x000b1; 3.8% of the DRG neurons from the right side (black bars) and 10.0 &#x000b1; 1.7% of the DRG neurons from the left side (white bars) showed labelling with SP&#x02013;gold. The proportion of SP&#x02013;gold-labelled neurons in immunized animals without knee injection (<italic>n</italic> = 3) was similar. By contrast, at days 1 (<italic>n</italic> = 2 rats) and 3 (<italic>n</italic> = 5 rats) of AIA in the right knee, approximately 50% of the DRG neurons exhibited labelling with SP&#x02013;gold, and this was seen both on the side of the injected knee and on the opposite side. At day 10 of AIA (<italic>n</italic> = 3 rats), 26.3 &#x000b1; 6.1% of the ipsilateral DRG neurons but only 15.7 &#x000b1; 0.6% of the contralateral neurons exhibited binding of SP&#x02013;gold. At days 21 (<italic>n</italic> = 5 rats), 42 (<italic>n</italic> = 3 rats) and 84 (<italic>n</italic> = 5 rats) of AIA, the proportion of SP&#x02013;gold-positive neurons had returned to the control values, although the inflammation was still present. In comparison with the DRG neurons of the untreated control rats, the increase in the proportion of labelled neurons was significant on both sides in the acute phase (days 1 and 3) and the intermediate phase (day 10) of AIA (Mann&#x02013;Whitney <italic>U</italic>-test). The size distribution of the neurons was similar in the DRG neurons of all experimental groups. The white bars in Figure <xref ref-type="fig" rid="F6">6</xref> show the distribution of the areas of all neurons measured, and the black insets show the subset of neurons with SP&#x02013;gold-binding sites. Under all conditions and at all time points, SP&#x02013;gold binding was found mainly in small and medium-sized (less than 700 &#x003bc;m²) neurons.</p><p>To show that SP-gold was bound to NK1 receptors we tried to suppress the binding of SP-gold to the neurons by the specific NK1 agonist [Sar⁹, Met(O₂)¹¹]-SP in three experiments. In Fig. <xref ref-type="fig" rid="F7">7</xref> the histograms on the left side show the grey density of the neurons from two experimental groups treated only with SP-gold. A proportion of these neurons exhibited grey values of more than 0.16, indicating binding of SP-gold. The histograms on the right side show that neurons with grey values of more than 0.16 were not found when SP-gold was administered together with the specific NK1 receptor agonist. Thus SP-gold specifically labelled NK1 receptors.</p></sec><sec><title>Cervical DRGs</title><p>In the cervical DRGs the expression of NK1 receptors did not change in the course of AIA (Table <xref ref-type="table" rid="T1">1</xref>). NK1 receptors were expressed in approximately 10% of the neurons in untreated control animals (<italic>n</italic> = 5), in immunized animals without knee injection (<italic>n</italic> = 5), in rats at 3 days of AIA in the knee (<italic>n</italic> = 5), and in 5 rats at 42 days of AIA in the knee.</p></sec></sec><sec><title>BK-gold-binding sites</title><sec><title>Lumbar DRGs</title><p>Figure <xref ref-type="fig" rid="F8">8</xref> shows the distribution of grey values in neurons in which labelling with BK&#x02013;gold was used. Figure <xref ref-type="fig" rid="F8">8a</xref> displays the grey values from neurons of all control incubations with BK&#x02013;gold and BK. The other graphs show specific BK&#x02013;gold-binding sites (grey values of more than 0.16) in neurons from untreated control animals (Fig. <xref ref-type="fig" rid="F8">8b</xref>), from immunized rats without knee injection (Fig. <xref ref-type="fig" rid="F8">8c</xref>) and from rats with AIA (Fig. <xref ref-type="fig" rid="F8">8d</xref> and <xref ref-type="fig" rid="F8">e</xref>).</p><p>Figure <xref ref-type="fig" rid="F2">2b</xref> shows the proportions of neurons with binding of BK&#x02013;gold in the different groups of rats. In untreated control rats (<italic>n</italic> = 5), 42.3 &#x000b1; 3.1% of the DRG neurons of the right side (black bars) and 39.6 &#x000b1; 2.6% of the DRG neurons of the left side (white bars) showed binding of BK&#x02013;gold. At days 1 (<italic>n</italic> = 2 rats) and 3 (<italic>n</italic> = 5 rats) of AIA, approximately 80% of the DRG neurons on the side of the knee injection (ipsilateral) and approximately 70% on the opposite side were labelled. In comparison with the untreated control group the increase in the proportion of labelled neurons was significant on both sides. The proportion of labelled neurons in the ipsilateral DRGs remained significantly increased in both the intermediate phase (day 10, <italic>n</italic> = 3 rats) and chronic phase (days 21, <italic>n</italic> = 5 rats, and 42, <italic>n</italic> = 3 rats) of inflammation. At 84 days after the induction of AIA, 51.0 &#x000b1; 12.7% of the neurons showed an expression of BK&#x02013;gold-binding sites and this was close to the prearthritic values. In the contralateral DRG of the same animals, however, the proportion of BK&#x02013;gold-labelled neurons declined in the intermediate phase (day 10) and chronic phase (days 21&#x02013;84) of AIA and was not significantly different from the control value. Thus the increase in BK&#x02013;gold-labelled neurons was persistent on the side where the inflammation had been induced, and transient on the opposite side.</p><p>The size distributions of the DRG neurons of the different experimental groups were similar (Fig. <xref ref-type="fig" rid="F9">9</xref>). Under all conditions and at all time points, BK&#x02013;gold binding was found mainly in small and medium-sized neurons. The binding of BK&#x02013;gold in these experiments was suppressed by the administration of a mixture of both the B1 and B2 agonists (data not shown).</p><p>In another series of experiments, we made an effort to determine the subtype(s) of BK receptor(s) that were expressed in DRGs L1&#x02013;L5 in different experimental groups. The data are displayed in Table <xref ref-type="table" rid="T2">2</xref>. In neither untreated control animals (<italic>n</italic> = 5) nor immunized rats without knee injection (<italic>n</italic> = 5) nor rats at 3 days (<italic>n</italic> = 5) and 42 days (<italic>n</italic> = 5) of AIA was the binding of BK-gold reduced by the coadministration of BK&#x02013;gold and the B1 agonist. By contrast, in these experimental groups the binding of BK&#x02013;gold was suppressed by the coadministration of the B2 agonist. These data show that in both normal animals and animals with AIA, B2 receptors but not B1 receptors were expressed.</p></sec><sec><title>Cervical DRGs</title><p>In the cervical DRGs the expression of BK receptors did not change in the course of AIA (Table <xref ref-type="table" rid="T1">1</xref>). BK receptors were expressed in approximately 40% of the neurons in untreated control animals (<italic>n</italic> = 5), in immunized animals without knee injection (<italic>n</italic> = 5), in rats at 3 days of AIA in the knee (<italic>n</italic> = 5), and in 5rats at 42 days of AIA in the knee.</p></sec></sec></sec><sec><title>Discussion</title><p>These results show that the expression of SP-binding and BK-binding sites in the perikarya of DRGs L1&#x02013;L5 is markedly upregulated in the course of inflammation in the knee joint. Several aspects are noteworthy. Firstly, although the inflammation was induced on one side only, the initial changes in the binding sites were found in the lumbar DRGs of both sides. No upregulation of SP-binding and BK-binding sites was observed in the cervical DRGs. Secondly, whereas the expression of SP-binding sites was upregulated only in the first days of AIA, that is, in the acute phase, the upregulation of BK-binding sites on the side of AIA persisted for up to 42 days. Thirdly, and unexpectedly, only the B2 receptor, not the B1 receptor, was upregulated. The enhanced expression of NK1 and BK receptors on sensory neurons might be an important mechanism by which arthritis causes pain and hyperalgesia.</p><sec><title>Inflammation and nociceptive behaviour</title><p>The limping during walking and the withdrawal responses to pressure on the joint show that the rats experienced pain during mechanical stimulation of the inflamed joint (mechanical hyperalgesia). Mechanical hyperalgesia is also a leading symptom of human arthritis, and thus AIA in the rat seems to be a suitable model for studying the mechanisms of arthritic pain. Hyperalgesia was most marked in the acute phase and less severe when the acute phase subsided. Interestingly, later stages of AIA were not obviously painful although the chronic inflammation and joint destruction persisted. Although hyperalgesia was pronounced in the knee with inflammation, signs of weak hyperalgesia were also noted on the contralateral side at 1&#x02013;9 days of AIA because the mechanical thresholds were lowered. However, the pain response was quite asymmetrical; the main symptoms were on the injected side.</p></sec><sec><title>Methodological considerations for the use of DRG</title><p>During inflammation, pain is elicited by the activation and sensitization of the sensory endings of nociceptors in the tissue. For technical reasons it is currently impossible to study molecular nociceptive processing and quantitative changes in receptor expression in the sensory endings <italic>in situ</italic>. However, the cell body of DRG neurons is thought to serve as a valid model for the sensory endings, for three reasons:</p><p>1. The perikarya of DRG neurons express ion channels that seem also to be present in the sensory endings, for example ion channels that are activated by noxious heat and other noxious stimuli [<xref ref-type="bibr" rid="B42">42</xref>,<xref ref-type="bibr" rid="B43">43</xref>].</p><p>2. The perikarya of DRG neurons express receptors for ligands that activate the sensory endings in preparations <italic>in vivo</italic> and <italic>in vitro</italic>, for example capsaicin and BK [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>,<xref ref-type="bibr" rid="B47">47</xref>].</p><p>3. Mediators can be released from the perikarya that are usually released from the sensory and spinal endings of the primary afferent fibres [<xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B49">49</xref>,<xref ref-type="bibr" rid="B50">50</xref>].</p><p>We therefore believe that the receptor expression in the cell bodies of DRG neurons also represents the receptor expression in the sensory terminals of the DRG neurons.</p><p>DRG neurons were cultured for 18 h after they had been removed from the rats. In the process of culturing, neurons might be lost, and cultured neurons might show changes compared with neurons <italic>in situ</italic>; this could influence the basal expression of receptors, for example. However, the neurons from different experimental groups were cultured in the same way, and the distribution of the cell sizes was similar in all preparations. Therefore changes in the expression of binding sites are most probably the outcome of pathophysiological changes in the rat during the inflammation.</p></sec><sec><title>Expression and upregulation of SP-binding sites during inflammation</title><p>Evidence was provided that some primary afferent neurons express receptors for SP. Low levels of NK1 receptor mRNA were identified in DRG of the mouse [<xref ref-type="bibr" rid="B30">30</xref>]. Immunohistochemical staining of NK1 receptors has been found in a proportion of unmyelinated axons of rat glabrous skin <italic>in situ</italic> [<xref ref-type="bibr" rid="B31">31</xref>]. In cultured DRG neurons, 10&#x02013;20% of the perikarya show binding of SP&#x02013;gold. Binding of SP&#x02013;gold can be suppressed by the coadministration of a specific NK1 receptor agonist but not of specific NK2 and NK3 receptor agonists [<xref ref-type="bibr" rid="B33">33</xref>]. A similar proportion of cultured DRG neurons show binding of an antibody directed against the carboxy terminus of the NK1 receptor (unpublished data). The application of SP to freshly isolated DRG neurons evokes an inward current [<xref ref-type="bibr" rid="B32">32</xref>], and changes in intracellular Ca²⁺ concentration were seen after the application of SP or NK1 agonists to cultured DRG neurons [<xref ref-type="bibr" rid="B51">51</xref>]. However, other studies failed to identify NK1 receptors in primary afferent neurons [<xref ref-type="bibr" rid="B52">52</xref>,<xref ref-type="bibr" rid="B53">53</xref>,<xref ref-type="bibr" rid="B54">54</xref>]. Thus the expression of functional NK1 receptors in the perikarya of DRG neurons is not entirely settled [<xref ref-type="bibr" rid="B55">55</xref>]. There is still debate on whether or not the effect of SP on sensory endings is important for the generation of pain (see below).</p><p>During inflammation in the periphery, the synthesis of SP in DRGs is upregulated [<xref ref-type="bibr" rid="B56">56</xref>,<xref ref-type="bibr" rid="B57">57</xref>,<xref ref-type="bibr" rid="B58">58</xref>,<xref ref-type="bibr" rid="B59">59</xref>,<xref ref-type="bibr" rid="B60">60</xref>,<xref ref-type="bibr" rid="B61">61</xref>,<xref ref-type="bibr" rid="B62">62</xref>,<xref ref-type="bibr" rid="B63">63</xref>]. More SP is released from the sensory endings in the periphery (producing for example neurogenic inflammation) and from the synaptic endings in the spinal cord [<xref ref-type="bibr" rid="B26">26</xref>]. An upregulation of NK1 receptors during inflammation has been observed in spinal cord neurons [<xref ref-type="bibr" rid="B64">64</xref>,<xref ref-type="bibr" rid="B65">65</xref>,<xref ref-type="bibr" rid="B66">66</xref>,<xref ref-type="bibr" rid="B67">67</xref>]. Here we show that there is also a marked upregulation of SP-binding sites in primary afferent neurons. The low proportion of neurons with SP-binding sites in the normal animal (approximately 10&#x02013;20%) could be the reason why SP was found to affect primary afferent neurons in some studies [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>] but not in others [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B37">37</xref>]. In the acute phase of AIA, approximately 50% of the lumbar DRG neurons showed an expression of SP-binding sites. Because the binding of SP&#x02013;gold can be suppressed by a specific NK1 receptor agonist (this study) but not by specific agonists at the NK2 and NK3 receptor [<xref ref-type="bibr" rid="B33">33</xref>], we believe that the SP&#x02013;gold-binding sites represent NK1 receptors.</p><p>Because peptide receptors are transported to the periphery, the marked upregulation of SP-binding receptors probably leads to an enhanced density of receptors in the sensory endings of the primary afferent units. This allows SP to sensitize more neurons under inflammatory conditions than under normal conditions. It is therefore likely that the upregulation of NK1 receptors contributes to hyperalgesia during arthritis. Indeed, the upregulation of NK1 receptors coincided with the marked pain responses that were seen in the acute phase of inflammation. However, the final proof of the involvement of NK1 receptors in the arthritic pain will require additional functional experiments. In the chronic phase NK1 receptors seem to be less important.</p></sec><sec><title>Expression and upregulation of BK-binding sites during inflammation</title><p>As shown by several methods, primary afferent neurons express BK receptors in their perikarya in the DRGs and also in their peripheral sensory and spinal synaptic endings [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B38">38</xref>,<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B67">67</xref>,<xref ref-type="bibr" rid="B68">68</xref>]. Although the functional effects of SP on nociceptive afferent neurons are still under discussion (see the last paragraph), there is full agreement that BK causes pain in animals and humans. In the present study, DRGs from AIA rats exhibited more neurons with BK-binding sites. Thus, during AIA more primary afferent neurons can be activated and sensitized by the inflammatory mediator BK, and thus the upregulation of the BK-binding sites is likely to be an important mechanism by which arthritis causes severe pain. Because the increased expression of BK-binding sites persisted for up to 42 days on the side of inflammation, BK-binding sites should be an interesting target for pain treatment in the acute and chronic phases of arthritis.</p><p>Only the specific B2 agonist, not the specific B1 agonist, suppressed the binding of BK&#x02013;gold, suggesting that the BK receptors in the DRG in the present study were B2 receptors. This is surprising because previous pharmacological studies have provided evidence that, during inflammation, B1 receptors are newly expressed [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. It is unlikely that the specific B1 agonist did not work in the present experiments for technical reasons, because this B1 agonist reduced the binding of BK&#x02013;gold in another study with the same technique [<xref ref-type="bibr" rid="B68">68</xref>]. Indeed, when BK receptors are upregulated after nerve lesion, a bilateral increase in BK receptors involves the upregulation of B2 receptors and an expression of B1 receptors <italic>de novo</italic> [<xref ref-type="bibr" rid="B68">68</xref>]. We therefore conclude that the increased expression of BK receptors in AIA is entirely due to an upregulation of the B2 receptor.</p></sec><sec><title>Bilateral effects and possible mechanisms of upregulation of receptors</title><p>Receptor upregulation in the acute phase of AIA was bilateral and almost symmetrical. However, the hyperalgesia was much more pronounced on the inflamed side. That the pain responses were asymmetrical in spite of a symmetrical receptor upregulation is most probably due to the fact that the receptors on the contralateral side were not readily activated because in the absence of gross inflammation the local concentration of the ligands BK and SP was probably quite low.</p><p>At present it is unclear which mechanisms cause the upregulation of BK-binding and SP-binding sites. Because the expression of BK receptors in cultured DRG neurons can be regulated, for example by nerve growth factor (NGF) [<xref ref-type="bibr" rid="B39">39</xref>], we assume that mediators of different sources (such as immune cells and nerve cells) could have a role. NGF could be a candidate because inflamed tissue shows a rapid enhancement of NGF [<xref ref-type="bibr" rid="B69">69</xref>,<xref ref-type="bibr" rid="B70">70</xref>,<xref ref-type="bibr" rid="B71">71</xref>]. NGF can also regulate the synthesis of neuropeptides such as SP [<xref ref-type="bibr" rid="B72">72</xref>,<xref ref-type="bibr" rid="B73">73</xref>]. Although such mechanisms could explain the upregulation of receptors on the side of inflammation, it is intriguing to consider which mechanisms are possibly involved in the upregulation of the BK-binding and SP-binding sites on the contralateral side. One possibility is that the process of disease was spreading to the contralateral side and thereby stimulated receptor upregulation in contralateral DRGs. Indeed, during unilateral inflammation, cellular infiltration [<xref ref-type="bibr" rid="B74">74</xref>], SP and calcitonin gene-related peptide upregulation [<xref ref-type="bibr" rid="B55">55</xref>], an increase in the interleukin-6 content [<xref ref-type="bibr" rid="B75">75</xref>], a loss of proteoglycans [<xref ref-type="bibr" rid="B10">10</xref>] and altered load-bearing [<xref ref-type="bibr" rid="B76">76</xref>] have been observed on the contralateral side. However, during AIA the contralateral knee did not show gross inflammatory alterations; it is therefore questionable whether the enhanced receptor expression in the contralateral DRGs was due to the spreading of the inflammatory disease to the contralateral side.</p><p>Because the upregulation was symmetrical and segmental, it is also possible that the upregulation was caused by the symmetrical innervation. In general, bilateral changes in the nervous system after unilateral pathologies such as nerve lesions and inflammation are not uncommon, although they are not always observed [<xref ref-type="bibr" rid="B77">77</xref>,<xref ref-type="bibr" rid="B78">78</xref>]. Symmetrical efferent effects on the tissue and/or the DRGs can be provided by the sympathetic nervous system [<xref ref-type="bibr" rid="B79">79</xref>,<xref ref-type="bibr" rid="B80">80</xref>] or by the primary afferent fibres in which dorsal root reflexes are generated. Dorsal root reflexes are action potentials that originate in the spinal cord entry zone of the afferent fibres and propagate from there to the periphery [<xref ref-type="bibr" rid="B81">81</xref>,<xref ref-type="bibr" rid="B82">82</xref>,<xref ref-type="bibr" rid="B83">83</xref>]. Both of these neuronal pathways depend on activity in the spinal cord; it has been shown that the development of joint inflammation can be attenuated by the spinal application of glutamate receptor antagonists that reduce the activity of the spinal cord neurons [<xref ref-type="bibr" rid="B84">84</xref>, <xref ref-type="bibr" rid="B85">85</xref>]. The involvement of the central nervous system in the generation of bilateral inflammation [<xref ref-type="bibr" rid="B86">86</xref>] and bilateral degeneration of articular cartilage [<xref ref-type="bibr" rid="B87">87</xref>] has been shown; it is therefore conceivable that efferent activity from the central nervous system to the periphery might also influence the expression of receptors in DRG neurons.</p></sec></sec>

Mesenchymal precursor cells in the blood of normal individuals

<sec><title>Introduction:</title><p>Adult human bone marrow contains a minority population of MSCs that contribute to the regeneration of tissues such as bone, cartilage, muscle, ligaments, tendons, fat, and stroma. Evidence that these MSCs are pluripotent, rather than being a mixture of committed progenitor cells each with a restricted potential, includes their rapid proliferation in culture, a characteristic morphology, the presence of typical marker proteins, and their consistent differentiation into various mesenchymal lineages. These attributes are maintained through multiple passages and are identifiable in individual stem cells.</p></sec><sec><title>Aims:</title><p>Since stem cells are present in both the bone marrow and other tissues, we thought it possible that cells with a similar appearance and pluripotent mesenchymal potential would be present in the blood. We applied techniques used successfully with marrow MSCs to identify similar cells in elutriation fractions of normal human blood.</p></sec><sec><title>Methods:</title><p>BMPCs were elutriated by diluting the buffy coats from 500 ml of anticoagulant-treated, platelet-depleted blood 1:4 in RPMI-1640 medium (RPMI) and layering 25-ml portions over 20 ml of Lymphoprep^&#x02122;. These samples were centrifuged at 2000 rpm for 20 min. The leukocyte-rich interface cells were collected, made up to 20 ml in RPMI, and separated by density-gradient centrifugation. The interface cells, now depleted of red blood cells, were collected, resuspended in 50 ml of sterile RMPI and 5% heat-inactivated FCS, and introduced into the sample line of the flow system of a Beckman JE-50 cell elutriator charged with elutriation buffer. The chamber was centrifuged at 25 000 rpm at 10&#x000b0;C and the flow rate adjusted to 12 ml/min. After about 150 ml had been collected, the flow rate was increased by 1 ml/min. Fractions nos. 1-6 (flow rates of 12-16 ml/min) contained most of the lymphocytes. Monocytes usually appeared in fractions 6 or 7 (as determined by flow cytometric analysis in a fluorescence-activated cell sorter (FACS). BMPCs were concentrated in fractions 7 and 8, along with monocytes and lymphocytes. Elutriation fractions with more than 50% and less than 75% monocytes were collected and concentrated by centrifugation at 1200 rpm for 5 min, and the cell pellets were combined, reconstituted in DMEM plus 20% sterile heat-inactivated FCS, counted, washed in medium, repelleted, and then resuspended in DMEM to 5 &#x000d7; 10⁶/ml and dispensed into either tissue-culture plastic slides or glass chamber slides. Cells thus obtained were observed in time-lapse cinematography, assayed for proliferation, and examined immunohistologically and histochemically, and their ability to become fibroblasts, osteoclasts, osteoblasts, and adipocytes was documented.</p></sec><sec><title>Results:</title><p>BMPCs were found in elutriation fractions containing less than 30% T cells and more than 60% monocytes from the blood of more than 100 normal persons. BMPCs adhered to plastic and glass and proliferated logarithmically in DMEM-20% FCS without added growth factors. The initial elutriate had only small, round, mononuclear cells; upon culture, these were replaced by fibroblast-like cells and large, round, stromal cells. The formation of cells with fibroblast-like and stromal morphology was not affected by eliminating CD34, CD3, or CD14 cells from the elutriation fraction. Osteogenic supplements (dexamethasone, ascorbic acid, and &#x003b2;-glycero-phosphate) added to the culture inhibited fibroblast formation, and BMPCs assumed the cuboidal shape of osteoblasts. After 5 days in supplemented medium, the elutriated cells displayed AP and its production was doubled by the addition of BMP2 (1 ng) (<italic>P</italic> &#x0003c; 0.04). Two weeks later, 30% of the cells were very large and reacted with anti-osteocalcin antibody. The same cultures contained two other types of cell: sudanophlic adipocytes and multinucleated giant cells, which stain for TRAP and vitronectin receptors (attributes of osteoclasts). Cultured BMPCs were immunostained by antibodies to vimentin, type I collagen, and BMP receptors (heterodimeric structures expressed on mesenchymal lineage cells). The cultured cells also stained strongly for the SH-2 (endoglin) antigen, a putative marker for marrow MSCs. BMPCs express the gene for SDF-1, a potent stroma-derived CXC&#x003b1; chemokine.</p></sec><sec><title>Discussion:</title><p>In the circulation of normal individuals is a small population of CD34^- mononuclear cells that proliferate rapidly in culture as an adherent population with a variable morphology, display cytoskeletal, cytoplasmic, and surface markers of mesenchymal precursors, and differentiate into several lineages (fibroblasts, osteoblasts, and adipocytes). These are all features found in bone-marrow-derived MSCs. Therefore, autologous blood could provide cells useful for tissue engineering and gene therapy. In addition, the demonstration of similar cells in the inflammatory joint fluids and synovium of patients with rheumatoid arthritis (RA) suggests that these cells may play a role in the pathogenesis of RA.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Zvaifler</surname><given-names>Nathan J</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>nzfaifler@ucsd.edu</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Marinova-Mutafchieva</surname><given-names>Lilla</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Adams</surname><given-names>Gill</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Edwards</surname><given-names>Christopher J</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Moss</surname><given-names>Jill</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Burger</surname><given-names>Jan A</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Maini</surname><given-names>Ravinder N</given-names></name><xref ref-type="aff" rid="I2">2</xref><xref ref-type="aff" rid="I3">3</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Bone marrow is a complex tissue containing hematopoietic cell progenitors and their progeny and a connective-tissue network of mesenchymally derived cells known as stroma. Marrow stroma includes a subpopulation of undifferentiated cells that are capable of becoming one of a number of phenotypes, including chondrocytes, osteoblasts, adipocytes, fibroblasts, possibly muscle cells, and the reticular cells that support hematopoietic cell differentiation [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. Extensive experimentation has defined conditions for the isolation, propagation, and differentiation <italic>in vitro</italic> and <italic>in vivo</italic> of the stromal cells referred to as MSCs. They are a population of firmly adherent cells with a high proliferative capacity and potential for self-renewal. Their developmental potential is retained even after repeated subcultivation <italic>in vitro</italic>, supporting their designation as stem cells [<xref ref-type="bibr" rid="B3">3</xref>].</p><p>Identification of MSCs <italic>in situ</italic> has been difficult, partly because they have few unique products or molecular markers. A series of monoclonal antibodies (SH antibodies) purportedly specific reagents have been used to isolate MSCs from a population of bone-marrow cells [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. The one used most often (SH-2) was recently shown to react with endoglin (CD105), a member of the transforming growth factor (TGF)-&#x003b2; receptor family usually found on the endothelium of postcapillary venules [<xref ref-type="bibr" rid="B4">4</xref>]. Two other reagents may be more specific. One consists of a group of antibodies to BMP receptors (BMPRs) present on embryonic mesenchyme and postnatally on osteoblasts and chondrocytes [<xref ref-type="bibr" rid="B5">5</xref>]. Another antibody, Stro-1 made against marrow fibroblastic cells, blocks hematopoiesis <italic>in vitro</italic> by interfering with the interaction of reconstituted human hematopoietic stem cells (HSCs) and stromal cells [<xref ref-type="bibr" rid="B6">6</xref>].</p><p>Attempts to demonstrate MSCs in peripheral blood have been unrewarding, except for a report by Fernandez <italic>et al</italic> [<xref ref-type="bibr" rid="B7">7</xref>], who identified cells with the features of MSCs in growth-factor-mobilized peripheral-blood cells from breast-cancer patients. Low-density mononuclear cells grown for 1 week in tissue culture with fetal calf serum (FCS) become adherent fibroblast-like cells and a few were large, flat, round cells. Immunohistology and flow cytometric analysis in a fluorescence-activated cell sorter (FACS) revealed fibronectin and three types of collagen (I, III, and VI) in the cytoplasm of the cultured cells. They expressed adhesion ligands and antigens recognized by SH-2 and SH-3 monoclonal antibodies. No stromal cells were demonstrated in normal peripheral-blood cells not mobilized by granulocyte-macrophage CSF [<xref ref-type="bibr" rid="B7">7</xref>]. Bucala <italic>et al</italic> [<xref ref-type="bibr" rid="B8">8</xref>] separated human blood cells by density centrifugation, cultured them on a fibronectin matrix, and identified a population of circulating cells that had fibroblast properties and a distinctive phenotype (collagen⁺/vimentin⁺/ CD34⁺). This novel circulating cell, termed a fibrocyte, has both mesenchymal and hematopoietic features.</p><p>Now we report for the first time that cells with the morphology and phenotype of mesenchymal precursors are normally present in the circulation. Hereafter these are referred to as `blood-derived mesenchymal precursor cells' (BMPCs). The observations that support these conclusions and the significance of the findings are discussed.</p></sec><sec sec-type="methods"><title>Methods</title><sec><title>Reagents and laboratory ware</title><p>Dexamethasone, ascorbic acid-2 phosphate, &#x003b2;-glycerophosphate, bovine serum albumin (BSA), and FCS were purchased from Sigma Diagnostics (St Louis, MO, USA); penicillin, streptomycin, DMEM, and RPMI-1640 from Biowhittaker (Watersville, MD, USA); Lymphoprep<sup>&#x02122;</sup> from Nycomed, Oslo, Norway.</p><p>Monoclonal antibodies were purchased from commercial vendors unless otherwise stated: CD3, IgG1; CD68, IgG1; CD34, IgG1; CD45, IgG1, D105, IgG1; and IgG2a controls (Dako Corporation, Carpinteria, CA, USA); anti-HLA-DR, IgG2a; CD14, IgG2b; CD34, IgG1 (Becton Dickinson, San Jose, CA, USA); anti-vimentin, IgG1; IgG2b control (Serotec, Kidlington, Oxfordshire, UK); anti-VCAM-1, IgG1; anti-&#x003b1;v&#x003b2;3 (vitronectin receptor), IgG1 (Pharmingen, San Diego, CA, USA); anti-collagen-type-1, IgG1 (Sigma Diagnostics); anti-osteocalcin, IgG1 (ABOC-5021, Haematologic Technologies, Essex Junction, VT, USA) anti-IgG1. Stro-1 is a culture supernatant, monoclonal IgM, from Developmental Studies Hybridoma Bank, University of Iowa (Iowa City, IA, USA). Biotinylated mouse Ig, streptavidin/horseradish peroxidase conjugate, diamino benzidine, and Vectastain ABC were from Vector (Burlingame, CA, USA).</p><p>Tissue-culture treated glass slides, Petri dishes, and six-well tissue-culture plates and eight-chamber tissue culture slides (Falcon) were from Becton Dickinson Labware (Franklin Lakes, NJ, USA); 12-well sterile glass slides were from ICN (Costa Mesa, CA, USA).</p></sec><sec><title>Elutriation procedure for BMPCs</title><p>Anticoagulant-treated, platelet-depleted buffy coat was obtained in sterile packages from the North London Blood Transfusion Service. About 50 ml of the buffy coat was diluted 1:4 in RPMI and 25 ml was layered over 20 ml of Lymphoprep<sup>&#x02122;</sup> in a 50-ml conical centrifuge tube. The tubes (approximately eight) were centrifuged at 2000 rpm for 20 min. The supernatant was discarded and leukocyte-rich interface cells were collected and combined. These were made up to 20 ml in RPMI and layered again over Lymphoprep<sup>&#x02122;</sup> and centrifuged at 2000 rpm for 20 min more. The buffy coat, now depleted of red blood cells, was collected from the interface, resuspended in 50 ml of sterile RMPI and 5% heat-inactivated FCS, and introduced into the sample line of the flow system of a Beckman JE-50 cell elutriator which had been charged with elutriation buffer. The chamber was centrifuged at 25 000 rpm at 10&#x000b0;C and the flow rate adjusted to 12 ml/min. The eluate fractions were collected in sterile, conical, 50-ml tubes. After about 150 ml had been collected, the flow rate was increased by 1 ml/min. Fractions nos. 1-6 (flow rates of 12-16 ml/min) contained most of the lymphocytes. Monocytes usually made their appearance (as determined by FACS analysis) in fraction 6 or 7. In fractions 7 and 8, monocytes constituted up to two thirds of the cells, and BMPCs were concentrated in these fractions. Elutriation fractions containing more than 50% and less than 75% monocytes were concentrated by centrifugation at 1200 rpm for 5 min, and the cell pellets were combined, reconstituted in DMEM plus 20% sterile heat-inactivated FCS (hereafter referred to as complete medium, unless otherwise stated), counted in a hemocytometer, washed in DMEM medium, repelleted, resuspended to 5 &#x000d7; 10<sup>6</sup>/ml, and dispensed into either plastic tissue-culture slides or glass chamber slides.</p><p>More than 100 consecutive buffy coats from normal individuals were processed by this method and cultured. In every case, the appropriate elutriation fractions had cells with the BMPC morphology.</p></sec><sec><title>Cell-proliferation assay</title><p>The BMPC-rich elutriation fractions were plated at 5 &#x000d7; 10<sup>5</sup> in 500 &#x003bc;l of complete medium in polystyrene chambers on treated glass tissue-culture slides (Falcon). At various times, cultures were rinsed twice with Tyrode's balanced salt solution, fixed with 1% glutaraldehyde (v/v) in Tyrode's for 15 min, rinsed twice with deionized water, and air-dried. Cultures were then stained with 0.1% crystal violet (w/v) in deionized water for 30 min and washed 3 times with deionized water; the crystal-violet dye was extracted by rocking the cultures gently in 1% Triton X-100 for 4 h at room temperature and read at 595 nm on a microplate reader (BioRad, Hercules, CA, USA). Absorbance values (optical densities; ODs) were converted into absolute cell numbers on the basis of established standard curves [<xref ref-type="bibr" rid="B9">9</xref>].</p></sec><sec><title>Immunohistochemistry</title><p>BMPC-rich elutriation fractions (500 &#x003bc;l, containing 5 &#x000d7; 10<sup>6</sup> cells per ml) or other sources of BMPCs were placed into the wells of sterile, 12-well multitest slides (ICN) in complete medium and left to adhere at 37&#x000b0;C for 4 h. The slides were then placed into 100 &#x000d7; 20 mm Petri dishes containing 5-7 ml DMEM-20% FCS. The nonadherent cells floated off, while mesenchymal cells adhered, spread, and grew. Their daily progress was assessed by phase-contrast microscopy. The medium was changed every 3 to 5 days and the cells were studied after 5-7 days. The growing, adherent cells were rinsed in phosphate-buffered saline (PBS), fixed in ice-cold 4% paraformaldehyde for 20 min, and then washed in PBS. All further incubations and washes were carried out using PBS. Endogenous peroxidase activity was blocked with 0.1 mol sodium azide containing 1% hydrogen peroxide, and the specimens were incubated with 10% normal goat serum, 2% normal rat serum, and 1% bovine serum albumin for 30 min at room temperature to eliminate non-specific binding. Specimens were then incubated with primary antibodies at 4&#x000b0;C overnight and were then incubated with a biotinylated secondary antibody (Vector). The antibody-biotin conjugates were detected with an avidin-biotin-peroxidase complex (Vector), applied for 30 min at room temperature. A color reaction was developed with 3-amino-9-ethylcarbazole and specimens were lightly counterstained with Mayer's hematoxylin.</p><p>Controls included normal rabbit or mouse IgG, 1% BSA in PBS, or, in the case of BMPR antibodies, preabsorbed with the respective peptide used for immunization.</p></sec><sec><title>Quantification of BMPCs by immunohistochemistry</title><p>BMPC-rich elutriation fractions (5 &#x000d7; 10<sup>5</sup> cells in 500 &#x003bc;l of complete medium) were placed into the chambers of sterile, eight-chamber, treated glass tissue-culture slides. Two to 4 h later, nonadherent cells were removed. Cultures were fed every 3 days. At regular intervals, the slides were rinsed in PBS, fixed in ice-cold 4% paraformaldehyde for 20 min, washed in PBS, stained with anti-BMPR antibodies, and visualized by the ABC immunoperoxidase method described above. The specimens were examined using an Olympus BH-2 microscope and analyzed by computer image analysis (AnalySIS, Soft Imaging System GmbH, M&#x000fc;nster, Germany). Six digital images (400&#x000d7;) per specimen were recorded and quantitative analysis was performed according to the color cell separation. Images chosen at random were analyzed and the data are presented as the mean of the total number of cells per six images examined at 400&#x000d7;. Slender cells with a small, centrally localized nucleus were scored as fibroblast-like. Large, round cells and intermediate-sized cells with more cytoplasm and a large, round nucleus were scored as large cells.</p></sec><sec><title>Anti-BMPR antibodies</title><p>Rabbit polyclonal antibodies to BMPRs were provided by K Funa (G&#x000f6;teborg University, Gothenburg, Sweden). Polyclonal rabbit antisera were prepared using synthetic peptides corresponding to the intracellular transmembrane portions of the types IA, IB, and II BMPRs [<xref ref-type="bibr" rid="B10">10</xref>]. The antisera were affinity-purified and tested for specificity by immunoprecipitation of cross-linked complexes of cultured cells transfected with receptor complementary deoxyribonucleic acids (cDNAs) [<xref ref-type="bibr" rid="B11">11</xref>].</p><p>Magnetic antibody-coated-bead separation (MACS) was performed in accordance with the manufacturer's recommendations (Miltenyi Biotec, Inc., Auburn, CA, USA). Elutriation fractions with 50-75% monocytes were centrifuged at 900 &#x000d7; <bold> <italic>g</italic> </bold>, washed with MACS buffer (PBS pH7.2, + 0.5% SSA + 2 mmol EDTA) and counted in a hemocytometer. The cell pellet was resuspended in 80 &#x003bc;l MACS buffer per 10<sup>7</sup> total cells, and 20 &#x003bc;l of MACS antibody-coated beads was added to the cells, mixed, and incubated for 15 min at 6-12&#x000b0;C. The cells were washed with a 20-fold volume of MACS buffer, spun, and resus-pended in 500 &#x003bc;l buffer. The cell suspension was applied to a positive selection column washed previously with 1 ml MACS buffer and placed in a magnetic separator and the cells were eluted. The column was rinsed four times with 500 &#x003bc;l buffer and the cells that passed through were combined as the antigen-free fraction. The column was removed from the magnetic separator, 1 ml of buffer was added to the column, and the positive cells were flushed out with a syringe plunger. This was repeated with another 1 ml of buffer. The elutriated cells were combined as the antigen-containing fraction.</p></sec><sec><title>Alkaline phosphatase activity of circulating BMPCs</title><p>BMPC-rich elutriation fractions were prepared from four individual blood packs as described and plated into four-well chamber slides (Lab-Tek) at 5 &#x000d7; 10<sup>6</sup> cells per ml in DMEM-10% FCS. After 24 h at 37&#x000b0;C, the nonadherent cells were removed and new medium was added containing BMP2 (a gift from the Genetics Institute, Cambridge, MA, USA) at concentrations of 0, 1, 10, or 100 ng/ml. The cells were incubated at 37&#x000b0;C in 5% CO<sub>2</sub> and the medium was changed every 5 days. Supernatants were taken at 5, 10, and 15 days and stored at -20&#x000b0;C for later analysis. AP activity in the supernatants was estimated using a <italic>p</italic>-nitro-phenol colorimetric assay. Cell supernatants were assayed for AP activity in 50 mmol glycine, 0.05% Triton X-100, 4 mmol MgCl<sub>2</sub>, and 5 mmol <italic>p</italic>-nitrophenol phosphate, pH10.3, for 15 min at 37&#x000b0;C (Sigma Diagnostics, St Louis, MO). OD was measured at 405 nm and compared with that of standards.</p></sec><sec><title>Stromal-cell-derived factor (SDF)-1 RT-PCR</title><p>RNA was isolated from BMPCs cultured for 7-12 days using the Qiagen RN Easy kit (Qiagen, Santa Clarita, CA, USA). RNA was then used for the first-strand cDNA synthesis in the SuperScript Preamplification System (GIBCO, BRL, Rockville, MD, USA) in accordance with the manufacturer's instructions. SDF-1 specific primers: 5'-GAGGATCCGACGGGAAGCCC-GTCAGC; 3'-GAA-TTCACATCTTGAACCTCTTG. The annealing temperature was 58&#x000b0;C and the reaction proceeded for 35 cycles. The glyceraldehyde-3-phosphate dehydrogenase (GA3PD) gene was included as a reverse transcriptase polymerase chain reaction (RT-PCR) control and performed under similar conditions to normalize for the amount of RNA. Reaction products were analyzed in 2% agarose gel containing 0.25 mg/ml ethidium bromide.</p></sec><sec><title>Data analysis and statistics</title><p>Results are shown as the standard error about the mean (SEM) of at least three experiments each. For statistical comparison between groups, the Student paired <italic>t</italic> test or Bonferroni <italic>t</italic> test was used. Analyses were performed using the Biostatistics software developed by Stanton A Glantz (UC San Francisco, CA, USA). Flow cytometry data were analyzed using the FlowJo software.</p></sec></sec><sec><title>Results</title><sec><title>BMPCs selected by elutriation of normal human blood</title><p>When the elutriated cells from the fractions between the smaller T cells and the larger, more granular monocytes were cultured in complete medium without any other supplements, they appeared small and round on examination by phase-contrast microscopy. Some of them were nonadherent (presumably T lymphocytes) and were removed with the initial feeding of the culture. After 72 h, elongated cells with a fibroblast-like morphology and large cells with a clear, thin, adherent cytoplasm around a central nucleus made their appearance. Over the ensuing 7-14 days, they became the predominant cells in the culture (Fig. <xref ref-type="fig" rid="F1">1a</xref>). At higher magnification they could often be seen to have a splayed, spreading cap at the end and a small, central nucleus. Another cell population, consisting of larger and wider cells, with more cytoplasm and a larger nucleus, was intermediate in morphology between the large, round cells and the thinner, fibroblast-like cells (Fig. <xref ref-type="fig" rid="F1">1b</xref>). Culture conditions modified the morphology of the elutriated cells. Adding dexamethasone (100 nmol) at the initiation of the culture significantly reduced (by 60% &#x000b1; 10%) the total number of cells at the 7th to 10th days and decreased the formation of fibroblast-like cells (data not shown). Cultures supplemented with a mixture of 100 nmol dexamethasone, 0.05 mmol ascorbic acid-2-phosphate, and 10 mmol &#x003b2;-glycerophosphate (conditions that favor the development of osteoblasts) developed only round or cuboidal cells, and not fibroblast-like ones (Fig. <xref ref-type="fig" rid="F1">1c</xref>). Dexamethasone alone added at days 6 to 8 reduced fibroblast numbers, but not the total cell numbers (data not shown).</p></sec><sec><title>BMPC-rich elutriation fractions observed in time-lapse cinematography</title><p>Clusters of small round cells formed within 24 h. Cell processes occasionally extended from them, but these retracted minutes later (Fig. <xref ref-type="fig" rid="F2">2a</xref>). Individual cells were motile and often left the field, but the clusters remained intact. After 72 h, a few cells with a fibroblast-like morphology could be seen beneath and at the edges of the clusters. The fibroblast-like cells were much larger than the initial cells and quite mobile, extending and retracting usually about a broad, fixed cup, or pseudopod. By 6 days, a significant portion of the cells retained their elongated form and looked like the cells in Fig. <xref ref-type="fig" rid="F2">2b</xref>. Large, round cells were also present. Thus, it appears that BMPCs in the circulation were present as small, round mononuclear cells and their subsequent morphology and function were dictated by culture conditions.</p></sec><sec><title>Cell numbers in the BMPC-rich elutriation fraction from 500 ml of normal human blood</title><p>Elutriation fractions were selected for quantification of BMPCs based on cell size (intermediate between lymphocytes and monocytes) and granularity (FACS). This population comprised less than 35% lymphocytes and more than 50% monocytes. Nineteen consecutive samples had an average total cell number of 2.14 &#x000b1; 0.22 (SEM) &#x000d7; 10<sup>7</sup>, of which 64.4% &#x000b1; 1.5% (SEM) were monocytes. A sub-population, estimated as 0.3-0.7% of the starting elutriation fractions, was judged to consist of BMPCs on the basis of their morphology, their strong adherence to plastic or glass, and their ability to proliferate in DMEM-20% FCS without added growth factors (ie &#x0003c;1% of the starting 2 &#x000d7; 10<sup>7</sup> elutriated cells represents 1000 to 10 000 BMPCs). Therefore, it is likely that 500 ml of normal blood will have several thousand BMPCs.</p><p>Cultures were established with 5 &#x000d7; 10<sup>5</sup> cells from the elutriation fractions and proliferation was measured on days 3, 10, and 17. Nonadherent cells were removed in the first 24 h and the cultures were fed twice weekly. The cells grew logarithmically, with an approximate doubling time of 2.5 days. By day 17, the initial 5 &#x000d7; 10<sup>5</sup> cells multiplied to 6.7 &#x000d7; 10<sup>7</sup> (Table <xref ref-type="table" rid="T1">1</xref>). The culture conditions are not conducive to growth of lymphocytes or monocytes; therefore, by week 3, most of the proliferating population in cultures were mesenchymal cells (&#x0003c;20% CD14-staining cells; data not shown).</p><p>Cells from a BMPC-rich elutriation fraction of healthy human blood were cultured in DMEM-20% FCS. At days 3, 5, 8, and 11 the cultured cells were fixed, stained with anti-BMPR antibodies, and quantified with an autoanalyzer (described in Methods). The results are presented in Fig. <xref ref-type="fig" rid="F3">3</xref> as the means of the total number of cells in six individual images. Big cells plus fibroblast-like cells constituted 37% to 59% of the total. Cells with a fibroblast-like morphology varied from 21% to 34% of the BMPCs. The 1:2 ratio of fibroblast-like to big cells remained relatively stable over 11 days of culture, even as the total cell numbers increased (see Table <xref ref-type="table" rid="T1">1</xref>).</p></sec><sec><title>Immunohistology of BMPCs</title><p>Elutriated cells were grown on sterile 12-well glass slides (ICN) in DMEM-20% FCS and 6 to 12 days later the cells were fixed, stained, and examined by immunohistology (Table <xref ref-type="table" rid="T2">2</xref>). The large cells and the fibroblast-like cells stained with antibodies to both vimentin and collagen type I. They were also identified by antibodies to one or the other chain (type IA and type II) of the BMPR (Fig. <xref ref-type="fig" rid="F4">4a</xref>), but did not react with an anti-type-IB-receptor-chain antibody. Monoclonal Stro-1 antibody stained most of the large BMPCs and a few of the fibroblast-like cells, while anti-CD105 reacted with both populations (Fig. <xref ref-type="fig" rid="F4">4b</xref>). BMPCs stained strongly with anti-CD44 antibody. Conventional T-cell (CD3), monocyte (CD14, CD68), and B-cell (CD20) antibodies stained neither of the two BMPC populations, nor did they react to anti-LCA (CD45), anti-VCAM (CD106), or MHC-Class II (anti-DR).</p></sec><sec><title>Cell separation by magnetic beads to analyze the contribution of CD34+ progenitors, monocytes, and T lymphocytes to BMPC formation</title><p>BMPC-rich elutriation fractions were incubated with magnetic beads coated with specific antibodies and subsequently separated into adherent (antigen-enriched) and nonadherent (antigen-depleted) populations. These were cultured in complete medium in six-well plates and observed daily by phase-contrast microscopy. The CD34-depleted fraction always developed many examples of both types of mesenchymal cells (seven experiments). The fibroblast-like cells appeared in the CD34-depleted cultures at the same time as in untreated controls (usually day 3 or 4). There were not enough CD34<sup>+</sup> cells to establish cultures. A representative experiment is shown in Table <xref ref-type="table" rid="T3">3</xref>. In two additional experiments, the cells from the elutriation fraction were exposed to CD34 beads and the negative fraction was separated again on fresh CD34 beads. There was no reduction in the time of appearance or number of fibroblast-like cells. Thus, although BMPCs were present in an elutriation fraction that contained CD34<sup>+</sup> cells, the two types of cell could be distinguished from one another.</p><p>Findings with anti-CD14 beads were somewhat different. Both CD14<sup>+</sup> and CD14<sup>-</sup> eluates developed fibroblast-like cells in culture, but the numbers were less than in unfractionated controls. In four of five experiments, the fibroblast-like cells in the CD14<sup>+</sup> fraction appeared sooner and in greater numbers than the CD14<sup>-</sup>fraction. In all instances, however, each fraction contained both large, round cells and fibroblast-like cells and their numbers become more equal with time (usually by day 13). A representative experiment is shown in Table <xref ref-type="table" rid="T3">3</xref>. When the CD14<sup>+</sup> and CD14<sup>-</sup> populations were combined, the number of fibroblast-like cells and their time of appearance was the same as in unfractionated controls.</p><p>T-cell depletion had no effect on the number, morphology, or time of appearance of mesenchymal cells (data not shown).</p></sec><sec><title>Pluripotent BMPC-fibroblast, osteoblast, and adipocyte formation</title><p>BMPC-rich elutriation fractions cultured in complete medium supported growth of fibroblasts (spindle-shaped cells stained by antibodies to vitronectin and type I collagen). When the same elutriated cells were supplemented with dexamethasone, ascorbic acid, and &#x003b2;-glycerophosphate their morphology altered, and they became uniform, polygonal cells reminiscent of osteoblasts (Fig. <xref ref-type="fig" rid="F1">1c</xref>). By 10 days, many of the cells stained for AP (not shown). Over the next 1 to 2 weeks, a subpopulation (approximately 30%) of large cells (which were 3- to 6-fold bigger than monocytes) developed. They accumulated an ill-defined pericellular matrix and the osteoblast-specific protein osteocalcin (Fig. <xref ref-type="fig" rid="F5">5a</xref> and <xref ref-type="fig" rid="F5">b</xref>). In the same supplemented cultures were sudanophilic adipocytes (Fig. <xref ref-type="fig" rid="F5">5a</xref>). Another large cell present after 1 week in the supplemented cultures had multiple nuclei (Fig. <xref ref-type="fig" rid="F6">6a</xref>) and stained for tartrate-resistant acid phosphatase (TRAP) and the vitronectin receptor (Fig. <xref ref-type="fig" rid="F6">6b</xref>), which are features of osteoclasts (OCs).</p><p>OCs developed in the supplemented BMPC cultures because both monocytes and stem cells (and/or pre-osteoblasts) were present together in the elutriated cells.</p></sec><sec><title>Alkaline phosphatase production</title><p>BMPC-rich elutriation fractions from four separate blood packs were cultured in complete medium with varying concentrations of BMP2 for 5 days and AP in the supernatants was measured (Fig. <xref ref-type="fig" rid="F7">7</xref>). The lowest concentration of BMP2 (1 ng/ml) caused a significant increase in AP activity (<italic>P</italic> = 0.004). This represents an increased AP production per cell, because BMP2 had no effect on BMPC proliferation over 5 days (data not shown).</p></sec><sec><title>Stromal-cell-derived factor 1</title><p>SDF-1 is a potent CXC&#x003b1; chemokine produced by bone-marrow spindle-shaped stromal cells and other cells of mesenchymal origin, but not blood leukocytes [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. RT-PCR analysis for SDF-1 mRNA expression in two cDNA samples of cultured BMPCs (lanes 2 and 3) and in a cDNA from RA synoviocytes (lane 4) are shown in Fig. <xref ref-type="fig" rid="F8">8</xref>.</p><p>Each sample displayed an amplification of a PCR fragment of the expected size for SDF-1 (296 bp) and was similar to the positive control (lane 5), a sequenced plasmid containing human SDF-1&#x003b2; cDNA.</p></sec></sec><sec><title>Discussion</title><p>The abilities to self-replicate and to give rise to daughter cells that undergo an irreversible terminal differentiation are features of stem cells [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. The best-characterized are HSCs and their progeny. Friedenstein <italic>et al</italic> proposed a similar scheme for mesenchymal cells and showed that bone marrow contained primitive cells that could generate progenitors committed to one or another mesenchymal line [<xref ref-type="bibr" rid="B2">2</xref>]. Such cells are called MSCs [<xref ref-type="bibr" rid="B1">1</xref>]. Conditions that direct marrow MSCs along myogenic [<xref ref-type="bibr" rid="B16">16</xref>], adipogenic [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>], osteogenic [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>], chondrogenic [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B21">21</xref>], and stromal pathways [<xref ref-type="bibr" rid="B22">22</xref>] have been defined. For instance, exposure of fibroblast-like marrow MSCs <italic>in vitro</italic> to optimal concentrations of dexamethasone, ascorbic acid and &#x003b2;-glycerophosphate induces a cuboidal morphology, upregulates AP and osteocalcin expression, and a mineralized (hydroxyapatite) matrix [<xref ref-type="bibr" rid="B20">20</xref>].</p><p>Lineage differentiation signals can be subtle. Dexamethasone at 10<sup>-9</sup>mol/l supports adipocyte differentiation, whereas osteogenesis is favored at 10<sup>-7</sup> mol/l [<xref ref-type="bibr" rid="B19">19</xref>]. Human marrow MSCs obtained by Ficoll density gradient fractionation (1.078 g/ml) and cultured in 25% serum (half horse and half fetal calf) supplemented with hydrocortisone (1 &#x003bc;mol) gives rise to a heterogeneous population, in which fibroblast-like cells do not predominate [<xref ref-type="bibr" rid="B22">22</xref>]. Separation of the same population of marrow MSCs on a Percoll density gradient (1.090 g/ml) and culture in carefully selected 10% FCS resulted in a homogeneous population of spindle-shaped fibroblast cells. The higher-density Ficoll may isolate cells that sediment through the Percoll solution used for the marrow MSC isolation. Elutriation, as used in this investigation, probably selects a somewhat different population.</p><p>The CD34 status of mesenchymal cells is disputed [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. A minority of adult marrow cells express CD34. The antigen is present on pluripotent HSCs, and all unipotent myeloid and erythroid-colony-forming cells [<xref ref-type="bibr" rid="B23">23</xref>], but CD34 is also recognized on vascular endothelial cells, basement-membrane structures, and dendritic and perifollicular cells in human skin [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Simmons and Torok-Storb separated human marrow cells on the basis of their CD34 expression [<xref ref-type="bibr" rid="B26">26</xref>]. More than 95% of the detectable colony forming unit fibroblasts were recovered in the adherent CD34<sup>+</sup> population, but their CD34 density was much less than on CD34 <sup>high</sup> HSCs. Furthermore, only 5% of the CD34<sup>+</sup> marrow cells reacted with the monoclonal antibody Stro-1, which identifies marrow MSCs [<xref ref-type="bibr" rid="B6">6</xref>]. All these studies were done on marrow MSCs before culture, because after culture <italic>in vitro,</italic> the same stromal cells no longer react with anti-CD34 antibodies [<xref ref-type="bibr" rid="B26">26</xref>]. Likewise, although CD34<sup>+</sup> cells are identified in tissue sections of human umbilical vein endothelial cells, they are not found <italic>in vitro</italic> [<xref ref-type="bibr" rid="B27">27</xref>]. These and other reports suggest the CD34 glycoprotein is either down-regulated or modified <italic>in vitro</italic> to a form that is not reactive with the usual anti-CD34 antibodies [<xref ref-type="bibr" rid="B23">23</xref>]. Therefore, the failure of Fernandez <italic>et al</italic> to demonstrate CD34 on circulating stromal cells mobilized by growth factors was probably because the cells had been in culture for 10 days [<xref ref-type="bibr" rid="B7">7</xref>]. Similarly, the inability of Majumdar <italic>et al</italic> to demonstrate CD34 staining was on first-passage marrow MSCs [<xref ref-type="bibr" rid="B22">22</xref>]. <italic>In vitro</italic> culture conditions, however, cannot explain our failure to eliminate BMPCs in fresh elutriation fractions with anti-CD34-coated magnetic beads, a technique widely used to harvest CD34<sup>+</sup> HSCs from growth-factor-mobilized human blood. The absorptions were performed on fresh BMPC-rich elutriation fractions. Therefore, BMPCs either lack CD34 or have only a very low density of this glycoprotein.</p><p>The CD34<sup>+</sup> cells called fibrocytes, which are present in monocyte fractions of human blood and develop a fibroblast morphology when grown on fibronectin, have features identical to those of circulating vascular endothelial cell progenitors [<xref ref-type="bibr" rid="B28">28</xref>] and are probably not BMPCs.</p><p>Bone morphogenetic proteins (BMPs) were originally identified as proteins that induced bone formation at extraskeletal sites [<xref ref-type="bibr" rid="B29">29</xref>]. Currently, there are 20 or more known BMPs, all members of a larger TGF-&#x003b2; superfamily. BMPs are involved in morphogenesis and embryogenesis, influencing bone, cartilage, and skeletal formation [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>]. Much of this information comes from animal cells and embryos, but the addition of BMP2 to cultured postnatal human bone marrow `preosteoblastic' cells in the presence of &#x003b2;-glycerophosphate and ascorbic acid increases the gene message and protein production of AP, osteopontin, bone sialoprotein, osteocalcin, and &#x003b1;-1 collagen [<xref ref-type="bibr" rid="B33">33</xref>]. BMPCs develop into osteocalcin-producing cells (Fig. <xref ref-type="fig" rid="F5">5a</xref>) and make increased amounts of AP in response to BMP2 (Fig. <xref ref-type="fig" rid="F7">7</xref>). This cannot be explained by proliferation, because BMP2 reduces the number of marrow MSCs in either serum or serum-free conditions [<xref ref-type="bibr" rid="B33">33</xref>].</p><p>BMPRs belong to the TGF-&#x003b2; receptor family of serine/threonine kinases [<xref ref-type="bibr" rid="B34">34</xref>]. Both type I and type II BMPRs bind their respective ligands, but heterodimerization is required for a signal to be transduced [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>]. For instance, coexpression of type II BMPR with either IA or IB BMPR increases ligand-binding affinity and dramatically enhances biologic activity [<xref ref-type="bibr" rid="B11">11</xref>]. Human marrow MSCs express BMP2/4 type I and II receptors as shown in studies employing BMP2 as ligand in the presence or absence of a 100-fold excess of a competitor [<xref ref-type="bibr" rid="B33">33</xref>]. BMP structure is conserved across species, and antibodies to type I and typeII receptors react equally well with murine and human mesenchymal cells, but not with hematopoietic cells [<xref ref-type="bibr" rid="B34">34</xref>]. This is consistent with our findings that polyclonal antibody to BMPRs can be used to identify BMPCs and constitutes strong evidence that the circulating cells described in this report are mesenchymal precursors.</p><p>BMPCs stain with the Stro-1 monoclonal antibody made against human bone-marrow stromal cells [<xref ref-type="bibr" rid="B6">6</xref>]. Stro-1⁺ cells cultured in an osteogenic medium exhibit three markers of differentiated bone: AP; 1,25-dihydroxyvitamin-D₃-dependent induction of osteocalcin; and a mineralized matrix (hydroxyapatite) [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B37">37</xref>]. Stro-1 is expressed by BMPCs, but the antibody also stains pericytes, cells that surround small-vessel endothelium. Pericytes, which are of mesodermal origin, can also differentiate into a variety of cell types, including osteoblasts and adipocytes (reviewed in [<xref ref-type="bibr" rid="B38">38</xref>]). The reactivity of pericytes with BMPR antibodies is not known, but we have used BMPR antibodies to analyze mesenchymal cells in synovial tissues. The antibodies identify large cells in the inflamed joint lining, but they do not stain blood vessels of either normal or inflamed synovium (Marinova-Mutafchieva, personal communication).</p><p>SH-2, a proprietary antibody developed against isolated bone marrow MSCs [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B3">3</xref>], was not available when these studies were performed. When SH-2 was identified as endoglin (CD105) [<xref ref-type="bibr" rid="B4">4</xref>], the cells in a BMPC-enriched elutriation fraction of blood were examined with an anti-CD105 monoclonal antibody. The strong cytoplasmic staining of the large blood mesenchymal cells (Fig. <xref ref-type="fig" rid="F4">4b</xref>) is more evidence of the great similarity of marrow MSCs and circulating BMPCs.</p><p>Progenitor and precursor B cells require close contact with marrow MSCs for growth and maturation. Mouse marrow MSCs contain the gene for a protein (termed either stromal-cell-derived factor 1 (SDF-1) or pre-B-cell-growth-stimulating factor (PBSF) [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. SDF-1 is a powerful CXC chemokine that recruits circulating lymphocytes, monocytes, and CD34⁺hematopoietic progenitors, but not neutrophils [<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. PBSF is responsible for converting `early' B cells into immunoglobulin-producing cells [<xref ref-type="bibr" rid="B41">41</xref>]. SDF-1 mRNA is constitutively expressed in many tissues, unlike other chemokines, which are only induced [<xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B42">42</xref>]. SDF-1 is expressed in marrow MSCs, dermal fibroblasts, and synovial fibroblasts, but not HSCs. The demonstration of constitutive expression of SDF-1 mRNA in cultured BMPCs (Fig. <xref ref-type="fig" rid="F8">8</xref>) and as protein in supernatants from cultured BMPCs (data not presented) is additional evidence that BMPCs are of mesenchymal lineage.</p><p>Human OCs arise from HSCs in close proximity to stromal cells or from blood monocytes. The OC is a TRAP-positive, large, multinucleated cell with receptors for calcitonin and vitronectin (&#x003b1;v&#x003b2;3) (Fig. <xref ref-type="fig" rid="F5">5a</xref>) and the capacity to form resorption lacunae in bone slices [<xref ref-type="bibr" rid="B43">43</xref>]. Osteoblast production and OC production are tightly linked and regulated. Osteoblasts facilitate OC formation by providing physical support and critical soluble factors [<xref ref-type="bibr" rid="B43">43</xref>]. Our observation of spontaneous formation of cells with the morphology and phenotype of OCs in monocyte-rich (65%) elutriation fractions is best explained by the simultaneous presence of BMPCs in the same fractions.</p><p>More than 100 normal individuals had CD34^- mononuclear cells in a fraction of elutriated blood cells that fulfilled criteria for mesenchymal precursors or stem cells. They proliferated rapidly in culture, had an adherent, spread morphology, displayed cytoskeletal, cytoplasmic, and surface markers of mesenchymal precursors, and had a capacity for differentiation into fibroblast, osteoblast, and adipocyte lineages. Thus, autologous blood could be an important source of cells for tissue engineering and gene therapy [<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>]. In addition, the finding of similar cells in the inflammatory joint fluids and synovial tissues of patients with RA suggests they may play a role in the pathogenesis of this disease [<xref ref-type="bibr" rid="B47">47</xref>].</p></sec>

FcgammaR expression on macrophages is related to severity and chronicity of synovial inflammation and cartilage destruction during experimental immune-complex-mediated arthritis (ICA)

<sec><title>Introduction:</title><p>Fc&#x003b3; receptors (Fc&#x003b3;Rs) present on cells of the haematopoietic lineage communicate with IgG-containing immune complexes that are abundant in the synovial tissue of patients with rheumatoid arthritis (RA). In mice, three classes of Fc&#x003b3;R (RI, RII, and RIII) have been described. Binding of these receptors leads to either activation (Fc&#x003b3;RI and RIII) or deactivation (Fc&#x003b3;RII) of intracellular transduction pathways. Together, the expression of activating and inhibitory receptors is thought to drive immune-complex-mediated diseases.</p><p>Earlier studies in our laboratory showed that macrophages of the synovial lining are of utmost importance in the onset and propagation of immune-complex-driven arthritic diseases. Selective depletion of macrophages in the joint downregulated both inflammation and cartilage destruction. As all three classes of Fc&#x003b3;R are expressed on synovial macrophages, these cells are among the first that come in contact with immune complexes deposited in the joint. Recently, we observed that when immune complexes were injected into the knee joints of mice, strains susceptible to collagen-type-II arthritis (DBA/1, B10.RIII) developed more severe arthritis than nonsusceptible strains did, or even developed chronic arthritis. One reason why these strains are more susceptible might be their higher levels of Fc&#x003b3;Rs on macrophage membranes. To test this hypothesis, we investigated the role of Fc&#x003b3;Rs in inflammation and cartilage damage during immune-complex-mediated arthritis (ICA). First, we studied arthritis and subsequent cartilage damage in mice lacking functional Fc&#x003b3;RI and RIII (FcR &#x003b3;-chain^-/- mice). Next, DBA/1 mice, which are prone to develop collagen-type-II arthritis (`collagen-induced arthritis'; CIA) and are hypersensitive to immune complexes, were compared with control C57BL/6 mice as regards cartilage damage and the expression and function of Fc&#x003b3;Rs on their macrophages.</p></sec><sec><title>Aims:</title><p>To examine whether Fc&#x003b3;R expression on macrophages is related to severity of synovial inflammation and cartilage destruction during immune-complex-mediated joint inflammation.</p></sec><sec><title>Methods:</title><p>ICA was induced in three strains of mice (FcR &#x003b3;-chain^-/-, C57BL/6, and DBA/1, which have, respectively, no functional Fc&#x003b3;RI and RIII, intermediate basal expression of Fc&#x003b3;Rs, and high basal expression of Fc&#x003b3;Rs) by passive immunisation using rabbit anti-lysozyme antibodies, followed by poly-L-lysine lysozyme injection into the right knee joint 1 day later. In other experiments, streptococcal-cell-wall (SCW)- or zymosan-induced arthritis was induced by injecting SCW (25 &#x003bc;g) or zymosan (180 &#x003bc;g) directly into the knee joint. At several time points after arthritis induction, knee joints were dissected and studied either histologically (using haematoxylin/eosin or safranin O staining) or immuno-histochemically. The arthritis severity and the cartilage damage were scored separately on an arbitrary scale of 0-3.</p><p>Fc&#x003b3;Rs were immunohistochemically detected using the monoclonal antibody 2.4G2, which detects both Fc&#x003b3;RII and RIII. Deposition of IgG and C3c in the arthritic joint tissue was also detected immunohistochemically. Expression of Fc&#x003b3;Rs by murine peritoneal macrophages was measured using a fluorescence-activated cell sorter (FACS).</p><p>Peritoneal macrophages were stimulated using heat-aggregated gamma globulins (HAGGs), and production of IL-1 was measured using a bioassay. To assess the levels of IL-1 and its receptor antagonist (IL-1Ra) during arthritis, tissue was dissected and washed in RPMI medium. Washouts were tested for levels of IL-1 and IL-1Ra using radioimmunoassay and enzyme-linked immunosorbent assay. mRNA was isolated from the tissue, and levels of macrophage inflammatory protein (MIP)-2, monocyte chemoattractant protein (MCP)-1, IL-1, and IL-1Ra were determined using semiquantitative reverse-transcription polymerase chain reaction (RT-PCR).</p></sec><sec><title>Results:</title><p>ICA induced in knee joints of C57BL/6 mice caused a florid inflammation at day 3 after induction. To investigate whether this arthritis was Fc&#x003b3;R-mediated, ICA was induced in FcR &#x003b3;-chain^-/- mice, which lack functional Fc&#x003b3;RI and RIII. At day3, virtually no inflammatory cells were found in their knee joints. Levels of mRNA of IL-1, IL-1Ra, MCP-1, and MIP-2, which are involved in the onset of this arthritis, were significantly lower in FcR &#x003b3;-chain^-/- mice than in control C57BL/6 mice. Levels of IL-1 protein were also measured. At 6 h after ICA induction, FcR &#x003b3;-chain^-/- mice and control C57BL/6 mice showed similar IL-1 production as measured by protein level. By 24 h after induction, however, IL-1 production in the FcR &#x003b3;-chain^-/- mice was below the detection limit, whereas the controls were still producing a significant amount. To investigate whether the difference in reaction to immune complexes between the DBA/1 and C57BL/6 mice might be due to variable expression of Fc&#x003b3;Rs in the knee joint, expression <italic>in situ</italic> of Fc&#x003b3;Rs in na&#x000ef;ve knee joints of these mice was determined. The monoclonal antibody 2.4G2, which detects both Fc&#x003b3;RII and RIII, stained macrophages from the synovial lining of DBA/1 mice more intensely than those from C57BL/6 mice. This finding suggests a higher constitutive expression of Fc&#x003b3;Rs by macrophages of the autoimmune-prone DBA/1 mice. To quantify the difference in Fc&#x003b3;R expression on macrophages of the two strains, we determined the occurrence of Fc&#x003b3;Rs on peritoneal macrophages by FACS analysis. The levels of Fc&#x003b3;R expressed by macrophages were twice as high in the DBA/1 mice as in the C57BL/6 mice (mean fluorescence, respectively, 440 &#x000b1; 50 and 240 &#x000b1; 30 intensity per cell). When peritoneal macrophages of both strains were stimulated with immune complexes (HAGGs), we found that the difference in basal Fc&#x003b3;R expression was functional. The stimulated macrophages from DBA/1 mice had significantly higher IL-1&#x003b1; levels (120 and 135 pg/ml at 24 and 48 h, respectively) than cells from C57BL/6 mice (45 and 50 pg/ml, respectively).</p><p>When arthritis was induced using other arthritogenic triggers than immune complexes (zymosan, SCW), all the mouse strains tested (DBA/1, FcR &#x003b3;-chain^-/-, and C57BL/6) showed similar inflammation, indicating that the differences described above are found only when immune complexes are used to elicit arthritis.</p><p>We next compared articular cartilage damage in arthritic joints of the three mouse strains FcR &#x003b3;-chain^-/-, C57BL/6 (intermediate basal expression of Fc&#x003b3;Rs), and DBA/1 (high basal expression of Fc&#x003b3;Rs). Three indicators of cartilage damage were investigated: depletion of PGs, chondrocyte death, and erosion of the cartilage matrix. At day 3 after induction of ICA, there was no PG depletion in FcR &#x003b3;-chain^-/- mice, whereas PG depletion in the matrix of the C57BL/6 mice was marked and that in the arthritic DBA/1 mice was even greater. PG depletion was still massive at days 7 and 14 in the DBA/1 mice, whereas by day 14 the PG content was almost completely restored in knee joints of the C57BL/6 mice. Chondrocyte death and erosion of cartilage matrix, two indicators of more severe cartilage destruction, were significantly higher in the DBA/1 than in the C57BL/6 mice, while both indicators were completely absent in the FcR &#x003b3;-chain^-/- mice. Again, when arthritis was induced using other triggers (SCW, zymosan), all strains showed similar PG depletion and no chondrocyte death or matrix erosion. These findings underline the important role of immune complexes and Fc&#x003b3;Rs in irreversible cartilage damage.</p></sec><sec><title>Discussion:</title><p>Our findings indicate that inflammation and subsequent cartilage damage caused by immune complexes may be related to the occurrence of Fc&#x003b3;Rs on macrophages. The absence of functional Fc&#x003b3;RI and RIII prevented inflammation and cartilage destruction after induction of ICA, whereas high basal expression of Fc&#x003b3;Rs on resident joint macrophages of similarly treated mice susceptible to autoimmune arthritis was correlated with markedly more synovial inflammation and cartilage destruction. The difference in joint inflammation between the three strains was not due to different susceptibilities to inflammation per se, since intra-articular injection of zymosan or SCW caused comparable inflammation. Although extensive inflammatory cell mass was found in the synovium of all strains after intra-articular injection of zymosan, no irreversible cartilage damage (chondrocyte death or matrix erosion) was found. ICA induced in C57BL/6 and DBA/1 mice did cause irreversible cartilage damage at later time points, indicating that immune complexes and Fc&#x003b3;Rs play an important role in inducing irreversible cartilage damage. Macrophages communicate with immune complexes via Fc&#x003b3; receptors. Absence of functional activating receptors completely abrogates the synovial inflammation, as was shown after ICA induction in FcR &#x003b3;-chain^-/- mice. However, the &#x003b3;-chain is essential not only in Fc&#x003b3;RI and RIII but also for Fc&#x003b5;RI (found on mast cells) and the T cell receptor (TcR)-CD3 (Tcells) complex of &#x003b3;&#x003b4;T cells. However, T, B, or mast cells do not play a role in this arthritis that is induced by passive immunisation. Furthermore, this effect was not caused by a difference in clearance of IgG or complement deposition in the tissue. In this study, DBA/1 mice, which are susceptible to collagen-induced autoimmune arthritis and in a recent study have been shown to react hypersensitively to immune complexes, are shown to express higher levels of Fc&#x003b3;Rs on both synovial and peritoneal macrophages. Because antibodies directed against the different subclasses of Fc&#x003b3;R are not available, no distinction could be made between Fc&#x003b3;RII and RIII. Genetic differences in DBA/1 mice in genes coding for or regulating Fc&#x003b3;Rs may be responsible for altered Fc&#x003b3;R expression. If so, these mouse strains would have a heightened risk for immune-complex-mediated diseases.</p><p>To provide conclusive evidence for the roles of the various classes of Fc&#x003b3;R during ICA, experiments are needed in which Fc&#x003b3;Rs are blocked with specific antibodies, or in which knockout mice lacking one specific class of Fc&#x003b3;R are used. The only available specific antibody to Fc&#x003b3;R (2.4G2) has a stimulatory effect on cells once bound to the receptor, and therefore cannot be used in blocking experiments. Experiments using specific knockout mice are now being done in our laboratory.</p><p>Macrophages are the dominant type of cell present in chronic inflammation during RA and their number has been shown to correlate well with severe cartilage destruction. Apart from that, in humans, these synovial tissue macrophages express activating FcRs, mainly Fc&#x003b3;IIIa, which may lead to activation of these macrophages by IgG-containing immune complexes. The expression of FcRs on the surface of these cells may have important implications for joint inflammation and severe cartilage destruction and therefore FCRs may constitute a new target for therapeutic intervention.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Blom</surname><given-names>Arjen B</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>A.Blom@reuma.azn.nl</email></contrib><contrib id="A2" contrib-type="author"><name><surname>van Lent</surname><given-names>Peter L</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>van Vuuren</surname><given-names>Hanneke</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Holthuysen</surname><given-names>Astrid E</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Jacobs</surname><given-names>Cor</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>van de Putte</surname><given-names>Leo B</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>van de Winkel</surname><given-names>Jan G</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>van den Berg</surname><given-names>Wim B</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Rheumatoid arthritis (RA) is characterised by chronic joint inflammation eventually leading to irreversible cartilage destruction. IgG-containing immune complexes, such as rheumatoid factors, are abundant in the synovial tissue of patients with RA [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. It is still debated what role these complexes play in the aetiology and pathology of the disease. Immune complexes communicate with haematopoietic cells via Fc&#x003b3; receptors (Fc&#x003b3;Rs) [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>]. In recent studies, the importance of these receptors in inflammation and tissue damage has been shown in various inflammatory diseases, eg autoimmune haemolytic anaemia and thrombocytopenia [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>], autoimmune glomerulonephritis [<xref ref-type="bibr" rid="B8">8</xref>], and induced glomerulonephritis [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Fc&#x003b3;Rs, which belong to the immunoglobulin superfamily, bind IgG. Three classes of leucocyte Fc&#x003b3;R have been described (Fc&#x003b3;RI, RII, and RIII). Binding of these receptors initiates signalling cascades that can lead to either activation or deactivation of effector cells. In mice, cross-linking of Fc&#x003b3;RI and RIII leads to activation [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>] of intracellular signalling transduction pathways, whereas stimulation of Fc&#x003b3;RIIb leads to their deactivation [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. Coordinate expression of activating and inhibitory receptors has been suggested to drive immune-complex-mediated diseases [<xref ref-type="bibr" rid="B15">15</xref>].</p><p>During RA, the dominant cell type in the joint is the macrophage. Elegant studies by Bresnihan's group have shown a correlation between the number of macrophages present in the joint and severe cartilage destruction [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. In other studies, selective elimination of synovial lining macrophages prior to induction of experimental arthritis prevented the onset of arthritis and cartilage destruction. Furthermore, selective removal of lining macrophages during chronic arthritis significantly downregulated synovial inflammation [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>] and partly prevented exacerbation [<xref ref-type="bibr" rid="B23">23</xref>]. Synovial macrophages, which cover the inside of diarthrodial joints and surround blood vessels in the synovium, are among the first cells that come in contact with immune complexes. All three classes of Fc&#x003b3;R are known to be expressed on macrophages in general and on synovial macrophages in particular [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>] and the amount and/or ratio of the various types of Fc&#x003b3;R expression might determine joint inflammation and cartilage destruction. Overexpression, downregulation, or functional impairment of activating Fc&#x003b3;RI and RIII on macrophages may have consequences for inflammation and cartilage destruction. Differences in basal expression levels of Fc&#x003b3;Rs may be genetically linked or related to age [<xref ref-type="bibr" rid="B27">27</xref>]. Cytokines such as IFN-&#x003b3; and TGF-&#x003b2; have been shown to upregulate, respectively, Fc&#x003b3;RI and RIII, whereas IL-4 and IL-13 down-regulate both receptor classes [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>].</p><p>Recently, we observed that when immune complexes were injected into the knee joints of mice, strains susceptible to collagen-type-II arthritis (DBA/1 and B10.RIII mice) developed more severe arthritis than nonsusceptible strains did, or even developed chronic arthritis [<xref ref-type="bibr" rid="B34">34</xref>]. Mouse strains that are prone to develop autoimmune arthritis may also express higher levels of Fc&#x003b3;Rs on their macrophages, thus driving joint inflammation and cartilage destruction.</p><p>To test this hypothesis, we have investigated the role of Fc&#x003b3;Rs in inflammation and cartilage damage during immune-complex-mediated arthritis (ICA). In this model, arthritis is induced by intra-articular injection of the antigen into knee joints of mice that were previously passively immunised against the antigen, by intravenous injection of specific antibodies [<xref ref-type="bibr" rid="B35">35</xref>]. First, synovial inflammation, cytokine production, and joint destruction were investigated in FcR &#x003b3;-chain knockout (FcR &#x003b3;-chain^-/-) mice, lacking functional FcRI and RIII, and their controls. Then DBA/1 mice, which are prone to develop collagen-type-II autoimmune arthritis and express a higher than usual sensitivity for immune complexes, were compared with non-susceptible C57BL/6 mice, and the expression and functional relation of Fc receptors on macrophages with inflammation and cartilage destruction were investigated. The findings indicate that FcR expression on synovial-lining macrophages is related to the severity and chronicity of synovial inflammation and cartilage destruction during joint inflammation elicited by immune complexes.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Animals</title><p>Male and female C57BL/6 mice were obtained from Jackson (Bar Harbor, Maine, USA). DBA/1 mice were obtained from Bomholdgard (Rye, Denmark). FcR &#x003b3;-chain^-/- mice were kindly given by T Saito [<xref ref-type="bibr" rid="B10">10</xref>]. FcR &#x003b3;-chain^-/- mice were backcrossed to C57BL/6 mice for 12 generations. FcR &#x003b3;-chain^-/- were also backcrossed to a DBA/1 background. Mice were fed a standard diet and tap water <italic>ad libitum</italic>. Mice were used between the ages of 8 and 12 weeks and weighed 25&#x02013;30 g.</p></sec><sec><title>Chemicals</title><p>Poly-L-lysine (PLL), lysozyme, 1-ethyl-3-(3-dimethylamino-propyl) carbodiimide (EDC), and zymosan A (from <italic>Saccharomyces cerevisiae</italic>) were obtained from Sigma Chemical Company, St Louis, MO, USA. N,N-dimethyl-1,3-propanediamine (DMPA) was obtained from BDH Chemicals Ltd, Poole, UK.</p></sec><sec><title>Lysozyme coupling to PLL</title><p>Lysozyme was coupled to PLL in accordance with the method of Danon <italic>et al</italic> [<xref ref-type="bibr" rid="B36">36</xref>]. As described by those authors, 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide was used as an activator and PLL as a nucleophil. Free carboxyl groups of the protein were then coupled to amino groups of PLL. The molecular mass was raised whereas the isoelectric point remained high, as was determined in a 5% polyacrylamide slab gel with 0.8% ampholines (pH gradient 3.5-9.5). The molecular mass appeared to be 74 kD on SDS-PAGE.</p></sec><sec><title>Induction of arthritis by immune complexes, zymosan, or SCW</title><p>Specific rabbit anti-lysozyme antisera, made as described elsewhere [<xref ref-type="bibr" rid="B35">35</xref>] and made complement-free by heating at 56&#x000b0;C for 30 min, were injected (0.2 ml) intravenously into mice. ICA was then induced by injecting 3 &#x003bc;g of PLL-coupled lysozyme into the right knee joint. The left knee joint was injected with saline solution and used as a control. As an additional control, some mice were given normal rabbit serum instead of specific anti-lysozyme. When PLL-lysozyme was injected into the knee joint without prior administration of specific anti-lysozyme antibodies, no inflammation or cartilage damage developed, in either C57BL/6 or DBA/1 mice.</p><p>Zymosan (30 mg) was dissolved in 1 ml saline solution by heating up to 100&#x000b0;C twice and was then sonicated to obtain a homogeneous suspension. Arthritis was induced in other mice by injecting 180 &#x003bc;g zymosan into both left and right knee joints. Fc&#x003b3;-chain knockout mice and the control strain (C57BL/6) were matched for age and sex.</p><p>SCW arthritis was induced by injecting 25 &#x003bc;g SCWs (rhamnose content), prepared as described elsewhere [<xref ref-type="bibr" rid="B37">37</xref>] and dissolved in 6 &#x003bc;l saline, into the right knee joint of C57BL/6 and DBA/1 mice.</p></sec><sec><title>Histology</title><p>Total knee joints of mice were isolated 3 days after induction of arthritis in Fc&#x003b3;-chain deficient mice and at days 3, 7, and 14 in C57BL/6 and DBA/1 mice. For standard histology, tissue was fixed in 4% formaldehyde, decalcified in formic acid, and subsequently dehydrated and embedded in paraffin. Paraffin sections were cut at 7 &#x003bc;m and mounted on gelatine-coated slides. Haematoxylin/eosin (H&#x00026;E) staining was performed to study the inflammatory cells.</p><p>Infiltrate and exudate were scored separately. The severity was determined by two blinded observers, using an arbitrary score (0&#x02013;3): 0 = none, 1 = mild, 2 = moderate, and 3 = maximal cellularity. To study PG depletion from the cartilage matrix, sections were stained with safranin O and then coun-terstained with fast green. PG depletion from several cartilage surfaces was scored (patella + adjacent femur surface, lateral condyle femur + adjacent surface tibia plateau, and medial femoral condyle + adjacent surface tibia plateau).</p><p>The severity of depletion was scored by two blinded observers, using an arbitrary score reflecting the level of destaining: 0 = none, 1 = mild, 2 = moderate, and 3 = maximal destaining of cartilage. Chondrocyte death was scored using H&#x00026;E-stained sections. The amount of empty lacunae was given as a percentage of total amount of cells (empty lacunae + viable chondrocytes). Cartilage erosion was scored by expressing the amount of eroded cartilage as a percentage of the cartilage surface.</p></sec><sec><title>Detection of Fc&#x003b3; receptors</title><p>To compare expression of Fc&#x003b3;RII/RIII by DBA/1 and C57BL/6 mice, we performed immunohistochemistry using a rat antibody against murine Fc&#x003b3;RII/RIII (2.4G2, Pharmin-gen, San Diego, CA, USA) [<xref ref-type="bibr" rid="B38">38</xref>]. Briefly, cryostat sections were fixed in acetone vapour for 3 min and endogenous peroxidase was blocked using 1% H₂O₂. Sections were incubated overnight with 2.4G2 (1 &#x003bc;g/ml) in phosphate-buffered saline (PBS)/0.1% bovine serum albumin (BSA). Sections were washed in PBS and then incubated with rabbit anti-rat IgG coupled to biotin (Vector Laboratories, Burlingame, CA, USA). Then the sections were incubated with avidin biotin complexes (ABC; Vectastain Elite-kit, Vector Laboratories) and developed using diaminobenzidine (Sigma Chemical Company, St Louis, MO, USA).</p></sec><sec><title>Detection of IgG and C3c</title><p>IgG deposited in the tissue by immune-complex formation was detected on paraffin-embedded knee-joint sections. Sections were incubated with biotinylated rat anti-rabbit antibodies (Vector Laboratories) for 1 h. Sections were washed thoroughly with PBS, incubated with ABC complexes, and developed using the method described above.</p><p>C3c was detected using cryostat sections. Sections were incubated with rat anti-murine C3c (Nordic, Tilburg, The Netherlands) after acetone fixation and inhibition of endogenous peroxidase. Isotype-matched antibodies directed against an irrelevant epitope (Vector) were used as a negative control. Staining was scored using image analysis by measuring the mean amount of blue light that passed through a well-defined location in the tissue section near the cruciate ligaments.</p></sec><sec><title>Analysis by fluorescence-activated cell sorter (FACS)</title><p>Peritoneal macrophages of different strains were isolated from the peritoneal cavity by lavage with ice-cold DMEM/ 10% fetal calf serum (FCS)/1% pyruvate. These cells (5 &#x000d7; 10⁶/100 &#x003bc;l) were incubated with 2.4G2 (5 &#x003bc;g/ml). They were then washed and incubated with mouse-adsorbed hamster anti-rat F(ab')₂ fragments labelled withfluorescein isothiocyanate (FITC). FACS analysis was performed using a Coulter Epics XL/XL-MCL (Coulter Electronics Ltd, Mijdrecht, The Netherlands). Omitting the first antibody and substitution of 2.4G2 for isotype-matched irrelevant antibodies (DAKO, Glostrup, Denmark) were used as negative controls. The window was set so that &#x0003e;95% of cells were F4/80-positive, indicating that &#x0003e;95% of cells were macrophages.</p></sec><sec><title>Stimulation of peritoneal macrophages with heat-aggregated gamma globulin</title><p>Peritoneal macrophages were isolated by peritoneal lavage using ice-cold DMEM/10% FCS. Cells were put into 24-well plates (Costar, Acton, MA, USA) at a concentration of 1 &#x000d7; 10⁶ cells/ml. After a 4-day adjustment period, the culture medium was changed to one containing heat-aggregated gamma globulin (HAGG) at 100 &#x003bc;g/ml, or a control medium. After 24 or 48 h, the culture medium was collected. HAGGs were obtained by heating 10 mg/ml rabbit IgG (Sigma Chemicals, St Louis, MO, USA) to 63&#x000b0;C for 30 min. After heating, the solution was centrifuged at 12 000 <bold> <italic>g</italic> </bold> for 10 min. The concentration of the HAGG in the supernatant was determined by reading the absorbance at 280 nm.</p></sec><sec><title>Production of inflammatory mediators by synovial tissue</title><p>Synovial tissue was isolated by dissection of patellar tendon and patellar plate containing the patella, tendons, and synovium, as described earlier [<xref ref-type="bibr" rid="B39">39</xref>]. Mediators were obtained by elution from synovial specimens derived from six knee joints in 2 ml Roswell Park Memorial Institute (RPMI) medium for 1 h at room temperature. Washouts were tested for their bioactive IL-1 levels in an IL-1-sensitive bioassay (NOB-1 assay) and in a radioimmunoassay (RIA) to determine the total protein levels.</p></sec><sec><title>Bioassay for IL-1</title><p>IL-1 activity was measured in the one-stage bioassay for IL-1 as described by Gearing <italic>et al</italic> [<xref ref-type="bibr" rid="B40">40</xref>]. The assay was performed as a culture of the IL-1-specific thymoma cell EL-4, designated NOB-1, which produces IL-2 and IL-4 in response to IL-1, in co-culture with the lymphokine-responder CTLL line. Briefly, EL-4 cells were washed twice and resuspended at 2.5 &#x000d7; 10⁵ cells per ml RPMI medium containing 5% FCS. The cells were distributed into 96-well microtitre plates at 2.5 &#x000d7; 10⁴ cells per well in 100-&#x003bc;l volumes. CTLL cells (4 &#x000d7; 10³) suspended in 50 &#x003bc;l RPMI medium were added, followed by appropriate dilutions of test sample to a final volume of 200 &#x003bc;l. After 20 h, 0.5 &#x003bc;Ci ³H-thymidine (specific activity 20 Ci/mmol; Amersham Nederland BV, 's Hertogenbosch, The Netherlands) was added to each well. After a 3-h incubation, the contents were harvested, and incorporated activity was determined. The EL-4 line, from which NOB-1 was derived, does not incorporate thymidine, because it is deficient in thymidine kinase, and therefore the incorporation of ³H-thymidine is a measure of only CTLL proliferation. Maximal ³H-thymidine incorporation in the bioassay in the presence of IL-1 was between 15 000 and 20 000 cpm. Culture media contained only minor concentrations of IL-2 or IL-4, as was assessed by testing samples with CTLL alone.</p></sec><sec><title>Measurement of IL-1&#x003b1; and IL-1&#x003b2; protein levels</title><p>IL-1 culture supernatants were measured in duplicate by a nonequilibrium RIA as described elsewhere [<xref ref-type="bibr" rid="B41">41</xref>]. Briefly, 100 &#x003bc;l polyclonal rabbit anti-murine IL-1&#x003b1; and IL-1&#x003b2; (diluted in RIA buffer, pH 7.4) was added to 100 &#x003bc;l of samples and standards and kept on ice. After vortexing, the tubes were incubated at 4&#x000b0;C. After 24 h, 100 &#x003bc;l of the appropriate ¹²⁵I-labelled IL-1&#x003b1; and &#x003b2; containing approximately 10 000 cpm was added to each tube, and incubation was continued for a further 24 h at 4&#x000b0;C. RIA buffer (750 &#x003bc;l) containing 9% (w/v) polyethylene glycol 6000 (Merck Diagnostica, Darmstadt, Germany) and 3% (w/v) goat anti-rabbit serum was added, to separate bound and free tracer. The tubes were incubated for 20 min at room temperature. After centrifugation at 1500 &#x000d7; <bold> <italic>g</italic> </bold> for 15 min, supernatants were quickly drained on adsorbent paper. A gamma-counter was used to count the radioactivity remaining on the paper. The radioactivity in control tubes (the nonspecific binding activity) was subtracted from samples and standards. The detection limit of the assay was 20 pg/ml for both IL-1&#x003b1; and IL-1&#x003b2;.</p></sec><sec><title>Measurement of IL-1 receptor antagonist</title><p>Protein levels of IL-1 receptor antagonist (IL-1Ra) in synovial washout specimens were detected using a specific sandwich enzyme-linked immunosorbent assay (ELISA). Briefly, microtitre plates were coated with unconjugated first antibody (anti-IL1Ra: MAB480, R&#x00026;D Systems, Abingdon, UK) overnight at 4&#x000b0;C. Subsequently, wells were blocked using 1% BSA followed by a 3-h incubation with patella washouts at 37&#x000b0;C. Wells were then incubated with a biotinylated specific antibody (IL-1Ra: BAF480, R&#x00026;D Systems) and a signal enhancement step was performed by incubating with PolyHRP (CLB, Amsterdam, The Netherlands). Microtitre plates were developed using o-phenylene-diamine and optical density was read at 492 nm. Between all steps, a washing episode was included using PBS/0.1% Tween 20.</p></sec><sec><title>Semiquantitative detection of mRNA using reverse-transcription polymerase chain reaction</title><p>Levels of mRNA for four different cytokines/chemokines were detected using a semiquantitative method. In brief, mice were killed by cervical dislocation, immediately followed by dissection of the patella with adjacent synovial tissue. Six patellae per group were obtained. Two biopsies with a diameter of 3 mm were punched from the synovial tissue, using a biopsy punch (Stiefel, Wachtersbach, Germany): one from the medial and one from the lateral aspect of the patella. Three lateral and three medial biopsies were pooled (six total), to obtain two samples per group. The samples were immediately frozen in liquid nitrogen. Thereafter samples were ground to powder using a Micro-dismembrator II (B Braun, Melsungen, Germany) and total RNA was extracted using 1 ml TRIzol reagent.</p><p>One microgram of RNA was used for reverse-transcription polymerase chain reaction (RT-PCR). Reverse transcription of mRNA was done using oligoDT primers, and 1/20 of the cDNA was used in one amplification by polymerase chain reaction (PCR). PCR was performed at a final concentration of 200 &#x003bc;M dNTPs, each primer at 0.1 &#x003bc;M, and 1unit <italic>Taq</italic> polymerase (Life Technologies, Breda, The Netherlands) in standard PCR buffer. Message for glycer-aldehyde-3-phosphate dehydrogenase (GAPDH), IL-1&#x003b2;, IL-1Ra, MCP-1, and MIP-2 was amplified using specific primers. The primer sequences for IL-1&#x003b2; were: 5'-TTGACGGACCCCAAAAGATG-3' (sense) and 5'-AGAAGGTGCTCATGTCCTCA-3' (antisense), for IL-1ra: 5'-TGCTGGGGACCCTACAGTCAC-3' (sense) and 5'-GCAAGTGCATCATCGTTGTTC-3' (antisense), for MCP-1: 5'-CTCACCTGCTGCTACTCATTC-3' (sense) and 5'-GCATGAGGTGGTTGTGAAAA-3' (antisense) and for MIP-2: 5'-GCTGGCCACCAACCACCAGG-3' (sense) and 5'-AGCGAGGCACATCAGGTACG-3' (antisense). The PCR-reaction was paused after 20, 23, 26, 29, 32, 35, 38, 41, and 44 cycles at the very end of the extension phase (72&#x000b0;C) and 5-&#x003bc;l samples were taken. PCR products were separated on 1.6% agarose gel and stained with ethidium bromide. The results are presented as cycle number in which the first detectable amount of DNA appears on the agarose gel. Samples were compared after correction for GAPDH content for each individual sample, to rule out confounding by variation of cellularity in the biopsies.</p></sec><sec><title>Statistical analysis</title><p>Differences between experimental groups were tested for significance using the Wilcoxon rank test. <italic>P</italic> values &#x0003c;0.05 were considered significant.</p></sec></sec><sec><title>Results</title><sec><title>Fc&#x003b3; receptors are essential for synovial inflammation during immune-complex-mediated arthritis</title><p>To investigate the involvement of Fc&#x003b3; receptors in locally induced ICA, we used FcR &#x003b3;-chain^-/- C57BL/6 mice, which lack functional Fc&#x003b3;RI and RIII. ICA was induced in knee joints of FcR &#x003b3;-chain^-/- mice and control C57BL/6 mice. Histology of total knee joints showed florid inflammation in C57BL/6 mice 3 days after induction of ICA (Fig. <xref ref-type="fig" rid="F1">1b</xref>). In knee joints of FcR &#x003b3;-chain^-/- mice, however, virtually no inflammatory cells were observed at day 3 of ICA (Fig. <xref ref-type="fig" rid="F1">1a</xref>). Joint inflammation, defined as cells present in the joint cavity (exudate) and synovium (infiltrate), was scored. Both exudate and infiltrate were substantial in C57BL/6 mice whereas, in FcR &#x003b3;-chain^-/- mice, cell influx was virtually absent (Table <xref ref-type="table" rid="T1">1</xref>).</p><p>To investigate whether Fc&#x003b3;Rs are involved in the upregulation of inflammatory mediators seen during the first phase of this arthritis, RT-PCR was performed on synovial specimens from arthritic knee joints. Levels of mRNA of inflammatory mediators (IL-1&#x003b2;, IL-1Ra, MCP-1, MIP-2) were detected semiquantitatively in knockout and control mice, 6 and 24 h after ICA induction (Fig. <xref ref-type="fig" rid="F2">2</xref>). IL-1&#x003b2; and IL-1Ra mRNA levels were high in control arthritis and appeared to be decreased (down to 2⁸) in FcR &#x003b3;-chain^-/- mice. MCP-1 and MIP-2 mRNA levels were also diminished in the early phase of ICA (6 h) relative to the levels of controls (respectively down to 2⁶ and 2⁴) suggesting downregula-tion of both polymorphonuclear neutrophils (PMNs) and monocyte-specific chemokines.</p><p>Subsequently, intra-articular production of IL-1 and IL-1Ra protein during arthritis was measured. Washouts of synovial specimens from joints of control C57BL/6 mice contained considerable amounts of IL-1 (240 pg/ml) at 6 h after induction of arthritis; at 24 h, IL-1 levels were reduced to 70 pg/ml (Table <xref ref-type="table" rid="T2">2</xref>). Washouts of arthritic FcR &#x003b3;-chain^-/- joints contained 200 ng/ml IL-1 at 6 h but no detected IL-1 at 24 h after induction, indicating that in the absence of functional Fc&#x003b3;RI and RIII, a rapid downregulation of IL-1 production is found. IL-1Ra production was comparable at 6 h in the two strains (respectively 180 and 215 pg/ml) and below the detection limit at 24 h after ICA induction in both strains (Table <xref ref-type="table" rid="T2">2</xref>).</p><p>To exclude the possibility that the strong reduction of inflammation during ICA in FcR &#x003b3;-chain^-/- mice is caused by a lower level of immune complexes which are formed or retained within the knee joint, deposition of IgG in the knee joint was detected using biotinylated goat anti-rabbit IgG. As Table <xref ref-type="table" rid="T3">3</xref> shows, at 6 and 24 h after ICA induction, no difference was found between FcR &#x003b3;-chain^-/- and C57BL/6 mice in the amount of immune complexes deposited within the knee joints. We compared deposition of complement (C3c) in the synovial tissue of both strains to rule out the possibility that a difference in complement deposition caused differences in inflammation. Control and knockout mice showed the same amounts of C3c in the tissue as was determined by immunohistochemistry (Table <xref ref-type="table" rid="T3">3</xref>).</p></sec><sec><title>Expression of Fc&#x003b3;RII/III is elevated in resident macrophages of na&#x000ef;ve CIA-sensitive mice that are hypersensitive to immune complexes</title><p>In a recent study [<xref ref-type="bibr" rid="B36">36</xref>], we noticed that certain mouse strains that are prone to develop collagen-type-II autoimmune arthritis are also particularly susceptible to immune-complex-induced inflammation. Induction of ICA in the DBA/1 knee joint resulted in a more severe arthritis than that in C57BL/6 mice, and this became chronic (Table <xref ref-type="table" rid="T1">1</xref>). Local production of IL-1 was also significantly higher and more prolonged than in C57BL/6 mice, both at 6 and 24 h after ICA induction (Table <xref ref-type="table" rid="T2">2</xref>). To investigate whether the enhanced inflammation in DBA/1 mice during ICA was still mediated by FcRs, FcR &#x003b3;-chain^-/- mice were backcrossed to DBA/1 (DBA&#x003b3;^-/-) mice. ICA was induced in these mice and the inflammation was compared with that in DBA&#x003b3;^+/+ littermates. We found that at day 3 after arthritis induction, arthritis was full-blown in DBA&#x003b3;^+/+mice (arbitrary scores of 1.0 &#x000b1; 0.4 for exudate and 1.6 &#x000b1; 0.3 for infiltrate) but was completely absent in &#x003b3;-chain-deficient DBA/1 mice (0.1 &#x000b1; 0.2 and 0.1 &#x000b1; 0.1, respectively; not shown).</p><p>To examine whether the different reactions in knee joints of C57BL/6 and DBA/1 mice to immune complexes might be due to variable local expression of Fc&#x003b3;R, expression <italic>in situ</italic> of Fc&#x003b3;Rs in na&#x000ef;ve knee joints was determined immunohistochemically, using the monoclonal antibody (mAb) 2.4G2 (Fc&#x003b3;RII/III). Synovial macrophage-like type A lining cells and deeper-lying synovial macrophages were the dominant cell types expressing these receptors, as was seen in double-labeling studies with F4/80 (data not shown). Interestingly, synovial macrophages of na&#x000ef;ve DBA/1 mice stained more intensely than those of C57BL/6 mice (Fig. <xref ref-type="fig" rid="F3">3</xref>), indicating a higher constitutive expression of Fc&#x003b3;Rs in lining macrophages of these autoimmune-prone mice.</p><p>To quantify the difference in Fc&#x003b3;R expression by macrophages of the two strains (DBA/1 and C57BL/6), we further investigated receptor expression on peritoneal macrophages, using FACS analysis. Macrophages of DBA/1 mice (mean fluorescence 440 &#x000b1; 50 intensity per cell) displayed almost twice the mean fluorescence with mAb 2.4G2 as those of C57BL/6 mice (240 &#x000b1; 30 intensity per cell) (Fig. <xref ref-type="fig" rid="F4">4</xref>). These findings also indicate that Fc&#x003b3;RII/III is significantly upregulated on the membrane of macrophage populations present in various body compartments of na&#x000ef;ve DBA/1 mice.</p><p>To further investigate whether a higher basal Fc&#x003b3;R expression on macrophages has consequences for the activation of cells upon interaction with immune complexes, peritoneal macrophages of both strains were incubated with aggregated IgG and production of IL-1 was determined by bioassay. Incubation of C57BL/6 cells for 24 or 48 h caused the release of considerable amounts of bioactive IL-1 (45 pg/ml at 24 h and 50 pg/ml at 48 h) (Fig. <xref ref-type="fig" rid="F5">5</xref>). Interestingly, DBA/1 macrophages produced significantly higher amounts of IL-1 (120 and 135 pg/ml, respectively) than C57BL/6 mice.</p></sec><sec><title>Arthritogenic triggers other than immune complexes cause comparable inflammation in knee joints of mice that differ in Fc&#x003b3;R expression levels</title><p>To substantiate the hypothesis that differences in FcR expression underlie the variation seen during immune-complex-induced joint inflammation, we injected various arthritogenic triggers (zymosan, SCWs) that do not act via Fc&#x003b3;Rs into the knee joints of three strains of mice (FcR &#x003b3;-chain^-/-, C57BL/6, and DBA/1). A previous study had shown that when these triggers were used, there was no difference between DBA/1 and C57BL/6 mice in joint inflammation [<xref ref-type="bibr" rid="B34">34</xref>]. In the present study, when zymosan was injected into knee joints of FcR &#x003b3;-chain^-/- mice, the inflammation was similar to that in control C57BL/6 mice. At day 3 after injection, inflammation in FcR &#x003b3;-chain^-/-mice was scored as 1.0 &#x000b1; 0.4 for infiltrate and 0.9 &#x000b1; 0.6 for exudate, versus 0.9 &#x000b1; 0.3 and 0.9 &#x000b1; 0.5, respectively, in control C57BL/6 mice. These findings indicate that knee joints of the three strains examined develop comparable, marked inflammation after injection of zymosan or SCWs. Moreover, IL-1&#x003b2; production in the knee joints during zymosan-induced arthritis was comparable in DBA/1 and C57BL/6 mice (Table <xref ref-type="table" rid="T2">2</xref>).</p></sec><sec><title>Fc&#x003b3;R expression on macrophages is correlated with cartilage destruction during immune-complex-mediated arthritis</title><p>Apart from inflammation, absence or overexpression of Fc&#x003b3;Rs by local macrophages may also have serious consequences for cartilage destruction. We investigated several parameters of cartilage destruction (depletion of proteogly-cans [PGs], chondrocyte death, and erosion of the cartilage matrix). Arthritic knee joints of the three strains were compared at different time points after ICA induction. At day 3, control C57BL/6 mice showed marked PG depletion in various cartilage surfaces (patella, femur adjacent to patella, lateral and medial condyle of the femur, and lateral and medial condyle of the tibia). Strikingly, PGs were not depleted in arthritic knee joints of the FcR &#x003b3;-chain^-/- mice (Fig. <xref ref-type="fig" rid="F6">6a</xref>,<xref ref-type="fig" rid="F6">b</xref>; Table <xref ref-type="table" rid="T4">4</xref>). In contrast, arthritic DBA/1 knee joints showed markedly greater depletion of PGs than those of C57BL/6 mice. Moreover, PG depletion was still severe at days 7 and 14 in the DBA/1 mice, whereas in the C57BL/6 mice it was almost fully restored by day 14 (Fig. 6b-e; Table <xref ref-type="table" rid="T4">4</xref>). When ICA was induced in FcR &#x003b3;^-/- mice with a DBA/1 background, PG depletion at day 3 was scored 2.6 &#x000b1; 0.3 in the Fc&#x003b3;-chain^+/+ mice and was totally absent in the FcR &#x003b3;-chain^-/- ones.</p><p>Chondrocyte death was measured by determining empty lacunae in the cartilage matrix and breaks in DNA strands as evaluated using an <italic>in situ</italic> cell detection kit (TUNEL method; data not shown). In knee joints of C57BL/6 mice, chondrocyte death at day 3 was 5% in the patella and 30% in the adjacent femur. In the knee joints of knockout mice, chondrocyte death was completely absent (Table <xref ref-type="table" rid="T4">4</xref>), whereas in the DBA/1 mice, it was strikingly higher in both the patella (50%) and the femur (90%) than in the control mice (C57BL/6).</p><p>Finally, erosion of the cartilage matrix in knee joints was measured. Whereas no erosion was found in either FcR &#x003b3;-chain^-/- or C57BL/6 mice at day 3 after arthritis induction, in DBA/1 mice erosion was already considerable by day 3, and complete loss of the cartilage surface layer up to the tidemark was observed at days 7 and day 14 (Fig. <xref ref-type="fig" rid="F6">6d</xref>,<xref ref-type="fig" rid="F6">e</xref>; Table <xref ref-type="table" rid="T4">4</xref>).</p></sec><sec><title>Severe cartilage destruction in DBA/1 mice is specific for immune-complex-mediated arthritis</title><p>To further substantiate the interpretation that immune complexes and Fc&#x003b3;Rs, rather than inflammatory cells, caused the observed severe joint cartilage destruction, various concentrations of zymosan were injected into the knee joints of all three mouse strains (FcR &#x003b3;-chain^-/-, C57BL/6, and DBA/1). Even large amounts of zymosan (up to 180 &#x003bc;g), although inducing a tremendous influx of inflammatory cells, failed to induce chondrocyte death or cartilage erosion. PG depletion at day 3, however, was scored as 3.6 &#x000b1; 0.4 in FcR &#x003b3;^-/- mice, versus 0.6 &#x000b1; 0.3 in C57BL/6 mice. At day 7, mean PG depletion was scored as 1.8 &#x000b1; 0.5 in C57BL/6 mice and 1.7 &#x000b1; 0.8 in DBA/1. PG depletion in the cartilage matrix in response to zymosan, therefore, was marked but did not differ between the three strains.</p></sec></sec><sec><title>Discussion</title><p>We have found that inflammation and subsequent cartilage destruction caused by immune complexes may be related to the expression of Fc&#x003b3;Rs on macrophages. Three strains differing in levels of expression of Fc&#x003b3;Rs were compared. Absence of functional Fc&#x003b3;RI and RIII prevents synovial inflammation and cartilage destruction after induction of immune complex arthritis, whereas elevated expression of Fc&#x003b3;R on resident joint macrophages of mice susceptible for development of autoimmune arthritis is correlated to markedly more synovial inflammation and cartilage destruction after induction of ICA. The difference in joint inflammation seen within the three strains was not due to genetic differences resulting in a different susceptibility to inflammation, since injection of zymosan or SCWs caused comparable inflammation in the three strains.</p><p>Mice lacking the FcR &#x003b3;-chain were used to investigate the role of Fc&#x003b3;Rs in ICA. The FcR &#x003b3; subunit is essential not only for functional FcRI and RIII IgG receptors but also for the high-affinity IgE receptor Fc&#x003b5;RI and TcR-CD3 complex of &#x003b3;&#x003b4;T cells. In this passively induced arthritis, B and T cells do not play a role [<xref ref-type="bibr" rid="B35">35</xref>]. Even during the chronic phase of ICA as seen in the DBA/1 knee joint, no effect on synovial inflammation was found after giving anti-&#x003b1;&#x003b2;TCR antibodies (data not shown). The high-affinity receptor Fc&#x003b3;RI, which binds IgE, is found on mast cells. A previous study showed, however, that experimentally induced ICA in knee joints of mice that are deficient for mast cells (WBB6F1-W/Wv) [<xref ref-type="bibr" rid="B42">42</xref>] was not significantly different from that in control mice, indicating that mast cells do not play an important role in ICA.</p><p>The findings described above suggest that the observed joint inflammation is locally regulated by synovial cells. Earlier studies revealed that macrophage-like type A cells, which form the dominant population (more than 80%) of the lining layer of diarthrodial joints in mice, regulate the onset and chronicity [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>] as well as the exacerbation [<xref ref-type="bibr" rid="B23">23</xref>] of experimental arthritis. Selective removal of type A cells by clodronate-containing liposomes prior to induction of several experimental arthritides prevented the onset of arthritis and also the larger part of cartilage destruction [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B43">43</xref>]. These studies suggested that synovial macrophages form an important source of chemotactic factors and that activation by immune complexes results in production of chemokines by these cells. It was also found that chronicity of synovial inflammation is mostly seen in models in which immune complexes are present [<xref ref-type="bibr" rid="B44">44</xref>]. Immune complexes communicate with macrophages via FcRs, especially the Fc&#x003b3;Rs, which bind IgG. IgG represents the most dominant immunoglobulin class in the blood and might be involved in activation of macrophages. Coordinate expression of activating Fc&#x003b3;RI and RIII and the inhibiting receptor Fc&#x003b3;RIIb exposed on joint macrophages probably determines the reaction to immune complexes [<xref ref-type="bibr" rid="B15">15</xref>].</p><p>Indeed, we found that the absence of functional Fc&#x003b3;RI and RIII completely abrogated the onset of ICA, as was seen in knee joints of FcR &#x003b3;-chain^-/- mice. The proinflammatory cytokine IL-1, which has been shown to regulate synovial inflammation within this model [<xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref>], was significantly decreased at protein level 24 h after induction. Semiquantitative determination of mRNA levels also suggested downregulation of IL-1. At 6 h after induction of ICA, no difference in cytokine production was found between FcR &#x003b3;-chain^-/- mice and controls. It is therefore possible that the first hours of ICA are mediated by a component other than Fc&#x003b3;Rs. This component may very well be complement. In addition to IL-1, complement split products have been shown to be important within this model. Depletion of complement generation by cobra venom factor treatment blocked ICA for the most part [<xref ref-type="bibr" rid="B35">35</xref>]. In the present study, complement C3c deposition did not differ in knee-joint structures of FcR &#x003b3;-chain^-/- and control mice, suggesting that complement activation does not differ between the two strains; this observation confirms earlier findings [<xref ref-type="bibr" rid="B8">8</xref>]. To further investigate whether complement or other mediators play a role in the enhanced inflammation in DBA/1 mice, we compared FcR&#x003b3;-chain^-/- mice in a DBA/1 background with FcR &#x003b3;-chain^+/+ mice. ICA did not develop in FcR &#x003b3;-chain deficient mice with a DBA/1 background, indicating that in the DBA/1 strain, also, FcRs are crucial mediators. Another difference may be the removal of immune complexes from the joint. Earlier studies revealed that absence of functional Fc&#x003b3;RI and RIII retarded the removal of insoluble immune complexes from the body [<xref ref-type="bibr" rid="B6">6</xref>]. We found, however, no differences in IgG deposition in the knee joints of the two mouse strains during the first days. The reason for this may be that the arthritogenic molecule we use to elicit ICA has an excellent retention, because of its positive charge. Cationic antigen will be bound electrostatically to the negatively charged structures of the joint [<xref ref-type="bibr" rid="B35">35</xref>,<xref ref-type="bibr" rid="B39">39</xref>]. Moreover, inflammation in the knee joint may be differently regulated than in other body compartments. The knee-joint cavity is covered mainly with macrophages, which are the first cells that come into contact with the immune complexes during ICA, whereas the kidney and skin contain a broader range of cell types handling immune complexes.</p><p>Next, we investigated ICA in knee joints of DBA/1 mice and found that the severity and the chronicity of arthritis corresponded with a higher level of expression of Fc&#x003b3;Rs on macrophages. In contrast, ICA in C57BL/6 mice was only mild, without a chronic phase. Macrophages in the knee-joint lining of normal DBA/1 mice stained more intensely than those of normal C57BL/6 mice, suggesting a higher basal level of Fc&#x003b3;RII/III expression on DBA/1 mice. Furthermore, higher Fc&#x003b3;R expression was not limited to joint macrophages. Peritoneal macrophages showed a twofold higher expression of Fc&#x003b3;Rs, suggesting that the higher basal expression of Fc&#x003b3;RII/III is found on all resident macrophages. Other cell types were also tested for differences in Fc&#x003b3;R expression. No significant differences of Fc&#x003b3;R expression were found on peripheral blood monocytes nor on a mixed B- and T-cell population, indicating that the differences in expression of Fc&#x003b3;Rs between C57BL/6 and DBA/1 are not found on all cells of the haematopoietic lineage (unpublished results). Antibodies to Fc&#x003b3;RI or Fc&#x003b3;RIII are not available and therefore we could not determine the different classes of Fc&#x003b3;R separately. Genetic differences in DBA/1 may be responsible for altered Fc&#x003b3;R expression, as Holmdahl <italic>et aI</italic> have pointed out [<xref ref-type="bibr" rid="B45">45</xref>]. This would imply that this mouse strain is at higher risk for immune-complex diseases in body compartments that contain macrophages. This is in line with findings in other studies [<xref ref-type="bibr" rid="B46">46</xref>]. The finding that FcR &#x003b3;-chain^-/-DBA/1 mice did not show any inflammation when ICA was induced strongly suggests that, in this strain, full-blown ICA is also mediated by Fc&#x003b3;Rs. Thus the differences in the severity and the chronicity of inflammation during ICA and in the severity of subsequent cartilage damage are likely to be caused by differences in Fc&#x003b3;R function or expression. The elevated expression of Fc&#x003b3;Rs on macrophages was also functional, as was found in experiments <italic>in vitro</italic> in which IL-1 production was measured after stimulation of macrophages with HAGGs. ICA induction caused much higher levels and longer-lasting production of IL-1 in the knee joints of DBA/1 mice than in the control C57BL/6 mice. IL-1 may well be the factor responsible for the chronicity of the synovial inflammation. Although IL-1 levels at day 3 were found to be below the detection limit of the RIA, blocking of IL-1 with anti-IL-1 antibodies starting at day 3 after ICA induction abrogated the chronic phase completely [<xref ref-type="bibr" rid="B34">34</xref>].</p><p>In addition to inflammation, cartilage destruction was also investigated. In earlier studies, we found that only in those experimental-arthritis models in which immune complexes are present was severe, irreparable cartilage damage found [<xref ref-type="bibr" rid="B44">44</xref>]. Mice injected with zymosan or SCW did not suffer irreparable cartilage damage, although a tremendous influx of inflammatory cells in the joint was observed.</p><p>In FcR &#x003b3;-chain^-/- mice, depletion of PG was entirely absent during ICA. In knee joints of DBA/1 mice, loss of PG from the cartilage matrix was more severe than in C57BL/6 mice. This finding is in line with the respective expression of Fc&#x003b3;R on macrophages in these strains and suggests that activation of Fc&#x003b3;R may be one of the mechanisms mediating PG loss. Activation of macrophages by Fc&#x003b3;RI and RIII leads to the release of IL-1 [<xref ref-type="bibr" rid="B47">47</xref>,<xref ref-type="bibr" rid="B48">48</xref>], which is the dominant cytokine involved in inhibition of PG synthesis [<xref ref-type="bibr" rid="B49">49</xref>,<xref ref-type="bibr" rid="B50">50</xref>] and production of aggrecan-degrading enzymes [<xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref>]. Depletion of PG within this model is probably mediated by aggrecanase [<xref ref-type="bibr" rid="B44">44</xref>], an enzyme that recently was identified as a metalloproteinase of the ADAM (A disintegrin and metalloproteinase) family [<xref ref-type="bibr" rid="B53">53</xref>]. Aggrecanase neoepitopes were abundantly found during the first phase of this arthritis [<xref ref-type="bibr" rid="B44">44</xref>]. Injection of zymosan into the knee joints of the three strains we studied was followed by a comparable depletion of PG; this finding suggests that the three strains have similar potencies to produce aggrecanase.</p><p>Severe cartilage destruction, like chondrocyte death and cartilage erosion, was found only in DBA/1 mice. During the acute phase, PMNs constitute the dominant cell type present in the joint, whereas during the chronic phase, macrophages are the predominant cell type present in the synovium and synovial cavity. Fc&#x003b3;R activation on both PMNs and macrophages may lead to the release of oxygen and nitrogen radicals, which in high concentrations can kill chondrocytes either by necrosis or apoptosis [<xref ref-type="bibr" rid="B54">54</xref>,<xref ref-type="bibr" rid="B55">55</xref>] or may activate latent enzymes in the matrix. Activated macrophages form an important source of latent metalloproteinases such as stromelysin and collagenase; after activation, these enzymes are involved in degradation of the collagen matrix, resulting in severe cartilage destruction [<xref ref-type="bibr" rid="B56">56</xref>,<xref ref-type="bibr" rid="B57">57</xref>]. Immune complexes activate macrophages in the synovium, causing the release of enzymes that are hindered in their activity by large amounts of inhibitors present in the synovial fluid. Immune complexes present on the cartilage surface may cause PMNs or macrophages to attach to this surface, as we have seen by clear pannus formation on the cartilage during chronic arthritis in the DBA/1 joint. This attachment to the cartilage surface may cause release of enzymes that directly penetrate the cartilage matrix thus escaping from their inhibitors (manuscript in preparation).</p><p>The present findings partly underline the findings in our previous study, in which an antigen-induced arthritis was produced in FcR &#x003b3;^-/-mice [<xref ref-type="bibr" rid="B58">58</xref>], in the sense that stimulation of Fc&#x003b3;Rs by immune complexes seems necessary in order to induce irreversible cartilage damage. However, in contrast to our present findings, antigen-induced arthritis in FcR &#x003b3;-chain^-/- mice did not significantly reduce joint inflammation. This indicates that within that model, FcRs do not play any role in the chronic phase of inflammation, but are of utmost importance in the induction of severe cartilage damage.</p><p>To provide conclusive evidence of the role of various classes of Fc&#x003b3;R during ICA, experiments are needed in which Fc&#x003b3;Rs are blocked with specific antibodies, or in which knockout mice are used that lack one of the specific classes of Fc&#x003b3;R. However, specific antibodies to Fc&#x003b3;Rs (2.4G2), once bound to the receptor, have a stimulatory effect and therefore cannot be used in blocking experiments. Experiments using specific knockout mice are therefore now being performed in our laboratory.</p><p>Macrophages are the dominant type of cell present in chronic inflammation during RA and their number has been shown to correlate well with severe cartilage destruction [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. Apart from that, in humans, these macrophages express activating Fc receptors, mainly Fc&#x003b3;IIIa [<xref ref-type="bibr" rid="B26">26</xref>], which may lead to activation of these cells by IgG-containing immune complexes. Fc receptor expression on the surface of these cells may have important implications for joint inflammation and severe cartilage destruction and may form new targets for therapeutic intervention.</p></sec>

Enhanced expression of genes involved in coagulation and fibrinolysis in murine arthritis

<sec><title>Introduction:</title><p>Accumulation of fibrin in the joints remains one of the most striking histopathological features of rheumatoid arthritis (RA). Recently, we have provided evidence of the deleterious role of synovial fibrin deposition in arthritic joints in antigen-induced arthritis (AIA), a well-established murine model of RA.</p><p>A local imbalance between fibrin formation and fibrin dissolution may result in fibrin deposition in the joints.</p><p>On the one hand, fibrin formation is mainly initiated by tissue factor (TF), a transmembrane protein serving as a receptor for factor VII. Under normal conditions, <italic>TF</italic> expression and activity are tightly regulated. Constitutive <italic>TF</italic> expression is restricted to perivascular and epithelial cells, and the catalytic activity of the TF/VIIa complex can be inhibited by tissue factor pathway inhibitor (TFPI). Pathological conditions can perturb the cell-type-restricted pattern of <italic>TF</italic> expression. In particular, recent reports have shown that transcriptional activation of <italic>TF</italic> can be mediated by molecular mechanisms involving induction of the early growth response gene 1 (<italic>EGR1</italic>) or of the protease-activated receptor (<italic>PAR1</italic>) or vascular endothelial growth factor (<italic>VEGF</italic>) genes.</p><p>On the other hand, fibrin degradation is mediated primarily by plasmin, which is the active form of the zymogen plasminogen. Conversion of plasminogen to plasmin is under the control of serine protease plasminogen activators, such as the urokinase plasminogen activator (uPA), and their inhibitors, such as the plasminogen activator inhibitor (PAI-1).</p></sec><sec><title>Aims:</title><p>We hypothesized that the deposition of fibrin in the joints may result from an imbalance in the local expression of key genes involved in coagulation and fibrinolytic pathways. To test this hypothesis, we investigated mRNA levels in arthritic versus nonarthritic joint tissues from two murine models of RA: AIA and collagen-induced arthritis (CIA). Genes that are directly implicated in coagulation (<italic>TF</italic>, <italic>TFPI</italic>) and fibrinolysis (<italic>UPA</italic>, <italic>PAI1</italic>), and other genes that may influence the expression of <italic>TF</italic> (<italic>EGR1</italic>, <italic>PAR1</italic>, <italic>VEGF</italic>), were investigated using a novel multiprobe RNase protection assay (RPA). Furthermore, we evaluated coagulation activity in arthritic and nonarthritic mice.</p></sec><sec><title>Methods:</title><p>Mice with AIA or CIA were sacrificed at different time points: 2, 4, and 16 h and 3, 7, and 14 d after intra-articular antigen injection for AIA; 42 d after the first immunization for CIA. Total RNA was prepared from arthritic and nonarthritic knees for AIA, or arthritic and nonarthritic hind paws for CIA. Messenger RNA (mRNA) levels of the genes described above were determined by RPA and normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA levels. Coagulation assays were performed on joint tissue extracts and concentrations of thrombin-antithrombin III (TAT) complex were measured in plasma.</p></sec><sec><title>Results:</title><p>In AIA, all the genes studied except <italic>VEGF</italic> were upmodulated as early as 2 h. <italic>PAR1</italic>, <italic>TFPI, EGR1,</italic> and <italic>UPA</italic> expression decreased to control levels by 16 h, whereas the expression of <italic>TF</italic> and <italic>PAI1</italic> remained elevated. At later times, only <italic>TF</italic>, <italic>PAI1</italic>, and <italic>UPA</italic> showed sustained overexpression. In CIA, gene expression was assayed at only one time point (42 d after immunization) and all genes showed higher mRNA levels in the affected paws than in control paws. In AIA mice, procoagulant activity and TF activity were significantly increased in arthritic joints, and in CIA mice, plasma TAT levels were significantly enhanced.</p></sec><sec><title>Discussion:</title><p>Fibrin deposition in synovia is prominent in both RA and experimental arthritis, suggesting that this protein may play a role in the pathogenesis of chronic inflammation. In this study, we have tried to shed some light on the molecular mechanisms leading to extravascular fibrin deposition, using two well-established mouse models of RA: AIA and CIA. The kinetics of gene expression was first analyzed in mice with AIA, because this model allows for an accurate, temporally controlled sampling of synovial inflammation. We then extended our observations by analyzing one time point in CIA, 42 d after immunization, when chronic inflammation is present. We found that in both models, coagulation and fibrinolysis in arthritic joints were significantly increased, and that the most significant increases were in TF and PAI-1.</p><p>Although the molecular mechanism or mechanisms responsible for the transcriptional changes observed are not completely understood, the increases in TF, PAI-1, and uPA are probably due to the production of proinflammatory cytokines such as IL-1 and TGF-&#x003b1;. These cytokines, whose presence in the inflamed synovium is well documented, are known to induce these genes through the activation of nuclear factor &#x003ba;B (NF-&#x003ba;B), a transcription factor. <italic>TF</italic> induction is also under the control of a proximal enhancer containing a binding site for the inducible transcription factor <italic>EGR1</italic>. Indeed, the early rise of <italic>EGR1</italic> expression in AIA is consistent with its classification as immediate-early gene and may be responsible for the induction of early expression of <italic>TF</italic>. Early <italic>TF</italic> stimulation in AIA can also be accounted for by the transient overexpression of <italic>PAR1</italic>. Contrary to what has been shown in RA, <italic>VEGF</italic> expression remained essentially unchanged throughout the progression of AIA, probably reflecting a peculiarity of this murine model.</p><p>The alteration of the patterns of gene expression was accompanied by increased functional coagulation activity, which was more marked in AIA than in CIA.</p></sec><sec><title>Conclusion:</title><p>Prominent fibrin deposition in two different animal models of RA &#x02013; AIA and CIA &#x02013; can be attributed to modulations in key regulatory genes for coagulation and fibrinolysis.</p></sec>

<contrib id="A1" contrib-type="author"><name><surname>Salvi</surname><given-names>Roberto</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>Nathalie.Busso@chuv.hospvd.ch</email></contrib><contrib id="A2" contrib-type="author"><name><surname>P&#x000e9;clat</surname><given-names>Veronique</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>So</surname><given-names>Alexander</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Busso</surname><given-names>Nathalie</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>natalie.busso@chuv.hospvd.ch</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Rheumatoid arthritis (RA) is a common autoimmune disease of unknown etiology, characterized by chronic synovial inflammation that leads to progressive destruction of cartilage and bone [<xref ref-type="bibr" rid="B1">1</xref>]. Immunological mechanisms are thought to initiate synovial inflammation, which becomes persistent with the disease progression. Among the many histopathological features described, one of the most striking is the accumulation of fibrin [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. We have recently provided evidence that synovial deposition of this protein plays a deleterious role in arthritic joints in antigen-induced arthritis (AIA), a well-established model of RA [<xref ref-type="bibr" rid="B4">4</xref>]. This accumulation of fibrin could result from a local imbalance between its formation and dissolution. Previous studies have revealed enhanced coagulation activity in rheumatoid synovial fluid and membrane [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B5">5</xref>] as well as increased activity of synovial urokinase plasminogen activator (uPA) in rheumatoid synovial membrane [<xref ref-type="bibr" rid="B6">6</xref>]. Little is known about the expression of procoagulant molecules in the arthritic synovial membrane, and the molecular events that tip the natural balance between synovial procoagulant and fibrinolysis in favor of coagulation remain to be elucidated.</p><p>Synovial fibrin deposition is mediated principally by tissue factor (TF), an activator of the extrinsic pathway of coagulation. TF is a transmembrane protein that initiates coagulation by serving as a cofactor for activated factor VII [<xref ref-type="bibr" rid="B7">7</xref>]. <italic>TF</italic> is constitutively expressed in perivascular and epithelial cells, but its expression can be induced on endothelial cells and monocytes by inflammation [<xref ref-type="bibr" rid="B8">8</xref>] and hypoxia [<xref ref-type="bibr" rid="B9">9</xref>]. Increased expression of <italic>TF</italic> during hypoxia is mediated by the transcription factor early growth response gene 1 (<italic>EGR1</italic>) product and leads to pulmonary fibrin deposition [<xref ref-type="bibr" rid="B9">9</xref>]. Tissue factor pathway inhibitor (TFPI), a naturally occuring Kunitz-type inhibitor, regulates coagulation by inhibiting the catalytic activity of the TF/VIIa complex [<xref ref-type="bibr" rid="B10">10</xref>]. Although little is known about its role in pathological processes, by virtue of its anticoagulant effect, under- or over-expression of this gene could alter the procoagulant/ anticoagulant balance.</p><p>Besides modulating fibrin deposition, activation of the coagulation pathway can also modulate genes that play an accessory role in inflammation by mediating cellular activation, for example thrombin receptor, protease-activated receptor 1 (PAR-1), and vascular endothelial growth factor (VEGF). PAR-1 is a G-protein-coupled receptor that binds thrombin [<xref ref-type="bibr" rid="B11">11</xref>] and is abundantly expressed in inflamed rheumatoid synovia [<xref ref-type="bibr" rid="B12">12</xref>]. Activation of <italic>PAR1</italic> by thrombin can lead to proliferation of synovial fibroblasts and rapidly induces the transcription of <italic>TF</italic> mRNA [<xref ref-type="bibr" rid="B13">13</xref>]. Since <italic>PAR1</italic> mRNA is itself upregulated by thrombin [<xref ref-type="bibr" rid="B14">14</xref>], <italic>PAR1</italic> may be part of a positive-feedback loop that potentiates the coagulation cascade. VEGF stimulates endothelial-cell proliferation <italic>in vitro</italic> and induces neovascularization <italic>in vivo</italic> [<xref ref-type="bibr" rid="B15">15</xref>]. Significant amounts of antigenic VEGF have been detected in synovial fluids and tissues from RA patients [<xref ref-type="bibr" rid="B16">16</xref>], and <italic>VEGF</italic> mRNA is abundantly expressed in highly vascularized areas of the RA synovial tissue [<xref ref-type="bibr" rid="B17">17</xref>]. A TF-dependent production of VEGF by human fibroblasts in response to activated factor VII binding has been reported [<xref ref-type="bibr" rid="B18">18</xref>]. Conversely, VEGF can induce the expression of <italic>TF</italic> [<xref ref-type="bibr" rid="B19">19</xref>] and of <italic>UPA</italic> [<xref ref-type="bibr" rid="B15">15</xref>], thereby influencing the coagulation/fibrinolysis equilibrium.</p><p>Degradation of extravascular fibrin is primarily mediated by plasmin, which is formed upon cleavage of plasminogen by plasminogen activators [<xref ref-type="bibr" rid="B20">20</xref>]. This fibrin degradation is supported by increased synovial uPA activity evidenced in human and experimental arthritis and by the observation that lack of uPA leads to increased fibrin deposition in the joint [<xref ref-type="bibr" rid="B4">4</xref>]. The activity of uPA is regulated by plasminogen activator inhibitors (PAIs), in particular PAI-1 [<xref ref-type="bibr" rid="B20">20</xref>]. Both <italic>UPA</italic> and <italic>PAI1</italic> expression are induced by cytokines that are abundant within the inflamed joint. [<xref ref-type="bibr" rid="B21">21</xref>].</p><p>In this study, we have analyzed the modulation of the mRNA levels of <italic>TF</italic>, <italic>TFPI</italic>, <italic>PAR1</italic>, <italic>EGR1</italic>, <italic>VEGF</italic>, <italic>UPA,</italic> and <italic>PAI1</italic> in the inflamed joint in two murine models of arthritis, namely AIA and collagen-induced arthritis (CIA). These two models recapitulate many features of RA, especially fibrin deposition in arthritic joints (for AIA, see [<xref ref-type="bibr" rid="B4">4</xref>]; for CIA, Marty <italic>et al</italic>, submitted). We have used a novel multiprobe RNase protection assay that was developed to facilitate the simultaneous analysis of all the genes studied from small amounts of tissue. Functional assays of procoagulant activity (PCA) and TF activity were also performed on these tissues. Evidence of ongoing coagulation was sought in arthritic mice by measuring the plasma concentration of TAT.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Induction of AIA</title><p>C57Bl/6 mice (Iffa-Credo, L'Arbresle, France) between 8 and 10 weeks old were immunized at day 0 and 7 with 100 &#x003bc;g of methylated bovine serum albumin (Sigma Chemical Company, Buchs, Switzerland) emulsified in 0.1 ml Freund's complete adjuvant containing 200 &#x003bc;g mycobacterial strain H37RA (Difco, Basel, Switzerland) by intradermal injection at the base of the tail. On the same day, 2 &#x000d7;10⁹ heat-killed <italic>Bordetella pertussis</italic> organisms (Berna, Bern, Switzerland) were injected intraperitoneally as additional adjuvant. Arthritis was induced at day 21 by intra-articular injection of 100 &#x003bc;g of methylated bovine serum albumin in 10 &#x003bc;l sterile phosphate-buffered saline solution (PBS) into the right knee, the left knee being injected with sterile PBS alone. Institutional approval was obtained for these experiments.</p></sec><sec><title>Induction of CIA</title><p>Male DBA/1J mice between 8 and 10 weeks of age were obtained from BRL/RCC Biotechnology &#x00026; Animal Breeding (F&#x000fc;llinsdorf, Switzerland). Native chicken type II collagen (MM Griffith, Salt Lake City, UT, USA) was dissolved in 0.1 M acetic acid overnight at 2 mg/ml. Collagen (100 &#x003bc;g) emulsified in Freund's complete adjuvant containing 5 mg/ml <italic>Mycobacterium tuberculosis</italic> was injected intradermally at the base of the tail. At day 24 after the first injection, a booster injection of 100 &#x003bc;g of native chicken collagen in Freund's incomplete adjuvant was given intradermally at the base of the tail. All immunization reagents were purchased from DIFCO (Basel, Switzerland). Clinical assesment of arthritis was performed in accordance with established protocols [<xref ref-type="bibr" rid="B22">22</xref>].</p></sec><sec><title>RNA extraction</title><p>Cryostat sections of synovial tissues from knee joints (AIA model) or cryostat sections of total paws (CIA model) were homogenized in Trizol reagent (Gibco BRL, Berne, Switzerland) and total RNA extractions were performed in accordance with the manufacturer's instructions.</p></sec><sec><title>DNA template set preparation for RNase protection assay (RPA)</title><p>Complementary DNA (cDNA) fragments of a different size for each of the chosen genes were recovered by reverse transcriptase-polymerase chain reaction (RT-PCR) and subsequently subcloned into the pGEM-T plasmid (Promega, Wallisen, Switzerland). Positive clones with antisense orientation with respect to the T7 promoter were identified by colony polymerase chain reaction (PCR) and then their complete sequence was verified by DNA sequencing. Table <xref ref-type="table" rid="T1">1</xref> summarizes the details for each probe.</p></sec><sec><title>Probe labeling and RPA</title><p>Radioactive multiriboprobe preparation and RPA were performed in accordance with standard protocols. Briefly, anti-sense a³²P-UTP-labeled riboprobes were synthesized by transcription <italic>in vitro</italic> of the DNA template set, which contained all the linearized plasmid cDNAs pooled in equimolar amounts. DNase I treatment was performed to remove the DNA templates and riboprobes were purified by phenol/chloroform extraction followed by ethanol precipitation using glycogen as carrier. For each sample, about 5 &#x003bc;g of total RNA was hybridized overnight at 52&#x000b0;C with 3 &#x000d7; 10⁵ cpm of the labeled multiprobe. Samples were first treated with an RNase cocktail (Ambion, Austin, TX, USA); RNase was then removed by treatment with proteinase K and samples were purified by extraction in phenol/chloroform followed by ethanol precipitation using glycogen as carrier. Protected fragments were resolved through a 5% sequencing gel. Precise quantification of all the experimental samples for all time points studied was determined by analyzing the gels with an InstantImager apparatus and software (Packard Instruments, Berne, Switzerland). For each protected band analyzed, a cpm/mm² value was obtained. This value was corrected for sample-loading errors by normalizing with the respective cpm/mm² value calculated for the constitutively expressed glyceraldehyde-3-phosphate dehydrogenase (<italic>GAPDH</italic>) gene, which was also included in the template set. Ratio values between arthritic/nonarthritic joints were then calculated for each time point.</p></sec><sec><title>Preparation of tissue extracts</title><p>Tissue extracts were prepared from cryostat sections (see RNA preparation above) as described before [<xref ref-type="bibr" rid="B4">4</xref>].</p></sec><sec><title>Measurement of procoagulant activity and <italic>TF</italic> activity</title><p>PCA was measured in tissue extracts from knee joints of mice with AIA or from paws of mice with CIA by a modified one-stage clotting assay. Briefly, tissue extracts (25 &#x003bc;l) were mixed with an equal volume of phospholipids (Dade-Behring, D&#x000fc;dingen, Switzerland). After adding 50 &#x003bc;l rabbit citrated plasma and 50 &#x003bc;l 0.02 M calcium chloride, time to thrombus formation was recorded using a micro-coagulometer (Dialine, Itingen, Switzerland). <italic>TF</italic> activity in tissue extracts was measured using a commercially available kit (American Diagnostica, Greenwich, CT, USA).</p></sec><sec><title>Determination of thrombin-antithrombin III</title><p>Citrated plasmas were obtained from arthritic and nonarthritic mice. Thrombin&#x02013;antithrombin III (TAT) concentration in plasma was measured using a commercially available ELISA (enzyme-linked immunosorbent assay) kit designed for human TAT (Enzygnost TAT, Dade-Behring, D&#x000fc;dingen, Switzerland), which cross-reacts also with murine TAT. The concentration of murine TAT was calculated according to the human TAT standard curve.</p></sec><sec><title>Statistical analysis</title><p>The unpaired Student's <italic>t</italic>-test was used to compare means for normally distributed values. A level of <italic>P</italic> &#x0003c; 0.05 was considered statistically significant.</p></sec></sec><sec><title>Results</title><sec><title>Detection of gene expression by multiprobe RNAse protection assay</title><p>A typical autoradiograph obtained with the multiprobe set is shown in Fig. <xref ref-type="fig" rid="F1">1a</xref>. It reveals the expression of eight probe-protected bands in RNA from mouse skin. The actual size of each of the protected fragments corresponded with their expected size (compare with Table <xref ref-type="table" rid="T1">1</xref>).</p></sec><sec><title>Confirmation of multiprobe specificity</title><p>To confirm the specificity of the bands detected, we analyzed RNA samples that either lacked a particular mRNA species or over- or under-expressed it.</p><p>In Fig. <xref ref-type="fig" rid="F1">1b</xref>, the patterns of the protected fragments detected in RNAs prepared from <italic>PAI1</italic>, <italic>UPA</italic>, and <italic>PAR1</italic> knockout mice were compared with those prepared from wild-type mice. As expected, no <italic>PAI1</italic>, <italic>UPA</italic>, or <italic>PAR1</italic> protected bands were detected in kidney RNA from the respective deficient mice. For <italic>TF</italic>, <italic>TFPI</italic>, and <italic>EGR1</italic>, we analyzed kidney RNA from lipopolysaccharide-treated mice (Fig. <xref ref-type="fig" rid="F1">1c</xref>), which confirmed the upregulation of <italic>TF</italic> and <italic>EGR1</italic> upon lipopolysaccharide treatment, while <italic>TFPI</italic> was downregulated [<xref ref-type="bibr" rid="B23">23</xref>]. For <italic>VEGF</italic>, we observed its striking upregulation in hypoxic endothelial cells using the multi-probe assay, whereas the mRNA was not detected in normoxic cells (Fig. <xref ref-type="fig" rid="F1">1d</xref>).</p></sec><sec><title>Synovial expression of procoagulant and fibrinolytic genes in AIA</title><p>Total RNAs were extracted from dissected synovial tissues of arthritic and nonarthritic joints (injected with PBS) at various time points (2, 4, and 16 h, and 3, 7, and 14 d) after AIA induction. A representative multiprobe assay autoradiograph showing the mRNA expression at 4 h and at 7 d in the arthritic and nonarthritic knees is shown in Fig. <xref ref-type="fig" rid="F2">2</xref>. As early as 4 h after AIA induction, there was a clear-cut upregulation of all the genes except <italic>VEGF</italic> in the arthritic (right) knee as compared with the uninvolved (left) knee. At 7 d after AIA induction, although increased expression of <italic>TF</italic> and <italic>PAI1</italic> mRNA was still evident in the arthritic joint, the increased expression of the other mRNA species was less marked than before. Analysis of the whole time course of expression yielded four major observations (Fig. <xref ref-type="fig" rid="F3">3</xref>):</p><p>1) <italic>PAR1</italic>, <italic>TFPI</italic>, <italic>EGR1</italic>, and <italic>UPA</italic> showed a rapid initial increase of mRNA levels, which peaked between 2 and 4 h, followed by a sharp decrease to near basal levels at 16 h.</p><p>2) The kinetics of <italic>PAI1</italic> and <italic>TF</italic> induction differed from that of the other genes studied. <italic>PAI1</italic> reached its peak level (a 3.5-fold increase in comparison with the control knee) at 16 h, followed by a slow decrease, so that the level was twofold by day 7 and 1.5-fold by day 14. <italic>TF</italic> expression increased gradually, peaking (to more than threefold) by day 7, and remaining at more than twofold by day 14.</p><p>3) At day 14, expression of <italic>TF</italic>, <italic>PAI1</italic>, and <italic>UPA</italic> was still increased (to about 1.4-fold each).</p><p>4) <italic>VEGF</italic> expression hardly changed during the experiment.</p></sec><sec><title>Increased expression of procoagulant and fibrinolytic genes changes in collagen-induced arthritis</title><p>The expression of the same genes during CIA was analyzed using the same multiprobe set. RNA samples from arthritic paws of collagen-immunized mice were compared with samples from nonimmunized control mice. For this experiment, we used only one time point. Animals were sacrified 42 d after the first immunization, when the incidence of arthritis reaches &#x0003e;80% and the severity, as evaluated by clinical scores, is maximal.</p><p>Figure <xref ref-type="fig" rid="F4">4a</xref> shows a representative autoradiograph demonstrating the upregulation of all the genes assayed in the arthritic paw in comparison with the control paw. <italic>TF</italic> and <italic>PAI1</italic> had the highest increases &#x02013; about fourfold and about threefold, respectively (Fig. <xref ref-type="fig" rid="F4">4b</xref>). All the other genes showed an approximately twofold increase over controls.</p></sec><sec><title>Increased procoagulant and TF activity in arthritic joints</title><p>Since the pattern of mRNA modulation observed in AIA and CIA strongly suggested a shift of the coagulation/fibrinolytic balance in favor of coagulation during joint inflammation, we next tried to confirm this prediction at a functional level. We first compared PCA in tissue extracts prepared from arthritic and nonarthritic joints (Fig. <xref ref-type="fig" rid="F5">5a</xref>). Time to reach thrombus formation was decreased in arthritic joints in both models (by 24% in CIA, and by 72% in AIA), although only the decrease in AIA was statistically significant. Because increased PCA in arthritic joints could result from increased <italic>TF</italic> activity, we also studied <italic>TF</italic> activity in these samples (Fig. <xref ref-type="fig" rid="F5">5b</xref>). It was increased in both models (by 78% in CIA, and by 365% in AIA), but only the increase in AIA was statistically significant.</p></sec><sec><title>Increased plasma TAT levels in arthritic mice</title><p>Because TAT is considered to be an indicator of ongoing activation of the clotting system, we measured its concentration in the plasma of arthritic mice. We found a statistically significant increase of TAT in the plasma of CIA animals (3.5 times its concentration in nonarthritic mice) but only a very modest increase (by 27%; not statistically significant) in AIA (Fig. <xref ref-type="fig" rid="F5">5c</xref>).</p></sec></sec><sec><title>Discussion</title><p>Excessive fibrin deposition has been shown to be deleterious in inflammation associated with various diseases such as hypoxia, sepsis, and RA [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Because the molecular mechanisms underlying fibrin deposition in RA have not been elucidated, we have characterized the alteration in coagulation and fibrinolytic gene expression in two murine models of RA. By analyzing the kinetics of gene expression in the AIA model, we were able to sample the very early phases of synovial inflammation. To extend our observations, we then analyzed total paw RNA prepared from CIA mice. In this model, synovial fibrin is also present and has a deleterious role in inflammation (Marty <italic>et al</italic>, submitted). In CIA, we studied gene expression at 42 d after immunization, when inflammation is clinically active. One possible drawback of using total paw RNA is that paws contain tissues in addition to synovia. Nevertheless, the use of total paw RNA has been shown to accurately reflect synovial RNA changes of cytokine levels in CIA [<xref ref-type="bibr" rid="B26">26</xref>].</p><p>An upregulation of both procoagulant and fibrinolytic genes was observed in both models. The pattern of expression in CIA at day 42 approximated that at the early time points in AIA, with <italic>TF</italic> and <italic>PAI1</italic> showing the most significant increases. If the protein expression of these molecules parallels their transcription, then the combined increase of TF and PAI-1 would favor fibrin formation and its persistence in the joint.</p><p>The transcriptional regulation of the changes noted are not completely understood. The increases in TF, PAI-1, and uPA are most likely to be due to the increased secretion of cytokines such as IL-1 and TNF-&#x003b1;, which are known inducers of these genes and whose presence in the synovium in AIA has been documented by other workers [<xref ref-type="bibr" rid="B27">27</xref>]. The transcriptional regulation of the <italic>TF</italic> gene probably involves activation of the transcription factor nuclear factor &#x003ba;B (NF-&#x003ba;B). Indeed, an enhancer in the <italic>TF</italic> gene containing an NF-&#x003ba;B site that is activated by endotoxin, TNF-&#x003b1;, and IL-1 has been found [<xref ref-type="bibr" rid="B28">28</xref>]. Moreover, findings in a recent study [<xref ref-type="bibr" rid="B29">29</xref>] suggest that induced <italic>TF</italic> and <italic>PAI1</italic> expression by vascular smooth muscle and endothelial cells may be simultaneously mediated by activation of NF-&#x003ba;B. In addition, <italic>TF</italic> expression is under the control of another proximal enhancer [<xref ref-type="bibr" rid="B28">28</xref>], containing overlapping binding sites for the constitutively expressed SP1 and the inducible <italic>EGR1</italic> transcription factor. The early rise of <italic>EGR1</italic> expression in AIA, at 2 to 4 h, is consistent with the immediate-early gene expression pattern of this gene and may be responsible for the induction of early expression of <italic>TF</italic>. However, by day 3, <italic>EGR1</italic> expression had declined to control levels, while <italic>TF</italic> expression remained elevated, an observation suggesting that <italic>EGR1</italic> does not contribute to the sustained overexpression of <italic>TF</italic> in AIA synovial tissues. Transient <italic>PAR1</italic> overexpression can also contribute to the early stimulation of <italic>TF</italic> [<xref ref-type="bibr" rid="B13">13</xref>]. <italic>VEGF</italic> expression remained essentially unchanged throughout the progression of AIA, with only a slight increase at the late time points. In CIA there was only a slight, not significant, increase of <italic>VEGF</italic> mRNA levels. This is somewhat surprising, in view of data showing increased expression in RA [<xref ref-type="bibr" rid="B17">17</xref>].</p><p>We found significantly increased functional coagulation activity in the AIA model but only a slight, nonsignificant increase in the CIA model. TF activity in AIA can be totally accounted for by changes in <italic>TF</italic> mRNA. By contrast, the discrepancy between TF activity and <italic>TF</italic> mRNA levels in CIA needs to be clarified. Furthermore, we were able to demonstrate increased plasma TAT levels in both models, though the difference was more striking in CIA than in AIA. Plasma TAT levels result most probably from the spillover into the systemic circulation of TAT that originated in arthritic joints. As multiple joints are affected in CIA, but only one in AIA, one could expect TAT values to be more elevated in CIA than in AIA. Altogether, these findings support the conclusion that articular inflammation is accompanied by significant upregulation of molecules that favor increased extravascular coagulation activity and fibrin deposition.</p><p>In conclusion, we have found that experimental arthritis is associated with a pronounced upregulation of both pro-coagulation and fibrinolytic genes, that in its later stages tends to favor the formation and persistence of fibrin. Workers using mice with targeted disruptions of either the fibrinogen gene and/or genes of the fibrinolytic pathway have observed that fibrin persistence may delay wound healing [<xref ref-type="bibr" rid="B30">30</xref>] and exacerbate progression of atherosclerosis [<xref ref-type="bibr" rid="B31">31</xref>], glomerulonephritis [<xref ref-type="bibr" rid="B32">32</xref>], and arthritis [<xref ref-type="bibr" rid="B4">4</xref>]. In the latter case, if efficient fibrinolytic mechanisms are not sufficiently activated in the course of joint inflammation, the persistence of joint fibrin may contribute to continued synovitis and subsequent joint damage.</p></sec>

Role of IL-12 in <italic>Staphylococcus aureus</italic>-triggered arthritis and sepsis

<p>The present study demonstrates that endogenous production of IL-12 is crucial for survival in <italic>Staphylococcus aureus</italic>-induced arthritis in mice. Staphylococcal load is enhanced in several organs, because of lack of IL-12. This might be due to decreased production of IFN-&#x003b3; in IL-12-deficient mice. Although IL-12-deficient mice were exposed to higher staphylococcal load, they demonstrated no increased severity of arthritis as compared with control animals.</p>

<contrib id="A1" contrib-type="author"><name><surname>Hultgren</surname><given-names>Olof H</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>olof.hultgren@immuno.gu.se</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Stenson</surname><given-names>Martin</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Tarkowski</surname><given-names>Andrzej</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Synopsis</title><sec><title>Introduction:</title><p>Septic arthritis is a severe disease, which is associated with high mortality and rapid destruction of affected joints. The most common bacterial agent in this condition is <italic>S aureus</italic>, which is the responsible pathogen in 37&#x02013;56% of all cases of septic arthritis. Despite eradication of bacteria from the joint cavity, destruction of the joint often continues, resulting in severe sequelae.</p><p>In our mouse model of sepsis and septic arthritis we use an intravenous inoculum of a <italic>S aureus</italic> strain (LS-1) that produces toxic shock syndrome toxin (TSST)-1. This bacterial strain was originally isolated in a mouse with spontaneous staphylococcal arthritis. IL-12 is a heterodimeric cytokine that is composed of the constitutively expressed p35 gene product and the inducible p40 subunit. It is primarily produced by monocytes/macrophages and dendritic cells.</p><p>IL-12 has a variety of effects on natural killer cells and T cells, including the ability to facilitate T-helper (Th1)-cell responses, and thereby IFN-&#x003b3; production. IFN-&#x003b3; has been shown to be arthritogenic in septic arthritis, but protective with regard to bacterial clearance and survival. IL-12 is protective in several experimental models of bacterial infections, as demonstrated by the fact that neutralization of this cytokine increases susceptibility and addition of recombinant IL-12 ameliorates the severity of the infection. However, IL-12 given at high doses induces a septic shock-like condition, and neutralization of IL-12 protects mice from lipopolysaccharide-induced shock. Thus, the role of IL-12 in cases of severe bacterial infection that culminate in septic shock is not established.</p><p>It has recently been shown that IL-12 deficiency exists in humans and that absence of IL-12 gives rise to recurrent infections with <italic>S aureus</italic>. The role of IL-12 in <italic>S aureus</italic> infection has not previously been assessed. Inoculation of IL-12 p40-deficient mice and their wild-type counterparts with a TSST-1-producing <italic>S aureus</italic> strain shows the critical importance of IL-12 for survival during <italic>S aureus</italic> sepsis, in that it mediates downregulation of staphylococcal growth.</p><p>The aim of the present study was to investigate the importance of IL-12 in <italic>S aureus</italic> arthritis, specifically its impact on survival, bacterial clearance and development of arthritis.</p></sec><sec><title>Materials and methods:</title><p>Inbred male C57BL/6 mice that were intact or defective with respect to IL-12 production were used throughout the study. At the time of the experiment, IL-12-deficient mice had undergone five backcrosses to C57BL/6. A TSST-1-producing <italic>S aureus</italic> strain (LS-1), which was originally isolated from a spontaneously arthritic NZB/W mouse, was prepared according to a previously described method and was injected into one of the tail veins.</p></sec><sec><title>Results:</title><p>Ten days after inoculation with staphylococci all of the IL-12^-/- mice were dead, as compared with 22% of the wild-type control animals (Fig. <xref ref-type="fig" rid="F1">1</xref>; <italic>P</italic> = 0.002). These results clearly indicate the critical importance of IL-12 production for survival during infection with <italic>S aureus</italic>.</p><p>Bacterial growth was determined in liver, kidneys and joints 20 days after intravenous inoculation with 1 &#x000d7; 10⁷ staphylococci/mouse. The clearance of bacteria was clearly diminished in IL-12-deficient mice in all three organs.</p><p>Serum levels of IFN-&#x003b3; were measured in IL-12-deficient mice and wild-type control animals after intravenous inoculation of 1 &#x000d7; 10⁷ staphylococci/mouse. IL-12-deficient mice displayed decreased IFN-&#x003b3; levels as measured at days 7 and 20 after the inoculation of bacteria as compared with levels in wild-type controls (day 7, 208 &#x000b1; 83 versus 764 &#x000b1; 288 U/ml; and day 20, 269 &#x000b1; 114 versus 444 &#x000b1; 170 U/ml; these results were not statistically significant).</p><p>The disruption of the <italic>p40</italic> gene did not result in any significant differences with regard to either the frequency or severity of arthritis as compared with wild-type controls. The clinically observed frequency of septic arthritis was 76% in IL-12^+/+ mice, as compared to 54% in IL-12^-/- animals 20 days after inoculation of 1 &#x000d7; 10⁷ staphylococci/mouse. Histopathological examination confirmed the clinical data (Table <xref ref-type="table" rid="T1">1</xref>).</p></sec><sec><title>Discussion:</title><p>The importance of functional IL-12 during <italic>S aureus</italic> arthritis is demonstrated by the present study. The major impact of IL-12 is on survival. This protection seems to be via more efficient bacterial clearance <italic>in vivo</italic>. All arthritic parameters, clinically and histopathologically, indicated decreased severity in IL-12-deficient mice as compared with wild-type controls; however, statistical significance was not reached. It is possible that differences in joint inflammation and erosion in response to staphylococci were more pronounced in IL-12^+/+ versus IL-12^-/- mice, if the same amount of bacteria were harboured in the joints. The levels of IFN-&#x003b3; are decreased in the absence of IL-12 production, although this was not statistically significant. This was expected, because IL-12 induces IFN-&#x003b3; production by natural killer cells and promotes type 1 T-cell differentiation, cells known to produce IFN-&#x003b3;. Because IFN-&#x003b3; is important for phagocytic functions as well as promoting development of septic arthritis, downregulation of IFN-&#x003b3; production in IL-12^-/- mice may explain the outcome of <italic>S aureus</italic>-induced arthritis.</p></sec></sec><sec><title>Introduction</title><p>Sepsis caused by <italic>S aureus</italic> is a life-threatening condition that may cause septic shock, resulting in multiple organ failure and ultimately death [<xref ref-type="bibr" rid="B1">1</xref>]. Septic arthritis is a severe disease, involving high mortality and rapid destruction of affected joints [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>]. The most common bacterial agent in this condition is <italic>S aureus</italic>, which is the responsible pathogen in 37-56% of all cases of septic arthritis [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. Despite eradication of bacteria from the joint cavity destruction of the joint often continues, resulting in severe sequelae [<xref ref-type="bibr" rid="B9">9</xref>]. Several efforts have been made to down-modulate the inflammatory responses that lead to septic death and arthritis [<xref ref-type="bibr" rid="B10">10</xref>]. One problem in these attempts is the risk of increasing staphylococcal growth <italic>in vivo</italic>. In our mouse model of sepsis and septic arthritis, we use an intravenous inoculum of a TSST-1-producing <italic>S aureus</italic> strain (LS-1) [<xref ref-type="bibr" rid="B11">11</xref>]. This bacterial strain was originally isolated from a mouse with spontaneous staphylococcal arthritis [<xref ref-type="bibr" rid="B12">12</xref>].</p><p>IL-12 is a heterodimeric cytokine that is composed of the constitutively expressed p35 gene product and the inducible p40 subunit [<xref ref-type="bibr" rid="B13">13</xref>]. It is primarily produced by monocytes/macrophages and dendritic cells [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. It has a variety of effects on natural killer cells and T cells, including the ability to faclitate Th1-cell responses and thereby IFN-&#x003b3; production [<xref ref-type="bibr" rid="B16">16</xref>].</p><p>IFN-&#x003b3; is a classical Th1 cytokine. Modulation of IFN-&#x003b3; levels using neutralizing monoclonal antibodies or addition of recombinant cytokines show the beneficial effect of IFN-&#x003b3; in terms of defence against <italic>S aureus</italic> infection, decreasing bacterial load in tissues and increasing survival. Results from these experiments also indicate an arthritogenic role of IFN-&#x003b3; in <italic>S aureus</italic> infections [<xref ref-type="bibr" rid="B17">17</xref>].</p><p>The role of IL-12 in several experimental models of bacterial infection is protective, as demonstrated by the fact that neutralization of this cytokine increases susceptibility and addition of recombinant IL-12 ameliorates the severity of the infection [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. However, IL-12 given at high doses induces a septic shock-like condition, and neutralization of IL-12 protects mice from lipopolysaccharide-induced shock [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>]. Thus, the role of IL-12 in cases of severe bacterial infection that culminate ending in septic shock is not established. It has recently been shown that IL-12 deficiency exists in humans [<xref ref-type="bibr" rid="B23">23</xref>] and that absence of IL-12 may give rise to recurrent infections with <italic>S aureus</italic> [<xref ref-type="bibr" rid="B24">24</xref>]. The role of IL-12 in <italic>S aureus</italic> infection has not previously been assessed. Inoculation of IL-12 p40-deficient mice and their wild-type counterparts with a TSST-1-producing <italic>S aureus</italic> strain demonstrates the critical importance of IL-12 for survival during <italic>S aureus</italic> sepsis due to IL-12-dependent downregulation of staphylococcal growth, and the arthritogenic role of IL-12 in septic arthritis.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Mice, bacteria and infection</title><p>Inbred male C57BL/6 mice that were intact or defective with respect to IL-12 production were used throughout the present study. IL-12 p40^-/- mice were kindly provided by Dr J Magram (Nutley, NJ, USA). The procedure of gene disruption and mouse phenotype has previously been described in detail [<xref ref-type="bibr" rid="B25">25</xref>]. At the time of experiment, IL-12-deficient mice had undergone five back-crosses to C57BL/6. Mice were maintained in the animal facility of the Department of Rheumatology, University of G&#x000f6;teborg, Sweden, and were kept under standard temperature and light conditions. The mice were housed in a pathogen-free environment. They were fed laboratory chow and water <italic>ad libitum</italic>.</p><p>A TSST-1-producing <italic>S aureus</italic> strain (LS-1), which was originally isolated from a spontaneously arthritic NZB/W mouse, was prepared according to a previously described method [<xref ref-type="bibr" rid="B26">26</xref>] and injected into one of the tail veins. Two different doses of bacteria (1 &#x000d7; 10⁷ and 4 &#x000d7; 10⁸ staphylococci/mouse) were used. Viable counts were used to check the numbers of staphylococci injected.</p></sec><sec><title>Clinical evaluation of arthritis and weight</title><p>All mice were labelled and monitored individually. Limbs were inspected by two observers at regular intervals (3, 7, 12 and 20 days). Arthritis was defined as visible erythema and/or swelling of at least one joint. In order to evaluate the intensity of arthritis, we used a clinical scoring system in which macroscopic inspection yields a score of 0&#x02013;3 for each paw (0, normal; 1, mild swelling and/or erythema; 2, moderate swelling and erythema; and 3, marked swelling and occasionally ankylosis), resulting in an arthritic score ranging from 0 to 12 for each mouse. An arthritic index was constructed by dividing the sum of scores from all four limbs in each mouse by the number of animals in each experimental group. Weight was checked at days 0, 3, 7, 12 and 20.</p></sec><sec><title>Experimental protocol</title><p>In the first experiment 16 IL-12^-/- mice and nine IL-12^+/+controls were inoculated intravenously with 1 &#x000d7; 10⁷ staphylococci/mouse. Arthritis and weight were checked at regular intervals. The experiment was finished after 20 days, when mice were bled and killed. Kidneys were aseptically removed and bacterial growth was determined by viable counts. All four limbs were histopathologically examined.</p><p>In the second experiment, 21 IL-12^-/- and 20 IL-12^+/+ controls were inoculated with 1 &#x000d7; 10⁷ staphylococci/mouse. Arthritis and weight were checked at regular intervals. Six hours after staphylococcal inoculation, five mice from each group were bled for bacterial examination and cytokine analyses. Two days after inoculation, five mice from each group were killed and bacterial growth in, liver, kidneys and joints were examined. The same procedure was performed at day 7 (IL-12^-/-, <italic>n</italic> = 6; IL-12^+/+, <italic>n</italic> = 5). The surviving mice were bled and killed at day 20.</p><p>In a third experiment, a high dose of <italic>S aureus</italic> (4 &#x000d7; 10⁸ staphylococci/mouse) were given intravenously to nine IL-12^-/- mice and nine IL-12 control animals. Survival was checked daily.</p></sec><sec><title>Cytokine and immunoglobulin analyses</title><p>Microtitre plates were coated with 2 &#x003bc;g/ml rat antimouse IFN-&#x003b3; monoclonal antibodies (Pharmingen, San Diego, CA, USA) dissolved in sodium bicarbonate (pH 9.6), and blocked with 1% bovine serum albumin dissolved in 0.05 mol/l Tris (pH 7.4) for 1 h before adding samples for a 2-h incubation. Recombinant mouse IFN-&#x003b3; (Genzyme, Cambridge, MO, USA) was used to create a standard curve. Biotinylated rat antimouse IFN-&#x003b3; (2 &#x003bc;g/ml; Pharmingen) was employed as the capture antibody. The plates were kept for 2 h at 37&#x000b0;C, and then incubated with streptavidin alkaline phosphatase (Dako A/S, Glostrup, Denmark) and alkaline phosphatase substrate 1 mg/ml (Sigma, St Louis, MO, USA). Absorbance was measured at 405 nm in a Titretec multiscan photometer (Flow Laboratories, McLean, WA, USA). IL-10 was analyzed using an optEIA ELISA (Pharmingen) Assays were performed in accordance with the manufacturer's instructions, except for the addition of 2 mol/l HCl, instead of 2 mol/l H₂SO₄.</p><p>Serum levels of IgG₁, IgG_{2<italic>a</italic>} and IgG₃ were measured using the radial immunodiffusion technique [<xref ref-type="bibr" rid="B27">27</xref>]. Antisera and immunoglobulin standards were purchased from Sigma.</p></sec><sec><title>Determination of bacterial load</title><p>Bacterial samples from talocrural and radiocarpal joints were obtained using cotton sticks. Bacterial presence was defined as 15 colony-forming units or more for each joint. The liver and both kidneys were removed aseptically, placed on ice, homogenized and diluted in 10 ml phosphate-buffered saline. Viable counts were performed to determine bacterial concentration. Colonies from every plate were then tested using a Staphaurex kit (Murex Diagnostcs, Dartford, UK).</p></sec><sec><title>Phagocytosis and intracellular killing</title><p>Intraperitoneal macrophages from noninfected mice were extracted, adjusted to 2 &#x000d7; 10⁶ cells/ml, and incubated in a 42-well plate (Nunc, Roskilde, Denmark) according to procedures detailed previously [<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>]. Adherent macrophages were incubated with 500 ml <italic>S aureus</italic> at a concentration of 5 &#x000d7; 10⁶ bacteria/ml for 50 min at 37&#x000b0;C, and were subsequently washed three times in Iscove's medium. Macrophage content of bacteria was then measured after two incubation intervals (0 and 4 h) in order to study phagocytosis and intracellular killing capacity, respectively. In order to avoid extracellular bacterial growth in the intracellular study, the incubation medium contained 10 mg/ml gentamycin. The antibiotic was washed away before lysing macrophages with distilled water. Macrophages were extracted from three wild-type mice and three IL-12-deficient mice. Macrophages from IL-12-deficient animals were preincubated for 12 h with 0, 10, or 100 ng/ml of recombinant IL-12 (R &#x00026; D Systems, Abingdon, Oxon, UK) before adding bacteria. Supernatants were collected from macrophage cultures before exposure to <italic>S aureus</italic>, and after 4 h of incubation with bacteria.</p></sec><sec><title>Histopathologic examination</title><p>Limbs were fixed in 4% paraformaldehyde, decalcified and embedded in paraffin. Tissue sections were prepared and stained with haematoxylin and eosin. Sections were examined by two blinded observers, who determined whether synovial hypertrophy (membrane thickness of more than two cell layers), pannus formation (joint cartilage covered with synovial tissue), or destruction of cartilage and bone were present.</p></sec><sec><title>Statistical analysis</title><p>Mann&#x02013;Whitney U test and Fisher's exact test were used for statistical analyses. <italic>P</italic> &#x0003c; 0.05 was considered statistically significant. Values are expressed as mean &#x000b1; standard error of the mean.</p></sec></sec><sec><title>Results</title><sec><title>IL-12 production protects the host from fatal <italic>S aureus</italic>-triggered sepsis</title><p>C57BL/6 mice defective (IL-12^-/-) or intact (IL-12^+/+) with respect to the <italic>p40</italic> gene, the inducible part of the heterodimeric IL-12 molecule, were injected intravenously with 4 &#x000d7; 10⁸ staphylococci/mouse. The high dose of bacteria used in this experiment was chosen in order to study the influence of IL-12 on Gram-positive bacteria-triggered septic death. No mice died within the first 4 days. The majority of deaths occurred more than 1 week after inoculation of staphylococci. Ten days after inoculation of staphylococci, all of the IL-12^-/- mice had died, as compared with 22% of the wild-type control animals (<italic>P</italic> = 0.002; Fig. <xref ref-type="fig" rid="F1">1</xref>). In order to determine whether the mortality rate of the IL-12^+/+ mice would increase during the later stage of infection, we terminated the experiment at 3 weeks after the bacterial inoculation. No further control animals died during the course of the experiment. These results clearly indicate the critical importance of IL-12 production for survival during infection with <italic>S aureus</italic>.</p><p>Using a smaller inoculum (1 &#x000d7; 10⁷ staphylococci/mouse) mortality rates were overall low (IL-12-deficient mice 6% [one out of 17] and controls 11% [one out of nine]). Interestingly, with this lower dose of bacteria the weight decrease was more pronounced in the wild-type mice as compared with the IL-12-deficient animals in the early phase (after 3 days) of infection (10.6% in IL-12^+/+ mice versus 7.3% in IL-12^-/- mice; <italic>P</italic> = 0.03). The differences between groups regarding weight change then declined, and at the end of the experiment the wild-type mice had regained their initial weight loss, whereas mice that lacked IL-12 production did not (-0.3% versus -4.2%; not significant).</p></sec><sec><title>Influence of IL-12 deficiency on <italic>S aureus</italic>-induced arthritis</title><p>The clinically observed frequency of septic arthritis was 76% in IL-12^+/+ mice as compared with 54% in IL-12^-/- animals 20 days after inoculation of 1 &#x000d7; 10⁷ staphylococci/mouse. At that time, the severity of arthritis was graded as 1.3 in IL-12^+/+ mice versus 0.9 in IL-12^-/- mice. Histopathological examination confirmed the clinical findings (Table <xref ref-type="table" rid="T1">1</xref>). Despite the fact that all parameters regarding development of joint inflammation and destruction of bone and cartilage indicated greater severity in mice with an intact IL-12 gene, none of the parameters tested reached the level of statistical significance.</p></sec><sec><title>Staphylococcal load is increased in several tissues due to the absence of IL-12 production</title><p>Bacterial content was determined in liver, kidneys and joints 20 days after intravenous inoculation with 1 &#x000d7; 10⁷ staphylococci/mouse. The clearance of bacteria was clearly diminished in IL-12-deficient mice in all three organs. A fivefold increase in bacterial counts was seen in the livers of IL-12^-/- mice as compared with wild-type controls (Fig. <xref ref-type="fig" rid="F2">2a</xref>). In kidneys, the bacterial load was 100-fold higher in IL-12^-/-mice as compared with IL-12^+/+ controls (Fig. <xref ref-type="fig" rid="F2">2b</xref>). Finally, <italic>S aureus</italic> was not detected in joints of IL-12^+/+ mice, whereas 75% of IL-12^-/- mice had staphylococci in their joints (Fig. <xref ref-type="fig" rid="F2">2c</xref>). We conclude that the decreased bacterial clearance in IL-12-deficient mice as compared with in control animals is probably the reason for the enhanced incidence of <italic>S aureus</italic>-induced death in the former.</p><p>We also determined staphylococcal load in blood, liver and kidneys earlier during the infection. Surprisingly, bacterial content in blood was decreased in IL-12-deficient mice as compared with that in the wild-type control animals at 6 h, as well as in joints, liver and kidneys at 2 days after inoculation of bacteria (Figs <xref ref-type="fig" rid="F2">2a</xref>,<xref ref-type="fig" rid="F2">b</xref>,<xref ref-type="fig" rid="F2">c</xref>,<xref ref-type="fig" rid="F2">d</xref>). This initial beneficial effect of IL-12 deficiency on the clearance of staphylococci may explain the significantly lower weight loss measured 3 days after bacterial inoculation in IL-12^-/-as compared with IL-12^+/+ mice.</p><p>In order to study the importance of IL-12 on phagocytosis and intracellular killing capacity, we extracted intraperitoneal macrophages from IL-12-deficient mice and their wild-type counterparts. To simplify our method, we conducted this part of the study in noninfected mice, because <italic>in vivo</italic> activated macrophages would have encountered different bacterial loads as a result of differences in the <italic>in vivo</italic> clearance of <italic>S aureus</italic>, and therefore would have been differentially activated. No differences in phagocytosis or intracellular killing were seen (data not shown). Furthermore, recombinant mouse IL-12 (0, 10 and 100 ng/ml) was added to the IL-12-deficient macrophage cultures 12 h before exposure to staphylococci. Addition of recombinant IL-12 did not affect either phagocytic activity or intracellular killing (data not shown).</p></sec><sec><title>Decreased levels but continuous production of IFN-&#x003b3; during <italic>S aureus</italic>-induced arthritis in IL-12-deficient mice</title><p>Serum levels of IFN-&#x003b3; were measured in IL-12-deficient mice and wild-type control animals after intravenous inoculation of 1 &#x000d7; 10⁷ staphylococci/mouse; Table <xref ref-type="table" rid="T2">2</xref>). IL-12-deficient mice displayed decreased IFN-&#x003b3; levels, measured at days 7 and 20 after the inoculation of bacteria, as compared with wild-type controls (day 7, 208 &#x000b1; 83 versus 764 &#x000b1; 288 U/ml; day 20, 269 &#x000b1; 114 versus 444 &#x000b1; 170 U/ml; these findings were not statistically significant). The difference in serum levels of IFN-&#x003b3; did not reach statistical significance, but results at 7 and 20 days after intravenous inoculation of <italic>S aureus</italic> indicate, as expected, higher IFN-&#x003b3; levels in IL-12 p40-intact mice, suggesting a stronger Th1 response in wild-type mice as compared with IL-12 p40-defective mice.</p><p>Because IL-10 production is induced by IL-12 [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>], and IL-12-induced IL-10 production could contribute to an explanation for the better condition of IL-12^-/- mice observed during the early phase of <italic>S aureus</italic> infection, we analyzed serum levels of IL-10. We did not find detectable levels of IL-10, either in IL-12^+/+ or in IL-12^-/- mice.</p><p>Twenty days after intravenous inoculation of staphylococci, immunoglobulin responses were higher in IL-12^-/-mice than in IL-12^+/+ mice (Table <xref ref-type="table" rid="T3">3</xref>). This was indicated by higher serum levels of IgG₁, IgG_{2<italic>a</italic>} and IgG₃ in the knock-out mice. Notably, levels of IgG_2a were not reduced, as seen in previous studies on IL-12-deficient mice, which is putatively due to lower levels of IFN-&#x003b3; [<xref ref-type="bibr" rid="B34">34</xref>]. Similar to the present results, IgG_2a :IgG₁ ratio was not substantially reduced in IFN-&#x003b3; receptor-deficient mice exposed to superantigen-producing staphylococci [<xref ref-type="bibr" rid="B35">35</xref>] or in IL-12-deficient mice subjected to viral infections [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B37">37</xref>]. The increased total IgG production might be a result of a more pronounced Th2 response, and thereby B-cell activation in absence of IL-12 production. Alternatively, an increased staphylococcal load in the IL-12^-/- mice might have triggered a more pronounced immunoglobulin production.</p></sec></sec><sec><title>Discussion</title><p>The present study demonstrates that IL-12 deficiency leads to striking increase in mortality during <italic>S aureus</italic> sepsis. These results suggest that the presence of IL-12 is required for efficient control of staphylococcal growth <italic>in vivo</italic>. Furthermore, although exposed to higher bacterial load, IL-12^-/- mice did not display increased joint pathology.</p><p>The increased mortality of IL-12^-/- mice as compared with their wild-type counterparts was seen in parallel with an increased staphylococcal load in the tissues examined. It is plausible that the increased bacterial growth was the main reason for the sharply increased death rates in the absence of IL-12 production.</p><p>In order to determine whether IL-12 itself had any direct effect on the phagocytosis and/or intracellular killing capacity, we employed two different approaches using intraperitoneal macrophages from IL-12^-/- and IL-12^+/+ mice. No differences in intracellular killing of bacteria were seen. Indeed, supplementing IL-12^-/- macrophages with recombinant IL-12 neither affected the uptake nor killing of staphylococci. This is in contrast to the beneficial antifungal effects that IL-12 p40^-/- macrophages exert during infection with <italic>Candida albicans</italic> [<xref ref-type="bibr" rid="B38">38</xref>].</p><p>IL-12 is known to induce development of Th1 cells and their subsequent differentation to produce IFN-&#x003b3; [<xref ref-type="bibr" rid="B39">39</xref>]. A recent study [<xref ref-type="bibr" rid="B17">17</xref>] showed the importance of IFN-&#x003b3; in recruitment of neutrophils and the increased phagocytosis during staphylococcal infection. Macrophages, despite not being a major producer of IFN-&#x003b3;, have been shown to synthesize this cytokine after stimulation with IL-12 and IL-18, but not with IL-12 alone [<xref ref-type="bibr" rid="B40">40</xref>]. This might have been the main reason for the lack of direct effect of IL-12 on staphylococcal growth <italic>in vitro</italic>. <italic>In vivo</italic> data showing decreased serum levels of IFN-&#x003b3; in IL-12^-/-mice as compared with IL-12^+/+ control animals support deficient IL-12-dependent IFN-&#x003b3; production as an explanation for the decreased bacterial clearance in IL-12^-/- mice. It is thus plausible that IL-12 is protective with regard to <italic>S aureus</italic>-triggered death via enhanced phagocytic activity mediated by IFN-&#x003b3;.</p><p>An alternative or parallel explanation for the observed effect of IL-12 deficiency on decreased survival during <italic>S aureus</italic>-induced sepsis is the previously described effects of this cytokine on IL-10 production. IL-12 induces IL-10 production, and thereby downregulates production of tumour necrosis factor (TNF), IFN-&#x003b3; and nitric oxide [<xref ref-type="bibr" rid="B41">41</xref>], with TNF and IFN-&#x003b3; being detrimental in septic shock [<xref ref-type="bibr" rid="B42">42</xref>,<xref ref-type="bibr" rid="B43">43</xref>,<xref ref-type="bibr" rid="B44">44</xref>]. Indeed, supplementation with recombinant IL-10 protects against death both from lipopolysaccharide-triggered enotoxemic shock [<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>] and staphylococcal enterotoxin B-induced shock [<xref ref-type="bibr" rid="B47">47</xref>]. This might be relevant because the higher bacterial cell counts in IL-12-deficient mice potentially lead to increased production of TSST-1 as compared with controls. However, we were not able to detect IL-10 production in IL-12^+/+ or IL-12^-/- mice, despite analyzing serum at several time points after bacterial inoculation. Downregulated IL-10 production in the absence of IL-12 could be a reason for the enhanced bacterial clearance seen in IL-12-deficient mice as compared with controls during the early phase of <italic>S aureus</italic> infection.</p><p>It is somewhat surprising that, although IL-12-deficient mice have a defect in bacterial clearance and therefore are exposed to a higher staphylococcal load than their wild-type counterparts, they did not develop a more severe septic arthritis, but rather milder joint involvement. A similar outcome has previously been shown using TNF/lymphotoxin-&#x003b1; double mutants [<xref ref-type="bibr" rid="B48">48</xref>]. The mechanisms that underlie these results may again be explained by the decreased production of IFN-&#x003b3; in IL-12^-/- mice versus IL-12^+/+ animals. Previous studies implicate IFN-&#x003b3; as a cytokine that promotes induction of septic arthritis, in addition to its importance in phagocytic activity.</p><p>The present results clearly demonstrate the critical role played by endogenous IL-12 in preventing death during <italic>S aureus</italic>-induced arthritis. The protection results from decreasing the bacterial load, an effect that is possibly due to IL-12-dependent IFN-&#x003b3; mediation of phagocytic activity.</p></sec>

Synovial cytokine mRNA expression during arthritis triggered by CpG motifs of bacterial DNA

<p>Our results show that cytokines derived from macrophages play an important role in pathogenesis of arthritis triggered by CpG oligodinucleotide (CpG ODN). IL-12 is in this respect an important immunomodulator during the development of joint inflammation.</p>

<contrib id="A1" contrib-type="author"><name><surname>Deng</surname><given-names>Guo-Min</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>guo-min@immuno.gu.se</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Tarkowski</surname><given-names>Andrej</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Synopsis</title><sec><title>Introduction:</title><p>Bacterial infections can be localized to the joints, causing septic arthritis, the most rapidly progressing joint disease. We have recently shown that unmethylated CpG motifs of bacterial DNA give rise to arthritis characterized by an influx of monocytic, Mac-1 antibody positive (Mac-1⁺ ) cells and by a scarcity of T lymphocytes. Cytokines have been shown to exert an important influence in the pathogenesis of arthritis in several mouse models. Tumor necrosis factor (TNF)-&#x003b1;, IL-1&#x003b2;, IFN-&#x003b3;, and IL-12 are all produced in various quantities in the joints of patients with rheumatoid arthritis and in experimental (eg collagen-induced and septic) arthritides.</p></sec><sec><title>Aims:</title><p>To investigate patterns of local cytokine mRNA expression for IL-1&#x003b2;, IL-12, TNF-&#x003b1;, and IFN-&#x003b3; in mice with arthritis induced by CpG ODN and the role of IL-12 in the development of CpG ODN-induced arthritis.</p></sec><sec><title>Methods:</title><p>CpG ODN was injected into the knee joints of mice, which were then killed after various intervals (0, 1, 3, 7, 14, or 21 d). At the end of each time interval, the synovial tissues were excised under an inverted microscope. We analyzed the mRNA expression of the cytokines in synovial tissue using a technique of hybridization <italic>in situ</italic>. IL-12 P40 knockout mice were used in an analysis of the role of IL-12 in the development of CpG ODN-induced arthritis.</p></sec><sec><title>Results:</title><p>None of the cytokine mRNAs studied was detected in the synovia of mice injected intra-articularly with calf thymus DNA or phosphate-buffered saline solution. In contrast, after intra-articular injection of CpG ODN, the monocyte/macrophage-derived cytokines TNF-&#x003b1;, IL-1&#x003b2;, and IL-12 were induced rapidly, being detected within the first day. The expression of TNF-&#x003b1; mRNA peaked on day 3 and then decreased, whereas IL-1&#x003b2; mRNA expression was high from day 1 onwards. IL-12 mRNA rose to peak values between days 3 and 21. The T helper (Th)1 cytokine IFN-&#x003b3; mRNA was undetected throughout the experiment. The arthritis had a lower incidence and was less severe in IL-12 knockout mice than in their congenic littermates. The frequency of TNF-&#x003b1; and IL-1&#x003b2; mRNA expression in synovia was lower at day 3 in IL-12 knockout mice than in wild-type mice, while IFN-&#x003b3; mRNA expression was not detectable. <italic>In vitro</italic>, TNF-&#x003b1; levels were lower in supernatants from mononuclear cells originating from IL-12 knockout (IL^{&#x02013;/&#x02013;}) mice and incubated with CpG ODN than in corresponding supernatants originating from IL-12^+/+ mice.</p></sec><sec><title>Discussion:</title><p>Previous studies have shown that bacterial DNA and CpG ODN directly activate macrophages to secrete proinflammatory cytokines, such as TNF-&#x003b1;, IL-1&#x003b2;, and IL-12. These cytokines exert proinflammatory activities in both septic and aseptic arthritides. TNF-&#x003b1; is produced primarily by monocytes and macrophages and stimulates macrophage production of IL-1. These two cytokines interact synergistically, stimulating each other's release and thereby amplifying the cascade of other inflammatory mediators.</p><p>IL-12 too is produced primarily by monocytes/macrophages, but mostly in response to microbial agents. It induces differentiation of Th1 cells and the production of IFN-&#x003b3; by natural killer and T cells and is involved in the inflammatory cascade as synovitis develops. Mice that are genetically unable to produce IL-12 have a decreased incidence of septic arthritis. The present study shows sustained expression of IL-12 mRNA in synovia of CpG ODN-triggered arthritis. Histopathological examination showed that arthritis had a lower incidence and was less severe in IL-12 knockout mice than in control mice. Although the differences were not statistically significant, these findings suggest that IL-12 may play a role in the pathogenesis of arthritis triggered by CpG ODN.</p><p>How does IL-12 participate in the induction of this arthritis? Our data suggest that TNF-&#x003b1; and IL-1&#x003b2; mRNA are expressed less strongly in synovial tissue of IL-12 knockout mice than in that of of wild-type controls. Furthermore, the levels of TNF-&#x003b1; released <italic>in vitro</italic> by spleen mononuclear cells in response to stimulation with CpG ODN were markedly lower in IL-12 knockout mice. IL-12, one of the cytokines produced in joints, has proinflammatory properties.</p></sec></sec><sec><title>Introduction</title><p>Several reports on the immunostimulatory properties of bacterial DNA have recently been published. Bacterial DNA directly activates B cells, monocytes, macrophages, and dendritic cells <italic>in vitro</italic> to upregulate their expression of costimulatory molecules that drive immune responses and secrete a variety of cytokines, including high levels of interleukin (IL)-12, IL-1, and tumor necrosis factor (TNF)-&#x003b1; [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Bacterial DNA indirectly activates natural killer (NK) cells and T cells [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>], whereas vertebrate DNA lacks immunostimulatory effects. Unmethylated CpG motifs are common in bacterial DNA and considerably less common in vertebrate DNA [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. In addition, whereas CpG motifs in bacterial DNA are unmethylated, the great majority of C and G nucleotides are methylated in all eukaryotic organisms, including mammals [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. Unmethylated CpG oligodinucleotides (CpG ODNs) are responsible for the immunostimulatory properties of bacterial DNA [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>].</p><p>Our group has recently reported that intra-articular bacterial DNA induces arthritis [<xref ref-type="bibr" rid="B10">10</xref>]. Histopathological signs of the arthritis were evident within two hours and lasted for at least three weeks, and it was characterized by an influx of monocytic, Mac-1⁺ cells and a scarcity of T lymphocytes. Unmethylated CpG motifs were responsible for the induction of this arthritis. This proinflammatory effect of bacterial DNA did not appear to be caused by contamination with endotoxins, since mice that did not respond to lipopolysaccharides developed arthritis in response to CpG ODNs but not in response to non-DNA bacterial contamination. Neither T cells, B cells, NK cells, nor neutrophils were found to be mandatory for induction of CpG ODN-mediated arthritis, whereas macrophages played a major role in induction of arthritis triggered by CpG motifs in bacterial DNA [<xref ref-type="bibr" rid="B10">10</xref>].</p><p>Cytokines have been shown to exert an important role in the pathogenesis of arthritis in several mouse models. TNF-&#x003b1;, IL-1&#x003b2;, IFN-&#x003b3;, and IL-12 are all produced in various quantities in the joints of patients with rheumatoid arthritis and in experimental arthritides such as collagen-induced [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>] and septic [<xref ref-type="bibr" rid="B15">15</xref>] arthritis. These cytokines play an important role in the induction and development of aseptic and septic arthritis [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. To better understand the pathogenesis of CpG ODN-mediated arthritis, we wanted to know more about the expression and role of these cytokines during its early phase. We therefore investigated the patterns of local cytokine mRNA using hybridization <italic>in situ</italic> and assessed the role of IL-12 in the induction of CpG ODN-mediated arthritis.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Mice</title><p>C57BL/6 mice were purchased from ALAB (Stockholm, Sweden). IL-12 P40 knockout mice were kindly provided by Dr J Magram (Nutley, NJ, USA) [<xref ref-type="bibr" rid="B21">21</xref>]. All mice were housed in the animal facility of the Department of Rheumatology, University of G&#x000f6;teborg. Male mice 6&#x02013;8 weeks of age were used in all the experiments.</p></sec><sec><title>Oligonucleotides and injection</title><p>Phosphorothioate-modified oligonucleotide (CpG ODN) 1668 were synthesized by Scandinavian Gene Synthesis AB (K&#x000f6;ping, Sweden). The sequence of oligodinucleotide (ODN) 1668 (containing the CpG motif) has been reported elsewhere [<xref ref-type="bibr" rid="B10">10</xref>]: 5'-TCC ATG ACG TTC CTG ATG CT-3'. CpG ODN (6 &#x003bc;g in a volume of 20 &#x003bc;l) was injected into the knee joints of the mice.</p></sec><sec><title>Tissue preparation</title><p>The mice were killed 0, 1, 3, 7, 14, or 21 d after inoculation. At the end of each time interval, the synovial tissues from the joints of four animals were excised under an inverted microscope. These tissues were then snap-frozen in OCT^TM compound (Tissue-TeK^&#x000ae;; Sakura Finetek Europe B.V., The Netherlands) by immersion in liquid nitrogen. Frozen tissue was stored at -70&#x000b0;C until use. Serial 6-&#x003bc;m sections were cut and thaw-mounted onto probe-on-slides (Fisher Scientific, Pittsburgh, PA, USA).</p></sec><sec><title>Hybridization <italic>in situ</italic></title><p>Hybridization was conducted <italic>in situ</italic> to analyze cytokine mRNA expression, as previously detailed [<xref ref-type="bibr" rid="B10">10</xref>]. Briefly, synthetic oligonucleotide probes (Table <xref ref-type="table" rid="T1">1</xref>) &#x02014; TNF-&#x003b1;, IFN-&#x003b3;, IL-1&#x003b2;, and IL-12 (the gift of Dr Tomas Olsson, Karolinska Institute, Stockholm, Sweden) &#x02014; were labeled at the 3' end using terminal deoxynucleotidyl transferase (Advanced Biotechnologies, Leatherhead, UK) and [³⁵S]ATP (Dupont Scandinavia, Stockholm, Sweden). Sections (6 &#x003bc;m thick) of freshly frozen synovial tissues were thaw-mounted onto slides and hybridized with 1 &#x000d7; 10⁶ cpm of labeled probe per 100 &#x003bc;l hybridization mixture. After emulsion autoradiography, development, and fixation, the coded slides were examined by dark-field microscopy for positive cells, which were defined as those containing &#x0003e;15 silver grains in a star-like distribution. The intracellular distribution of the grains was checked by light microscopy at higher magnification. The data were expressed as the number of cells (mean &#x000b1; SEM) expressing mRNA per mm² of the tissue section.</p></sec><sec><title>Histopathological examination</title><p>Joints were fixed, decalcified, and embedded in paraffin for histopathological examination. Tissue sections from knee joints were cut and stained with hematoxylin&#x02013;eosin. All the slides were coded and evaluated blind. The specimens were evaluated with regard to synovial hypertrophy, pannus formation, and destruction of cartilage and sub-chondral bone [<xref ref-type="bibr" rid="B22">22</xref>].</p></sec><sec><title>TNF-&#x003b1; levels in supernatant and serum</title><p>Spleen mononuclear cells were prepared as described previously [<xref ref-type="bibr" rid="B20">20</xref>]. The cells (1&#x000d7;10⁶/ml) were cultured in Iscove's complete medium (10% fetal calf serum, 5 &#x000d7; 10^-5 m 2-mercaptoethanol, 2 mm L-glutamine, and 50 &#x003bc;g/ml gentamicin) and stimulated with 1 &#x003bc;g/ml lipopolysaccharides or 1 &#x003bc;m CpG ODN. The cultures were maintained in 24-well plates (Nunc; Roskilde, Denmark) at 37&#x000b0;C in 5% CO₂ and 95% humidity. The supernatants were collected after 18 h for determination of TNF-&#x003b1;. TNF-&#x003b1; levels in supernatant and serum taken from IL-12 knockout and wild-type mice at day 3 after intra-articular inoculation with 6 &#x003bc;g CpG ODN were analyzed using a TNF-&#x003b1; ELISA kit (Genzyme, Cambridge, MA, USA).</p></sec><sec><title>Statistical analysis</title><p>The differences between mean values were tested for significance with the Fisher's Exact test and the Mann&#x02013;Whitney U test.</p></sec></sec><sec><title>Results</title><sec><title>Kinetics of synovial cytokine mRNA expression</title><p>Various workers have suggested that locally released cytokines are a key mediator of the inflammation and joint destruction observed in inflammatory arthritis [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. To assess the local induction of cytokines, we analyzed the expression of their mRNA in synovia, using hybridization <italic>in situ</italic>. Figure <xref ref-type="fig" rid="F1">1</xref> shows a typical result. None of the cytokine mRNAs studied were detected in the synovia of mice given intra-articular injections of calf-thymus DNA or phosphate-buffered saline solution. In contrast, after intra-articular inoculation with CpG ODN, the monocyte/macrophage-derived cytokines TNF-&#x003b1;, IL-1&#x003b2;, and IL-12 were induced rapidly, being detected within the first day. The expression of TNF-&#x003b1; mRNA peaked on day 3 and then decreased, whereas that of IL-1&#x003b2; mRNA rose on day 1 and remained high from then onwards. IL-12 mRNA rose to peak values between days 3 and 21. The Th1 cytokine IFN-&#x003b3; mRNA was undetectable throughout the experiment (Fig. <xref ref-type="fig" rid="F2">2a</xref>,<xref ref-type="fig" rid="F2">b</xref>).</p></sec><sec><title>Histopathological examination</title><p>Since IL-12 mRNA expression was high during arthritis induced by CpG ODN, we studied its role further using IL-12 knockout mice. The incidence (62.5%) and severity (0.75 &#x000b1; 0.75) of arthritis were decreased in IL-12 knockout mice as compared with their congenic littermates (100% and 1.33 &#x000b1; 0.5, respectively) (Fig. <xref ref-type="fig" rid="F3">3a</xref>,<xref ref-type="fig" rid="F3">b</xref>). These differences do not reach statistical significance.</p></sec><sec><title>Proinflammatory cytokine mRNA levels</title><p>To explore how IL-12 influenced the CpG ODN-triggered arthritis, synovial tissues taken from the knees of wild-type and IL-12 knockout mice at day 3 after intra-articular inoculation with CpG ODN were evaluated for TNF-&#x003b1;, IL-1&#x003b2;, and IFN-&#x003b3; mRNA expression using hybridization <italic>in situ.</italic> As shown in Fig. <xref ref-type="fig" rid="F4">4</xref>, the frequency of TNF-&#x003b1; and IL-1&#x003b2; mRNA expression in synovia was lower at day 3 in IL-12 knockout mice than in wild-type mice, and IFN-&#x003b3; mRNA expression was undetectable.</p></sec><sec><title>TNF-&#x003b1; levels in supernatant and serum</title><p>To assess why the incidence and severity of CpG ODN-triggered arthritis were decreased in IL-12 knockout mice, we compared TNF-&#x003b1; levels in IL-12 knockout mice with those in wild-type mice, using CpG ODN as stimulus <italic>in vitro</italic> or <italic>in vivo</italic>. The reason for this approach is that TNF-&#x003b1; is an essential mediator of CpG ODN-mediated arthritis [<xref ref-type="bibr" rid="B10">10</xref>]. TNF-&#x003b1; levels in supernatants from mononuclear cells stimulated with CpG ODN were lower in IL-12 knockout mice than in the wild-type control mice (Fig. <xref ref-type="fig" rid="F5">5</xref>). In contrast, the TNF-&#x003b1; levels were similar in supernatants from mononuclear cells stimulated with lipopolysaccharides, irrespective of IL-12 phenotype (Fig. <xref ref-type="fig" rid="F5">5</xref>). This suggests that the TNF-&#x003b1; response by macrophages to CpG ODN stimulation is at least partly influenced by expression of IL-12. Serum TNF-&#x003b1; levels collected at day 3 after intra-articular inoculation with 6 &#x003bc;g CpG ODN from wild-type and IL-12 knockout mice did not show detectable amounts of this cytokine.</p></sec></sec><sec><title>Discussion</title><p>Bacterial DNA in general, and unmethylated CpG ODN in particular, trigger joint inflammation and thus may play a pathogenic role in septic arthritis. Macrophages participate in innate cellular immunity and initiate many host defense responses. How does CpG ODN activate macrophages and induce arthritis? Previous studies have demonstrated that bacterial DNA and CpG ODN activate macrophages directly [<xref ref-type="bibr" rid="B23">23</xref>]. The first step of activation comprises the uptake of bacterial DNA or synthetic oligonucleotides by macrophages in a saturable, sequence-independent, temperature- and energy-dependent fashion [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>] into an acidified intracellular compartment, where DNA becomes degraded to oligodeoxynucleotides [<xref ref-type="bibr" rid="B26">26</xref>]. Once there, unmethylated CpG dinucleotides activate the stress-kinase/jun pathway within minutes, yielding transcriptionally active activating protein-1 and NF-&#x003ba;B [<xref ref-type="bibr" rid="B27">27</xref>]. These transcription factors control mRNA expression of a variety of cytokines and secretion of proinflammatory cytokines, such as TNF-&#x003b1;, IL-1&#x003b2;, and IL-12 [<xref ref-type="bibr" rid="B28">28</xref>]. These cytokines are considered to exert proinflammatory activities both in septic and aseptic arthritides [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>].</p><p>TNF-&#x003b1; is produced primarily by monocytes and macrophages and acts as a mediator of inflammation and host responses to invasion by microbes [<xref ref-type="bibr" rid="B29">29</xref>]. It activates endothelial cells, upregulates expression of adhesion molecules, and stimulates macrophage production of IL-1 [<xref ref-type="bibr" rid="B29">29</xref>]. IL-1 and TNF-&#x003b1; act synergistically, stimulating each other's release and thereby amplifying the cascade of other inflammatory mediators [<xref ref-type="bibr" rid="B16">16</xref>]. High levels of TNF-&#x003b1; and IL-1 are found in the joints of patients with rheumatoid arthritis and such experimental arthritides as collagen-induced arthritis and septic arthritis [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>]. A single intra-articular injection of IL-1&#x003b2; or TNF-&#x003b1; can induce acute synovitis [<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>]. In collagen-induced arthritis, neutralization of TNF-&#x003b1; and IL-1 lessens inflammation and joint destruction [<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>]. The incidence and severity of this arthritis are reduced significantly in TNF-&#x003b1; knockout mice [<xref ref-type="bibr" rid="B10">10</xref>].</p><p>IL-12 is a heterodimer consisting of two disulfide-linked subunits (P35 and P40). It is produced mainly by monocytes and macrophages, mostly in response to microbes. It induces differentiation of Th1 cells and the production of IFN-&#x003b3; by NK and T cells. It takes part in the inflammatory cascade during the development of synovitis. Previous studies have shown that the role of IL-12 in the development of collagen-induced arthritis appears to depend upon a number of factors, including the timing of its administration, its dosage, and the immunization protocol used [<xref ref-type="bibr" rid="B33">33</xref>,<xref ref-type="bibr" rid="B34">34</xref>]. In addition, mice that are genetically unable to produce IL-12 display a decreased incidence of septic arthritis [<xref ref-type="bibr" rid="B35">35</xref>]. The present study shows sustained expression of IL-12 mRNA in synovia of mice with CpG ODN-triggered arthritis. Histopathological examination revealed that both the incidence and the severity of arthritis in IL-12 knockout mice were approximately half those in control animals, although the large within-group variation kept this difference from reaching statistical significance (see Fig. <xref ref-type="fig" rid="F3">3b</xref>). Taken altogether, these data suggest that IL-12 plays a role in the pathogenesis of arthritis triggered by CpG ODN. Further experiments should be performed to confirm this relation.</p><p>How might IL-12 contribute to the induction of CpG ODN-triggered arthritis? Previous studies showed that it might exert its influence in three ways. First, IFN-&#x003b3; is an important intermediate for the action of IL-12, which is known to be able to induce IFN-&#x003b3; production by NK and T cells [<xref ref-type="bibr" rid="B36">36</xref>]. However, our data show that IFN-&#x003b3; mRNA was expressed neither in wild-type nor in IL-12 knockout mice. This finding suggests that the amelioration of CpG ODN-triggered arthritis in IL-12 knockout mice was not due to downregulation of IFN-&#x003b3; production. A second possible means by which IL-12 exerts its influence is through its upregulation of B-cell production of autoantibodies [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B37">37</xref>]. Again, our previous studies revealed that B cells are not important for induction of this arthritis [<xref ref-type="bibr" rid="B10">10</xref>]. Finally, IL-12 could be thought to promote arthritis by favoring the production of proinflammatory cytokines other than IFN-&#x003b3;. Indeed, previous studies in a model of pulmonary mycobacterial infection suggested that IL-12 is necessary for the local release of TNF-&#x003b1; [<xref ref-type="bibr" rid="B38">38</xref>,<xref ref-type="bibr" rid="B39">39</xref>], and that neutralization of IL-12 using monoclonal antibodies lowers production of TNF-&#x003b1;. We have recently shown that TNF-&#x003b1; is of importance in mediation of both septic [<xref ref-type="bibr" rid="B19">19</xref>] and CpG ODN-triggered arthritis [<xref ref-type="bibr" rid="B10">10</xref>]. In contrast to macrophage-derived cytokines, IFN-&#x003b3; mRNA expression was not detected in arthritic synovia. This finding is not surprising, since there is a scarcity of T cells, a major source of IFN-&#x003b3;, in CpG ODN-triggered arthritis [<xref ref-type="bibr" rid="B10">10</xref>], and NK cells are not important mediators for this arthritis [<xref ref-type="bibr" rid="B40">40</xref>].</p><p>Our data present <italic>in situ</italic> cytokine expression in CpG ODN-triggered arthritis. In the light of these data, we believe that IL-12, one of the cytokines produced in joints, has pro-inflammatory properties.</p></sec>

Rheumatoid synovial CD4⁺ T cells exhibit a reduced capacity to differentiate into IL-4-producing T-helper-2 effector cells

<p>CD4⁺ memory T cells (Tm) from rheumatoid arthritis peripheral blood (RAPB) or peripheral blood from normal donors produced IL-2, whereas fewer cells secreted IFN-&#x003b3; or IL-4 after a brief stimulation. RAPB Tm contained significantly more IFN-&#x003b3; producers than normal cells. Many rheumatoid arthritis (RA) synovial Tm produced IFN-&#x003b3; alone (40%) and fewer cells produced IL-2 or IL-4. An <italic>in vitro</italic> model was employed to generate polarized T-helper (Th) effectors. Normal and RAPB Tm differentiated into both IFN-&#x003b3;- and IL-4-producing effectors. RA synovial fluid (RASF) Tm demonstrated defective responsiveness, exhibiting diminished differentiation of IL-4 effectors, whereas RA synovial tissue (RAST) Tm exhibited defective generation of IFN-&#x003b3; and IL-4 producers.</p>

<contrib id="A1" contrib-type="author"><name><surname>Davis</surname><given-names>Laurie S</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>laurie.davis@UTSouthwestern.edu</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Cush</surname><given-names>John J</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Schulze-Koops</surname><given-names>Hendrik</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Lipsky</surname><given-names>Peter E</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>RA is an autoimmune disease that is characterized by persistent local and systemic inflammation [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Aggressive forms of RA are thought to be driven by T-lymphocyte activity, leading to the rapid erosion of bone and cartilage. Activated Tm are found in the perivascular regions of the inflamed synovium, and T cells accumulate in the synovial fluid [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. It appears that memory T cells retain the potential to traffic through the synovium and lymphatics and recirculate in the blood, as evidenced by the increased numbers of activated memory T cells found in the blood of some RA patients [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. It is not known, however, whether Tm migrate into the synovium and undergo activation and further differentiation, or whether activation occurs in regional lymph nodes before migration into the inflammatory site.</p><p><italic>In vitro</italic> studies [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>] have demonstrated that synovial CD4⁺ T cells proliferate poorly and produce decreased levels of IL-2 in response to mitogen or antigen, as compared with peripheral blood T cells from healthy individuals or RA patients. Although such findings suggest that synovial T cells might be partly anergic, more recent studies have indicated that the T cells found in the RA synovium were postactivated cells. Synovial T cells were activated as assessed by expression of CD40 ligand/CD154, and were also efficient helpers for B-cell immunoglobulin production [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>]. Previous studies in animal models [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>] suggested that a balance exists between immunoregulatory IFN-&#x003b3;-producing Th1 cells and IL-4-producing Th2 cells. Dominance of either subset can result in a chronic disease state. Moreover, induction of Th2 cells in a Th1-mediated disease model of collagen-induced arthritis led to amelioration of autoimmune disease [<xref ref-type="bibr" rid="B16">16</xref>]. Thus, it is possible that, in RA, there is biased differentiation of IFN-&#x003b3;-secreting proinflammatory Th1 CD4⁺ T cells, and insufficient differentiation of immunoregulatory IL-4-producing Th2 cells.</p><p>Human memory T cells acquire the ability to secrete IL-4 late during <italic>in vivo</italic> differentiation. Our previous studies and work by others have demonstrated that all of the <italic>in vivo</italic> differentiated IL-4-producing T cells reside within the mature memory (CD45RO⁺, CD27^&#x02013;) subset of CD4⁺ T cells [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Thus, early memory (CD45RO⁺, CD27⁺) CD4⁺ T cells cannot secrete IL-4 after a brief <italic>in vitro</italic> stimulation. Previous studies [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>] suggested that early memory T cells represent an uncommitted precursor population, from which both IL-4-producing and IFN-&#x003b3;-producing effector cells can be generated. The RA synovium appears to contain an increased number of phenotypically mature memory T cells as compared with the blood. However, the majority of memory T cells found in the synovium bear the phenotype of early Tm [<xref ref-type="bibr" rid="B7">7</xref>].</p><p>Therefore, the present study was undertaken to examine the cytokine effector status of synovial memory T cells and the functional status of immature memory cell precursors of cytokine-producing effector cells. For these experiments, Tm were isolated from the blood of normal subjects, and from RAPB, RASF and RAST. Intracellular cytokine expression was assessed by flow cytometry immediately after isolation and a brief <italic>in vitro</italic> stimulation, or after <italic>in vitro</italic> priming designed to generate IL-4-producing effector T cells.</p><p>The results suggest that a deficiency in the generation of adequate numbers of regulatory IL-4-producing effector cells in the synovium might be a contributing factor to the perpetuation of chronic inflammation.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Antibodies and reagents</title><p>MAbs used included the following: anti-CD3 (OKT3; American Type Culture Collection, Rockville, MD, USA), anti-CD8 (OKT8; American Type Culture Collection), anti-CD16 (B73.1; generous gift of Dr G Trinchieri, The Wistar Institute, Philadelphia, PA, USA), anti-CD19 (Dako, Glostrup, Denmark), anti-CD45RO (UCHL1; Dako), and anti-CD45RA (2H4; gift of Dr C Morimoto, Boston, MA, USA) [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. Anti-CD28 (28.2), Phycoerythrin (PE)-labeled anti-IL-2, PE-labeled anti-IL-4, and fluorescein isothiocyanate (FITC)-labeled anti-IFN-&#x003b3; was from Pharmingen (San Diego, CA, USA). FITC-labeled anti-CD3, and PE-labeled or Quantum red-labeled anti-CD4 was from Sigma (St Louis, MO, USA). Recombinant IL-4 and recombinant IL-15 were from R&#x00026;D Systems (Minneapolis, MN, USA). Neutralizing antihuman IFN-&#x003b3; antibody was from Endogen (Cambridge, MA, USA). Culture medium contained RPMI1640 with penicillin G (200 U/ml), gentamicin (10 &#x003bc;g/ml), L-glutamine (0.3 mg/ml), and 10% normal human serum [<xref ref-type="bibr" rid="B17">17</xref>].</p></sec><sec><title>Patients</title><p>All patients had an established diagnosis of RA, as defined by the 1987 revised criteria of the American College of Rheumatology for the classification of RA. All patients had long-standing, active RA. All samples were obtained after informed consent, as approved by the UT Southwestern Institutional Review Board. The data shown were obtained from 14 RA patients (age range 29&#x02013;78 years). Matching blood and synovial fluid were obtained from six patients. Peripheral blood alone was obtained from one patient and synovial fluid samples were obtained from three patients. Matching blood and synovial tissue samples were obtained from three patients at the time of surgery. An additional synovial tissue sample was obtained from one patient.</p></sec><sec><title>Cell preparation</title><p>Mononuclear cells were isolated from synovial tissue by treating the synovial tissue with collagenase briefly at 37&#x000b0;C with constant shaking. The cells were washed, filtered through nylon mesh, and briefly adhered to plastic petri dishes at 37&#x000b0;C to deplete residual fibroblast and monocytes [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. Mononuclear cells were obtained from blood or synovial samples of RA patients, or sex-matched and age-matched healthy adult individuals by ficoll density centrifugation. Cells were washed and incubated with neuraminidase-treated sheep red blood cells, and purified by negative selection panning as previously described [<xref ref-type="bibr" rid="B17">17</xref>]. The cells were incubated with anti-CD8, anti-CD19, B73.1, and 2H4 mAbs. After panning, the recovered cells were routinely greater than 99% CD3⁺, 95% CD4⁺, and 90% enriched for CD45RO⁺ cells.</p></sec><sec><title>Cell culture</title><p>T cells were immediately stimulated after isolation with ionomycin (0.5 &#x003bc;M; Sigma), phytohemagglutinin (2 &#x003bc;g/ml; Wellcome Diagnostics, Greenville, NC, USA), and phorbol myristate acetate (20 ng/ml; Sigma) in the presence of monensin (2 &#x003bc;M; Sigma) for 6 h and were fixed in paraformaldehyde for subsequent intracellular cytokine analysis. In some experiments, T cells were primed for 1 week before cytokine analysis [<xref ref-type="bibr" rid="B17">17</xref>]. T cells (5 &#x000d7;10⁵ cells/ml) were cultured in 96-well plates (Costar, Cambridge, MA, USA). Where indicated, wells were coated with anti-CD3 mAb (OKT3) before addition of T cells. Medium contained 10 U/ml recombinant IL-2, 5 ng/ml recombinant IL-15, 50 ng/ml recombinant IL-4, anti-CD28 mAb (0.5 &#x003bc;g/ml) and, where indicated, neutralizing anti-IFN-&#x003b3; antibody (5 &#x003bc;g/ml). After 1 week in culture the cells were harvested and counted. The cells were washed, and incubated for 60 h in medium containing IL-2 (10 U/ml) and IL-15 (5 ng/ml) before restimulation. In preliminary experiments the presence of IL-4 during the rest phase did not increase the number of IL-4-producing cells detected on subsequent restimulation. Restimulation and intracellular cytokine analysis were carried out as described above for immediately stimulated T cells [<xref ref-type="bibr" rid="B17">17</xref>].</p></sec><sec><title>Flow cytometry</title><p>T cells (3 &#x000d7; 10⁵) were stained with saturating amounts of directly conjugated mAb and analyzed on the fluorescence-activated cell sorter (FACScan; Becton Dickinson, San Jose, CA, USA). For intracellular staining of cytokines, cells were stimulated as described above. The cells were harvested, washed, fixed with paraformaldehyde, perme-abilized with saponin, and blocked with rodent serum. The cells were stained with directly conjugated mAbs to human cytokines (Pharmingen) and analyzed by flow cytometry [<xref ref-type="bibr" rid="B17">17</xref>]. Stimulated cells that were stained with irrelevant mAbs were used as controls. Analysis was carried out on 10⁴ cells for each sample.</p></sec><sec><title>Statistical analysis</title><p>Statistically significant differences in the percentages of cells that expressed each cytokine were determined using the two-tailed Student's <italic>t</italic>-test, or the Mann&#x02013;Whitney Rank Sum when appropriate. <italic>P</italic> &#x0003c; 0.05 was considered statistically significant.</p></sec></sec><sec><title>Results</title><sec><title>Enrichment of IFN-&#x003b3; producers in freshly isolated RA synovial CD4⁺ CD45RO⁺ T cells</title><p>Initial experiments delineated cytokine production by freshly isolated CD4⁺ CD45RO⁺ T cells obtained from matching samples of RAPB and RASF. A representative sample is shown in Fig. <xref ref-type="fig" rid="F1">1</xref> (RA patient1). The majority of peripheral blood T cells from RA patients secreted IL-2 alone, whereas fewer cells secreted IFN-&#x003b3; alone or in combination with IL-2. By contrast, the CD4⁺ T cells isolated from the synovial fluid demonstrated a significant increase in the percentage of cells that produced IFN-&#x003b3; alone or in combination with IL-2, whereas a decrease was observed in the percentage of cells that secreted IL-2 alone. Few cells from peripheral blood or synovial fluid produced IL-4.</p><p>Matching samples obtained from the peripheral blood and synovial tissue of RA patients were also assessed for cytokine expression (Fig. <xref ref-type="fig" rid="F1">1</xref>; RA patient 7). As was observed in the synovial fluid, memory T cells isolated from the synovial tissue contained a decreased percentage of IL-2-producing cells and an increased percentage of IFN-&#x003b3;-producing cells. Thus, T cells isolated from the rheumatoid synovial tissue and fluid demonstrated a polarized Th1 cytokine profile.</p><p>As a control, the cytokine profiles of Tm from normal adult volunteers were examined (Fig. <xref ref-type="fig" rid="F1">1</xref>; normal1). As observed for blood from RA patients, the majority of cells produced IL-2 alone, whereas fewer cells produced the combination of IL-2 and IFN-&#x003b3;. Relatively infrequent cells expressed a highly polarized phenotype by producing either IFN-&#x003b3; alone or IL-4 alone.</p><p>A compilation of the frequency of cells that produced each cytokine from a number of donors is shown in Fig. <xref ref-type="fig" rid="F2">2</xref>. After a brief <italic>in vitro</italic> stimulation, the majority of Tm obtained from normal donors (<italic>n</italic> = 11) produced IL-2 alone (75 &#x000b1; 4%; mean &#x000b1; SEM). Fewer cells produced the combination of IL-2 and IFN-&#x003b3; (14 &#x000b1; 3%). Relatively infrequent cells expressed a highly polarized phenotype by producing either IFN-&#x003b3; (6 &#x000b1; 1%) or IL-4 alone (4 &#x000b1; 1%). Less than 1% of the cells produced the combination of IFN-&#x003b3; and IL-4.</p><p>Cytokine secretion profiles of freshly isolated Tm obtained from RAPB (<italic>n</italic> = 9) also demonstrated that most secreted IL-2 alone (75 &#x000b1; 3%), whereas fewer cells produced IL-2 in combination with IFN-&#x003b3; (8 &#x000b1; 1%). There was no significant difference between the normal T cells and RA T cells for these two groups. Of note, there was a significant (<italic>P</italic> &#x0003c; 0.05) increase in the percentage of cells that secreted IFN-&#x003b3; alone compared with normal cells (11 &#x000b1; 2% versus 6 &#x000b1; 1%), whereas there was no difference between normal blood and RAPB in the number of cells that secreted IL-4 alone (4 &#x000b1; 1%) or in the frequency of those rare cells that produced the combination of IL-4 and IFN-&#x003b3; (1 &#x000b1; 0%).</p><p>There was a significant decrease in the number of cells isolated from RASF (<italic>n</italic> = 8) that produced IL-2 alone (31 &#x000b1; 9%) when compared with normal blood (<italic>P</italic> &#x0003c; 0.001) or RAPB (<italic>P</italic> &#x0003c; 0.001) T cells. Moreover, an increase was observed in the number of cells that secreted the combination of IL-2 and IFN-&#x003b3; (27 &#x000b1; 3%), which was significantly greater than the percentage of the same subset in normal blood (<italic>P</italic> &#x0003c; 0.01) or RAPB (<italic>P</italic> &#x0003c; 0.01). The percentage of RASF cells that produced IFN-&#x003b3; alone (41 &#x000b1; 8%) was also significantly increased over the percentage of such cells in normal blood (<italic>P</italic> &#x0003c; 0.002) or RAPB (<italic>P</italic> &#x0003c; 0.002) T cells. Significantly fewer cells produced IL-4 alone (1 &#x000b1; 0%) as compared with normal blood (<italic>P</italic> &#x0003c; 0.05) or RAPB (<italic>P</italic> &#x0003c; 0.05) T cells, whereas no significant difference was observed in the number of cells that produced the combination of IL-4 and IFN-&#x003b3; (2 &#x000b1; 0%).</p><p>RAST Tm (<italic>n</italic> = 4) also contained a decreased percentage of IL-2-secreting cells (37 &#x000b1; 9%), which did not differ significantly from that in the RASF, but did differ significantly from that in normal blood (<italic>P</italic> &#x0003c; 0.001) and RAPB (<italic>P</italic> &#x0003c; 0.001). The percentage of cells that produced both IFN-&#x003b3; and IL-2 (25 &#x000b1; 1%) did not differ significantly from that in RASF, but was significantly greater than the frequencies observed in normal blood (<italic>P</italic> &#x0003c; 0.05) or RAPB (<italic>P</italic> &#x0003c; 0.05). In RAST, 34 &#x000b1; 9% of cytokine effectors secreted IFN-&#x003b3; alone. This was not different from the percentage found in RASF, but was significantly greater than the percentage of IFN-&#x003b3;-producing cells in normal blood (<italic>P</italic> &#x0003c; 0.005) or RAPB (<italic>P</italic> &#x0003c; 0.005). There was no significant difference in the number of synovial tissue cells that produced IL-4 alone (2 &#x000b1; 1%) or IL-4 and IFN-&#x003b3; (2 &#x000b1; 0%), as compared with RASF, normal blood or RAPB.</p><p>Thus, T cells isolated from RAST and RASF demonstrated a decrease in IL-2-producing cells, with a concomitant increase in IFN-&#x003b3;-producing cells when compared with normal blood and RAPB Tm.</p></sec><sec><title>Rheumatoid CD4⁺ T cells demonstrate defective responsiveness to differentiation signals <italic>in vitro</italic></title><p>Peripheral blood and synovial Tm were than examined in an <italic>in vitro</italic> culture system in order to determine the relative capacity of differentiation signals to modulate cytokine production. Our previous studies with normal T cells [<xref ref-type="bibr" rid="B17">17</xref>] demonstrated that this system effectively generates polarized IFN-&#x003b3;-producing or IL-4-producing effector cells. Fig. <xref ref-type="fig" rid="F3">3</xref> shows the composite cytokine profiles of CD4⁺, CD45RO⁺ T cells obtained from cultures primed with anti-CD28 and cytokines in the absence or presence of anti-CD3 mAb. As seen in Fig. <xref ref-type="fig" rid="F3">3</xref>, stimulation of normal T cells after priming with anti-CD28 alone resulted in a significant decrease in the percentage of IL-2 producers (35% versus 75%) compared with cells immediately stimulated after isolation, and a significant increase in the percentages of IFN-&#x003b3; (31% versus 6%) and IL-4 producers (17% versus 4%). After priming with a low dose of anti-CD3 (Fig. <xref ref-type="fig" rid="F3">3</xref>; middle panel), the percentage of IL-2 producers decreased further and the percentage of IFN-&#x003b3;-secreting cells increased. As previously shown, the generation of IL-4-secreting cells was optimally induced by priming with anti-CD28 alone and was inhibited by costimulation with anti-CD3 mAb [<xref ref-type="bibr" rid="B17">17</xref>].</p><p>T cells derived from RAPB demonstrated a distinct pattern of responsiveness. Whereas the percentage of IL-2-secreting cells decreased after <italic>in vitro</italic> priming with anti-CD28 as compared with freshly isolated cells (52% versus 75%), a greater decrease was observed after priming with low dose anti-CD3 (29%), and the cells were resistant to further modulation by higher concentrations of anti-CD3. Of note, the loss of IL-2 production by <italic>in vitro</italic> primed RA T cells was less marked than observed with normal T cells. An increased percentage of IFN-&#x003b3; producers was observed after priming, and the peak response was achieved in the presence of low concentrations of anti-CD3 (45% versus 11%). Again, the induction of IFN-&#x003b3; production in RAPB memory T cells was less than noted in normal blood. IL-4-producing cells were generated from RA memory T cells by stimulation with anti-CD28, and these were inhibited by the presence of anti-CD3 mAb. The percentage of IL-4 producers (12%) generated from RA memory T cells was less than was generated from normal T cells (17%).</p><p>T cells from RASF primed with anti-CD28 exhibited a decrease in the percentage of IL-2-producing cells (15% versus 31%) and an increase in the percentage of IFN-&#x003b3; secreting cells (67% versus 41%) as compared with freshly isolated cells. Synovial fluid T cells were deficient in the ability to generate IL-4-producing cells (1%), which was in contrast to the blood. Whereas RAPB T cells were refractory to priming by high concentrations of anti-CD3 mAb, the synovial fluid T cells responded to anti-CD3 in a dose-dependent manner. T cells derived from synovial tissue were also relatively refractory to <italic>in vitro</italic> priming as compared with normal blood T cells. Synovial tissue T cells contained an increased percentage of IL-2 producers (48% versus 37%) and a small decrease in IFN-&#x003b3; producers (23% versus 34%) after <italic>in vitro</italic> priming with anti-CD28. Importantly, IL-4-secreting cells (8% versus 2%) could be generated from synovial tissue memory T cells after <italic>in vitro</italic> differentiation, although in lesser numbers than from RAPB and normal blood. Priming in the presence of anti-CD3 reduced the percentages of IL-2-producing and IL-4-producing cells, while increasing the percentage of IFN-&#x003b3;-secreting cells.</p></sec><sec><title>Effect of anti-IFN-&#x003b3; antibody on the generation of IL-4-secreting effector cells from RA synovial fluid T cells</title><p>In order to determine whether cytokine production by rheumatoid T cells was resistant to modulation in the <italic>in vitro</italic> priming cultures because of the secretion of IFN-&#x003b3;, the <italic>in vitro</italic> priming was also carried out in the presence of anti-IFN-&#x003b3; antibody. As shown in Fig. <xref ref-type="fig" rid="F4">4</xref>, the addition of increasing concentrations of a neutralizing antibody to IFN-&#x003b3; had little effect on the generation of cytokine-producing cells from either normal synovial fluid or RASF CD4⁺ T cells. Then we examined whether the combined effects of anti-IFN-&#x003b3; antibody and IL-4 could influence the effector phenotype of cells isolated from the synovium. As shown in Table <xref ref-type="table" rid="T1">1</xref>, Tm were isolated from matching RAPB and RASF, and were cultured under conditions that optimized the generation of IL-4-secreting effector cells [<xref ref-type="bibr" rid="B17">17</xref>]. RAPB Tm generated a marked increase in the percentage of IL-4-producing cells when primed with the combination of IL-4 and anti-IFN-&#x003b3; antibody. In contrast, T cells from the synovial fluid of the same patient generated virtually no IL-4-producing cells and few IL-2-producing cells, despite supplemental IL-4 and anti-IFN-&#x003b3; antibody.</p><p>As a control, cell growth during the priming cultures was monitored in the presence or absence of antibody to IFN-&#x003b3; and IL-4. As shown in Fig. <xref ref-type="fig" rid="F5">5</xref>, synovial T-cell growth was not inhibited by the addition of anti-IFN-&#x003b3; antibody or IL-4. These data suggest that RAPB Tm have the capacity to generate IL-4-producing or IFN-&#x003b3;-producing effector cells, whereas memory T cells from RASF are inhibited in their capacity to become IL-4-producing effector cells. Further studies indicated that anti-CD28 mAb and cytokines enhanced synovial fluid Tm growth approximately fourfold above the initial input cell number. Synovial fluid T cells cultured in medium alone underwent a fourfold reduction in cell number, whereas cells cultured with cytokines alone increased in number by an average of twofold (data not shown). These data further support the conclusion that synovial fluid T cells were responsive to signals delivered by the combination of anti-CD28 and cytokines.</p></sec></sec><sec><title>Discussion</title><p>RAST and RASF were deficient in IL-4-secreting cells as compared with RAPB after a brief <italic>in vitro</italic> stimulation. Thus, mature IL-4-secreting effector cells were found to be decreased in the RA synovium but not in the blood.</p><p>There are at least three major mechanisms by which IL-4 production might be inhibited in the RA synovium. One mechanism could be through the selective recruitment of IFN-&#x003b3;-producing effector cells. For example, it has been suggested that the T cells found in the RA synovium have been selectively recruited on the basis of expression of chemokine receptors [<xref ref-type="bibr" rid="B20">20</xref>]. Thus, T cells found in the rheumatoid synovium express CCR5, whereas T cells found at sites of parasitic infection express CCR4 and CCR3 [<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. However, the role of chemokine receptor expression in the recruitment of T cells to these sites is unknown. This possibility cannot be the full explanation of the current observations in light of the finding that synovial tissue T cells had the capacity to differentiate into IL-4-secreting effector cells in <italic>in vitro</italic> cultures. Thus, some synovial cells remained responsive to IL-4- and anti-CD28-mediated signaling in the priming cultures. A second mechanism for suppression of IL-4 production in the rheumatoid synovium could be through regulatory signals received in the local microenvironment. This remains a likely possibility because studies in murine models have shown that only the most highly differentiated T cells obtain a polarized cytokine secretion profile that cannot be altered by external stimuli [<xref ref-type="bibr" rid="B22">22</xref>]. A third possibility is that precursor cells of IL-4 producers rapidly migrate out of the synovium, perhaps being unresponsive to retention signals.</p><p>Although the exact mechanism that is operative in the RA synovium remains to be elucidated, our ongoing studies are focused on determining the phenotype and response defect of these highly polarized Th cells.</p><p>Tm isolated from rheumatoid blood or synovium contained increased numbers of cells with the capacity to produce IFN-&#x003b3; after a brief <italic>in vitro</italic> stimulation, compared with the same subset isolated from the periphery of normal donors. A strong mitogenic stimulation was employed in order to obtain the maximum cytokine-secreting potential of the effector population. We have previously demonstrated that staining for intracellular cytokines correlated with cytokine secretion [<xref ref-type="bibr" rid="B17">17</xref>]. However, intracellular cytokine analysis proved to be a more sensitive method for detecting cytokines, such as IL-2 and IL-4, and also made it possible to determine the percentage of cells producing each. It should be noted that longer incubations did not result in a substantial increase in the number of cytokine-secreting cells detected in this assay. These studies corroborate previous studies that demonstrated that synovial T cells are enriched in IFN-&#x003b3;-producing cells on initial isolation [<xref ref-type="bibr" rid="B11">11</xref>]. The results suggest that these memory cells appear to have undergone biased differentiation into Th1 cells, perhaps as a result of signals received in the synovial microenvironment. The small increase in the number of IFN-&#x003b3;-producing cells found in rheumatoid blood may well reflect the recirculation of memory T cells previously activated in the blood [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B23">23</xref>].</p><p>Tm were used for these studies because it has been shown [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B9">9</xref>] that the majority of CD4⁺ cells in the RA synovium are of the memory subset. Therefore, a potential concern was that the <italic>in vitro</italic> priming protocol might select for cells that are already committed to IFN-&#x003b3; or IL-4 production, rather than generate new effector cells. In humans, however, highly polarized effector T cells arise late in the differentiation pathway [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Whereas early memory T cells, identified by the CD45RO⁺, CD27⁺ phenotype, have the potential to secrete IL-2 and IFN-&#x003b3;, these cells cannot produce IL-4 and few produce IFN-&#x003b3; in the absence of IL-2 [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. Only mature memory cells, identified by the presence of CD45RO and the loss of CD27 cell-surface expression, have the capacity to secrete IFN-&#x003b3; or IL-4 alone. We have previously shown that the <italic>in vitro</italic> priming protocol used in the present study induces the differentiation of early uncommitted CD4⁺, CD27⁺ memory T cells into IL-4-producing effector cells [<xref ref-type="bibr" rid="B17">17</xref>]. Thus, it was hypothesized that the uncommitted early CD4⁺, CD27⁺ memory cells that were isolated from the RA synovium or blood maintained the capacity to differentiate into IL-4-producing effector cells. The majority of T cells that are found in the RA synovium belong to the early memory subset [<xref ref-type="bibr" rid="B7">7</xref>]. Therefore, the defective generation of IL-4 producers from RA synovial fluid under conditions that induced IL-4-producing effector cells from blood memory T cells suggested that the rheumatoid microenviroment played an important role in selecting or modifying the precursor effector memory population.</p><p>CD28 is expressed by most human T cells and is thought to be critical for T-cell differentiation. CD28 is also important during T-cell receptor-mediated activation, IL-2 production, and the prevention of anergy [<xref ref-type="bibr" rid="B24">24</xref>]. Recent studies [<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>] have indicated that there were increased numbers of CD4⁺, CD28^&#x02013; T cells in the periphery and synovium of RA patients. Therefore, the lack of expression of CD28 on RA CD4⁺ T cells represented a potential explanation for the abnormal memory T-cell differentiation observed in RA blood and tissue and the deficiency in IL-4-producing effector cells from the synovial fluid after <italic>in vitro</italic> priming. However, CD4⁺, CD28^&#x02013; cells are a minor subset of T cells, representing less than 3% of peripheral CD4⁺ T cells from the most chronically active RA patients versus less than 1% from healthy control individuals [<xref ref-type="bibr" rid="B26">26</xref>], and are therefore unlikely to account for the broad defect in memory T-cell differentiation found here. We have observed that CD4⁺, CD28^&#x02013; cells in the periphery represent a unique subset that is restricted to the CD4⁺, CD27^&#x02013; population (Davis L, unpublished observation). As noted above, although there were increased numbers of well-differentiated CD4⁺, CD27^&#x02013; cells in the RA synovium, these cells represented the minority of Tm [<xref ref-type="bibr" rid="B7">7</xref>]. Therefore, the inability to modulate effector cell responses in RASF samples could not be explained by the absence of CD28. Moreover, RASF T cells were capable of cell growth in response to anti-CD28 mAbs and cytokines. This finding is in agreement with previous studies [<xref ref-type="bibr" rid="B28">28</xref>] demonstrating that the CD28 signaling pathway was intact in RA synovial CD4⁺ T cells.</p><p>The present studies also demonstrated that synovial fluid Tm lacked the capacity to generate IL-4-secreting effector cells in numbers similar to those found in the periphery. Our previous studies [<xref ref-type="bibr" rid="B17">17</xref>] demonstrated that IL-4 was required during the <italic>in vitro</italic> priming cultures to generate IL-4-producing effector cells from non-IL-4-producing early memory precursor cells. Therefore, one concern in the current studies was that RA T cells might require exogenous IL-4 to generate IL-4-producing effector cells.</p><p>An additional concern was that IFN-&#x003b3;-producing effector cells might inhibit potential IL-4 precursors from differentiating or expanding in these cultures, although previous studies [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>] demonstrated that IFN-&#x003b3; indirectly affected the generation of IL-4-secreting cells through the activities of antigen-presenting cells. It should be noted that antigen-presenting cells were depleted from the memory T-cell population before priming in the present studies, and therefore this possibility was only a minor concern.</p><p>Both the addition of IL-4 to the priming cultures and the presence of anti-IFN-&#x003b3; antibody ensured that potential IL-4-secreting precursor populations had optimal conditions for differentiation. Blocking IFN-&#x003b3; activity during <italic>in vitro</italic> priming with a neutralizing anti-IFN-&#x003b3; antibody had little effect on the number of IFN-&#x003b3;-secreting effector cells detected on subsequent restimulation. Moreover, the availability of IFN-&#x003b3; had little impact on the generation of either subset. Thus, the inability to generate IL-4-producing effector cells from RASF precursor cells was not simply explained by the presence of IFN-&#x003b3;-producing effector cells or the lack of a subset of fully differentiated IL-4-producing T cells in the initial memory cell population. Recent studies [<xref ref-type="bibr" rid="B32">32</xref>] have suggested that there is an intrinsic defect in the development of IL-4-producing effector cells in RA patients. Therefore, it is possible that the apparent disordered differentiation of effector cells becomes more marked at inflammatory sites such as in the rheumatoid synovium [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B32">32</xref>].</p><p>It is interesting to note that both IL-2 production and IL-4 production appear to be downregulated in RASF T cells. Previous studies [<xref ref-type="bibr" rid="B9">9</xref>] have shown that mature memory (CD45RO⁺, CD27^&#x02013;) CD4⁺ T cells isolated from the blood have the same capacity as early memory (CD45RO⁺, CD27⁺) CD4⁺ T cells to produce IL-2. In the RA synovium, IL-2 appears to be downregulated [<xref ref-type="bibr" rid="B9">9</xref>]. It should be noted that IL-2 production was not decreased in T cells obtained from osteoarthritic joints [<xref ref-type="bibr" rid="B9">9</xref>]. In addition, Tm obtained from other tissues, such as inflamed tonsil, expressed levels of IL-2 that were similar to those in blood (Davis L, unpublished observation). Therefore, migration into tissue <italic>per se</italic> does not induce downregulation of IL-2 production. Recent studies [<xref ref-type="bibr" rid="B33">33</xref>] have attributed the lack of IL-2 production by synovial CD4⁺ T cells to a defect in the T-cell receptor-mediated signal transduction cascade. However, in the present studies we made use of stimuli that bypassed the early steps in the T-cell receptor-mediated signaling cascade and found that there was defective IL-2 production in cells stimulated immediately on isolation from the RA synovium, whereas these same cells were completely competent for IFN-&#x003b3; production. Thus, it is interesting to speculate that IL-2 and IL-4 are downregulated in the rheumatoid synovium by some active repressor mechanism induced by the synovial microenvironment. Although recent studies have indicated that redox balance alterations were critical in determining whether cells can produce IL-2 [<xref ref-type="bibr" rid="B33">33</xref>,<xref ref-type="bibr" rid="B34">34</xref>], it is not known whether this influences IL-4 production, or has it been determined whether readjustment of redox balance in RA synovial T cells could correct the deficiency in IL-4.</p><p>Few studies have used similar techniques to assess cytokine production in blood, synovial tissue, and synovial fluid obtained from RA patients. Immunohistology is routinely carried out on synovial tissue sections, whereas enzyme-linked immunosorbent assays are employed to determine cytokine profiles of serum and synovial fluid [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B31">31</xref>]. Therefore, it has remained difficult to determine the potential impact of these compartments on T-cell function in rheumatoid inflammation. In the current studies, RA T cells from all three compartments produced increased amounts of IFN-&#x003b3; on immediate stimulation. However, the cytokine profiles diverged when assessed after <italic>in vitro</italic> priming. Whereas, synovial tissue Tm were deficient in IL-4 producers compared with matching blood, these cells retained the ability to generate IL-4-producing effector cells. In this regard, recent studies [<xref ref-type="bibr" rid="B31">31</xref>] have suggested that synovial tissue cells were selectively responsive to IL-12 produced in the synovium, because IL-12 induced increased production of IFN-&#x003b3;, but not IL-2 and IL-4. Those studies support the hypothesis that the synovial microenvironment may play a role in skewing T cells toward a Th1 phenotype. The finding that IL-4-producing effector cells could be generated from synovial tissue T cells once removed from that environment suggests that all of the cells were not yet terminally polarized to Th1-like effector cells. Importantly, synovial fluid T cells appeared to have been more affected by the synovial microenvironment than synovial tissue T cells because, regardless of the <italic>in vitro</italic> priming conditions, these cells yielded increased percentages of IFN-&#x003b3; producers and were deficient in IL-4 producers.</p><p>The present data suggest that synovial fluid T cells are those that have passed through the synovium, whereas synovial tissue T cells are a mixed population of recently migrated cells and those that have been retained in the synovium. Whether the inability of synovial fluid memory T cells to generate IL-4 producers is the result of activation, differentiation, or prolonged exposure to the synovial microenvironment, the data clearly indicate that a majority of synovial fluid memory T cells appear to be polarized IFN-&#x003b3; effector cells.</p><p>In summary, the present studies show that RA blood and synovial T cells contain increased numbers of polarized IFN-&#x003b3; effector cells. RA synovial T cells also demonstrated a deficiency in the ability to generate IL-4-producing effector cells. The diminished ability to generate IL-4 effector cells from synovial T cells <italic>in vitro</italic> suggests that the rheumatoid microenvironment alters T-cell effector function and thereby perpetuates the chronic inflammatory disease state.</p></sec>

The development of clinical signs of rheumatoid synovial inflammation is associated with increased synthesis of the chemokine CXCL8 (interleukin-8)

<p>Paired synovial tissue samples were obtained from both clinically uninvolved (CU) and clinically involved (CI) knee joints of eight rheumatoid arthritis (RA) patients. In addition, biopsies were taken from five control subjects. We observed the expression of the chemokines CXCL8, CXCL9, CXCL10, CCL2 and CCL4 in CI and CU joints of RA patients. In particular, CXCL8 protein levels were specifically increased in CI joints compared with CU joints, which was confirmed by immunohistochemistry and <italic>in situ</italic> hybridization.</p>

<contrib id="A1" contrib-type="author"><name><surname>Kraan</surname><given-names>Maarten C</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>m.c.kraan@amc.uva.nl</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Patel</surname><given-names>Dhavalkumar D</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Haringman</surname><given-names>Jasper J</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Smith</surname><given-names>Malcolm D</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Weedon</surname><given-names>Helen</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Ahern</surname><given-names>Michael J</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Breedveld</surname><given-names>Ferdinand C</given-names></name><xref ref-type="aff" rid="I4">4</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Tak</surname><given-names>Paul P</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib>

Arthritis Research

<sec><title>Introduction</title><p>Rheumatoid arthritis (RA) is a chronic inflammatory disease of unknown etiology affecting diarthrodial joints. Macrophages are major components in the inflammatory cascade [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>] and are also believed to be important mediators of joint destruction [<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>]. Large numbers of macrophages are present in synovial tissue (ST) [<xref ref-type="bibr" rid="B6">6</xref>] and their cell numbers are associated with scores for local disease activity in clinically involved (CI) joints of RA patients [<xref ref-type="bibr" rid="B7">7</xref>]. Interestingly, increased numbers of macrophages can also be observed in clinically uninvolved (CU) joints of RA patients, although their number is lower than in CI joints [<xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. The observation that macrophage numbers are increased in joints that are still clinically quiescent could be explained by a difference in functional activity between macrophages in CU joints and in CI joints [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B10">10</xref>].</p><p>Macrophage activity in the synovial compartment includes the production of chemotactic cytokines called chemokines [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>]. Chemokines can be generally divided into four groups: the CXC, C, CX₃C and CC chemokine receptor families [<xref ref-type="bibr" rid="B13">13</xref>]. The CXC subfamily includes CXCL-8 [interleukin (IL)-8], CXCL-9 [monokine induced by &#x003b3;-interferon (Mig)], and CXCL10 [interferon-&#x003b3; inducible protein-10 (IP-10)]. The best-studied chemokine of this family is CXCL8, which is produced constitutively by macrophages in the synovial compartment [<xref ref-type="bibr" rid="B14">14</xref>] and is capable of inducing synovial inflammation in an animal model [<xref ref-type="bibr" rid="B15">15</xref>]. CXC chemokines are mostly chemotactic factors for neutrophils [<xref ref-type="bibr" rid="B16">16</xref>], although CXCL10 and CXCL9 attract monocytes and T lymphocytes [<xref ref-type="bibr" rid="B17">17</xref>]. The CC chemokines include CCL2 [monocyte chemoattractant protein 1] [<xref ref-type="bibr" rid="B18">18</xref>] and CCL4 [macrophage-inhibiting protein-1&#x003b2; (MIP-1&#x003b2;)] [<xref ref-type="bibr" rid="B19">19</xref>]. The main function of both CCL2 and CCL4 seems to be the recruitment of macrophages [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>].</p><p>To investigate the role of chemokines in relation to clinical signs of synovitis, we determined chemokine expression in paired CU and CI RA knee joints.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patients</title><p>Eight patients with RA (ACR 1987 criteria [<xref ref-type="bibr" rid="B22">22</xref>]) and both a CI and a CU knee joint were investigated. The knee was considered uninvolved if the patient noticed no pain or limitation of range of motion and two independent observers could not detect signs of inflammation such as swelling, warmth, effusion, or pain on examination. The knee was considered involved if there were signs of inflammation, namely joint effusion, synovial swelling, and pain. Five subjects without inflammatory joint disease and meniscus pathology served as control subjects. All patients gave informed consent, and the Medical Ethics Committee of the Leiden University Medical Center approved the study protocol. The mean ages were similar for patients and controls. The mean disease duration of the RA patients was 73 months (range 3&#x02013;252 months). Most RA patients had active disease in other joints in addition to the knee joint from which the synovial biopsies were obtained. The mean Ritchie articular index [<xref ref-type="bibr" rid="B23">23</xref>] was 11 (range 3&#x02013;24) and the mean serum level of C-reactive protein (CRP) was 69 mg/l (range 4&#x02013;108 mg/l). In six of the eight RA patients the CU joint had shown clinical signs of arthritis at previous phases of the disease.</p></sec><sec><title>Arthroscopy</title><p>In all RA patients an arthroscopy procedure was performed in both knees with a small-bore 2.4 mm arthroscope (Storz, Tuttlingen, Germany) by a single skin portal in the suprapatellar pouch both for macroscopic examination of the synovium and for the biopsy procedure. At each arthroscopy, synovial biopsies were taken from the supra-patellar pouch, the synovium&#x02013;cartilage junction, the patellar gutters, and the tibia&#x02013;femur junction with a 2.0 mm grasping forceps (Storz). If there was macroscopic variation of synovitis, samples were taken from both macroscopically inflamed and macroscopically non-inflamed regions. The ST from all five control patients with meniscus pathology was obtained by arthroscopy with a 5.0 mm grasping forceps before the meniscectomy. On average, 20 ST samples from one knee joint were snap-frozen in methylbutane (&#x02013;80&#x000b0;C) and used for tissue enzyme-linked immunosorbent assay (ELISA). Biopsy samples were stored in liquid nitrogen. On average, five ST samples from one knee joint were fixed in formalin and subsequently embedded in paraffin and then used for immunohistochemistry and <italic>in situ</italic> hybridization.</p></sec><sec><title>Tissue ELISA</title><p>ST protein extracts were prepared by mincing tissues in 2 volumes of extraction buffer [1% Tween 20, 1 M NaCl, 0.1% NaN₃ in phosphate-buffered saline (PBS)], incubating the mixture on ice for 1 h and sonicating for 5 min in a water bath sonicator. After centrifugation at 10,000 <italic>g</italic> for 10 min, the supernatant was collected, diluted with 9 volumes of PBS and stored at &#x02013;80&#x000b0;C or immediately assayed for protein concentration by the bicinchoninic acid assay (Pierce). Chemokine protein levels were quantified by sandwich ELISA with matched antibody pairs. ELISA plates (Costar) were coated with 100&#x02013;400 ng per well of capture antibody [6217.11 for CXCL8 (R&#x00026;D Systems, Minneapolis, Minnesota, USA); B8-11 for CXCL9 (PharMingen); 4D5/A7/C5 for CXCL10 (PharMingen); polyclonal rabbit antibodies against monocyte chemoattractant protein, for CCL2 (Endogen, Woburn, Massachusetts, USA); and 24006.111 for CCL4 (R&#x00026;D Systems)] in PBS overnight at 4&#x000b0;C, blocked with block buffer [PBS containing 1% (w/v) BSA, 5% (w/v) sucrose and 0.05% NaN₃] for 2 h at room temperature, and washed with wash buffer (0.05% Tween 20 in PBS). Samples (100 &#x003bc;l) and standards were incubated for 2 h at room temperature, washed, and incubated with 2 ng per well of polyclonal biotinylated antibodies against CXCL8 (R&#x00026;D Systems) or CCL4 (R&#x00026;D) or 20&#x02013;400 ng per well of monoclonal antibodies 6D4/D6/G2 against CXCL10 (PharMingen), T-MCAF2 against CCL2 (Endogen) or B8-6 against CXCL-9 (PharMingen) for 2 h at room temperature. After washing, 0.1 ng/ml horseradish peroxidase-conjugated streptavidin (Zymed, South San Francisco, California, USA) was added for 20 min at room temperature, washed and developed with tetramethyl benzidine (TMB) peroxidase substrate (Kirkegaard and Perry Laboratories, Gaithersburg, Maryland, USA) for 20 min. The peroxidase reaction was stopped by the addition of 0.5 volume of 1 M phosphoric acid and the plate was read at 450 nm in an ELISA plate reader (Anthos, Durham, North Carolina, USA).</p></sec><sec><title>Immunohistochemistry</title><p>Formalin-fixed, paraffin-embedded ST was cut into 5 &#x003bc;m sections and placed on slides coated with 2% 3-amino-propyltriethoxysilane (Sigma, St Louis, Missouri, USA). Immunohistochemical staining was performed with a monoclonal mouse antibody against CXCL8 (R&#x00026;D Systems), as described previously [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. In brief, we used alkaline phosphatase-conjugated antibodies (Dako, Glostrup, Denmark), napthol-AS-MX-phosphate, Fast Red Violet LB and levamisole (Sigma) for detection of the monoclonal antibody. All ST samples were processed in the same run. Negative control experiments were performed with irrelevant isotype control antibodies (X63 and FMC41) or normal donkey serum alone, or by leaving out the secondary antibody. A positive control (ST with known staining characteristics) was also used in each run.</p></sec><sec><title>Hybridization <italic>in situ</italic></title><p>The cDNA used to produce CXCL8 riboprobes was obtained from Dr Phil Auron (Boston, Massachusetts, USA) and was restricted to 800 base pairs within the coding sequence. The generation of riboprobes was performed as described previously [<xref ref-type="bibr" rid="B26">26</xref>]. The RNA copies were run off with a commercial riboprobe generation kit, which included digoxigenin-labeled dideoxyuridine triphosphate (Boehringer-Mannheim, Mannheim, Germany) as part of the nucleotide mix. Subsequently, paraffin-embedded tissue sections were processed as described previously [<xref ref-type="bibr" rid="B26">26</xref>]. In brief, after washing steps with 0.2 M HCl and diethyl pyrocarbonate-treated water the sections were digested in proteinase K. After washing steps with SSC (0.3 M NaCl and 0.03 M sodium citrate, pH 7.0), sections were incubated with prehybridization solution. The prehybridization solution was replaced by a previously tested amount of labeled RNA anti-sense probe (or labeled RNA sense probe as a control). Sections were covered with coverslips, sealed with nail polish, and incubated overnight at 55&#x000b0;C. For visualization of the probes, sections were incubated with digoxigenin labeling solution and labeled with nitro blue tetrazolium and 5-bromo-4-chloroindol-3-yl phosphate. All sections were counterstained with hematoxylin. Slides were mounted in Aquamount (BDH Laboratories, Poole, UK) with a coverslip.</p></sec><sec><title>Digital image analysis</title><p>One observer (MCK) performed the processing of all images. Five randomly selected high-power fields were chosen for the evaluation of each section, as described previously [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>]. Each high-power field contained both intimal lining layer and synovial sublining. The high-power field images were acquired on a Olympus microscope (Olympus, Japan, Tokyo), and were captured with a CCD (charge-coupled device) three-chip video camera (Sony, Tokyo, Japan), and were digitized with a PV100 multimedia 16-bit color video digitizer card with a standardized macro program to both simplify and standardize the acquisition process. The resultant color images were in a 640 &#x000d7; 480 pixel RGB (red, green, and blue) format with a 24-bit resolution, enabling the use of 16,581,375 colors. For each acquisition session the microscope, camera, and computer were calibrated with a standardized procedure. The images obtained were stored as bitmaps without compression with a Zip disk and portable driver (Iomega, Roy, Utah, USA).</p><p>All sections were examined with a Qwin (Qwin Pro V2.2; Leica, Cambridge, UK) computer-assisted color video image analysis systems. This system consisted of a PC with software. Two binary masks were applied with threshold values for the red, green, and blue channels. These thresholds were kept constant for all measurements with the same marker. A first binary mask identified the counter-stained areas as reference for the total region of tissue, and a second binary mask covered the positively stained areas. Both binary masks were processed individually to decrease the signal:noise ratio with erode, open, and dilate commands. Overlapping areas between the two binary masks were identified and they were not included in the analysis. Analysis was performed on the absolute area stained as indicated by the secondary binary mask. For the assessment of CXCL8 staining, the area was measured in pixels, the mean optical density was measured and the integrated optical density (IOD) was calculated by multiplication of the relative or absolute stained area by the mean optical density.</p></sec><sec><title>Statistical analysis</title><p>Data were analyzed with the following nonparametric tests: Wilcoxon signed ranks test for matched pairs comparing CI knee joints with CU knee joints, and Mann&#x02013;Whitney test comparing CU knee joints with control knee joints.</p></sec></sec><sec><title>Results</title><sec><title>Tissue ELISA</title><p>Differences in chemokine expression between control, CI, and CU joints were screened by determining protein expression in synovial extracts. The CXC chemokines CXCL8, CXCL9, and CXCL10, and the CC chemokines CCL2 and CCL4 could be detected in rheumatoid ST. In three of the eight patients, CXCL8 expression was detected in the CU joint (0.1 &#x000b1; 0.06 ng of chemokine/mg of tissue extract; mean &#x000b1; SEM) and it was increased in the CI joint (0.3 &#x000b1; 0.13 ng/mg), but CXCL8 was not detectable in the two control patients that were tested (Fig. <xref ref-type="fig" rid="F1">1</xref>). CCL2 expression was detected in five of six CU joints (0.9 &#x000b1; 0.5 ng/mg), two of four patients showed increased expression in the CI joint (1.0 &#x000b1; 0.2 ng/mg), and one of two controls had detectable levels (0.1 &#x000b1; 0.0 ng/mg) (Fig. <xref ref-type="fig" rid="F1">1</xref>). CXCL9 expression was detected in four of the six CU joints (9.1 &#x000b1; 8.3 ng/mg), two of four patients showed increased expression in CI joints (8.0 &#x000b1; 2.7 ng/mg), and CXCL9 was not detectable in two controls (Fig. <xref ref-type="fig" rid="F1">1</xref>). CXCL10 expression was detected in two of the six CU joints (0.2 &#x000b1; 0.2 ng/mg), whereas two of four patients showed increased expression in CI joints (0.4 &#x000b1; 0.2 ng/mg), and CXCL10 was not detectable in two controls (Fig. <xref ref-type="fig" rid="F1">1</xref>). CCL4 expression was detected in three of six CU joints (0.1 &#x000b1; 0.1 ng/mg); one of four patients showed increased expression in the CI joint (0.2 &#x000b1; 0.1 ng/mg), and CCL4 was not detectable in two controls (Fig. <xref ref-type="fig" rid="F1">1</xref>). Differences between CI knees, CU knees and control patients were not statistically significant, possibly owing to the small numbers of patients.</p></sec><sec><title>Immunohistochemistry</title><p>To confirm and extend the observation that CXCL8 expression was higher in CI joints than in CU joints, immunohistologic analysis was performed. In one of the patients, the ST from the CI joint was not assessable. Expression of CXCL8 was observed in both the intimal lining layer and the synovial sublining (Fig. <xref ref-type="fig" rid="F2">2</xref>). There was markedly increased CXCL8 protein expression in six of the seven CI joints (IOD 65,961 &#x000b1; 5364) compared with CU joints (IOD 28,552 &#x000b1; 5749) (<italic>P</italic> &#x0003c; 0.05), whereas the eight CU joints showed similar results compared with the five control subjects (IOD 34,599 &#x000b1; 5144) (Figs <xref ref-type="fig" rid="F3">3</xref> and <xref ref-type="fig" rid="F4">4</xref>).</p></sec><sec><title><italic>In situ</italic> hybridization</title><p>The increased expression of CXCL8 in CI joints compared with CU joints was confirmed by <italic>in situ</italic> hybridization to quantify CXCL8 mRNA expression. In one of the patients the ST from the CI joint was not assessable. CXCL8 mRNA expression in all seven CI joints (IOD 17,109 &#x000b1; 6944) was specifically increased compared with the eight CU joints (IOD 2774 &#x000b1; 1415) (<italic>P</italic> &#x0003c; 0.05) and the five control subjects (IOD 180 &#x000b1; 66) (Figs <xref ref-type="fig" rid="F3">3</xref> and <xref ref-type="fig" rid="F4">4</xref>). The values for CU and normal controls were essentially the same, as shown in Fig. <xref ref-type="fig" rid="F3">3</xref>.</p></sec></sec><sec><title>Discussion</title><p>The results presented here show protein expression of the CXC chemokines CXCL8, CXCL9, and CXCL10, and the CC chemokines CCL2 and CCL4 in rheumatoid ST by tissue ELISA. CXCL8 protein levels were increased in CI joints compared with CU joints. Immunohistochemistry and <italic>in situ</italic> hybridization confirmed the results for CXCL8.</p><p>Macrophages are major components of the inflammatory cascade involved in synovitis [<xref ref-type="bibr" rid="B1">1</xref>], and increasing evidence indicates that there is a distinct role for macrophage-derived cytokines in this process. Tumor necrosis factor-&#x003b1; (TNF-&#x003b1;) and IL-1&#x003b2; are important mediators of synovial inflammation and joint destruction [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref>]. In addition, the release of chemokines from inflammatory cells has a pivotal role in the development of inflammation [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B33">33</xref>]. The coordinated production of chemokines and proinflammatory cytokines is also important for the orchestration of the inflammatory response observed in patients with active RA [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B34">34</xref>]. In RA, chemokines are involved in the infiltration of the synovium by leukocytes [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B35">35</xref>,<xref ref-type="bibr" rid="B36">36</xref>]. Chemokines such as CXCL8, CCL2, and CCL4 [<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B37">37</xref>] are readily detectable in rheumatoid synovial fluid, and their expression is correlated with disease activity [<xref ref-type="bibr" rid="B38">38</xref>,<xref ref-type="bibr" rid="B39">39</xref>]. CXCL8 expression in rheumatoid ST is also associated with disease activity [<xref ref-type="bibr" rid="B39">39</xref>,<xref ref-type="bibr" rid="B40">40</xref>]. In addition, the chemokines CCL2 [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B35">35</xref>,<xref ref-type="bibr" rid="B41">41</xref>] and CCL4 [<xref ref-type="bibr" rid="B42">42</xref>] have been detected in ST in RA. CXCL9 and CXCL10 are also abundantly expressed in RA synovium [<xref ref-type="bibr" rid="B43">43</xref>]. The increased presence of the C&#x02013;C chemokines CCL2 and CCL4 in both CU and CI knee joints compared with controls is consistent with the increased macrophage numbers present in both CU and CI knee joints [<xref ref-type="bibr" rid="B8">8</xref>]. Both CCL2 and CCL4 are chemoattractants that act mainly on monocytes/macrophages and also regulate their expression of adhesion molecules [<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>]. We have previously shown that increased macrophage numbers can be found in CI as well as CU joints of RA patients [<xref ref-type="bibr" rid="B8">8</xref>].</p><p>The CXC chemokines CXCL10 and CXCL9 were undetectable in control patients and expressed equally in CU and CI knee joints. Both are associated mainly with the attraction of activated T cells [<xref ref-type="bibr" rid="B17">17</xref>], which are thought to have a role in the pathogenesis of RA [<xref ref-type="bibr" rid="B47">47</xref>,<xref ref-type="bibr" rid="B48">48</xref>]. In line with our previous studies, the results presented here show no direct relationship between the number of T cells in the synovium and clinical signs of arthritis [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>].</p><p>The observed difference in CXCL8 expression between CU and CI knee joints described here is a close reflection of the observations in animal models of crystal-induced arthritis where CXCL8 was the single factor determining the development of clinical signs and symptoms of arthritis [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B49">49</xref>]. Furthermore, neutralization of CXCL8-attenuated crystal-induced arthritis in an animal model [<xref ref-type="bibr" rid="B49">49</xref>] and RA patients treated with a high dose of methylprednisolone displayed a significant decrease in CXCL8 expression in ST biopsies together with an excellent clinical response [<xref ref-type="bibr" rid="B50">50</xref>]. This phenomenon could be explained by the attraction of neutrophils towards the synovial compartment under the influence of CXCL8 [<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref>,<xref ref-type="bibr" rid="B53">53</xref>,<xref ref-type="bibr" rid="B54">54</xref>]. Once neutrophils have arrived at the site of inflammation they are activated and contain a variety of proteinases and other enzymes, such as collagenase, elastase, gelatinase, myeloperoxidase, prostaglandins, and leukotrienes [<xref ref-type="bibr" rid="B55">55</xref>]. They also release proteins such as fibronectin, and cytokines including IL-1&#x003b2; [<xref ref-type="bibr" rid="B56">56</xref>], TNF-&#x003b1; [<xref ref-type="bibr" rid="B57">57</xref>], and CXCL8 [<xref ref-type="bibr" rid="B58">58</xref>]. Thus, the increased ingress of neutrophils into the synovial compartment and their prolonged lifespan [<xref ref-type="bibr" rid="B59">59</xref>] in response to CXCL8 might be important in the development of clinical symptoms, including swelling and pain. Therefore, targeting CXCL8 could be an effective anti-rheumatic treatment. This notion is supported by the observation that CXCL8 mRNA expression is decreased after treatment of antigen-induced arthritis [<xref ref-type="bibr" rid="B41">41</xref>]. Specific inhibition of CXCL8 by therapy with monoclonal antibody has been shown to be effective in an acute model of arthritis [<xref ref-type="bibr" rid="B60">60</xref>]. In RA patients, decreased CXCL8 production has been observed after treatment with methotrexate [<xref ref-type="bibr" rid="B61">61</xref>]. Furthermore, treatment with corticosteroids results in decreased expression of CXCL8 in ST and CXCL8 levels in synovial fluid, decreased migration of neutrophils into the joint, and diminished arthritis activity [<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B51">51</xref>]. Studies on the effects of specific targeting of CXCL8 in RA patients are not yet available.</p><p>In conclusion, this study demonstrates the expression of CCL2, CXCL10, CCL4, CXCL9, and CXCL8 in ST of RA patients. Increased synthesis and expression of CXCL8 is associated with clinical signs and symptoms of arthritis. Future research should be directed toward elucidating which factors determine the specific upregulation of CXCL8 synthesis by macrophages in clinically inflamed joints of RA patients.</p></sec>

Isolation and characterization of rheumatoid arthritis synovial fibroblasts from primary culture &#x02014; primary culture cells markedly differ from fourth-passage cells

<p>To reduce culture artifacts by conventional repeated passaging and long-term culture <italic>in vitro</italic>, the isolation of synovial fibroblasts (SFB) was attempted from rheumatoid arthritis (RA) synovial membranes by trypsin/collagenase digest, short-term <italic>in vitro</italic> adherence (7 days), and negative isolation using magnetobead-coupled anti-CD14 monoclonal antibodies. This method yielded highly enriched SFB (85% prolyl-4-hydroxylase⁺/74% Thy-1/CD90⁺ cells; &#x0003c;2% contaminating macrophages; &#x0003c;1% leukocytes/endothelial cells) that, in comparison with conventional fourth-passage RA-SFB, showed a markedly different phenotype and significantly lower proliferation rates upon stimulation with platelet-derived growth factor and IL-1&#x003b2;. This isolation method is simple and reliable, and may yield cells with features closer to the <italic>in vivo</italic> configuration of RA-SFB by avoiding extended <italic>in vitro</italic> culture.</p>

<contrib id="A1" contrib-type="author"><name><surname>Zimmermann</surname><given-names>Thomas</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Kunisch</surname><given-names>Elke</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Pfeiffer</surname><given-names>Robert</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Hirth</surname><given-names>Astrid</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Stahl</surname><given-names>Hans-Detlev</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Sack</surname><given-names>Ulrich</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A7" contrib-type="author"><name><surname>Laube</surname><given-names>Anke</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A8" contrib-type="author"><name><surname>Liesaus</surname><given-names>Eckehard</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A9" contrib-type="author"><name><surname>Roth</surname><given-names>Andreas</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A10" contrib-type="author"><name><surname>Palombo-Kinne</surname><given-names>Ernesta</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A11" contrib-type="author"><name><surname>Emmrich</surname><given-names>Frank</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A12" contrib-type="author"><name><surname>Kinne</surname><given-names>Raimund W</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>raimund.w.kinne@rz.uni-jena.de</email></contrib>

Arthritis Research

<sec><title>Synopsis</title><sec><title>Introduction:</title><p>Activated SFB in the invasive pannus tissue appear to play a major role in the crippling destruction of cartilage and bone in the joints of patients with RA [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>]. Several studies indicate that RA-SFB are morphologically altered [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B4">4</xref>], grow in an anchorage-independent fashion [<xref ref-type="bibr" rid="B5">5</xref>], and strongly express markers of activation, including major histocompatibility complex (MHC)-II and other surface molecules, proto-oncogenes, or matrix degrading enzymes (reviewed in [<xref ref-type="bibr" rid="B2">2</xref>]). Finally, activated RA-SFB elicit erosive arthritis upon intra-articular injection or engraftment into severe combined immunodeficiency mice [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>].</p></sec><sec><title>Aims:</title><p>To investigate <italic>ex vivo</italic> the pathophysiological properties of activated FB from the RA synovial membrane (SM), it is desirable to obtain freshly isolated SFB with phenotypic features as close as possible to the configuration observed <italic>in vivo</italic>. Because the generation of cell lines requires a number of passages <italic>in vitro</italic> to eliminate contaminating cells (especially macrophages), the risk of substantial <italic>in vitro</italic> alteration/growth selection exists.</p><p>We attempted to isolate FB from primary cultures of synovial cells, using a negative separation technique to eliminate macrophages, to minimize these artifacts.</p></sec><sec><title>Methods:</title><p>Synovial tissue was obtained from a total of 16 patients fulfilling the American Rheumatism Association criteria for RA [<xref ref-type="bibr" rid="B8">8</xref>] and 21 patients with osteoarthritis (OA) under approval of the local Ethics Committees. The tissue was placed in cell culture medium at ambient temperature and subjected to tissue digestion within 2 h. Synovectomy samples of RA and OA SM were finely minced, digested for 30 min at 37&#x000b0;C in phosphate-buffered saline (PBS) containing 0.1% trypsin (Sigma, Deisenhofen, Germany), and thereafter digested in 0.1% collagenase P (Boehringer Mannheim, Mannheim, Germany) in Dulbecco's modified Eagle medium (DMEM)/10% fetal calf serum (FCS) for 2 h at 37&#x000b0;C, 5% CO₂. The cell suspension was then filtered and the cells collected by centrifugation. Cells were kept in primary culture for 7 days (DMEM/10% FCS, 25 mM HEPES, 100 U/ml penicillin, 100 &#x003bc;g/ml streptomycin, and 2.5 &#x003bc;g/ml amphotericin B [Gibco BRL, Eggenstein, Germany], including removal of non-adherent cells on days 1, 3, 5, and 7) and subsequently used for SFB isolation. The samples were randomly tested to exclude <italic>Mycoplasma</italic> contamination. For negative isolation of SFB from primary culture, adherent synovial cells were detached by short-term trypsinization for 2 min (0.25% trypsin/0.2% EDTA; Gibco) and 10⁷/ml synovial cells were incubated with 4 &#x000d7; 10⁷/ml Dynabeads^&#x000ae; M-450 CD14 (clone RMO52; Dynal, Hamburg, Germany) in PBS/2% FCS for 1 h at 4&#x000b0;C. Nine milliliters of PBS/2% FCS were then added and the conjugated cells collected using the Dynal magnetic particle concentrator^&#x000ae;. The compositions of magnetobead-conjugated cells and unconjugated cells were analyzed by flow cytometry. Phenotype analysis of the expression of FB markers, as well as that of SFB features previously reported at a tissue level, was conducted by flow cytometry in RA-SFB, either negatively isolated from primary culture or obtained from conventional fourth passage. The findings were compared with those of normal skin-FB (lineage control) and OA-SFB (disease control). The proliferation of RA-SFB, either negatively isolated from primary culture or obtained from conventional fourth passage, was assayed by [³H]-thymidine incorporation.</p></sec><sec><title>Results:</title><p>The primary culture of RA synovial cells resulting from trypsin/collagenase digestion of the RASM contained large, spindle-shaped Thy-1⁺ SFB (CD90⁺; Fig. <xref ref-type="fig" rid="F1">1C</xref>) (monoclonal antibody [mAb] AS02; Dianova, Hamburg, Germany) and small, round CD14⁺ cells, most probably macrophages (Fig. <xref ref-type="fig" rid="F1">1D</xref>) (mAb Tyk4; Dako, Hamburg, Germany), as detected by immunohistochemical staining [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. Endothelial cells were absent, as confirmed by lack of staining for von Willebrand Factor (Fig. <xref ref-type="fig" rid="F1">1F</xref>) (mAb 4F9; Immunotech, Hamburg, Germany) and CD144 (Fig. <xref ref-type="fig" rid="F1">1G</xref>) (mAb Cadherin 5; Immunotech), which clearly identified human umbilical vein endothelial cells (HUVEC) (data not shown). The FB nature of the spindle-shaped cells was confirmed by intracellular staining for procollagen I and III (Fig. <xref ref-type="fig" rid="F1">1E</xref>,<xref ref-type="fig" rid="F1">H</xref>) (rabbit antibodies MP I and MP III; Prof. Schuppan, Berlin, Germany). An average of 62% of the cells stained with the anti-Thy-1 mAb AS02 (<italic>n</italic> = 4 RA patients; Table <xref ref-type="table" rid="T1">1a</xref>) in flow cytometry (FACS) [<xref ref-type="bibr" rid="B10">10</xref>]; the average percentage of CD14⁺ cells was approximately 15% (<italic>n</italic> = 4; Table <xref ref-type="table" rid="T1">1a</xref>). There were &#x0003c;1% T cells (CD3⁺) (mAb UCHT-1; ATCC, Manassas, VA, USA), B cells (CD19⁺/20⁺) (mAbs HD 37 and B-Ly 1; Dako), plasma cells (CD38⁺) (mAb AT 13/5; Dako), natural killer (NK) cells (CD56⁺) (mAb NKH/1; Immunotech), dendritic cells (CD83⁺) (mAb HB 15a; Immunotech), endothelial cells (CD144⁺), or PMN (CD15⁺) (mAb80H5; Immunotech), indicating that non-adherent cells had been efficiently removed during primary culture. The total yield of cells following 7 days of primary culture averaged (5.2 &#x000b1; 1.1) &#x000d7; 10⁷ cells (mean &#x000b1; SEM; <italic>n</italic> = 7).</p><p>Negative isolation of SFB from primary culture resulted in RA-SFB that were Thy-1⁺ (on average, approximately 74%; <italic>n</italic> = 9; Fig. <xref ref-type="fig" rid="F2">2A</xref> and Table <xref ref-type="table" rid="T1">1b</xref>) and, more importantly, prolyl-4-hydroxylase⁺ (on average, approximately 85%; <italic>n</italic> = 9; Table <xref ref-type="table" rid="T1">1b</xref>) (mAb3-2B12; Dianova), as shown by FACS analysis and confirmed by immunohistochemistry in chamber slides. There were very few contaminating non-specific esterase⁺ (three RA patients; Fig. <xref ref-type="fig" rid="F2">2C</xref>), CD14⁺, CD68⁺, and/or CD11b⁺ macrophages (&#x0003c;2%; Fig. <xref ref-type="fig" rid="F2">2B</xref>,<xref ref-type="fig" rid="F2">D</xref> and Table <xref ref-type="table" rid="T1">1b</xref>), as well as &#x0003c;1% T cells (CD3⁺), B cells (CD19⁺/20⁺), plasma cells (CD38⁺), NK cells (CD56⁺), dendritic cells (CD83⁺), PMN (CD15), or endothelial cells (CD144⁺; von Willebrand factor-positive). The positive fraction, in turn, contained virtually no SFB, indicating minimal cell loss, and thereby also excluding major subset selection. The average yield of RA-SFB negatively isolated from primary culture was (2.8 &#x000b1; 0.9) &#x000d7; 10⁷ cells (mean &#x000b1; SEM; <italic>n</italic> = 7). Phenotype analyses yielded two main results. The first, regarding expression of SFB features previously reported at a tissue level, was that, strikingly, approximately 66% of the cells expressed MHC-II molecules (normal skin-FB, 2%; OA-SFB, 17%). A low or moderate and variable percentage of RA-SFB, but also normal skin-FB and OA-SFB, expressed vascular cell adhesion molecule-1 (VCAM-1) (using two different anti-VCAM-1 mAbs) without statistically significant differences between the three different FB preparations. RA-SFB showed a higher (although non-significant) mean fluorescence intensity (MFI) for the cytoskeletal protein vimentin than normal skin-FB. Approximately 45% and 50% of the cells, respectively, expressed procollagen I and procollagen III (similar to normal skin-FB); however, the MFI for procollagen III was significantly higher in RA-SFB. Approximately 57% of RA-SFB expressed <italic>c</italic>-Fos. Neither the percentage nor the MFI were significantly different from those of normal skin-FB (approximately 87%) or OA-SFB (approximately 54%). In general, the differences between RA-SFB and normal skin-FB were not specific to RA, since they were also observed in the comparison between the disease control OA-SFB and normal skin-FB. The second outcome of phenotype analysis was that the percentages of RA-SFB positive for MHC-II as well as the MFI for VCAM-1 and <italic>c</italic>-Jun were significantly decreased in conventional fourth passage compared with isolated primary RA-SFB. In contrast, the percentages of cells positive for MHC-I, CD13, prolyl-4-hydroxylase, vimentin, procollagen I and III, <italic>c</italic>-Fos, and Jun-D were significantly increased in conventional fourth passage.</p><p>Proliferation assays showed that, at rest, the proliferation rates of isolated first-passage RA-SFB did not significantly differ from those of the corresponding cells in conventional fourth passage. Upon stimulation with platelet-derived growth factor (PDGF) (2.5 and 5 U/ml) and IL-1&#x003b2; (150 U/ml), however, the mean proliferation rates of conventional fourth-passage RA-SFB were significantly higher than those observed in first-passage cells (3.9-fold and 4.2-fold, respectively, in the case of 2.5 and 5 U/ml PDGF; 2.1-fold in the case of 150 U/ml IL-1&#x003b2;).</p></sec><sec><title>Discussion:</title><p>The advantages of negative isolation are as follows: first, there was minimal contamination with other cells, especially macrophages (&#x0003c;2%); second, negative isolation avoided direct contact of SFB with mAbs and/or magnetobeads, thereby limiting possible functional alterations of the cells; and third, the FB marker mAb AS02 (anti-Thy-1; used in the parallel attempt to positively isolate SFB) identified 91% of primary-culture/first-passage normal skin-FB but only 74% of isolated primary RA-SFB, probably due the variable density of Thy-1 molecules on the cell surface (Table <xref ref-type="table" rid="T1">1</xref>). Thus, Thy-1-independent isolation appears to avoid selection of Thy-1⁺ SFB subpopulations, a critical point since Thy-1⁺ and Thy-1^- FB diverge in several characteristics potentially relevant to RA.</p><p>Potential disadvantages of the negative isolation method include the limited number of cells obtained (approximately 2.8 &#x000d7; 10⁷), due to the lack of <italic>in vitro</italic> expansion, and the necessity for reliable access to freshly obtained synovial specimens. However, the applicability of this procedure not only to joint replacement samples, but also to arthroscopic synovectomy samples, augments the potential sources of fresh synovial tissue. The use of enzymes for tissue digestion may also represent at least a transient stress for the cells. On the other hand, this method may avoid selection of premitotic FB, a problem usually linked to tissue-outgrowth techniques [<xref ref-type="bibr" rid="B11">11</xref>].</p><p>The advantages of limited-passage isolation include the following. The present technique avoids a high number of passages <italic>in vitro</italic>; indeed, even a low number of passages increased the proliferation rates in response to cytokine stimulation and altered the cellular phenotype of RA-SFB (as exemplified by a strong decrease of the percentage of MHC-II⁺ cells, on one hand, and a striking increase of the percentage of procollagen I and III⁺, as well as <italic>c</italic>-Fos and Jun-D⁺ cells, on the other). Caution should thus be applied to the interpretation of data from passaged cells, especially in the case of proto-oncogenes. A second advantage is that this isolation procedure reduces the time of exposure (7 days) to surviving macrophages (unlike the weeks of exposure upon conventional isolation), thus possibly minimizing long-term <italic>in vitro</italic> changes due to monokine secretion or cell&#x02013;cell contact between SFB and synovial macrophages. Finally, a high percentage of RA-SFB from limited-passage isolation expressed MHC-II molecules (in analogy to findings in synovial tissue) without <italic>in vitro</italic> stimulation with cytokines, which in conventionally isolated cells requires stimulation with interferon-gamma. Therefore, the present <italic>in vitro</italic> system may be particularly useful in studies on the inter-relationship of SFB with T cells or macrophages, in which the influence of exogenous mediators may complicate the design and/or interpretation of the experiments.</p><p>Negative isolation of cells from synovial primary culture with magnetobead-conjugated anti-CD14 mAbs thus yielded highly enriched RA-SFB. This technique, by avoiding repeated passaging and numerous cumulative population doublings, may minimize the risk of <italic>in vitro</italic> alteration/growth selection; in RA-SFB, this includes increased proliferation rates and changes of several phenotypic features.</p></sec></sec><sec><title>Introduction</title><p>Activated SFB play a major role in the crippling destruction of cartilage and bone in the joints of patients with RA [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>], either autonomously [<xref ref-type="bibr" rid="B1">1</xref>] or in concert with T cells and macrophages [<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B19">19</xref>]. RA-SFB are a major part of the invasive pannus, a vascular and fibrous granulation tissue arising from the joint recessus and extending onto the surface of cartilage [<xref ref-type="bibr" rid="B20">20</xref>]. Several studies indicate that RA-SFB are morphologically and functionally altered [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B12">12</xref>], and strongly express markers of activation, including surface molecules (eg MHC-II, CD13, or VCAM-1 and other integrins) [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>] and proto-oncogenes [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>].</p><p>Activation of SFB <italic>in vitro</italic> generates several functional responses that may considerably contribute to joint pathology in RA; that is, the production of matrix components [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B26">26</xref>], soluble mediators [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B27">27</xref>], or matrix degrading enzymes [<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>,<xref ref-type="bibr" rid="B30">30</xref>]. Because of the capability to act as antigen-presenting cells, activated SFB can also potentially fuel local antigen-driven processes relevant to the immunopathogenesis of RA [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B31">31</xref>]. Finally, stimulated RA-SFB elicit erosive arthritis upon intra-articular injection or engraftment into severe combined immunodeficiency mice [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>].</p><p>To investigate <italic>ex vivo</italic> the pathophysiological properties of activated FB from the RASM, it is desirable to obtain freshly isolated SFB with phenotypic features as close as possible to the configuration observed <italic>in vivo</italic>. Because the generation of cell lines requires a number of passages <italic>in vitro</italic> to eliminate contaminating cells (especially macrophages), the risk of substantial <italic>in vitro</italic> growth selection exists [<xref ref-type="bibr" rid="B32">32</xref>]. Indeed, increasing the number of passages modifies the surface phenotype of synoviocytes [<xref ref-type="bibr" rid="B19">19</xref>] and increases the frequency of genotype changes [<xref ref-type="bibr" rid="B33">33</xref>]. We attempted to isolate FB from primary cultures of synovial cells to minimize these artifacts, with careful characterization of each preparation step and comparison of the phenotypic and functional features of primary-culture SFB with those of repeated-passage cells. Both positive and negative isolation approaches were pursued.</p><p>Positive identification/isolation of SFB was attempted using the mAb AS02, an antibody initially described as FB-specific [<xref ref-type="bibr" rid="B9">9</xref>] and subsequently shown to recognize the Thy-1 molecule (CD90, highly glycosylated membrane-bound protein; molecular weight, 30&#x02013;35 kDa; core protein, 17 kDa; anchored via glycosylphosphatidyl-inositol [<xref ref-type="bibr" rid="B34">34</xref>]).</p><p>Negative isolation was performed using magnetobead-conjugated anti-CD14 mAbs. The choice of CD14 was based on the consideration that this is regarded as a reliable surface marker for RA synovial macrophages [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B35">35</xref>]; CD14 is also the macrophage marker most clearly correlated with disease activity and radiologic progression of joint destruction in RA [<xref ref-type="bibr" rid="B36">36</xref>]. The use of the most valid alternative (ie anti-CD68 mAbs) is limited by predominant expression of the CD68 antigen in the lysosomal compartment (requiring permeabilization of all cells for negative selection), and by the fact that CD68 epitopes recognized by certain mAbs are not specific for macrophages, but are also expressed by SFB [<xref ref-type="bibr" rid="B37">37</xref>,<xref ref-type="bibr" rid="B38">38</xref>]. The possibility of using anti-CD11b mAbs was also discarded, because CD11b is not specific for monocytes/macrophages and is also found on granulocytes and NK cells. CD11b expression on subpopulations of RA synovial macrophages may also be variable [<xref ref-type="bibr" rid="B39">39</xref>], possibly due to the local influence of cytokine stimulation [<xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref>]. Therefore, although the expression of CD14 on RA synovial macrophages may also be influenced by cytokines [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B42">42</xref>,<xref ref-type="bibr" rid="B43">43</xref>], the anti-CD14 mAb appeared the most appropriate choice for the purpose of this study.</p><p>Negative isolation from primary culture with magnetobead-conjugated anti-CD14 mAbs yielded highly enriched RA-SFB, positive for the known FB marker prolyl-4-hydroxylase (85%) and the recently suggested FB marker Thy-1 (74% [<xref ref-type="bibr" rid="B9">9</xref>]), with &#x0003c;2% macrophage contamination and &#x0003c;1% T cells, B cells, plasma cells, NK cells, dendritic cells, endothelial cells, or PMN. These cells bore surface and intracellular characteristics typical of the FB lineage when compared with primary-culture normal skin-FB (lineage control) or OA-SFB (disease control).</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Patients</title><p>Synovial tissue was obtained during open joint replacement surgery or arthroscopic synovectomy from a total of 16 patients with the clinical diagnosis of RA (13 knee joints, one hip joint, one wrist, one metacarpo-phalangeal joint) as well as 21 patients with the clinical diagnosis of OA (all knee joints; Table <xref ref-type="table" rid="T2">2</xref>) from the Department of Orthopedics, University of Leipzig, Germany, the Clinic of Orthopedics, Bad D&#x000fc;ben, Germany, and the Clinic of Orthopedics, Eisenberg, Germany, as well as the Department of Orthopedic Surgery, University of Michigan, Ann Arbor, MI, USA. All RA patients fulfilled the American Rheumatism Association criteria for RA [<xref ref-type="bibr" rid="B8">8</xref>]. The study was approved by the Ethics Committees of the University of Leipzig and the University of Jena, Germany, and the University of Michigan, MI, USA. One portion of each sample was immediately frozen in isopentane (Merck, Darmstadt, Germany), cooled in liquid nitrogen and stored at &#x02013;70&#x000b0;C for immunohistochemistry. The remaining tissue was placed in cell culture medium at ambient temperature and subjected to tissue digestion within 2 h.</p></sec><sec><title>Histochemical detection of non-specific esterase</title><p>Non-specific esterase activity was analyzed in methanol-fixed (10 min at room temperature) first-passage RA-SFB and OA-SFB on chamber slides, and in air-dried, unfixed rat kidney sections or rat peritoneal macrophages, by incubation for 1 h at 37&#x000b0;C with a reaction medium containing 0.02% 5-bromo-4-chloro-3-indoxylacetate (initially dissolved in dimethylformamide), 1 mM potassiumferricyanide, and 1 mM potassiumferrocyanide (final concentrations) (all from Sigma) in 2 mM Tris7#150;HCl buffer (pH 6.8). The positive controls (rat kidney or peritoneal macrophages) were stained in all experiments, while negative control cell preparations/kidney sections incubated with substrate-free medium did not show any positive reaction (respective data not shown). The cells/sections were then washed in PBS (see later) and mounted in glycerin gelatin.</p></sec><sec><title>Immunohistochemistry</title><p>Immunohistochemical analysis was performed on cryostat sections of RASM (fixed with acetone for 10 min at 4&#x000b0;C) or chamber slides (fixed with 100% methanol for 10 min at room temperature; Nunc, Wiesbaden, Germany) using the antibodies presented in Table <xref ref-type="table" rid="T3">3</xref>. The mouse mAbs, diluted in Tris&#x02013;buffered saline (TBS)/1% bovine serum albumin for per-oxidase/alkaline phosphatase detection, were added for 30 min in a humid chamber at room temperature following pre-incubation of the sections/chambers with 20% normal human serum in TBS for 20 min (pH 7.4). A peroxidase-coupled rabbit anti-mouse antibody (Dako) and a peroxi-dase-coupled swine anti-rabbit antibody (Dako; both 1:30 in TBS/20% human serum) were added for 30 min each. The peroxidase was revealed using diaminobenzidine (5 min, 0.5 mg diaminobenzidine in 1 ml PBS [pH 7.4] containing 30 &#x003bc;l H₂O₂). Blocking of endogenous peroxidase (0.03% H₂O₂) before immunohistochemical analysis, as performed in some experiments, did not alter the staining results.</p><p>Alternatively, either the indirect alkaline phosphatase or the alkaline phosphatase anti-alkaline phosphatase (APAAP) techniques were used for detection of the monoclonal or polyclonal primary antibodies [<xref ref-type="bibr" rid="B35">35</xref>]. An alkaline phosphatase-coupled goat-anti-mouse antibody (Dako) was applied for 30 min (1:30 in TBS/20% human serum; pH 7.4) in the case of the indirect alkaline phosphatase, and the alkaline phosphatase was revealed in substrate buffer (TBS; pH 9.5) containing 0.16 mg/ml 5-chromo-4-chloro-3-indoxylphosphate and 0.15 mg/ml nitroblue tetra-zolium chloride (both from Boehringer Mannheim). In the case of the APAAP, a rabbit anti-mouse antibody (Dako) was applied for 45 min (1:100 in TBS/20% human serum; pH 7.4). The sections/chamber slides were washed, incubated for 45 min with APAAP complex (1:80 in TBS/1% bovine serum albumin; Dianova), and washed again with TBS. To amplify the staining, incubation with the rabbit anti-mouse mAb and the APAAP was repeated once. The alkaline phosphatase was revealed in Tris&#x02013;HCl buffer (pH 8.6) containing 0.3 mg/ml Naphthol AS-MX phosphate (first dissolved at 15 mg/ml in dimethylformamide; both from Sigma), 1 mg/ml Fast Red TR salt or Fast Blue BB salt, and 0.24 mg/ml levamisole (all from Sigma). The solution was filtered and added to the sections/chamber slides for 30 min in a humid chamber. For the APAAP technique, polyclonal primary antibodies (against procollagen I and III) were incubated with a secondary mouse anti-rabbit antibody (Dako) before using the secondary detection systems.</p><p>Double-staining experiments on cryostat sections of the RA SM were performed by combining the peroxidase technique with the APAAP technique. The sections were blocked with 20% normal human serum in TBS for 30 min following completion of peroxidase staining. Isotype-matched mouse mAbs, rabbit sera, or purified rabbit immunoglobulins at identical concentrations to the primary antibodies were used as controls for all immunohistochemical analyses (Table <xref ref-type="table" rid="T3">3</xref>), which gave no positive results. Cross-reactivity of the peroxidase and the alkaline phosphatase detection (theoretically possible on the basis of employing mouse primary mAbs and anti-mouse secondary antibodies in both systems) was never observed when using control Ig or when omitting one or both primary mAbs, presumably because the peroxidase reaction with diaminobenzidine eliminates all immunoreactive epitopes from this step.</p><p>For comparative immunohistochemical and FACS analyses, the percentage of cells stained for the individual markers following immunohistochemistry on chamber slides was determined (total of 100 cells analyzed) and compared with the percentage of positive cells in FACS analysis (see later).</p></sec><sec><title>Tissue digestion and cell culture</title><p>Samples of RA or OA SM were finely minced with scissors (tissue pieces of approximately 30 mm³) and digested for 30 min at 37&#x000b0;C in 20 ml PBS containing 0.1% trypsin (Sigma). After removal of trypsin/PBS, the samples were digested in 20 ml 0.1% collagenase P (Boehringer Mannheim) in DMEM/10% FCS for 2 h at 37&#x000b0;C, 5% CO₂. The cell suspension was then filtered through a sterile sieve (Sigma), the cells collected by centrifugation, washed twice with serum-free DMEM medium, and subsequently cultured for 7 days in DMEM/10% FCS, 25 mM HEPES, 100 U/ml penicillin, 100 &#x003bc;g/ml streptomycin, and 2.5 &#x003bc;g/ml amphotericin B (all from Gibco BRL, Karlsruhe, Germany). Primary-culture skin-FB from normal donors (kindly provided by Dr Sauer, Leipzig, Germany) were prepared by first incubating skin samples with 0.5 U/ml Dispase II (Boehringer Mannheim) overnight at 4&#x000b0;C and, following removal of the epidermis, by digesting with 0.25% collagenase P (Boehringer Mannheim) in DMEM/1% FCS at 37&#x000b0;C and 5% CO₂ for 4 h. The resulting cell suspension was then cultured for 14 days as already described. Repeated-passage HUVEC (kindly provided by Dr C. Syring, Institute of Clinical Immunology and Transfusion Medicine, University of Leipzig) were cultured in DMEM/10% FCS/25 mM HEPES containing 10 U/ml heparin, 10 ng/ml basic fibroblast growth factor (Sigma), and 100 &#x003bc;g/ml recombinant endothelial growth factor (Sigma). The medium was changed every 2-3 days in all cases. Adherent RA-SFB or OA-SFB were subjected to very short trypsinization (2 min, 0.25% trypsin/0.2% EDTA; Gibco), removed from the culture dish by mechanical dislocation, washed in PBS/2% FCS, and used for negative isolation. The samples were randomly tested to exclude <italic>Mycoplasma</italic> contamination.</p></sec><sec><title>Negative isolation from primary culture</title><p>Trypsinized and washed RA and OA synovial cells from primary culture (10⁷/ml) were incubated with 4 &#x000d7; 10⁷/ml Dynabeads^&#x000ae; M-450 CD14 (clone RMO52; Dynal) in PBS/2% FCS for 1 h at 4&#x000b0;C under bidirectional rotation. Nine milliliters of PBS/2% FCS were then added and the conjugated cells collected using the Dynal magnetic particle concentrator^&#x000ae;. Magneto-bead-conjugated cells and unconjugated cells were collected and washed twice in PBS/2% FCS; cell composition and phenotype were analyzed by flow cytometry using the antibodies presented in Table <xref ref-type="table" rid="T3">3</xref>. For the purpose of comparison with fourth-passage cells obtained by conventional isolation (ie repeated passaging; see later), negatively isolated RA-SFB and OA-SFB were then passaged four times by culture in complete DMEM/10% FCS (see earlier) with a 1:3 split of confluent cells in each passage.</p></sec><sec><title>Conventional isolation of RA-SFB and OA-SFB by repeated passage</title><p>Synovial cells were obtained by trypsin/collagenase digestion of the RASM as already described. The cells were subsequently cultured for four passages in DMEM/10% FCS containing the aforementioned additives by splitting confluent cells in each passage at a ratio of 1:3.</p></sec><sec><title>Flow cytometry</title><p>The antibodies presented in Table <xref ref-type="table" rid="T3">3</xref> were used for FACS analyses of primary-culture RA and OA synovial cells (following 7 days of culture), isolated RA-SFB and OA-SFB (immediately following isolation or at fourth-passage), conventional fourth-passage RA-SFB and OA-SFB, normal skin-FB (primary-culture or fourth-passage), or repeated-passage endothelial cells (HUVEC). Primary antibodies were used at concentrations of 10&#x02013;20 &#x003bc;g/ml. Standard single and double-staining procedures for surface molecules were performed as previously described [<xref ref-type="bibr" rid="B10">10</xref>]. The specificity of staining was confirmed using isotype-matched control mAbs, rabbit serum, or rabbit Ig at identical concentrations (Table <xref ref-type="table" rid="T3">3</xref>). The FACS analysis in RA-SFB and OA-SFB was not always performed with the complete set of antibodies due to the initial establishment of the isolation procedure.</p><p>Trypsinized cells were washed twice with serum-free medium and fixed for 10 min at 4&#x000b0;C in 4% paraformalde-hyde (Fluka, Deisenhofen, Germany) for detection of antigens in cytoplasm and/or nucleus (eg prolyl-4-hydroxylase; Table <xref ref-type="table" rid="T3">3</xref>). After washing twice in PBS/2% FCS, the pellet was resuspended in permeabilization buffer (PBS/1% FCS, 0.01% NaN₃, and 0.5% saponine; Serva, Heidelberg, Germany) and incubated for 10 min at room temperature. Unlabeled primary mAbs were added at saturating concentrations and detected with a secondary phycoerythrine-labeled goat anti-mouse antibody (Dako), both for 45 min at 4&#x000b0;C in permeabilization buffer. Polyclonal antibodies were incubated with a secondary mouse anti-rabbit antibody (Dako) before using the phycoerythrine-labeled goat anti-mouse antibody. After washing twice in permeabilization buffer, the cells were incubated for 10 min in PBS/5% FCS and 0.1% NaN₃ without saponine. In the special case of the CD68 epitope recognized by the mAb PG-M1, present on the cell surface and in the cytoplasm, FACS analysis was performed in both non-permeabilized and permeabilized cells. For double-staining analyses, fluoresceine isothiocyanate (FITC)-labeled Anti-Thy-1 mAb AS02 (20 &#x003bc;g/ml in PBS/2% FCS) or a matched concentration of the FITC-labeled IgG₁ isotype control mAb were added for 30 min at 4&#x000b0;C.</p><p>Analyses were performed on a FACSCAN^&#x000ae; using the software Cell Quest (Becton Dickinson, San Jose, CA, USA). Forward and side scatter gates were set to include all viable cells. In single-staining experiments, a gate was set to exclude 99% of the cells stained with control Ig. The gates for control Ig in double-staining experiments were placed to limit, to less than 1%, not only the individual percentages in the upper left, upper right, and lower right quadrants, but also the sum of the percentages in the upper left and upper right quadrant, or in the upper right and lower right quadrant.</p></sec><sec><title>Proliferation assays</title><p>The proliferation rates of normal skin-FB as well as OA-SFB and RA-SFB, derived/negatively isolated from primary culture, from the fourth passage of the isolated cells (in the case of OA-SFB and RA-SFB), or from conventional fourth passage, were assessed by seeding the cells at a density of 1.7 &#x000d7; 10³/well in 96-well plates and subsequent culture in DMEM/10% FCS with the aforementioned additives for 24 h at 37&#x000b0;C and 5% CO₂. The cells were then starved in DMEM/1% FCS for 72 h and subsequently stimulated for 18 h by addition of 50, 100, or 150 U/ml IL-1&#x003b2; (Genzyme, R&#x000fc;sselsheim, Germany) or 2.5, 5, or 10 U/ml PDGF-BB (R&#x00026;D Systems, Wiesbaden, Germany). A total of 1 &#x003bc;Ci [³H]-thymidine (Amersham, Braunschweig, Germany) was added to each well, the cells cultured for an additional 18 h, harvested, and the incorporated radioactivity determined in a &#x003b2;-counter (Canberra-Packard, Frankfurt/Main, Germany).</p></sec><sec><title>Statistical analysis</title><p>The data were first subjected to the multi-group Kruskal-Wallis test because of multiple comparisons. Only those parameters showing significant differences (<italic>P</italic> &#x02264; 0.05) underwent further analysis. The non-parametric Mann&#x02013;Whitney U test was then applied for analysis of the phenotypic and functional features. The Spearman Rank correlation test was used to analyze correlations among phenotypic features of SFB and between these features and the clinical status/treatment of individual patients. Differences were considered statistically significant in all cases for <italic>P</italic> &#x02264; 0.05. Analyses were performed using the SPSS 9.0^&#x02122; program (SPSS Inc., Chicago, IL, USA).</p></sec></sec><sec><title>Results</title><sec><title>Expression of Thy-1 in RA synovial tissue</title><p>The mAb AS02 was tested by immunohistochemistry on cryostat sections of RASM to verify the feasibility of this anti-Thy-1 mAb for positive identification/isolation of SFB. The mAb AS02 stained connective tissue cells, while largely sparing the lining layer (see Fig. <xref ref-type="fig" rid="F3">3A</xref> for details), therefore reproducing the known distribution of FB within the SM [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>]. The mAb AS02 showed no overlap with an anti-CD14 mAb in double-staining experiments, thereby excluding crossreactivity with monocytes/macrophages (Fig. <xref ref-type="fig" rid="F3">3C</xref>,<xref ref-type="fig" rid="F3">E</xref>). However, the mAb AS02 stained endothelial cells, as demonstrated by double-staining with rabbit Ig against von Willebrand factor (Fig. <xref ref-type="fig" rid="F3">3D</xref>,<xref ref-type="fig" rid="F3">F</xref>,<xref ref-type="fig" rid="F3">G</xref>). Cultured, non-stimulated HUVEC, in contrast, did not express Thy-1 (data not shown). These results, in line with cytokine induction of Thy-1 expression on endothelial cells in culture [<xref ref-type="bibr" rid="B44">44</xref>,<xref ref-type="bibr" rid="B45">45</xref>,<xref ref-type="bibr" rid="B46">46</xref>], suggest that Thy-1 expression on RASM endothelial cells may reflect ongoing inflammation [<xref ref-type="bibr" rid="B46">46</xref>].</p></sec><sec><title>Phenotype analysis of RA synovial cells in primary culture</title><p>RA synovial cells were subjected to extensive analysis to carefully characterize the starting population before negative isolation (ie the 7-day primary culture of synovial cells resulting from trypsin/collagenase digestion of the RASM). The primary culture contained large, spindle-shaped Thy-1⁺ SFB (Fig. <xref ref-type="fig" rid="F1">1C</xref>) and small, round CD14⁺ cells, most probably macrophages (Fig. <xref ref-type="fig" rid="F1">1D</xref>; mAb Tyk4), as detected by immunohistochemical staining [<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B9">9</xref>]. Endothelial cells were absent, as confirmed by lack of staining for von Willebrand factor (Fig. <xref ref-type="fig" rid="F1">1F</xref>; mAb 4F9) and CD144 (Fig. <xref ref-type="fig" rid="F1">1G</xref>; mAb Cadherin 5), which clearly identified HUVEC (data not shown). The FB nature of the spindle-shaped cells was confirmed by intracellular staining for procollagen I and III (Fig. <xref ref-type="fig" rid="F1">1E</xref>,<xref ref-type="fig" rid="F1">H</xref>; rabbit antibodies MP I and MP III). An average of approximately 62% of the cells stained with the anti-Thy-1 mAb AS02 (<italic>n</italic> = 4 RA patients; Table <xref ref-type="table" rid="T1">1a</xref> and Fig. <xref ref-type="fig" rid="F4">4A</xref>) in FACS analysis [<xref ref-type="bibr" rid="B10">10</xref>]. The average percentage of CD14⁺ cells was approximately 15% (<italic>n</italic> = 4; Table <xref ref-type="table" rid="T1">1a</xref> and Fig. <xref ref-type="fig" rid="F4">4B</xref>). There were &#x0003c;1% T cells (CD3⁺; mAb UCHT-1), B cells (CD19⁺/20⁺; mAbs HD 37 and B-Ly 1), plasma cells (CD38⁺; mAb AT 13/5), NK cells (CD56⁺; mAb NKH/1), dendritic cells (CD83⁺; mAb HB 15a), endothelial cells (CD144⁺), or PMN (CD15⁺; mAb 80H5), indicating that non-adherent cells had been efficiently removed during primary culture. Endothelial cells presumably did not adhere to the culture dish due to the absence of gelatin coating and unfavorable medium composition (see Materials and Methods).</p><p>Adherent synovial cells were then detached by short-term trypsinization for 2 min (0.25% trypsin/0.2% EDTA; Gibco) and used for negative isolation. The total yield of cells following 7 days of RA primary culture averaged (5.2 &#x000b1; 1.1) &#x000d7; 10⁷ cells (mean &#x000b1; SEM; <italic>n</italic> = 7; open synovectomy samples). The yield of cells from arthroscopic synovectomy samples ([3.1 &#x000b1; 0.6] &#x000d7; 10⁷ cells; <italic>n</italic> = 3) was comparable. There was no significant difference between the yield of cells from the RA or OA SM.</p></sec><sec><title>Flow cytometry/histochemical analysis of cells negatively isolated from primary culture</title><p>Negative isolation from primary culture using Dynabeads^&#x000ae; M-450 CD14 (clone RMO52) resulted in cells that were Thy-1⁺ (on average, approximately 74%; <italic>n</italic> = 9; Fig. <xref ref-type="fig" rid="F2">2A</xref> and Table <xref ref-type="table" rid="T1">1b</xref>) and, more importantly, prolyl-4-hydroxylase⁺ (on average, approximately 85%; <italic>n</italic> = 9; Table <xref ref-type="table" rid="T1">1b</xref>) (mAb 3-2B12, Dianova), as shown by FACS analysis (Table <xref ref-type="table" rid="T1">1b</xref>) and confirmed by immunohistochemistry in chamber slides. There were very few contaminating non-specific esterase⁺ (<italic>n</italic> = 3 RA and <italic>n</italic> = 3 OA patients; Fig. <xref ref-type="fig" rid="F2">2C</xref>), CD14⁺, CD68⁺, and/or CD11b⁺ macrophages (&#x0003c;2%; Fig. <xref ref-type="fig" rid="F2">2B</xref>,<xref ref-type="fig" rid="F2">D</xref> and Table <xref ref-type="table" rid="T1">1b</xref>), as well as &#x0003c;1% T cells (CD3⁺), B cells (CD19⁺/20⁺), plasma cells (CD38⁺), NK cells (CD56⁺), dendritic cells (CD83⁺), PMN (CD15), or endothelial cells (CD144⁺; von Wille-brand factor-positive).</p><p>The average yield of RA-SFB negatively isolated from primary culture was (2.8 &#x000b1; 0.9) &#x000d7; 10⁷ cells (mean &#x000b1; SEM; <italic>n</italic> = 7; open synovectomy samples). A similar yield was observed upon isolation of cells from arthroscopic synovectomy samples ([1.7 &#x000b1; 0.6] &#x000d7; 10⁷ cells; <italic>n</italic> = 3). There was no significant difference between the yield from the primary culture of the RASM and that of the OA SM.</p><p>Inclusion of the trypsin component in the initial tissue digestion led to a considerably higher yield of cells following negative isolation from primary culture (1.2-fold to 5-fold; <italic>n</italic> = 3) than without trypsin or with DNAse instead of trypsin.</p><p>The composition of the cells was very similar to that obtained with primary-culture or first-passage normal skin-FB (Table <xref ref-type="table" rid="T1">1g</xref>). Comparable results, although with a considerably lower percentage of Thy-1⁺ cells, were also obtained when OA-SFB were negatively isolated from primary culture (Table <xref ref-type="table" rid="T1">1e</xref>). Conventional fourth-passage RA-SFB (Table <xref ref-type="table" rid="T1">1d</xref>) or OA-SFB (Table <xref ref-type="table" rid="T1">1f</xref>) showed FB markers on a high percentage of cells (&#x0003e;98% prolyl-4-hydroxylase⁺ cells; &#x0003e;80% Thy-1⁺ cells) and contained virtually no contaminating macrophages (&#x0003c;2% CD14⁺-positive or CD68⁺-positive cells). The same was true for isolated primary RA-SFB kept in culture until fourth passage (Table <xref ref-type="table" rid="T1">1c</xref>).</p></sec><sec><title><italic>In vitro</italic> morphology of negatively isolated RA-SFB upon reculture</title><p>Negatively isolated RA-SFB showed almost exclusively spindle-shaped or stellate, flat morphology when recultured (Fig. <xref ref-type="fig" rid="F5">5A</xref>). Recultured CD14⁺ cells (Fig. <xref ref-type="fig" rid="F5">5B</xref>) exhibited small, round morphology with attached Dynabeads^&#x000ae; and contained only very few cells with FB morphology.</p></sec><sec><title>Phenotype characterization of negatively isolated RA-SFB</title><p>To verify whether isolated RA-SFB displayed the features observed <italic>in situ</italic> in the RASM [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>, <xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B28">28</xref>], the expression of several surface or intracellular/nuclear molecules was investigated by FACS analysis (Fig. <xref ref-type="fig" rid="F6">6</xref> and Table <xref ref-type="table" rid="T4">4</xref>). The phenotype was compared with that of primary-culture normal skin-FB and isolated primary-culture OA-SFB (Fig. <xref ref-type="fig" rid="F6">6</xref> and Table <xref ref-type="table" rid="T4">4</xref>).</p><sec><title>Surface antigens</title><p>Approximately 74% of RA-SFB expressed the Thy-1 antigen (Fig. <xref ref-type="fig" rid="F6">6G</xref> and Table <xref ref-type="table" rid="T4">4</xref>), compared with 91% of normal skin-FB (Fig. <xref ref-type="fig" rid="F6">6A</xref> and Table <xref ref-type="table" rid="T4">4</xref>). Approximately 77% of the cells expressed MHC-I molecules, compared with 90% of normal skin-FB (Table <xref ref-type="table" rid="T4">4</xref>). Strikingly, approximately 66% of the cells expressed MHC-II molecules (Fig. <xref ref-type="fig" rid="F6">6H</xref> and Table <xref ref-type="table" rid="T4">4</xref>), significantly more than OA-SFB (17%; Fig. <xref ref-type="fig" rid="F6">6E</xref> and Table <xref ref-type="table" rid="T4">4</xref>) and normal skin-FB (2%; Fig. <xref ref-type="fig" rid="F6">6B</xref> and Table <xref ref-type="table" rid="T4">4</xref>). At the same time, however, the MFI for MHC-II did not significantly differ between RA-SFB and normal skin-FB (Table <xref ref-type="table" rid="T4">4</xref>).</p><p>The expression of CD13 (aminopeptidase N; EC 3.4.11.2) on 85% of RA-SFB was comparable with that of normal skin-FB (84%; Table <xref ref-type="table" rid="T4">4</xref>).</p><p>A low or moderate and variable percentage of normal skin-FB, as well as OA-SFB and RA-SFB, expressed VCAM-1 (using two different anti-VCAM-1 mAbs; see Table <xref ref-type="table" rid="T3">3</xref>) without statistically significant differences among these three different FB preparations (Table <xref ref-type="table" rid="T4">4</xref>).</p></sec><sec><title>Cytoplasmic antigens</title><p>Approximately 85% of the RA-SFB expressed the FB marker prolyl-4-hydroxylase (Fig. <xref ref-type="fig" rid="F6">6M</xref>, and Tables <xref ref-type="table" rid="T1">1</xref> and <xref ref-type="table" rid="T4">4</xref>), comparable with the expression in normal skin-FB (71%; Tables <xref ref-type="table" rid="T1">1</xref> and <xref ref-type="table" rid="T4">4</xref>).</p><p>RA-SFB showed a similar percentage but a higher MFI for the cytoskeletal protein vimentin (Fig. <xref ref-type="fig" rid="F6">6I</xref>) than normal skin-FB (Fig. <xref ref-type="fig" rid="F6">6C</xref> and Table <xref ref-type="table" rid="T4">4</xref>), although the MFI difference did not attain statistical significance.</p><p>A mean of approximately 45% and 50% of the cells, respectively (Table <xref ref-type="table" rid="T4">4</xref>), expressed procollagen I (Fig. <xref ref-type="fig" rid="F6">6J</xref>) and procollagen III (Fig. <xref ref-type="fig" rid="F6">6K</xref>), as confirmed by double-staining (Fig. <xref ref-type="fig" rid="F6">6L</xref>). The percentage of cells expressing procollagen I and III was similar to that of normal skin-FB (68% and 55%, respectively; Table <xref ref-type="table" rid="T4">4</xref>); however, the MFI for procollagen III was significantly higher in RA-SFB (26 versus 15; Table <xref ref-type="table" rid="T4">4</xref>).</p></sec><sec><title>Cytoplasmic/intranuclear antigens</title><p>Approximately 57% of RA-SFB expressed <italic>c</italic>-Fos (Fig. <xref ref-type="fig" rid="F6">6N</xref> and Table <xref ref-type="table" rid="T4">4</xref>). Neither the percentage nor the MFI were significantly different from those of normal skin-FB.</p><p>The proto-oncogenes <italic>c</italic>-Jun and Jun-D (both percentage of positive cells and MFI) were also comparably expressed in RA-SFB and normal skin-FB (Fig. <xref ref-type="fig" rid="F6">6O</xref> and Table <xref ref-type="table" rid="T4">4</xref>).</p><p>The differences between RA-SFB and normal skin-FB were, in general, not specific to RA since they were also observed in the comparison between the disease control OA-SFB and normal skin-FB (Table <xref ref-type="table" rid="T4">4</xref>). In fact, some of the differences in OA were even more pronounced than in RA, attaining statistical significance in the direct comparison between OA and RA (Table <xref ref-type="table" rid="T4">4</xref>). Briefly, significantly higher MFI levels in OA-SFB were noted for MHC-I, MHC-II and procollagen I, as well as <italic>c</italic>-Fos. Opposite differences among the different FB preparations, when considering the percentage of positive cells or the MFI (eg for the comparison of MHC-II and procollagen I and III expression in RA-SFB versus OA-SFB), appeared due to large variations in one or both parameters.</p></sec><sec><title>Comparison of FACS analysis and immunohistochemistry</title><p>To verify whether the short-term trypsin digestion (2 min at 37&#x000b0;C) employed for removal of the cells from the culture dishes and subsequent FACS analysis altered the expression of the antigens, selected experiments were carried out to compare the expression of surface and intracellular antigens in FACS analysis (Table <xref ref-type="table" rid="T1">1b</xref>,<xref ref-type="table" rid="T1">e</xref>, and Table <xref ref-type="table" rid="T4">4</xref>) and immunohistochemistry on chamber slides. Negatively isolated SFB from <italic>n</italic> = 3 OA and <italic>n</italic> = 3 RA patients showed comparable percentages of positive cells for surface antigens (RA-Thy-1, 95.3% FACS versus 85% immunohistochemistry; CD14, 1.1% versus 0%; VCAM-1, 0.7% versus 0%) and intracellular antigens (RA-prolyl-4-hydroxylase, 94.7% versus 85%; CD68, 1.7% versus 0%) with both methods, confirming previously published results [<xref ref-type="bibr" rid="B47">47</xref>] and excluding the possibility that short-term trypsinization significantly influences antigen expression.</p></sec></sec><sec><title>Phenotype comparison of primary-culture and fourth-passage RA-SFB</title><p>In order to assess differences between the primary RA-SFB derived from the isolation technique developed in the present study (immediately following isolation or at fourth passage) and those obtained by conventional repeated passaging, the phenotypic features of the three preparations were compared.</p><sec><title>Isolated primary RA-SFB versus conventional fourth-passage RA-SFB</title><p>The percentages of RA-SFB positive for MHC-II, as well as the MFI for VCAM-1 and <italic>c</italic>-Jun, were significantly decreased in conventional fourth passage in comparison with isolated primary RA-SFB (Fig. <xref ref-type="fig" rid="F7">7C</xref>,<xref ref-type="fig" rid="F7">D</xref>,<xref ref-type="fig" rid="F7">K</xref>,<xref ref-type="fig" rid="F7">L</xref> and Table <xref ref-type="table" rid="T5">5</xref>). The percentages of cells positive for MHC-I, CD13, prolyl-4-hydroxylase, vimentin, procollagen I and III, <italic>c</italic>-Fos and Jun-D were, in contrast, significantly increased in conventional fourth passage (Fig. <xref ref-type="fig" rid="F7">7E</xref>,<xref ref-type="fig" rid="F7">F</xref>,<xref ref-type="fig" rid="F7">G</xref>,<xref ref-type="fig" rid="F7">H</xref>,<xref ref-type="fig" rid="F7">I</xref>,<xref ref-type="fig" rid="F7">J</xref> and Table <xref ref-type="table" rid="T5">5</xref>).</p><p>The upregulation of the proto-oncogenes <italic>c</italic>-Fos and Jun-D was limited to RA, since the percentages of cells positive for these molecules were significantly decreased upon passaging in OA-SFB (<italic>c</italic>-Fos: isolated primary-culture OA-SFB, 53.9 &#x000b1; 10.4%; conventional fourth-passage OA-SFB, 15.0 &#x000b1; 6.6%; Jun-D: isolated primary-culture OA-SFB, 37.6 &#x000b1; 17.7%; conventional fourth-passage OA-SFB, 25.1 &#x000b1; 4.2%; <italic>n</italic> =3 for both SFB preparations) and numerically decreased upon passaging in normal skin-FB (<italic>c</italic>-Fos: primary-culture/first-passage skin-FB, 87.2 &#x000b1; 2.1%; conventional fourth-passage skin-FB, 59.4 &#x000b1; 26.0%; Jun-D: primary-culture/first-passage skin-FB, 87.4 &#x000b1; 1.5%; conventional fourth-passage skin-FB, 66.2 &#x000b1; 18.6%; <italic>n</italic> = 3 for both). These decreases were also noted for the MFI (data not shown). The percentage of positive cells and/or MFI for <italic>c</italic>-Fos and Jun-D in conventional fourth-passage RA-SFB were significantly higher than in OA-SFB as a consequence of these reciprocal changes, under these circumstances confirming previously published data [<xref ref-type="bibr" rid="B25">25</xref>].</p></sec><sec><title>Isolated primary RA-SFB versus isolated fourth-passage RA-SFB</title><p>Some of the differences observed between isolated primary RA-SFB and conventional fourth passage also arose when the isolated primary RA-SFB were passaged until the fourth passage; this applied to the significantly increased percentage of cells positive for MHC-I, prolyl-4-hydroxylase, vimentin, and procollagen I and III, as well as to the decreased percentage of MHC-II-positive cells (Table <xref ref-type="table" rid="T5">5</xref>). In addition, isolated fourth-passage RA-SFB showed a significantly decreased MFI (but a numerically increased percentage) for Thy-1 in comparison with isolated primary RA-SFB (Table <xref ref-type="table" rid="T5">5</xref>).</p></sec><sec><title>Conventional fourth-passage RA-SFB versus isolated fourth-passage RA-SFB</title><p>The most important difference between these two preparations was the significantly increased percentage of cells positive for the proto-oncogenes <italic>c</italic>-Fos and Jun-D, observed only in conventional fourth-passage RA-SFB (Table <xref ref-type="table" rid="T5">5</xref>). The MFI for Thy-1 was also significantly higher in conventional fourth-passage cells.</p></sec></sec><sec><title>Proliferation rates of first- and fourth-passage FB</title><p>The proliferation rates of skin-FB and SFB were determined and the different preparations compared.</p><sec><title>Isolated first-passage FB versus conventional fourth-passage FB</title><p>At rest, the proliferation rates of conventional fourth-passage normal skin-FB did not significantly differ from those of the corresponding cells in first passage (Fig. <xref ref-type="fig" rid="F8">8A</xref>). The same was true for conventional fourth passage OA-SFB and RA-SFB (Fig. <xref ref-type="fig" rid="F8">8B</xref>,<xref ref-type="fig" rid="F8">C</xref>). Upon stimulation with PDGF (2.5 and 5 U/ml for all cells, 10 U/ml only for OA-SFB), however, the proliferation rates of skin, RA, and OA conventional fourth-passage FB were significantly higher (maximum mean increase, 5.4-fold) than those observed in first-passage cells (Fig. <xref ref-type="fig" rid="F8">8A</xref>,<xref ref-type="fig" rid="F8">B</xref>,<xref ref-type="fig" rid="F8">C</xref>). Following stimulation with IL-1&#x003b2;, in contrast, only conventional fourth-passage RA-SFB (150 U/ml IL-1&#x003b2;) showed significantly higher proliferation rates than first-passage cells (Fig. <xref ref-type="fig" rid="F8">8C</xref>).</p></sec><sec><title>Isolated first-passage versus isolated fourth-passage OA-SFB and RA-SFB</title><p>The proliferation rates of isolated fourth-passage cells in OA-SFB differed significantly from those of isolated first-passage cells only upon stimulation with IL-1&#x003b2; (50 and 150 U/ml; Fig. <xref ref-type="fig" rid="F8">8B</xref>). Such differences were not observed at rest or following stimulation with PDGF.</p><p>Isolated fourth-passage RA-SFB, in contrast, showed significantly higher proliferation rates than first-passage cells upon stimulation with all concentrations of IL-1&#x003b2; and PDGF, but not at rest (Fig. <xref ref-type="fig" rid="F8">8C</xref>).</p></sec><sec><title>Conventional fourth-passage versus isolated fourth-passage OA-SFB and RA-SFB</title><p>The proliferation rates of isolated fourth-passage SFB in both OA and RA were comparable with those of conventional fourth-passage SFB, whether at rest or PDGF-stimulated (Fig. <xref ref-type="fig" rid="F8">8B</xref>,<xref ref-type="fig" rid="F8">C</xref>). Stimulation with IL-1&#x003b2; (at all concentrations), in contrast, induced significantly higher proliferation in isolated fourth-passage RA-SFB than in conventional fourth-passage RA-SFB (Fig. <xref ref-type="fig" rid="F8">8C</xref>). This difference was specific for RA-SFB, since it was not observed in OA-SFB (Fig. <xref ref-type="fig" rid="F8">8B</xref>).</p><p>As a consequence of differential IL-1&#x003b2; sensitivity in RA-SFB and OA-SFB, there were significant differences between isolated fourth-passage RA-SFB and OA-SFB (<italic>P</italic> &#x02264; 0.05; RA&#x0003e;OA, for 100 U/ml IL-1&#x003b2;), as well as between conventional fourth-passage RA-SFB and OA-SFB (<italic>P</italic> &#x02264; 0.05; OA&#x0003e;RA, for 50 and 150 U/ml IL-1&#x003b2;; Fig. <xref ref-type="fig" rid="F8">8B</xref>,<xref ref-type="fig" rid="F8">C</xref>).</p></sec></sec><sec><title>Alternative attempts to isolate RA-SFB</title><p>The following isolation techniques were also tested to compare the quality and reliability of negative isolation from primary culture with possible alternative methods. The first technique was positive isolation directly from the trypsin/collagenase digest using the FB-directed anti-Thy1/CD90 mAb AS02 and Dynabeads^&#x000ae; M-450 goat anti-mouse IgG. This approach, however, considered legitimate on the basis of the immunostaining in RA synovial tissue (Fig. <xref ref-type="fig" rid="F3">3A</xref>,<xref ref-type="fig" rid="F3">E</xref>), remained unsatisfactory (yield, only approximately 6% of the initial cell numbers). Another technique involved negative isolation directly from the trypsin/collagenase digest. This approach proved technically prohibitive for the amount of different anti-leukocyte and anti-endothelial mAbs and Dynabeads^&#x000ae; M-450 goat anti-mouse IgG necessary for complete removal of all contaminating cell types. Positive isolation from primary culture with the anti-Thy-1 mAb AS02 and Dynabeads^&#x000ae; M-450 goat anti-mouse IgG was also tested. This method was unsuitable because: first, the percentages of contaminating macrophages were too high/variable (2-15%); second, the isolated SFB showed a significantly reduced surface density for Thy-1 (MFI 19 versus 152 before isolation; <italic>n</italic> = 4, <italic>P</italic> &#x02264; 0.05), presumably due to contact with the mAb used for isolation; third, the magnetobeads remained attached to the positively isolated SFB for as long as 1 week, thus requiring further manipulation for their removal (eg with anti-idiotypic mAbs); fourth, a substantial proportion of SFB (approximately 30%) was lost into the negative fraction; and, finally, a Thy-1-dependent isolation technique bore the risk of selecting for SFB subpopulations, since the prolyl-4-hydroxylase⁺ cells obtained by negative isolation with Dynabeads^&#x000ae; M-450 CD14 contained both a Thy-1⁺ fraction (RA, 49.3&#x000b1; 14.9%, <italic>n</italic> = 6; OA, 43.6 &#x000b1; 12.0%, <italic>n</italic> = 4) and a Thy-1^&#x02013; fraction, as demonstrated by double-staining FACS analysis.</p></sec></sec><sec><title>Discussion</title><sec><title>Negative isolation from primary culture</title><p>Negative isolation of SFB from primary cultures of RASM using magnetobead-coupled anti-CD14 mAbs proved successful, providing satisfactory numbers of cells and avoiding repeated passaging and numerous cumulative population doublings. With culture times as short as 7 days, fresh SFB express surface and intracellular antigens typical of the FB lineage and/or SFB features as reported at a tissue level (Fig. <xref ref-type="fig" rid="F6">6</xref> and Table <xref ref-type="table" rid="T4">4</xref>) [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>, <xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B28">28</xref>]. This approach may also reduce the risk of <italic>in vitro</italic> growth selection. In the case of conventionally isolated RA-SFB, for example, the higher the number of passages, the higher the percentage of cells with trisomy 7 [<xref ref-type="bibr" rid="B33">33</xref>]. Some SFB functions (eg cell density at saturation and expression of transcription factors), in contrast, decrease at later passages [<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B32">32</xref>,<xref ref-type="bibr" rid="B48">48</xref>].</p><p>While the use of enzymes for the initial tissue digestion represents at least a transient stress for the cells, this method may also represent an alternative to tissue-outgrowth techniques, known to select for premitotic FB [<xref ref-type="bibr" rid="B26">26</xref>]. This is important since the relative contribution of pre-mitotic or postmitotic FB to the pathogenetic features of RA is presently unclear.</p></sec><sec><title>Advantages of negative isolation</title><p>Negative isolation with magnetobead-coupled anti-CD14 mAbs shows the following advantages. There is minimal contamination with macrophages (&#x0003c;2%) and other inflammatory cells (all &#x0003c;1%). Thus, although CD14 may be expressed to a lower degree than CD68 on RA synovial lining macrophages [<xref ref-type="bibr" rid="B36">36</xref>,<xref ref-type="bibr" rid="B39">39</xref>], the degree of CD14-positivity appears more than sufficient to conduct the isolation procedure. The positive fraction, in turn, contains virtually no SFB, indicating minimal cell loss, and thereby also excluding major subset selection. Another advantage is that negative isolation with Dynabeads^&#x000ae; M-450 CD14 avoids direct contact of SFB with mAbs and/or magnetobeads, thereby reducing possible functional alterations of the cells. Finally, the anti-Thy-1 mAb AS02 (used in the parallel attempt to positively isolate SFB) identifies 91% of normal skin-FB, but not more than 74% of RA-SFB (Tables <xref ref-type="table" rid="T1">1</xref> and <xref ref-type="table" rid="T4">4</xref>). In addition, the prolyl-4-hydroxylase⁺ cells obtained by negative isolation with Dynabeads^&#x000ae; M-450 CD14 contain both a Thy-1⁺ and a Thy-1^&#x02013; fraction. Thy-1-independent isolation thus circumvents the danger of selecting for Thy-1⁺ SFB subpopulations, especially relevant since Thy-1⁺ and Thy-1^- FB diverge in several characteristics potentially relevant to RA; for example, cytokine and matrix production, as well as antigen presentation [<xref ref-type="bibr" rid="B49">49</xref>] (reviewed in [<xref ref-type="bibr" rid="B50">50</xref>,<xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref>]).</p><p>The negative isolation technique, however, also bears possible disadvantages: the limited number of cells obtained (approximately 2.8 &#x000d7; 10⁷), due to the lack of <italic>in vitro</italic> expansion, and the necessity for reliable access to fresh synovial specimens. On the other hand, the applicability of this procedure not only to joint replacement samples, but also to arthroscopic synovectomy samples, augments the potential sources of fresh synovial tissue.</p></sec><sec><title>Advantages of limited-passage isolation</title><p>Limited-passage isolation shows the following advantages. It avoids phenotypic and functional <italic>in vitro</italic> alterations due to repeated passaging; indeed, even a low number of passages increases the proliferation rates in response to cytokine stimulation (Fig. <xref ref-type="fig" rid="F8">8</xref>). Passaging also alters the cellular phenotype of RA-SFB, as exemplified by a strong decrease of the percentage of MHC-II⁺ cells, on one hand, and a striking increase of the cells positive for procollagen I and III, and the proto-oncogenes <italic>c</italic>-Fos and Jun-D, on the other (Figs. <xref ref-type="fig" rid="F7">7</xref> and <xref ref-type="fig" rid="F8">8</xref>, and Table <xref ref-type="table" rid="T5">5</xref>). These results are relevant especially in view of the question whether proliferation and some phenotypic features of RA-SFB (in particular, the expression of proto-oncogenes) can be assimilated to those of dysregulated growth [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>]. As these features seem to depend on the length of <italic>in vitro</italic> culture, caution should be applied to the interpretation of data from passaged cells [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>, <xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. Another advantage is that this isolation procedure reduces the time of exposure to surviving macrophages to 7 days (in contrast to the weeks typical of conventional passaging), thus limiting long-term <italic>in vitro</italic> changes resulting from monokine secretion or cell-cell contact between SFB and synovial macrophages. Limited-passage isolation may therefore allow one to differentiate between FB-inherent changes (eg growth selection and cell senescence) and the changes mediated by the degree or length of contact with surviving macrophages during the initial passages. A high percentage of RA-SFB (and, to a lesser degree, also OA-SFB) from limited-passage isolation express MHC-II molecules without <italic>in vitro</italic> stimulation with cytokines (Table <xref ref-type="table" rid="T4">4</xref>), which in conventionally isolated cells requires stimulation with interferon-gamma (used not only to explore the antigen-presentation potential of RA-SFB, but also to inhibit collagen biosynthesis [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B53">53</xref>]). In this aspect, the present <italic>in vitro</italic> system may be very useful in studies on the interrelationship of SFB with T cells or macrophages, in which the influence of exogenous mediators complicates the design and/or interpretation of the experiments.</p></sec><sec><title>Similarities/discrepancies between the <italic>in situ</italic> and <italic>in vitro</italic> phenotype of RA-SFB</title><sec><title>Similarities</title><p>The cells obtained by negative isolation generally showed features similar to those reported in <italic>in situ</italic> analyses. The expression of Thy-1, CD13 (aminopeptidaseN), and vimentin exemplify this. The expression of these molecules, particularly their local upregulation by cytokines [<xref ref-type="bibr" rid="B54">54</xref>,<xref ref-type="bibr" rid="B55">55</xref>] and/or heterogeneous expression at different anatomical sites [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B49">49</xref>], may be pathogenetically relevant, either in terms of defining functionally heterogeneous SFB subpopulations [<xref ref-type="bibr" rid="B49">49</xref>] (reviewed in [<xref ref-type="bibr" rid="B50">50</xref>,<xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref>]), by providing pro-inflammatory enzymes [<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B47">47</xref>,<xref ref-type="bibr" rid="B56">56</xref>,<xref ref-type="bibr" rid="B57">57</xref>,<xref ref-type="bibr" rid="B58">58</xref>,<xref ref-type="bibr" rid="B59">59</xref>], or as targets of autoimmune reactions [<xref ref-type="bibr" rid="B60">60</xref>].</p><p><italic>MHC-II.</italic> The constitutive MHC-II expression in isolated RA-SFB and OA-SFB, confirming previous <italic>in situ</italic> observations in the RA and OA SM [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B28">28</xref>] and in systemic scleroderma [<xref ref-type="bibr" rid="B61">61</xref>], is probably related to the inflammatory microenvironment, since normal skin-FB (Table <xref ref-type="table" rid="T4">4</xref>) and normal SFB [<xref ref-type="bibr" rid="B31">31</xref>] hardly express MHC-II. This is further supported by the a strong decrease of the percentage of MHC-II⁺ cells upon repeated passage (both conventional and following negative isolation), possibly due to lack of or a progressive decrease of external stimulation with pro-inflammatory mediators (Fig. <xref ref-type="fig" rid="F7">7</xref> and Table <xref ref-type="table" rid="T5">5</xref>). Notably, there was a significant, negative correlation between the percentage of MHC-II-positive SFB and treatment of the RA patients with Methotrexate (&#x003c1;=-0.866, <italic>P</italic> = 0.01, <italic>n</italic> = 7), indicating that effective antirheumatic therapy may be reflected by decreased MHC-II expression on RA-SFB [<xref ref-type="bibr" rid="B28">28</xref>].</p><p><italic>Procollagen III.</italic> The significant increase of the MFI for intracellular procollagen III in RA-SFB and OA-SFB (Table <xref ref-type="table" rid="T4">4</xref>) suggests that the pattern of SFB activation may be functional (among other things) to the increase of collagen metabolism; this is supported by the increased expression of collagen &#x003b1;₂I and &#x003b1;₁III mRNA in the RASM <italic>in situ</italic> [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>], in comparison with normal or non-RA synovial tissue. Also, in as much as procollagen III is the fetal form of collagen used in wound healing and tissue repair (reviewed in [<xref ref-type="bibr" rid="B62">62</xref>,<xref ref-type="bibr" rid="B63">63</xref>]), ongoing fibrosis may be a considerable component of the disease process in RA (and OA), in analogy to systemic scleroderma [<xref ref-type="bibr" rid="B61">61</xref>] or interstitial kidney fibrosis [<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B26">26</xref>]. A striking increase of procollagen I and III expression in RA-SFB from primary culture to fourth passage (increase of positive cells by &#x02265; 38%; Fig. <xref ref-type="fig" rid="F7">7</xref> and Table <xref ref-type="table" rid="T5">5</xref>), on the other hand, indicates that primary-culture cells do not exploit their full potential of matrix production.</p></sec><sec><title>Discrepancies</title><p>The expression of VCAM-1 [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B64">64</xref>,<xref ref-type="bibr" rid="B65">65</xref>] and Jun and Fos proto-oncogenes [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B66">66</xref>,<xref ref-type="bibr" rid="B67">67</xref>] showed some <italic>in situ</italic> / <italic>in vitro</italic> discrepancies.</p><p><italic>VCAM-1.</italic> Surprisingly, only a moderate and variable percentage of RA-SFB (whether conventionally passaged or negatively isolated from primary culture) expressed the adhesion molecule VCAM-1 (Tables <xref ref-type="table" rid="T4">4</xref> and <xref ref-type="table" rid="T5">5</xref>). There were also no significant differences between RA-SFB and OA-SFB or normal skin-FB (Table <xref ref-type="table" rid="T4">4</xref>). This is in apparent contrast to the enhanced expression of VCAM-1 reported in the lining layer of RA and OA synovial tissue [<xref ref-type="bibr" rid="B64">64</xref>,<xref ref-type="bibr" rid="B65">65</xref>]; however, it is well compatible with the large variability in VCAM-1 expression observed <italic>in vitro</italic> (Table <xref ref-type="table" rid="T4">4</xref>) [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B68">68</xref>] and <italic>in situ</italic> [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. The significant, positive correlation between VCAM-1 expression in isolated primary-culture RA-SFB and the erythrocyte sedimentation rate in RA patients (&#x003c1;=1.00, <italic>P</italic> = 0.000, <italic>n</italic> = 4) indicates that the variability may depend on disease activity, as also suggested in recent reports [<xref ref-type="bibr" rid="B68">68</xref>,<xref ref-type="bibr" rid="B69">69</xref>,<xref ref-type="bibr" rid="B70">70</xref>].</p><p><italic>Proto-oncogene expression.</italic> Negatively isolated RA-SFB expressed <italic>c</italic>-Fos, <italic>c</italic>-Jun, and Jun-D (Table <xref ref-type="table" rid="T4">4</xref>), in analogy to the features of SFB in RASM [<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B66">66</xref>]. The expression of <italic>c</italic>-Jun and Jun-D in the present study could be unequivocally demonstrated by FACS analysis (Table <xref ref-type="table" rid="T4">4</xref>), while previous immunohistochemical studies had failed to detect these molecules <italic>in situ</italic> [<xref ref-type="bibr" rid="B23">23</xref>]. Notably, however, the degree of proto-oncogene expression (both percentage of positive cells and MFI) did not significantly differ from that of normal skin-FB (Fig. <xref ref-type="fig" rid="F6">6</xref> and Table <xref ref-type="table" rid="T4">4</xref>) or, as previously reported, of SFB from traumatic joint injury [<xref ref-type="bibr" rid="B67">67</xref>]. This finding supports <italic>in situ</italic> observations that, at a single-cell level, the degree of proto-oncogene expression in RA, OA, and joint trauma is similar [<xref ref-type="bibr" rid="B23">23</xref>], thereby questioning whether proto-oncogene expression in RA reflects <italic>per se</italic> cell transformation and/or severe metabolic abnormalities.</p><p>Notably, the expression of the proto-oncogenes <italic>c</italic>-Fos and Jun-D was strikingly increased in conventional fourth-passage RA-SFB as compared with isolated primary-culture RA-SFB (increase of positive cells &#x02265; 38%; Fig. <xref ref-type="fig" rid="F7">7</xref> and Table <xref ref-type="table" rid="T5">5</xref>). This increase was limited to RA, since the percentage of cells positive for these molecules was significantly decreased in OA-SFB and numerically decreased in normal skin-FB upon passaging. As a consequence of these reciprocal changes, the percentage of positive cells and/or MFI for <italic>c</italic>-Fos and Jun-D in conventional fourth-passage RA-SFB was significantly higher than in OA-SFB, under these circumstances confirming previously published data [<xref ref-type="bibr" rid="B25">25</xref>].</p><p>Finally, a significant, positive correlation between the percentage of <italic>c</italic>-fos⁺ and procollagen I and III⁺ isolated primary-culture SFB in both RA and OA patients (RA: <italic>c</italic>-fos/procollagen I, &#x003c1;=1.000, <italic>P</italic> = 0.00, <italic>n</italic> = 7; <italic>c</italic>-fos/procollagen III, &#x003c1;=0.964, <italic>P</italic> = 0.00, <italic>n</italic> = 7) suggests a link between <italic>c</italic>-fos expression and augmented matrix production by SFB in rheumatic disorders, in line with the known regulation of collagen expression by the activator protein-1 transcription factor (reviewed in [<xref ref-type="bibr" rid="B13">13</xref>]).</p></sec></sec><sec><title>Suitability of fibroblast markers for unequivocal identification of SFB</title><p>None of the markers used to recognize SFB [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B62">62</xref>] proved to be universal (Table <xref ref-type="table" rid="T4">4</xref>) or stable upon passaging (Fig. <xref ref-type="fig" rid="F7">7</xref> and Table <xref ref-type="table" rid="T5">5</xref>). Cellular uridine diphosphoglucose dehydrogenase activity, a useful histochemical marker for intimal SFB [<xref ref-type="bibr" rid="B37">37</xref>], while clearly detectable by immunohistochemistry in the lining layer of RASM tissue sections, was also only positive in a small subpopulation of cultured SFB (90% negative, 9% weakly positive, 1% strongly positive). Thus, despite its parallel recognition of endothelial cells [<xref ref-type="bibr" rid="B62">62</xref>], prolyl-4-hydroxylase at present remains the most suitable FB marker (see Tables <xref ref-type="table" rid="T1">1</xref>,<xref ref-type="table" rid="T4">4</xref> and <xref ref-type="table" rid="T5">5</xref>).</p><p>The variable degree of positivity for different markers observed in the present study supports the existence of different SFB subpopulations or maturational stages (Tables <xref ref-type="table" rid="T4">4</xref> and <xref ref-type="table" rid="T5">5</xref>, and Fig. <xref ref-type="fig" rid="F6">6E</xref>,<xref ref-type="fig" rid="F6">H</xref>,<xref ref-type="fig" rid="F6">I</xref>) [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B15">15</xref>]. By providing isolated SFB that do not undergo major subset selection, the present technique may allow, on one hand, the further characterization of these subpopulations and, on the other, the identification of more universal and/or less regulated pan-FB markers.</p><p>In summary, the present technique allows the isolation of highly enriched SFB with very limited <italic>in vitro</italic> culture time. These cells clearly show different phenotypic and functional properties in comparison with conventional fourth-passage RA-SFB. In conjunction with simultaneously derived synovial macrophages from the same patient (eg the positive fraction of the CD14 negative isolation), normally lost upon repeated passage due to their inability to divide <italic>in vitro</italic> (reviewed in [<xref ref-type="bibr" rid="B1">1</xref>]), and non-adherent, viable cells from the supernatant of the primary culture, recreation of a simplified 'rheumatoid' system can be attempted to investigate the relative contribution of the different cells to joint pathology [<xref ref-type="bibr" rid="B18">18</xref>,<xref ref-type="bibr" rid="B71">71</xref>].</p></sec></sec><sec><title>Note added in proof</title><p>Increased levels of c-Fos protein in late-passage versus early-passage RA-SFB, but decreased levels of c-Fos in late-passage OA-SFB, ie an RA-specific increase of c-Fos transcription/translation in late-passage SFB, has recently been confirmed by Shimizu <italic>et al</italic> [<xref ref-type="bibr" rid="B72">72</xref>].</p></sec>

Synovial stromal cells from rheumatoid arthritis patients attract monocytes by producing MCP-1 and IL-8

<p>Macrophages that accumulate in the synovium of rheumatoid arthritis patients play an important role in the pathogenesis of this inflammatory disease. However, the mechanism by which macrophages are attracted into the inflamed synovium and accumulate there has not been completely delineated. The results of this study show that rheumatoid arthritis synovial stromal cells produce the chemokines monocyte chemotactic protein-1 and IL-8, and these have the capacity to attract peripheral monocytes. These results suggest that one of the mechanisms by which macrophages accumulate in the inflamed synovium is by responding to the chemokines produced locally.</p>

<contrib id="A1" contrib-type="author"><name><surname>Hayashida</surname><given-names>Kenji</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A2" contrib-type="author"><name><surname>Nanki</surname><given-names>Toshihiro</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Girschick</surname><given-names>Hermann</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Yavuz</surname><given-names>Sule</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Ochi</surname><given-names>Takahiro</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A6" corresp="yes" contrib-type="author"><name><surname>Lipsky</surname><given-names>Peter E</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>lipskyp@mail.nih.gov</email></contrib>

Arthritis Research

<sec><title>Introduction</title><p>The synovial tissue in rheumatoid arthritis (RA) contains synovial fibroblasts and stromal cells as well as macrophages. Synovial stromal cells and fibroblasts are thought to proliferate <italic>in situ</italic> [<xref ref-type="bibr" rid="B1">1</xref>]. On the contrary, macrophages do not proliferate, but rather differentiate from monocytes that migrated from the peripheral blood, and are activated to differentiate in the synovial tissue. Macrophages in RA synovium secrete many inflammatory mediators (IL-1, IL-6, IL-8, tumor necrosis factor-&#x003b1; [TNF-&#x003b1;] and PGE₂) and a variety of matrix metalloproteinases, and are thought to play a central role in the inflammation and joint destruction characteristic of RA [<xref ref-type="bibr" rid="B2">2</xref>]. The number of macrophages in RA synovium correlates significantly with clinical symptoms and the degree of joint damage [<xref ref-type="bibr" rid="B3">3</xref>]. Understanding the regulation of macrophage accumulation in the RA synovium should therefore provide insight into the inflammatory nature of rheumatoid synovitis.</p><p>Migration of peripheral blood monocytes is likely to be influenced by chemokines. A number of chemokines, including IL-8, monocyte chemotactic protein-1 (MCP-1), macrophage inflammatory protein-1&#x003b1; (MIP-1&#x003b1;), MIP-1&#x003b2;, epithelial-derived neutrophil attractant 78, and regulated upon activation, normal T cell expressed and secreted (RANTES), are known to be found in RA synovial fluid [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>]. A number of specific chemokine receptors including CCR1, CCR2, CCR5, CCR8, and CXCR4 are also known to be expressed by peripheral blood monocytes [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>]. Despite this information, the specific chemokine&#x02013;chemokine receptor interactions involved in the recruitment of monocytes into the rheumatoid synovium have not been fully delineated.</p><p>To address this issue, we examined chemokine receptor expression by peripheral blood monocytes, and also analyzed the capacity of supernatants from RA synovial stromal cells to induce monocyte migration. The data indicate that MCP-1 secreted by synovial stromal cells plays a major role in attracting monocytes to the synovium, and that IL-8 may also contribute.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Antibodies and reagents</title><p>Biotinylated mouse anti-human CCR1 monoclonal antibody (mAb), mouse anti-human CCR2 mAb conjugated with phycoerythrin (PE), mouse anti-human CCR6 mAb conjugated with PE, mouse anti-human CXCR1 mAb conjugated with PE, mouse anti-human CXCR2 mAb conjugated with PE, mouse anti-human CXCR5 mAb conjugated with PE, mouse anti-human CCR3 mAb conjugated with FITC, mouse anti-human MCP-1 mAb (24822.111), mouse anti-human IL-8 mAb (6217.111), mouse anti-human IP-10 mAb (33036.211), mouse anti-human CCR5 mAb (45531.111), and biotinylated goat IgG anti-human IP-10 were purchased from R&#x00026;D Systems Inc (Miami, FL). Mouse anti-human CCR5 mAb conjugated with FITC and mouse anti-human CXCR4 mAb conjugated with PE were purchased from Pharmingen (San Diego, CA). Mouse anti-human CD14 mAb conjugated with FITC or PE, and Streptavidin conjugated with PE were obtained from Sigma (St Louis, MO). Mouse IgG1 mAb was prepared from hybridoma cell lines purchased from ATCC (Rockville, MD). Mouse IgG2A conjugated with PE (Pharmingen), mouse IgG2B conjugated with PE (R&#x00026;D Systems), and mouse IgG1 conjugated with FITC (Becton Dickinson, San Jose, CA) were used for negative controls for flow cytometry.</p><p>Recombinant human TNF-&#x003b1;, RANTES, interferon-gamma inducible protein 10 (IP-10), and stromal cell derived factor-1&#x003b1; were purchased from R&#x00026;D Systems. Trizol reagent, deoxyribonucleaseI, and SuperScriptII reverse transcriptase were purchased from Gibco BRL (Rockville, MD). Taq polymerase was purchased from Promega (Madison, WI), and Oligo dT and ficoll/isopaque were purchased from Pharmacia (Piscataway, NJ). DMEM with high-glucose, RPMI-1640, and FBS were purchased from Gibco BRL.</p><p>Enzyme-linked immunosorbent assays (ELISAs) for MCP-1, IL-8, MIP-1&#x003b2;, and RANTES were purchased from R&#x00026;D Systems, and transwell membranes (5 &#x003bc;m pore size in 24 wells) were purchased from Costar (Cambridge, MA). The TMB microwell peroxidase substrate system was purchased from KPL (Gaithersburg, MD).</p></sec><sec><title>Stromal cell lines and fibroblast lines</title><p>One RA stromal cell line (SCL) was established from synovium as previously described [<xref ref-type="bibr" rid="B11">11</xref>]. In brief, synovial tissue of a patient with RA who met American College of Rheumatology criteria [<xref ref-type="bibr" rid="B12">12</xref>] was obtained following informed consent, and dissociated with collagenase and trypsin. Dissociated cells were cultured in DMEM supplemented with 10% FBS and 10% conditioned medium, which was prepared by incubation of peripheral blood mononuclear cells from 10 healthy donors in RPMI-1640 medium with 10% FBS for 48 h. The cultures were then maintained for more than 2 months, and the SCLs were cloned by limiting dilution. Clones were thereafter maintained and replenished with fresh DMEM medium with 10 or 20% FBS every 3&#x02013;4 days. One clone (Sy77) was used in the present experiments. RA tissues from two other patients were also dissociated with collagenase and trypsin. Dissociated cells were cultured in DMEM supplemented with 10% FBS and 10% conditioned medium and used as RA SCLs (RA6/1 and RA8/3) after 4&#x02013;10 passages. An SCL was also established from osteoarthritis synovium (OA5/26) using the same procedure. A skin fibroblast line (FBHG) was also established from a healthy human skin sample.</p></sec><sec><title>SCL and fibroblast culture</title><p>RA SCL, osteoarthritis (OA) SCL, and skin fibroblasts (1 &#x000d7; 10⁵) were seeded in six-well culture plates (Costar) with DMEM containing 10% FBS, 200 U/ml penicillinG, 10 &#x003bc;g/ml gentamicin and 0.3 mg/ml <sc>L</sc>-glutamine, and cultured for 3 days. After culturing, the cells were washed once and the medium was changed to DMEM containing 5% FBS, 200 U/ml penicillinG, 10 &#x003bc;g/ml gentamicin, and 0.3 mg/ml <sc>L</sc>-glutamine with or without 2 ng/ml TNF-&#x003b1;. To assess chemokine mRNA expression, cell lines were harvested after 4 h of culture, then suspended in Trizol and stored at -80&#x000b0;C. Supernatants were collected after 24 h and stored at -80&#x000b0;C until use for migration assay and quantification of chemokines.</p></sec><sec><title>Preparation of mononuclear cells</title><p>Peripheral blood mononuclear cells (PBMC) were isolated from heparinized blood of healthy adult volunteers by density sedimentation using ficoll/isopaque. PBMC were cultured with sheep red blood cells (SRBC) treated for 70 min at 4&#x000b0;C, and rosette negative cells were collected by ficoll/isopaque sedimentation.</p></sec><sec><title>Chemokine receptor expression by monocytes</title><p>Chemokine receptor expression by monocytes was assessed by dual immunofluorescence flow cytometry following staining with anti-CD14 mAb and anti-chemokine receptor mAbs (FACScan; Becton Dickinson). For staining chemokine receptors, 10 &#x003bc;l of each mAb was used for 1 &#x000d7; 10⁵ cells suspended in 100 &#x003bc;l staining buffer (PBS + 2% FBS) in accordance with the manufacturer's instructions.</p></sec><sec><title>Migration assay</title><p>Cell migration was assessed in 24-well chemotaxis chambers fitted with 5.0 &#x003bc;m transwell membranes. Supernatants (550 &#x003bc;l) of cell lines cultured in various conditions were added to the lower wells, and 5 &#x000d7; 10⁵ SRBC rosette negative PBMC in 100 &#x003bc;l DMEM containing 5% FBS were added to the upper wells. In blocking experiments, the supernatants were cultured with blocking mAbs specific for chemokines overnight at 4&#x000b0;C before assay. SRBC rosette negative PBMC were incubated with blocking mAb to chemokine receptors for 1 h at 4&#x000b0;C before assay in some experiments. The transwell membranes were removed after 90 min of incubation at 37&#x000b0;C, and migrated cells in the lower chamber were harvested by pipetting. The cells were stained with anti-CD14 mAb conjugated with FITC, and they were suspended in 110 &#x003bc;l PBS containing 3% FBS. All of the cells were analyzed by flow cytometry and CD14⁺ cells were counted.</p></sec><sec><title>RNA isolation and reverse transcriptase-polymerase chain reaction</title><p>RNA was extracted from SCLs and fibroblasts using Trizol reagent in accordance with the manufacturer's instructions. Two micrograms of the extracted RNA was treated with 2 U DNaseI to eliminate DNA, and reverse transcribed with 200 U SuperScriptII reverse transcriptase at 42&#x000b0;C for 70 min using oligo dT primers in accordance with the manufacturer's instructions. Polymerase chain reaction (PCR) was carried out with Taq polymerase using 0.1&#x02013;0.3 &#x003bc;l cDNA (1.5 mM MgCl₂). Denaturation and extension conditions were 94&#x000b0;C for 1 min and 72&#x000b0;C for 1 min, respectively. The annealing period was 1 min for each PCR, and the temperature was 62&#x000b0;C for RANTES and 56&#x000b0;C for the other chemokines and &#x003b2;-actin. PCR products were resolved by electrophoresis on 1.5% agarose gels and identified with ethidium bromide staining. Firstly, &#x003b2;-actin expression was examined using 26-30 cycles of reverse transcriptase-PCR (RT-PCR) to amplify 0.1, 0.2, and 0.3 &#x003bc;l cDNA to adjust the amount of cDNA of each sample precisely. After resolving the PCR products on agarose gels and identifying the relevant bands with ethidium bromide, the optimal amounts of cDNA for analysis were then determined. Chemokine expression in this amount of cDNA was examined using 30, 32, 35, 38, and 40 cycles of PCR amplification, and the results in the linear part of the amplification curve are reported in the figures.</p></sec><sec><title>PCR primers</title><p>The primer pair for &#x003b2;-actin was GTC CTC TCC CAA GTC CAC ACA (forward) and CTG GTC TCA AGT CAG TGT ACA GGT AA (reverse), that of IL-8 was CTG CGC CAA CAC AGA AAT TA (forward) and ATT GCA TCT GGC AAC CCT AC (reverse), that of MCP-1 was GCC TCC AGC ATG AAA GTC TC (forward) and TAA AAC AGG GTG TCT GGG GA (reverse), that of IP-10 was CCA CGT GTT GAG ATC ATT GC (forward) and TGG AAG ATG GGA AAG GTG AG (reverse), that of RANTES was CGC TGT CAT CCT CAT TGC TA (forward) and GCT GTC TCG AAC TCC TGA CC (reverse), that of MIP-1&#x003b1; was TGC AAC CAG TTC TCT GCA TC (forward) and ACA GGG GAA CTC TCA GAG CA (reverse), and that of MIP-1&#x003b2; was CTG GGT CCA GGA GTA CGT GT (forward) and ACA GTG GAC CAT CCC CAT AG (reverse).</p></sec><sec><title>ELISA</title><p>Concentrations of chemokines were measured by sandwich ELISA according to the manufacturer's instructions (R&#x00026;D Systems).</p></sec><sec><title>Statistical analysis</title><p>The Student paired <italic>t</italic> test was used to compare the effect of CD14 cell migration, and the blocking effect of mAbs. The Student <italic>t</italic> test was used to evaluate chemokine production by SCLs and the fibroblast line.</p></sec></sec><sec><title>Results</title><sec><title>Chemokine receptor expression by peripheral blood monocytes</title><p>Chemokine receptor expression by CD14⁺ monocytes was examined using flow cytometry. Representative examples (Fig. <xref ref-type="fig" rid="F1">1</xref>) and <xref ref-type="fig" rid="F1">a</xref> summary of staining results (Table <xref ref-type="table" rid="T1">1</xref>) are presented. CXCR3 and CXCR4 were expressed by a high frequency of monocytes in all healthy donors. More than one-half of monocytes expressed CCR2, CCR5, CXCR1, and CXCR2 in seven of nine cases, although a minimal number of monocytes from two donors expressed these receptors. Expression of CCR6 and CXCR5 by monocytes was minimal in all subjects. CCR1 was expressed by few monocytes in three donors, whereas 20-40% of monocytes of the other donors expressed CCR1. These results suggest that CCR1, CCR2, CCR5, and CXCR1, CXCR2, CXCR3, and CXCR4 are candidates to be involved in chemokine-mediated trafficking of monocytes.</p></sec><sec><title>Supernatants of cell lines attract monocytes</title><p>To examine whether SCLs derived from RA synovium can attract monocytes efficiently, migration of monocytes in response to culture supernatants of RA and OA SCLs and fibroblasts was examined (Fig. <xref ref-type="fig" rid="F2">2</xref>). Supernatants of all cell lines attracted significantly more monocytes than medium alone (<italic>P</italic> &#x0003c; 0.01). Those supernatants from RA SCLs induced the migration of significantly more monocytes than the supernatants of the OA SCL and skin fibroblasts.</p></sec><sec><title>Chemokine mRNA production by cell lines</title><p>Production of proinflammatory chemokine mRNA by the various cell lines was assessed using RT-PCR (Fig. <xref ref-type="fig" rid="F3">3</xref>). RA SCLs expressed a number of proinflammatory chemokine mRNAs, including MCP-1, IL-8, MIP-1&#x003b1;, MIP-1&#x003b2;, RANTES, and IP-10. OA SCL also expressed these same chemokine mRNAs with the exception of MIP-1&#x003b2;. The fibroblast line expressed mRNAs for IL-8, MCP-1, and IP-10, but not RANTES, MIP-1&#x003b1;, and MIP-1&#x003b2;.</p></sec><sec><title>Chemokine production by cell lines</title><p>Chemokines secreted into the culture supernatants of cell lines were measured by ELISA (Fig. <xref ref-type="fig" rid="F4">4</xref>). Production of MCP-1 and IL-8 by RA SCLs was greater than that by OA SCL and the fibroblast line. The cell lines produced low levels of RANTES, whereas IP-10 and MIP1-&#x003b2; were not detected (data not shown).</p></sec><sec><title>MCP-1 and IL-8 play a role in migration of monocytes</title><p>Blocking experiments were carried out using mAbs to determine the chemokines that contribute to monocyte migration induced by RA SCLs (Fig. <xref ref-type="fig" rid="F5">5</xref>). Anti-MCP-1 and anti-IL-8 mAbs decreased monocyte migration significantly, with the effect of anti-MCP-1 being consistently greater than that of anti-IL-8. Anti-CCR5 blocking mAb did not effect monocyte migration. The anti-CCR5 mAb did have blocking activity, however, since migration of T cells induced by RANTES (500 ng/ml) was blocked by 95% (data not shown).</p></sec><sec><title>Stimulation with TNF-&#x003b1; induces RA SCL to produce more chemokines and attract more monocytes</title><p>RA SCLs were stimulated with TNF-&#x003b1;, and chemokine production and the capacity of supernatants to influence monocyte migration were examined to assess the influence of inflammatory cytokines on chemokine production and monocyte trafficking. Supernatants from RA SCLs stimulated with 2 ng/ml TNF-&#x003b1; produced more MCP-1, IL-8, RANTES and IP-10 (Fig. <xref ref-type="fig" rid="F6">6</xref>), and attracted more monocytes (Fig. <xref ref-type="fig" rid="F7">7</xref>) than those that were unstimulated.</p></sec><sec><title>MCP-1 and IL-8 play major roles in monocyte migration by RA SCLs after TNF-&#x003b1; stimulation</title><p>Anti-MCP-1 mAb and anti-IL-8 mAb significantly inhibited monocyte migration by supernatants of RA SCLs stimulated with TNF-&#x003b1;. Anti-IP-10 mAb and anti-CCR5 mAb did not inhibit monocyte migration significantly (Fig. <xref ref-type="fig" rid="F8">8</xref>).</p></sec></sec><sec><title>Discussion</title><p>The results of this study indicate that supernatants of SCL derived from RA synovial tissue can attract more monocytes from peripheral blood than OA SCL and skin fibroblasts. MCP-1 and, to a lesser degree, IL-8 played the major roles in SCL-induced trafficking of monocytes. These results begin to provide an explanation for the extensive accumulation of myeloid cells in rheumatoid synovium. Importantly, RA SCL induced more monocyte migration after stimulation with TNF-&#x003b1;, one of the major inflammatory cytokines produced in the rheumatoid synovium.</p><p>Peripheral blood monocytes express many chemokine receptors [<xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref>], including CCR1, CCR2, CCR5, CCR8, and CXCR4. The current analysis indicates that, along with these receptors, CXCR1, CXCR2, and CXCR3 are also expressed by monocytes. CCR1, CCR2, CCR5, CXCR1, CXCR2, CXCR3, and CXCR4 notably appear to be expressed by monocytes obtained from most donors.</p><p>Ligands of these chemokine receptors are candidates to be involved in monocyte trafficking. These ligands would include RANTES, MIP-1&#x003b1; and MCP-3 (CCR1), MCP-1, MCP-2, MCP-3, MCP-4, MCP-5 (CCR2), RANTES, MIP-1&#x003b1;, and MIP-1&#x003b2; (CCR5), IL-8 and CGP-2 (CXCR1), IL-8, GRO&#x003b1;, and epithelial-derived neutrophil attractant 78 (CXCR2), IP-10 and MIG (CXCR3), and stromal cell derived factor-1 (CXCR4) [<xref ref-type="bibr" rid="B13">13</xref>]. Migration assays using supernatants of RA SCLs were employed to determine which of these chemokine and receptor interactions might be involved in monocyte migration in the RA synovium.</p><p>Stromal cells are one of the important cell populations in RA synovium. SCL can produce many cytokines, chemokines, and can promote viability and functional activation of T cells and B cells [<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>]. The current findings also indicate that supernatants of RA SCLs attract more monocytes than the supernatants of OA SCL and skin fibroblasts. This result suggests that the supernatants from RA SCLs may contain more chemokines than or different chemokines to those of OA SCL or fibroblast lines and, therefore, are able to attract additional monocytes. This is likely to contribute to the more marked accumulation of monocytes in RA compared with OA synovium.</p><p>An analysis of chemokines produced by RA SCL indicated that these cells expressed MCP-1 and IL-8 mRNAs, and also secreted MCP-1 and IL-8. It is noteworthy that there were some discrepancies between mRNA expression and protein production. From the mRNA expression, we expected RA SCLs to produce reasonable amounts of IP-10, RANTES, MIP-1&#x003b1;, and MIP-1&#x003b2;. RA SCLs, however, secreted only small amounts of RANTES, and IP-10 and MIP-1&#x003b2; secretion was not detected. According to previous reports, mRNAs of MCP-1, IL-8, and RANTES were found in RA SCL or fibroblasts, but only MCP-1 was reported to be secreted by RA SCL or fibroblasts without cytokine stimulation [<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B17">17</xref>,<xref ref-type="bibr" rid="B18">18</xref>]. The current study indicates a wider profile of chemokine expression and secretion by RA SCL, and also a greater production of IL-8 and MCP-1 by RA SCL compared with OA SCL, implying a greater proinflammatory potential.</p><p>Experiments with blocking mAbs were carried out to examine the chemokines in the supernatants of SCL that accounted for monocyte migration. The data were consistent with the conclusion that MCP-1 and, to a lesser extent, IL-8 accounted for the capacity of RA SCL supernatants to stimulate monocyte migration. Whether additional cytokines produced by RA SCL also contributed to monocyte migration is currently not known, but the combination of the chemokine data and the mAb blocking results suggests that MCP-1 and IL-8 play a dominant role in monocyte migration measured by SCL.</p><p>TNF-&#x003b1; is one of the major cytokines produced in inflammatory sites such as RA synovium, and it is thought to play a central proinflammatory role [<xref ref-type="bibr" rid="B2">2</xref>]. We examined the influence of TNF-&#x003b1; on monocyte migration induced by RA SCL. RA SCLs stimulated by TNF-&#x003b1; secreted more MCP-1, IL-8, and RANTES than those that were unstimulated, and began to produce IP-10. Moreover, supernatants of TNF-&#x003b1;-stimulated RA SCLs attracted more monocytes in migration assay. MCP-1 and IL-8 played the main roles in monocyte migration induced by TNF-&#x003b1;-stimulated RA-SCL, as documented by blocking experiments.</p><p>It has recently been reported that T cells expressing CCR5 gather at inflammatory sites such as RA or multiple sclerosis, and that interactions between CCR5 and its ligands (RANTES, MIP-1&#x003b1;, MIP-1&#x003b2;) are thought to be important in the accumulation of inflammatory cells at these sites [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B21">21</xref>]. CCR5 expressed by monocytes, however, was not apparently active in transmitting transmigratory signals. RANTES is also the ligand of CCR1, but the major receptor for RANTES is thought to be CCR5. In this regard, expression of CCR1 by monocytes was lower than that of CCR5. Moreover, antibody to MCP-1 and IL-8 blocked 75% and 20% of migration of monocytes, respectively. According to this information, RANTES is unlikely to play a major role in monocyte migration by RA SCL supernatants.</p><p>MCP-1 has been shown to play an important role in the development of arthritis in MRL-lpr mice [<xref ref-type="bibr" rid="B22">22</xref>], whereas the migration of monocytes into inflammatory sites was reduced in the CCR2-deficient mouse [<xref ref-type="bibr" rid="B23">23</xref>]. These results are consistent with the current findings, demonstrating the important role of MCP-1 in monocyte migration. Blocking the effect of RANTES has, however, also been shown to be effective in ameliorating collagen-induced arthritis in DAB/1 mice and adjuvant-induced arthritis in rats [<xref ref-type="bibr" rid="B24">24</xref>,<xref ref-type="bibr" rid="B25">25</xref>]. RANTES might, therefore, play a different role in the initiation of inflammatory arthritis in experimental animals compared with propagating chronic inflammation in RA.</p><p>The overlapping roles of MCP-1 and IL-8 in the migration of inflammatory cells have recently been highlighted in studies of neutrophil trafficking. It was reported that MCP-1 plays a role in neutrophil trafficking in inflammation, although neutrophil trafficking was previously believed to be regulated only by IL-8, and not by MCP-1 [<xref ref-type="bibr" rid="B26">26</xref>]. These previous results, together with the current data, indicate that both MCP-1 and IL-8 play major roles in regulating the trafficking of myeloid cells into inflammatory sites.</p><p>The role of specific chemokines in arthritis is still controversial, and further investigation is necessary to delineate to specific roles of these effector molecules. Moreover, the role of tissue cells in regulating the migration of inflammatory cells into the synovium is also not fully established. The current data, however, strongly imply that monocyte accumulation in rheumatoid synovium is regulated by SCL via production of the chemokines MCP-1 and IL-8.</p></sec><sec><title>Abbreviations</title><p>ELISA = enzyme-linked immunosorbent assay; IP-10 = interferon-gamma inducible protein 10; mAb = monoclonal antibody; MCP-1 = monocyte chemotactic protein-1; MIP = macrophage inflammatory protein; OA = osteoarthritis; PBMC = peripheral blood mononuclear cells; PE = phycoerythrin; RA = rheumatoid arthritis; RANTES = regulated upon activation, normal T cell expressed and secreted; RT-PCR = reverse transcriptase-polymerase chain reaction; SCL = stromal cell line; SRBC = sheep red blood cells; TNF-&#x003b1; = tumor necrosis factor-&#x003b1;.</p></sec>

Phenotypic characteristics of human monocytes undergoing transendothelial migration

<p>In our study we characterised the immunophenotype of monocytes that migrated through an endothelial cell (EC) monolayer <italic>in vitro</italic>. We found that monocyte migration led to an enhanced expression of CD11a, CD33, CD45RO, CD54 [intercellular cell-adhesion molecule (ICAM)-1] and human leucocyte antigen-DR. The most striking increase was observed for ICAM-1 when ECs were activated with tumour necrosis factor-&#x003b1; and interleukin-1&#x003b1;. The results of our study indicate the following: (1) there is a characteristic immunophenotype on the surface of monocytes after transendothelial migration; (2) this phenotype seems to be induced by interactions between monocytes and ECs; and (3) this change is enhanced by the pretreatment of ECs with cytokines. Taken together, the results suggest that local cytokine production activating ECs is sufficient to enhance monocyte migration and that this, in turn, can induce changes consistent with an activated phenotype known to be interactive between antigen-presenting cells and T cells. These results have implications for our pathogenetic insights into rheumatoid arthritis.</p>

<contrib id="A1" corresp="yes" contrib-type="author"><name><surname>Grisar</surname><given-names>Johannes</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>johannes.grisar@akh-wien.ac.at</email></contrib><contrib id="A2" contrib-type="author"><name><surname>Hahn</surname><given-names>Philipp</given-names></name><xref ref-type="aff" rid="I2">2</xref></contrib><contrib id="A3" contrib-type="author"><name><surname>Brosch</surname><given-names>Susanne</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A4" contrib-type="author"><name><surname>Peterlik</surname><given-names>Meinrad</given-names></name><xref ref-type="aff" rid="I3">3</xref></contrib><contrib id="A5" contrib-type="author"><name><surname>Smolen</surname><given-names>Josef S</given-names></name><xref ref-type="aff" rid="I1">1</xref></contrib><contrib id="A6" contrib-type="author"><name><surname>Pietschmann</surname><given-names>Peter</given-names></name><xref ref-type="aff" rid="I1">1</xref><xref ref-type="aff" rid="I2">2</xref><xref ref-type="aff" rid="I3">3</xref></contrib>

Arthritis Research

<sec><title>Synopsis</title><sec><title>Introduction:</title><p>Acute and chronic inflammation are characterised by the enhanced migration of leucocytes from the blood vessels through the endothelium into the extravascular tissue. There are ample data on the transendothelial migration of lymphocytes [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. In addition, T cells capable of transendothelial migration have a characteristic immunophenotype; for instance, migrating T cells express significantly higher amounts of CD29 (&#x003b2;₁-integrin) than cells that do not interact with endothelial cells (ECs) [<xref ref-type="bibr" rid="B7">7</xref>]. In contrast, much less is known about the extravasation of cells of the monocyte/macrophage lineage. Under normal conditions, only a few monocytes are able to migrate from the bloodstream into healthy tissue, where they differentiate into macrophages [<xref ref-type="bibr" rid="B8">8</xref>]. In inflammatory diseases such as rheumatoid arthritis (RA), monocytes accumulate in the synovial membrane and contribute significantly to the pathogenesis of the disease, mainly by secreting cytokines such as tumour necrosis factor-&#x003b1; (TNF-&#x003b1;) and interleukin-1 (IL-1) [<xref ref-type="bibr" rid="B9">9</xref>]. Transendothelial migration of monocytes might also be an important step in the pathogenesis of non-inflammatory diseases, for example atherosclerosis [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>].</p><p>Several pairs of receptors and counter-receptors that mediate the interaction of monocytes with ECs have been described [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>], but the phenotype of monocytes capable of transendothelial migration was not defined. Previously we found an enhanced expression of CD54 [intercellular cell-adhesion molecule (ICAM)-1] on monocytes that had migrated through an unstimulated EC monolayer [<xref ref-type="bibr" rid="B18">18</xref>].</p></sec><sec><title>Aims:</title><p>In our study we characterised the immunophenotype of monocytes that migrated through an EC monolayer in an <italic>in vitro</italic> model. We found that monocyte migration led to an enhanced expression of CD11a, CD33, CD45RO, CD54 (ICAM-1) and human leucocyte antigen (HLA)-DR. The most striking increase was observed for CD54 when ECs were activated with TNF-&#x003b1; and IL-1&#x003b1;. Our findings of increased CD54 expression on monocytes that migrated through endothelium pretreated with TNF-&#x003b1; and IL-1&#x003b1; might be useful in elucidating the mode of action of therapy with antibodies against TNF-&#x003b1; and IL-1 in RA and thus might lead to new avenues of RA therapy in future.</p></sec><sec><title>Methods:</title><p>ECs were isolated from human umbilical cord veins by digestion with collagenase as described previously [<xref ref-type="bibr" rid="B7">7</xref>] and then cultured. ECs in the third to fifth passage were used. Peripheral blood mononuclear cells (PBMCs) were prepared from buffy coats of healthy blood donors by centrifugation over gradients of Ficoll-Hypaque. PBMCs were prepared immediately before starting the experiments.</p><p>Interactions with ECs in PBMC populations were examined on hydrated bovine collagen gels in the wells of 16 mm macrowell tissue culture plates, as described previously [<xref ref-type="bibr" rid="B7">7</xref>]. To form a confluent monolayer on the collagen gels, 5 &#x000d7;10⁵ ECs per well were incubated overnight.</p><p>To measure monocyte interaction with ECs, PBMCs (3 &#x000d7;10⁶) were resuspended in fresh culture medium, layered on top of collagen gels with and without ECs and incubated at 37&#x000b0;C. The range of the incubation period was 15 minutes to 24 hours. Nonadherent cells (NAD) were harvested by washing twice with culture medium. Cells bound to the surface (BND) were enriched by washing each well twice with warm (37&#x000b0;C) Puck's EDTA, twice with warm (37&#x000b0;C) EGTA [0.5 mM EGTA in phosphate-buffered saline (PBS)] and once with cold (4&#x000b0;C) Puck's EDTA. Finally, for the recovery of those cells that had migrated into the collagen gels (MIG), 0.7 ml of a solution containing 0.1% collagenase, 1% (v/v) fetal calf serum and 50 mM Hepes buffer was added per well. The collagen gels were then minced gently with a pipette and incubated for 60 minutes at 37&#x000b0;C, after which the migrated PBMCs were removed by washing the wells twice with PBS. Each population (NAD, BND and MIG) was washed, resuspended in culture medium and counted under a microscope. In some experiments we studied monocyte migration into plain collagen gels. In these experiments no ECs were layered on the collagen gels; in other respects the experiments were performed exactly as described above.</p><p>In some experiments the EC monolayer was preactivated by incubation with TNF-&#x003b1;, IL-1&#x003b1;, macrophage inflammatory protein (MIP)-1&#x003b1; or interferon-&#x003b3; (IFN-&#x003b3;). To this end, the medium in each well was removed and the ECs were incubated for 5 hours at 37&#x000b0;C with or without the respective cytokines or chemokines (100 IU/ml TNF-&#x003b1;, IL-1&#x003b1; or IFN-&#x003b3;, or 50 ng/ml MIP-1&#x003b1;). After this 5-hour pretreatment, each well was washed thoroughly and the migration assay was performed as described above.</p><p>Staining of monocytes was performed with fluorescein isothiocyanate (FITC)-conjugated antibodies against CD11a, CD33, CD45RA, CD45RO, CD49d (&#x003b1;₄-integrin), CD54 (ICAM-1), CD86, HLA-DR, CD45RB and CD62L (L-selectin) (functions and ligands of these surface markers are shown in Table <xref ref-type="table" rid="T1">1</xref>). To distinguish monocytes from other immune cells, all samples were counterstained with a phycoerythrin-labelled anti-CD14 monoclonal antibody. Cells (3 &#x000d7; 10⁵ to 4 &#x000d7; 10⁵ per sample) were incubated at 4&#x000b0;C for 30 minutes. Cells were then pelleted and resuspended in 250 &#x003bc;l of PBS before analysis was performed on a flow cytometer. All results are expressed as the respective mean fluorescence intensity among CD14-positive cells. Because not only monocytes but also ECs express CD54 (ICAM-1), in the analyses of the expression of CD54 on monocytes by fluorescence-activated cell sorting, monocytes were defined by both the scatter profile and the expression of CD14. In addition, ECs, which are considerably larger, were excluded by size.</p><p>All data are presented as means &#x000b1; SD. Paired Student's <italic>t</italic>-tests were used for comparisons.</p></sec><sec><title>Results:</title><p>In initial experiments we studied the time course of PBMC migration into plain or EC-coated collagen gels, respectively. As shown in Fig. <xref ref-type="fig" rid="F1">1</xref>, the presence of an endothelium clearly facilitated the migration of PBMCs: after 30 minutes the percentage of PBMCs that had migrated was twice as high as that in the absence of ECs. After 2 hours, about 40% of the PBMCs could be recovered from collagen gels coated with an EC layer, whereas only 24% of PBMCs had migrated into plain collagen gels. Prolonging the incubation time to 24 hours allowed further migration of PBMCs only in the absence of ECs, but did not significantly increase the extent of EC-mediated migration.</p><p>The results and the statistical evaluation of the phenotypic analysis of monocytes recovered in various fractions of the migration assay are shown in Table <xref ref-type="table" rid="T2">2</xref>. The expression of CD11a, CD33, CD45RO, CD54 and HLA-DR was significantly higher in MIG than in NAD. When compared with BND, these markers, and also CD45RB and CD62L, were significantly elevated in MIG. NAD, BND and MIG were incubated with collagenase for the same durations to control for possible cell activation by the collagenase treatment; the expression of adhesion molecules was similar to that on untreated cells.</p><p> We also studied the capacity for monocyte migration into plain collagen gels, that is, in the absence of an endothelium. No significant difference in surface marker expression was observed when migrated monocytes were compared with any other fraction.</p><p>It has been reported that cytokines such as TNF-&#x003b1;, IL-1&#x003b1; and IFN-&#x003b3; and also the chemokine MIP-1&#x003b1; can enhance the transendothelial migration of monocytes [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. We were therefore interested to investigate whether pretreatment of the ECs with these factors would be sufficient to enhance the transendothelial migration of monocytes and/or to induce changes in their expression of surface markers. Figure <xref ref-type="fig" rid="F2">2</xref> shows that pretreatment of the endothelium with any of the cytokines led to a consistent and significant increase in the number of migrated mononuclear cells in comparison with simultaneously performed control experiments in which the endothelium was not pretreated. Pretreatment with IL-1&#x003b1; was the most effective, resulting in a 132% increase in migrated cells (<italic>P</italic> &#x0003c; 0.001). The respective values for the other cytokines were as follows: MIP-1&#x003b1;, 194% (<italic>P</italic> = 0.043); TNF-&#x003b1;, 193% (<italic>P</italic> = 0.006); IFN-&#x003b3;, 136% (<italic>P</italic> = 0.016).</p><p>Pretreatment of ECs with TNF-&#x003b1; led to a significant decrease in CD45RO and HLA-DR on migrated monocytes. In contrast, CD54 (ICAM-1) was significantly increased on monocytes that migrated through endothelium pretreated with TNF-&#x003b1; or IL-1&#x003b1; in comparison with migration through untreated endothelium (Figs <xref ref-type="fig" rid="F3">3</xref> and <xref ref-type="fig" rid="F4">4</xref>, Table <xref ref-type="table" rid="T3">3</xref>).</p><p>When ECs were pretreated with IFN-&#x003b3; or MIP-1&#x003b1;, no statistically significant change in monocyte surface markers was observed (Table <xref ref-type="table" rid="T3">3</xref>).</p></sec><sec><title>Discussion:</title><p>Our experiments characterised the immuno-phenotype of monocytes that migrated through EC monolayers. Several surface molecules (CD11a, CD33, CD45RO, ICAM-1 and HLA-DR) were significantly increased on the migrated monocytes in comparison with the nonadherent population. The differences in the immunophenotype between migrating and nonadherent monocytes could be explained either by the preferential migration of a particular subset of monocytes or as a result of the interaction of monocytes with the endothelium. If the observed alterations in the immunophenotype of migrating monocytes had been due to the preferential migration of a particular subset, we would have expected a depletion of this subset in the nonadherent population. This did not occur; all surface markers studied were expressed to similar extents on the nonadherent and initial populations (see Table <xref ref-type="table" rid="T2">2</xref>). The notion that the process of transendothelial migration leads to an upregulation of certain surface markers on monocytes is also supported by the fact that the expression of most markers was significantly higher in the migrated cells than in the bound cells (see Table <xref ref-type="table" rid="T2">2</xref>). Taken together, our results suggest that transendothelial migration induces the activation or maturation of monocytes.</p><p>As described previously [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>], we found that pretreatment of the endothelium with certain chemokines or cytokines enhanced transendothelial migration. The most striking phenotypic changes were seen for ICAM-1 expression, when ECs were pretreated with TNF-&#x003b1; and IL-1&#x003b1; (see Table <xref ref-type="table" rid="T3">3</xref> and Figs <xref ref-type="fig" rid="F3">3</xref> and <xref ref-type="fig" rid="F4">4</xref>). In contrast, MIP-1&#x003b1; pretreatment did not change the monocyte phenotype investigated here. In the light of enhanced migration through MIP-1&#x003b1;-prestimulated endothelium, these results suggest a dichotomy of cytokine/chemokine effects on migration compared with surface marker expression: ECs activated with TNF-&#x003b1; and IL-1&#x003b1; seem to lead to an upregulation of both monocyte migration and surface marker expression, whereas MIP-1&#x003b1; only enhances migration, a finding that is compatible with the chemotactic chemokine nature of MIP-1&#x003b1;. Alternatively, MIP-1&#x003b1; could have been trapped in the collagen gel, acting as a chemotactic gradient directly on monocytes rather than via ECs. Thus TNF-&#x003b1; and IL-1&#x003b1; seem to mediate different proinflammatory events from those mediated by MIP-1&#x003b1;.</p><p>Our observations of increased expression of ICAM-1 on migrated monocytes after the pretreatment of ECs with TNF-&#x003b1; and IL-1&#x003b1; are especially remarkable because these cytokines are important in the pathogenesis of inflammation. In RA, TNF-&#x003b1; and IL-1 blockade showed an unequivocal therapeutic effect [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>]. In addition, ICAM-1 and E-selectin levels of patients with RA who had received anti-TNF-&#x003b1; therapy decreased within a few days of the initiation of therapy [<xref ref-type="bibr" rid="B24">24</xref>].</p><p>Previous findings indicate that TNF-&#x003b1; and IL-1&#x003b1; induce an upregulation of the ICAM-1 counter-receptor on ECs [<xref ref-type="bibr" rid="B25">25</xref>]. This is consistent with the increase in cell migration found in our experiments and the altered expression of ICAM-1 on monocytes. Because the classic ICAM-1 counter-receptors LFA-1 and Mac-1 have not been detected on ECs, the existence of another ICAM-1 ligand, one that facilitates the transendothelial migration of monocytes, remains possible.</p><p>The reported changes indicate that, after migration, monocytes could become more liable to interact with T cells (which are known to enhance LFA-1 expression in the presence of TNF-&#x003b1;). This interaction might lead to a further mutual stimulation of T cells and macrophages. In fact the ligand-counterligand system consisting of LFA-1 and ICAM-1 also is one co-stimulatory pathway involved in interactions between antigen-presenting cells (APCs) and T cells [<xref ref-type="bibr" rid="B26">26</xref>]. Because our results demonstrate that both ICAM-1 and HLA-DR are upregulated on migrated monocytes, their function as APCs and possible ability to communicate with T cells, might be facilitated after transendothelial migration. Thus, this observation also supports previous notions on the importance of T cells in the pathogenesis of RA [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>].</p><p>In summary, our findings indicate that monocyte migration is accompanied by changes in function-associated surface antigens and that TNF-&#x003b1; and IL-1&#x003b1; in particular increase the number of migrating monocytes and lead to an enhanced expression of certain surface markers involved in cell-cell interactions. These events might not only be partly responsible for the high inflammatory activity in RA synovium; they also suggest that ECs have a pivotal role in these processes and thus might constitute an important therapeutic target.</p></sec></sec><sec><title>Introduction</title><p>Acute and chronic inflammation are characterised by an enhanced migration of leucocytes from the blood vessels through the endothelium into the extravascular tissue. There are ample data on the transendothelial migration of lymphocytes [<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B6">6</xref>]. In addition, T cells capable of transendothelial migration have a characteristic immunophenotype; for instance, migrating T cells express significantly higher amounts of CD29 (&#x003b2;₁-integrin) than cells that do not interact with endothelial cells (ECs) [<xref ref-type="bibr" rid="B7">7</xref>]. In contrast, much less is known about the extravasation of cells of the monocyte/macrophage lineage. Under normal conditions, only a few monocytes are able to migrate from the bloodstream into healthy tissue, where they differentiate into macrophages [<xref ref-type="bibr" rid="B8">8</xref>]. In inflammatory diseases such as rheumatoid arthritis (RA), monocytes accumulate in the synovial membrane and contribute significantly to the pathogenesis of the disease, mainly by secreting cytokines such as tumour necrosis factor-&#x003b1; (TNF-&#x003b1;) and interleukin-1 (IL-1) [<xref ref-type="bibr" rid="B9">9</xref>]. Transendothelial migration of monocytes might also be an important step in the pathogenesis of non-inflammatory diseases, for example atherosclerosis [<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref>].</p><p>Several pairs of receptors and counter-receptors that mediate the interaction of monocytes with ECs have been described [<xref ref-type="bibr" rid="B13">13</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref>], but the phenotype of monocytes capable of transendothelial migration was not defined. Previously we found an enhanced expression of CD54 [intercellular cell-adhesion molecule (ICAM)-1] on monocytes that had migrated through an unstimulated EC monolayer [<xref ref-type="bibr" rid="B18">18</xref>].</p><p>It was the aim of the present study to perform a detailed characterisation of the immunophenotype of transendothelially migrated monocytes.</p></sec><sec sec-type="materials|methods"><title>Materials and methods</title><sec><title>Cell cultures</title><p>ECs were isolated from human umbilical cord veins by digestion with collagenase as described previously [<xref ref-type="bibr" rid="B7">7</xref>]. The culture medium consisted of MCDB-M 104 glutamine (Gibco, Paisley, UK) supplemented with 20% (v/v) fetal calf serum (FCS), 24 &#x003bc;g/ml EC growth supplement (TC Laevosan, Vienna, Austria), 50 IU/ml heparin, 2 mM L-glutamine (Gibco), penicillin (100 IU/ml; Gibco) and streptomycin (100 &#x003bc;g/ml; Gibco). ECs in the third to fifth passage were used.</p></sec><sec><title>Preparation of peripheral blood mononuclear cells</title><p>Peripheral blood mononuclear cells (PBMCs) were prepared from buffy coats of healthy blood donors by centrifugation over gradients of Ficoll-Hypaque (Histopaque R, Sigma, Vienna, Austria). PBMCs were prepared immediately before starting the experiments.</p></sec><sec><title>Monocyte-EC binding and transendothelial migration</title><p>Interactions with ECs in PBMC populations were examined on hydrated bovine collagen gels in the wells of 16 mm macrowell tissue culture plates, as described previously [<xref ref-type="bibr" rid="B7">7</xref>]. In brief, the collagen gels were made of 50% (v/v) bovine collagen (Vitrogen 100; Collagen Biomaterials, Palo Alto, California, USA), 7% (v/v) NaOH, 10% (v/v) 10 &#x000d7; phosphate-buffered saline (PBS; Gibco) and 33% (v/v) distilled water. To form a confluent monolayer on the collagen gels, 5 &#x000d7; 10⁵ ECs per well were incubated overnight.</p><p>To measure monocyte interaction with ECs, PBMCs (3 &#x000d7; 10⁶) were resuspended in fresh culture medium, layered on top of collagen gels with and without ECs and incubated at 37&#x000b0;C. The range of the incubation period was 15 minutes to 24 hours. Nonadherent cells (NAD) were harvested by washing twice with culture medium. Cells bound to the surface (BND) were enriched by washing each well twice with warm (37&#x000b0;C) Puck's EDTA, twice with warm (37&#x000b0;C) EGTA (0.5 mM EGTA in PBS) and once with cold (4&#x000b0;C) Puck's EDTA. Finally, for the recovery of those cells that had migrated into the collagen gels (MIG), 0.7 ml of a solution containing 0.1% collagenase (Sigma), 1% (v/v) FCS and 50mM Hepes buffer (Gibco) was added per well. The collagen gels were then gently minced with a pipette and incubated for 60 minutes at 37&#x000b0;C, after which the migrated PBMCs were removed by washing the wells twice with PBS. Each population (NAD, BND and MIG) was washed, resuspended in culture medium and counted by microscope. In some experiments monocyte we studied migration into plain collagen gels. In these experiments no ECs were layered on the collagen gels; in other respects the experiments were performed exactly as described above.</p></sec><sec><title>Pretreatment of ECs</title><p>In some experiments the EC monolayer was preactivated by incubation with TNF-&#x003b1; (Pharma Biotechnology, Hannover, Germany), IL-1&#x003b1;, macrophage inflammatory protein-1&#x003b1; (MIP-1&#x003b1;) or interferon-&#x003b3; (IFN-&#x003b3;) (all purchased from Serotec, Oxford, UK). To this end, the medium in each well was removed and the ECs were incubated for 5 hours at 37&#x000b0;C with or without the respective cytokines or chemokines (100 IU/ml TNF-&#x003b1;, IL-1&#x003b1; or IFN-&#x003b3;, or 50 ng/ml MIP-1&#x003b1;). After this 5-hour pretreatment, each well was washed thoroughly and the migration assay was performed as described above.</p></sec><sec><title>Analysis of monocyte surface markers by dual colour flow cytometry</title><p>Staining of monocytes was performed with fluorescein isothiocyanate (FITC)-conjugated antibodies against CD11a, CD33, CD45RA, CD45RO, CD49d (&#x003b1;₄-integrin), CD54 (ICAM-1), CD86, HLA-DR (all purchased from Serotec), CD45RB (Dako, Glostrup, Denmark) and CD62L (L-selectin) (Becton-Dickinson, San Jose, California, USA) (functions and ligands of these surface markers are shown in Table <xref ref-type="table" rid="T1">1</xref>). To distinguish monocytes from other immune cells, all samples were counterstained with a phycoerythrin-labelled anti-CD14 monoclonal antibody. Cells (3 &#x000d7; 10⁵ to 4 &#x000d7; 10⁵ per sample) were incubated at 4&#x000b0;C for 30 minutes. Cells were then pelleted and resus-pended in 250 &#x003bc;l of PBS before analysis was performed on a flow cytometer (FACScan; Becton-Dickinson). All results are expressed as the respective mean fluorescence intensity among CD14-positive cells. Because not only monocytes but also ECs express CD54 (ICAM-1), in the analyses of the expression of CD54 on monocytes by fluorescence-activated cell sorting, monocytes were defined by both the scatter profile and the expression of CD14. In addition, ECs, which are considerably larger, were excluded by size.</p></sec><sec><title>Statistics</title><p>All data are presented as means &#x000b1; SD. Paired Student's <italic>t</italic>-tests were used for comparisons.</p></sec></sec><sec><title>Results</title><sec><title>Transendothelial migration of PBMCs</title><p>In initial experiments we studied the time course of PBMC migration into plain or EC-coated collagen gels, respectively. As shown in Fig. <xref ref-type="fig" rid="F1">1</xref>, the presence of an endothelium clearly facilitated the migration of PBMCs: after 30 minutes the percentage of PBMCs that had migrated was twice as high as that in the absence of ECs. After 2 hours, about 40% of the PBMCs could be recovered from collagen gels coated with an EC layer, whereas only 24% PBMCs had migrated into plain collagen gels. Prolonging the incubation time to 24 hours allowed further migration of PBMCs only in the absence of ECs, but did not significantly increase the extent of EC-mediated migration.</p></sec><sec><title>Phenotypic analysis of monocytes capable of transendothelial migration</title><p>The results and the statistical evaluation of the phenotypic analysis of monocytes recovered in various fractions of the migration assay are shown in Table <xref ref-type="table" rid="T2">2</xref>. The expression of CD11a, CD33, CD45RO, CD54 and HLA-DR was significantly higher in MIG than in NAD. When compared with BND, these markers, and also CD45RB and CD62L, were significantly elevated in MIG. NAD, BND and MIG were incubated with collagenase for the same durations to control for possible cell activation by the collagenase treatment; the expression of adhesion molecules was similar to that on untreated cells.</p></sec><sec><title>Phenotypic analysis of monocytes migrated into plain collagen gels</title><p>We also studied the capacity for monocyte migration into plain collagen gels, that is, in the absence of an endothelium. No significant difference in surface marker expression was observed when migrated monocytes were compared with any other fraction.</p></sec><sec><title>Effect of pretreatment of ECs on the transendothelial migration of monocytes</title><p>It has been reported that cytokines such as TNF-&#x003b1;, IL-1&#x003b1; and IFN-&#x003b3; and also the chemokine MIP-1&#x003b1; can enhance the transendothelial migration of monocytes [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>]. We were therefore interested to investigate whether pretreat-ment of the ECs with these factors would be sufficient to enhance the transendothelial migration of monocytes and/or to induce changes in their expression of surface markers. Figure <xref ref-type="fig" rid="F2">2</xref> shows that pretreatment of the endothelium with any of the cytokines led to a consistent and significant increase in the number of migrated mononuclear cells in comparison with simultaneously performed control experiments in which the endothelium was not pretreated. Pretreatment with IL-1&#x003b1; was the most effective, resulting in a 132% increase in migrated cells (<italic>P</italic> &#x0003c; 0.001). The respective values for the other cytokines were as folows: MIP-1&#x003b1;, 194% (<italic>P</italic> = 0.043); TNF-&#x003b1;, 193% (<italic>P</italic> = 0.006); IFN-&#x003b3;, 136% (<italic>P</italic> = 0.016).</p></sec><sec><title>Effect of pretreatment of ECs on the immunophenotype of migrated monocytes</title><p>Pretreatment of ECs with TNF-&#x003b1; led to a significant decrease in CD45RO and HLA-DR on migrated monocytes. In contrast, CD54 (ICAM-1) was significantly increased on monocytes that migrated through endothelium pretreated with TNF-&#x003b1; or IL-1&#x003b1; in comparison with migration through untreated endothelium (Figs <xref ref-type="fig" rid="F3">3</xref> and <xref ref-type="fig" rid="F4">4</xref>, Table <xref ref-type="table" rid="T3">3</xref>).</p><p>When ECs were pretreated with IFN-&#x003b3; or MIP-1&#x003b1;, no statistically significant change in monocyte surface markers was observed (Table <xref ref-type="table" rid="T3">3</xref>).</p></sec></sec><sec><title>Discussion</title><p>Our experiments characterised the immunophenotype of monocytes that migrated through EC monolayers. Several surface molecules (CD11a, CD33, CD45RO, ICAM-1 and HLA-DR) were significantly increased on the migrated monocytes in comparison with the nonadherent population. The differences in the immunophenotype between migrating and nonadherent monocytes could be explained either by the preferential migration of a particular subset of monocytes or as a result of the interaction of monocytes with the endothelium. If the observed alterations in the immunophenotype of migrating monocytes had been due to the preferential migration of a particular subset, we would have expected a depletion of this subset in the non-adherent population. This did not occur; all surface markers studied were expressed to similar extents on the nonadherent and initial populations (see Table <xref ref-type="table" rid="T2">2</xref>). The notion that the process of transendothelial migration leads to an upregulation of certain surface markers on monocytes is also supported by the fact that the expression of most markers was significantly higher in the migrated cells than in the bound cells (see Table <xref ref-type="table" rid="T2">2</xref>). Taken together, our results suggest that transendothelial migration induces the activation or maturation of monocytes.</p><p>The fact that monocytes that migrated into collagen gels without an endothelium failed to change their immunophenotype indicates that monocyte differentiation might be signalled not by the migration step itself or by the presence of collagen but by interactions between monocytes and ECs. Additional control experiments proved that collagenase itself did not lead to any of the observed changes. Recent observations with lymphocytes indicate that at least some of the changes could be due to the transfer of surface molecules from ECs to monocytes [<xref ref-type="bibr" rid="B30">30</xref>]. It is not known at present whether this is also true of monocytes. For T cells a preferential migration of a CD45RO-positive subset is already known [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B7">7</xref>]. Here we found an upregulation of CD45RO in the whole population of migrated monocytes; the significance of this finding is not clear and also merits further elucidation.</p><p>CD11a on monocytes, and its ligand ICAM-1 on ECs, are known to be important for adhesion in monocyte migration [<xref ref-type="bibr" rid="B31">31</xref>]; our observation of upregulated CD11a on monocytes after transendothelial migration is in line with these data.</p><p>Audran <italic>et al</italic> [<xref ref-type="bibr" rid="B32">32</xref>] compared the difference in the expression of certain adhesion molecules between monocytes and differentiated macrophages. They found that ICAM-1 increased during differentiation and showed stronger expression on macrophages than on monocytes. Moreover, monocytes and macrophages that were recovered from inflammatory sites, such as from the synovial fluid of patients with RA [<xref ref-type="bibr" rid="B33">33</xref>], from bronchial biopsies of patients with asthma [<xref ref-type="bibr" rid="B34">34</xref>] or from peritoneal fluid of patients with peritonitis [<xref ref-type="bibr" rid="B35">35</xref>], had a high expression of ICAM-1. These findings and our results therefore suggest that the immunophenotype of macrophages or monocytes recovered from inflammatory lesions is determined, at least in part, by the process of transendothelial migration.</p><p>As described previously [<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>], we found that pretreatment of the endothelium with certain chemokines or cytokines enhanced transendothelial migration. The most striking phenotypic changes were seen for ICAM-1 expression, when ECs were pretreated with TNF-&#x003b1; and IL-1&#x003b1; (see Table <xref ref-type="table" rid="T3">3</xref> and Figs <xref ref-type="fig" rid="F3">3</xref> and <xref ref-type="fig" rid="F4">4</xref>). In contrast, MIP-1&#x003b1; pretreatment did not change the monocyte phenotype investigated here. In the light of enhanced migration through MIP-1&#x003b1;-prestimulated endothelium, these results suggest a dichotomy of cytokine/chemokine effects on migration compared with surface marker expression: ECs activated with TNF-&#x003b1; and IL-1&#x003b1; seem to lead to an upregulation of both monocyte migration and surface marker expression, whereas MIP-1&#x003b1; only enhances migration, a finding that is compatible with the chemotactic chemokine nature of MIP-1&#x003b1;. Alternatively, MIP-1&#x003b1; could have been trapped in the collagen gel, acting as a chemotactic gradient directly on monocytes rather than via ECs. Thus TNF-&#x003b1; and IL-1&#x003b1; seem to mediate different proinflammatory events from those mediated by MIP-1&#x003b1;.</p><p>Generally, the influence of ECs on monocytes could be explained in two ways: either such signals are provided by cell-cell interaction via pairs of receptors and counter-receptors during the process of migration, or, alternatively, ECs could secrete chemoattractants. However, in the latter case we would instead expect similar effects on the bound subpopulation as well as a higher percentage of EC-bound monocytes.</p><p>Our observations of increased expression of ICAM-1 on migrated monocytes after the pretreatment of ECs with TNF-&#x003b1; and IL-1&#x003b1; are especially remarkable because these cytokines are important in the pathogenesis of inflammation. In RA, TNF-&#x003b1; and IL-1 blockade showed an unequivocal therapeutic effect [<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B22">22</xref>,<xref ref-type="bibr" rid="B23">23</xref>]. In addition, ICAM-1 and E-selectin levels of patients with RA who had received anti-TNF-&#x003b1; therapy decreased within a few days of the initiation of therapy [<xref ref-type="bibr" rid="B24">24</xref>].</p><p>Previous findings indicate that TNF-&#x003b1; and IL-1&#x003b1; induce an upregulation of the ICAM-1 counter-receptor on ECs [<xref ref-type="bibr" rid="B25">25</xref>]. This is consistent with the increase in cell migration found in our experiments and the altered expression of ICAM-1 on monocytes. Because the classic ICAM-1 counter-receptors LFA-1 and Mac-1 have not been detected on ECs, the existence of another ICAM-1 ligand, one that facilitates the transendothelial migration of monocytes, remains possible.</p><p>The reported changes indicate that, after migration, monocytes could become more liable to interact with T cells (which are known to enhance LFA-1 expression in the presence of TNF-&#x003b1;). This interaction might lead to a further mutual stimulation of T cells and macrophages. In fact the ligand-counterligand system consisting of LFA-1 and ICAM-1 also is one co-stimulatory pathway involved in interactions between antigen-presenting cells (APCs) and T cells [<xref ref-type="bibr" rid="B26">26</xref>]. Because our results demonstrate that both ICAM-1 and HLA-DR are upregulated on migrated monocytes, their function as APCs and possible ability to communicate with T cells, might be facilitated after transendothelial migration. Thus, this observation also supports previous notions on the importance of T cells in the pathogenesis of RA [<xref ref-type="bibr" rid="B27">27</xref>,<xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref>].</p><p>We also observed an upregulation of CD33 on migrated monocytes. In this respect it is noteworthy that Randolph et al. reported monocyte differentiation into dendritic cells after reverse transmigration [<xref ref-type="bibr" rid="B36">36</xref>]; furthermore, it was shown that dendritic cells isolated from RA synovial fluid expressed high levels of CD33 [<xref ref-type="bibr" rid="B37">37</xref>]. Because it is known that the synovial lesions in RA are enriched in dendritic cells [<xref ref-type="bibr" rid="B37">37</xref>], it is tempting to speculate that some CD33 high monocytes might differentiate into dendritic cells in inflammatory sites.</p><p>Previous studies investigated the transendothelial migration of lymphocytes in particular T cells. For both monocytes and T cells the presence of an endothelium facilitated the migration into collagen gels; however, in contrast with the monocyte situation, the capacity of T cells is an intrinsic ability of certain subpopulations, for example CD4-positive memory cells [<xref ref-type="bibr" rid="B7">7</xref>]. Interestingly, the activation of ECs by IL-1 and IFN-&#x003b3; shows the same migration-enhancing effect on T cells as seen here for monocytes [<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B38">38</xref>]. Oppenheimer-Marks <italic>et al</italic> [<xref ref-type="bibr" rid="B2">2</xref>,<xref ref-type="bibr" rid="B39">39</xref>] reported that ICAM-1 on T cells promoted binding and migration through non-activated endothelium. This conclusion is consistent with our results: ICAM-1 seems to be a pivotal adhesion molecule in the transendothelial migration of both monocytes and T cells. Moreover, in a study <italic>in vivo</italic> it was observed that the application of monoclonal antibodies against ICAM-1 induced T cell hyporesponsiveness in patients with RA [<xref ref-type="bibr" rid="B40">40</xref>] and led to a clinical benefit in patients with early or subacute RA.</p><p>In summary, our findings indicate that monocyte migration is accompanied by changes in function-associated surface antigens and that TNF-&#x003b1; and IL-1&#x003b1; in particular increase the number of migrating monocytes and lead to an enhanced expression of certain surface markers involved in cell-cell interactions. These events might not only be partly responsible for the high inflammatory activity in RA synovium; they also suggest that ECs have a pivotal role in these processes and thus might constitute an important therapeutic target.</p></sec><sec><title>Abbreviations</title><p> APC = antigen-presenting cell; BND = population of cells bound to the surface; EC = endothelial cell; ICAM = intercellular cell-adhesion molecule; IFN-&#x003b3; = interferon-&#x003b3;; IL = interleukin; MIG = population of cells migrated into collagen gel; MIP = macrophage inflammatory protein; NAD = non-adherent cell population; PBMC = peripheral blood mononuclear cells; PBS = phosphate-buffered saline; RA = rheumatoid arthritis; TNF-&#x003b1; = tumour necrosis factor-&#x003b1;. </p></sec>

On the typology and the worship status of sacred trees with a special reference to the Middle East

<p>This article contains the reasons for the establishment of sacred trees in Israel based on a field study. It includes 97 interviews with Muslim and Druze informants. While Muslims (Arabs and Bedouins) consider sacred trees especially as an abode of righteous figures' (Wellis') souls or as having a connection to their graves, the Druze relate sacred trees especially to the events or deeds in the lives of prophets and religious leaders. A literary review shows the existence of 24 known reasons for the establishment of sacred trees worldwide, 11 of which are known in Israel one of these is reported here for the first time. We found different trends in monotheistic and polytheistic religions concerning their current worship of sacred trees.</p>

<contrib id="A1" corresp="yes" contrib-type="author"><name><surname>Dafni</surname><given-names>Amots</given-names></name><xref ref-type="aff" rid="I1">1</xref><email>adafni@research.haifa.ac.il</email></contrib>

Journal of Ethnobiology and Ethnomedicine

<sec><title>Background</title><p>Frese and Gray [[<xref ref-type="bibr" rid="B1">1</xref>]: 26] write: "Trees are a form of nature that represent life and the sacred continuity of the spiritual, cosmic, and physical worlds. A tree is often used to symbolize a deity or other sacred beings, or it may stand for what is sacred in general... Trees represent certain deities or ancestors, serve as mediators or as a link to the religious realm, and are associated with cultural beliefs in heaven or the afterlife... Through association with particular religious or historical events, an individual tree or species of tree acquires the symbolic significance of the events as part of its meaning. A society's religious beliefs about the kinds of trees that are sacred generally depend on the nature and number of trees found in the territory. If trees are plentiful, the forest as a whole will also be an important part of the religion's spiritual beliefs and rituals".</p><p>Hughes and Chandran [[<xref ref-type="bibr" rid="B2">2</xref>]:78] have already noticed that sacred groves developed originally in traditional societies, which considered themselves linked in a web of spiritual relationships with their biophysical environments. Trees have always been regarded as the first temple of the gods, and sacred groves as their first place of worship; they were held in the utmost reverence [[<xref ref-type="bibr" rid="B3">3</xref>]:12.2.; [<xref ref-type="bibr" rid="B4">4</xref>]:471; [<xref ref-type="bibr" rid="B5">5</xref>]:203; [<xref ref-type="bibr" rid="B6">6</xref>]:190; [<xref ref-type="bibr" rid="B7">7</xref>]:45; [<xref ref-type="bibr" rid="B8">8</xref>], I: 87]. Sacred individual trees and groups of trees characterized almost every culture and religion where trees were capable of growing [[<xref ref-type="bibr" rid="B9">9</xref>]:4; [<xref ref-type="bibr" rid="B4">4</xref>]:467; [<xref ref-type="bibr" rid="B10">10</xref>], I: 109&#x02013;135, [<xref ref-type="bibr" rid="B11">11</xref>]:414; [<xref ref-type="bibr" rid="B12">12</xref>]:57; [<xref ref-type="bibr" rid="B13">13</xref>]:30]. Thus, it is not surprising to find traces of tree worship in the Middle East [[<xref ref-type="bibr" rid="B14">14</xref>]:187]. There is even evidence about the magnitude of tree worship in Palestine in the 19^thcentury "yet in no country are the people more awed by trees than in Palestine" [[<xref ref-type="bibr" rid="B15">15</xref>]:54].</p><p>It is recognized that trees are not worshipped for themselves but for what is revealed through them, what is implied and signified [[<xref ref-type="bibr" rid="B16">16</xref>]:268; [<xref ref-type="bibr" rid="B17">17</xref>]:28], especially some kind of power that they express [[<xref ref-type="bibr" rid="B18">18</xref>]:35; [<xref ref-type="bibr" rid="B19">19</xref>]:359; [<xref ref-type="bibr" rid="B12">12</xref>]:57] or their being the abode of supreme beings [[<xref ref-type="bibr" rid="B20">20</xref>]:91]. Eliade [[<xref ref-type="bibr" rid="B21">21</xref>]:149] noted, "The image of the tree was not chosen only to symbolize the cosmos but also to express life, youth, immortality, wisdom. In other words, the tree came to express everything that religious man regarded as <italic>pre-eminently real and sacred</italic>" (italics in the original). It seems that worship of trees began, among other reasons, because of their long life. In most ancient civilizations trees, whose life-span is several times greater that man's, were treated with the same respect as the elders among men Several generations lived in the shade of the same venerable tree, almost as if it were eternal [[<xref ref-type="bibr" rid="B22">22</xref>]:276].</p><p>Strangely enough, despite the long list of specific books devoted solely to tree worship [[<xref ref-type="bibr" rid="B23">23</xref>]:9; [<xref ref-type="bibr" rid="B24">24</xref>]:25; [<xref ref-type="bibr" rid="B20">20</xref>]:26; [<xref ref-type="bibr" rid="B27">27</xref>]:28; [<xref ref-type="bibr" rid="B29">29</xref>]:6], we were unable to find any definition of a "sacred tree". Several authors recognize some "categories" of sacred trees (e.g., [[<xref ref-type="bibr" rid="B30">30</xref>]:448; [<xref ref-type="bibr" rid="B27">27</xref>]:5&#x02013;20; 12:58&#x02013;67]), some of which are not mutually exclusive. To clarify the "hallowed" status of trees, in Israel, we must elucidate the conceptual difference between "blessed", "sacred", and "holy" trees. The difference is not merely semantic but reflects the religious attitude to the adoration of trees. According to the Druze religion only people like prophets could be "sacred"; physical objects like trees may be regarded only as "blessed" [<xref ref-type="bibr" rid="B31">31</xref>]. A plant species all of whose specimens are worshipped owing to religious tradition (regardless the exact background) has to be treated as "holy". Simoons [1998:293] distinguishes tree rituals wherein a certain species of tree is considered "holy", as in the case of the sacred fig (Bo tree, <italic>Ficus religiosa </italic>L.), from rituals in which individual trees are "sacred" because of special characteristics or have won respect through their location in a holy place or their association with a holy person (see also [[<xref ref-type="bibr" rid="B33">33</xref>]:150] for a similar distinction).</p><p>The literature survey shows that the definitions of "sacred tree/wood/grove/forest" are by no means mutually exclusive due to the complexity of reasons for, and the history of, the sanctification of the individual tree, a tree species, or a group of trees. Several authors supply definitions of sacred forest/grove/wood: for example, "... a sacred grove is a stand of trees in a religious context" [[<xref ref-type="bibr" rid="B34">34</xref>]:1]; " Sacred trees... describe individual trees or woods which were treated with a certain reverence which, normally, protected them from a wilful damage "[[<xref ref-type="bibr" rid="B35">35</xref>]:16]; "Clusters of forest vegetation that honour a deity, provide sanctuary for spirits, remind present generations of ancestors or protect a sanctified place from exploitation. They are treated as sacred by virtue of their location, cultural meaning and history", [[<xref ref-type="bibr" rid="B36">36</xref>]:30]; "Sacred grove is a patch of forest or natural vegetation protected and managed by the community considering it to be the reside place of the deities or ancestral spirits" [[<xref ref-type="bibr" rid="B37">37</xref>]:2]; for similar definitions see also [[<xref ref-type="bibr" rid="B38">38</xref>]:225; [<xref ref-type="bibr" rid="B40">40</xref>]:49]; "Sacred groves are more or less patches of climax vegetation... preserved on a religious ground" ([[<xref ref-type="bibr" rid="B41">41</xref>]:272] see also [[<xref ref-type="bibr" rid="B42">42</xref>]:1063 ; [<xref ref-type="bibr" rid="B43">43</xref>]:1204; [<xref ref-type="bibr" rid="B44">44</xref>]:1541&#x02013;1542]). Freeman [[<xref ref-type="bibr" rid="B45">45</xref>]:262]) criticized these "ecological" definitions because they were derived from a botanical ideal (climax) and not based on local understanding; his definition is "a piece of garden or forest land... that is dedicated for the exclusive use of particular deities".</p><p>Hughes and Chandran [[<xref ref-type="bibr" rid="B2">2</xref>]:69] supply a comprehensive definition for a sacred grove as follows: "Segments of landscape containing trees and other forms of life and geographical features that are delimited and protected by human activities believing that preserving such a patch of vegetation in relatively undisturbed state is necessary for expressing one's relation to the divine or to nature".</p><p>Gupta [[<xref ref-type="bibr" rid="B20">20</xref>]:19] distinguishes a "tree-god", whose worship became organized into a definite religion, from a "tree spirit", whose propitiation degraded the level of sorcery and incarnation. In practice it is impossible to discern, "spirits", "demons", and "jinns" (general supernatural agents) as against "goddesses, "gods", and "the deity" (religiously established worshipped elements). In the Middle East and North Africa, specific trees may be considered the abode of jinns, demons, or spirits, but these supernatural powers are never worshipped as a kind of "god". No religious ceremonies are associated with or performed near these trees; these are regarded as heathen rites and are strictly prohibited.</p><p>The various definitions of "sacred tree/grove/wood/forest" may be classified according to four groups of criteria: <bold>A. </bold>natural elements: the physical characters of the tree; <bold>B. </bold>supernatural elements believed to reside in the tree and act upon humans. <bold>C</bold>. human ritual behaviours related to the trees and <bold>D</bold>. botanical criteria such as climax and high biodiversity. A sacred tree/grove/wood/forest may contain the following seven elements: 1. It is the abode of a supernatural power. 2. It is well delineated physically/geographically. 3. The trees are protected by taboos from cutting/exploitation/disrespect/secular behaviours. 4. It is related to historical/cultural/religious issues. 5. The area is protected to please the supernatural powers so as to ensure their benevolence or to avert their malevolent power. 6. It is a piece of natural vegetation (in most cases). 7. It is a ubiquitous phenomenon not limited to any specific religion or geographic territory.</p><p>As a practical working definition we suggest treating "sacred trees" as "trees that are subjected to practical manifestations of worship, adoration, and/or veneration that are not practised with ordinary trees". These trees could be single units, groves, forests, or all the specimens of a certain botanical species.</p><p>Many religions relate to "metaphysical" trees such as "cosmic tree", "sky tree", "inverted tree", "tree of life", "celestial trees", " tree of wisdom", and "tree of knowledge". These "types" are not mutually exclusive [[<xref ref-type="bibr" rid="B6">6</xref>]:273&#x02013;278; [<xref ref-type="bibr" rid="B1">1</xref>]:27&#x02013;28, [[<xref ref-type="bibr" rid="B9">9</xref>]:1&#x02013;23] Some of these "spiritual trees" are identified with specific species: The Indo-European cosmic and tree of life with oak [[<xref ref-type="bibr" rid="B6">6</xref>]:278], the Indian "sky tree" with <italic>Ficus religiosa </italic>[[<xref ref-type="bibr" rid="B1">1</xref>]:27], the Assyrian tree of life with date palm [[<xref ref-type="bibr" rid="B46">46</xref>]:8&#x02013;13; [<xref ref-type="bibr" rid="B47">47</xref>]:132&#x02013;133]; while the Egyptian "tree of life" is identified as a date or as sycamore (<italic>Ficus sycamorus</italic>) [[<xref ref-type="bibr" rid="B48">48</xref>]:76]. Belief in "metaphysical trees" is not necessarily evidence of practical tree worship [[<xref ref-type="bibr" rid="B6">6</xref>]:272].</p><p>In this study we exclude "metaphysical trees", while taking into consideration that sometimes the boundary between these trees and reality is not clearly delineated. The present paper is an attempt to elucidate which causes of tree sanctification are characteristic of the Middle East as against other regions, and if differences exist among ethnic groups in Israel on this matter.</p></sec><sec sec-type="methods"><title>Methodology</title><p>The field study (1999&#x02013;2005) centred on Arab, Bedouin and Druze villages in Galilee. Informants were asked about the reasons why certain trees became sacred. The survey covered 97 informants, consisting of 38 Druze, 59 Muslims (36 Arabs and 23 Bedouins). We consider "Arabs" people who have been settled in their villages for several centuries, "Bedouins" people who originated from the deserts of Israel and Jordan, migrated to the Galilee at the last three centuries and were nomadic till the end of the 20^thcentury [[<xref ref-type="bibr" rid="B49">49</xref>]:30]. The Druze are an East Mediterranean group adhering to a religion that was established in Egypt the 11^thcentury [[<xref ref-type="bibr" rid="B50">50</xref>]:3]. Today they are concentrated in Lebanon Syria and Israel [[<xref ref-type="bibr" rid="B50">50</xref>]:8&#x02013;14]. The belief in the revelation of God in the form of a human being is considered the most important fundamental principle of the Druze faith [[<xref ref-type="bibr" rid="B50">50</xref>]:15]. Druze faith is not ritual-ceremonial religion in essence, but rather a neo-platonic philosophy [[<xref ref-type="bibr" rid="B9">9</xref>]:17].</p><p>We distinguish "Arabs" and "Bedouins" in attempt to discern different traditions regarded sacred trees which may reflect the different origin of nomads and settled village people. The survey excluded Christians, who hardly believe in sacred trees, while in the Jewish sector the adoration of trees is a new trend of the last two decades and almost all the worshiped trees are already known as old Muslim sacred ones in the vicinity of graves of supposed historical Jewish righteous personalities.</p><p>In each village we made a preliminary survey to locate the knowledgeable people in advance and we also approached important religious leaders to examine theirs attitudes to the veneration of sacred trees, then informants were chosen according to their knowledge of common traditions and/or religious status. The average age of the informants was 57.7 (+/-14.8) years. Respondents were 95 males and two females (in general women are reluctant to be interviewed, and when they agreed the interview was held in the presence of other family members). Because of the refusal of most of the informants to be videotaped or recorded all the study is based on oral interviews and field notes that were taken on the spot. The interviewees were asked, "Why have specific trees [especially in their home village] become sacred?" and "How is al-Khader [= Elijah] related to trees?" this question was introduced because this prophet is the most popular one (see below) and at not less that 30 sacred trees are named in his honour [50, Dafni, unpubl.].</p></sec><sec><title>Results</title><p>The answers for the question "Why are trees venerated?" are presented in Table <xref ref-type="table" rid="T1">1</xref></p><table-wrap position="float" id="T1"><label>Table 1</label><caption><p>Reasons for sanctification of trees Numbers indicate the percentage of informants in each ethnic group that gave a certain reason.</p></caption><table frame="hsides" rules="groups"><thead><tr><td align="left">Reason</td><td align="center">Druze</td><td align="center">Arab</td><td align="center">Bedouin</td><td align="left"><bold>References from Palestine</bold></td><td align="left"><bold>Other regions (selected references).</bold></td></tr></thead><tbody><tr><td align="left">1. Tree is the abode of a Welli's/saint's spirit.</td><td align="center">0</td><td align="center">58.3</td><td align="center">52.0</td><td align="left">[60:303,331;55:91; 57:264; 56:152; 106:89]</td><td align="left">[Iraq (107:92); Iran (87,I:375;108:317); Turkey (65:41)]</td></tr><tr><td align="left">2. A saint is buried near/under the tree.</td><td align="center">2.6</td><td align="center">52.7</td><td align="center">21.7</td><td align="left">[55:93, 61:27, 62:242]</td><td align="left">[Morocco (59:67; Turkey (58:227); Algeria (109:311); Turkey (110:215)]</td></tr><tr><td align="left">3. Religious and social meetings take place under the tree; well-known figures are associated with the tree.</td><td align="center">71.0</td><td align="center">44.4</td><td align="center">30.4</td><td align="left">[56:71; 111:36]</td><td align="left">[Ancient Assyria (112:42); Iran (92:142); Morocco (59:68); India (73:19&#x02013;28)]</td></tr><tr><td align="left">4. The tree is dedicated to a prophet.</td><td align="center">47.3</td><td align="center">22.2</td><td align="center">34.7</td><td align="left">[56:65]</td><td></td></tr><tr><td align="left">5. The species is religiously blessed.</td><td align="center">26.3</td><td align="center">19.4</td><td align="center">17.3</td><td align="left">[31]</td><td></td></tr><tr><td align="left">6. The tree/forest commemorates events in the lives of saints, heroes, kings, in the tribe's history etc..</td><td align="center">36.8</td><td align="center">41.6</td><td align="center">26.0</td><td align="left">[106:51; 55:93; 113:95]</td><td align="left">[Assyria (112:42); Ancient Britain (26:44,104); Ireland (35, 19, 21, 25, 33, 35:29, 32, 36); England (118:184); Morocco (114:27); Mozambique (103:229); Uganda (115:37); Ghana (38:225&#x02013;232;120:366; 121:41); Iran (87, I: 372, 92:142); India (116:26); Nepal; (119:248; 122:335); Okinawa (96:57); New Zealand (117:220)]</td></tr><tr><td align="left">7. The tree shows the way to a sacred place.</td><td align="center">92.0</td><td align="center">0</td><td align="center">0</td><td></td><td></td></tr><tr><td align="left">8. The tree sprouted from saints' staffs.</td><td align="center">7.8</td><td align="center">11.1</td><td align="center">4.3</td><td align="left">[56:27; 124:407; 125:378&#x02013;379]</td><td align="left">[England (26:21); Ireland (35:35, 38, 40); Poland (6:55); Morocco (59:68); Iran (125:41); India (126 (127:287 in 21:27); New Zealand (128:68)]</td></tr><tr><td align="left">9. Tree provides shade in the desert</td><td align="center">0</td><td align="center">0</td><td align="center">17.3</td><td></td><td align="left">[Ancient Egypt (129:121); Iran (87,I:373); India (130:43)]</td></tr><tr><td align="left">10. The Tree Grows/is planted over the grave of the saint.</td><td align="center">2.6</td><td align="center">2.7</td><td align="center">0</td><td align="left">[106:89]</td><td align="left">[Ancient Celts (5:202); Egypt (66:56); Morocco (63:80); Zanzibar (131:36); Mongolia (79:2830]</td></tr><tr><td align="left">11. The tree is sacred because of the tree's size, age, shape; being evergreen or having a strange form.</td><td align="center">0</td><td align="center">5.5</td><td align="center">0</td><td></td><td align="left">[Ancient Greece (76:26; 34:10); Ancient Rome (3, 12.2.3.; 132:67); Pagan Europe (74:38; 132:67); Russia (133, I:194); Siberia (141:58); Armenia (135:320); East Africa (78:4); Kenya (139:151); Cameroon (140:100); Sierra Leone (136:47); Iran (92:142); Inner Mongolia (80:279,283); India (137:239); China (138:759); Taiwan (134:5,III,1)]</td></tr><tr><td align="left">12. The tree has healing powers.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Britain (26:44); Uganda (143, II: 832); China (142:412&#x02013;413)]</td></tr><tr><td align="left">13. The tree is the abode of supernatural beings: tree spirits, djinns, demons, deities, goddess, angels, divine beings, dragon, monsters, etc.,</td><td align="center">0</td><td align="center">2.7</td><td align="center">0</td><td></td><td align="left">Ancient Egypt (151:89; 70:40); Minoan (152:141); Ancient Greece (34:10, 16, 20); Ancient Celts (5:198); Old Scandinavia (6:52); Teutonic mythology (8,I:71); East Africa (78:4); Uganda (115:37); Zimbabwe (101:6); India (98:9,15; 137:240,242); Ethiopia (154:3); Zanzibar (131:35); Mozambique, 103:229); Sierra Leone (156:102); Nigeria (157:292); West Africa (158:44); East Africa (78:4); Central Africa (159:317); Siberia (155:117); Iran (144:125); Inner Mongolia (80:277); Nepal (150:334); India (137:242; 40:280; 43:1578; 153:47); Thailand (147:330,147:66); China (138:759;161:131); Japan (145:34,35; 60:176); Indonesia (148:317);Okinawa (96:46,57,63;149:177);Micronesia (102:9); Papua (162:72)]</td></tr><tr><td align="left">14. All the individual trees of the species are religiously sacred.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Dahomey (172:54); India (165); Armenia (166:5); Tibet (168:325); China (138:759); Micronesia (167:13); Kiribati Pacific Islands (102:46); Chile (169:146); American Indians (164:121)]</td></tr><tr><td align="left">15. To commemorate miracles occurred near the tree or related to a saint.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td align="left">[(56:70]</td><td align="left">[Ireland (35:37,43); Egypt (90:17); ]Morocco (59:67,259); Iran (108:317); India (137:243); Tibet (171:276); Taiwan (134:5,III,1); China (172:133)]</td></tr><tr><td align="left">16. Tree is abode of ancestors' souls.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Old Celts (5:103); Mozambique (103:231); Ghana (121:41); Madagascar (176:19,20, 177:61); Zimbabwe(101:6,9;126, :378); Sierra Leone (136:47); S. Rhodesia, (178:102); Ghana (120:366; 179:149, 18:35); Kenya (180:135; French Guinea (181:14); Inner Mongolia (80:277,280); India (137:242; 173:332; 38:593); Laos (174:4; 182:7); Indonesia (175:310,318); Papua (162:72); Australia (183:163)]</td></tr><tr><td align="left">17. The tree is located near holy water source.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Ancient Greece (34:20,25,34); Ancient Syria (82:115); Ireland (35:40&#x02013;42; 83:30); Egypt (90:17); Algeria (184:72); Tanganyika (185:39); Turkey (110:215); China (138:759)]</td></tr><tr><td align="left">18. Tree/grove is grown/planted in sacred places, burial sites, graveyards, temple worship sites.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Ireland (35:29,44); East Africa (189:414,415,); Ghana (120:366;121:41); Kenya (180:134); Madagascar (191:83); Central Asia (190:99); Sri-lanka (150:176); Nepal (151:270,334); India (137:242; 32:293; 188:105;192:433; 193:237); Japan (186:180, 185:133) ; Taiwan (134:5, III, 1); China (161:131; 194:704)]</td></tr><tr><td align="left">19. The tree is useful.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td align="left">Sinai (195:184)</td><td align="left">[Ancient Assyria (112:42); Europe (5:199; 6:149); Ethiopia (154:4); Nigeria (157:292); S. Rhodesia (178:126&#x02013;127); India (137:252, 196:157, 196:400; 73:21); Himalaya (38:58)]</td></tr><tr><td align="left">20. Tree creates a religious atmosphere</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Ancient Rome (196:IV, 12, 3: 6:52); India (201:67; 199:153, 1981:281)]</td></tr><tr><td align="left">21. The tree attracts lightning</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Europe (202:676; 203:79; 6:161); India (137:239); American Indians (204:15)]</td></tr><tr><td align="left">22. God is transformed into a tree.</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Ancient Egypt (205: 15.16; 14:191); Russia (100:700)]</td></tr><tr><td align="left">23. The grove/forest declared sacred to insure the community resources/water catchments</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">[Uganda(115:38);Mozambique (207:12);India ([206:8); China (163:131)]</td></tr><tr><td align="left">24. Planting a tree in high places as an abode for gods</td><td align="center">0</td><td align="center">0</td><td align="center">0</td><td></td><td align="left">Mongolia [79:282].</td></tr></tbody></table></table-wrap><p>When the interviewees were asked "why trees are venerated", they gave the following answers (the bold numbers are related to certain informants, see Appendix):</p><p>1."The sacredness of the tree originated only from a sacred tomb or the place of a Welli" (23 infor-mants).</p><p>2."A man that prays to a tree is a heathen; we need to pray only to God. The tree is temporal, only God is eternal. If the sacred tree is not able to protect itself against being cutting down, how will it protect humans? We have to pray to God, who is the Creator. We have to worship God, not trees" (<bold>1</bold>).</p><p>3."We have to believe in God, not in trees; it is against the religious law" (<bold>2</bold>).</p><p>4."God protects sacred trees" (<bold>3</bold>).</p><p>5."Sanctity rests only on prophets; it is forbidden to sanctify objects like stones and trees" (<bold>4</bold>).</p><p>6."People have prayed to trees since Roman times, and [the tendency] remains inside the human. I believe only in one God" (<bold>5</bold>).</p><p>7." The sacred tree is a path between man and god" (<bold>6</bold>).</p><p>8. "The importance of the sacred trees depends on the saint; the holier the man, the holier is the tree " (<bold>7</bold>).</p><p>9."Sacred trees are the memento of a saint or holy man, so their importance is relative to the man's holiness" (<bold>8</bold>).</p><p>10." The blessed tree is a symbol of the prophet, not an object per se" (<bold>9</bold>).</p><p>11."Holy is that which is sanctified by God, blessed it is by humans'" (<bold>10</bold>).</p><p>12."Each blessed tree has an angel or jinn or demon that resides inside and protects it" (<bold>11</bold>).</p><p>13."Religious people give the tree their sacredness; maybe the saint planted the tree or was buried underneath it" (<bold>12</bold>).</p><p>14."Holiness is accorded only to the prophets and it is forbidden to sanctify objects like trees and stones" (<bold>13</bold>).</p><p>15."Faith gives the strength and a tree near a sacred grave gives blessing" (<bold>14</bold>,<bold>15</bold>).</p><p>16."In our religion (Druze) we don't sanctify people or trees, only God" (<bold>15</bold>). "We love the Prophet but we don't sanctify him" (<bold>16</bold>).</p><p>17."Blessed trees are memorials to singular figures in Druze history and religion; because the Druze tradition forbids any tomb sign or offerings, special people are remembered by large blessed trees" (<bold>17</bold>).</p><p>18."The man (a religious figure) is sacred and the tree is blessed. The tree belongs to the people and the saints are the prophets and God". "The blessed trees are a monument to special figures in the history of the Druze religion. In our religion there are no gravestones or offerings in graveyards; the trees commemorate the deeds of these special personalities" (<bold>11</bold>).</p><p>When our informants were asked, "How is al-Khader related to trees?" we received the following answers:</p><p>1. "Every tree that is dedicated to al-Khader is a blessing; he is even closer to God than Nabi Shu'ayb (the Druze's most important prophet)" (<bold>15</bold>).</p><p>2. " Al-Khader has the power of 70 prophets" (<bold>18</bold>).</p><p>3. "Every place where al-Khader sat became green" (12 informants).</p><p>4. "The place is sacred and green because al-Khader rested there" (11 informants).</p><p>5. "The colour of al-Khader is green" (21 informants).</p><p>6. "al-Khaher had the habit of sitting under trees", "al-Khader is related to trees", or "al-Khader loves trees" (12 informants).</p><p>Based on Table <xref ref-type="table" rid="T1">1</xref>, the analysis of the interviews, additional field observations and the literature survey we find the following trends: (figures in parentheses are the numbers as in Table <xref ref-type="table" rid="T1">1</xref>).</p><p>1. The most common reasons for the sanctity of a tree are its being the abode of a saint's (Welli's) spirit and the location in its vicinity being the grave or shrine of a saint ("Makam"); these two reasons were strictly confined to Muslims and were never given by Druze.</p><p>2. Dedication to a prophet, the most popular of whom is al-Khader.</p><p>3. Religious and social meetings: The most common reason given was that the tree was a place beneath which religious leaders used to gather for preaching/discussions/meetings/court sessions/judgements. This reason was more represented among Druze than among Muslims.</p><p>4. Events in lives of saints. These may include praying, preaching, resting, living or visiting under the specific tree. This reason seemed similarly represented among the three ethnic groups.</p><p>5. The tree species is adored because of a religious dedication: the only tree that in this category was <italic>Ziziphus spina christi </italic>(see [<xref ref-type="bibr" rid="B31">31</xref>] and below).</p><p>6. The tree shows the way to a sacred place: this reason was given only by Druze, all of whom</p><p>mentioned one specific tree (see below).</p><p>7. Some reasons (13, 14, 16, 18) which seem to be connected especially to polytheism are absent from the Middle East.</p><p>8. Some reasons that are known from the classical world, as well as from pagan Europe, are very rare or not known today in the Middle East (10,11,13,14,15,16,17,19,20,21,22).</p><p>9. All the recorded reasons (except "showing the way to a sacred place") are already mentioned in the literature especially in ancient Europe, North Africa, and the Fertile Crescent.</p><p>10. Some reasons, as far as the author is aware, have never been recorded in the Middle East: (12&#x02013;14, 16&#x02013;18, 20&#x02013;24).</p><sec><title>New and abandoned sacred trees</title><p>During our study we came upon four cases in which the "sacredness" of a tree could be declared "de novo" or annulled for political as well as religious reasons.</p><p>On the Jewish Day of Atonement in 2003 several people from the Druze village of Dalyat al-Karmel (Mt. Carmel) decided to annex an area that was in dispute with Israel's Nature Reserve Authority. In an organized operation an asphalt road was paved and a large oak tree (<italic>Quercus calliprinos</italic>) in the at Nahal Alon Nature reserve, was declared blessed and a new religious building ("Khalva") was erected at this place. By this act the authorities were forced to accept the Druze claim to the area (<bold>19, 20</bold>). After the religious ceremony the local people accepted the tree as blessed and since that day they regularly place put rags and flags on it, and honour the tree as is customary with every "sacred tree". The action was "covered" by a religious tradition (<bold>21</bold>). The local keeper of the "Khalva" related that a note dating back 180 years was found in one of the holy books. It told of a pious virgin, "Sit Khadra", (the green lady) who lived in the area and was famous as a successful farmer. After this note was found it was decided to sanctify the area in which she was lived to commemorate her (<bold>22</bold>).</p><p>A case of withdrawal of sanctification occurred in the town of 'Arr&#x00101;be (Lower Galilee). There stands a very famous ancient tree of <italic>Pistacia atlantica</italic>, which is "the" sacred tree of the village. It is named "the tree of the Saddik (righteous man)" believed buried under it. In 1961 an archaeological excavation discovered that this was a Jewish grave (the burial direction was north-south, not west-south as in the Arab tradition). The authorities constructed an old-style building and named it officially after "Khanina Ben Dossa". A famous Talmudic sage (the tradition concerning this burial place is dated to the 10^thcentury [[<xref ref-type="bibr" rid="B52">52</xref>]:324&#x02013;325; [<xref ref-type="bibr" rid="B53">53</xref>]:152&#x02013;155]. From the time the tomb was declared Jewish, the appeal of the tree as sacred to the Muslims declined and fewer people go there to ask the help of the righteous man buried underneath. The various informants (n = 10) held conflicting views as to whether the change was influenced by the local religious leaders.</p><p>Similarly, recently (around 2002) religious leaders in the village of Mes'hed (near Nazareth) were unhappy about the veneration by Jewish people of a tree in its midst (a large <italic>Pistacia atlantica </italic>that is a centre for vows and is abundantly visited by local people as a "wishing tree"). Nevertheless, people continue to visit the tree as usual (<bold>23</bold>). In the town of Sakhnin (Lower Galilee) there is an ancient cave containing the tomb of Rabbi Yehoshua D'Sikhnin [[<xref ref-type="bibr" rid="B52">52</xref>]:302&#x02013;303, [[<xref ref-type="bibr" rid="B53">53</xref>]:150&#x02013;151]. The place is renowned throughout the region for its miraculous powers to cure sick people, and especially barren woman. Very close to the cave there is a large sacred Ziziphus <italic>spina christii </italic>tree stands in the middle of an Arab cemetery. In 1980 a local Muslim religious leader decided to forbid the attachment of rags to the tree as well as prayer there, because the place was sacred to the Jews but not the Arabs. The ban held for five or six years, after which the Muslim people returned to their old tradition, saying that the Jewish saint helped them (<bold>24</bold>).</p></sec></sec><sec><title>Discussion</title><p>All the known reasons for the establishment of the sacred trees/groves/forests may apparently be sorted into several generalized categories: dedication to supernatural beings/powers, relation to established religious rituals and ceremonies, dedication to people, commemoration of historical or miraculous events and practical, economic and conservational reasons. It is noteworthy that in a single community/religion more than one reason, for the establishment of a sacred grove (and we suggest to extending this to all types of sacred/holy trees) may be established. More than one aspect was caused by a combination of economic, religious, social and environmental factors, to yield of social, environmental, economic as well as religious reasons [[<xref ref-type="bibr" rid="B55">55</xref>]:30]. Political as well as religious reasons could lead to the declaration of new "sacred tree" or to the denial of well known existing ones, as was shown in this study.</p><sec><title>Trees as an abode of a saint's spirit</title><p>As it can be seen (Table <xref ref-type="table" rid="T1">1</xref>) the most common "function" of the sacred trees in the Middle East is to serve as the abode of the spirit/soul of a saint (Welli). Curtiss [[<xref ref-type="bibr" rid="B56">56</xref>]:75, 77, 79], regarding the status of saints in the Muslim world, noted "... orthodox Moslems insist that the saints are only mediators that a worshipper asks his Welli to intercede for him with God... These saints are really departed spirits, connected with some particular shrine, chosen because they revealed themselves there in the past, and where they are wont to reveal themselves now to these who seek their favour .... The worship of the saints is like that of the ancient Baalim. They are the deities whom people fear, love, serve and adore". Cannan [[<xref ref-type="bibr" rid="B57">57</xref>]:151] held that "The present-day peasant does not venerate the trees themselves but the divine-power which acts in them and which is derived from the godly person whose soul is supposed to be still inhabiting the shrine, tomb, cave or spring with which they have become associated. Often these holy men have appeared either in the tree itself or near by". The objection of the religious leaders, and role of the tree as mediator, were also stressed by our informants.</p></sec><sec><title>Trees and saints' graves</title><p>In the Middle East, as in North Africa, a saint's grave is closely connected to a sacred tree; trees beneath which saints are buried are regarded as "sacred trees" [[<xref ref-type="bibr" rid="B56">56</xref>]:93]. The identification of the sacred tree with the saint's grave imparts to it the miraculous and magical powers of the holy man [[<xref ref-type="bibr" rid="B58">58</xref>]:264; [<xref ref-type="bibr" rid="B56">56</xref>]:94; [<xref ref-type="bibr" rid="B57">57</xref>]:71; [<xref ref-type="bibr" rid="B59">59</xref>]:176&#x02013;177]. Westermarck [[<xref ref-type="bibr" rid="B60">60</xref>]:74] comments that the existence of sacred groves around saints' tombs maybe related to the people's avoidance of cutting down these trees for fear of the saint's retribution, even in places where there is no tomb but only a tradition of the holiness of the location, especially in cases where it is not at all clear who the saint is. Thus it seems (in Morocco) that the worship site exists owing to the grove more that the grove exists owing to the site. Canaan comments that sacred trees that are not connected with graves never bear the name of a specific personality [[<xref ref-type="bibr" rid="B6">6</xref>]:70]. Our data show that this is not a rule, and today many sacred trees in Israel are not related to tombs; however, these structures might have disappeared in the course of time. Because tradition relates the tree to a saint, it is respected accordingly [[<xref ref-type="bibr" rid="B61">61</xref>]:331; [<xref ref-type="bibr" rid="B62">62</xref>]:27&#x02013;28]. In practice it is impossible to determine which came first, the tree or the grave, because of the customs of burying important people near sacred trees [[<xref ref-type="bibr" rid="B63">63</xref>]:242] and of planting trees on saints' graves (Morocco [[<xref ref-type="bibr" rid="B64">64</xref>]:80]. Hasluck [[<xref ref-type="bibr" rid="B59">59</xref>]:238] concluded on this subject in Turkey, "It is often impossible to say whether the sacredness of these groves is primitive and their connection with saints evolved from it, or whether it is secondary and due to their proximity to saint's graves"; the same situation exists in Syria [[<xref ref-type="bibr" rid="B65">65</xref>]:179] (at that time Palestine was a part of "Great Syria").</p><p>Zarcone [[<xref ref-type="bibr" rid="B66">66</xref>]:41] notes that certain trees are sacred because of their connection to a specific figure in the Islamic tradition, and tree veneration may be the fusion of tree worship in general as a part of the supernatural. So sometimes the tree confers holiness on a specific site or a sheikh's tomb as a part of his hagiography. Blackman [[<xref ref-type="bibr" rid="B67">67</xref>]:57] mentioned that (in Egypt) trees were sanctified because they grew at a place where a saint was murdered and they bear his soul. He also argued that this is parallel to the ancient Egyptian myth that the sycamore <italic>(Ficus sycamorus</italic>) grew out of the dead body of Osiris [[<xref ref-type="bibr" rid="B68">68</xref>]:29,339].</p><p>The conclusion is that it is not clear if the tree became sacred because of the saint, or the personality became sacred because of the tree.</p></sec><sec><title>Social and religious meetings</title><p>Especially among the Druze, trees acquired their sacredness through the habit of historical religious leaders meeting under them, to preach or/and to discuss religious issues. When leaders from Lebanon used to visit their faithful in the Galilee they customarily met their local colleagues beneath these trees, which came therefore to be considered "blessed". Sacred trees can't be considered as an abode of a soul because the Druze believe in the transmigration of souls, a person's body is a kind of clothing for the soul, and with the person's demise the soul passes over to the body of a newborn child" [[<xref ref-type="bibr" rid="B9">9</xref>]:60]. Thus, souls cannot reside in a tree and graves are not revered, trees were blessed on account of the visits of the religious leader; but souls are never connected with trees.</p></sec><sec><title>Dedication to a prophet</title><p>In all the ethnic groups we found sacred trees that were dedicated to a prophet. When we asked specifically who the most common, people mentioned al-Khader, who is highly respected by Muslims as well as by Druze. Not less than 30 places (some of which contain "sacred trees") in the Holy Land are named after him [[<xref ref-type="bibr" rid="B69">69</xref>]:13&#x02013;34]. This prophet is adored by Druze as well by Muslims. Al-Khader (also Al-Khidr or Al-Khudr), who has common features that characterize Elijah and St. George [[<xref ref-type="bibr" rid="B8">8</xref>]:48&#x02013;65], is the most popular of all saints in the Middle East [[<xref ref-type="bibr" rid="B56">56</xref>]:84; [<xref ref-type="bibr" rid="B70">70</xref>]:288; [<xref ref-type="bibr" rid="B59">59</xref>]:319&#x02013;336]. This prophet is closely connected with sacred trees, as was also found in this study. This notion is revealed in the name: Al-Khader means "the green one" [[<xref ref-type="bibr" rid="B70">70</xref>]:288; [<xref ref-type="bibr" rid="B69">69</xref>]:9]. It is believed that every place on which Al-Khader sat became green. This concept may explain why so many trees are dedicated specifically to this prophet.</p></sec><sec><title>Religious species ("holy trees")</title><p>In many cultures (see Table <xref ref-type="table" rid="T1">1</xref>) all the individual trees of certain species are sacred, the most famous being <italic>Ficus religiosa </italic>underneath which Buddha received his enlightenment [[<xref ref-type="bibr" rid="B71">71</xref>]:24; [<xref ref-type="bibr" rid="B32">32</xref>]:41&#x02013;100]. The only tree in the Middle East that can be regarded as close to "holy tree" is <italic>Ziziphus spina christi</italic>, which is mentioned in the Quran. Individual trees of this species are highly respected, by Muslims, but are worshipped only in connection with a saintly person, and not <italic>per se</italic>. The Druzes treat this species at the same manner, but it is still regarded as a "blessed" tree [<xref ref-type="bibr" rid="B31">31</xref>]. All the other categories of worshipped trees (Table <xref ref-type="table" rid="T1">1</xref>) can thus be considered as "sacred trees".</p></sec><sec><title>Events under the tree</title><p>In our survey Muslims as well as Druze mentioned events in the life of the saints/prophet/religious figures as one of the main reason for the sanctification of trees. Curtiss [[<xref ref-type="bibr" rid="B56">56</xref>]:93] noted that "trees under which saints rested are considered holy". We can add that it was sufficient for the saint to teach, preach, or pray under a tree to make it sacred. In Israel we failed to find even a single sacred tree that commemorates a specific historical event. In Britain, for example, many such cases are found, although it is hard to discern what comes first, the event or the sanctification of the tree [[<xref ref-type="bibr" rid="B26">26</xref>]:44,104], see also Table <xref ref-type="table" rid="T1">1</xref> for more eamples.I.</p></sec><sec><title>Showing the way to a sacred place</title><p>Near village of Mghar (Lower Galilee), on the main road to Nabi Shu'ayb (believed to be the grave of the prophet Jethro), the holiest place for Druze in Israel [[<xref ref-type="bibr" rid="B50">50</xref>]:217&#x02013;218], there is a huge Christ's Thorn Jujube (<italic>Z. spina christi</italic>) tree. In the past this important tree served as a meeting point for pilgrims before approaching the holy place for the festival of Nabi Sua'yb (on 25 April, every year). Whoever arrived first waited for the others under that tree. Over the years the tradition of the first meeting point took root, and this specific tree became a station for praying as well. It is the only individual tree of his kind that reached the status of a "blessed tree". When the pilgrims reached the tree they became very excited, and this is how the tree came to be named "Sidrat Nebi Shu'ayb" (the Prophet's Jujube) [<xref ref-type="bibr" rid="B31">31</xref>].</p></sec><sec><title>Sprouting from a saint's staff</title><p>In some places we heard that sacred trees had sprung from staffs carried by saints or religious pleaders. Similar stories are known also from other countries and are not endemic to our region (Table <xref ref-type="table" rid="T1">1</xref>).</p></sec><sec><title>Shadow in the desert</title><p>This reason was mentioned only by the Bedouins and can be looked as vestiges of old traditions reflecting their history.</p></sec><sec><title>Planting trees in sacred places/groves growing in sacred places and tree's as having healing powers</title><p>These reasons are rare in the Middle East and are not common worldwide.</p></sec><sec><title>Tree characters</title><p>Only two of our interviewees mentioned tree size as a reason for its sanctification. In the literature (Table <xref ref-type="table" rid="T1">1</xref>) large tree size and evergreen-ness are mentioned as important characters that lead to tree veneration. In Israel at least two of the common sacred trees are deciduous (<italic>Quercus ithaburensis </italic>and <italic>Pistacia atlantica</italic>). These two species can grow to a considerable size, and it seems that this is the very reason why they were venerated. These observations are run counter to Wilson [[<xref ref-type="bibr" rid="B62">62</xref>]:6] who maintains that all the oaks on saint's graves are evergreen. In one village (Sajur, Upper Galilee) there is a blessed <italic>Styrax officinalis </italic>tree; the keeper of the tree is convinced that this specific tree is Venerated (named al-Mubarakeh, meaning "the blessed") because it is said that it is the only evergreen individual of this deciduous species (<bold>24, </bold>it is a well known story in the village, n = 12). Our observation failed to corroborate this, although it is in a protected garden and the leaf fall is somewhat shorter in comparison with other <italic>Styrax </italic>trees growing in an exposed habitat. This case brings to mind a sacred platanus (<italic>Platanus orientalis) </italic>in Gortyna that was sanctified in Ancient Greece because it was an individual evergreen plant (the species is generally deciduous) and connected with the abduction of Europa [[<xref ref-type="bibr" rid="B3">3</xref>]: 7.1; [<xref ref-type="bibr" rid="B72">72</xref>]:176].</p><p>Huge trees were objects of veneration and a manifestation of the Almighty [[<xref ref-type="bibr" rid="B23">23</xref>]:2; [<xref ref-type="bibr" rid="B73">73</xref>]:29]. The most famous "great tree" is the oak, which is the foremost tree in European mythologies and tree worship [<xref ref-type="bibr" rid="B75">75</xref>]; [<xref ref-type="bibr" rid="B76">76</xref>]:188&#x02013;191; [<xref ref-type="bibr" rid="B77">77</xref>]:23]. In the words of Folkrad [[<xref ref-type="bibr" rid="B77">77</xref>]:21] "The Oak, the strongest of all trees, has been revered as the emblem of the Supreme Being by almost all the nations of heathendom" Porteous [[<xref ref-type="bibr" rid="B6">6</xref>]:150] explains why these trees were venerated: "As year after year passed with the same continual changefulness, trees, or perhaps one outstanding tree on account of its size and age, would come to be regarded with a special reverence, and primitive imagination would people it with all sorts of beings, such as Gods, Nymphs, and Demons". The Kikuyus in East Africa select large trees for veneration. A sacred tree must be high because it is deemed nearer to the god as a medium through which prayers are to ascend [[<xref ref-type="bibr" rid="B78">78</xref>]:4]. Sometimes the reason for sanctity is the strange or the unusual appearance of the tree rather it size [Japan [79:188; Inner Mongolia [[<xref ref-type="bibr" rid="B80">80</xref>]:279,283].</p></sec><sec><title>Trees as an abode of super natural beings</title><p>Trees that are the abode of deities, gods and ancestor' souls were never recognized in the Middle East, and it seems that these reasons are confined today to polytheistic religions. The recurrent pattern of this kind of sacred tree/forest/grove is highly typical: the tree/grove/forest is dedicated to a certain god/deity or ancestral soul, which is in charge of the welfare and well being of the community/village. To ensure supernatural blessing, certain rituals must be performed to please the god or the ancestors to win their favour [[<xref ref-type="bibr" rid="B80">80</xref>]:1578; [<xref ref-type="bibr" rid="B18">18</xref>]:35; [<xref ref-type="bibr" rid="B81">81</xref>]:350&#x02013;351].</p></sec><sec><title>The tree is located near holy water source</title><p>The survey of sacred trees in Israel ([<xref ref-type="bibr" rid="B77">77</xref>] and our observations) revealed that they are rarely adjacent/related to water sources; we were unable to find any evidence that this vicinity is the reason for the sanctification. In ancient Greece [[<xref ref-type="bibr" rid="B34">34</xref>]:20, 25, 34] sacred trees were sometimes associated with holy water sources. Concerning ancient Syria White states, "this early nature worship, whether of numerous Baalim of the Syrian oases or the local nymphs of the sacred springs of Hellas, required only the marked-off enclosure of holy ground beside the spring, or about the circle of trees in the sacred grove" [[<xref ref-type="bibr" rid="B82">82</xref>]:115]. Sacred trees associated with sacred wells are common in Britain even today [[<xref ref-type="bibr" rid="B35">35</xref>]:40&#x02013;42; [<xref ref-type="bibr" rid="B83">83</xref>]:30].</p></sec></sec><sec><title>Conclusion</title><p>Of the 24 known reasons for the creation of sacred trees/groves/woods/forests 11 reasons (Table <xref ref-type="table" rid="T1">1</xref>, reasons 1 to 11) were recorded by us in the field while the other 13 were compiled from the literature and were almost never recorded from the Middle East. Only one reason (showing the way to a sacred place) was never recorded before and was found to be endemic to the Druze and is related to one specific very famous tree. Generally, most of the local reasons seem to be more confined to the monotheistic religions as "vestiges" of the old paganism (see below) and appear very rarely in polytheistic religions. Some of the remaining reasons (especially 13, 16, and 18) are typical of polytheistic religions as a part of the regular rituals and are seldom mentioned in monotheistic areas. The evidence from ancient Europe, Egypt, and the Middle East may reflect the prehistoric pagan heritage that was partly adopted later by the monotheistic faiths. These replaced tree-dwelling spirits, gods, and souls by saint worship performed through the trees. Owing to the long and aggressive struggle of the monotheistic establishments (Jewish, Christian, as well as Muslim) with the old paganism, "tree adoration" became marginal and prevails today mainly in rural areas of the Middle East and North Africa.</p><p>In the pre-Islamic pagan world tree worship was quite common [[<xref ref-type="bibr" rid="B84">84</xref>]: 169&#x02013;107; [<xref ref-type="bibr" rid="B14">14</xref>]:185; [<xref ref-type="bibr" rid="B85">85</xref>]:181]. These trees were worshipped as the abode of jinns and spirits, and were treated as possessing godly characters [[<xref ref-type="bibr" rid="B1">1</xref>]:30]. Several authors consider tree veneration in the Muslim world a relic of old heathen worship of tree-spirits or gods, which has survived, in a thinly disguised form, throughout all the ages of Christian and Islamic supremacy [[<xref ref-type="bibr" rid="B77">77</xref>]:242; [<xref ref-type="bibr" rid="B14">14</xref>]:199; [<xref ref-type="bibr" rid="B93">93</xref>]:50; [<xref ref-type="bibr" rid="B86">86</xref>]:34; [<xref ref-type="bibr" rid="B77">77</xref>]:92]. After the Arab conquest the tree spirits/deities/gods of the early heathen inhabitants were replaced, after the Arab conquest, by the spirits of the Muslim saints, the "Awlia" (= plural of Welli), which may survive and appear in sacred trees (Palestine [[<xref ref-type="bibr" rid="B77">77</xref>]:151]; Iran [[<xref ref-type="bibr" rid="B87">87</xref>], I:378]; Morocco [[<xref ref-type="bibr" rid="B88">88</xref>]:97]). A completely different view is expressed by Albright [[<xref ref-type="bibr" rid="B89">89</xref>]:284&#x02013;286] who considers the Welli cult of Palestine and Syria as merely as a phase of the saint-cult of the Mediterranean region and differing only in detail from the saint-cult of the lower classes in other Mediterranean lands. Moreover, he argues that this saint-cult goes back to the Christian saint-cult of the Eastern Roman Empire in the early Byzantine centuries and is Hellenistic-Roman, not Semitic, in origin. Hornblower [[<xref ref-type="bibr" rid="B90">90</xref>]:19] similarly notes, concerning sacred trees in Egypt, that the local gods were replaced by local saints, first Christian and then Muslim. With the crystallization of Islam, these old venerated trees were cut down and this kind of worship was strictly forbidden [[<xref ref-type="bibr" rid="B91">91</xref>]:318; [<xref ref-type="bibr" rid="B92">92</xref>]:243&#x02013;244]. The practical result, as can be seen today (Table <xref ref-type="table" rid="T1">1</xref>), may be considered a kind of "functional religious replacement": no longer are the trees regarded as the abode of tree-spirits, deities, or gods, as in earlier heathen times, but as the abode of saints, who are regarded as the messengers of God himself. This kind of "softened idolatry" exists to the present day, despite the Islamic regime, which has proved too weak to eliminate it, especially in rural areas. In consequence of our survey, we fully agree with Frazer [[<xref ref-type="bibr" rid="B93">93</xref>]:43] who, very pithily, summarizes the status of sacred trees in the Middle East: "Thus the worship at the high places and green trees, which pious Hebrew kings forbade and prophets thundered against thousands of years ago, persists apparently in the same places to this day".</p><p>A review of the reasons for the creation of sacred trees/groves (Table <xref ref-type="table" rid="T1">1</xref>) shows some kind of dichotomy between the monotheistic legacy of the sacred trees in Europe, the Middle East and North Africa as against the polytheistic traditions in the world. In polytheistic religions, especially in Africa and Asia, people still see the sacred tree/grove as the abode of deities, ancestors' spirits, etc. (see Table <xref ref-type="table" rid="T1">1</xref>), which may reflect the old "pagan" customs that prevailed in Europe, the Middle East, and North Africa at the remote past. Fergusson [[<xref ref-type="bibr" rid="B94">94</xref>]:62] noted that while all the monotheistic religions fought ancient tree worship, Buddhism elevated it to a higher level of veneration. According to Avasthi [[<xref ref-type="bibr" rid="B95">95</xref>]:7] polytheism in India allows multiplicity in worshipping objects like trees, rivers, or the village deity. These objects vary from person to person and festival to festival. The celebration is public whereas the worship of gods and goddesses takes place in the family.</p><p>In the polytheistic world the sacred grove/wood it is a centre of common tribal activities. Sometimes access is limited to certain people and/or certain occasions, and the grove is kept by the community or by a special priest, additionally to the general taboo not to harm the tree [Sierra Leone (36:311); Okinawa (96:5,18); East Timor ([<xref ref-type="bibr" rid="B97">97</xref>]:224); India ([<xref ref-type="bibr" rid="B41">41</xref>]:49; [<xref ref-type="bibr" rid="B98">98</xref>]:9; [<xref ref-type="bibr" rid="B99">99</xref>]:712); Russia ([<xref ref-type="bibr" rid="B100">100</xref>]:699); Zimbabwe ([<xref ref-type="bibr" rid="B101">101</xref>]:6); Northern Ghana ([<xref ref-type="bibr" rid="B18">18</xref>]:35); Vanatua (Pacific islands, [<xref ref-type="bibr" rid="B102">102</xref>]:9); Mozambique ([<xref ref-type="bibr" rid="B103">103</xref>]:229); West Africa ([<xref ref-type="bibr" rid="B104">104</xref>]:45)]. In the present-day Middle East the sacred tree is a centre for individual ritual behaviour with free access.</p><p>There is no doubt that the present-day vestigial tree worship in Europe is a result of "Christianization" of the old pagan religions. Elworthy [[<xref ref-type="bibr" rid="B105">105</xref>]:107&#x02013;108] noted " The remarkable similarity in the customs (of tree veneration) all over Europe points to the conclusion that tree worship was once an important element in the early religion of mankind, especially of the Aryan stock, and that the singular uniformity of the rites and ceremonies which can easily be shown to exist in widely separated countries, fully warrants us in believing that they have not much changed from very remote ages, and that the practices continued down to a very recent period by peasantry.... were substantially identical with the same rites and ceremonies observed by Egyptians, Etruscans, Greeks and Romans". At this point it is fitting to cite Lucas [[<xref ref-type="bibr" rid="B35">35</xref>]:34] "At first, in brief, the church came to the [sacred] tree, not the tree to the church". Almost the same view is expressed by Robertson-Smith [[<xref ref-type="bibr" rid="B14">14</xref>]:186&#x02013;187: "The worship of solitary trees survived the fall of the great gods of Semitic heathenism... The solitary tree may in certain cases be the last relic of a ruined heathen sanctuary". Likewise Porteous [[<xref ref-type="bibr" rid="B6">6</xref>]:162] comments that "It [tree worship] was... so deeply ingrained in the human hearts that in many cases it was utilized by the Church for its own ends by blessing the most ancient and venerated trees, and by erecting Christian altars and placing crucifixes and images the people had sacrificed to the heathen divinities" (see also [[<xref ref-type="bibr" rid="B8">8</xref>]:I,86&#x02013;87]).</p><p>The present survey shows that tree veneration is still quite common in Israel among Muslims and Druze. The reasons for sanctification of trees are mainly connected with the adoration of saints and prophets. While the Muslims connected sacred trees with saints' souls and graves, the Druze, who believe in transmigration of souls, relate the blessed tree mainly to the events and activities of prophets and historical religious leaders.</p><p>A worldwide comparison shows the great similarity of the monotheistic religions, in which saint adoration is the main focus for tree worship. In polytheistic religions sacred trees are mainly connected with local gods, spirits, demons and ancestor veneration, none of which is found in the present-day Middle East.</p></sec><sec><title>Declaration of competing interest</title><p>The author(s) declares that he has no competing interests.</p></sec><sec><title>Appendix</title><p>List of the informants who are cited personally according to theirs appearance order in the text (Bold numbers in the text to differ from the literature sources) the given data are: The name of the informant, his age, ethnic group, place, and the date of the interview.</p><p>1. Zaki Abu Bilal Hashad, 67, Muslim imam, Tars&#x00127;&#x000ee;ha, 19 Dec. 2005</p><p>2. Q&#x00101;sim Bader, 45, Druze, Keeper of the sacred place of Nabi Sabal&#x00101;n sanctuary, 9 June</p><p>2002.</p><p>3. Faraj Kiblawi, 67, Muslim, Tarsh&#x0012b;&#x00127;a, 19 Dec. 2005.</p><p>4. &#x00126;&#x00101;med abu Mustafa, 45, Muslin, 'Arr&#x00101;be, 21. Dec. 2003.</p><p>5. Ruqqiya Maghis, 50, Joreih, Bedouin. 27 March 2005.</p><p>6. Q&#x00101;sem Shibli, 32, Bedouin, Shibli 21. Oct. 2004.</p><p>7. Sa'&#x0012b;d Mu&#x00127;mmad, 90, Druze, Y&#x00101;nu&#x00127;, 25 May 2003</p><p>8. Sa'&#x0012b;d Ma&#x00127;am&#x000fb;d Sa'&#x0012b;d, 69, Druze, Y&#x00101;nu&#x00126;, 25 May 2003.</p><p>9. Akab 'Amashe, 45, Druze Sheikh, Buq' &#x00101;tha, 12 Dec. 2001.</p><p>10. Sheikh N&#x000fb;r Rifa'&#x0012b;yye, 40, Majdl Kr&#x000fb;m, Sufi Muslim, 24. June 2000.</p><p>11. Sheikh &#x00160;hah&#x0012b;n &#x00126;ussein, 70, Druze a religious leader, Beit Jan, 12. Sept. 2000.</p><p>12. '&#x000c2;del Abu &#x00126;amid, 57, Muslim, Kufr Manda, 16 April 2004.</p><p>13. &#x00126;&#x00101;med abu Mustafa, 45, Muslim, 'Arr&#x00101;be, 31. Dec. 2003.</p><p>14. Sa'di Qaram&#x00101;n, 75, Muslim, Dam&#x000fb;n, 13 Sept. 2000.</p><p>15. Jam&#x0012b;l Abu R&#x00101;'id Araf&#x00101;t, 71, Muslim, Mashhad, 23. Sept. 2004.</p><p>16. S&#x00101;leh &#x00126;at&#x0012b;b, 50, Mghar, Druze, 18 March 2003.</p><p>17. Karmel Na'ama, 52, Muslim, 'Arr&#x00101;be, 6 June 2004.</p><p>18. Sueid Hussein,68, Druze, Peqee'in, 8 Nov. 2005.</p><p>19. Sheikh Tawkf&#x0012b;q Amashe, 70, Druze, Mas'ade, 12 Dec. 2001.</p><p>20. Salm&#x00101;n Abu Rukan, 55, Druze, 'Isfia, 12 Dec. 2003.</p><p>21. Mustafa &#x00126;al&#x00101;wi, 48, Druze, 'Isfia, 15 Dec. 2003.</p><p>22. Ma&#x00127;mud Abu L'Ra&#x00127;man Mar'&#x0012b;n, 67, Muslim, Meshhed, 20 Sep. 2004.</p><p>23. Yuval Avidor, 35, Jew, Yodfat. 21 Dec. 2003.</p><p>24. Zi'ad Tall&#x00101;l, Druze, 35, Druze, Sajur, 3 Oct. 2003.</p></sec>

Carrageenan Is a Potent Inhibitor of Papillomavirus Infection

<p>Certain sexually transmitted human papillomavirus (HPV) types are causally associated with the development of cervical cancer. Our recent development of high-titer HPV pseudoviruses has made it possible to perform high-throughput in vitro screens to identify HPV infection inhibitors. Comparison of a variety of compounds revealed that carrageenan, a type of sulfated polysaccharide extracted from red algae, is an extremely potent infection inhibitor for a broad range of sexually transmitted HPVs. Although carrageenan can inhibit herpes simplex viruses and some strains of HIV in vitro, genital HPVs are about a thousand-fold more susceptible, with 50&#x00025; inhibitory doses in the low ng/ml range. Carrageenan acts primarily by preventing the binding of HPV virions to cells. This finding is consistent with the fact that carrageenan resembles heparan sulfate, an HPV cell-attachment factor. However, carrageenan is three orders of magnitude more potent than heparin, a form of cell-free heparan sulfate that has been regarded as a highly effective model HPV inhibitor. Carrageenan can also block HPV infection through a second, postattachment heparan sulfate&#x02013;independent effect. Carrageenan is in widespread commercial use as a thickener in a variety of cosmetic and food products, ranging from sexual lubricants to infant feeding formulas. Some of these products block HPV infectivity in vitro, even when diluted a million-fold. Clinical trials are needed to determine whether carrageenan-based products are effective as topical microbicides against genital HPVs.</p>

<contrib contrib-type="author"><name><surname>Buck</surname><given-names>Christopher B</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Thompson</surname><given-names>Cynthia D</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Roberts</surname><given-names>Jeffrey N</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>M&#x000fc;ller</surname><given-names>Martin</given-names></name><xref ref-type="aff" rid="aff2">2</xref></contrib><contrib contrib-type="author"><name><surname>Lowy</surname><given-names>Douglas R</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Schiller</surname><given-names>John T</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="corresp" rid="cor1">&#x0002a;</xref></contrib>

PLoS Pathogens

<sec id="s1"><title>Introduction</title><p>Papillomaviruses are a diverse group of nonenveloped DNA viruses that infect the skin and mucosal tissues of a range of vertebrate species, including humans. A group of genital mucosotropic human papillomavirus (HPV) types are etiologic agents responsible for virtually all cases of cancer of the uterine cervix, as well as a substantial fraction of other ano-genital and head-and-neck cancers (reviewed in [<xref rid="ppat-0020069-b001" ref-type="bibr">1</xref>]). Cancer-associated genital HPV types, as well as another subset of HPV types associated with the development of benign genital warts (condyloma accuminata), are generally transmitted through sexual contact. Infection with genital HPV types is very common, with an estimated lifetime risk of infection of about 75&#x00025; [<xref rid="ppat-0020069-b002" ref-type="bibr">2</xref>]. Although most genital HPV infections are subclinical and self-limiting, a subset of persistently infected individuals have lesions that progress to premalignancy or cancer.</p><p>Recent meta-analyses have suggested that condoms are, at best, only marginally effective for preventing the sexual transmission of HPV [<xref rid="ppat-0020069-b003" ref-type="bibr">3</xref>,<xref rid="ppat-0020069-b004" ref-type="bibr">4</xref>]. However, a highly effective group of prophylactic HPV vaccines are expected to become publicly available in the near future [<xref rid="ppat-0020069-b005" ref-type="bibr">5</xref>]. Two possible drawbacks to these vaccines are that they are expected to be relatively expensive (at least initially) and are likely to be papillomavirus type-restricted in their protection. Thus, the vaccines may not initially be available to women in all parts of the world and may not offer protection against all cancer-associated HPV types. Inexpensive condom-compatible compounds that could function as broad-spectrum topical microbicides targeting sexually transmitted HPVs might therefore serve as useful adjuncts to vaccination programs.</p><p>In vitro analysis of papillomavirus infection has historically been hampered by the fact that key events in the late phase of the viral lifecycle, such as the expression of the capsid proteins L1 and L2, require cellular differentiation in the upper layers of the stratified squamous epithelial tissues that the viruses inhabit (reviewed in [<xref rid="ppat-0020069-b006" ref-type="bibr">6</xref>]). As a consequence, papillomaviruses cannot replicate in conventional monolayer cell cultures. Investigation of the assembly and entry phases of the papillomavirus lifecycle has recently been simplified by the development of high-yield methods for producing papillomavirus-based gene transfer vectors, known as pseudoviruses (PsV), using conventional monolayer cell lines [<xref rid="ppat-0020069-b007" ref-type="bibr">7</xref>,<xref rid="ppat-0020069-b008" ref-type="bibr">8</xref>]. We have used PsV to develop a high-throughput screening method to identify and compare compounds with the potential to block papillomavirus infectivity in vitro [<xref rid="ppat-0020069-b009" ref-type="bibr">9</xref>].</p><p>Previous studies have shown that sulfated polysaccharides, such as heparin, cellulose sulfate, and dextran sulfate, can block the infectivity of papillomaviruses [<xref rid="ppat-0020069-b010" ref-type="bibr">10</xref>&#x02013;<xref rid="ppat-0020069-b012" ref-type="bibr">12</xref>]. For many classes of virus, including papillomaviruses, initial attachment of the virion to cultured cell lines is thought to be mediated mainly by interactions between the virion and a type of cell-surface glycosaminoglycan (GAG) known as heparan sulfate (reviewed in [<xref rid="ppat-0020069-b013" ref-type="bibr">13</xref>]). In general, sulfated polysaccharides are thought to block viral infection by chemically mimicking heparan sulfate, thereby competing against initial virion attachment to the cell surface. Many previous studies have used various types of heparin, a highly sulfated form of heparan sulfate produced by mast cells, as a model inhibitor of papillomavirus infectivity. In this report we show that carrageenan, a class of sulfated polysaccharide extracted from marine red algae (seaweed), inhibits the infectivity of genital HPV PsV in vitro with nearly a thousand-fold greater potency than heparin.</p><p>Carrageenan is in widespread commercial use as a thickening agent for a variety of foods and cosmetic products, including some brands of sexual lubricant. Since high-quality carrageenan preparations (reviewed in [<xref rid="ppat-0020069-b014" ref-type="bibr">14</xref>]) appear to have a good safety profile for long-term vaginal use [<xref rid="ppat-0020069-b015" ref-type="bibr">15</xref>,<xref rid="ppat-0020069-b016" ref-type="bibr">16</xref>], carrageenan might be useful as an inexpensive topical microbicide for blocking the sexual transmission of HPV.</p></sec><sec id="s2"><title>Results</title><sec id="s2a"><title>Testing of Candidate Microbicides, including Carrageenan</title><p>Various candidate HPV infection inhibitors were tested using a PsV-based inhibition assay [<xref rid="ppat-0020069-b009" ref-type="bibr">9</xref>]. The assay uses flow cytometric analysis to assess the inhibition of PsV-mediated delivery of a green fluorescent protein (GFP) reporter plasmid into HeLa cells. HPV16, an exceptionally oncogenic genital HPV type [<xref rid="ppat-0020069-b017" ref-type="bibr">17</xref>], was chosen as a model HPV for initial experiments.</p><p>A wide variety of compounds were screened using the inhibition assay (<xref ref-type="table" rid="ppat-0020069-t001">Tables 1</xref> and <xref ref-type="table" rid="ppat-0020069-t002">2</xref>). Although a variety of nonsulfated polysaccharides failed to inhibit the infectivity of the HPV16 PsV, most types of sulfated polysaccharides we tested were inhibitory (<xref ref-type="table" rid="ppat-0020069-t002">Table 2</xref>). A standard heparin preparation was inhibitory at doses similar to previous reports [<xref rid="ppat-0020069-b011" ref-type="bibr">11</xref>,<xref rid="ppat-0020069-b018" ref-type="bibr">18</xref>]. However, preparations of intestine- and kidney-derived heparan sulfate displayed no detectable inhibition of PsV infectivity (<xref ref-type="table" rid="ppat-0020069-t002">Table 2</xref>). Since heparan sulfate modification is complex and varies depending on tissue source (reviewed in [<xref rid="ppat-0020069-b013" ref-type="bibr">13</xref>]), it is possible that the noninhibitory heparan sulfate preparations simply lack the proper chemical features (e.g., a particular pattern of sulfation) required to inhibit infection.</p><table-wrap id="ppat-0020069-t001" content-type="2col" position="float"><label>Table 1</label><caption><p>Inhibitory Effects of Nonpolysaccharide Compounds</p></caption><graphic xlink:href="ppat.0020069.t001"/></table-wrap><table-wrap id="ppat-0020069-t002" content-type="2col" position="float"><label>Table 2</label><caption><p>Inhibitory Effects of Polysaccharide Compounds</p></caption><graphic xlink:href="ppat.0020069.t002"/></table-wrap><p>GAGs are unbranched polysaccharides primarily composed of a monotonous series of characteristic disaccharide repeats. GAGs can be divided into two broad categories: glucosaminoglycans, such as heparan sulfate, and galactosaminoglycans, such as chondroitin sulfate. As their names imply, a defining difference between the two GAG classes is their initial incorporation of <italic>N</italic>-acetyl-glucosamine or <italic>N</italic>-acetyl-galactosamine saccharide units, respectively. Another important difference between the two classes of GAG is that glucosaminoglycans are linked in a series of 1,4 saccharide bonds, whereas galactosaminoglycans are linked in an alternating series of 1,3 and 1,4 bonds. Average structures for examples of both types of GAG are shown in <xref ref-type="fig" rid="ppat-0020069-g001">Figure 1</xref>.</p><fig id="ppat-0020069-g001" position="float"><label>Figure 1</label><caption><title>Polysaccharide Structures</title><p>Idealized average structures of the disaccharide repeat units of various sulfated polysaccharides. Ac = acetyl, R = [H or SO₃ ^&#x02212;], R&#x02032; = [H, Ac, or SO₃ ^&#x02212;]. Structures adapted from [<xref rid="ppat-0020069-b013" ref-type="bibr">13</xref>], <ext-link ext-link-type="uri" xlink:href="http://www.lsbu.ac.uk/water/hycar.html">http://www.lsbu.ac.uk/water/hycar.html</ext-link>, and <ext-link ext-link-type="uri" xlink:href="http://www.sigmaaldrich.com">http://www.sigmaaldrich.com</ext-link>.</p></caption><graphic xlink:href="ppat.0020069.g001"/></fig><p>Despite the substantial chemical differences between the average structures of chondroitin 6-sulfate and heparin, preparations of the two polysaccharides displayed similar inhibitory effects against the HPV16 PsV (<xref ref-type="table" rid="ppat-0020069-t002">Table 2</xref>). In contrast, a chondroitin 4-sulfate preparation was noninhibitory (<xref ref-type="table" rid="ppat-0020069-t002">Table 2</xref> and [<xref rid="ppat-0020069-b011" ref-type="bibr">11</xref>]), despite its high degree of chemical similarity to chondroitin 6-sulfate. The results are consistent with a previous report demonstrating that chondroitin 6-sulfate (but not chondroitin 4-sulfate) competes against the interaction of noninfectious capsids with immobilized heparin [<xref rid="ppat-0020069-b010" ref-type="bibr">10</xref>].</p><p>Various types of carrageenan were by far the most potent inhibitors identified in the screen (<xref ref-type="table" rid="ppat-0020069-t002">Table 2</xref>). Carrageenan is an unbranched polysaccharide composed of galactose derivatives arranged in an alternating series of 1,3 and 1,4 saccharide linkages reminiscent of the pattern seen in galactosaminoglycans (<xref ref-type="fig" rid="ppat-0020069-g001">Figure 1</xref>). In addition to having a similar pattern of saccharide linkages, the typical sulfation pattern of &#x003ba;-type carrageenan also closely resembles chondroitin 4-sulfate, with an average of one 4-O-linked sulfate group per disaccharide repeat in both types of polysaccharide (<xref ref-type="fig" rid="ppat-0020069-g001">Figure 1</xref>). Despite these apparent chemical similarities to the noninhibitory chondroitin 4-sulfate, &#x003ba;-carrageenan was an extremely potent inhibitor, with 50&#x00025; inhibitory concentration (IC₅₀) values in the low ng/ml range.</p><p>Like heparin, &#x003bb;- and &#x003b9;-carrageenan types are more heavily sulfated than most tissue-derived heparan sulfate (reviewed in [<xref rid="ppat-0020069-b013" ref-type="bibr">13</xref>]). On average, these carrageenan types exhibited somewhat greater inhibitory potency than &#x003ba;-carrageenan (<xref ref-type="table" rid="ppat-0020069-t002">Table 2</xref>).</p></sec><sec id="s2b"><title>The Influence of Capsid Dose on Carrageenan Inhibitory Effects</title><p>Assuming a typical carrageenan chain length of about a thousand saccharide units (reviewed in [<xref rid="ppat-0020069-b014" ref-type="bibr">14</xref>]), the roughly 5 ng/ml IC₅₀ of &#x003b9;-carrageenan preparations corresponds to a concentration of about 20 pM. The standard PsV-based inhibition assay uses an HPV16 capsid inoculum of 1 ng of the major capsid protein, L1, per milliliter (also ~20 pM). Thus, the observed IC₅₀s for the various types of carrageenan occurred under conditions where there was only a slight mass (or molar) excess of carrageenan over L1. If the inhibitory effects of carrageenan were due to its direct binding to the capsid, the IC₅₀ might be expected to shift if excess capsids were added to the assay. Consistent with this hypothesis, the addition of &#x0201c;cold&#x0201d; capsids (i.e., PsV produced in the absence of a GFP reporter plasmid) resulted in a dose-dependent reduction in the apparent inhibitory potency of &#x003b9;-carrageenan preparation C4014 (<xref ref-type="fig" rid="ppat-0020069-g002">Figure 2</xref>). This was not true for the inhibitory effects of heparin, presumably because IC₅₀s for heparin occur under conditions of substantial inhibitor excess, thus satisfying the law of mass action.</p><fig id="ppat-0020069-g002" position="float"><label>Figure 2</label><caption><title>Capsid Dose Influences Carrageenan IC₅₀ </title><p>Inhibition assays were performed using a standard dose of GFP-expressing HPV16 PsV in the presence of increasing doses of cold capsids. Overall capsid dose is given as the final concentration of the major capsid protein, L1, in the culture medium.</p></caption><graphic xlink:href="ppat.0020069.g002"/></fig><p>The observation that excess capsids can increase the IC₅₀ for carrageenan strongly implies that the primary inhibitory mechanism involves the direct binding of carrageenan to the capsid. To further investigate this hypothesis, we performed capsid pull-down experiments using &#x003b9;-carrageenan beads. The results show that purified HPV16 capsids directly bind carrageenan in phosphate buffer containing &#x02264; 0.4 M NaCl (<xref ref-type="fig" rid="ppat-0020069-g003">Figure 3</xref>, top panel).</p><fig id="ppat-0020069-g003" position="float"><label>Figure 3</label><caption><title>Capsids Bind Carrageenan</title><p>Carrageenan beads were incubated with HPV16 or HPV5 capsids in buffers with the NaCl concentration shown. The beads were washed, then bound capsids were eluted and visualized in stained SDS-PAGE gels. Bovine serum albumin (BSA, 65 kDa) was used as a control for nonspecific binding to the beads. For HPV16, an L2 band can be seen above the L1 band.</p></caption><graphic xlink:href="ppat.0020069.g003"/></fig><p>Taken together, the results suggest the possibility that carbohydrate-binding motifs on the capsid surface are highly selective for a limited subset of sulfated polysaccharide sequences. Although it is tempting to speculate that the &#x003b9;-carrageenan structure depicted in <xref ref-type="fig" rid="ppat-0020069-g001">Figure 1</xref> represents an ideal (or near-ideal) binding substrate, it should be noted that the biochemistry of GAG and carrageenan modification is complex, and preparations of these molecules can contain localized sequences of alternatively modified saccharide residues that differ substantially from the idealized structures depicted in <xref ref-type="fig" rid="ppat-0020069-g001">Figure 1</xref>. Thus, the possibility that inhibition of the PsV is mediated by a subset of atypical saccharide sequences cannot be ruled out.</p></sec><sec id="s2c"><title>Inhibitory Effects of &#x003b9;-Carrageenan against Various Papillomavirus Types</title><p>The effectiveness of sulfated polysaccharides for blocking the in vitro infectivity of other viruses, such as HIV-1 and dengue viruses, varies dramatically according to which viral strain is used [<xref rid="ppat-0020069-b019" ref-type="bibr">19</xref>&#x02013;<xref rid="ppat-0020069-b021" ref-type="bibr">21</xref>]. Thus, it seemed possible that the inhibitory effects of carrageenan might also vary for different papillomavirus types.</p><p>A variety of papillomavirus types are available as PsV. The particle-to-infectivity ratios for stocks of different PsV types vary substantially. We therefore used HPV16 cold capsids to generate a standard curve comparing the IC₅₀ of &#x003b9;-carrageenan inhibition to total capsid dose (<xref ref-type="fig" rid="ppat-0020069-g004">Figure 4</xref>). Compared to this standard curve, PsV based on three other cancer-associated genital HPV types, 18, 31, and 45, exhibited a similar degree of susceptibility to inhibition by &#x003b9;-carrageenan. PsV based on HPV6, a relatively nononcogenic type that can cause genital warts, also showed similar susceptibility to &#x003b9;-carrageenan when compared to the HPV16 PsV.</p><fig id="ppat-0020069-g004" position="float"><label>Figure 4</label><caption><title>Standardized Carrageenan IC₅₀ for Various Papillomavirus Types</title><p>Points represent carrageenan IC₅₀ of infectivity, except for empty red triangles, which represent the carrageenan IC₅₀ of cell binding for HPV16 capsids covalently linked to a fluorescent dye. Empty circles represent carrageenan IC₅₀ of infectivity observed using HaCaT cells instead of HeLa cells. Error bars represent the 95&#x00025; CI for the IC₅₀.</p></caption><graphic xlink:href="ppat.0020069.g004"/></fig><p>Papillomaviruses are divided phylogenetically into more than a dozen genera, with genital HPV types occupying a single genus, alpha [<xref rid="ppat-0020069-b022" ref-type="bibr">22</xref>]. PsV are currently available for papillomaviruses from three other genera: delta (bovine papillomavirus type 1 [BPV1]), kappa (cottontail rabbit papillomavirus [CRPV]) and beta (HPV5). In contrast to the genital HPVs, PsV based on BPV1 and CRPV, which cause nongenital skin lesions in host animals, were about a hundred-fold less susceptible to inhibition by carrageenan (<xref ref-type="fig" rid="ppat-0020069-g004">Figure 4</xref>).</p><p>Surprisingly, HPV5, which typically infects nongenital skin without causing overt symptoms, was not inhibited by &#x003b9;-carrageenan or heparin, even at doses of up to 100 &#x003bc;g/ml. Despite this extreme resistance to inhibition by carrageenan, HPV5 PsV were found to bind directly to &#x003b9;-carrageenan beads (<xref ref-type="fig" rid="ppat-0020069-g003">Figure 3</xref>, bottom panel).</p><p>The same hierarchy of inhibition was seen for the various papillomavirus types when the assay was performed using other cell lines, including the spontaneously immortalized human keratinocyte line HaCaT, human 293TT cells, or murine C127 cells (<xref ref-type="fig" rid="ppat-0020069-g004">Figure 4</xref> and unpublished data).</p><p>To verify that &#x003b9;-carrageenan is active against an authentic papillomavirus, we performed assays examining the focal transformation of C127 cells by BPV1 [<xref rid="ppat-0020069-b023" ref-type="bibr">23</xref>,<xref rid="ppat-0020069-b024" ref-type="bibr">24</xref>]. These experiments confirmed that &#x003b9;-carrageenan can inhibit the infectivity of an authentic papillomavirus with an IC₅₀ of between 1 and 10 &#x003bc;g/ml (unpublished data), consistent with capsid dose-adjusted IC₅₀s observed for BPV1 PsV.</p></sec><sec id="s2d"><title>The Influence of pH on Carrageenan Inhibitory Effects</title><p>Since carrageenan might have utility as a topical microbicide for preventing the sexual transmission of HPVs, it is important to consider the fact that human vaginal pH is typically below 4.5 (reviewed in [<xref rid="ppat-0020069-b025" ref-type="bibr">25</xref>]). We therefore performed inhibition assays examining the inhibitory effects of &#x003b9;-carrageenan in culture medium buffered to pH 4.5 or 5.0 with lactic or acetic acid, respectively. &#x003b9;-Carrageenan remained effective for blocking infectivity under acidic conditions, with IC₅₀ 95&#x00025; confidence intervals (CIs) of 17&#x02013;20, 13&#x02013;16, or 2.4&#x02013;6.1 ng/ml at pH 4.5, 5.0, and 7.4, respectively.</p></sec><sec id="s2e"><title>&#x003b9;-Carrageenan Can Block Postbinding Events</title><p>To investigate the concept that &#x003b9;-carrageenan might block infectivity by preventing the initial attachment of capsids to cells, we performed flow cytometric analysis of cells exposed to fluorescently labeled HPV16 capsids in the presence of various doses of carrageenan. As expected, &#x003b9;-carrageenan blocked the binding of labeled capsids at concentrations similar to those capable of blocking infectivity when standardized for capsid dose (<xref ref-type="fig" rid="ppat-0020069-g004">Figure 4</xref>). Similar results were obtained using HaCaT cells as a binding target (unpublished data).</p><p>In addition to blocking the initial attachment of virions to cells, heparin can also block the infectivity of cell-bound PsV for many hours after initial attachment to cells [<xref rid="ppat-0020069-b011" ref-type="bibr">11</xref>]. To investigate the possibility that &#x003b9;-carrageenan also exerts additional, postattachment inhibitory effects on PsV infectivity, we performed time course experiments in which cells with prebound PsV were exposed to &#x003b9;-carrageenan. As seen in the 2-h timepoint in <xref ref-type="fig" rid="ppat-0020069-g005">Figure 5</xref>, cell-bound HPV16 PsV remained entirely susceptible to inhibition by somewhat higher doses of &#x003b9;-carrageenan. Half of the infectious titer remained susceptible to inhibition by &#x003b9;-carrageenan for up to 12 h after initial binding to cells. Similar results were observed when HaCaT cells were used as an infection target (unpublished data). The results demonstrate that, in addition to blocking the initial interaction of capsids with cells, &#x003b9;-carrageenan also exerts a postattachment inhibitory effect on infectivity.</p><fig id="ppat-0020069-g005" position="float"><label>Figure 5</label><caption><title>Carrageenan Addition Time Course</title><p>Cells were incubated with HPV16 PsV for 2 h, followed by washout of the virus inoculum. Carrageenan was added at the timepoints shown, where time zero represents initial PsV inoculation.</p></caption><graphic xlink:href="ppat.0020069.g005"/></fig></sec><sec id="s2f"><title>Postbinding Inhibitory Effects of &#x003b9;-carrageenan Do Not Involve Heparan Sulfate</title><p>The postattachment inhibitory effects of &#x003b9;-carrageenan could be due either to displacement of capsids from heparan sulfate proteoglycans (HSPGs), or due to disruption of HSPG-independent steps in the viral infectious pathway. To address this issue, we made use of a GAG-negative cell line known as pgsA-745. The line was created by chemical mutagenesis of Chinese hamster ovary (CHO) cells, resulting in the disruption of a xylosyltransferase gene that is required for the first peptide glycosylation step in the synthesis of all GAGs, including heparan sulfate [<xref rid="ppat-0020069-b026" ref-type="bibr">26</xref>]. Fluorescent capsid binding studies (unpublished data) confirmed a previous report showing that pgsA-745 cells bind HPV capsids relatively poorly compared to parental CHO cells [<xref rid="ppat-0020069-b010" ref-type="bibr">10</xref>]. Despite this reduction in bulk capsid binding, pgsA-745 cells could be infected to the same extent as parental CHO cells if a 50-fold higher inoculum of HPV16 PsV was used (data not shown).</p><p>Using these conditions, we examined the effects of &#x003b9;-carrageenan on HPV16 PsV prebound to the two cell lines for 3 h. Both cell lines displayed similar infectious IC₅₀ values, with 95&#x00025; CIs of 14&#x02013;61 ng/ml for CHO and 13&#x02013;21 ng/ml for pgsA-745. The result directly demonstrates that &#x003b9;-carrageenan can disrupt steps in the HPV16 infectious pathway that do not involve HSPGs.</p></sec><sec id="s2g"><title>PsV Inhibition by Consumer Products Containing Carrageenan</title><p>The widespread use of carrageenan in various consumer products led us to wonder whether lubricants intended for sexual use might employ carrageenan as a gelling agent. Internet searches revealed a number of sexual lubricant products that list carrageenan, or other algal polysaccharides, as ingredients. Although most lubricated condom brands (and some sexual lubricant brands) do not publicize their ingredients, one condom brand, Chapeau Crystal Carrageenan (Fuji Latex Co., Tokyo, Japan), advertises its use of a carrageenan-based lubricating gel [<xref rid="ppat-0020069-b027" ref-type="bibr">27</xref>].</p><p>Various sexual lubricant gels were subjected to testing in the HPV16 PsV inhibition assay. Several of the lubricant products were extremely potent inhibitors, with IC₅₀ values occurring at dilutions of a few million&#x02013;fold (<xref ref-type="table" rid="ppat-0020069-t003">Table 3</xref>). Two highly inhibitory European brands, Bioglide and Bioglide Anal, list carrageenan as an ingredient. The other components in these two products, water, glycerol, and xanthan gum, were noninhibitory when tested individually (<xref ref-type="table" rid="ppat-0020069-t002">Tables 2</xref> and <xref ref-type="table" rid="ppat-0020069-t003">3</xref>).</p><table-wrap id="ppat-0020069-t003" content-type="2col" position="float"><label>Table 3</label><caption><p>Inhibitory Effects of Consumer Products</p></caption><graphic xlink:href="ppat.0020069.t003"/></table-wrap><p>One US product, Divine N&#x000b0; 8, uses the term &#x0201c;natural kelp extract&#x0201d; in its ingredient list. Although it is not clear what type of algal polysaccharide this term refers to, the high potency of Divine N&#x000b0; 8 in the HPV16 PsV inhibition assay strongly suggests that the product contains carrageenan. An unscented version of the product, Divine N&#x000b0; 9, does not list ingredients on its packaging or at its manufacturer's website, but its high potency suggests that it, too, may contain carrageenan. The inhibition curves for the Bioglide and Divine lubricants were similar to what would be expected if the products were composed of roughly 1&#x00025;&#x02013;3&#x00025; carrageenan, a typical concentration range used to achieve gelation.</p><p>The packaging of a third US product, ForPlay Gel Plus, lists carrageenan as its fifth ingredient, behind several nonsulfated thickening agents. ForPlay Gel Plus did not display detectable inhibitory effects, even at dilutions as little as a hundred-fold. It is possible that ForPlay Gel Plus either contains very little carrageenan, or contains a less-inhibitory type of carrageenan. Alternatively, interactions between carrageenan and other compounds in ForPlay Gel Plus may abrogate the inhibitory effects of the carrageenan.</p><p>Lubricant brands containing other algal polysaccharides, such as agar and algin, were less effective for blocking the HPV16 PsV in the inhibition assay, consistent with the observation that these compounds are less inhibitory than carrageenan when tested individually (<xref ref-type="table" rid="ppat-0020069-t002">Table 2</xref>).</p><p>Several lubricant gels that do not contain sulfated polysaccharides were ineffective for blocking the HPV16 PsV at tested doses. Ortho Options Conceptrol, a contraceptive gel containing the detergent spermicide nonoxynol-9, does not contain sulfated polysaccharides. The fact that Conceptrol was ineffective for blocking the HPV16 PsV at noncytotoxic doses (<xref ref-type="table" rid="ppat-0020069-t003">Table 3</xref>) is consistent with a previous report demonstrating that nonoxynol-9 is not effective for blocking papillomavirus infectivity [<xref rid="ppat-0020069-b028" ref-type="bibr">28</xref>].</p><p>Another common use for carrageenan is as a stabilizing agent in milk-based products, including infant feeding formulas. The use of infant formulas containing carrageenan might thus be a factor in vertical transmission of HPVs, since such transmission could involve establishment of initial infection in infants' oral mucosa. We therefore tested several brands of infant formula using the HPV16 PsV inhibition assay. Formulas containing carrageenan displayed significant inhibitory effects, while fresh milk and infant formulas without carrageenan did not display detectable inhibitory effects at tested doses (<xref ref-type="table" rid="ppat-0020069-t003">Table 3</xref>).</p></sec></sec><sec id="s3"><title>Discussion</title><p>In this report we demonstrate that carrageenan, an inexpensive commercial thickening agent extracted from seaweed, is an exceptionally potent inhibitor of papillomavirus infectivity in vitro. Carrageenan was found to be active against a range of common sexually transmitted HPV types that can cause cervical cancer and genital warts. Since carrageenan is generally recognized as safe for food and topical applications, it is an appealing candidate for use as a broad-spectrum topical microbicide to block HPV transmission.</p><p>Some, but not all, carrageenan-containing over-the-counter sexual lubricant gels we tested were extremely effective for blocking the infectivity of an HPV16 reporter pseudovirus in vitro. These results raise the possibility that use of such lubricant products, or condoms lubricated with carrageenan-based gels, could block the sexual transmission of HPV. However, in the absence of clinical efficacy data, it would be inappropriate to recommend currently available products for use as topical microbicides.</p><p>Carrageenan is also active in vitro and in murine model systems against other viruses, including herpes simplex viruses and some strains of HIV-1 [<xref rid="ppat-0020069-b029" ref-type="bibr">29</xref>&#x02013;<xref rid="ppat-0020069-b034" ref-type="bibr">34</xref>]. However, in vitro IC₅₀ values for carrageenan inhibition of herpes simplex virus and HIV-1 infectivity are about a thousand-fold higher than the IC₅₀s we have observed for carrageenan inhibition of genital HPVs in vitro.</p><p>It is important to emphasize that cell culture systems may not fully represent some aspects of HPV infection of keratinocytes in vivo. However, our group has recently developed a pseudovirus-based murine genital challenge model for initial HPV infection (unpublished data). This animal model system should be useful for investigating of the potential efficacy of carrageenan for blocking HPV transmission in vivo.</p><p>A clinical trial focused on the effectiveness of a &#x003ba;/&#x003bb;-carrageenan preparation as a topical microbicide is currently in progress in South Africa. A recent patent application by the trial's organizers (<ext-link ext-link-type="uri" xlink:href="http://www.popcouncil.org">http://www.popcouncil.org</ext-link>) contains a claim of carrageenan as a papillomavirus inhibitor, but the potency of the inhibitory effect was not indicated [<xref rid="ppat-0020069-b035" ref-type="bibr">35</xref>]. Since the principal focus of the ongoing trial is the efficacy of carrageenan against HIV-1, it may be necessary to develop additional clinical trials specifically focused on the in vivo efficacy of well-defined carrageenan preparations against HPVs. The high rate of acquisition of genital HPV infection in young adult populations (reviewed in [<xref rid="ppat-0020069-b002" ref-type="bibr">2</xref>]) might make it possible to perform short-duration clinical efficacy trials with relatively small numbers of human subjects.</p><p>Our results show that the principal mechanism by which carrageenan blocks papillomavirus infectivity is via the direct binding of carrageenan to the viral capsid. The binding of carrageenan appears to block interactions between the capsid and cell-surface HSPG attachment factors. Although the presence of HSPGs on the cell surface significantly enhances papillomavirus binding to and infection of most types of cultured cell lines [<xref rid="ppat-0020069-b010" ref-type="bibr">10</xref>,<xref rid="ppat-0020069-b011" ref-type="bibr">11</xref>,<xref rid="ppat-0020069-b036" ref-type="bibr">36</xref>,<xref rid="ppat-0020069-b037" ref-type="bibr">37</xref>], in this report we have used GAG-negative cells to demonstrate conclusively that HSPG attachment factors are not strictly required for infection to occur. A similar situation has been described for certain strains of HIV-1, particularly lab-adapted HIV-1 strains, for which HSPGs are thought to serve as attachment factors that facilitate (but are ultimately dispensable for) the in vitro infection of some cultured cell lines [<xref rid="ppat-0020069-b038" ref-type="bibr">38</xref>,<xref rid="ppat-0020069-b039" ref-type="bibr">39</xref>].</p><p>In addition to blocking the initial interaction between papillomavirus virions and HSPGs, carrageenan also exerts a second HSPG-independent inhibitory effect. This secondary inhibitory effect could be due to occlusion of virion surfaces involved in binding to cellular proteins involved in the infectious process. Alternatively, carrageenan might interfere with the development of needed conformational changes within the virion. Since it is possible that HPVs use alternative, non-HSPG attachment factors in vivo [<xref rid="ppat-0020069-b036" ref-type="bibr">36</xref>,<xref rid="ppat-0020069-b037" ref-type="bibr">37</xref>], the existence of a postattachment, GAG-independent inhibitory effect increases the likelihood that carrageenan might ultimately be effective as a topical microbicide against HPVs.</p><p>Although carrageenan was highly effective for neutralizing five different genital HPV types in vitro, it was substantially less potent against several papillomavirus types tropic for nongenital skin. Since common genital HPVs occupy a single genus, and the three nongenital papillomavirus types we have tested are phylogenetically distant from the genital types [<xref rid="ppat-0020069-b022" ref-type="bibr">22</xref>], it is tempting to speculate that all HPVs tropic for the genital mucosa would be comparably susceptible to inhibition by carrageenan. However, the possibility that some genital HPVs might exhibit natural resistance to inhibition by carrageenan would be an important factor to consider in the design of clinical efficacy trials.</p><p>Recurrent respiratory papillomatosis is a rare but debilitating HPV-induced condition involving the formation of large benign tumors on airway surfaces. The main treatment for the disorder is surgical removal of recurring obstructive masses. Juvenile onset recurrent respiratory papillomatosis (JORRP) is thought to be the result of vertical transmission of genital wart-associated HPV types 6 or 11 during birth (reviewed in [<xref rid="ppat-0020069-b040" ref-type="bibr">40</xref>]). The fact that some common infant formulas contain carrageenan raises the possibility that such formulas might function as inhibitors of the initial establishment of papillomavirus infection in newborns' oro-laryngeal epithelium [<xref rid="ppat-0020069-b041" ref-type="bibr">41</xref>]. Since human breast milk offers infants a wide variety of health benefits and JORRP is rare (roughly four cases per 100,000 births), it would be inappropriate to consider using infant formula for the purpose of preventing JORRP. However, the apparent safety of infant formulas containing carrageenan suggests that a pharmaceutical-grade carrageenan-based gel might be safe for perinatal cervico-vaginal application. The infrequency of JORRP, and its long initial latency period, would make it an impractical endpoint to use for clinical efficacy trials investigating the application of carrageenan-based gels. However, asymptomatic vertical transmission of genital HPV types, which is thought to be relatively common (reviewed in [<xref rid="ppat-0020069-b042" ref-type="bibr">42</xref>]), might be used as a surrogate endpoint that could readily be monitored by HPV DNA testing of infant buccal swabs.</p></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Cell culture.</title><p>All cell lines were cultured in DMEM (Invitrogen, Carlsbad, California, United States) supplemented with 10&#x00025; 56 &#x000b0;C heat-inactivated fetal calf serum (Hyclone, South Logan, Utah, United States), nonessential amino acids, and Glutamax-I (Invitrogen) (DMEM-10). CHO-K1 and pgsA-745 (American Type Culture Collection, Manassas, Virginia, United States) [<xref rid="ppat-0020069-b026" ref-type="bibr">26</xref>] were maintained in DMEM-10 supplemented with an additional 100 &#x003bc;M proline.</p></sec><sec id="s4b"><title>Pseudovirus production.</title><p>Nucleotide maps of plasmids used in this work, as well as detailed protocols, are available at our laboratory website (<ext-link ext-link-type="uri" xlink:href="http://home.ccr.cancer.gov/lco/default.asp">http://home.ccr.cancer.gov/lco/default.asp</ext-link>). Various types of GFP-expressing pseudoviruses were produced according to previously described methods [<xref rid="ppat-0020069-b007" ref-type="bibr">7</xref>&#x02013;<xref rid="ppat-0020069-b009" ref-type="bibr">9</xref>,<xref rid="ppat-0020069-b043" ref-type="bibr">43</xref>,<xref rid="ppat-0020069-b044" ref-type="bibr">44</xref>]. Briefly, 293TT cells were transfected with plasmids expressing the papillomavirus major and minor capsid proteins, L1 and L2, together with a GFP-expressing reporter plasmid, pfwB [<xref rid="ppat-0020069-b008" ref-type="bibr">8</xref>]. All PsV were produced using codon-modified L1 and L2 genes, except for HPV31 PsV, which used expression constructs based on wild-type L1 and L2 open reading frames. The high particle-to-infectivity ratio of HPV31 PsV stocks (<xref ref-type="fig" rid="ppat-0020069-g004">Figure 4</xref>) is likely due to relatively poor expression of L2 (unpublished data). Codon-modified HPV45 L1 and L2 genes (p45L1w and p45L2w) were constructed based on sequencing of an HPV45 molecular clone. HPV16 PsV were produced using a previously unreported bicistronic L1/L2 expression plasmid, p16sheLL. Capsids were allowed to mature overnight in cell lysate, then purified using Optiprep gradients. The L1 protein content of PsV stocks was determined by comparison to bovine serum albumin standards in Coomassie-stained NuPAGE gels.</p><p>Fluorescently tagged capsids were generated by covalently conjugating Alexa Fluor 488 carboxylic acid, succinimidyl ester (Molecular Probes, Eugene, Oregon, United States) to HPV16 PsV, according to the manufacturer's instructions. Cell-binding inhibition results were also confirmed using fluorescent capsids generated by incorporation of an L2-GFP fusion protein [<xref rid="ppat-0020069-b009" ref-type="bibr">9</xref>]. Both types of fluorescent capsid displayed particle-to-infectivity ratios similar to wild-type HPV16 PsV (unpublished data).</p></sec><sec id="s4c"><title>Inhibition assays.</title><p>Nonpolysaccharide candidate inhibitors were purchased from the suppliers shown in <xref ref-type="table" rid="ppat-0020069-t001">Table 1</xref>. Polysaccharides were purchased from Sigma-Aldrich (St. Louis, Missouri, United States), except for agarose (Invitrogen), agar (Difco, Detroit, Michigan, United States), and cellulose sulfate (Acros, Geel, Belgium).</p><p>Compounds were tested using a previously described PsV-based papillomavirus inhibition assay [<xref rid="ppat-0020069-b009" ref-type="bibr">9</xref>]. Briefly, HeLa cells were plated at 6,000 cells/well in 50 &#x003bc;l of medium in 96-well plates. Candidate inhibitors were dissolved at 1&#x02013;10 mg/ml (or reconstituted as directed by manufacturer) in sterile water or appropriate solvent, then subjected to a ten-point three-fold serial dilution covering an appropriate concentration range. Diluted candidate inhibitor (50 &#x003bc;l) was added to preplated HeLa cells, followed by 50 &#x003bc;l of diluted PsV stock. The cells were incubated for 24 h, fed by addition of 100 &#x003bc;l of DMEM-10, then subjected to flow cytometric analysis 48&#x02013;56 h after infection. IC₅₀ values and 95&#x00025; CI were determined using Prism (GraphPad Software, San Diego, California, United States) to calculate a variable slope sigmoidal dose-response curve.</p><p>PsV doses were calibrated such that between 5&#x00025; and 20&#x00025; of cells scored as GFP^&#x0002b; when no inhibitors were added. For HPV16, a standard dose with a final concentration of 1 ng/ml L1 (about 750 capsid equivalents of L1 per cell) resulted in fluorescence of about 10&#x00025; of cells in the &#x0201c;no inhibitor&#x0201d; condition. Other papillomavirus types were used at the final L1 concentrations shown in <xref ref-type="fig" rid="ppat-0020069-g004">Figure 4</xref>. HPV5 PsV infection of HeLa cells was performed with a dose of 35 ng/ml L1. Relative to HeLa cells, appropriate infection of HaCaT cells required a two- to three-fold higher capsid dose for HPVs 16, 18, and 45 and a two- to three-fold lower capsid dose for HPVs 5 and 6.</p><p>Cytotoxicity was defined as a &#x0003e;50&#x00025; reduction in the net turnover of the metabolic substrate WST-1 (Roche, Indianapolis, Indiana, United States) and/or by the appearance of dramatic alterations in the microscopic appearance of cell morphology at the time of harvest.</p><p>The effect of acidic conditions on the inhibitory activity of &#x003b9;-carrageenan was analyzed by inoculating HeLa cells with HPV16 PsV (1 ng/ml L1) together with various doses of &#x003b9;-carrageenan in bicarbonate-free RPMI (Invitrogen) buffered to pH 7.4 with 10 mM HEPES, pH 5.0 with 10 mM acetic acid, or pH 4.5 with 10 mM lactic acid. The virus inoculum was removed after 2 h at 37 &#x000b0;C (ambient CO₂) and replaced with DMEM-10 without carrageenan. Cell viability and PsV infectivity under acidic conditions were not significantly different from the neutral RPMI control.</p><p>Inhibition of capsid-to-cell binding was analyzed using HeLa and HaCaT cells detached using Cellstripper (Mediatech, Herndon, Virginia, United States), a proprietary, nonenzymatic chelating buffer. Suspended cells (3 &#x000d7; 10⁴ per condition) were incubated for 1 h at 37 &#x000b0;C with fluorescent capsids in a 150 &#x003bc;l volume of DMEM-10 supplemented various doses of &#x003b9;-carrageenan. The cells were then washed and subjected to flow cytometric analysis. Binding IC₅₀ was calculated based on net geometric mean fluorescence intensity using Prism software.</p><p>Time course analyses were performed in 96-well plates by incubating HPV16 PsV (1 ng/ml L1) with preplated HeLa cells for 2 h at 37 &#x000b0;C, followed by two washes to remove unbound virus. Various doses of carrageenan in DMEM-10 were then added to the cultures at the timepoints shown in <xref ref-type="fig" rid="ppat-0020069-g005">Figure 5</xref>. For the 0-h timepoint, carrageenan was added to cultures immediately prior to virus inoculation, and a second carrageenan dose was added after washout of the inoculum. For the 2-h timepoint, carrageenan was added immediately after washout of the inoculum.</p><p>CHO-K1 and pgsA-745 cells were preplated overnight at 50,000 cells/well in 24-well plates. Cells were inoculated with 250 &#x003bc;l of HPV16 PsV at 45 ng/ml L1 (pgsA-745) or 0.9 ng/ml L1 (CHO-K1). The plates were incubated at 37 &#x000b0;C and gently swirled every 20 min for 3 h. The cells were then washed, and placed in 0.5 ml of medium containing various doses of carrageenan. The cultures were fed by addition of 2 ml of medium with no carrageenan after 24 h, and subjected to flow cytometric analysis 56 h after initial PsV inoculation.</p></sec><sec id="s4d"><title>Capsid pull-down experiments.</title><p>Crosslinked 3&#x00025; &#x003b9;-carrageenan beads (Bioscience Beads, West Warwick, Rhode Island, United States) were used to perform HPV5 and HPV16 carrageenan direct-binding studies. HPV5 or HPV16 L1 (2 &#x003bc;g in the form of purified PsV) were incubated for 1 h at room temperature with 50 &#x003bc;l of carrageenan bead slurry pre-equilibrated into 1 ml of Dulbecco's PBS supplemented with 0.01&#x00025; Tween-80 [<xref rid="ppat-0020069-b045" ref-type="bibr">45</xref>] and various concentrations of NaCl. BSA (4 &#x003bc;g) was used as a negative control. The beads were washed 3 &#x000d7; 10 min in Dulbecco's PBS containing appropriate concentrations of NaCl, followed by a final wash with plain Dulbecco's PBS. The beads were eluted by incubation at 65 &#x000b0;C in NuPage Load Dye (Invitrogen) with 8&#x00025; 2-mercaptoethanol. Samples (10 &#x003bc;l out of about 75 &#x003bc;l total) were separated on NuPage gels (Invitrogen) and visualized using SYPRO Ruby Stain (Molecular Probes). Comparison of the samples to 1 &#x003bc;l of BenchMark Protein Ladder (Invitrogen) suggested an overall L1 recovery of roughly 75&#x00025; in the 0.15 M NaCl condition (<xref ref-type="fig" rid="ppat-0020069-g003">Figure 3</xref>).</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting Information</title><sec id="s5a"><title>Accession Numbers</title><p>The GenBank (<ext-link ext-link-type="uri" xlink:href="http://www.ncbi.nlm.nih.gov/Genbank">http://www.ncbi.nlm.nih.gov/Genbank</ext-link>) accession number for our sequencing of the L1 and L2 genes of an HPV45 molecular clone is DQ080002.</p></sec></sec>

DC-SIGN on B Lymphocytes Is Required For Transmission of HIV-1 to T Lymphocytes

<p>Infection of T cells by HIV-1 can occur through binding of virus to dendritic cell (DC)-specific ICAM-3 grabbing nonintegrin (DC-SIGN) on dendritic cells and transfer of virus to CD4^&#x0002b; T cells. Here we show that a subset of B cells in the blood and tonsils of normal donors expressed DC-SIGN, and that this increased after stimulation in vitro with interleukin 4 and CD40 ligand, with enhanced expression of activation and co-stimulatory molecules CD23, CD58, CD80, and CD86, and CD22. The activated B cells captured and internalized X4 and R5 tropic strains of HIV-1, and mediated <italic>trans</italic> infection of T cells. Pretreatment of the B cells with anti&#x02013;DC-SIGN monoclonal antibody blocked <italic>trans</italic> infection of T cells by both strains of HIV-1. These results indicate that DC-SIGN serves as a portal on B cells for HIV-1 infection of T cells in <italic>trans</italic>. Transmission of HIV-1 from B cells to T cells through this DC-SIGN pathway could be important in the pathogenesis of HIV-1 infection.</p>

<contrib contrib-type="author"><name><surname>Rappocciolo</surname><given-names>Giovanna</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="corresp" rid="cor1">&#x0002a;</xref></contrib><contrib contrib-type="author"><name><surname>Piazza</surname><given-names>Paolo</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Fuller</surname><given-names>Craig L</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="author-notes" rid="n104">&#x000a4;</xref></contrib><contrib contrib-type="author"><name><surname>Reinhart</surname><given-names>Todd A</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Watkins</surname><given-names>Simon C</given-names></name><xref ref-type="aff" rid="aff2">2</xref></contrib><contrib contrib-type="author"><name><surname>Rowe</surname><given-names>David T</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Jais</surname><given-names>Mariel</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Gupta</surname><given-names>Phalguni</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff3">3</xref></contrib><contrib contrib-type="author"><name><surname>Rinaldo</surname><given-names>Charles R</given-names><suffix>Jr.</suffix></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff3">3</xref></contrib>

PLoS Pathogens

<sec id="s1"><title>Introduction</title><p>HIV-1 can bind to the type II C-type lectin receptor, dendritic cell (DC)-specific ICAM-3 grabbing nonintegrin (DC-SIGN; CD209), on myeloid DC and be transferred to CD4^&#x0002b; T cells [<xref rid="ppat-0020070-b001" ref-type="bibr">1</xref>,<xref rid="ppat-0020070-b002" ref-type="bibr">2</xref>]. An important feature of this <italic>trans</italic> pathway is that the virus does not establish an efficient, productive infection in these DC. Rather, it is captured by the DC and internalized in distinct intracellular compartments, and then transmitted to CD4^&#x0002b; T cells wherein it undergoes productive replication [<xref rid="ppat-0020070-b003" ref-type="bibr">3</xref>]. This is considered to be an alternative pathway to HIV-1 <italic>cis</italic> infection of T cells, macrophages, and DC that occurs through binding to the primary CD4 receptor and either of the chemokine coreceptors CXCR4 or CCR5.</p><p>B lymphocytes have also been implicated in <italic>trans</italic> infection of T cells with HIV-1 [<xref rid="ppat-0020070-b004" ref-type="bibr">4</xref>]. Given the intimate association of B cells and T cells in lymphoid tissue, B cell&#x02013;mediated <italic>trans</italic> infection pathways could be important in the spread of virus to T cells. B cells derived either from lymphoid tissue or from the peripheral blood of HIV-1&#x02013;infected persons carry replication-competent virus of either the CXCR4 (X4) or CCR5 (R5) tropic strain [<xref rid="ppat-0020070-b005" ref-type="bibr">5</xref>]. The mechanism by which B cells have been shown to transmit the virus involves binding of HIV-1 immune complexes to CR2 or CD21 on the surface of B cells and subsequent passage to the T cells [<xref rid="ppat-0020070-b006" ref-type="bibr">6</xref>&#x02013;<xref rid="ppat-0020070-b010" ref-type="bibr">10</xref>]. Other reports have proposed a role for B cells in HIV-1 infection involving B cell activation processes induced by <italic>nef</italic>-expressing macrophages [<xref rid="ppat-0020070-b011" ref-type="bibr">11</xref>].</p><p>B cells express some C-type lectin receptors [<xref rid="ppat-0020070-b012" ref-type="bibr">12</xref>&#x02013;<xref rid="ppat-0020070-b015" ref-type="bibr">15</xref>], with conflicting reports on expression of DC-SIGN [<xref rid="ppat-0020070-b016" ref-type="bibr">16</xref>,<xref rid="ppat-0020070-b017" ref-type="bibr">17</xref>]. Transfer of HIV-1 by B cells via a C-type lectin receptor pathway described for DC could be of significance in HIV-1 pathogenesis. We therefore investigated B cells for a DC-SIGN&#x02013;mediated, <italic>trans</italic> pathway for HIV-1 infection of CD4^&#x0002b; T cells. We found that a subset of B cells in the peripheral blood and tonsils of healthy, HIV-1 seronegative donors expressed DC-SIGN, and that DC-SIGN expression increased after stimulation with interleukin 4 (IL-4) and CD40 ligand (CD40L). The stimulated, DC-SIGN^&#x0002b; B cells mediated <italic>trans</italic> infection of T cells.</p></sec><sec id="s2"><title>Results</title><sec id="s2a"><title>DC-SIGN Expression in B cells and Enhancement by Stimulation with IL-4 and CD40L</title><p>Our initial, three-color flow cytometric analysis of peripheral blood mononuclear cells (PBMC) showed that DC-SIGN was expressed on a small but distinct subset of CD19^&#x0002b; B cells within the CD45^&#x0002b;/CD19^&#x0002b; gated population of PBMC of normal, HIV-1&#x02013;negative persons (representative example, <xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>A). To extend this finding, we analyzed DC-SIGN expression on B cells that were purified from PBMC of 33 normal donors by sorting with anti-CD19 monoclonal antibody (mAb)-coated magnetic beads. Flow cytometric results showed that DC-SIGN was expressed on 7.9 &#x000b1; 1.8&#x00025; (mean &#x000b1; standard error [SE]) of purified, peripheral blood CD20^&#x0002b; B cells (representative example, <xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>B, T0; cumulative results, <xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>C, T0).</p><fig id="ppat-0020070-g001" position="float"><label>Figure 1</label><caption><title>Expression of DC-SIGN on B Cells</title><p>(A) B cells within the PBMC population expressed DC-SIGN. PBMC were assessed for expression of CD45, CD19, and DC-SIGN by flow cytometry. Cells coexpressing CD45 and CD19 (left) were analyzed for DC-SIGN coexpression (right). Quadrant positions were determined with the isotype controls.</p><p>(B) Increase in coexpression of DC-SIGN on CD20^&#x0002b; cells detected by flow cytometry in unstimulated, purified, fresh B cells (0 h; T0) and IL-4&#x02013; and CD40L-stimulated, purified B cells at 24 h (T24). Data are from a representative, healthy HIV-1&#x02013;seronegative donor.</p><p>(C) DC-SIGN expression on B cells from healthy, HIV-1&#x02013;seronegative donors (<italic>n</italic> = 33, left), HIV-1&#x02013;seropositive, ART therapy&#x02013;naive subjects (<italic>n</italic> = 10, middle), and HIV-1&#x02013;seropositive subjects receiving ART (<italic>n</italic> = 10, right). Quadrants were positioned on the isotype controls.</p><p>(D) CD40L and IL-4 acted synergistically in inducing enhanced DC-SIGN expression in B cells by 24 h. In this representative experiment (top), purified B cells were cultured in the presence of IL-4 or CD40L or a combination of the two. Untreated B cells were used as controls (NT). DC-SIGN and CD20 coexpression on activated B cells was greatest using a combination of 1,000 U/ml of IL4 and 1 &#x003bc;g/ml of CD40L for 24 h (bottom). Single concentrations of 100, 500, and 2,000 U/ml of IL-4 and 0.1, 0.5, and 10 &#x003bc;g/ml of CD40L induced similar, low levels of DC-SIGN expression (unpublished data); various combinations of these concentrations of IL-4 and CD40L induced less DC-SIGN expression than this combination (unpublished data).</p></caption><graphic xlink:href="ppat.0020070.g001"/></fig><p>We next addressed whether expression of DC-SIGN on this subset of B cells was related to a state of cellular activation. For this, we determined the proportion of B cells that expressed CD23, another type II C-type lectin receptor that is a low-affinity Fc receptor for IgE (Fc epsilon RII) and is associated with B cell activation and differentiation [<xref rid="ppat-0020070-b015" ref-type="bibr">15</xref>]. We found that DC-SIGN and CD23 were coexpressed on 8.1 &#x000b1; 1.7&#x00025; of B cells in normal persons (<xref ref-type="supplementary-material" rid="ppat-0020070-sg001">Figure S1</xref>A, T0). Indeed, 80&#x00025; of DC-SIGN^&#x0002b; B cells also expressed CD23 (unpublished data). To further delineate the relationship of B cell activation to DC-SIGN, we examined whether stimulation of B cells with the T helper type 2 (Th2) cytokine interleukin 4 (IL-4) alone or in combination with the CD4^&#x0002b; T cell activation factor CD40L could alter expression of DC-SIGN and CD23, a procedure that mimics activation of B cells by CD4^&#x0002b; T helper cells during antigen processing [<xref rid="ppat-0020070-b018" ref-type="bibr">18</xref>,<xref rid="ppat-0020070-b019" ref-type="bibr">19</xref>]. Also, IL-4 has been shown to up-regulate expression of DC-SIGN on monocyte-derived DC [<xref rid="ppat-0020070-b020" ref-type="bibr">20</xref>], breast-milk macrophages [<xref rid="ppat-0020070-b021" ref-type="bibr">21</xref>], and the THP-1 cell line [<xref rid="ppat-0020070-b022" ref-type="bibr">22</xref>]. Our data demonstrate that treatment of purified B cells with a range of concentrations of either IL-4 or CD40L alone had little effect on the expression of DC-SIGN or CD23 by 24 h (<xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>D, top; <xref ref-type="supplementary-material" rid="ppat-0020070-sg001">Figure S1</xref>B, top) or 48 h (unpublished data). In contrast, treatment with a combination of different concentrations of IL-4 and CD40L had a synergistic effect on expression of DC-SIGN and CD23 on B cells, with the greatest increase in the number of B cells expressing DC-SIGN being induced by stimulation with 1,000 U/ml of IL-4 and 1 &#x003bc;g/ml of CD40L for 24 h (<xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>D, top and bottom, respectively; <xref ref-type="supplementary-material" rid="ppat-0020070-sg001">Figure S1</xref>B, bottom). This treatment was therefore used for stimulation of B cells in subsequent experiments.</p><p>Cumulative data from 33 normal donors showed that expression of DC-SIGN by purified, peripheral blood CD20^&#x0002b; B cells significantly increased after stimulation of the cells for 24 h with IL-4 and CD40L (representative example, <xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>B) from a mean level of 7.9 &#x000b1; 1.8&#x00025; to 28.2 &#x000b1; 3.3&#x00025; (<italic>p</italic> = 0.0001) (<xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>C). Moreover, the mean fluorescence intensity (MFI) of DC-SIGN on the B cells increased 3-fold after stimulation, from 10.8 &#x000b1; 2.4 to 30.2 &#x000b1; 6.6 (<italic>p</italic> &#x0003c; 0.005). The DC-SIGN^&#x0002b; CD23^&#x0002b; B cell population also increased from 8.1 &#x000b1; 1.7&#x00025; at 0 h to 25.7 &#x000b1; 3.2&#x00025; at 24 h (<italic>n</italic> = 33; <italic>p</italic> &#x0003c; 0.0002; <xref ref-type="supplementary-material" rid="ppat-0020070-sg001">Figure S1</xref>A, T24), with 86&#x00025; of DC-SIGN^&#x0002b; cells expressing CD23 after 24 h of stimulation (unpublished data).</p><p>To confirm the increase in DC-SIGN expression in B lymphocytes stimulated by IL-4 and CD40L, we measured the level of DC-SIGN mRNA by real-time reverse transcriptase PCR [<xref rid="ppat-0020070-b023" ref-type="bibr">23</xref>,<xref rid="ppat-0020070-b024" ref-type="bibr">24</xref>] in B cells stimulated with IL-4 and CD40L compared to unstimulated B cells. B cells from five of seven normal donors (donors 1, 2, 4, 5, and 7) had large, 6.1- to 122.4-fold increases in levels of DC-SIGN mRNA, accompanied by an increase in surface expression of DC-SIGN (<xref ref-type="table" rid="ppat-0020070-t001">Table 1</xref>). B lymphocytes from the other two donors (donors 3 and 6) had low-to-moderate, 1.1- to 2.1-fold increases in DC-SIGN mRNA levels together with increases in surface expression of DC-SIGN after stimulation.</p><table-wrap id="ppat-0020070-t001" content-type="2col" position="float"><label>Table 1</label><caption><p>Real-Time RT-PCR Measurement of DC-SIGN mRNA Levels in B cells of Seven Healthy, HIV-1 Seronegative Donors</p></caption><graphic xlink:href="ppat.0020070.t001"/></table-wrap><p>We next examined whether HIV-1 infection altered the number of B cells expressing DC-SIGN in the peripheral blood of HIV-1&#x02013;infected persons, and the capacity of their B cells to respond to stimulation with IL-4 and CD40L. We found that the percentage of DC-SIGN^&#x0002b; B cells in the blood of HIV-1&#x02013;infected persons with chronic HIV-1 infection who were antiretroviral therapy (ART) naive or those who had suppressed viral infection on ART was similar to uninfected persons (<xref ref-type="fig" rid="ppat-0020070-g001">Figure 1</xref>C). However, expression of DC-SIGN was not enhanced on B cells from the HIV-1&#x02013;infected subjects on ART in response to stimulation with IL-4 and CD40L, whereas DC-SIGN expression was enhanced in B cells from HIV-1&#x02013;uninfected persons and ART-naive, HIV-1&#x02013;infected subjects.</p><p>Next we compared the level of expression of DC-SIGN to other surface molecules that are known to increase during B cell activation and that play an important role in the interaction between B and T lymphocytes. Flow cytometry analysis of purified, IL-4&#x02013; and CD40L-stimulated B cells for expression of DC-SIGN, CD23, the B cell signal transduction molecule CD22, and T cell co-stimulatory molecules CD58, CD80, and CD86 showed that there was an increase in coexpression of all of these markers with DC-SIGN at 24 h compared to 0 h (<italic>p</italic> &#x0003c; 0.005; T0 and T24, <xref ref-type="fig" rid="ppat-0020070-g002">Figure 2</xref>A) or mock-stimulated B cells at 24 h (<italic>p</italic> &#x0003c; 0.05; unpublished data).</p><fig id="ppat-0020070-g002" position="float"><label>Figure 2</label><caption><title>Coexpression of DC-SIGN and Markers of Activation on B Cells</title><p>Expression of CD20, CD22, CD23, CD58, CD80, and CD86 in purified, unstimulated ([A] time 0 h; T0) and IL-4&#x02013; and CD40L-stimulated ([B] time 24 h; T24) B cells. The quadrants were set on the isotype controls. Expression of CD4, CXCR4, and CCR5 is shown for unstimulated (full histograms) and stimulated (24 h; red profile histograms), purified B lymphocytes. Isotype controls are indicated by gray lines. Data are from a single experiment representative of five independent experiments.</p></caption><graphic xlink:href="ppat.0020070.g002"/></fig><p>Finally, we determined the expression of CD4 and the chemokine receptors CXCR4 and CCR5 on B cells, which are the primary receptor and coreceptors involved in HIV-1 <italic>cis</italic> infection. Although a small subset of B cells from the blood expressed low levels of CD4, this marker was not detectable after 24 h (<xref ref-type="fig" rid="ppat-0020070-g002">Figure 2</xref>B) or 48 h (unpublished data) of stimulation of the B cells with IL-4 and CD40L. High levels of expression of the CXCR4 coreceptor for HIV-1 were evident on most B cells before and after stimulation with IL-4 and CD40L, whereas the CCR5 coreceptor was not expressed.</p><p>Taken together, these results indicate that a distinct population of activated B lymphocytes constitutively expressed DC-SIGN in the blood of normal donors. Moreover, there was a significant increase in the frequency and intensity of DC-SIGN&#x02013;expressing B cells derived from HIV-1&#x02013;negative subjects and HIV-1&#x02013;infected persons not on ART, which was not observed in HIV-1&#x02013;infected persons on ART, after 24 h of stimulation with IL-4 and CD40L. Finally, the activated B cells expressed the CXCR4 coreceptor for HIV-1, but not the primary CD4 receptor or the CCR5 coreceptor for the virus.</p></sec><sec id="s2b"><title>B Cells Express Sufficient Levels of DC-SIGN for HIV-1 <italic>trans</italic> Infection of T Cells</title><p>It has been shown in both DC and transfected cell lines that surface expression of a minimum of approximately 60,000 molecules of DC-SIGN is necessary for transmission of HIV-1 to T cells [<xref rid="ppat-0020070-b025" ref-type="bibr">25</xref>]. We therefore determined the level of expression of DC-SIGN on blood-derived B cells as compared to DC and Raji&#x02013;DC-SIGN cells by antibody binding capacity (ABC) and the number of molecules of equivalent soluble fluorochrome (MESF). The results show that IL-4&#x02013; and CD40L-stimulated B cells expressed 137,870 &#x000b1; 22,432 ABC (<italic>n</italic> = 6) for DC-SIGN compared to mock-stimulated B cells (36,946 &#x000b1; 1,125, <italic>n</italic> = 6), DC (213,350 &#x000b1; 43,370 ABC; <italic>n</italic> = 6), and Raji&#x02013;DC-SIGN cells (225,750 &#x000b1; 19,880 ABC; <italic>n</italic> = 4) (<xref ref-type="fig" rid="ppat-0020070-g003">Figure 3</xref>A). A similar quantitative expression of DC-SIGN was observed on activated B cells by MESF (<xref ref-type="fig" rid="ppat-0020070-g003">Figure 3</xref>B). The MESF and ABC histogram profiles for DC-SIGN on activated B cells compared to DC and Raji&#x02013;DC-SIGN cells are shown in <xref ref-type="fig" rid="ppat-0020070-g003">Figure 3</xref>C. These data support that IL-4&#x02013; and CD40L-stimulated B cells expressed a sufficient number of DC-SIGN molecules for transfer of HIV-1 to T cells.</p><fig id="ppat-0020070-g003" position="float"><label>Figure 3</label><caption><title>Quantitation of DC-SIGN Expression on B Cells</title><p>ABC and MESF (panels [A] and [B], respectively) of unstimulated (resting [rB]) and activated (aB) B cells and DC from six donors, and four independent samplings of Raji&#x02013;DC-SIGN cells, were determined as described in Materials and Methods. <italic>p</italic> &#x0003c; 0.008 for ABC and <italic>p</italic> &#x0003c; 0.001 for MESF comparing rB to aB cells; <italic>p</italic> = NS comparing aB cells to DC or Raji&#x02013;DC-SIGN cells. The red lines indicate the mean values for ABC and MESF. In panel (C), the histogram profiles of expression of DC-SIGN on activated B cells and DC from one representative donor are shown, as well as the profile of expression of DC-SIGN on Raji&#x02013;DC-SIGN cells for comparison (full histograms: DC-SIGN positive cells, black line overlay histograms: isotype control).</p></caption><graphic xlink:href="ppat.0020070.g003"/></fig></sec><sec id="s2c"><title>Activated B Cells Transmit HIV-1 X4 and R5 Strains to T Lymphocytes</title><p>Our finding that B lymphocytes expressed DC-SIGN led us to investigate whether these cells could be exploited by HIV-1 as a means for enhanced infection of T cells, as has been demonstrated for DC [<xref rid="ppat-0020070-b001" ref-type="bibr">1</xref>,<xref rid="ppat-0020070-b026" ref-type="bibr">26</xref>]. We first examined whether HIV-1 was able to bind to B cells and be transmitted to T lymphocytes through DC-SIGN using a low concentration of HIV-1 (multiplicity of infection [MOI] = 10^&#x02212;4, corresponding to 10 pg of p24 per 10⁶ cells), similar to that used in studies of DC-SIGN&#x02013;related transmission of HIV-1 from DC to T cells [<xref rid="ppat-0020070-b001" ref-type="bibr">1</xref>,<xref rid="ppat-0020070-b027" ref-type="bibr">27</xref>]. This low amount of virus usually does not result in efficient <italic>cis</italic> infection of T cells. As displayed in <xref ref-type="fig" rid="ppat-0020070-g004">Figure 4</xref>A, when the purified IL-4&#x02013; and CD40L-stimulated B cells were incubated with X4 tropic HIV-1 (strain IIIb) and then co-cultured with autologous CD4^&#x0002b; T cells, virus replicated in the cultures as shown by an increase from undetectable, &#x0003c;1 &#x000d7; 10¹ pg of p24 per ml at day 4, to &#x0003e;4 &#x000d7; 10³ pg at day 16. Purified B cells not stimulated with IL-4 and CD40L, and loaded with HIV-1 X4 and cultured with T cells for 24 h, were not capable of enhancing HIV-1 infection in the co-cultures. Levels of HIV-1 X4 remained below detection in IL-4&#x02013; and CD40L-stimulated B cells, mock-stimulated B cells, and T cells alone (<xref ref-type="fig" rid="ppat-0020070-g004">Figure 4</xref>A). This indicates that virus replication in the B&#x02013;T cell co-cultures was not a result of direct, <italic>cis</italic> infection of the T cells or B cells by HIV-1. For these HIV-1 transmission experiments, we used B cells obtained from magnetic bead&#x02013;purified fractions with &#x0003e;96&#x00025; CD20^&#x0002b; cells and &#x0003c;1&#x00025; T cells and monocytes. However, to ensure that the observed results were not due to contamination with other cell types, we also found that fractions of &#x0003e;96&#x00025; pure B cells obtained by flow cytometry sorting were able to transmit HIV-1 to T cells with the same efficiency as the magnetic bead&#x02013;purified B cells (unpublished data).</p><fig id="ppat-0020070-g004" position="float"><label>Figure 4</label><caption><title>Activated B Cells Transmit HIV-1 X4 and R5 to T Cells</title><p>(A) Levels of HIV-1 p24 in B cells stimulated with IL-4 and CD40L (activated B cells [aB]) or mock stimulated (resting B cells [rB]) for 24 h and loaded with 10^&#x02212;4 MOI of HIV-1 X4 (IIIb) for 2 h at 37 &#x000b0;C, then extensively washed in cold medium and incubated with purified, autologous CD4^&#x0002b; T cells (T). B cells and T cells directly loaded with HIV-1 served as controls.</p><p>(B) Dose response of HIV-1 <italic>trans</italic> infection shown by levels of p24 in cultures of autologous T cells (T) mixed with activated B cells (B) that were loaded with 10^&#x02212;5, 10^&#x02212;4, or 10^&#x02212;3 MOI of HIV-1 X4 (IIIb). The results shown are representative of three independent experiments with B and T cells from different donors.</p><p>(C) Activated B cells (B) transmitted both HIV-1 X4 (IIIb) and R5 (Ba-L) strains to autologous T cells (T) as shown by increases in levels of p24. T cells and B cells were separately cultured with either strain of virus as controls. Data are represented as the means of triplicates &#x000b1; SE, and are from single experiments representative of five or more independent experiments.</p></caption><graphic xlink:href="ppat.0020070.g004"/></fig><p>Further experiments demonstrated that IL-4&#x02013; and CD40L-activated B cells loaded with three different concentrations of HIV-1 X4 resulted in a dose response&#x02013;related level of infection in the B&#x02013;T cell co-cultures (<xref ref-type="fig" rid="ppat-0020070-g004">Figure 4</xref>B). Levels of HIV-1 p24 increased from undetectable, &#x0003c;1 &#x000d7; 10¹ pg per ml on day 4, to &#x02265;5 &#x000d7; 10⁴ pg of p24 per ml on day 16 at the highest input MOI (i.e., 10^&#x02212;3), and &#x0003e;1.7 &#x000d7; 10⁴ and &#x0003e;8 &#x000d7; 10³ pg per ml at the lower, 10^&#x02212;4 and 10^&#x02212;5 MOI, respectively. Virus did not replicate at any of these input concentrations in B cells or T cells alone (<xref ref-type="fig" rid="ppat-0020070-g004">Figure 4</xref>B).</p><p>We next investigated whether expression of CD4, CXCR4, or CCR5 by B cells was related to <italic>trans</italic> infection of T cells. As shown in <xref ref-type="fig" rid="ppat-0020070-g002">Figure 2</xref>B and in the literature [<xref rid="ppat-0020070-b028" ref-type="bibr">28</xref>], low numbers of unstimulated B cells in blood express very low levels of CD4. Moreover, expression of CD4 was down-regulated after stimulation of the B cells with IL-4 and CD40L for 24 h (<xref ref-type="fig" rid="ppat-0020070-g002">Figure 2</xref>B). We next demonstrated that activated B cells enhanced <italic>trans</italic> infection of CD4^&#x0002b; T cells with a low MOI of the HIV-1 R5 tropic, Ba-L strain (<xref ref-type="fig" rid="ppat-0020070-g004">Figure 4</xref>C), comparable to HIV-1 X4, even though B cells do not express CCR5 (<xref ref-type="fig" rid="ppat-0020070-g002">Figure 2</xref>B) [<xref rid="ppat-0020070-b028" ref-type="bibr">28</xref>]. Moreover, <italic>cis</italic> infection of either unstimulated B cells (unpublished data), stimulated B cells (<xref ref-type="fig" rid="ppat-0020070-g004">Figure 4</xref>C), or T cells (<xref ref-type="fig" rid="ppat-0020070-g004">Figure 4</xref>C) with HIV-1 R5 at this low MOI did not result in virus replication.</p><p>Taken together, our data indicate that IL-4&#x02013; and CD40L-activated B cells from HIV-1&#x02013;negative persons mediated efficient <italic>trans</italic> infection of autologous CD4^&#x0002b; T cells with either X4 or R5 tropic HIV-1. Furthermore, this <italic>trans</italic> infection was not related to CD4, CXCR4, or CCR5 expression on the B cells. The fact that these were purified B cells from HIV-1&#x02013;negative persons indicates that HIV-1 immune complexes and other types of cells were not involved in this process. It should also be noted that we did not include the lectin polybrene, a commonly used, receptor-independent enhancer of HIV-1 infection [<xref rid="ppat-0020070-b029" ref-type="bibr">29</xref>], in any of the cultures.</p></sec><sec id="s2d"><title>B Cell&#x02013;Mediated Transmission of HIV-1 to T Cells Is Dependent on Expression of DC-SIGN by the B Cells</title><p>Based on the above results, we reasoned that DC-SIGN could be involved in B cell&#x02013;mediated <italic>trans</italic> infection of T cells. To address this hypothesis, we first examined the relationship between DC-SIGN expression on purified B cells and uptake of HIV-1. Virus was expressed together with DC-SIGN in activated B cells after 2 h of exposure of the cells to virus as detected by immunofluorescent microscopy (<xref ref-type="fig" rid="ppat-0020070-g005">Figure 5</xref>A). Quantitative immunofluorescence analysis by flow cytometry showed that 80&#x00025; of B cells expressing HIV-1 p24 were DC-SIGN^&#x0002b; (<xref ref-type="fig" rid="ppat-0020070-g005">Figure 5</xref>B). Our results also demonstrated that there was no alteration of DC-SIGN expression by B cells or DC loaded with HIV-1, although there was minimal down-regulation of DC-SIGN expression on HIV-1&#x02013;infected, Raji&#x02013;DC-SIGN cells by 24 h, using an MOI of 10^&#x02212;4 (unpublished data) or 10^&#x02212;3 (<xref ref-type="fig" rid="ppat-0020070-g005">Figure 5</xref>C). This is in contrast to our recent findings that human herpesvirus 8 infects DC and macrophages via DC-SIGN and results in loss of surface expression of DC-SIGN [<xref rid="ppat-0020070-b030" ref-type="bibr">30</xref>]. Thus, after exposure of IL-4&#x02013; and CD40L-activated B cells to HIV-1, the virus is associated with DC-SIGN^&#x0002b; but not DC-SIGN^&#x02212; B cells, and there is no alteration in DC-SIGN expression in these cells as detected by flow cytometry.</p><fig id="ppat-0020070-g005" position="float"><label>Figure 5</label><caption><title>Expression of DC-SIGN and HIV-1 in B Cells</title><p>(A) Immunofluorescent microscopy showing coexpression of DC-SIGN and HIV-1 in IL-4 and CD40L purified, activated B cells after 2 h of exposure to HIV-1 X4 (MN) at 37 &#x000b0;C. B cells not exposed to HIV-1 were used as controls; these included a subset that expressed DC-SIGN and were negative for HIV-1 (unpublished data). (600&#x000d7; magnification) Green indicates DC-SIGN, red indicates p24, and blue indicates DAPI.</p><p>(B) Flow cytometry histograms showing coexpression of DC-SIGN and HIV-1 in IL-4 and CD40L purified, activated B cells after 2 h of exposure to HIV-1 X4 (MN) at 37 &#x000b0;C.</p><p>(C) HIV-1 R5 (Ba-L) had no effect on expression of DC-SIGN by purified, activated B cells (left) or DC (middle), and minimally inhibited DC-SIGN expression in Raji&#x02013;DC-SIGN cells (right) after 24 h incubation with the virus. Full histogram represents uninfected B cells; purple overlay histogram represents HIV-1&#x02013;infected cells; and black overlay histogram represents isotype controls. Data are from single experiments representative of five independent experiments.</p></caption><graphic xlink:href="ppat.0020070.g005"/></fig><p>We next determined the role of DC-SIGN on B cells in <italic>trans</italic> infection of T cells by treating IL-4&#x02013; and CD40L-activated B cells with anti&#x02013;DC-SIGN mAb or two nonspecific IgG as controls, one with the same isotype as the anti&#x02013;DC-SIGN mAb (IgG2b) and the other of a different isotype (IgG1), prior to incubation with HIV-1 X4 or HIV-1 R5 and co-culture with T cells. The results show that pretreatment of B cells with anti&#x02013;DC-SIGN mAb inhibited HIV-1 X4 and R5 replication in the B&#x02013;T cell co-cultures, whereas pretreatment with the control IgG had no effect (<xref ref-type="fig" rid="ppat-0020070-g006">Figures 6</xref>A and <xref ref-type="fig" rid="ppat-0020070-g006">6</xref>B). Also, inhibition of HIV-1 transmission by anti&#x02013;DC-SIGN mAb was dose-dependent (<xref ref-type="fig" rid="ppat-0020070-g006">Figure 6</xref>C). Because T cells do not express DC-SIGN [<xref rid="ppat-0020070-b017" ref-type="bibr">17</xref>], this virus-inhibitory effect was related to blocking of DC-SIGN on the activated B cells. These results indicate that HIV-1 X4 and R5 strains can be associated with B cells via DC-SIGN and that this leads to <italic>trans</italic> infection of T cells.</p><fig id="ppat-0020070-g006" position="float"><label>Figure 6</label><caption><title>B Cell&#x02013;Mediated Transmission of HIV X4 and R5 to T Cells Is Blocked by Anti&#x02013;DC-SIGN mAb</title><p>(A) and (B) HIV-1 p24 levels in cultures of activated B cells that were treated with anti&#x02013;DC-SIGN mAb, incubated with either HIV-1 X4 (IIIb) (<xref ref-type="fig" rid="ppat-0020070-g006">Figure 6</xref>A) or R5 (Ba-L) (<xref ref-type="fig" rid="ppat-0020070-g006">Figure 6</xref>B) for 2 h at 37 &#x000b0;C, washed and co-cultured with autologous T cells. B and T cells cultured alone with each virus were used as controls.</p><p>(C) HIV-1 p24 levels in cultures of activated B cells that were incubated with decreasing amounts of anti&#x02013;DC-SIGN mAb prior to exposure to HIV-1 R5 (Ba-L) and culture with autologous T cells for 8 d. Treatment with mouse IgG or an unrelated mAb (anti-CD11a) had no effect on HIV-1 R5 (Ba-L) <italic>trans</italic> infection.</p><p>(D) HIV-1 p24 levels in cultures of activated B cells incubated with anti&#x02013;DC-SIGN mAb or anti-CXCR4 mAb, washed, loaded with HIV-1 X4 (IIIb), and co-cultured with autologous T cells for 12 d.</p><p>Data are represented as mean of triplicates &#x000b1; SE. Data are from single experiments representative of eight independent experiments.</p></caption><graphic xlink:href="ppat.0020070.g006"/></fig><p>Although activated B cells did not express the primary CD4 receptor or the CCR5 coreceptor necessary for conventional infection by HIV-1 R5 (<xref ref-type="fig" rid="ppat-0020070-g002">Figure 2</xref>B), they did express high levels of the CXCR4 coreceptor for HIV-1 (<xref ref-type="fig" rid="ppat-0020070-g002">Figure 2</xref>B) [<xref rid="ppat-0020070-b028" ref-type="bibr">28</xref>]. Even though this is an inefficient receptor for HIV-1 infection in the absence of CD4 [<xref rid="ppat-0020070-b031" ref-type="bibr">31</xref>], we examined whether HIV-1 X4 <italic>trans</italic> infection of T cells was related to CXCR4 expression on the B cells. We treated activated B cells with anti-CXCR4 mAb or an IgG control, and showed that this did not inhibit HIV-1 X4 <italic>trans</italic> infection of T cells, whereas <italic>trans</italic> infection was blocked by pretreatment of the B cells with anti&#x02013;DC-SIGN mAb (<xref ref-type="fig" rid="ppat-0020070-g006">Figure 6</xref>D).</p><p>Taken together, these results show that DC-SIGN expression on purified, IL-4&#x02013; and CD40L-activated, peripheral blood B cells is directly related to <italic>trans</italic> infection of T cells. We hypothesized that this process could occur during B&#x02013;T cell interactions in lymphatic tissues. In support of this, we found that DC-SIGN was expressed by 26.4 &#x000b1; 6&#x00025; (<italic>n</italic> = 5) of single-cell suspensions of purified tonsil B cells (representative example, T0, <xref ref-type="fig" rid="ppat-0020070-g007">Figure 7</xref>A). In contrast to blood B cells, there was little coexpression of CD23 by freshly purified, tonsil B cells, but comparable up-regulation of CD23 and coexpression with DC-SIGN after 24-h stimulation with IL-4 and CD40L (T24, <xref ref-type="fig" rid="ppat-0020070-g007">Figure 7</xref>A). Importantly, <italic>trans</italic> infection of T cells was mediated by the stimulated tonsil B cells, as demonstrated by an increase in p24 levels from &#x0003c;1 &#x000d7; 10¹ pg per ml at day 4 to &#x0003e;3 &#x000d7; 10³ pg per ml at days 8 to12 in the B&#x02013;T cell co-cultures (<xref ref-type="fig" rid="ppat-0020070-g007">Figure 7</xref>B). Over 90&#x00025; of this virus replication was blocked by pretreatment of the B cells with anti&#x02013;DC-SIGN mAb. Thus, both DC-SIGN expressing blood and lymphatic tissue B cells can mediate efficient <italic>trans</italic> infection of T cells.</p><fig id="ppat-0020070-g007" position="float"><label>Figure 7</label><caption><title>DC-SIGN&#x02013;Expressing Tonsil B Cells Mediate <italic>trans</italic> Infection of T Cells</title><p>(A) Expression of CD20 and DC-SIGN or CD23 and DC-SIGN in fresh (time 0 h; T0) and IL-4&#x02013; and CD40L-activated (time 24 h; T24) B cells derived from tonsils. Data are from one experiment representative of five independent experiments.</p><p>(B) Levels of HIV-1 p24 in co-cultures of B and T cells derived from tonsils. B cells were stimulated with IL-4 and CD40L for 24 h, treated with anti&#x02013;DC-SIGN mAb or mouse Ig (20 &#x003bc;g/ml) or left untreated, and loaded with 10^&#x02212;4 MOI of HIV-1 R5 (Ba-L) for 2 h at 37 &#x000b0;C, then extensively washed in cold medium and incubated with purified, autologous CD4^&#x0002b; T cells (T). B cells and T cells directly loaded with HIV-1 served as controls. Amount of HIV-1 p24 in the cell culture supernatants was determined by ELISA. Data are from one experiment representative of two independent experiments.</p></caption><graphic xlink:href="ppat.0020070.g007"/></fig></sec><sec id="s2e"><title>HIV-1 Is Internalized in Activated B Cells</title><p>HIV-1 is predominantly found within cytoplasmic vacuoles in DC after binding to DC-SIGN [<xref rid="ppat-0020070-b003" ref-type="bibr">3</xref>]. This could relate to different pathways of HIV-1 that are required for subsequent <italic>trans</italic> infection of T cells. We therefore conducted a series of experiments to assess whether HIV-1 was internalized by B cells. We first examined the physical association of HIV-1 with purified, activated B cells and DC by electron microscopy. As expected, no viral particles were observed in activated B cells not loaded with HIV-1 (<xref ref-type="fig" rid="ppat-0020070-g008">Figure 8</xref>A). Activated B cells incubated with HIV-1 for 1 h at 4 &#x000b0;C had viral particles bound only to their outer surface (<xref ref-type="fig" rid="ppat-0020070-g008">Figure 8</xref>B). However, activated B cells exposed to HIV-1 for 2 h at 37 &#x000b0;C had viral particles internalized in vacuoles of the cell cytoplasm (<xref ref-type="fig" rid="ppat-0020070-g008">Figure 8</xref>C), with no intracellular virus apparent outside of vacuoles. Both intact viral particles and possible degraded virions were present in the vacuoles, similar to those observed after internalization of HIV-1 in DC (<xref ref-type="fig" rid="ppat-0020070-g008">Figure 8</xref>D; [<xref rid="ppat-0020070-b003" ref-type="bibr">3</xref>]).</p><fig id="ppat-0020070-g008" position="float"><label>Figure 8</label><caption><title>HIV-1 Is Internalized in B Cells</title><p>(A&#x02013;D) Transmission electron microscopy of HIV-1&#x02013;loaded, activated B cells and DC. Activated B cells were left untreated (panel A) or loaded with AT-2&#x02013;inactivated HIV-1 X4 (MN) for 1 h at 4 &#x000b0;C (panel B) or at 37 &#x000b0;C for 2 h (panel C). Immature DC (panel D) that were loaded with the same preparation of HIV-1 and incubated at 37 &#x000b0;C for 2 h were used as controls. Arrows indicate viral particles that have been internalized in cytoplasmic vesicles.</p><p>(E) HIV-1 p24 levels in post-trypsin treatment supernatants (sup) and cell pellets of activated B cells (lys) that had been incubated with AT-2&#x02013;inactivated HIV-1 X4 (MN) or HIV-1 R5 (Ba-L) for 2 h at 37 &#x000b0;C, and then incubated with 0.25&#x00025; trypsin solution for 10 min at 37 &#x000b0;C. Controls included activated B cells without virus, which were negative for p24 (unpublished data).</p><p>(F) Activated B cells were incubated with HIV-1 R5 (Ba-L) on wet ice, then washed and shifted at 37 &#x000b0;C for the times indicated. Cells were then treated with trypsin or mock treated and added to T cells. Supernatants were collected every 4 d and tested for p24. Data are from single experiments representative of three independent experiments.</p></caption><graphic xlink:href="ppat.0020070.g008"/></fig><p>To prove further that HIV-1 particles were localized inside B lymphocytes and not merely bound to their surface, we incubated activated B cells with AT-2 inactivated HIV-1 X4 particles for 2 h at 37 &#x000b0;C to allow internalization, and then treated the cells with trypsin to cleave virions still bound on the cell surface. The culture medium was removed, and the cells were washed and lysed. HIV-1 p24 contained in the wash supernatants and cell lysates was measured by enzyme-linked immunosorbent assay (ELISA). Under these conditions, essentially all of the p24 was found in the whole cell lysate, as very little was detected in the cell culture supernatants collected after the trypsin treatment (<xref ref-type="fig" rid="ppat-0020070-g008">Figure 8</xref>E). To prove that internalized virus and not surface-bound virus was transferred from B cells to the T cells, we next performed a protease protection transmission assay. Cells were loaded with HIV-1 R5 at 4 &#x000b0;C (wet ice) for 1 h, washed, and then shifted to 37 &#x000b0;C for the times indicated prior to treatment with trypsin or medium (mock treatment). As shown in <xref ref-type="fig" rid="ppat-0020070-g008">Figure 8</xref>F, very brief proteolysis of the B cells that had been exposed to HIV-1 at 4 &#x000b0;C to inhibit virus entry and then shifted to 37 &#x000b0;C (i.e., 0 min incubation) prevented transmission of HIV-1 to the T cells. In contrast, proteolysis of the HIV-1&#x02013;loaded B cells after shifting to 37 &#x000b0;C for 30 min did not prevent <italic>trans</italic> infection of the T cells. These data support that HIV-1 was internalized by the activated B cells prior to transmission to T cells.</p><p>Finally, we examined the ability of HIV-1 to remain infectious in B cells over time, as has been shown for HIV-1 infection of DC [<xref rid="ppat-0020070-b032" ref-type="bibr">32</xref>]. For this, we exposed activated B cells to infectious X4 or R5 strains of HIV-1 for 2 h at 37 &#x000b0;C, and then either mixed them immediately with T cells or kept them at 37 &#x000b0;C for 2 d before mixing with T cells. As shown in <xref ref-type="fig" rid="ppat-0020070-g009">Figure 9</xref>A and <xref ref-type="fig" rid="ppat-0020070-g009">9</xref>B, B cells that were loaded with HIV-1 and cultured alone for 2 d could still transmit virus to T cells, although at a lower efficiency then when T cells were added immediately to the HIV-1&#x02013;exposed B cells. Taken together, these results indicate that HIV-1 was captured and internalized in activated B cells, where it remained infectious for T cells for at least 2 d.</p><fig id="ppat-0020070-g009" position="float"><label>Figure 9</label><caption><title>HIV-1 Is Maintained in an Infectious Form by B Cells</title><p>Activated B cells were loaded with HIV-1 X4 (IIIb) or HIV-1 R5 (Ba-L) for 2 h at 37 &#x000b0;C, extensively washed and then either cultured immediately with autologous T cells (panel A) or incubated for 2 d prior to adding autologous T cells (panel B). B cells and T cells directly loaded with HIV-1 served as controls.</p></caption><graphic xlink:href="ppat.0020070.g009"/></fig></sec></sec><sec id="s3"><title>Discussion</title><p>Our results demonstrate that activated B lymphocytes derived from peripheral blood and lymphatic tissue express DC-SIGN, and that these cells mediate HIV-1 <italic>trans</italic> infection of T lymphocytes. Evidence for <italic>trans</italic> infection was an increase of HIV-1 p24 from undetectable levels (&#x0003c;1 &#x000d7; 10¹ pg/ml) to &#x0003e;10³&#x02013;10⁴ pg/ml after 12&#x02013;16 d of co-culture of the activated B cells with autologous CD4^&#x0002b; T cells. The <italic>trans</italic> infection was not due to direct <italic>cis</italic> infection of the B cells, as virus replication was not detectable in B cells alone after exposure to HIV-1. Likewise, there was little or no detectable replication of HIV-1 after direct infection of T cells alone with the low input of HIV-1 used in our experiments. Expression of DC-SIGN by the B cells was associated with this enhanced HIV-1 infection of CD4^&#x0002b; T cells in <italic>trans</italic>. That is, treatment with IL-4 and CD40L, which are mediators of B cell activation by CD4^&#x0002b; T cells [<xref rid="ppat-0020070-b019" ref-type="bibr">19</xref>], enhanced DC-SIGN expression on the B cells. A total of 80&#x00025; of the HIV-1&#x02013;infected B cells were DC-SIGN^&#x0002b; by flow cytometry. This was supported by immunofluorescence microscopy analysis showing that DC-SIGN and HIV-1 were coexpressed in the B cell cytoplasm. Association of virus with a small portion of non-DC-SIGN&#x02013;expressing B cells could be related to expression of other C-type lectin receptors by B cells that bind gp120 [<xref rid="ppat-0020070-b016" ref-type="bibr">16</xref>]. Most importantly, DC-SIGN expression by the B cells was required for <italic>trans</italic> infection of CD4^&#x0002b; T cells with both X4 and R5 tropic strains of HIV-1, as we could block essentially all of the <italic>trans</italic> infection by pretreating the B cells with mAb specific for DC-SIGN.</p><p>In further support of a central role for DC-SIGN on B cells in HIV-1 <italic>trans</italic> infection, we found that the number of DC-SIGN molecules expressed on activated B cells was similar to that known to be sufficient to sustain capture of HIV-1 and <italic>trans</italic> infection of T cells by DC [<xref rid="ppat-0020070-b025" ref-type="bibr">25</xref>]. Notably, the mechanism of <italic>trans</italic> infection of T cells by B cells did not involve HIV-1 infection of B cells by the conventional, CD4-CXCR4/CCR5 pathway. This was supported by the fact that B cells expressed little or no CD4 and did not express CCR5. Moreover, although B cells expressed high levels of CXCR4, treatment with anti-CXCR4 mAb prior to virus binding to the B cells did not inhibit <italic>trans</italic> infection of the T cells. Thus, although other C-type lectin receptors are expressed on B cells and can bind gp120 [<xref rid="ppat-0020070-b016" ref-type="bibr">16</xref>], our results support a requirement for DC-SIGN in B cell&#x02013;mediated, <italic>trans</italic> infection of T cells.</p><p>HIV-1 was internalized by B cells as determined by resistance of B cell&#x02013;associated virus to treatment with trypsin, and predominance of virus particles in cytoplasmic vacuoles of B cells after binding to DC-SIGN. Both intact and apparently degraded particles were present within the vacuoles, similar to HIV-1 internalization in DC via DC-SIGN [<xref rid="ppat-0020070-b003" ref-type="bibr">3</xref>]. Internalized and not extracellularly bound virus resulted in <italic>trans</italic> infection of T cells, which was demonstrated by lack of effect of trypsin treatment of HIV-1&#x02013;loaded B cells on their ability to mediate <italic>trans</italic> infection. These results suggest a role for internalization of HIV-1 by B cells that is similar to that of DC in <italic>trans</italic> infection of T cells [<xref rid="ppat-0020070-b033" ref-type="bibr">33</xref>]. We further observed that virus was maintained for at least 2 d in an infectious form in B cells, comparable to its association with DC [<xref rid="ppat-0020070-b032" ref-type="bibr">32</xref>]. It is not clear how this virus infectivity persists. Recent studies indicate that most of the captured virions are destroyed in DC-SIGN&#x02013;expressing DC and B cell lines engineered to express DC-SIGN [<xref rid="ppat-0020070-b003" ref-type="bibr">3</xref>,<xref rid="ppat-0020070-b034" ref-type="bibr">34</xref>]. However, a portion of the input virus can rapidly <italic>trans</italic> infect T cells [<xref rid="ppat-0020070-b003" ref-type="bibr">3</xref>,<xref rid="ppat-0020070-b035" ref-type="bibr">35</xref>]. In contrast, the longer term, persistent infectivity of DC-SIGN&#x02013;expressing DC and B cell lines has been related to low levels of de novo HIV-1 replication [<xref rid="ppat-0020070-b003" ref-type="bibr">3</xref>,<xref rid="ppat-0020070-b035" ref-type="bibr">35</xref>]. It is possible that very low, subdetectable levels of HIV-1 replication occurred in our IL-4&#x02013; and CD40L-stimulated B cells, and resulted in persistence of infectious virus involved in <italic>trans</italic> infection of T cells. Further work is in progress to delineate the mechanisms by which DC-SIGN&#x02013;expressing B cells lead to HIV-1 <italic>trans</italic> infection of T cells.</p><p>The DC-SIGN&#x02013;mediated, B-to-T cell <italic>trans</italic> infection pathway appears to be distinct from previously described B&#x02013;T cell infectious processes. That is, it has been reported that transmission of HIV-1 from B cells to T cells involves virus trapped in immune complexes on the surface of the B cells [<xref rid="ppat-0020070-b006" ref-type="bibr">6</xref>&#x02013;<xref rid="ppat-0020070-b010" ref-type="bibr">10</xref>]. This is not involved in our system, since in our studies, HIV-1 <italic>trans</italic> infection of T cells was mediated by B cells from normal, HIV-1 antibody&#x02013;negative persons. Finally, in contrast to Swingler et al. [<xref rid="ppat-0020070-b011" ref-type="bibr">11</xref>], where <italic>nef</italic>-induced soluble factors released by infected macrophages work together with B cells to lead to <italic>trans</italic> infection of T cells, we observed <italic>trans</italic> infection of T cells by purified, DC-SIGN^&#x0002b; B cells in the absence of macrophages.</p><p>Notably, we found that approximately 8&#x00025; of B cells in the blood of normal donors expressed DC-SIGN. Stimulation for 24 h with IL-4 and CD40L resulted in an increase in the number of B cells expressing DC-SIGN (i.e., 28&#x00025;) and the level of DC-SIGN expression on the B cells. Approximately 80&#x00025; of the DC-SIGN^&#x0002b; B cells in blood also expressed the type II C-type lectin receptor and B cell activation marker, CD23, which increased to 86&#x00025; after stimulation in vitro with IL-4 and CD40L. Activation of these stimulated, DC-SIGN^&#x0002b; B cells was confirmed by elevated levels of T cell coreceptors CD58, CD80, and CD86, as well as increases in coexpression of the B cell signal transduction molecule CD22. Expression of DC-SIGN was not restricted to blood B cells, as we found that approximately 26&#x00025; of tonsil B cells constitutively expressed DC-SIGN, and that this increased to 39&#x00025; after stimulation of the cells with IL-4 and CD40L. Interestingly, our study showed that few B cells directly isolated from the tonsils expressed CD23, but that this number increased after stimulation with IL-4 and CD40L. Expression of CD23 on some but not all DC-SIGN^&#x0002b; B cells could be related to differential expression of CD23 by B cells in distinct areas of tonsils [<xref rid="ppat-0020070-b036" ref-type="bibr">36</xref>], and a pronounced cleavage and shedding of soluble CD23 by activated B cells [<xref rid="ppat-0020070-b037" ref-type="bibr">37</xref>].</p><p>Of interest is that there was a comparable number of DC-SIGN&#x02013;expressing B cells in the blood of uninfected persons as in those with chronic HIV-1 infection who were treatment naive or receiving ART. In support of these findings, although B cells can harbor HIV-1 in blood [<xref rid="ppat-0020070-b006" ref-type="bibr">6</xref>], we found that HIV-1 did not kill B cells or inhibit DC-SIGN expression within 24&#x02013;48 h of exposure to virus in vitro. Our observation that IL-4 and CD40L stimulation of B cells from persons receiving ART failed to enhance expression of DC-SIGN suggests a negative effect of ART on this process. This could be related to the reported inhibitory effect of protease inhibitors, which are common components of ART, on expression of DC-SIGN [<xref rid="ppat-0020070-b038" ref-type="bibr">38</xref>], and requires further study.</p><p>Expression of DC-SIGN on B cells suggests that it is operative in their normal function. This type II C-type lectin receptor could be involved in B cell trafficking and antigen presentation to T cells, similar to its function in DC [<xref rid="ppat-0020070-b039" ref-type="bibr">39</xref>]. Indeed, high levels of expression of DC-SIGN were induced in B cells by a combination of T cell factors IL-4 and CD40L and not by either alone. These results support the concept that DC-SIGN is involved in the interaction of activated B and T cells during antigen processing and presentation.</p><p>Our data indicate that DC-SIGN&#x02013;expressing B cells could become vehicles for HIV-1 infection of T cells during their cognate interactions in the lymphatics. High concentrations of HIV-1 have been found associated with B cells and CD4^&#x0002b; T cells in lymph nodes [<xref rid="ppat-0020070-b040" ref-type="bibr">40</xref>], where B cells are activated and proliferate through cytokine and CD40L-CD40 interactions [<xref rid="ppat-0020070-b041" ref-type="bibr">41</xref>]. This has been related to virus-containing immune complexes on the surface of B cells that could lead to infection of CD4^&#x0002b; T cells during &#x0201c;cross talk&#x0201d; in the microenvironment of lymphoid tissues. The presence of a large amount of unspliced simian immunodeficiency virus RNA in B cell&#x02013;rich areas of lymphoid tissues [<xref rid="ppat-0020070-b042" ref-type="bibr">42</xref>], is also consistent with virions being associated with B cells and follicular DC in germinal centers. B cells have therefore been proposed as an important, lymphatic reservoir in the pathogenesis of HIV-1 [<xref rid="ppat-0020070-b028" ref-type="bibr">28</xref>]. Our results suggest that B cells play a previously unrecognized role in replication of HIV-1 in T cells and viral pathogenesis through a DC-SIGN&#x02013;dependent mechanism.</p></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Donors.</title><p>PBMC were obtained from healthy, HIV-1&#x02013;seronegative (<italic>n</italic> = 33) and HIV-1&#x02013;seropositive (<italic>n</italic> = 20) adult volunteers and subjects enrolled in the Multicenter AIDS Cohort Study. Informed consent was obtained following institutional guidelines. ART-naive, HIV-1&#x02013;seropositive subjects (<italic>n</italic> = 10) had mean (&#x000b1; SE) of 658 &#x000b1; 124 CD4^&#x0002b; T cells per ml of blood, and HIV-1 RNA loads of 6,410 &#x000b1; 3,701 copies per ml of plasma; HIV-1&#x02013;seropositive subjects on ART (<italic>n</italic> = 10) had CD4^&#x0002b; T cell counts of 939 &#x000b1; 82 and viral RNA loads of 66 &#x000b1; 20. Human tonsils were obtained from patients (<italic>n</italic> = 5) undergoing therapeutic surgery, in accordance with institutional guidelines.</p></sec><sec id="s4b"><title>Preparation of B and T cells.</title><p>PBMC were isolated by Ficoll-Hypaque density gradient separation and used immediately for surface phenotype staining or further purification. For B cell purification, monocytes were depleted with two rounds of anti-CD14 mAb-coated immunomagnetic microbeads (Miltenyi) according to the manufacturer's instructions. B cells (CD19^&#x0002b; cells) were then isolated from the CD14^&#x02212; cell fraction by incubation with anti-CD19 mAb-coated, magnetic microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). The purity of the fractionated B cells was 96.4&#x00025; &#x000b1;0.4 (mean &#x000b1; SE; <italic>n</italic> = 33) as determined by staining with anti-CD20 mAb, with &#x0003c;1&#x00025; CD14^&#x0002b; and CD3^&#x0002b; cells. The immunomagnetic bead purification procedure did not alter expression of DC-SIGN or CD23 (unpublished data). Autologous CD4^&#x0002b; T cells were obtained from the remaining CD14^&#x02212; CD19^&#x02212; fraction by CD4^&#x0002b; cell purification using anti-CD4 mAb-coated microbeads (Miltenyi). Activated B cells were generated by culture of CD19^&#x0002b; cells in RPMI 1640 medium supplemented with 10&#x00025; heat-inactivated FCS, 1,000 U/ml of rhIL-4 (R &#x00026; D Systems, Minneapolis, Minnesota, United States), and 1 ug/ml of soluble trimeric CD40 L (Amgen, Thousand Oaks, California, United States). CD4^&#x0002b; T lymphocytes were cultured in RPMI1640 medium supplemented with 20&#x00025; FCS and phytohemagglutinin (PHA) and IL-2, as described [<xref rid="ppat-0020070-b043" ref-type="bibr">43</xref>].</p><p>Tonsils were immediately transferred to the laboratory in cold PBS supplemented with penicillin (100 U/ml), streptomycin (100 &#x003bc;g/ml), gentamicin (5 &#x003bc;g/ml), ampothericin B (0.5 &#x003bc;g/m), and 5&#x00025; FCS. The tissues were cut into fragments and pushed through a stainless steel sieve with a 250-&#x003bc;m mesh, using the flat end of a plastic syringe plunger, to remove the connective capsula. Lymphocytes from the collected cell suspension were isolated by Ficoll-Hypaque density gradient centrifugation. The cells collected at the interface were stained with anti-CD20, anti-CD3, and anti-CD14 mAb to determine the relative percentage of B and T lymphocytes, and were shown to contain 60&#x00025;&#x02013;70&#x00025; B lymphocytes, 30&#x00025;&#x02013;40&#x00025; T lymphocytes, and 1 &#x00025; or less monocytes. B and T cells were purified by immunomagnetic bead separation and activated as described above.</p></sec><sec id="s4c"><title>Preparation of DC.</title><p>DC were generated from CD14^&#x0002b; cells as described [<xref rid="ppat-0020070-b044" ref-type="bibr">44</xref>]. Briefly, CD14^&#x0002b; cells were enriched from PBMC with anti-CD14 mAb magnetic beads (&#x0003e;95&#x00025; lineage negative, HLA DR^&#x0002b; cells) and cultured for 5 d in the presence of 1,000 U/ml of rhIL4 and rGM-CSF.</p></sec><sec id="s4d"><title>Antibodies.</title><p>The following mAbs were used in this study: anti&#x02013;DC-SIGN (CD209) (clone 120507; R &#x00026; D Systems); anti-CD4, anti-CD14, anti-CD19, anti-CD20, anti-CD22, anti-CD45, anti-CD58, anti-CD80, anti-CD86, anti-CXCR4, and CCR5 (BD Pharmingen, San Diego, California, United States); anti-CD23 (Caltag Laboratories, Burlingame, California, United States); anti&#x02013;HIV-1 p24 (KC57; Beckman-Coulter, Fullerton, California, United States). These mAbs were used either unlabeled or conjugated with FITC, PE, PE-Cy5, or PE-Cy7 as indicated below. Appropriate isotype-matched controls were used for background staining evaluation. For the virus receptor blocking experiments, we used sodium azide&#x02013;free, low-endotoxin, purified, anti&#x02013;DC-SIGN clone 120507 mAb and mouse IgG of the relevant isotype (Becton Dickinson, Palo Alto, California, United States) reconstituted in sterile PBS.</p></sec><sec id="s4e"><title>Flow cytometry and quantification of DC-SIGN expression.</title><p>Expression of cell surface molecules was examined by flow cytometry with a Beckman Coulter XL flow cytometer. Cells were incubated with the desired antibodies or isotype controls for 30 min at 4 &#x000b0;C in buffer consisting of PBS supplemented with 0.1&#x00025; FCS and 0.1&#x00025; NaN₃. After extensive washing with the buffer, the cells were resuspended in 1&#x00025; paraformaldehyde in PBS for flow cytometric analysis. HIV-1 intracellular p24 was identify using PE-labeled KC57 anti&#x02013;HIV-1 Gag mAb, following cell permeabilization using Permiflow (Invirion Diagnostics, Oak Brook, Illinois, United States) according to manufacturer's instructions. Results were expressed either as percent positive cells above the isotype control threshold or as MFI above the isotype controls.</p><p>For quantification of DC-SIGN expression on DC, Raji&#x02013;DC-SIGN cells, and activated B cells, we used the Quantum Simply Cellular kit (Bangs Laboratories, Fishers, Indiana, United States) according to the manufacturer's instructions. Quantification was performed by converting the geometric mean channel fluorescence (GMCF) into ABC. The kit contains five microbeads of uniform size coated with different amounts of goat anti&#x02013;mouse IgG (Fc-specific) on their surface that have different abilities to bind mouse antibodies (ranging from 0 to about 250,000 molecules). Both beads and cells were labeled with saturating amounts of FITC-conjugated, anti&#x02013;DC-SIGN mAb, processed, and analyzed by flow cytometry under identical conditions. A calibrating curve was derived from the bead samples using QuickCal (Bangs Laboratories). The GMCF of the samples was converted to ABC per cell by comparison with the regression curve generated with the beads. Samples were also evaluated for fluorescence intensity as expressed by MESF units, using the Quantum MESF kit (Bangs Laboratories). The kit consists of five bead populations having different levels of FITC fluorescence intensity. As described above, a regression curve was generated by plotting the GMCF of the beads against their known MESF using QuickCal. The MESF of the cell samples was then determined as described above for ABC.</p></sec><sec id="s4f"><title>Real-time RT-PCR measurement of DC-SIGN mRNA.</title><p>Total RNA was extracted from cells using Trizol (Invitrogen Life Sciences, Carlsbad, California, United States) according to the manufacturer's instructions, DNAse-treated (Ambion, Austin, Texas, United States) and affinity column purified (RNAeasy, Qiagen, Valencia, California, United States). The sequences of the primers and probe for Taqman amplification and detection of DC-SIGN mRNA were kindly provided by B. Lee (University of California, Los Angeles) and were DC-SIGN.F1, 5&#x02032;-GCTGAGAGGCCTTGGATTCC-3&#x02032;; DC-SIGN.R1, 5&#x02032;-AGAGCGTGAAGGAGAGGAGTTG-3&#x02032;; and DC-SIGN.probe, 5&#x02032;-6FAM-ACCATGGCCAAGACACCCTGCTA -MGB-3&#x02032;. The comparative threshold cycle (Ct) method [<xref rid="ppat-0020070-b023" ref-type="bibr">23</xref>,<xref rid="ppat-0020070-b024" ref-type="bibr">24</xref>] was used to determine relative mRNA expression levels. The primer and probe set for &#x003b2;-glucoronidase (&#x003b2;-GUS) (Applied Biosystems, Foster City, California, United States) was used as the endogenous control. Across all samples, mean &#x003b2;-GUS Ct values were 29 &#x000b1; 1 (unpublished data). cDNA synthesis with 400-ng input RNA was performed in duplicate with Superscript II reverse transcriptase (Invitrogen) and random hexamers as described [<xref rid="ppat-0020070-b024" ref-type="bibr">24</xref>], in parallel with control reactions lacking RT. Amplification and detection for 40 cycles was performed on a Prism 7000 Sequence Detection System (Applied Biosystems). The fold-changes in DC-SIGN mRNA expression were calculated relative to appropriate calibrator samples, including untransfected K562 cells and unstimulated CD19^&#x0002b; B cells.</p></sec><sec id="s4g"><title>Virus.</title><p>HIV-1 IIIb (X4 tropic virus) and HIV-Ba-L (R5 tropic virus) were propagated in PHA- and IL-2&#x02013;activated, normal donor PBMC and purified as described [<xref rid="ppat-0020070-b043" ref-type="bibr">43</xref>]. Virus titers were determined as pg/ml by p24 ELISA (DuPont, Wilmington, Delaware, United States), with a lower limit of 1 &#x000d7; 10¹ pg/ml and upper limit of &#x02265; 5 &#x000d7; 10⁴ pg/ml<bold>.</bold> AT-2&#x02013;inactivated HIV-1 MN (CL.4/SUPT1; X4 tropic), and ADA (M/SUPT1-CCR5 CL.30; R5 tropic) were a gift from J. D. Lifson, National Cancer Institute, Frederick, Maryland.</p></sec><sec id="s4h"><title>HIV-1 infection and transmission assay.</title><p>Purified, IL-4&#x02013; and CD40L-stimulated B lymphocytes (1 &#x000d7; 10⁶) were incubated with different amounts of HIV-1 IIIb or HIV-1 Ba-L, i.e., 10^&#x02212;3, 10^&#x02212;4, or 10^&#x02212;5 MOI at 37 &#x000b0;C for 2 h. The MOI was based on tissue culture infectious dose 50&#x00025; determined with susceptible CD8^&#x02212;, PHA-stimulated human lymphocytes and confirmed by spectrophotometric analysis of 10-fold serial dilutions on TZM-bl indicator cell line. Unless otherwise specified, cell-free supernatants were taken at various time intervals for titration of virus by p24 ELISA. No difference was observed in viability of mock-treated and HIV-1&#x02013;treated B cells as measured by trypan blue dye exclusion. In some experiments, stimulated B cells were incubated with anti&#x02013;DC-SIGN mAb (clone 120507) or CD11a/LFA-1 (clone HI111, BD Pharmingen) or mouse IgG (R &#x00026; D Systems) for 30 min at 4 &#x000b0;C prior to incubation with virus. The specificity of anti&#x02013;DC-SIGN mAb clone 120507 was confirmed by binding to DC-SIGN&#x02013;transfected K-562 cells and lack of binding to K562 cells transfected with the DC-SIGN&#x02013;related, type II C-type lectin, DC-SIGNR [<xref rid="ppat-0020070-b030" ref-type="bibr">30</xref>] .</p></sec><sec id="s4i"><title>Loading of DC and activated B cells with HIV-1 for electron microscopy.</title><p>A total of 1 &#x000d7; 10⁶ DC or purified, activated B cells were incubated in a 1.5-ml Eppendorf tube with AT-2 HIV-1 (3 ng of p24 Ag /10⁶ cells) in a total volume of 100 &#x003bc;l at 37 &#x000b0;C for up to 2 h. After the incubation, cells were extensively washed in cold medium using a refrigerated microfuge, and the cell pellets were fixed in PBS with 2.5&#x00025; glutaraldehyde for 1 h. Pellets were washed three times in PBS and then post-fixed in 1&#x00025; aqueous osmium tetroxide supplemented with 1&#x00025; K₃Fe(CN)₆ for 1 h. Pellets were then washed three times in PBS and then dehydrated through a series of 30&#x00025;&#x02013;100&#x00025; ethanol, 100&#x00025; propylene oxide, and then infiltrated with 1:1 mixture of propylene oxide&#x02013;Polybed 812 epoxy resin (Polysciences, Warrington, Pennsylvania, United States) for 1 h. After several changes of 100&#x00025; resin over 24 h, the pellet was embedded in a final change of resin, cured at 37 &#x000b0;C overnight, followed by additional hardening at 65 &#x000b0;C for two more days. Ultrathin (70 nm) sections were collected onto 200-mesh copper grids, stained with 2&#x00025; uranyl acetate in 50&#x00025; methanol for 10 min, followed by 1&#x00025; lead citrate for 7 min. Sections were viewed using a JEM 1210 electron microscope (JEOL, Peabody, Massachusetts, United States).</p></sec><sec id="s4j"><title>Immunofluorescence microscopy.</title><p>Purified, activated B cells loaded with AT-2 HIV-1 MN were spotted on poly-L-lysine&#x02013;coated slides, fixed with 4&#x00025; paraformaldehyde for 20 min, and then permeabilized with buffer (0.5&#x00025; BSA, 0.1&#x00025; saponin, 0.1&#x00025;NaN₃) for 20 min at room temperature. Cells were stained with FITC-conjugated, anti&#x02013;DC-SIGN mAb and PE-labeled anti&#x02013;HIV-1 p24 mAb. To avoid nonspecific binding of IgG, all incubations and dilutions of reagents were done in Super-Block blocking buffer (Pierce Biotechnology, Rockford, Illinois, United States). Controls included activated B cells not exposed to HIV-1.</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting Information</title><supplementary-material content-type="local-data" id="ppat-0020070-sg001"><label>Figure S1</label><caption><title>CD40L and IL-4 Act Synergistically in Inducing Enhanced DC-SIGN and CD23 Expression on B Cells</title><p>(A) Coexpression of DC-SIGN and CD23 on B cells from healthy, HIV-1&#x02013;seronegative donors (<italic>n</italic> = 20). Hour 0 [T0] = 0 h; hour 24 [T24] = 24 h.</p><p>(B) Time-dependent expression of DC-SIGN and CD23 in response to IL-4 and CD40L. Purified B cells from a normal donor were cultured in the presence of IL-4 or CD40L or a combination of the two. Untreated cells were used as controls (NT). DC-SIGN and CD23 coexpression on activated B cells was greatest using a combination of 1,000 U/ml of IL4 and 1 &#x003bc;g/ml of CD40L for 24 h. Single concentrations of 100, 500, and 10,000 U/ml of IL-4 and 0.1, 0.5, and 10 &#x003bc;g/ml of CD40L induced similar, low levels of DC-SIGN expression (unpublished data). Also, various combinations of these concentrations of IL-4 and CD40L induced less DC-SIGN expression than this combination (unpublished data).</p><p>(380 KB JPG)</p></caption><media xlink:href="ppat.0020070.sg001.jpg"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><sec id="s5a"><title>Accession Numbers</title><p>The GenBank accession number (<ext-link ext-link-type="uri" xlink:href="http://www.ncbi.nlm.nih.gov/Genbank">http://www.ncbi.nlm.nih.gov/Genbank</ext-link>) for the gene mentioned in this paper is <italic>CD209</italic> (NM_021155).</p></sec></sec>

<named-content content-type="genus-species">E. coli</named-content> Microcosms Indicate a Tight Link between Predictability of Ecosystem Dynamics and Diversity

<p>The diversity-stability hypothesis proposes that ecosystem diversity is positively correlated with stability. The impact of ecosystem diversity is, however, still debated. In a microcosm experiment using diverged <named-content content-type="genus-species">Escherichia coli</named-content> cells, we show that the fitness of community members depends on the complexity (number of participants) of the system. Interestingly, the spread of a community member with a superior genotype is mostly stochastic in low-complexity systems, but highly deterministic in a more complex environment. We conclude that system complexity provides a buffer against stochastic effects.</p>

<contrib contrib-type="author"><name><surname>Imhof</surname><given-names>Marianne</given-names></name><xref ref-type="author-notes" rid="n110">&#x000a4;</xref><xref ref-type="aff" rid="aff1"/></contrib><contrib contrib-type="author"><name><surname>Schl&#x000f6;tterer</surname><given-names>Christian</given-names></name><xref ref-type="corresp" rid="cor1">*</xref><xref ref-type="aff" rid="aff1"/></contrib>

PLoS Genetics

<sec id="s1"><title>Introduction</title><p>As early as 1872 Darwin [<xref rid="pgen-0020103-b001" ref-type="bibr">1</xref>] had envisioned the critical impact that species diversity has on ecosystem dynamics. Elton [<xref rid="pgen-0020103-b002" ref-type="bibr">2</xref>] explicitly formulated this thought with the diversity-stability hypothesis, which proposes that ecosystem diversity is positively correlated with stability. The relationship between ecosystem functioning and species diversity is widely discussed [<xref rid="pgen-0020103-b003" ref-type="bibr">3</xref>&#x02013;<xref rid="pgen-0020103-b006" ref-type="bibr">6</xref>]. While early empirical studies suggested that more diverse communities enhance ecosystem stability [<xref rid="pgen-0020103-b002" ref-type="bibr">2</xref>,<xref rid="pgen-0020103-b007" ref-type="bibr">7</xref>], subsequent ecological models indicated that diversity tends to destabilize community dynamics [<xref rid="pgen-0020103-b008" ref-type="bibr">8</xref>]. Since then, more realistic models have been proposed that reconcile community complexity with ecosystem stability [<xref rid="pgen-0020103-b009" ref-type="bibr">9</xref>,<xref rid="pgen-0020103-b010" ref-type="bibr">10</xref>]. Food web structure has been discussed as centrally important in the relationship between ecosystem stability and diversity [<xref rid="pgen-0020103-b011" ref-type="bibr">11</xref>]. If the distribution of consumer-resource interaction is skewed to weak interactions, ecosystem diversity is positively linked with stability (weak-interaction effect [<xref rid="pgen-0020103-b012" ref-type="bibr">12</xref>]). Nevertheless, the intrinsic complication of measuring ecosystem stability has resulted in opposing outcomes, depending on how ecosystem stability is defined [<xref rid="pgen-0020103-b013" ref-type="bibr">13</xref>].</p><p>The majority of experimental ecosystems have focused on assemblies of different species, often covering a range of trophic levels. Attempts to study intraspecific variation as a way to work under more controlled experimental conditions imply the main drawback: the difficulty of distinguishing intraspecific variants. Nevertheless, the importance of intraspecific variation for ecosystem functioning should not be underestimated. One particularly good example of the effect of intraspecific variation on ecological dynamics is the analysis of predator-prey cycles of a system consisting of one species in each group (rotifers and algae) [<xref rid="pgen-0020103-b014" ref-type="bibr">14</xref>]. The authors demonstrated that genetic diversity in the prey population (algae) significantly altered the predator-prey cycle in length and synchronisation. Other examples of the ecological impact of intraspecific variation were provided by recent studies on eelgrass diversity [<xref rid="pgen-0020103-b015" ref-type="bibr">15</xref>].</p><p>In this report, we focus on intraspecific variation generated in an evolving <named-content content-type="genus-species">E. coli</named-content> population, testing how diversity affects the evolutionary trajectory of the population. If the evolutionary trajectory is repeatable (and thus predictable), we consider the system of evolved <named-content content-type="genus-species">E. coli</named-content> cells to be stable. Within a recently introduced classification system of definitions for ecosystem stability [<xref rid="pgen-0020103-b016" ref-type="bibr">16</xref>], our use corresponds best with the term &#x0201c;resilience&#x0201d; [<xref rid="pgen-0020103-b017" ref-type="bibr">17</xref>]. However, rather than testing for a return to a reference state after a disturbance (the formal definition of resilience), we tested for attainment of the reference state, namely the spread of a beneficial mutation. Using a highly informative marker system, we monitored the spread of a beneficial genotype in systems with different levels of complexity.</p></sec><sec id="s2"><title>Results</title><p>We used experimental evolution to generate an evolved community of diverged <named-content content-type="genus-species">E. coli</named-content> lineages. The stability of the system was measured by the reproducibility of the dynamics of a beneficial mutation that occurred in the intact system.</p><p>Beneficial mutations regularly arise in experimental <named-content content-type="genus-species">E. coli</named-content> populations [<xref rid="pgen-0020103-b018" ref-type="bibr">18</xref>,<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>]. We used a microsatellite marker to infer the spread of beneficial mutations in the evolving <named-content content-type="genus-species">E. coli</named-content> population [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>]. <xref ref-type="fig" rid="pgen-0020103-g001">Figure 1</xref> shows the deterministic spread of a genotype carrying the beneficial mutation (indicated by the red bar, 33 repeats). This selective sweep was reproduced in a recent study of five independent replicate cultures [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>]. Cells with (sweepers) and without (competitors) the beneficial mutation were isolated at generation 324, as at this point in time the sweeper had already reached a considerable frequency, but a large diversity of competitor cells was still present (<xref ref-type="fig" rid="pgen-0020103-g001">Figure 1</xref>). Among the competitors, isolated cells differed by size of the microsatellite marker as well as by tetracycline resistance and levels of adherence, indicating that the harvested cells had already diversified (<xref ref-type="supplementary-material" rid="pgen-0020103-st001">Tables S1</xref> and <xref ref-type="supplementary-material" rid="pgen-0020103-st002">S2</xref>).</p><fig id="pgen-0020103-g001" position="float"><label>Figure 1</label><caption><title>Changes in Allele Frequency at a Microsatellite Marker during the Spread of a Beneficial Mutation</title><p>A &#x0201c;snapshot&#x0201d; of the allele distribution in the evolving <named-content content-type="genus-species">E. coli</named-content> population is shown for every eighteenth generation. The number of generations after the start of the experiment is given on the upper right corner of each graph. Bars represent the frequency of the corresponding microsatellite allele. The microsatellite allele carried by the cell with the beneficial mutation (sweeper) is shown in red. The red ellipse indicates the generation at which we isolated the cells used for the competition experiments. Note that for better resolution the scale of the y-axis has been modified between generations 324 and 342.</p></caption><graphic xlink:href="pgen.0020103.g001"/></fig><p>For three competitor genotypes (360 experiments using the genotypes with clone numbers 902, 903, and 1139; for further details see <xref ref-type="table" rid="pgen-0020103-t001">Tables 1</xref>, <xref ref-type="supplementary-material" rid="pgen-0020103-st001">S1</xref>, and <xref ref-type="supplementary-material" rid="pgen-0020103-st002">S2</xref>) we performed a detailed analysis on the influence of the frequency of a given genotype on the outcome of the competition experiment. The frequency of the competitors at the onset of the competition experiments ranged from 0.04 to 0.96. Two competitors showed no frequency dependence, but for one competitor we detected a significant correlation between the starting frequency and fitness (two-tailed Spearman's Rank correlation, <italic>r</italic> ² = 0.45, <italic>p</italic> &#x02264; 0.001, <italic>n</italic> = 84 [unpublished data]).</p><table-wrap id="pgen-0020103-t001" content-type="1col" position="float"><label>Table 1</label><caption><p>Clone Combinations of Experimental Groups</p></caption><graphic xlink:href="pgen.0020103.t001"/></table-wrap><p>We determined the effect of system complexity by competition experiments using different levels of complexity: The extreme cases were either individual competitors (low diversity) or the whole population consisting of the entire collection of genotypes (high diversity). Intermediate levels of diversity were obtained by gradually increasing the number of competitors. Consistent with phenotypic and genotypic divergence among the competitor clones, we also found that the outcome of low diversity experiments was dependent on the genotype of the competitor cell (<xref ref-type="table" rid="pgen-0020103-t001">Tables 1</xref> and S1). To account for this heterogeneity, we always considered the average fitness of the competitor genotypes (or combinations of genotypes; see <xref ref-type="table" rid="pgen-0020103-t001">Table 1</xref> and Materials and Methods for more details).</p><p>Mean fitness was significantly lower in experiments involving a single competitor than the entire population (<italic>p</italic> &#x0003c; 0.003, permutation test based on 300 replicates, <xref ref-type="fig" rid="pgen-0020103-g002">Figure 2</xref>). When intermediate levels of complexity were also considered, we found a strong correlation between mean fitness and diversity (two-tailed Spearman's rank correlation, <italic>&#x003c1;</italic> = 0.94; <italic>p</italic> = 0.005; <italic>n</italic> = 6 [<xref ref-type="fig" rid="pgen-0020103-g002">Figure 2</xref>]). Thus, the fitness of the cells carrying the beneficial mutation depends on the level of complexity of the system.</p><fig id="pgen-0020103-g002" position="float"><label>Figure 2</label><caption><title>Relationship between Genetic Diversity of Competitors (Complexity) and Mean Fitness of the Clone Carrying the Beneficial Mutation (Sweeper)</title><p>Fitness of the sweeper was determined by competition against a single competitor (lowest level of complexity, far left bar) increasing up to the entire population (highest level of complexity, very right bar). The number of experiments performed for each experimental group and each combination of competitors can be found in <xref ref-type="table" rid="pgen-0020103-t001">Table 1</xref>. Error bars indicate the standard deviation of the Malthusian fitness parameter determined by 100 bootstrap pseudoreplicates. The mean of the means and standard deviations of these values are plotted.</p></caption><graphic xlink:href="pgen.0020103.g002"/></fig><p>We further discovered that in 78 (~22%) of 360 of the competition experiments involving single competitors, the clone carrying the beneficial mutation was not just less fit than in higher complexity experiments but even lost (exhibited negative fitness values) against the competitor cell. To further quantify this effect, we performed three replicate experiments for each competitor and determined the heterogeneity in Malthusian fitness of the sweeper among the three replicates. We observed the lowest variation for those competition experiments with the highest complexity level (entire population) and the highest variation among replicate experiments at the lowest complexity level (one competitor genotype, <italic>p</italic> &#x0003c; 0.003, permutation test based on 300 replicates [<xref ref-type="fig" rid="pgen-0020103-g003">Figure 3</xref>]). The comparison to intermediate complexity levels indicated that the decrease in heterogeneity is not linear, as the coefficient of variation for two competitors was reduced to about 30% of the single competitor experiments. Nevertheless, the highest complexity level resulted in the lowest coefficient of variation, indicating that reproducibility increases with complexity.</p><fig id="pgen-0020103-g003" position="float"><label>Figure 3</label><caption><title>Heterogeneity among Replicate Experiments</title><p>For each level of complexity (number of competitors) we determined the mean coefficient of variation of three replicate experiments. The number of experiments performed for each experimental group and each combination of competitors can be found in <xref ref-type="table" rid="pgen-0020103-t001">Table 1</xref>. Error bars indicate the standard deviation of 100 bootstrap values obtained by resampling experiments (and the corresponding coefficient of variation).</p></caption><graphic xlink:href="pgen.0020103.g003"/></fig></sec><sec id="s3"><title>Discussion</title><sec id="s3a"><title>To What Extent Are Microbial Models Suitable for Making Inferences about Ecosystem Dynamics?</title><p>Microbial models offer a variety of experimental advantages, such as short generation times, low cost, and the possibility of preserving genotypes by freezing. Several ecological issues such as succession, the diversity-stability relationship, predator-prey dynamics, the coexistence of competitors, and the coexistence of generalists and specialists are readily addressed with microbial model systems [<xref rid="pgen-0020103-b020" ref-type="bibr">20</xref>].</p><p>Nevertheless, it is also well understood that adaptation differs profoundly between prokaryotes and eukaryotes [<xref rid="pgen-0020103-b021" ref-type="bibr">21</xref>]. Although in bacteria, beneficial mutations are mainly fixed sequentially, in sexually reproducing eukaryotes, recombination allows different beneficial mutations to combine in the same genotype [<xref rid="pgen-0020103-b022" ref-type="bibr">22</xref>,<xref rid="pgen-0020103-b023" ref-type="bibr">23</xref>]. Furthermore, in general bacteria exhibit little homologous gene recombination, but do exhibit high rates of horizontal gene transfer (frequently considered an indication that a new species concept for prokaryotes is needed [<xref rid="pgen-0020103-b024" ref-type="bibr">24</xref>&#x02013;<xref rid="pgen-0020103-b028" ref-type="bibr">28</xref>]), whereas sexual eukaryotes frequently reshuffle their genes but rarely acquire genes from other species [<xref rid="pgen-0020103-b029" ref-type="bibr">29</xref>]. Nevertheless, horizontal gene transfer should play a minor role in our single-species <named-content content-type="genus-species">E. coli</named-content> experiments.</p><p>Our data suggest that high diversity of (nonrecombining) genotypes favours stability. Interestingly, two recent studies on ecosystem recovery and dependence on diversity come to similar conclusions using a eukaryotic system. The authors demonstrated that a higher number of eelgrass genotypes result in significantly higher resistance against disturbance (grazing geese) [<xref rid="pgen-0020103-b030" ref-type="bibr">30</xref>]. Furthermore, a higher genotypic diversity of the eelgrass also resulted in an increase in the number of invertebrates after perturbation (extreme heat wave) [<xref rid="pgen-0020103-b015" ref-type="bibr">15</xref>]. This similarity suggests that bacterial models are probably well suited to the study of ecological processes, in particular to study the importance of intraspecific diversity.</p></sec><sec id="s3b"><title>Diversity and Fitness</title><p>We found that mean fitness of the sweeper changed significantly in experiments involving a single competitor compared to those with more competitors. The highest level of complexity is most similar to the environment in which the beneficial mutation originated. This indicates that fitness of individual clones depends strongly on the system in which they have evolved. We think that our approach of using a coevolved community reflects real ecosystems better than randomly assembled systems, as these are thought not to be realistic [<xref rid="pgen-0020103-b031" ref-type="bibr">31</xref>].</p><p>If our observation is extrapolated to other systems, it may be concluded that attempts aiming to restore disturbed communities by using a small number of founder genotypes, whose individual performance is known only in systems with high complexity, may therefore not be the optimal strategy. Nevertheless, given the obvious simplification of our experimental system, further work is required to validate this conclusion.</p></sec><sec id="s3c"><title>Does Diversity Buffer against Stochastic Effects?</title><p>The experiments involving two genotypes&#x02014;the sweeper and one competitor&#x02014;were found to be highly stochastic. In some experiments, the fitness of the sweeper was actually lower than that of the competitor genotype. In population genetics, such observations are attributed to stochastic effects during the early phase of a selective sweep, when the frequency of the beneficial mutation is low. If the beneficial allele reaches a higher frequency, the stochastic phase is followed by a deterministic phase at which random effects can be safely ignored. In our experiments at least 4% of the cells carried the beneficial mutation, thus a deterministic outcome of the sweep was expected (given a population size &#x0003e;10⁶). Further evidence against genetic drift is provided by our high-complexity experiments, which were highly reproducible despite the fact that the frequency of the sweeper was similar to that in the single-competitor experiments. Hence, we conclude that the highly stochastic outcome of the competition experiments in a low-diversity setting characterizes an intrinsic property of the experimental system: genetic diversity buffers against the stochastic outcome of the competition experiments.</p><p>What might be the basis of the buffering effect of diversity seen in our experiments? One possible explanation could be gleaned from other microcosm experiments. By reducing the number of members of the community, the balance of the system is disturbed, leading to considerable stochastic noise, possibly due to the loss of redundancy [<xref rid="pgen-0020103-b032" ref-type="bibr">32</xref>,<xref rid="pgen-0020103-b033" ref-type="bibr">33</xref>]. An alternative scenario assumes that the co-occurring competitor clones are already functionally diverged. It is conceivable that through coevolved trophic interactions the system is stabilized. In experiments with reduced complexity, such interactions are diminished, which could explain the higher stochasticity in our experiments. Previous studies on yeast and bacteria indicated that trophic interactions based on secondary metabolites can occur during the cultivation of cells derived from the same ancestor [<xref rid="pgen-0020103-b034" ref-type="bibr">34</xref>&#x02013;<xref rid="pgen-0020103-b036" ref-type="bibr">36</xref>]. Interestingly, such trophic interactions either could have detrimental effects on co-occurring genotypes [<xref rid="pgen-0020103-b037" ref-type="bibr">37</xref>&#x02013;<xref rid="pgen-0020103-b039" ref-type="bibr">39</xref>] or they could be utilized by community members via cross-feeding, establishing the basis for simple food webs [<xref rid="pgen-0020103-b040" ref-type="bibr">40</xref>&#x02013;<xref rid="pgen-0020103-b042" ref-type="bibr">42</xref>]. While we do not know whether such trophic interactions had already emerged in our experiment, the genetic and phenotypic divergence among the competitors suggest that this possibility should be considered. Further work is required to test if the reduction of stochastic effects with an increasing number of competitor genotypes is limited to co-evolved competitors or if similar effects could be obtained by independently evolved genotypes.</p><p>Assuming that our findings from <named-content content-type="genus-species">E. coli</named-content> can be extrapolated to other communities, our results imply that disturbed ecosystems characterized by reduced diversity (compared to undisturbed systems, which contain more functional groups) might be more affected by stochastic effects of population dynamics than are complex (undisturbed) systems.</p></sec></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Experimental background.</title><p>Starting from a single <named-content content-type="genus-species">E. coli</named-content> cell we performed an evolution experiment to build a simple community consisting of diverged <named-content content-type="genus-species">E. coli</named-content> lineages with possible interactions at different levels. The population evolved in rich medium to avoid restrictions in adaptability due to the culture medium. Thus, the population could develop in a complex medium (environment) that fostered the possibility of a broad spectrum of niches and trophic interactions among the members of the evolving community.</p></sec><sec id="s4b"><title>Bacterial strain, culture conditions, and detection of the adaptive event.</title><p>In brief, the evolution experiment was performed with the common laboratory strain <named-content content-type="genus-species">Escherichia coli</named-content> XL1 blue (<italic>recA1 end A1 gyr A96 thi-1 hsdR17 sup E44 relA1 lac</italic> [F&#x02032; <italic>pro AB lacI^q Z&#x00394;M15 Tn10</italic>]) (Stratagene, La Jolla, California, United States). Cells were maintained by serial transfer in 5 ml of rich medium (Lennox L Broth Base, GIBCO BRL, San Diego, California, United States) at 37 &#x000b0;C and 250 rpm. Every 12 h the population was diluted 1:500, allowing about nine generations per transfer. Bacterial density at transfer was ~5 &#x000d7; 10⁸ cells/ml. The number of generations per growth cycle (we use the variable <italic>g</italic> to indicate growth cycle number) was taken from [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>].</p><p> <named-content content-type="genus-species">E. coli</named-content> cells carry a highly variable (GA)_n microsatellite marker. We determined the length of this marker by a restriction digest that cleaved in the sequence flanking the microsatellite. Subsequent electrophoresis separated the microsatellite alleles of different sizes [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>]. The frequency of each microsatellite allele was estimated by the relative intensity of the corresponding allele.</p><p>For the competition experiments, we measured the frequency of a bacterial genotype by quantitating the intensity of the microsatellite allele associated with that genotype. The frequency change was determined by a restriction analysis at the beginning and end of the experiment. For further details on the analytical procedure, see Imhof [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>].</p></sec><sec id="s4c"><title>Isolation of clones from the sweeper lineage.</title><p>At generation 324 we isolated four clones carrying the microsatellite allele that had rapidly increased in frequency. We performed a series of competition experiments using the competitor genotypes 8, 13, and 21 (see below; <xref ref-type="table" rid="pgen-0020103-t001">Table 1</xref>), and all three clones from the sweeper lineage yielded a similar selection coefficient (Kruskal-Wallis H test, &#x003c7;² = 1.150, <italic>p</italic> = 0.765, <italic>n</italic> = 53; unpublished data). On the basis of these results, we concluded that no heterogeneity among the four sweeper genotypes exists; thus they were used interchangeably for the remaining experiments.</p></sec><sec id="s4d"><title>Derivation and characterization of competitors.</title><p>A subsample of the evolving population was plated at generation 324, a few generations before the advantageous genotype was fixed, but its increase in frequency was already recognizable (<xref ref-type="fig" rid="pgen-0020103-g001">Figure 1</xref>). At that stage the heterogeneity of the population was still high. Thus, sufficiently differentiated genotypes could be isolated for the consecutive competition experiments.</p><p>Based on the microsatellite allele, we categorized the isolated clones into the groups of competitors (not carrying the beneficial mutation and the microsatellite allele was different from 33 repeats) and sweepers (carrying the beneficial mutation and the 33-repeat microsatellite allele). In total, we isolated 18 competitor cells with microsatellite alleles sized between 11 and 28 repeats (<xref ref-type="supplementary-material" rid="pgen-0020103-st001">Table S1</xref>).</p><p>In addition to the genetic heterogeneity at the microsatellite, we characterized tetracycline resistance and adherence (floating versus adherent cells) to test for genetic heterogeneity of the competitors (<xref ref-type="supplementary-material" rid="pgen-0020103-st002">Table S2</xref>).</p><p>To account for potential heterogeneity among derivatives from the sweeper lineage, four distinct clones carrying allele 33 were isolated and characterized. We observed no phenotypic differences (for example, tetracycline resistance and levels of adherence, <xref ref-type="supplementary-material" rid="pgen-0020103-st001">Table S1</xref>) among the four sweeper genotypes.</p></sec><sec id="s4e"><title>Competition experiments.</title><p>Competing cells were grown separately for one growth cycle (about 12 h) to reach comparable physiological states. LB medium (5 ml) was inoculated with 5 &#x003bc;l of the sweeper and a total of 5 &#x003bc;l of the competitor(s), and the culture allowed to grow for one growth cycle (12 h). At that time point, we started the competition experiment by serial transfer of a 1:500 dilution every 12 h. The rest of the starting culture was harvested, and the frequency of each competitor was determined by the intensity of the microsatellite alleles (for details see [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>]). At the end of the competition experiment, we determined the frequencies of the competing genotypes in the same way. Competition time was on average 45 generations. This time span was long enough to achieve unambiguous results, while the probability of new positive mutations arising and spreading to a detectable frequency was rather low (with a beneficial mutation rate <italic>&#x003bc;</italic> = 4 &#x000d7; 10^&#x02212;9 per cell generation [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>]). Similarly, other mutations diversifying the competitors and thus interfering with our results should have been negligible.</p><p>Competition experiments were performed between the sweeper clone carrying beneficial mutation(s) against an increasing diversity level of competitor cells. Therefore, experiments were started with single competitors and continued with different arbitrarily selected strain combinations from a total pool of 18 isolated genotypes (<xref ref-type="supplementary-material" rid="pgen-0020103-st001">Table S1</xref>). As highest diversity treatment, the entire collection of diverging genotypes (represented by the entire population, consisting of more than 18 genotypes) was used (<xref ref-type="table" rid="pgen-0020103-t001">Table 1</xref>).</p><p>With 18 different competitor cells, an extremely large number of combinations is possible, preventing a systematic investigation of the effect of the genotype on the competition experiments.</p><p>First we performed a pre-test for the highest diversity treatment. In this treatment the carrier of the beneficial mutation (sweeper) competed against the entire collection of diverging genotypes (entire population) when the beneficial mutation was absent (generation 162). To ensure the absence of the beneficial mutation at generation 162, we incubated cells from this generation without adding a clone carrying the beneficial mutation. There was no increase in frequency of any of the microsatellite allele with 33 repeats (the size of the clone with the beneficial mutation) detected.</p><p>To rule out artefacts and frequency dependence effects, competition experiments were performed with the same competitor(s)/sweeper combinations on different experimental days, starting with different overnight cultures and frequencies. The number of experiments performed for each of the 13 combinations studied is given in <xref ref-type="table" rid="pgen-0020103-t001">Table 1</xref>. Each of these experiments was done in triplicate. Triplicates were started from the same culture (thus from the same overnight culture and same frequency of each competitor genotype and the sweeper) but independently cultivated, harvested, and analysed.</p><p>In none of the competition experiments was a new selective sweep detected. This observation is consistent with previous results in the same system in which adaptive events exhibited selection coefficients in the range of Malthusian fitness parameter <italic>m</italic> = 0.01 to 0.06. With a starting frequency of a new mutation of <italic>P</italic> = 1/2<italic>N</italic> and population size of 5 &#x000d7; 10⁶ at transfer it would take approximately 3,085 to 514 generations, respectively, for a new advantageous mutation to become fixed [<xref rid="pgen-0020103-b019" ref-type="bibr">19</xref>]. Also no substantial diversification was noted within the rather short experimental time.</p><p>Next, it can be assumed that fitness was transitive in our study, although our approach did not allow a detailed analysis of the fitness between all nonsweeping competitors. Nevertheless, a recent study with a similar experimental setting found no evidence for nontransitivity and thus validates our assumptions [<xref rid="pgen-0020103-b043" ref-type="bibr">43</xref>].</p></sec><sec id="s4f"><title>Determination of fitness parameter <italic>m.</italic> </title><p>The Malthusian fitness parameter <italic>m</italic> can be determined from the frequency change of the carrier of the advantageous mutation [<xref rid="pgen-0020103-b044" ref-type="bibr">44</xref>]: <disp-formula id="pgen-0020103-e001"><graphic xlink:href="pgen.0020103.e001.jpg" position="anchor" mimetype="image"/></disp-formula>where <italic>m</italic> _ij is given per generation. <italic>P</italic> _i is the frequency of the selected lineage at time point <italic>t</italic> of the experiment. The frequency of genotypes that do not belong to the selected lineage <italic>P</italic> _j = 1 &#x02212; <italic>P_i</italic>. Competition time measured in generations is specified by <italic>g</italic>. The competition time differed among experiments, with a mean of 44 generations and a standard deviation of 9 (minimum = 36; maximum = 90). </p></sec><sec id="s4g"><title>Data analysis.</title><p>As the number of experiments performed for each competitor genotype and combination of competitors differed (see <xref ref-type="table" rid="pgen-0020103-t001">Table 1</xref>) and some competitor genotypes had a strong effect on the outcome of the competition experiments (see <xref ref-type="supplementary-material" rid="pgen-0020103-sg001">Figure S1</xref>), we devised a special resampling strategy to account for the heterogeneity. For each level of complexity we performed multiple competition experiments for each competitor/combination of competitors. Each of these experiments was performed in triplicate (that is, it was performed in parallel, from the identical overnight culture under the identical conditions). Hence, we had a nested experimental design with multiple competition experiments each consisting of three parallel replicates.</p><p>We calculated the mean fitness <italic>m&#x000af;</italic> as follows: <disp-formula id="pgen-0020103-e002"><graphic xlink:href="pgen.0020103.e002.jpg" position="anchor" mimetype="image"/></disp-formula> </p><p>The unit of resampling is one competition experiment with the corresponding three replicate cultures (<italic>m</italic> _i1 to <italic>m</italic> _i3). For each of the <italic>g</italic> different competitors (competitor combinations) that were used for a given level of complexity, one competition experiment (with its three replicates) was selected randomly with replacement (bootstrapping [<xref rid="pgen-0020103-b045" ref-type="bibr">45</xref>]). We performed 100 bootstrap replications. The standard deviation was calculated accordingly. Note, that this procedure gives equal weight to each different competitor genotype (combination of genotypes).</p><p>The variance in competition experiments (reproducibility) was measured slightly differently. Rather than averaging the replicates <italic>m</italic> _i1 to <italic>m</italic> _i3 we calculated their coefficient of variation and averaged it over <italic>g</italic> different competitors (competitor combinations).</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting Information</title><supplementary-material content-type="local-data" id="pgen-0020103-sg001"><label>Figure S1</label><caption><title>Mean Fitness of the Clone Carrying the Beneficial Mutation (Sweeper) against Individual Competitors and the Entire Population</title><p>(46 KB PPT)</p></caption><media xlink:href="pgen.0020103.sg001.ppt"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020103-st001"><label>Table S1</label><caption><title>Attributes of 22 Genotypes Isolated for the Competition Experiments</title><p>(39 KB DOC)</p></caption><media xlink:href="pgen.0020103.st001.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020103-st002"><label>Table S2</label><caption><title>Criteria for Levels of Adherence</title><p>(27 KB DOC)</p></caption><media xlink:href="pgen.0020103.st002.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material></sec>

Modeling Chromosomes in Mouse to Explore the Function of Genes, Genomic Disorders, and Chromosomal Organization

<p>One of the challenges of genomic research after the completion of the human genome project is to assign a function to all the genes and to understand their interactions and organizations. Among the various techniques, the emergence of chromosome engineering tools with the aim to manipulate large genomic regions in the mouse model offers a powerful way to accelerate the discovery of gene functions and provides more mouse models to study normal and pathological developmental processes associated with aneuploidy. The combination of gene targeting in ES cells, recombinase technology, and other techniques makes it possible to generate new chromosomes carrying specific and defined deletions, duplications, inversions, and translocations that are accelerating functional analysis. This review presents the current status of chromosome engineering techniques and discusses the different applications as well as the implication of these new techniques in future research to better understand the function of chromosomal organization and structures.</p>

<contrib contrib-type="author"><name><surname>Brault</surname><given-names>V&#x000e9;ronique</given-names></name></contrib><contrib contrib-type="author"><name><surname>Pereira</surname><given-names>Patricia</given-names></name></contrib><contrib contrib-type="author"><name><surname>Duchon</surname><given-names>Arnaud</given-names></name></contrib><contrib contrib-type="author"><name><surname>H&#x000e9;rault</surname><given-names>Yann</given-names></name><xref ref-type="corresp" rid="cor1">*</xref></contrib>

PLoS Genetics

<sec id="s1"><title>Introduction</title><p>Recent strategies to produce mouse lines that contain large genomic rearrangements represent a major advance to accelerate functional genomics and to provide animal models for developmental processes and human diseases such as contiguous gene and gene dosage effect syndromes. The first chromosomal rearrangements were obtained in the mouse using X-ray irradiation [<xref rid="pgen-0020086-b001" ref-type="bibr">1</xref>] or chemicals [<xref rid="pgen-0020086-b002" ref-type="bibr">2</xref>]. However, the size and position of the induced rearrangements cannot be predetermined even though the use of specific selectable markers and embryonic stem (ES) cells has improved the irradiation strategy, making it possible to generate a series of interstitial deletions at a given locus [<xref rid="pgen-0020086-b003" ref-type="bibr">3</xref>&#x02013;<xref rid="pgen-0020086-b006" ref-type="bibr">6</xref>]. Homologous recombination in ES cells using replacement vectors has generated deletion of genomic fragments up to 30 kb [<xref rid="pgen-0020086-b007" ref-type="bibr">7</xref>,<xref rid="pgen-0020086-b008" ref-type="bibr">8</xref>], but inducing large defined chromosomal rearrangements was only achieved in the mid-1990s by taking advantages of the Cre/<italic>loxP</italic> recombinase system [<xref rid="pgen-0020086-b009" ref-type="bibr">9</xref>&#x02013;<xref rid="pgen-0020086-b012" ref-type="bibr">12</xref>] and defining the chromosomal engineering strategy.</p><p>This technology led to the creation of new genetic tools for functional analysis of the mouse genome. Generation of deletions with engineered visible markers provides segmental haploidy to study recessive mutations [<xref rid="pgen-0020086-b010" ref-type="bibr">10</xref>,<xref rid="pgen-0020086-b013" ref-type="bibr">13</xref>], whereas inversions can serve as balancer chromosomes to prevent crossing-over and to facilitate large-scale mutagenesis screens for recessive lethal mutations [<xref rid="pgen-0020086-b014" ref-type="bibr">14</xref>,<xref rid="pgen-0020086-b015" ref-type="bibr">15</xref>]. The possibility of manipulating large chromosome fragments or whole chromosomes using microcell-mediated chromosome transfer (MMCT) offers the opportunity to study the function of large genes or clusters of genes and provides more and more mouse models to study human pathologies such as contiguous gene syndromes. In this review, we describe the panel of techniques available for chromosome engineering in the mouse, some of their applications for studying gene function and genomic organization and for modeling human diseases, and the implications for future research.</p></sec><sec id="s2"><title>Chemical and Radiation-Induced Chromosome Rearrangements</title><p>Historically, various types of rearrangements including deletions, inversions, and reciprocal translocations were obtained through irradiation or chemical mutagenesis. Such chromosomal configurations are important tools for looking at recessive lethal mutations in mice [<xref rid="pgen-0020086-b016" ref-type="bibr">16</xref>] or to obtain mouse models of partial aneuploidy. For example, Ts65Dn is a well-known model to study human trisomy 21 that recapitulates several phenotypic features of people with Down syndrome [<xref rid="pgen-0020086-b017" ref-type="bibr">17</xref>,<xref rid="pgen-0020086-b018" ref-type="bibr">18</xref>].</p><p>A major improvement of the radiation strategy was developed to induce a panel of deletions at a defined region [<xref rid="pgen-0020086-b005" ref-type="bibr">5</xref>]. Starting from the integration of a negative selection marker such as the <italic>tk</italic> gene into a predetermined locus by homologous recombination in a hybrid ES cell line, You et al. [<xref rid="pgen-0020086-b005" ref-type="bibr">5</xref>,<xref rid="pgen-0020086-b019" ref-type="bibr">19</xref>] were able to select for the loss of the negative marker after irradiation. The characterization of the deletions is then achieved by taking advantage of genetic markers that are polymorphic in the hybrid ES cells (<xref ref-type="fig" rid="pgen-0020086-g001">Figure 1</xref>). Using this approach, Schimenti et al. [<xref rid="pgen-0020086-b003" ref-type="bibr">3</xref>] generated three overlapping deletion complexes spanning about 40 cM of mouse chromosome 5 (MMU5), allowing the systematic characterization of the functional regions of this chromosome by pairwise combination of the deletions through mating and serving as a model of the Wolf-Hirschhorn contiguous gene syndrome that resides in this region [<xref rid="pgen-0020086-b003" ref-type="bibr">3</xref>,<xref rid="pgen-0020086-b020" ref-type="bibr">20</xref>].</p><fig id="pgen-0020086-g001" position="float"><label>Figure 1</label><caption><title>Generation and Characterization of Radiation-Induced Deletion Complexes in Mouse ES Cells</title><p>(A) Insertion of a negative selectable marker (cassette Neo-tk: Hsv-thymidine kinase/neomycin resistance) into a predetermined locus by homologous recombination in F1 hybrid (129/SvJae x C57BL/6J) ES cells (C57BL/6J chromosome represented with a black centromere; 129/SvJae chromosome represented with a white centromere).</p><p>(B) Treatment of the neomycin-resistant targeted cells with radiation to induce the deletions.</p><p>(C) Selection in medium containing 1,2&#x02032;-deoxy-2&#x02032;-fluoro-&#x003b2;-D-arabinofuranosyl-5-iodouracil of the colonies having lost the <italic>tk</italic> gene.</p><p>(D) Characterization of the deletion breakpoints by amplification of the DNA from these clones using primers corresponding to genetic polymorphic markers (represented under the chromosome map by letters a&#x02013;j) flanking the site of the targeted integration. Deletions are represented as solid boxes.</p></caption><graphic xlink:href="pgen.0020086.g001"/></fig></sec><sec id="s3"><title>Using Site-Specific Recombinase to Generate Megabase Chromosome Rearrangements</title><sec id="s1a"><title/><sec id="s1a1"><title>In vitro.</title><p>More defined rearrangements require the use of the Cre/<italic>loxP</italic> technology in mouse ES cells [<xref rid="pgen-0020086-b009" ref-type="bibr">9</xref>&#x02013;<xref rid="pgen-0020086-b011" ref-type="bibr">11</xref>]. By two consecutive events one can target the <italic>loxP</italic> sites to predefined loci in the genome by homologous recombination. Expression of the Cre recombinase induces the desired rearrangement through site-specific recombination between the two <italic>loxP</italic> sites, depending on their relative position and orientation (<xref ref-type="fig" rid="pgen-0020086-g002">Figure 2</xref>). For small regions (less than 100 kb), deletions and inversions can be easily generated, starting from ES cells carrying two <italic>loxP</italic> sites inserted in a <italic>cis</italic> configuration, that are treated with Cre recombinase [<xref rid="pgen-0020086-b021" ref-type="bibr">21</xref>,<xref rid="pgen-0020086-b022" ref-type="bibr">22</xref>]. However, the rearrangement occurs at a lower efficiency for larger regions. Thus, targeting vectors containing the 5&#x02032; or 3&#x02032; part of a positive selection cassette with <italic>loxP</italic> located downstream or upstream, respectively, were designed. Expression of the Cre recombinase results in the desired rearrangement through site-specific recombination between the two <italic>loxP</italic> sites, enabling the selection of the recombined allele through the restoration of the selection marker upon rearrangement. Different markers such as <italic>Hprt</italic> [<xref rid="pgen-0020086-b010" ref-type="bibr">10</xref>] or the resistance genes for neomycin [<xref rid="pgen-0020086-b023" ref-type="bibr">23</xref>], puromycin [<xref rid="pgen-0020086-b024" ref-type="bibr">24</xref>,<xref rid="pgen-0020086-b025" ref-type="bibr">25</xref>], or hygromycin [<xref rid="pgen-0020086-b026" ref-type="bibr">26</xref>] were used. As an alternative, a <italic>tk</italic> negative selection marker can be deleted in the rearranged locus [<xref rid="pgen-0020086-b027" ref-type="bibr">27</xref>&#x02013;<xref rid="pgen-0020086-b030" ref-type="bibr">30</xref>].</p><fig id="pgen-0020086-g002" position="float"><label>Figure 2</label><caption><title>The Different Types of Chromosomal Rearrangements Produced by the Cre/<italic>loxP</italic> Recombinase System</title><p>Deletions, duplications, or inversions can be produced depending on the relative orientation of the <italic>loxP</italic> sites, on their position on the homologous chromosome (i.e., in <italic>cis</italic> or in <italic>trans</italic>), and on the cell cycle stage during which the Cre-mediated recombination occurs (G1 or G2).</p><p>(A) Recombination between <italic>loxP</italic> sites (green arrowhead) integrated in the same orientation in a <italic>cis</italic> configuration during the G1 phase can only generate a deletion of the region of interest (red arrow); the same configuration in the G2 phase can result in the creation of a deletion and a duplication.</p><p>(B) The deletion and the corresponding duplication can also be obtained from a <italic>trans</italic> configuration in both G1 and G2 phases. This represents the best configuration to establish the deleted and duplicated chromosomes in the mouse, with both chromosomes compensating for each other with regard to genetic dosage, thus reducing the potential consequence of haploinsufficiency.</p><p>(C and D) When the <italic>loxP</italic> sites are oriented in opposite directions in a <italic>cis</italic> configuration, an equilibrium with two forms, inverted and non-inverted, is obtained if the Cre is expressed in G1, while a more likely unstable recombined pair of acentric and dicentric chromosomes is generated if Cre reacts on <italic>loxP</italic> sites after the S phase or from a <italic>trans</italic> configuration. From all these recombinant alleles, only those containing the reconstituted mini-gene, however, will be retained during the selection in vitro. Recombinant ES-cell clones should be extensively characterized to verify the engineered chromosome (by Southern blot analysis, normal and quantitative PCR, or FISH).</p><p>Del, deletion; Dup, duplication; Inv, inversion; Rec, one of the original recombinant alleles.</p></caption><graphic xlink:href="pgen.0020086.g002"/></fig><p>The efficiency of the in vitro technique, and hence its feasibility, depends on the chromosomal context and more dramatically on the design of the experiment [<xref rid="pgen-0020086-b031" ref-type="bibr">31</xref>&#x02013;<xref rid="pgen-0020086-b034" ref-type="bibr">34</xref>]. Cre-mediated recombination efficiency is dependent on different factors such as the increasing distance between the <italic>loxP</italic> sites for a <italic>cis</italic> configuration (10% to 0.1% efficiency), the level of recombinase activity, and the region of interest that could induce ES-cell lethality after deletion [<xref rid="pgen-0020086-b010" ref-type="bibr">10</xref>,<xref rid="pgen-0020086-b031" ref-type="bibr">31</xref>,<xref rid="pgen-0020086-b032" ref-type="bibr">32</xref>,<xref rid="pgen-0020086-b035" ref-type="bibr">35</xref>] (unpublished data). Increasing the efficiency of the technique by a factor of 10 can be achieved by using a GFP/Cre expressing vector or classical co-transfection, followed by fluorescent-activated cell sorting of GFP+ cells that also express Cre [<xref rid="pgen-0020086-b036" ref-type="bibr">36</xref>,<xref rid="pgen-0020086-b037" ref-type="bibr">37</xref>].</p><p>Another dimension of the chromosomal engineering in vitro is the induction of mitotic recombination in G2 phase to produce selectable homozygous daughter cells from a double heterozygous mother for genetic mosaics. Such a powerful method can selectively produce ES-cell clones carrying homozygous mutations for functional recessive mutations screens in vitro, speeding up the analysis of a gene's function in cells [<xref rid="pgen-0020086-b038" ref-type="bibr">38</xref>,<xref rid="pgen-0020086-b039" ref-type="bibr">39</xref>].</p><p>Using the in vitro technique requires two targeting vectors for each bordering loci that could contain different types of selectable cassettes for inducing and selecting chromosomal rearrangements in various types of ES cells. But in the case of the restoration of the HPRT function, the use of <italic>Hprt</italic>-deficient ES cells such as AB2.2 [<xref rid="pgen-0020086-b010" ref-type="bibr">10</xref>] or HM-1 [<xref rid="pgen-0020086-b040" ref-type="bibr">40</xref>] is required. Nevertheless, this is a method of choice given the number of ready-to-use targeting vectors for the <italic>Hprt</italic> selection system available from the Mutagenic Insertion and Chromosome Engineering Resource [<xref rid="pgen-0020086-b041" ref-type="bibr">41</xref>] (MICER; http://www.sanger.ac.uk/micer/), the use of retroviral vectors to target the second <italic>loxP</italic> site [<xref rid="pgen-0020086-b042" ref-type="bibr">42</xref>&#x02013;<xref rid="pgen-0020086-b044" ref-type="bibr">44</xref>], and the panel of rearranged chromosomes that has been engineered [<xref rid="pgen-0020086-b015" ref-type="bibr">15</xref>,<xref rid="pgen-0020086-b041" ref-type="bibr">41</xref>]. Furthermore, the presence of coat color markers in those targeting vectors allows an easy discrimination of mice carrying the recombined chromosome [<xref rid="pgen-0020086-b013" ref-type="bibr">13</xref>].</p><p>The in vitro strategy was developed extensively to study specific loci containing large genes or clusters of genes (see <xref ref-type="table" rid="pgen-0020086-t001">Table 1</xref>) [10,21,22,25&#x02013;28,45&#x02013;56], to generate translocations or deletions similar to those found in cancer [9,11,30&#x02013;32,57&#x02013;62], or to create models of contiguous gene syndromes, such as the Smith-Magenis [<xref rid="pgen-0020086-b044" ref-type="bibr">44</xref>,<xref rid="pgen-0020086-b063" ref-type="bibr">63</xref>&#x02013;<xref rid="pgen-0020086-b066" ref-type="bibr">66</xref>], Prader-Willy [<xref rid="pgen-0020086-b067" ref-type="bibr">67</xref>], and DiGeorge syndromes [<xref rid="pgen-0020086-b043" ref-type="bibr">43</xref>,<xref rid="pgen-0020086-b068" ref-type="bibr">68</xref>,<xref rid="pgen-0020086-b069" ref-type="bibr">69</xref>]. Similarly, new mouse models of Down syndrome have been generated to study the impact of a critical region identified in human Chromosome 21 [<xref rid="pgen-0020086-b023" ref-type="bibr">23</xref>]. Taken together, these data illustrate how large-scale chromosomal engineering is a powerful tool to unravel genes with dosage effects and their contribution to aneuploid syndromes in the mouse model.</p><table-wrap id="pgen-0020086-t001" content-type="2col" position="float"><label>Table 1</label><caption><p>Genomic Rearrangements Induced in the Mouse Genome by Chromosome Engineering</p></caption><graphic xlink:href="pgen.0020086.t001"/></table-wrap><p>Unraveling lethal mutations is difficult in classical mutagenesis screens, unless a chromosome carrying an inversion is used as a balancer chromosome [<xref rid="pgen-0020086-b015" ref-type="bibr">15</xref>]. The production of inversions by chromosome engineering in the mouse offers the advantage of controlling the size of the inverted region and of generating inversions that will induce lethality in homozygous state, such as those obtained on mouse Chromosome 11 that disrupt the <italic>Wnt3</italic> gene [<xref rid="pgen-0020086-b014" ref-type="bibr">14</xref>,<xref rid="pgen-0020086-b031" ref-type="bibr">31</xref>,<xref rid="pgen-0020086-b032" ref-type="bibr">32</xref>,<xref rid="pgen-0020086-b070" ref-type="bibr">70</xref>] or on Chromosome 4 [<xref rid="pgen-0020086-b071" ref-type="bibr">71</xref>]. Using such a balancer chromosome approach associated with chemical mutagenesis, two mutagenesis screens led to the characterization of, respectively, 59 recessive lethal mutations on mouse Chromosome 11, and 19 mutations on mouse Chromosome 4 [<xref rid="pgen-0020086-b014" ref-type="bibr">14</xref>,<xref rid="pgen-0020086-b072" ref-type="bibr">72</xref>].</p></sec><sec id="s3X0X2"><title>In vivo.</title><p>Cre-mediated recombination is widely used to generate conditional mutagenesis for a gene of interest in a tissue- or time-dependent manner [<xref rid="pgen-0020086-b073" ref-type="bibr">73</xref>]. Similarly, several groups succeeded in generating in vivo large deletions, duplications, inversions, and translocations (<xref ref-type="table" rid="pgen-0020086-t001">Table 1</xref>). This can be carried out between two <italic>loxP</italic> sites in <italic>cis</italic> either inserted in ES cells [<xref rid="pgen-0020086-b022" ref-type="bibr">22</xref>,<xref rid="pgen-0020086-b074" ref-type="bibr">74</xref>&#x02013;<xref rid="pgen-0020086-b076" ref-type="bibr">76</xref>] or selected after classical crossing-over between two original founder lines by the sequential targeted recombination induced genomic rearrangement (STRING) (<xref ref-type="fig" rid="pgen-0020086-g003">Figure 3</xref>), [<xref rid="pgen-0020086-b024" ref-type="bibr">24</xref>,<xref rid="pgen-0020086-b077" ref-type="bibr">77</xref>,<xref rid="pgen-0020086-b078" ref-type="bibr">78</xref>]. Furthermore, the targeted meiotic recombination (TAMERE) strategy allows to obtain both a deletion and a duplication of a region of up to 150 kb, starting from two <italic>loxP</italic> sites, in a <italic>trans</italic> configuration, with a frequency that can vary from 1% to 10% (<xref ref-type="fig" rid="pgen-0020086-g003">Figure 3</xref>) [<xref rid="pgen-0020086-b012" ref-type="bibr">12</xref>,<xref rid="pgen-0020086-b079" ref-type="bibr">79</xref>]. Similarly to in vitro, the recombination frequency decreases in vivo for large regions. Nevertheless it is still effective for <italic>cis</italic> configuration (0.3%&#x02013;1%) separated by up to 28 Mb [<xref rid="pgen-0020086-b078" ref-type="bibr">78</xref>], but seems to be not workable with <italic>loxP</italic> in <italic>trans</italic> distant from 3.9 Mb [<xref rid="pgen-0020086-b034" ref-type="bibr">34</xref>]. Thus, the STRING approach [<xref rid="pgen-0020086-b078" ref-type="bibr">78</xref>] offers an interesting alternative to the MICER strategy, without the need for extensive technological investments, but still requires an efficient and large genotyping program.</p><fig id="pgen-0020086-g003" position="float"><label>Figure 3</label><caption><title>Strategies for In Vivo Cre-Mediated Recombination</title><p>(A) The general principle of TAMERE is based on two successive breedings in order to have in one male, named the trans-loxer, the <italic>Sycp1Cre</italic> (<italic>Synaptonemal Complex protein 1</italic>) transgene and the two <italic>loxP</italic> sites in a <italic>trans</italic> configuration, inserted previously in the same orientation at each targeted locus, that define the genetic interval. The <italic>Synaptonemal Complex protein 1</italic> promoter drives Cre expression at prophase of meiosis in male spermatocytes when chromatid pairs are closely aligned, in order to facilitate the chromatid exchange, leading to the formation of the deletion and the duplication of the interval delimited by the two <italic>loxP</italic> sites. The last step consists in mating trans-loxer males with wild-type females to generate, in the progeny, individuals carrying the deletion or the duplication of the targeted region.</p><p>(B) The STRING approach takes advantage of a classical crossing-over to bring the two <italic>loxP</italic> sites into a <italic>cis</italic> configuration to generate a deletion. Two parental mice (F₀) carrying <italic>loxP</italic> sites flanking a selected region are crossed. The F₁ progeny containing the two <italic>loxP</italic> sites are then mated to wild-type mice. The offspring are screened for meiotic crossing-over between both sites leading to mice carrying the <italic>loxP</italic> sites in a <italic>cis</italic> configuration. In the subsequent cross, a ubiquitously expressed Cre transgene is introduced, generating the deletion that is established in the next F₄ generation.</p><p>Del, deletion; Dup, duplication.</p><p>red arrow, region of interest; green arrow, <italic>loxP</italic> site.</p></caption><graphic xlink:href="pgen.0020086.g003"/></fig><p>In vivo chromosomal engineering during mitosis often leads to mosaic individuals in the first progeny in which the <italic>Cre</italic> transgene is combined with <italic>loxP</italic> sites in <italic>cis</italic> [<xref rid="pgen-0020086-b074" ref-type="bibr">74</xref>,<xref rid="pgen-0020086-b080" ref-type="bibr">80</xref>]. The level of mosaicism found depends on a balance between the size of the targeted region and Cre activity. For example, Mersher et al. (2001) noticed 17% of mosaic mice for a 1.5-Mb deletion using the <italic>ZP3-Cre</italic> line expressed in the oocytes, and we found 26% of mice showing a mosaic profile for a 0.7-Mb induced deletion using a <italic>CMV-Cre</italic> expressing line [<xref rid="pgen-0020086-b081" ref-type="bibr">81</xref>]. Genetic mosaicism could be an advantage for lineage analysis as Cre-induced rearrangements can be used to label and trace a cell population in vivo [<xref rid="pgen-0020086-b082" ref-type="bibr">82</xref>]. Thus, it is crucial to well characterize the Cre transgenic lines, by analyzing the expression of Cre mRNA, or the Cre activity by mating with a mouse carrying a reporter gene such as the R26R [<xref rid="pgen-0020086-b083" ref-type="bibr">83</xref>] and ACZL lines [<xref rid="pgen-0020086-b084" ref-type="bibr">84</xref>].</p><p>Large rearrangements can lead to early embryonic lethality. Hence, absence of mice carrying the new engineered chromosome in the in vivo approach could be caused by a failure of Cre-mediated recombination but could also depend on the lethal effect of the new genetic configuration [<xref rid="pgen-0020086-b085" ref-type="bibr">85</xref>]. To address such questions, it is important to test for the presence of the new chromosomal configuration in the carrier double-transgenic animal [<xref rid="pgen-0020086-b078" ref-type="bibr">78</xref>]. An alternative would be to compensate the deletion with a balanced duplication [<xref rid="pgen-0020086-b085" ref-type="bibr">85</xref>]. Unfortunately, an efficient strategy needs to be explored to induce such <italic>trans</italic> recombination in vivo for large regions. An additional way to overcome this problem is to trigger the recombination in a time- and/or tissue-specific manner, but only a few reports so far have described this option [<xref rid="pgen-0020086-b031" ref-type="bibr">31</xref>,<xref rid="pgen-0020086-b061" ref-type="bibr">61</xref>,<xref rid="pgen-0020086-b062" ref-type="bibr">62</xref>].</p><p>In vivo approaches to making deletions and inversions were largely used to explore the function and regulation of genes (<xref ref-type="table" rid="pgen-0020086-t001">Table 1</xref>). The best example for such studies is the extensive work done over the last ten years by D. Duboule and collaborators to analyze the molecular mechanisms that modulate expression of <italic>Hox</italic> genes encoding transcription factors controlling positional information along the trunk and limb axes [22,45,46,76,86&#x02013;88]. In particular, they have combined the in vivo strategy with knock-in of a <italic>lacZ</italic> reporter gene, to trace the consequences of the induced rearrangements on the expression of the <italic>Hoxd</italic> genes during embryonic development. Similar deletion approaches were used to study <italic>cadherin 1</italic> gene regulation [<xref rid="pgen-0020086-b075" ref-type="bibr">75</xref>]. Herault et al. [<xref rid="pgen-0020086-b089" ref-type="bibr">89</xref>] extended this strategy by replacing the <italic>LacZ</italic> by the <italic>Cre</italic> to provide a tool to specifically inactivate <italic>Hox</italic> genes in progressively more extended domains of <italic>Hox</italic> expression. On the whole, the in vivo strategy appears very attractive as it is less expensive, doesn't need the settling of a large breeding program, and avoids a few complicated steps of cell culturing thanks to the increasing availability of mouse or ES cell lines containing <italic>loxP</italic> sites (<xref ref-type="table" rid="pgen-0020086-t002">Table 2</xref>).</p><table-wrap id="pgen-0020086-t002" content-type="1col" position="float"><label>Table 2</label><caption><p>Resources Available Online for Chromosomal Engineering in Mice</p></caption><graphic xlink:href="pgen.0020086.t002"/></table-wrap><p>The mosaic analysis with double markers [<xref rid="pgen-0020086-b082" ref-type="bibr">82</xref>] combines chimeric fluorescent constructs and G2-induced mitotic recombination in vivo. Such a strategy allows determination of the consequence of mutation by visualization of double-colored mutated cells, enabling high-resolution lineage tracing in vivo to elucidate biological processes, such as cell fate in the nervous system. The development of such a strategy will lead to genetic mosaic analysis, bypassing the problem of the lethality linked to some mutations by exploring the mosaic condition.</p></sec></sec></sec><sec id="s4"><title>Transchromosomal Lines</title><p>The Cre/<italic>loxP</italic> system allows the modeling of native chromosomes while MMCT enables manipulation of large fragments or whole chromosomes from various species and particularly human with the making of transchromosomic lines. This technique originally developed in the 1970s, is based on the fusion of microcells, containing single or small numbers of chromosomes, with whole cells in order to transfer exogenous chromosome material into host cells (for a review, see [<xref rid="pgen-0020086-b090" ref-type="bibr">90</xref>,<xref rid="pgen-0020086-b091" ref-type="bibr">91</xref>]). MMCT can be carried out with somatic cells, embryonic carcinoma cells or ES cells as recipients. Applications of MMCT are numerous, ranging from hunting for tumor suppressor, DNA repair or senescence-inducing genes, assessing genomic instability, imprinting and chromatin modification, constructing and manipulating artificial chromosomes for potential gene therapies to act as vectors for specific genes or genomic regions, and expressing proteins or studying aspects of chromosome behaviour in mitosis and meiosis [<xref rid="pgen-0020086-b092" ref-type="bibr">92</xref>&#x02013;<xref rid="pgen-0020086-b095" ref-type="bibr">95</xref>].</p><p>A further application is the creation of mouse lines that carry fragments or whole human chromosomes as freely segregating extra chromosomes. To achieve this, transchromosomic mouse ES cells were injected into mouse blastocysts to produce mouse chimeras. The first transchromosomal animals were produced in the late 1990s and carried different human chromosome fragments as freely segregating extra chromosomes [<xref rid="pgen-0020086-b096" ref-type="bibr">96</xref>]. Using this technique, models of Down syndrome were obtained in chimeric mice with transchromosomal ES cells containing different parts of human Chromosome 21 (Hsa21), ranging from ~50 to ~0.2 Mb [97.98]. But the chromosome fragments tended to be lost during development, leading to phenotypic variations. In spite of this, the chimeric animals exhibited behavioral impairments and cardiac defects similar to those described in humans with Down syndrome [<xref rid="pgen-0020086-b099" ref-type="bibr">99</xref>,<xref rid="pgen-0020086-b100" ref-type="bibr">100</xref>]. E. Fisher and collaborators succeeded in producing a trans-species aneuploid mouse line, called Tc1, that stably transmitted almost a complete Hsa21, generating a more full model of Down syndrome [<xref rid="pgen-0020086-b101" ref-type="bibr">101</xref>]. Indeed, the Tc1 line displays a large set of deficits similar to those observed in trisomic 21 patients that were not found in earlier models. The previously observed failure of transchromosomal germline transmission was overcame by using female mouse ES cell lines to derive female chimeras that support the transmission of the aneuploid chromosome through the germline [<xref rid="pgen-0020086-b097" ref-type="bibr">97</xref>]. Nevertheless such transchromosomic lines are difficult to obtain and to work with. A large number of ES cell clones should be controlled to detect any rearrangements and should be injected to recover a germline transmission. In addition, transmission from Tc1 females only could be achieved with a rate of 40% in a hybrid genetic background, and a large panel of animals should be studied as the aneuploid chromosome tends to be lost during mitosis with variable rates [<xref rid="pgen-0020086-b101" ref-type="bibr">101</xref>]. Nonetheless, transchromosomic lines offer a promising substitute to mouse segmental aneuploidies, as the complete genomic sequence is included, and hence modeling a more complete human aneuploidy.</p></sec><sec id="s5"><title>Conclusion</title><p>Chromosome engineering combined with the transchromosomic approach is widely used for dissecting the function, the regulation, and the contribution of genes to genetic disorders, such as contiguous gene and aneuploidy syndromes. For example, the increasing number of mouse models for trisomy 21 that have been created recently [<xref rid="pgen-0020086-b101" ref-type="bibr">101</xref>,<xref rid="pgen-0020086-b102" ref-type="bibr">102</xref>] provides the necessary tools to understand how dosage imbalance results in the abnormal phenotypes observed in the human patients. Breeding of the different mouse strains carrying either the duplication or the deletion of human Chromosome 21 syntenic regions will not only allow the creation of a full model for human trisomy 21, but also the deciphering of genetic interaction between regions and enable one to look for candidate genes for each phenotype. This approach is also valid for evaluating the effect of copy number variation of genomic regions observed in the human population that might contribute to the susceptibility to certain diseases [<xref rid="pgen-0020086-b103" ref-type="bibr">103</xref>&#x02013;<xref rid="pgen-0020086-b105" ref-type="bibr">105</xref>]. Manipulating chromosomes on a large scale is likely to be increasingly developed in the future also to identify genes underlying complex phenotypes of polygenic diseases such as diabetes, cancer, asthma, or obesity caused by a combination of environmental and multiple genetic factors. To this end, transchromosomic lines combined with deletions might be a promising way to obtain more humanized models to study specific human genes and pathologies.</p><p>Even though both techniques still require considerable effort, they benefit from the development of the MICER resource, with more than 18,000 targeting vectors referenced in the mouse genome database, the increasing number of mouse carrying <italic>loxP</italic> sites at various loci and of Cre transgenic lines (<xref ref-type="table" rid="pgen-0020086-t002">Table 2</xref>). Together with the in vivo approaches TAMERE and STRING [<xref rid="pgen-0020086-b012" ref-type="bibr">12</xref>,<xref rid="pgen-0020086-b078" ref-type="bibr">78</xref>], any laboratory can now manipulate large genomic regions without any additional work on ES cells, in order to analyze further the genomic organization or to derive new tools for genetic analysis (deletion, duplication, and inversion). These different technologies are further supported by the combined effort of the scientific community to establish large-scale gene trap mutagenesis programs, integrating Cre/<italic>loxP</italic> technology and offering the possibility to create a variety of alleles [<xref rid="pgen-0020086-b106" ref-type="bibr">106</xref>&#x02013;<xref rid="pgen-0020086-b108" ref-type="bibr">108</xref>].</p><p>Emerging from these strategies of chromosome engineering is the fascinating aspect of chromosome organization, structure, and function, implicated in the as yet poorly understood code for genetic instructions. Indeed, the human genome contains about 2% of coding sequence, RNA genes, and regulatory regions. In some cases, control regions lay at a great distance from the gene, and hence the long-range interaction is only detected by manipulation of large chromosomal regions [<xref rid="pgen-0020086-b109" ref-type="bibr">109</xref>&#x02013;<xref rid="pgen-0020086-b112" ref-type="bibr">112</xref>]. Surprisingly, a significant amount of the non-coding portion of the genome is under active selection, suggesting that it is also functionally important, yet little is known about it [<xref rid="pgen-0020086-b113" ref-type="bibr">113</xref>&#x02013;<xref rid="pgen-0020086-b116" ref-type="bibr">116</xref>]. Large desert gene regions also are found and are starting to be investigated by using chromosome engineering [<xref rid="pgen-0020086-b030" ref-type="bibr">30</xref>]. We speculate that the modeling of chromosome will be more commonly used to better understand the role of non-coding sequences, and may be used to decipher the &#x0201c;C-value enigma&#x0201d; [<xref rid="pgen-0020086-b117" ref-type="bibr">117</xref>] and the chromosome architecture lying at the heart of genetic instructions. Our increased ability to manipulate the mammalian genome provides us with new tools for the development of a functional chromoso&#x02013;genomic approach, bringing an exciting new dimension into biological and biomedical research.</p></sec><sec sec-type="supplementary-material" id="s6"><title>Supporting Information</title><sec id="s6a"><title>Accession Numbers</title><p>The accession numbers of genes mentioned in this paper from the GenBank database (<ext-link ext-link-type="uri" xlink:href="http://www.ncbi.nlm.nih.gov/entrez">http://www.ncbi.nlm.nih.gov/entrez</ext-link>) are <italic>Wnt3</italic> gene (NM_009521.1) and <italic>cadherin 1</italic> (NM_009864.1).</p></sec></sec>

Genetic Regulation of Unsaturated Fatty Acid Composition in <named-content content-type="genus-species">C. elegans</named-content>

<p>Delta-9 desaturases, also known as stearoyl-CoA desaturases, are lipogenic enzymes responsible for the generation of vital components of membranes and energy storage molecules. We have identified a novel nuclear hormone receptor, NHR-80, that regulates delta-9 desaturase gene expression in <named-content content-type="genus-species">Caenorhabditis elegans</named-content>. Here we describe fatty acid compositions, lifespans, and gene expression studies of strains carrying mutations in <italic>nhr-80</italic> and in the three genes encoding delta-9 desaturases, <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic>. The delta-9 desaturase single mutants display only subtle changes in fatty acid composition and no other visible phenotypes, yet the <italic>fat-5;fat-6;fat-7</italic> triple mutant is lethal, revealing that endogenous production of monounsaturated fatty acids is essential for survival. In the absence of FAT-6 or FAT-7, the expression of the remaining desaturases increases, and this ability to compensate depends on NHR-80. We conclude that, like mammals, <named-content content-type="genus-species">C. elegans</named-content> requires adequate synthesis of unsaturated fatty acids and maintains complex regulation of the delta-9 desaturases to achieve optimal fatty acid composition.</p>

<contrib contrib-type="author"><name><surname>Brock</surname><given-names>Trisha J</given-names></name><xref ref-type="aff" rid="aff1"/></contrib><contrib contrib-type="author"><name><surname>Browse</surname><given-names>John</given-names></name><xref ref-type="aff" rid="aff1"/></contrib><contrib contrib-type="author"><name><surname>Watts</surname><given-names>Jennifer L</given-names></name><xref ref-type="corresp" rid="cor1">*</xref><xref ref-type="aff" rid="aff1"/></contrib>

PLoS Genetics

<sec id="s1"><title>Introduction</title><p>Monounsaturated fatty acids (MUFAs) are key components of membrane phospholipids and triglycerides that play important roles in diverse cellular processes such as membrane function, energy storage, and signaling. MUFAs are synthesized from saturated fatty acids by delta-9 (&#x00394;9) desaturases, also known as stearoyl-CoA desaturases (SCDs), which introduce a double bond between the 9^th and 10^th carbon of a saturated fatty acyl chain. Alterations in the ratio of MUFAs to saturated fatty acids are implicated in heart disease and cancer [<xref rid="pgen-0020108-b001" ref-type="bibr">1</xref>], the two leading causes of death in the United States [<xref rid="pgen-0020108-b002" ref-type="bibr">2</xref>]. The appropriate ratio between MUFAs and saturated fatty acids is maintained by the activity of the &#x00394;9 desaturases, which are subject to complex regulation [<xref rid="pgen-0020108-b003" ref-type="bibr">3</xref>]. As a key control point in metabolic regulation, &#x00394;9 desaturases could be therapeutic targets for treatment of obesity, diabetes, and cardiovascular disease.</p><p>The &#x00394;9 desaturases are ubiquitous enzymes in eukaryotes, found in organisms from yeast to humans. Yeast have one &#x00394;9 desaturase, Ole1p, and mutants that lack this activity are not able to survive without exogenous supplementation of unsaturated fatty acids [<xref rid="pgen-0020108-b004" ref-type="bibr">4</xref>]. Mice have four &#x00394;9 desaturases, each having a unique expression pattern [<xref rid="pgen-0020108-b005" ref-type="bibr">5</xref>,<xref rid="pgen-0020108-b006" ref-type="bibr">6</xref>]. Mutant analysis has revealed distinct roles for SCD1 and SCD2. SCD1 is important for adult energy metabolism and lipid synthesis [<xref rid="pgen-0020108-b007" ref-type="bibr">7</xref>], while SCD2 is involved in lipid synthesis during embryonic development [<xref rid="pgen-0020108-b008" ref-type="bibr">8</xref>]. In humans, two SCD isoforms, hSCD1 and hSCD5, have been described [<xref rid="pgen-0020108-b009" ref-type="bibr">9</xref>,<xref rid="pgen-0020108-b010" ref-type="bibr">10</xref>]. A variety of environmental and physiological signals affect the expression of &#x00394;9 desaturases. Diets rich in unsaturated fatty acids decrease &#x00394;9 desaturase expression, while high carbohydrate consumption increases expression [<xref rid="pgen-0020108-b003" ref-type="bibr">3</xref>]. Decreased temperature leads to increases in &#x00394;9 desaturase gene expression in poikilotherms [<xref rid="pgen-0020108-b011" ref-type="bibr">11</xref>]. In addition, endogenous hormones such as leptin and glucagon cause a decrease in &#x00394;9 desaturase gene expression, while insulin has the opposite effect [<xref rid="pgen-0020108-b003" ref-type="bibr">3</xref>].</p><p>Sterol regulatory element binding proteins (SREBPs) and peroxisome proliferator-activator receptor protein-alpha (PPAR&#x003b1;) have been identified as key transcriptional regulators of SCD1 gene expression in mammals [<xref rid="pgen-0020108-b005" ref-type="bibr">5</xref>]. The SREBP-1 gene encodes a transcription factor that stimulates expression of genes involved in fatty acid biosynthesis, including SCD1 [<xref rid="pgen-0020108-b012" ref-type="bibr">12</xref>], while the SREBP-2 gene product stimulates genes involved in cholesterol biosynthesis [<xref rid="pgen-0020108-b013" ref-type="bibr">13</xref>]. PPAR&#x003b1; is one of a family of nuclear hormone receptors (NHRs), that, upon ligand binding, acts as a heterodimer with the retinoid X receptor to induce transcription of target fat metabolism genes [<xref rid="pgen-0020108-b014" ref-type="bibr">14</xref>]. PPAR&#x003b1;, like all NHRs, contains a hydrophobic pocket for ligand binding and a DNA binding domain for interacting with the promoters of target genes. The targets of PPAR&#x003b1; include genes for the &#x003b2;-oxidation enzymes, SCDs, and other fatty acid desaturases [<xref rid="pgen-0020108-b015" ref-type="bibr">15</xref>,<xref rid="pgen-0020108-b016" ref-type="bibr">16</xref>]. The other members of the PPAR family, PPAR&#x003b4; and PPAR&#x003b3; are also involved in regulation of fat metabolism [<xref rid="pgen-0020108-b017" ref-type="bibr">17</xref>]. These regulators have unique roles due to differences in their gene expression patterns and regulatory activities.</p><p> <named-content content-type="genus-species">Caenorhabditis elegans</named-content> is becoming recognized as an important model for the study of fat metabolism. These animals synthesize a wide variety of fatty acids using a &#x00394;12 desaturase, an &#x00394;3 desaturase, a &#x00394;5 desaturase, a &#x00394;6 desaturase, and three &#x00394;9 desaturases [<xref rid="pgen-0020108-b018" ref-type="bibr">18</xref>,<xref rid="pgen-0020108-b019" ref-type="bibr">19</xref>]. <italic>C. elegans</italic> can also incorporate dietary fatty acids into lipids, allowing researchers to modify the fatty acid composition of live animals [<xref rid="pgen-0020108-b020" ref-type="bibr">20</xref>,<xref rid="pgen-0020108-b021" ref-type="bibr">21</xref>]. In an RNAi (RNA interference) screen, genes were identified that altered fat storage and many of these genes have mammalian counterparts known to function in fat metabolism [<xref rid="pgen-0020108-b022" ref-type="bibr">22</xref>]. In addition, mutant analysis offers insight into pathways known to regulate fat storage in both nematodes and mammals such as the insulin-signaling pathway [<xref rid="pgen-0020108-b023" ref-type="bibr">23</xref>]. A recent study established a role for NHR-49, as a regulator of lipid homeostasis [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. The <italic>nhr-49</italic> mutants have increased levels of the saturated fatty acid 18:0, higher fat accumulation, and a shorter lifespan than wild-type animals. NHR-49 is also required for inducing &#x00394;9 desaturase expression in well-fed animals [<xref rid="pgen-0020108-b025" ref-type="bibr">25</xref>].</p><p>To gain a deeper understanding of fatty acid metabolism in <named-content content-type="genus-species">C. elegans</named-content> we have characterized the three &#x00394;9 desaturase mutants using biochemistry, gene expression, and phenotypic analysis. While the three &#x00394;9 desaturase single mutants, <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> display few differences from wild type, we show that they compensate for loss of one isoform by regulated induction of the remaining &#x00394;9 desaturase genes. This induction depends on NHR-80, a novel NHR that we have identified as a regulator of desaturase expression. Furthermore, the <italic>fat-5;fat-6;fat-7</italic> triple mutant is unable to survive, revealing that endogenous production of monounsaturated fatty acids is essential for survival under standard growth conditions. The &#x00394;9 desaturase genes and their transcriptional regulators are vital for maintaining optimal fatty acid unsaturation and proper membrane composition.</p></sec><sec id="s2"><title>Results/Discussion</title><sec id="s2a"><title>Identification of NHR-80 as a Regulator of Fatty Acid Metabolism</title><p>In our search to identify the desaturases and elongases involved in generation of unsaturated fatty acids in <italic>C. elegans,</italic> we performed a genetic screen to identify mutants with altered fatty acid profiles [<xref rid="pgen-0020108-b018" ref-type="bibr">18</xref>]. In the process of identifying the molecular nature of one mutation, we used RNAi against 156 genes at the end of Chromosome III to determine the fatty acid composition of animals when each of these genes was inactivated. We found an RNAi clone, <italic>nhr-80,</italic> that caused <named-content content-type="genus-species">C. elegans</named-content> to accumulate increased levels of 18:0. NHR-80 is a member of the NHR family of transcription factors in <named-content content-type="genus-species">C. elegans</named-content> [<xref rid="pgen-0020108-b026" ref-type="bibr">26</xref>]. To further examine this gene we obtained a deletion allele from the National BioResource Program for the Experimental Animal <italic>C. elegans,</italic> Japan. The <italic>nhr-80(tm1011)</italic> mutant carries a 446-bp deletion that eliminates approximately half of the nucleotides in the second exon and all of the third exon (<xref ref-type="fig" rid="pgen-0020108-g001">Figure 1</xref>). Like the <italic>nhr-80(RNAi)</italic> worms, these mutants also showed an accumulation of 18:0 and reduction of 18:1 &#x00394;9 (<xref ref-type="fig" rid="pgen-0020108-g002">Figure 2</xref>) indicating that <italic>nhr-80(tm1011)</italic> is likely to be a loss of function mutation. In the <italic>nhr-80</italic> mutants, 18:0 accounts for about 10.2 &#x000b1; 0.3% of the total fatty acids and 18:1 &#x00394;9 accounts for 2.2 &#x000b1; 0.1%, as compared with 6.8 &#x000b1; 0.2% and 3.2 &#x000b1; 0.1%, respectively, in the wild type. The difference between these fatty acids in the <italic>nhr-80</italic> mutants and wild-type animals is significant, with <italic>p</italic> &#x0003c; 0.01 for both fatty acids. The changes in fatty acid composition shown for the <italic>nhr-80</italic> mutants in <xref ref-type="fig" rid="pgen-0020108-g002">Figure 2</xref> are similar to those reported for the <italic>nhr-49</italic> mutants. In those mutants the ratio of 18:0 to 18:1 &#x00394;9 was 4.3 compared to a ratio of 1.9 in wild-type animals [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. In our analysis of the <italic>nhr-80</italic> mutants the 18:0 to 18:1 &#x00394;9 ratio was 4.6 compared to the wild-type ratio of 2.2. The <italic>nhr-80</italic> mutants are viable and fertile indicating this change in fatty acid composition, though significant, does not affect essential functions of the animal.</p><fig id="pgen-0020108-g001" position="float"><label>Figure 1</label><caption><title>Diagram of <italic>nhr-80, fat-5, fat-6,</italic> and <italic>fat-7</italic> Genes and Mutations</title><p>(A) <italic>nhr-80</italic> is composed of a zinc finger domain (green boxes) and a ligand-binding domain (light blue boxes)<italic>. nhr-80(tm1011)</italic> contains a 446-bp deletion (light grey bar).</p><p>(B) <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> all contain four trans-membrane domains (dark blue boxes) and three histidine boxes (red boxes)<italic>. fat-5(tm420)</italic> consists of a 779-bp deletion (light grey bar). <italic>fat-6(tm331)</italic> contains a 1,232-bp deletion (light grey bar), and a 428-bp insertion (purple bar). The <italic>fat-7</italic> alleles are point mutations with <italic>fat-7(wa36),</italic> creating a premature stop codon and <italic>fat-7(wa37)</italic> changing a conserved histidine into a tyrosine.</p></caption><graphic xlink:href="pgen.0020108.g001"/></fig><fig id="pgen-0020108-g002" position="float"><label>Figure 2</label><caption><title>Fatty Acid Composition of <italic>nhr-80</italic> </title><p>Relative abundance of selected fatty acid species expressed as percentage of total fatty acid as determined by gas chromatography analysis. The <italic>nhr-80</italic> mutants have significantly higher levels of 18:0 and lower levels of 16:0 and 18:1 &#x00394;9 than wild type. Error bars represent the standard error. *significant differences between wild type and <italic>nhr-80</italic> mutant, <italic>p</italic> &#x0003c; 0.01.</p></caption><graphic xlink:href="pgen.0020108.g002"/></fig><p>Although two NHR mutant lines, <italic>nhr-49</italic> and <italic>nhr-80,</italic> show increased 18:0 as compared with wild-type worms, not all NHR mutants cause these changes in fat metabolism [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. Both of these transcription factors are proposed to be derived from the same ancestral gene that also is the progenitor of the mammalian gene encoding hepatocyte nuclear factor 4 [<xref rid="pgen-0020108-b027" ref-type="bibr">27</xref>], which in mammals, binds to fatty acids as ligands and is a key activator of lipid and cholesterol metabolism genes [<xref rid="pgen-0020108-b028" ref-type="bibr">28</xref>].</p><p>In addition to the change in fatty acid composition, the <italic>nhr-49</italic> mutants display an increase in fat storage based on staining of whole worms with the lipophilic dye Nile red [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. In the <italic>nhr-80</italic> mutants, we observed no increase in Nile red staining as compared to wild type (unpublished data) indicating no increase in fat storage. To confirm this, we tested fat storage in the <italic>nhr-80</italic> mutants by measuring the percent triglycerides in the total lipids. In the <italic>nhr-80</italic> mutants triglycerides comprised 44 &#x000b1; 1% of the total lipids as compared to 45 &#x000b1; 1% in wild type signifying no increase in fat storage. Thus the increased 18:0 accumulation and increased 18:0 to 18:1 &#x00394;9 ratio does not cause increased triglyceride synthesis. However, the altered fatty acid profile of <italic>nhr-80</italic> mutants indicates a role for NHR-80 in the regulation of fatty acid metabolism in <italic>C. elegans.</italic> </p></sec><sec id="s2b"><title>NHR-80 Is Required for Normal Expression of &#x00394;9 Desaturases</title><p>As NHR-80 is a transcription factor expressed in the intestine [<xref rid="pgen-0020108-b026" ref-type="bibr">26</xref>], the major site of fat metabolism in <italic>C. elegans,</italic> the increased 18:0 accumulation in the <italic>nhr-80</italic> mutants may be due to a reduced expression of the &#x00394;9 desaturase genes. To test this we used quantitative RT-PCR (QPCR) to measure gene expression with primers designed to amplify <italic>fat-5, fat-6,</italic> and <italic>fat-7,</italic> along with the control genes <italic>tbb-2</italic> (&#x003b2;-tubulin) and <italic>ubc-2</italic> (ubiquitin-conjugating enzyme, E2). Relative expression of these genes was examined in wild-type and <italic>nhr-80</italic> mutant adult populations and we found that expression of all three &#x00394;9 desaturases was decreased in the <italic>nhr-80</italic> mutants relative to wild type for eight experimental replicates (<xref ref-type="fig" rid="pgen-0020108-g003">Figure 3</xref>). On average, <italic>fat-5</italic> and <italic>fat-6</italic> expression were reduced by 66% and 22% respectively, while <italic>fat-7</italic> expression was almost completely eliminated in the <italic>nhr-80</italic> mutants.</p><fig id="pgen-0020108-g003" position="float"><label>Figure 3</label><caption><title>Expression of the &#x00394;9 Desaturase Genes in <italic>nhr-80</italic> </title><p>(A) Gene expression by QPCR in the <italic>nhr-80</italic> mutant reveals a decrease in expression of the &#x00394;9 desaturase genes relative to wild type. Error bars represent standard error.</p><p>(B) Transformed lines expressing &#x00394;9 desaturase gene GFP fusions grown to adulthood on empty vector control bacteria or <italic>nhr-80(RNAi)</italic> bacteria. Exposure times for photographs were adjusted due to different GFP expression in the three genes, although the exposure time for the two treatments was kept the same for each genotype. The exposure time for the <italic>fat-5::GFP</italic> worms was 1/4 s, for the <italic>fat-6::GFP</italic> worms was 1/30 s, and for the <italic>fat-7::GFP</italic> worms was 1/8 s. After 4 d, there is a dramatic reduction in &#x00394;9 desaturase gene expression in the intestine for <italic>fat-5::GFP</italic> and <italic>fat-7::GFP</italic> lines grown on <italic>nhr-80(RNAi)</italic>.</p></caption><graphic xlink:href="pgen.0020108.g003"/></fig><p>To determine if the expression pattern of <italic>nhr-80</italic> overlapped with the expression pattern of the &#x00394;9 desaturases we created two green fluorescent protein (GFP)-fusion expressing lines for each of the &#x00394;9 desaturase genes. Like <italic>nhr-80,</italic> all three &#x00394;9 desaturase genes were expressed in the intestine in adult worms (<xref ref-type="fig" rid="pgen-0020108-g003">Figure 3</xref>B), and in all four larval stages (unpublished data). The <italic>fat-5 promoter</italic>::<italic>GFP</italic> expressing lines showed additional expression in the pharynx and tail cells after hatching and throughout the lifespan. The <italic>fat-6 whole gene</italic>::<italic>GFP</italic> expressing lines displayed additional expression in the hypodermis in all life stages. The overlapping intestinal expression for all three &#x00394;9 desaturase genes indicates possible functional redundancy. The potential role for <italic>fat-5</italic> in the pharynx and <italic>fat-6</italic> in the hypodermis remain to be determined; however, the constitutive expression of these genes in the intestine is consistent with a central role for &#x00394;9 desaturation in normal <named-content content-type="genus-species">C. elegans</named-content> function.</p><p>To confirm the regulation of the &#x00394;9 desaturases by NHR-80, lines expressing the GFP fusions were grown on <italic>nhr-80(RNAi)</italic> bacteria. Transformed adults were allowed to lay eggs on <italic>nhr-80(RNAi)</italic> and control bacteria. The adults were removed and about 20 of the progeny were examined for GFP expression after 4 d of growth. Representative samples are shown in <xref ref-type="fig" rid="pgen-0020108-g003">Figure 3</xref>B. Expression of <italic>fat-7 whole gene</italic>::<italic>GFP</italic> was completely eliminated by the RNAi treatment. Expression of <italic>fat-5 promoter</italic>::<italic>GFP</italic> was decreased but only in the intestine, not in the pharynx. Expression of <italic>fat-6 whole gene::GFP</italic> was also slightly decreased. The reduction of <italic>fat-5</italic> and <italic>fat-6</italic> expression and the elimination of <italic>fat-7</italic> expression likely accounts for the changes in fatty acid composition observed in the <italic>nhr-80</italic> mutant. Similar to the <italic>nhr-80</italic> mutants, the <italic>nhr-49</italic> mutants exhibited an increased level of 18:0 accumulation and a decrease in expression of the &#x00394;9 fatty acid desaturase genes by QPCR with <italic>fat-5</italic> and <italic>fat-7</italic> as the most reduced [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. However, the <italic>nhr-49</italic> mutants have increased fat storage, which is not seen in the <italic>nhr-80</italic> mutants, and show decreased expression of two genes that encode proteins that participate in the mitochondrial &#x003b2;-oxidation pathway, an enoyl-CoA hydratase gene (C29F3.1, <italic>ech-1</italic>) and an acyl-CoA synthetase gene (F28F8.2, <italic>acs-2</italic>). We tested the expression of <italic>ech-1</italic> and <italic>acs-2</italic> in the <italic>nhr-80</italic> mutants by QPCR and found that there was no change in expression levels relative to wild-type expression. This is consistent with the normal level of fat storage seen in the <italic>nhr-</italic>80 mutants. Though both NHR-49 and NHR-80 are required for &#x00394;9 desaturase expression, their effects on fatty acid metabolism in <named-content content-type="genus-species">C. elegans</named-content> are not identical; NHR-49 appears to regulate a wider range of lipid homeostasis pathways.</p></sec><sec id="s2c"><title> <italic>nhr-80</italic> Mutants Do Not Die Early like <italic>nhr-49</italic> Mutants</title><p>It has been suggested that shifts in the ratio of saturated fatty acids to MUFAs in <named-content content-type="genus-species">C. elegans</named-content> may lead to a decreased lifespan. For example, the change in the ratio of 18:0 to 18:1 &#x00394;9 from 1.9 in wild type to 4.3 in <italic>nhr-49</italic> mutants has been proposed to cause a substantial reduction in lifespan from 15&#x02013;18 d in wild type, to 6&#x02013;8 d in <italic>nhr-49</italic> mutants [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. We examined the lifespan of the <italic>nhr-80</italic> mutants (<xref ref-type="fig" rid="pgen-0020108-g004">Figure 4</xref>) and found that they may have slightly shorter lifespans than wild type but live considerably longer than <italic>nhr-49</italic> mutants despite having a similar fatty acid composition. In this experiment, the average lifespan of the <italic>nhr-80</italic> mutant was 12.5 &#x000b1; 0.5 d as compared to 13.9 &#x000b1; 0.4 d in wild-type animals and 8.2 &#x000b1; 0.2 d in <italic>nhr-49</italic> mutants when grown at 25 &#x000b0;C. These data indicate a 10% decrease in mean lifespan between wild type and <italic>nhr-80</italic> mutants, the difference between wild type and <italic>nhr-49</italic> mutants is much greater with a 41% reduction in mean lifespan. The early death of the <italic>nhr-49</italic> does not seem to be caused solely by an elimination of <italic>fat-7</italic> expression or an increase in the ratio of 18:0 to 18:1 &#x00394;9 since <italic>nhr-80</italic> mutants also show these characteristics but do not have a dramatically shortened lifespan. It is possible that the shorter lifespan of the <italic>nhr-49</italic> mutants is caused by metabolic changes due to other targets of NHR-49 regulation.</p><fig id="pgen-0020108-g004" position="float"><label>Figure 4</label><caption><title>Aging of Adult Mutant Populations</title><p>(A) Life span of <italic>nhr-49, nhr-80,</italic> and wild type at 25 &#x000b0;C. The <italic>nhr-80</italic> and wild type display a considerably longer life than the <italic>nhr-49</italic> mutants. All lifespan data are presented as mean lifespan &#x000b1; standard error (total number of animals scored). Wild type: 13.9 &#x000b1; 0.4 (83); <italic>nhr-49</italic>: 8.2 &#x000b1; 0.2 (80); <italic>nhr-80</italic>: 12.5 &#x000b1; 0.5 (70).</p><p>(B) Life span of <italic>fat-5, fat-6, fat-7,</italic> and wild type at 25 &#x000b0;C. The <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> mutants exhibit a lifespan similar to wild type. Wild type: 13.9 &#x000b1; 0.4 (83); <italic>fat-5</italic>: 15.9 &#x000b1; 0.6 (82); <italic>fat-6</italic>: 14.2 &#x000b1; 0.5 (82); <italic>fat-7</italic>: 15.0 &#x000b1; 0.5 (75).</p></caption><graphic xlink:href="pgen.0020108.g004"/></fig></sec><sec id="s2d"><title> <named-content content-type="genus-species">C. elegans</named-content> &#x00394;9 Desaturases Are Redundant under Standard Growth Conditions</title><p>Previous studies revealed that the three <named-content content-type="genus-species">C. elegans</named-content> &#x00394;9 desaturase isozymes display different substrate specificities. While FAT-6 and FAT-7 preferentially desaturate stearic acid (18:0), similar to most of the characterized SCDs, FAT-5 prefers palmitic acid (16:0) and has little or no activity on stearic acid [<xref rid="pgen-0020108-b019" ref-type="bibr">19</xref>]. We obtained &#x00394;9 desaturase single mutants to further characterize the roles of these three desaturases. We obtained <italic>fat-5(tm420)</italic> and <italic>fat-6(tm331)</italic> deletion alleles from the National BioResource Program for the Experimental Animal <italic>C. elegans,</italic> Japan (<xref ref-type="fig" rid="pgen-0020108-g001">Figure 1</xref>B). The <italic>fat-5</italic> allele has a 779-bp deletion early in the coding sequence that eliminates two of the conserved histidine boxes and two of the trans-membrane domains. The <italic>fat-6</italic> allele has a 1,232-bp deletion and a 428-bp insertion. The deletion is early in the coding sequence and also eliminates two of the conserved histidine-rich regions and two trans-membrane domains. Both of these mutations are likely null. The <italic>fat-7(tm326)</italic> deletion allele is available but molecular analysis of this allele led us to believe that a more extensive genetic disruption had occurred that affects other genes in addition to <italic>fat-7</italic>. Alternative <italic>fat-7</italic> alleles were isolated by TILLING (Targeting Induced Local Lesions IN Genomes) [<xref rid="pgen-0020108-b029" ref-type="bibr">29</xref>] and are single base pair changes (<xref ref-type="fig" rid="pgen-0020108-g001">Figure 1</xref>B). The <italic>fat-7(wa36)</italic> allele is a C to T mutation that leads to a premature stop codon that eliminates two trans-membrane domains and one of the conserved histidine boxes required for activity of the rat SCD enzyme [<xref rid="pgen-0020108-b030" ref-type="bibr">30</xref>], indicating that this allele is, at a minimum, a strong reduction-of-function allele. The <italic>fat-7(wa37)</italic> allele is a C to T mutation that replaces a conserved histidine with tyrosine [<xref rid="pgen-0020108-b019" ref-type="bibr">19</xref>]. Because these histidines are expected to be required for &#x00394;9 desaturase activity we expressed this allele in mutant yeast that lack &#x00394;9 desaturase activity (<italic>ole1</italic> mutants). The mutant <italic>fat-7(wa37)</italic> did not support growth of the <italic>ole1</italic> mutant yeast, whereas expression of wild-type <italic>fat-7</italic> did allow growth [<xref rid="pgen-0020108-b019" ref-type="bibr">19</xref>]. Phenotypic characterization including fatty acid composition and lifespan with <italic>fat-7(wa37)</italic> showed no difference from <italic>fat-7(wa36)</italic> therefore only data from <italic>fat-7(wa36)</italic> are reported here.</p><p>The <named-content content-type="genus-species">C. elegans</named-content> &#x00394;9 desaturase mutants show subtle differences from wild type in their fatty acid profile when grown on an <named-content content-type="genus-species">Escherichia coli</named-content> lawn on NGM plates at 20 &#x000b0;C (<xref ref-type="fig" rid="pgen-0020108-g005">Figure 5</xref>). Compared to wild type (4.1 &#x000b1; 0.2%), the <italic>fat-5</italic> mutants display decreased 16:1 &#x00394;9 (3.4 &#x000b1; 0.1%), which is the product of FAT-5 desaturation based on the substrate specificity exhibited in yeast [<xref rid="pgen-0020108-b019" ref-type="bibr">19</xref>]. The <italic>fat-6</italic> mutants exhibit a significant increase in their accumulation of the predicted substrate of FAT-6, 18:0 (9.6 &#x000b1; 0.2%), over wild type (7.0 &#x000b1; 0.2%).</p><fig id="pgen-0020108-g005" position="float"><label>Figure 5</label><caption><title>Fatty Acid Composition of the &#x00394;9 Desaturase Mutants</title><p>(A&#x02212;C) There is little change in fatty acid composition for <italic>fat-5</italic> (A), <italic>fat-6</italic> (B), and <italic>fat-7</italic> (C) mutants compared to wild-type worms when grown under standard growth conditions with OP50 <named-content content-type="genus-species">E. coli</named-content> as the sole food source.</p><p>(D&#x02212;F) Axenic growth conditions for wild-type worms and <italic>fat-5</italic> (D), <italic>fat-6</italic> (E), and <italic>fat-7</italic> (F) mutants reveal major changes in fatty acid composition for <italic>fat-5</italic> and <italic>fat-6</italic> mutants compared to wild-type worms. In all graphs relative abundance of selected fatty acid species is expressed as percentage of total fatty acid as determined by gas chromatography analysis. Error bars represent the standard error. *significant difference from wild type, <italic>p</italic> &#x0003c; 0.01</p></caption><graphic xlink:href="pgen.0020108.g005"/></fig><p>The &#x00394;9 desaturase mutants are indistinguishable from wild type in other characteristics tested including growth rate, reproduction, and behavior. The lack of phenotype indicates that subtle changes in fatty acid composition have no apparent effect and that the desaturases are functionally redundant. To determine if gene expression changes are involved in compensating for the lack of one isozyme, we examined expression of the &#x00394;9 desaturase genes in the <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> mutants (<xref ref-type="fig" rid="pgen-0020108-g006">Figure 6</xref>). In the <italic>fat-6</italic> mutants, <italic>fat-7</italic> expression is increased approximately 4-fold over wild type and <italic>fat-5</italic> expression is increased 2&#x02013;3-fold over wild type. In the <italic>fat-7</italic> mutant, expression of <italic>fat-6</italic> and <italic>fat-5</italic> is also slightly increased over wild type. The <italic>fat-5</italic> mutant shows little difference from wild type in <italic>fat-6</italic> and <italic>fat-7</italic> expression.</p><fig id="pgen-0020108-g006" position="float"><label>Figure 6</label><caption><title>Expression of the &#x00394;9 Desaturase Genes in the <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> Mutants</title><p>Gene expression by QPCR in the <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> mutants relative to wild type reveals an increase in &#x00394;9 desaturase gene expression in the <italic>fat-6</italic> and <italic>fat-7</italic> mutants, relative to wild type. Error bars represent standard error of 7&#x02013;12 experiments.</p></caption><graphic xlink:href="pgen.0020108.g006"/></fig></sec><sec id="s2e"><title>Axenic Growth Reveals Substrate Specificity of the &#x00394;9 Desaturases</title><p>The standard strain of <named-content content-type="genus-species">E. coli</named-content> on which <named-content content-type="genus-species">C. elegans</named-content> are maintained in the laboratory contains palmitic (16:0), palmitoleic (16:1 &#x00394;9), and vaccenic (18:1 &#x00394;11), but not oleic acid (18:1 &#x00394;9) or polyunsaturated fatty acids [<xref rid="pgen-0020108-b031" ref-type="bibr">31</xref>]. When worms eat these bacteria they incorporate the fatty acids in their lipids. To test the fatty acid composition of the &#x00394;9 desaturase mutants grown on a different food source we grew the <named-content content-type="genus-species">C. elegans</named-content> strains in axenic media devoid of bacteria. This liquid media provides amino acids, vitamins, growth factors, and heme [<xref rid="pgen-0020108-b032" ref-type="bibr">32</xref>]. Our measurements reveal that the axenic media contains palmitic, palmitoleic, oleic and linoleic acids, but no vaccenic acid (unpublished data). Wild-type worms grow considerably more slowly under the axenic growth conditions, and the fatty acid profile is also dramatically different. In axenic culture, wild-type worms accumulate higher levels of 16:0, 18:0, and 18:1 &#x00394;9, while they produce lesser amounts of 20:5 (<xref ref-type="fig" rid="pgen-0020108-g005">Figure 5</xref>A and <xref ref-type="fig" rid="pgen-0020108-g005">5</xref>D).</p><p>The &#x00394;9 desaturase mutants show greater differences in fatty acid composition when grown axenically than when grown on <named-content content-type="genus-species">E. coli</named-content> plates (<xref ref-type="fig" rid="pgen-0020108-g005">Figure 5</xref>D&#x02013;<xref ref-type="fig" rid="pgen-0020108-g005">5</xref>F). Comparing the fatty acid composition of the <italic>fat-5</italic> mutant with wild type we observe an increase in 16:0 (19 &#x000b1; 1% versus 12 &#x000b1; 1%) and a decrease in 16:1 &#x00394;9 (1.1 &#x000b1; 0.4% versus 3.0 &#x000b1; 0.3%) and 18:1 &#x00394;11 (3.8 &#x000b1; &#x02212;0.1% versus 17 &#x000b1; 1%) in the <italic>fat-5</italic> mutants. The <italic>fat-6</italic> mutants also display dramatic differences from wild type, with an increase in 18:0 (16.7 &#x000b1; 0.8% versus 10.9 &#x000b1; 0.7%) and a decrease in 18:1 &#x00394;9 (11.3 &#x000b1; 0.6% versus 21.8 &#x000b1; 0.5%) in the <italic>fat-6</italic> mutants. The fatty acid composition of <italic>fat-7</italic> mutants does not differ significantly from wild type, indicating that <italic>fat-6</italic> can completely compensate for <italic>fat-7</italic> in axenic culture and therefore that FAT-7 does not play an important role in maintaining proper fatty acid composition under axenic conditions. The dramatic reduction of 16:1 &#x00394;9 and 18:1 &#x00394;11 fatty acids in <italic>fat-5</italic> mutants and 18:1&#x00394;9 in <italic>fat-6</italic> mutants grown in axenic culture is the first evidence that these enzymes have the same substrate specificity in <named-content content-type="genus-species">C. elegans</named-content> as they do when expressed in yeast [<xref rid="pgen-0020108-b019" ref-type="bibr">19</xref>].</p><p>To determine whether the levels of &#x00394;9 desaturase gene expression are modulated in response to diet we examined the expression of <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> genes in axenic media and on <named-content content-type="genus-species">E. coli</named-content> seeded plates using QPCR. We found that compared to worms grown on <italic>E. coli, fat-5</italic> expression increases about 6-fold in axenic media. In contrast, <italic>fat-6</italic> expression is maintained at similar levels while <italic>fat-7</italic> expression is dramatically decreased in axenic media (<xref ref-type="supplementary-material" rid="pgen-0020108-sg001">Figure S1</xref>).</p></sec><sec id="s2f"><title>Single &#x00394;9 Desaturase Mutants Have No Early-Death Phenotype</title><p>Previous studies investigating the <named-content content-type="genus-species">C. elegans</named-content> &#x00394;9 desaturases have used RNAi to deplete <italic>fat-7</italic> expression and have suggested that <italic>fat-7</italic> expression is required to maintain a normal lifespan [<xref rid="pgen-0020108-b023" ref-type="bibr">23</xref>,<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. Based on these results, it was proposed that the reduced expression of <italic>fat-7</italic> was the cause of the short lifespan in the <italic>nhr-49</italic> mutants [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. However, the <italic>fat-7</italic>mutants used in our experiment as well as the other &#x00394;9 desaturase mutants, <italic>fat-5</italic> and <italic>fat-6,</italic> do not exhibit an early death phenotype (<xref ref-type="fig" rid="pgen-0020108-g004">Figure 4</xref>B). The average lifespan of the <italic>fat-5</italic> mutants is 15.8 &#x000b1; 0.6 d, the <italic>fat-6</italic> mutant is 14.2 &#x000b1; 0.5 d, and the <italic>fat-7</italic> mutant is 15.0 &#x000b1; 0.5 d, as compared with a lifespan of 13.9 &#x000b1; 0.4 d in wild-type animals. In this experiment, the <italic>fat-5</italic> mutant displayed a slight but significant (<italic>p</italic> &#x0003c; 0.01) increase in lifespan over wild type, while the <italic>fat-6</italic> and <italic>fat-7</italic> mutants were not significantly different from wild type in average lifespan.</p><p>Our experiments with the <italic>fat-7</italic> mutant do not support the requirement for <italic>fat-7</italic> for normal lifespan as proposed from studies using <italic>fat-7(RNAi)</italic> [<xref rid="pgen-0020108-b023" ref-type="bibr">23</xref>,<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. Additionally, <italic>fat-7(RNAi)</italic> revealed major changes in fatty acid composition and a reduction of fat storage [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>] that was not observed in the <italic>fat-7</italic> mutants. The RNAi phenotype observed could be due to transitive secondary RNAi effect [<xref rid="pgen-0020108-b033" ref-type="bibr">33</xref>] as <italic>fat-7</italic> has 84% nucleotide identity with <italic>fat-6</italic> including eight regions of 21&#x02013;44 nucleotides with 100% identity. Van Gilst et al. report that <italic>fat-7(RNAi)</italic> did not reduce <italic>fat-6</italic> expression when measured by QPCR [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]; however, we observe an elimination of <italic>fat-6</italic> expression when <italic>fat-6 whole gene::GFP</italic> lines were grown on <italic>fat-7(RNAi)</italic> (unpublished data). In addition, it is possible that compensation by the third &#x00394;9 desaturase, <italic>fat-5</italic>, is inhibited in the <italic>fat-7(RNAi).</italic> Because the <italic>fat-7</italic> loss-of-function mutant is wild type for fatty acid composition and lifespan it must be concluded that <italic>fat-7(RNAi)</italic> is having off-target effects on the worm.</p></sec><sec id="s2g"><title>&#x00394;9 Desaturase Activity and Monounsaturated Fatty Acids Are Required for Survival</title><p>Because the &#x00394;9 desaturase genes appear to compensate for each other, we constructed a <italic>fat-5;fat-6;fat-7</italic> triple mutant lacking all three &#x00394;9 desaturases. We expected these mutants would be unable to survive under standard growth conditions, so we supplemented the worms with a combination of 18:1 &#x00394;9, 18:2 &#x003c9;6, and 20:5 &#x003c9;3 dietary fatty acids. After identifying the <italic>fat-5;fat-6;fat-7</italic> triple mutant, we moved the worms to plates without fatty acid supplementation and found that indeed these worms could not survive. Larvae that hatch from eggs laid on unsupplemented plates arrest in the L1 stage, while L3 and L4 stage larvae that are moved from supplemented to unsupplemented plates develop into thin, sterile adults with reduced movement and early death. The MUFAs provided by the standard <named-content content-type="genus-species">E. coli</named-content> diet are not sufficient for survival in the <italic>fat-5;fat-6;fat-7</italic> triple mutant. Thus <named-content content-type="genus-species">C. elegans</named-content> have a requirement for a certain level of &#x00394;9 desaturation that cannot be met by the standard <named-content content-type="genus-species">E. coli</named-content> diet. The yeast &#x00394;9 desaturase mutant, <italic>ole1,</italic> is also unable to grow without supplementation [<xref rid="pgen-0020108-b004" ref-type="bibr">4</xref>]. The <italic>fat-5;fat-6;fat-7</italic> triple mutant is the first multicellular organism generated that lacks all endogenous &#x00394;9 desaturase activity.</p><p>To examine genetic interaction between <italic>nhr-80</italic> and <italic>fat-6,</italic> the most highly expressed &#x00394;9 desaturase, we constructed the <italic>fat-6;nhr-80</italic> double mutant using plates supplemented with dietary fatty acids. When we removed the <italic>fat-6;nhr-80</italic> double mutants to unsupplemented plates we found that these worms also did not survive. Since the <italic>nhr-80(RNAi)</italic> phenotype resembles the <italic>nhr-80</italic> mutants, we used RNAi in combination with the &#x00394;9 desaturase mutants to study this interaction further. The <italic>fat-6</italic> mutants, when grown on <italic>nhr-80(RNAi)</italic> from eggs, become thin, slow growing, and reproductively inviable after 4 d of growth (<xref ref-type="fig" rid="pgen-0020108-g007">Figure 7</xref>). They also accumulate very high levels of 18:0 (<xref ref-type="fig" rid="pgen-0020108-g007">Figure 7</xref>B). The <italic>fat-6</italic> mutants grown on <italic>nhr-80(RNAi)</italic> accumulate 34 &#x000b1; 1% of their fatty acids as 18:0 as compared to 9.1 &#x000b1; 0.1% when <italic>fat-6</italic> is grown on control bacteria or 14 &#x000b1; 2% when wild-type worms are grown on <italic>nhr-80(RNAi)</italic> bacteria. Although 18:0 also accumulates in the <italic>fat-5</italic> and <italic>fat-7</italic> mutants grown on <italic>nhr-80(RNAi),</italic> the extent of 18:0 accumulation is not as dramatic as observed in <italic>fat-6</italic> (<xref ref-type="fig" rid="pgen-0020108-g007">Figure 7</xref>B) and they do not show a synthetic lethality (<xref ref-type="fig" rid="pgen-0020108-g007">Figure 7</xref>A).</p><fig id="pgen-0020108-g007" position="float"><label>Figure 7</label><caption><title>Effects of <italic>nhr-80(RNAi)</italic> in the &#x00394;9 Desaturase Mutant Background</title><p>(A) Photographs showing adult worms after 4 d of growth on <italic>nhr-80(RNAi)</italic> and empty vector control bacteria. The <italic>fat-6</italic> mutants grown on <italic>nhr-80(RNAi)</italic> are thin, pale, and produce no viable progeny.</p><p>(B) Relative abundance of 18:0 expressed as a percentage of total fatty acid as determined by gas chromatography analysis. The <italic>fat-6</italic> mutants grown on <italic>nhr-80(RNAi)</italic> (<italic>n</italic> = 5) accumulate much higher levels than <italic>fat-6</italic> mutants grown on control (<italic>n</italic> = 7) and wild type grown on <italic>nhr-80(RNAi)</italic> (<italic>n</italic> = 6). *significant differences from growth on control bacteria, <italic>p</italic> &#x0003c; 0.01.</p><p>(C) Effects of <italic>nhr-80</italic> on &#x00394;9 desaturase gene expression in <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> mutants. QPCR in <italic>fat-5, fat-6, fat-7,</italic> and wild type for worms grown on empty vector control bacteria (ev) and <italic>nhr-80(RNAi)</italic> (80i) (<italic>n</italic> = 6)<italic>.</italic> Values are expressed relative to <italic>fat-6</italic> expression in wild-type worms grown on control bacteria. For all graphs error bars represent standard error.</p></caption><graphic xlink:href="pgen.0020108.g007"/></fig><p>One explanation for the synthetic lethality of <italic>fat-6;nhr-80</italic> double mutants is that NHR-80 is required for the increased <italic>fat-5</italic> and <italic>fat-7</italic> expression in the <italic>fat-6</italic> mutant. To test this we examined the expression of the &#x00394;9 desaturase genes in the <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> mutants grown on <italic>nhr-80(RNAi).</italic> We found that expression of <italic>fat-7</italic> in the <italic>fat-6;nhr-80(RNAi)</italic> is less than 10% of the expression of <italic>fat-7</italic> in the <italic>fat-6</italic> mutants grown on control bacteria, consistent with the notion that NHR-80 is required to induce the expression of <italic>fat-7</italic> (<xref ref-type="fig" rid="pgen-0020108-g007">Figure 7</xref>C). We graphed the relative expression values, setting <italic>fat-6</italic> expression in wild-type worms grown on control bacteria as 100%. In wild-type worms on control bacteria <italic>fat-6</italic> is the most highly expressed &#x00394;9 desaturase gene and <italic>fat-5</italic> and <italic>fat-7</italic> are expressed at 3.6 &#x000b1; 0.2% and 6.5 &#x000b1; 0.6% of the level of <italic>fat-6</italic> respectively<italic>.</italic> When the wild-type worms are grown on <italic>nhr-80(RNAi)</italic> we observe a similar relative decrease in &#x00394;9 desaturase gene expression seen in the <italic>nhr-80</italic> mutants (<xref ref-type="fig" rid="pgen-0020108-g003">Figure 3</xref>A). Comparing the <italic>fat-5</italic> and <italic>fat-7</italic> mutants grown on control with those grown on <italic>nhr-80(RNAi)</italic> reveals a decrease in &#x00394;9 desaturase gene expression. However, the biggest difference is seen in the <italic>fat-6</italic> mutants. When these animals are grown on control bacteria <italic>fat-7</italic> is increased in expression 37-fold over wild type. When <italic>fat-6</italic> is grown on <italic>nhr-80(RNAi)</italic> the <italic>fat-7</italic> relative expression is only a 3-fold increase over wild type.</p><p>The overall amount of &#x00394;9 desaturase gene expression is approximately equal for all worms grown on <italic>nhr-80(RNAi),</italic> but only <italic>fat-6</italic> displays synthetic lethality with <italic>nhr-80.</italic> This could be due to the composition of the &#x00394;9 desaturase gene expression. When wild type, <italic>fat-5,</italic> or <italic>fat-7</italic> are grown on <italic>nhr-80(RNAi) fat-6</italic> is the major gene expressed suggesting its central importance for &#x00394;9 desaturation activity. The <italic>fat-6</italic> mutants lack <italic>fat-6</italic> expression and compensate by substantially increasing <italic>fat-7</italic> expression when grown on control bacteria. It is noteworthy that under these conditions <italic>fat-7</italic> expression is increased 37-fold, perhaps indicating that <italic>fat-7</italic> is not as effective at &#x00394;9 desaturation as <italic>fat-6</italic> due to differences in tissue specific expression, translation efficiency or protein stability. When the <italic>fat-6</italic> mutants are grown on <italic>nhr-80(RNAi)</italic> they are unable to compensate with an increase in <italic>fat-7</italic> expression to an appropriate level and this may cause their reduced survival. Thus NHR-80 is required for increasing <italic>fat-7</italic> expression in situations where higher <italic>fat-7</italic> levels are necessary and consequently defines a critical regulator of fatty acid metabolism.</p><p>Our characterization of the novel NHR-80 and the family of <named-content content-type="genus-species">C. elegans</named-content> &#x00394;9 desaturase mutants enhances our understanding of the regulation of lipid homeostasis. Maintaining appropriate fatty acid composition is essential and without sufficient &#x00394;9 desaturase activity both the <italic>fat-5;fat-6;fat-7</italic> triple mutants and the <italic>fat-6;nhr-80</italic> double mutants are unable to survive. The integration of endogenous and environmental signals by NHRs such as NHR-80 precisely regulates the expression of the &#x00394;9 desaturase genes and the production of monounsaturated fatty acids leads to optimal membrane fluidity and fat storage.</p></sec></sec><sec id="s3"><title>Materials and Methods</title><sec id="s3a"><title>Culture of nematodes.</title><p>Unless otherwise noted, <named-content content-type="genus-species">C. elegans</named-content> were grown on nematode growth media (NGM) plates with OP50 strain of <named-content content-type="genus-species">E. coli</named-content> as a food source [<xref rid="pgen-0020108-b034" ref-type="bibr">34</xref>]. The wild-type strain used is strain N2. Mutant strains obtained from Shohei Mitani and Edwin Cuppen were outcrossed at least four times to the N2 strain. The <italic>nhr-80(RNAi)</italic> construct, as well as the others used in the screen of Chromosome III, are from the Ahringer RNAi library [<xref rid="pgen-0020108-b035" ref-type="bibr">35</xref>] and were used as described [<xref rid="pgen-0020108-b036" ref-type="bibr">36</xref>]. As a control for RNAi experiments, nematodes were grown on NGM plates with the HT115 strain of <named-content content-type="genus-species">E. coli</named-content> transformed with pPD129.36 (L4440) empty vector plasmid. The axenic culture media consisted of 3% soy peptone, 3% yeast extract, 0.5 mg/ml hemoglobin in 1M KOH, and 20% ultra-high temperature pasteurized skim milk [<xref rid="pgen-0020108-b032" ref-type="bibr">32</xref>]. Worms were grown in this liquid culture at room temperature (22&#x02013;23 &#x000b0;C) with constant shaking. To make plates supplemented with dietary fatty acids a 0.1 M stock solution of fatty acid sodium salts (NuCheck Prep, Elysian, Minnesota, United States) in water was prepared fresh for each supplementation experiment. The fatty acid stock was added slowly to NGM containing 0.1% tergitol. Plates were poured, covered and allowed to dry in the dark at room temperature overnight. The OP50 strain of <named-content content-type="genus-species">E. coli</named-content> was added to each plate and allowed to dry for at least one night [<xref rid="pgen-0020108-b021" ref-type="bibr">21</xref>].</p></sec><sec id="s3b"><title>Fatty acid and lipid analysis.</title><p>For fatty acid analysis, adult nematodes were washed from plates and allowed to settle. The excess water was removed from the worm pellet and 1 ml of 2.5% methanolic H₂SO₄ was added and incubated at 80 &#x000b0;C for 1 h to generate fatty acid methyl esters, which were extracted by adding 1.5 ml water and 0.2 ml hexane. The hexane was sampled for determination of fatty acid composition by gas chromatography on an SP-2380 fused silica capillary column (Supelco, Bellefonte, Pennsylvania, United States) using an Agilent (Palo Alto, California, United States) 6890 series gas chromatograph [<xref rid="pgen-0020108-b018" ref-type="bibr">18</xref>].</p><p>For lipid analysis, about 0.5 ml of adult nematodes were collected in a glass tube and frozen. Lipids were extracted by incubation in (1:1) chloroform/methanol overnight at &#x02212;20 &#x000b0;C. The samples were washed with 2.2 ml Hajra's solution (0.2M H₃PO₄, 1M KCl) and the chloroform phase containing the lipids was isolated. The silica gel HL plates (Analtech, Newark, Delaware, United States) were activated by incubation at 110 &#x000b0;C for 1 h and 15 min. The samples were loaded onto the thin layer chromatography plates along with lipid standards (Sigma, St. Louis, Missouri, United States). The plates were run with a 65:43:3:2.5 chloroform/methanol/water/acetic acid solvent mixture until the solvent front was three-fourths of the way up the plate. The plate was dried, a new solvent mixture of 80:20:2 hexane/diethyl ether/acetic acid was added, and the plate was run until the solvent front reached the top of the plate. The marker lanes were visualized using iodine vapor and the corresponding bands for triglycerides and individual phospholipids in the silica gel were scraped into individual tubes. To quantitate, 50 &#x003bc;g of 15:0 free fatty acid was added to each tube as an internal standard and fatty acid analysis was performed by gas chromatography as described above [<xref rid="pgen-0020108-b022" ref-type="bibr">22</xref>].</p></sec><sec id="s3c"><title>QPCR analysis.</title><p>Adult nematodes were harvested and frozen in liquid nitrogen. RNA was prepared using TRIzol Reagent (Invitrogen, Carlsbad, California, United States). A DNA-FREE RNA kit (Zymo Research, Orange, California, United States) was used for Dnase treatment and purification. After quantification, 1 &#x003bc;g of RNA was used in a reverse-transcription reaction with SuperScriptIII (Invitrogen) to generate cDNA. Primer sequences for the &#x00394;9 desaturase genes and the reference genes were designed using PrimerQuest software at <ext-link ext-link-type="uri" xlink:href="http://www.idtdna.com">http://www.idtdna.com</ext-link>. Other primer sequences were obtained from Dr. Marc Van Gilst [<xref rid="pgen-0020108-b024" ref-type="bibr">24</xref>]. Primer sequences are listed in <xref ref-type="supplementary-material" rid="pgen-0020108-st001">Table S1</xref>. The PCR mixture consisted of 0.3 &#x003bc;M primers, cDNA, ROX, and 1&#x000d7; SYBR green mix (Invitrogen Platinum SYBR green qPCR Supermix UDG). The QPCR was run and monitored on a MX3000P (Stratagene, La Jolla, California, United States). Relative abundance was determined using the &#x00394;&#x00394;Ct method and an average of the expression of the reference genes <italic>tbb-2</italic> and <italic>ubc-2</italic> to control for template levels [<xref rid="pgen-0020108-b037" ref-type="bibr">37</xref>].</p></sec><sec id="s3d"><title>Construction of GFP fusions and microinjection.</title><p>Fusion PCR was used to create translational <italic>fat-5, fat-6,</italic> and <italic>fat-7</italic> GFP constructs. The promoters and coding sequences of <italic>fat-6</italic> and <italic>fat-7</italic> and the promoter and first exon of <italic>fat-5</italic> were amplified from genomic DNA. The upstream regulatory region for <italic>fat-5</italic> was 4 kb, for <italic>fat-</italic>6 was 2.6 kb, and for <italic>fat-7</italic> was 3.0 kb. GFP was amplified from the Fire vector pPD95.75 including the entire coding sequence and a termination sequence. These PCR products were fused together in a final PCR using nested primers [<xref rid="pgen-0020108-b038" ref-type="bibr">38</xref>]. These fusions were microinjected into <italic>lin-15</italic> mutant <named-content content-type="genus-species">C. elegans</named-content> along with a rescuing plasmid, pJM23, containing the wild-type <italic>lin-15</italic> gene [<xref rid="pgen-0020108-b039" ref-type="bibr">39</xref>,<xref rid="pgen-0020108-b040" ref-type="bibr">40</xref>]. Multiple independent lines of nematodes without the <italic>lin-15</italic> phenotype were selected and examined for GFP expression using fluorescence microscopy on an Olympus IX70 microscope.</p></sec><sec id="s3e"><title>Lifespan analysis.</title><p>Aging experiments were performed on adult nematodes grown at 25 &#x000b0;C. Worms were moved to plates containing 5-fluoro-2&#x02032;-deoxyuridine (Sigma) at the fourth larval stage of development (L4). Live animals were assayed for movement in response to touch every 1&#x02013;2 d [<xref rid="pgen-0020108-b041" ref-type="bibr">41</xref>].</p></sec><sec id="s3f"><title>Generation of <italic>fat-5;fat-6;fat-7</italic> triple mutants and <italic>fat-6;nhr-80</italic> double mutants.</title><p>The <italic>fat-6(tm331);fat-7(wa36)</italic> hermaphrodites were crossed with <italic>fat-5(tm420);fat-7(wa36)</italic> males on plates supplemented with 18:1 &#x00394;9. The F1 generation was moved to new 18:1 &#x00394;9 supplemented plates and their progeny were moved to plates supplemented with a combination of 18:1 &#x00394;9, 18:2 &#x003c9;6, and 20:5 &#x003c9;3. After the F2 generation reproduced, the adults were harvested for single worm PCR to determine the genotype [<xref rid="pgen-0020108-b042" ref-type="bibr">42</xref>]. The <italic>fat-5</italic> and <italic>fat-6</italic> mutations were monitored using the difference in amplicon size between wild-type and mutant alleles due to the large deletions. The wild-type products were 1,100 bp for <italic>fat-5</italic> and 1,457 bp, for <italic>fat-6</italic> compared with the mutant products of 321 bp and 652 bp, respectively. All cross-progeny were homozygous for the <italic>fat-7</italic> single base pair mutation.</p><p>To generate <italic>nhr-80;fat-6</italic> double mutants we crossed <italic>fat-6</italic> males with <italic>nhr-80</italic> hermaphrodites on 18:1 &#x00394;9 supplemented plates and isolated the F1 generation onto new supplemented plates. The F2s were moved to fresh 18:1 &#x00394;9 supplemented plates and allowed to reproduce then single worm PCR was used to identify <italic>nhr-80;fat-6</italic> double mutants. The <italic>nhr-80</italic> wild-type allele generated a PCR product of 745 bp, whereas the <italic>nhr-80(tm1011)</italic> mutant allele generated a product 298 bp in length.</p></sec></sec><sec sec-type="supplementary-material" id="s4"><title>Supporting Information</title><supplementary-material content-type="local-data" id="pgen-0020108-sg001"><label>Figure S1</label><caption><title>Expression of &#x00394;9 Desaturase Genes in Wild-Type Worms Grown on <named-content content-type="genus-species">E. coli</named-content> (OP50) Seeded Plates or Axenic Liquid Media</title><p>The percent expression shown is relative to <italic>fat-6</italic> expression on <named-content content-type="genus-species">E. coli</named-content> plates, which is set at 100%. In wild-type worms grown in axenic culture the expression of <italic>fat-5</italic> is increased and the <italic>fat-7</italic> expression is nearly eliminated relative to expression in wild-type worms grown on <named-content content-type="genus-species">E. coli</named-content> (OP50) plates. Relative to <italic>fat-6</italic> expression, <italic>fat-5</italic> and <italic>fat-7</italic> expression is higher in wild-type worms grown on <named-content content-type="genus-species">E. coli</named-content> (OP50) compared to wild-type worms grown on <named-content content-type="genus-species">E. coli</named-content> (HT115) (<xref ref-type="fig" rid="pgen-0020108-g007">Figure 7</xref>C). Error bars are SEM, <italic>n</italic> = 3 replicates for plate grown and <italic>n</italic> = 6 replicates for axenic cultured nematodes.</p><p>(56 KB TIF)</p></caption><media xlink:href="pgen.0020108.sg001.tif"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020108-st001"><label>Table S1</label><caption><title>Sequence of DNA Primers Used in These Studies</title><p>(34 KB DOC)</p></caption><media xlink:href="pgen.0020108.st001.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><sec id="s4a"><title>Accession Numbers</title><p>The GenBank (<ext-link ext-link-type="uri" xlink:href="http://www.ncbi.nlm.nih.gov/Genbank">http://www.ncbi.nlm.nih.gov/Genbank</ext-link>) accession numbers for genes used in this study are <italic>nhr-80</italic> (H10E21.3) (AY204179), <italic>fat-5</italic> (W06D12.3) (AF260242), <italic>fat-6</italic> (VZK822L.1) (AF260244), and <italic>fat-7</italic> (F10D2.9) (AF260243).</p></sec></sec>

Neofunctionalization in Vertebrates: The Example of Retinoic Acid Receptors

<p>Understanding the role of gene duplications in establishing vertebrate innovations is one of the main challenges of Evo-Devo (evolution of development) studies. Data on evolutionary changes in gene expression (i.e., evolution of transcription factor-<italic>cis</italic>-regulatory elements relationships) tell only part of the story; protein function, best studied by biochemical and functional assays, can also change. In this study, we have investigated how gene duplication has affected both the expression and the ligand-binding specificity of retinoic acid receptors (RARs), which play a major role in chordate embryonic development. Mammals have three paralogous <italic>RAR</italic> genes&#x02014;<italic>RAR&#x003b1;, &#x003b2;,</italic> and <italic>&#x003b3;</italic>&#x02014;which resulted from genome duplications at the origin of vertebrates. By using pharmacological ligands selective for specific paralogues, we have studied the ligand-binding capacities of RARs from diverse chordates species. We have found that RAR&#x003b2;-like binding selectivity is a synapomorphy of all chordate RARs, including a reconstructed synthetic RAR representing the receptor present in the ancestor of chordates. Moreover, comparison of expression patterns of the cephalochordate amphioxus and the vertebrates suggests that, of all the RARs, RAR&#x003b2; expression has remained most similar to that of the ancestral RAR. On the basis of these results together, we suggest that while RAR&#x003b2; kept the ancestral RAR role, RAR&#x003b1; and RAR&#x003b3; diverged both in ligand-binding capacity and in expression patterns. We thus suggest that neofunctionalization occurred at both the expression and the functional levels to shape RAR roles during development in vertebrates.</p>

<contrib contrib-type="author"><name><surname>Escriva</surname><given-names>Hector</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="author-notes" rid="n105">&#x000a4;a</xref></contrib><contrib contrib-type="author" equal-contrib="yes"><name><surname>Bertrand</surname><given-names>St&#x000e9;phanie</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author" equal-contrib="yes"><name><surname>Germain</surname><given-names>Pierre</given-names></name><xref ref-type="aff" rid="aff2">2</xref></contrib><contrib contrib-type="author"><name><surname>Robinson-Rechavi</surname><given-names>Marc</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="author-notes" rid="n106">&#x000a4;b</xref></contrib><contrib contrib-type="author"><name><surname>Umbhauer</surname><given-names>Muriel</given-names></name><xref ref-type="aff" rid="aff3">3</xref></contrib><contrib contrib-type="author"><name><surname>Cartry</surname><given-names>J&#x000e9;r&#x000f4;me</given-names></name><xref ref-type="aff" rid="aff3">3</xref></contrib><contrib contrib-type="author"><name><surname>Duffraisse</surname><given-names>Marilyne</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Holland</surname><given-names>Linda</given-names></name><xref ref-type="aff" rid="aff4">4</xref></contrib><contrib contrib-type="author"><name><surname>Gronemeyer</surname><given-names>Hinrich</given-names></name><xref ref-type="aff" rid="aff2">2</xref></contrib><contrib contrib-type="author"><name><surname>Laudet</surname><given-names>Vincent</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib>

PLoS Genetics

<sec id="s1"><title>Introduction</title><p>The origin of organismal complexity is generally thought to be tightly linked to the evolution of new gene functions. Susumu Ohno proposed in 1970 that, in contrast to mutations, gene duplications can create evolutionary novelties [<xref rid="pgen-0020102-b001" ref-type="bibr">1</xref>]. He also proposed, based on the genome weight of different deuterostomes, that two periods of genome duplication occurred during evolution of the vertebrate lineage [<xref rid="pgen-0020102-b001" ref-type="bibr">1</xref>]. This hypothesis has been revisited and discussed by different authors, and even if the precise timing and mechanisms of these gene duplications are still under discussion, their general importance during vertebrate evolution is now widely accepted [<xref rid="pgen-0020102-b002" ref-type="bibr">2</xref>&#x02013;<xref rid="pgen-0020102-b004" ref-type="bibr">4</xref>].</p><p>Cephalochordates (specifically the small marine animals called amphioxus) have been generally considered the closest extant invertebrates to vertebrates. Although recent studies place urochordates as the sister group of vertebrates [<xref rid="pgen-0020102-b005" ref-type="bibr">5</xref>,<xref rid="pgen-0020102-b006" ref-type="bibr">6</xref>], it remains accepted that amphioxus diverged from the vertebrate lineage before the vertebrate genome duplications occurred (<xref ref-type="fig" rid="pgen-0020102-g001">Figure 1</xref>A). In general, for each gene paralogy group in vertebrates, amphioxus contains a single copy of its respective orthologue&#x02014;amphioxus contains a single <italic>Hox</italic>-cluster [<xref rid="pgen-0020102-b007" ref-type="bibr">7</xref>] instead of four in mammals, and a single retinoic acid receptor (RAR), AmphiRAR [<xref rid="pgen-0020102-b008" ref-type="bibr">8</xref>], instead of three as in mammals (RAR&#x003b1;, RAR&#x003b2; and RAR&#x003b3;). Many data suggest that the duplications took place at two distinct periods during evolution, one before the split of agnathans (hagfish and lampreys) and one before the split of cartilaginous fishes [<xref rid="pgen-0020102-b009" ref-type="bibr">9</xref>&#x02013;<xref rid="pgen-0020102-b011" ref-type="bibr">11</xref>]. The lamprey genome has probably experienced only one of these large-scale gene duplications, although some independent duplications also occurred in the lamprey <italic>Hox</italic>-cluster [<xref rid="pgen-0020102-b012" ref-type="bibr">12</xref>].</p><fig id="pgen-0020102-g001" position="float"><label>Figure 1</label><caption><title>Phylogenetic View of Deuterostomes and RARs</title><p>(A) Current view of deuterostome phylogeny with amphioxus representing the basal chordate [<xref rid="pgen-0020102-b005" ref-type="bibr">5</xref>]. RARs used in the present study are indicated at their respective taxonomic positions&#x02014;for mouse, <italic>Xenopus,</italic> zebrafish, lamprey, amphioxus, and tunicates. The position of the synthetic ancestral sequence is indicated by a red circle. The two proposed periods of whole genome duplications in vertebrates are indicated as Phase I and Phase II, occurring respectively before and after the divergence of lampreys.</p><p>(B) Phylogenetic tree showing the placement of the RARs used in this study. Branch length is proportional to evolutionary change (bar = 0.1 substitutions per site); numbers at nodes are bootstrap support, in percent of 1,000 replicates. Branches supported by bootstrap lower than 70% have been polytomised. The tree was rooted by the amphioxus sequence, in agreement with [<xref rid="pgen-0020102-b005" ref-type="bibr">5</xref>]. Species abbreviations and their groups are indicated as follows. Amphioxus: Amphi, <italic>Branchiostoma floridae.</italic> Tunicates: Pm, <italic>Polyandrocarpa misakiensis;</italic> Ci, <italic>Ciona intestinalis.</italic> Lampreys: Lamp, <named-content content-type="genus-species">Petromyzon marinus</named-content>. Teleost fish: Takifugu, <italic>Takifugu rubripes;</italic> Tetraodon, <italic>Tetraodon nigroviridis;</italic> and Danio, <named-content content-type="genus-species">Danio rerio</named-content>. Amphibians: Xenopus, <italic>Xenopus laevis;</italic> Ambystoma, <italic>Ambystoma mexicanum;</italic> and Notophthalmus, <italic>Notophthalmus viridescens.</italic> Birds: Gallus, <italic>Gallus gallus;</italic> and Coturnix, <italic>Coturnix coturnix.</italic> Mammals: Homo, <italic>Homo sapiens;</italic> Mus, <italic>Mus musculus;</italic> and Rattus, <named-content content-type="genus-species">Rattus norvegicus</named-content>.</p></caption><graphic xlink:href="pgen.0020102.g001"/></fig><p>The contribution of duplicated genes to the origin of evolutionary novelties has been formalized by the &#x0201c;duplication-degeneration-complementation&#x0201d; model [<xref rid="pgen-0020102-b013" ref-type="bibr">13</xref>]. This model establishes three possible fates for duplicate genes: (i) one member of the duplicated pair degenerates by accumulating deleterious mutations, while the other retains the original gene function; (ii) the ancestral function is partitioned and shared by the two members of the duplicated pair (subfunctionalization); or (iii) one duplicate acquires a new function while the other retains the original function (neofunctionalization).</p><p>Two paths for the generation of evolutionary novelties have been proposed: (i) changes in the noncoding moiety of the gene (i.e., evolution of <italic>cis</italic>-regulatory elements) and (ii) changes in the coding moiety of the gene (i.e., evolution of protein function). Changes in transcriptional regulation of the genes can underlie the evolution of body plan diversity. Thus, spatial and temporal changes in gene expression of orthologous <italic>Hox</italic> genes in different vertebrates are correlated with morphological innovations. For example, a change in the expression domains of <italic>Hox</italic> genes correlates with anatomical differences among vertebrae in tetrapods [<xref rid="pgen-0020102-b014" ref-type="bibr">14</xref>]. Similarly, the three mammalian RARs have overlapping but somewhat different expression domains and their functions are not entirely redundant [<xref rid="pgen-0020102-b015" ref-type="bibr">15</xref>,<xref rid="pgen-0020102-b016" ref-type="bibr">16</xref>]. Sequence changes in proteins and consequent alterations in their biochemical functions could also underlie the diversification of body patterns. For example, changes in DNA-binding specificity of a transcription factor, its interactions with cofactors, or the posttranslational regulation of its activity could evolve in concert with more complex developmental roles (reviewed in [<xref rid="pgen-0020102-b017" ref-type="bibr">17</xref>]). Unfortunately, given the experimental limitations in characterizing the protein functions of developmental genes, little evidence to date supports the functional diversification of relevant genes during the chordate-to-vertebrate transition. RARs are particularly well suited for such a goal, since their developmental function is well studied and it is possible to characterize their DNA- and ligand-binding properties as well as their transcriptional and dimerisation activities [<xref rid="pgen-0020102-b018" ref-type="bibr">18</xref>]. Thus, RARs combine the potential for both classical Evo-Devo (evolution of development) and &#x0201c;Evo-Fun&#x0201d; (evolution of function) studies.</p><p>To decipher how gene duplications affected the ligand binding function, in the present work we studied RARs of several chordates as well as a reconstructed RAR representing the hypothetical sequence present in the ancestor of all vertebrates (AncRAR, <xref ref-type="fig" rid="pgen-0020102-g001">Figure 1</xref>A; <xref ref-type="table" rid="pgen-0020102-t001">Table 1</xref>). The ligand-binding domain (LBD) of nuclear receptors, including RARs, is structured in a three-layered &#x003b1;-helical antiparallel sandwich of 12 helices (H1&#x02013;H12), forming a hydrophobic ligand-binding pocket (LBP, <xref ref-type="fig" rid="pgen-0020102-g002">Figure 2</xref>). In RARs, this LBD is composed of about 270 amino-acids, with about 25 localized in H1, H3, H5, the &#x003b2;-turn, loop 6&#x02013;7, H11, loop 11&#x02013;12, and H12, which all make direct contact with the ligand (<xref ref-type="fig" rid="pgen-0020102-g002">Figure 2</xref>). To date, only one in vivo ligand of all vertebrate RARs is known&#x02014;all-<italic>trans</italic> retinoic acid (ATRA)&#x02014;and genetic evidence in mice has suggested that retinoic acid (RA) metabolites do not play a significant developmental role [<xref rid="pgen-0020102-b019" ref-type="bibr">19</xref>]. However, the LBPs of human RARs differ from each other in three amino acid positions [<xref rid="pgen-0020102-b020" ref-type="bibr">20</xref>,<xref rid="pgen-0020102-b021" ref-type="bibr">21</xref>], which cause different binding specificities in vitro, with differential binding and transactivation of each paralogue induced by different synthetic retinoids [<xref rid="pgen-0020102-b022" ref-type="bibr">22</xref>]. It is not known whether this difference in specificity has a role in vivo. Since the ligand-binding selectivity and the LBP structure are directly correlated [<xref rid="pgen-0020102-b023" ref-type="bibr">23</xref>], we used these synthetic retinoids as markers of the LBP structure in RARs. Thus, the comparison of the ligand-binding abilities of different RARs from organisms at key phylogenetic positions and of RARs with mutated LBPs provides information about the evolution of LBP structure and function.</p><table-wrap id="pgen-0020102-t001" content-type="1col" position="float"><label>Table 1</label><caption><p>cDNA Sequences of RARs Used in the Present Study</p></caption><graphic xlink:href="pgen.0020102.t001"/></table-wrap><fig id="pgen-0020102-g002" position="float"><label>Figure 2</label><caption><title>Protein Sequence Alignment of Selected Gnathostome RARs</title><p>RARs are represented from lamprey (LampRAR, <named-content content-type="genus-species">Petromyzon marinus</named-content>), amphioxus (AmphiRAR, <named-content content-type="genus-species">Branchiostoma floridae</named-content>), tunicate (RAR_POLM1, <named-content content-type="genus-species">Polyandrocarpa misakiensis</named-content>), and the synthetic predicted ancestral RAR (Ancestor). The position of the 12 helices is indicated above the alignment (H1&#x02013;H12). Residues implicated in direct contacts with the ligand are numbered from 1 to 25 below the alignment. The three divergent residues within the LBP between vertebrate RARs are within vertical rectangles in helices 3, 5, and 11. Gnathostome and <italic>Polyandrocarpa</italic> sequences are named with the nomenclature code used in the nuclear receptor database NUREBASE (<ext-link ext-link-type="uri" xlink:href="http://www.ens-lyon.fr/LBMC/laudet/nurebase/nurebase.html">http://www.ens-lyon.fr/LBMC/laudet/nurebase/nurebase.html</ext-link>) [<xref rid="pgen-0020102-b039" ref-type="bibr">39</xref>].</p></caption><graphic xlink:href="pgen.0020102.g002"/></fig><p>To explore how gene function evolved in vertebrates after gene duplications we studied and compared two properties of RARs in different chordates to decipher changes in the coding and non coding moieties of the gene during evolution&#x02014;namely, the ligand-binding capacity and the gene expression pattern during embryonic development, respectively.</p></sec><sec id="s2"><title>Results</title><sec id="s2a"><title>Vertebrate RARs Arose from Duplications at the Origin of Vertebrates</title><p>The phylogeny of RARs (<xref ref-type="fig" rid="pgen-0020102-g001">Figure 1</xref>B), including xRAR&#x003b2; <italic>(Xenopus),</italic> the first report of an <italic>RAR&#x003b2;</italic> gene outside amniotes, is consistent with known chordate phylogeny, and with the hypothesis that <italic>RAR&#x003b1;, &#x003b2;,</italic> and <italic>&#x003b3;</italic> arose from duplications at the origin of vertebrates. Although a number of branchings are not well resolved, support for the nodes that are important to our discussion is strong (RAR&#x003b1;, &#x003b2;, &#x003b3; are supported by bootstrap &#x0003e; 90%).</p></sec><sec id="s2b"><title>RARs Bind ATRA in All Chordates</title><p>Transactivation of a luciferase reporter gene with Gal4-RAR(LBD) constructs in transient transfections, in parallel with a limited proteolysis assay, was used to ascertain both transcriptional and ligand-binding activities of the receptors. Our results show that all the chordate RARs, including the reconstructed AncRAR, are able to transactivate in a dose-dependent manner with ATRA (red bars, <xref ref-type="fig" rid="pgen-0020102-g003">Figures 3</xref> and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>). The EC₅₀ values (~10^&#x02212;9 to 10^&#x02212;8 M) for all the chordate RARs are in a similar range, suggesting that binding of ATRA to RAR is a shared ancestral function in all chordates.</p><fig id="pgen-0020102-g003" position="float"><label>Figure 3</label><caption><title>Transcriptional Activity and Binding Selectivity of Vertebrate RARs</title><p>Transcriptional activity is shown in (A&#x02013;C), (G&#x02013;I), (M), and (N), and corresponding binding selectivity in (D&#x02013;F), (J&#x02013;L), (O), and (P). Identities of the vertebrate RARs for each activity-selectivity pair are indicated above each bar graph. In each case, a chimera comprising the RAR LBD fused to the GAL4 DNA-binding domain (GAL-RAR(LBD)) has been used. The analysis of transcriptional activity in (A&#x02013;C), (G&#x02013;I), (M), and (N) shows transient transactivation assays in Cos1 cells with the indicated GAL-RAR(LBD) expression vector and the cognate (17m)5x-G-luc reporter plasmid, in the presence of increasing concentrations (10^&#x02212;10 to 10^&#x02212;6 M) of ATRA (red bars), BMS753 (yellow bars), BMS641 (light green bars), and BMS961 (dark green bars) respectively. The black bars indicate transactivation in the absence of hormone. Partial proteolysis maps of different in vitro-translated RARs are shown in (D&#x02013;F), (J&#x02013;L), (O), and (P). For each proteolysis gel lane 1 represents the undigested protein, lane 2 shows digestion of the receptor in the absence of ligand, lanes 3 and 4 show digestion of the receptor in the presence of ATRA (10^&#x02212;4 to 10^&#x02212;5 M), lanes 5 and 6 show digestion in the presence of BMS753 (10^&#x02212;4 to 10^&#x02212;5 M), lanes 7 and 8 show digestion in the presence of BMS641 (10^&#x02212;4 to 10^&#x02212;5 M), and lanes 9 and 10 show digestion in the presence of BMS961 (10^&#x02212;4 to 10^&#x02212;5 M). Protected bands in the presence of BMS641 are indicated by an asterisk, and slightly protected bands are indicated by arrowheads.</p></caption><graphic xlink:href="pgen.0020102.g003"/></fig><fig id="pgen-0020102-g004" position="float"><label>Figure 4</label><caption><title>Transcriptional Activity and Binding Selectivity of Chordate RARs</title><p>Transcriptional activity is shown in (A&#x02013;D) and (I&#x02013;L), and corresponding binding selectivity in (E&#x02013;H) and (M&#x02013;P). Identities of the chordate RARs for each activity-selectivity pair are indicated above each bar graph. Transcriptional activity is shown in (A&#x02013;D) for LampRAR, AmphiRAR, PmRAR, and AncRAR, and that of AmphiRAR mutants is shown in (I&#x02013;L). Partial proteolysis maps of the different in vitro-translated RARs are shown in (E&#x02013;H) and (M&#x02013;P). Chimeric GAL-RAR(LBD) transactivation methods, colour code of the transactivation figures, and contents of each proteolysis gel are as in <xref ref-type="fig" rid="pgen-0020102-g003">Figure 3</xref>. Protected bands in the presence of BMS641 are indicated by an asterisk, and slightly protected bands are indicated by arrowheads.</p></caption><graphic xlink:href="pgen.0020102.g004"/></fig></sec><sec id="s2c"><title>The Structure of the LBP Directs the Differential Recognition of Synthetic Monospecific Retinoids</title><p>As previously shown, only three amino acid positions differ between the LBPs of mammalian RARs [<xref rid="pgen-0020102-b022" ref-type="bibr">22</xref>]. These positions account for the different binding specificities with synthetic retinoids (various Bristol-Myers Squibb synthesized retinoids [BMS] compounds [<xref rid="pgen-0020102-b022" ref-type="bibr">22</xref>,<xref rid="pgen-0020102-b024" ref-type="bibr">24</xref>]) to the receptors (<xref ref-type="fig" rid="pgen-0020102-g003">Figures 3</xref> and <xref ref-type="supplementary-material" rid="pgen-0020102-sg001">S1</xref>A). However, the LBP of each RAR differs between several vertebrate orthologues. For example, RAR&#x003b3;s of zebrafish and <italic>Xenopus</italic> have an amino acid in H3 that is found in mammalian RAR&#x003b1; (Ser) but not in mRAR&#x003b3; (mouse; Ala), while the two other positions are the same as those of mRAR&#x003b3; (Met and Ala) (see <xref ref-type="fig" rid="pgen-0020102-g002">Figures 2</xref> and <xref ref-type="supplementary-material" rid="pgen-0020102-sg001">S1</xref>A). Since the overall structure of the receptor can influence the LBP, we tested the selective ligand recognition of zebrafish and <italic>Xenopus</italic> RARs (<xref ref-type="table" rid="pgen-0020102-t001">Table 1</xref>). Transactivation assays in the presence of increasing concentrations of the BMS compounds showed that all the RAR&#x003b1;s from vertebrates that share the same three key amino acid positions (Ser, Ile, Val) have a comparable transactivation pattern as the synthetic compounds (<xref ref-type="fig" rid="pgen-0020102-g003">Figure 3</xref>A, <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>G, <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>H, and <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>M) (i.e., high transactivation with BMS753 and low transactivation with BMS641). However, xRAR&#x003b3; and zfRAR&#x003b3; (zebrafish), which differ at a key amino acid position (Ser in H3) from mRAR&#x003b3; (Ala in H3) (see <xref ref-type="supplementary-material" rid="pgen-0020102-sg001">Figure S1</xref>A), exhibit a pattern intermediate between that of mRAR&#x003b1; and mRAR&#x003b3; (<xref ref-type="fig" rid="pgen-0020102-g003">Figure 3</xref>I, <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>L, <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>N, and <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>P)&#x02014;they transactivate with both BMS753 and BMS961. We noted that the binding abilities of the receptors tested by a limited proteolytic digestion always paralleled the transactivation patterns except for the BMS641 compound (<xref ref-type="fig" rid="pgen-0020102-g003">Figure 3</xref>D&#x02013;<xref ref-type="fig" rid="pgen-0020102-g003">3</xref>F, <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>J&#x02013;<xref ref-type="fig" rid="pgen-0020102-g003">3</xref>L, <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>O, and <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>P), since all the vertebrate RARs can bind this retinoid, but only RAR&#x003b2;, and to a lesser extent RAR&#x003b1;, can transactivate in its presence.</p><p>These data suggest that changes in the LBPs of RARs may have played a functional role during vertebrate evolution. They also show that the use of transactivation and/or binding assays in the presence of synthetic monospecific compounds is an excellent tool for studying the structure-function relationships of different RARs, and potentially of other nuclear receptors, since the transactivation and binding pattern reflect the LBP structure.</p></sec><sec id="s2d"><title>Chordate RARs Share an LBD Structure Able to Bind At Least ATRA and the &#x003b2;-Specific Compound BMS641</title><p>AmphiRAR diverged evolutionarily before the vertebrate-specific genome duplications occurred and represents one of the closest invertebrate RARs to the vertebrate RARs (<xref ref-type="fig" rid="pgen-0020102-g001">Figure 1</xref>). The AmphiRAR sequence has a high percentage of identity with the vertebrate RARs (~88% DNA-binding domain, ~58% LBD). As previously shown, AmphiRAR functions in a dose-response manner in the presence of ATRA (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>) [<xref rid="pgen-0020102-b008" ref-type="bibr">8</xref>]. However, one of the three key amino acid positions within the LBP of AmphiRAR (C225, I263, and V388, <xref ref-type="supplementary-material" rid="pgen-0020102-sg001">Figure S1</xref>A) diverges from those of vertebrate RARs (<xref ref-type="fig" rid="pgen-0020102-g002">Figures 2</xref> and S1A): the position at H3 (Cys) does not correspond to any of the three vertebrate RARs, while the two other key positions (H5 and H11) are conserved with both the mammalian &#x003b1; and &#x003b2; paralogues (Ile and Val). With the synthetic monospecific retinoids, AmphiRAR is able to transactivate the reporter gene in the presence of the mammalian &#x003b2;-specific compound (BMS641) (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>B), which it binds strongly. It also binds the &#x003b1;-specific compound (BMS753) (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>F). This is reminiscent of the specificity exhibited by mRAR&#x003b2;, suggesting that AmphiRAR and mRAR&#x003b2; LBPs share a similar structure (compare <xref ref-type="fig" rid="pgen-0020102-g003">Figure 3</xref>B and <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>E with <xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>B and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>F).</p><p>Although both the tunicate RAR (PmRAR) and the ancestral RAR (AncRAR) transactivate the reporter gene in the presence of increasing amounts of ATRA, neither is able to activate transcription in mammalian cells in the presence of any synthetic monospecific retinoid (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>C and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>D). However, both PmRAR and AncRAR are able to bind weakly the &#x003b2;-specific retinoid (BMS641) (asterisks, <xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>G and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>H), suggesting once again that a similar structure of the LBD is shared by mRAR&#x003b2;, PmRAR, and AncRAR (compare <xref ref-type="fig" rid="pgen-0020102-g003">Figure 3</xref>E with <xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>G and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>H).</p><p>LampRAR is able to bind and transactivate the reporter gene in the presence of increasing concentrations of ATRA and the synthetic compounds BMS753 and BMS641 (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>A and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>E). Twenty-five residues within the LBP of RARs, including the three variable residues of RAR&#x003b1;, &#x003b2;, and &#x003b3; in mammals, make direct contact with the ligand [<xref rid="pgen-0020102-b020" ref-type="bibr">20</xref>] (<xref ref-type="fig" rid="pgen-0020102-g002">Figures 2</xref> and <xref ref-type="supplementary-material" rid="pgen-0020102-sg002">S2</xref>). These 25 positions are strictly conserved between LampRAR LBP and mRAR&#x003b1;. However, the LampRAR transactivation and binding pattern in the presence of the BMS compounds is a composite of those of mRAR&#x003b1; and mRAR&#x003b2; (i.e., high transactivation and binding with both BMS753 and BMS641; compare <xref ref-type="fig" rid="pgen-0020102-g003">Figure 3</xref>A and <xref ref-type="fig" rid="pgen-0020102-g003">3</xref>D with <xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>A and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>E). This result suggests that the overall structure of the receptor can influence the LBP.</p></sec><sec id="s2e"><title>Vertebrate RARs Acquired Different Monospecific Ligand Specificities by Accumulating Mutations in Their LBPs following Gene Duplications</title><p>It is known that single point mutations at the three key positions of the LBP of mammalian RARs suffice to change their specificities for the synthetic monospecific retinoids [<xref rid="pgen-0020102-b022" ref-type="bibr">22</xref>]. Since only the first of the three key positions of the AmphiRAR (Cys225) is divergent compared to mRAR&#x003b1; and &#x003b2;, we mutated it either to Ser (like the corresponding position in mRAR&#x003b1;) or to Ala (like the corresponding position in mRAR&#x003b2;) and determined the capacity of the mutant proteins to bind different synthetic monospecific retinoids. We also asked whether mutating Cys-Ile-Val to Ala-Met-Ala (like the corresponding positions in mRAR&#x003b3;) would confer a &#x003b3;-like specificity to AmphiRAR. The two first mutants (C225A and C225S) conferred &#x003b1;-like and &#x003b2;-like transactivation and binding patterns respectively, (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>I, <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>J, <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>M, and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>N). However, the triple mutant (C225A, I263M, V388A) did not confer the &#x003b3;-like pattern. Instead, this mutant completely lost its capacity to bind any of the monospecific retinoids (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>K and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>O). When we compared the sequence of the 25 amino acid positions of the LBP in AmphiRAR and mRAR&#x003b3; LBPs, we found that the AmphiRAR LBP contains not three but nine divergent positions (seven when compared to mRAR&#x003b1; and mRAR&#x003b2;; see <xref ref-type="fig" rid="pgen-0020102-g002">Figures 2</xref> and <xref ref-type="supplementary-material" rid="pgen-0020102-sg002">S2</xref>). Mutating all of these nine positions to those of the mRAR&#x003b3; LBP recovered the BMS &#x003b3;-like binding behaviour (<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>L and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>P). These results show that a relatively small number of mutations in key residues of the LBP can change the specificity of the RARs.</p></sec><sec id="s2f"><title>Vertebrate RARs Show New Expression Territories When Compared with AmphiRAR</title><p>We previously showed that during amphioxus development, AmphiRAR is strongly expressed at 16 to 24 hours postfertilization in the middle third of the neural tube, somites, and endoderm but not in the cerebral vesicle or notochord [<xref rid="pgen-0020102-b008" ref-type="bibr">8</xref>]. Thus AmphiRAR gene expression decreases strongly at the anterior and posterior parts of the larvae [<xref rid="pgen-0020102-b008" ref-type="bibr">8</xref>] (<xref ref-type="fig" rid="pgen-0020102-g005">Figures 5</xref> and <xref ref-type="supplementary-material" rid="pgen-0020102-sg003">S3</xref>). Expression of the three mammalian and <italic>Xenopus</italic> RARs at comparable stages (at embryonic day 9 [E9] in mouse and at stage 30 in <italic>Xenopus</italic>) is diagrammed in <xref ref-type="fig" rid="pgen-0020102-g005">Figure 5</xref>. In mouse, RAR&#x003b1; is ubiquitously expressed but is at particularly high levels in both the neuroectoderm and mesenchyme of the head, RAR&#x003b3; is strongly expressed in the tail and forebrain, and RAR&#x003b2; transcripts are present in the head mesenchyme, in the trunk tissues, and in the mesonephric duct, but are not detectable in the forebrain and tail (<xref ref-type="fig" rid="pgen-0020102-g005">Figures 5</xref>A&#x02013;<xref ref-type="fig" rid="pgen-0020102-g005">5</xref>F and <xref ref-type="supplementary-material" rid="pgen-0020102-sg004">S4</xref>) [<xref rid="pgen-0020102-b015" ref-type="bibr">15</xref>,<xref rid="pgen-0020102-b016" ref-type="bibr">16</xref>]. A comparable gene expression pattern is observed in <italic>Xenopus,</italic> with a ubiquitous expression of RAR&#x003b1; (especially in the neuroectoderm and head regions), a polarized expression of RAR&#x003b3; in the brain and posterior parts of the embryo, and expression of RAR&#x003b2; in the posterior hindbrain and anterior spinal chord, as well as in branchial arches (<xref ref-type="fig" rid="pgen-0020102-g005">Figures 5</xref>G&#x02013;<xref ref-type="fig" rid="pgen-0020102-g005">5</xref>L and <xref ref-type="supplementary-material" rid="pgen-0020102-sg005">S5</xref>).</p><fig id="pgen-0020102-g005" position="float"><label>Figure 5</label><caption><title>Schematic Representation of the Expression Territories of RARs</title><p>Staining of embryos indicates expression of mRAR&#x003b1; (A), mRAR&#x003b2; (B), and mRAR&#x003b3; (C) in mouse embryos at E9; of xRAR&#x003b1; (G), xRAR&#x003b2; (H), and xRAR&#x003b3; (I) in stage 30 <italic>Xenopus</italic> embryos, and of AmphiRAR (M) in 20 h old amphioxus larvae. Schematic representations are shown of the expression territories of mRARs (D&#x02013;F), xRARs (J&#x02013;L), and AmphiRAR (N) in mouse, <italic>Xenopus,</italic> and amphioxus embryos, respectively. Regions with high levels of expression are red and those with lower levels of expression are pink. Arrowheads indicate regions in mouse and <italic>Xenopus</italic> embryos where the RAR expression cannot be correlated with AmphiRAR expression and can be described as &#x0201c;new expression territories.&#x0201d;</p></caption><graphic xlink:href="pgen.0020102.g005"/></fig></sec></sec><sec id="s3"><title>Discussion</title><p>A general overview of the binding and transactivation capacities of chordate RARs is shown in <xref ref-type="fig" rid="pgen-0020102-g006">Figure 6</xref>. Despite their different transactivation patterns, all the chordate RARs bind the &#x003b2;-specific retinoid BMS641. This suggests that the LBPs of all the chordate RARs share common features. Using AmphiRAR as a model, we have also shown that just a few mutations in the LBP are sufficient to change the binding selectivity of the receptor. Even if possible evolutionary scenarios can be drawn in which a position mutates back and forth between two alternative amino acids, the demonstration of the presence of endogenous RA in amphioxus, the high-affinity binding of ATRA to AmphiRAR, and the activity of ATRA during amphioxus embryonic development [<xref rid="pgen-0020102-b008" ref-type="bibr">8</xref>,<xref rid="pgen-0020102-b025" ref-type="bibr">25</xref>] lead us to propose that the most parsimonious explanation of all these results is that chordate RARs evolved from a common ancestor that was already able to bind ATRA and had an LBP similar to that of modern mammalian RAR&#x003b2;. This model has two interesting implications. First, the mammalian <italic>RAR&#x003b2;</italic> gene has conserved structural and functional aspects of the ancestral <italic>RAR</italic>. Second, following the vertebrate-specific genome duplications, other vertebrate <italic>RAR</italic> genes accumulated mutations in the LBP that changed the structure and the specificity of the protein they encode. Of note, each vertebrate lineage evolved differently, since fish and <italic>Xenopus</italic> RARs have different structure-function relationships from those of the mammalian RARs. Thus, of the three paralogues resulting from vertebrate genome duplications, RAR&#x003b2; retained the ancestral binding specificity, leaving RAR&#x003b1; and &#x003b3; free to &#x0201c;explore&#x0201d; other functionalities. An important constraint on RAR evolution was that all paralogues had to bind the major ligand ATRA. In addition to this constraint, it appears that one duplicate, RAR&#x003b2;, kept the ancestral LBP functionality. This allowed the other paralogues to be selected for new functions. While RAR&#x003b1; remained relatively close to the ancestral RAR (weak transactivation by BMS641), RAR&#x003b3; evolved the most divergent LBP (nine point mutations needed to recover a &#x003b3;-like functional LBP in AmphiRAR [<xref ref-type="fig" rid="pgen-0020102-g004">Figure 4</xref>L and <xref ref-type="fig" rid="pgen-0020102-g004">4</xref>P]). Phylogenetic analysis confirms unambiguously that the three paralogues RAR&#x003b1;, &#x003b2;, and &#x003b3; result from vertebrate-specific gene duplications after the divergence from tunicates and amphioxus and before that of teleost fish (<xref ref-type="fig" rid="pgen-0020102-g001">Figure 1</xref>B). Resolution of the order of duplication events, and the exact position of LampRAR, is less clear, since support for the different topologies of the tree is limited (&#x0003c; 70% bootstrap), perhaps because of rapid evolution of RAR&#x003b3; and of the outgroup RARs.</p><fig id="pgen-0020102-g006" position="float"><label>Figure 6</label><caption><title>Representation of the Transactivation and Binding Capacities of the RARs Used in the Present Study</title><p>The three synthetic retinoids are shown as &#x003b1;, BMS753; &#x003b2;, BMS641; and &#x003b3;, BMS961. The phylogenetic relationships between the RARs have been schematized by a phylogenetic tree (the tunicate and amphioxus RARs have been polytomised, LampRAR is also polytomised with the vertebrate RARs). The putative position in the tree of the ancestral sequence is indicated by a dashed branch in red.</p></caption><graphic xlink:href="pgen.0020102.g006"/></fig><sec id="s3a"><title>Is the RAR Ligand-Binding Specificity the Only Functional Characteristic That Changed following the Vertebrate-Specific Gene Duplications?</title><p>Just as mRAR&#x003b2; and AmphiRAR have similar ligand-binding selectivity, they also have similar expression patterns. In contrast, although mRAR&#x003b1; and &#x003b3; share some expression domains with RAR&#x003b2;, they are also expressed in other embryonic territories (forebrain and caudal regions). Thus, expression of amphioxus RAR is either repressed in the anterior and posterior parts of the embryo and this repression has been lost in vertebrate RAR&#x003b1; and &#x003b3;, or activation of these receptors in anterior and posterior tissues has been acquired during vertebrate evolution.</p><p>Moreover, we have shown that RA directly patterns the pharyngeal endoderm in amphioxus, an invertebrate that lacks neural crest [<xref rid="pgen-0020102-b008" ref-type="bibr">8</xref>,<xref rid="pgen-0020102-b026" ref-type="bibr">26</xref>]. In mammals a role of RA during development of the branchial region was long ascribed to defects in the migration of the neural crest cells. However, it has been demonstrated with isoform-specific retinoids and knockout mice that RA has a direct role on development of the branchial region in the mouse, and that this function is carried exclusively by RAR&#x003b2; and not RAR&#x003b1; or &#x003b3; [<xref rid="pgen-0020102-b027" ref-type="bibr">27</xref>]. This shows that RAR&#x003b2; has not only conserved the expression pattern and the ligand-binding selectivity of the ancestral RAR but also a central biological role during embryonic development.</p><p>Taken together, these findings support a model in which an ancestral RAR containing an LBP close to that of mRAR&#x003b2; was expressed in the trunk region of the putative ancestral chordate. This ancestral RAR patterned the anterior-posterior axis of both the neuroectoderm and endoderm. Following vertebrate-specific gene duplications, neofunctionalization events generated new RAR functions. One vertebrate paralogue, RAR&#x003b2;, was constrained by natural selection and kept most of the ancestral functions, allowing the two other paralogues to take on new possible functions. Thus, RAR&#x003b1; and &#x003b3; gained new expression territories (forebrain and tail regions of the embryo) and, in parallel, they also diverged in their LBP structure.</p><p>Although it has been shown that oxidative derivatives of RA (i.e., degradation products) are not in vivo ligands of RARs [<xref rid="pgen-0020102-b019" ref-type="bibr">19</xref>], the evolutionary scenario presented here leaves open the possibility that vertebrate RARs could bind different ligands in vivo since their LBPs also evolved by neofunctionalization. Research of possible natural ligands with different affinities for each vertebrate RAR paralogue should address this question in the future.</p></sec></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Phylogenetic analysis and ancestral sequence estimation.</title><p>The RAR LBDs used are presented in <xref ref-type="table" rid="pgen-0020102-t001">Table 1</xref>. Amino acid sequences of chordate RARs were aligned using the CLUSTAL W program [<xref rid="pgen-0020102-b028" ref-type="bibr">28</xref>] and manually corrected with SEAVIEW [<xref rid="pgen-0020102-b029" ref-type="bibr">29</xref>]. Phylogenetic trees were inferred by maximum likelihood as implemented in PhyML [<xref rid="pgen-0020102-b030" ref-type="bibr">30</xref>] with the JTT+&#x003b3; model. The 245 complete sites (no gap, no X) were used. Robustness was assessed by bootstrap analysis (1,000 repetitions) [<xref rid="pgen-0020102-b031" ref-type="bibr">31</xref>].</p><p>The ancestral RAR of vertebrates, before duplication, was reconstructed by maximum likelihood as implemented in PAML [<xref rid="pgen-0020102-b032" ref-type="bibr">32</xref>], under the JTT+&#x003b3; substitution model. Sites with indels were treated as follows. (i) If the indel was due to partially sequenced genes, the partial sequences were excluded for reconstruction of these specific sites. (ii) The ancestral state (gap or not) of other indels was estimated manually by parsimony. (iii) If this parsimony analysis predicted that the sites were present in the ancestor, they were reconstructed by maximum likelihood, excluding the sequences with the deletion. Few sites in the LBD were affected by this problem. Overall, the reconstruction was very good, with an average confidence in predicted sites of 0.990.</p></sec><sec id="s4b"><title>Cloning of LampRAR, xRAR&#x003b2;, and mutation constructs.</title><p> <italic>LampRAR</italic> was obtained by semi-nested RT-PCR with degenerate primers based on vertebrate RARs and first-strand cDNA synthesized from total RNA of <named-content content-type="genus-species">Petromyzon marinus</named-content> liver, brain, and muscle. <italic>xRAR&#x003b2;</italic> was obtained by PCR using specific primers based on the incomplete genome of <named-content content-type="genus-species">Xenopus tropicalis</named-content> found in the ENSEMBL database (<ext-link ext-link-type="uri" xlink:href="http://www.ensembl.org/Xenopus_tropicalis/index.html">http://www.ensembl.org/Xenopus_tropicalis/index.html</ext-link>) and first-strand cDNA synthesized from total RNA of <named-content content-type="genus-species">Xenopus laevis</named-content> embryos.</p><p>The Gal4 mutants correspond to a fusion between residues 1 and 147 of Gal4 and the LBDs of the different chordate RARs. The starting position for the LBD was the Ser154 for AmphiRAR (29 amino acids before H1 of the LBD; see <xref ref-type="fig" rid="pgen-0020102-g002">Figure 2</xref>) and the corresponding conserved serine residues of the other chordate RARs. Mutants were constructed by PCR-assisted site-directed mutagenesis. In this procedure mutagenesis is performed by creating an oligonucleotide primer that is complementary to the normal DNA sequence except for the mutant base, which is generally positioned near the 5&#x02032; end of the oligonucleotide to ensure adequate priming. The mutant primer is incorporated by PCR into the newly synthesized DNA. This procedure is repeated as many times as mutations are introduced, and the final DNA fragment is sequenced to confirm the presence of the mutated positions. To subclone in phase with the Gal4 protein ORF, the 5&#x02032; ends of the primers for amplifying each sequence contained restriction sites corresponding to the specific insertion site of the pG4MpolyII vector [<xref rid="pgen-0020102-b033" ref-type="bibr">33</xref>]. The primers with the desired point mutation were designed to overlap the corresponding region within the wild-type sequence. The chimeric and mutant constructs were sequenced to confirm their identity.</p></sec><sec id="s4c"><title>Transactivation assays in mammalian cells.</title><p>Cos-1 (monkey kidney) cells were maintained in DMEM supplemented with 5% charcoal-treated FCS. The cells were transfected at 70% confluence in 24-well plates using 4 &#x003bc;l of ExGen 500 (Euromedex, Souffelweyersheim, France) with 1.0 &#x003bc;g of total DNA including 0.1 &#x003bc;g of reporter plasmid (17m)x5-tk-luc, and 10 ng of CMV-&#x003b2;GAL as an internal control to account for variations of transfection efficiency. The culture medium was changed 6 h after transfection and, when appropriate, ATRA or the RA agonists BMS753, BMS641, and BMS961 in ethanol were added to different final concentrations (10^&#x02212;10 to 10^&#x02212;6 M). Cells were lysed 24 h after transfection and assayed for luciferase activity.</p></sec><sec id="s4d"><title>Limited proteolytic digestion.</title><p>These assays were done as described [<xref rid="pgen-0020102-b034" ref-type="bibr">34</xref>] using in vitro-translated ³⁵S-labelled RARs (TNT kit; Promega, Madison, Wisconsin, United States). Briefly, after incubating at room temperature for 15 min with ligands, receptor proteins were digested at room temperature for 10 min with 25 &#x003bc;g/ml trypsin. The proteolytic fragments were separated on a 10% SDS polyacrylamide gel and visualized by autoradiography.</p></sec><sec id="s4e"><title>In situ hybridization.</title><p> <italic>AmphiRAR, mRAR&#x003b1;, mRAR&#x003b2;, mRAR&#x003b3;, xRAR&#x003b1;, xRAR&#x003b2;,</italic> and <italic>xRAR&#x003b3;</italic> partial cDNAs cloned into the pBluescript vector (Stratagene, La Jolla, California, United States) and linearized with appropriate enzymes were used for synthesis of antisense riboprobes. For <italic>AmphiRAR,</italic> fixation and whole-mount in situ hybridization were done as described [<xref rid="pgen-0020102-b035" ref-type="bibr">35</xref>]. Two probes were combined&#x02014;one synthesized to the 3&#x02032; UTR plus a 735 bp probe to the 5&#x02032; end of the cDNA. For mRARs, whole-mount in situ hybridization was done using standard methods [<xref rid="pgen-0020102-b036" ref-type="bibr">36</xref>]. Probes correspond to the DNA-binding domain for mRAR&#x003b1; and the LBD region for mRAR&#x003b2; and mRAR&#x003b3;. Labelling of the probes was performed using the digoxigenin-UTP labelling kit (Roche, Basel, Switzerland). <named-content content-type="genus-species">Xenopus laevis</named-content> eggs were obtained from females injected with 500 IU of human chorionic gonadotropin, artificially fertilized, dejellied with 2% cysteine hydrochloride (pH 7.8), and cultured in 0.1&#x000d7; modified Barth's saline. Embryos were staged as described [<xref rid="pgen-0020102-b037" ref-type="bibr">37</xref>]. In situ hybridization was carried out as previously reported in [<xref rid="pgen-0020102-b038" ref-type="bibr">38</xref>].</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting Information</title><supplementary-material content-type="local-data" id="pgen-0020102-sg001"><label>Figure S1</label><caption><title>Conservation/Divergence of the Sequences Used in the Present Study</title><p>(A) Amino acids within the LBP corresponding to the three variable positions in H3, H5, and H11 of mammalian RARs, for the chordate RARs used in this study.</p><p>(B) Distance matrix of the sequences used in the alignment shown in <xref ref-type="fig" rid="pgen-0020102-g002">Figure 2</xref>. Gnathostome and <italic>Polyandrocarpa</italic> sequences are named with the nomenclature code used in the nuclear receptor database NUREBASE (<ext-link ext-link-type="uri" xlink:href="http://www.ens-lyon.fr/LBMC/laudet/nurebase/nurebase.html">http://www.ens-lyon.fr/LBMC/laudet/nurebase/nurebase.html</ext-link>) [<xref rid="pgen-0020102-b039" ref-type="bibr">39</xref>].</p><p>(3.8 MB TIF)</p></caption><media xlink:href="pgen.0020102.sg001.tif"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020102-sg002"><label>Figure S2</label><caption><title>Schematic Representation of the 25 Residues Present in the LBP of AmphiRAR Contacting ATRA</title><p>The three key residues that differ between vertebrate RARs are indicated by a circle. The other six residues that differ between mRAR&#x003b3; and AmphiRAR are indicated by a square. Adapted from [<xref rid="pgen-0020102-b020" ref-type="bibr">20</xref>].</p><p>(220 KB TIF)</p></caption><media xlink:href="pgen.0020102.sg002.tif"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020102-sg003"><label>Figure S3</label><caption><title>Expression of AmphiRAR in Amphioxus Embryos</title><p>Whole mounts are shown with anterior toward the left. Blastula stage shows no expression and gastrula shows ubiquitous expression (unpublished data). In early neurula (15 h) expression is down-regulated in the cerebral vesicle (arrow), anterior endoderm and non-neural ectoderm. In 18- to 22-h neurula the expression is down-regulated in the anterior third of the nerve cord and in the pharyngeal endoderm and is up-regulated in the middle third of the embryo. In 24- to 30-h embryos the expression is strong in the nerve cord posterior to the cerebral vesicle and in a small region of endoderm, but is largely down-regulated elsewhere.</p><p>(3.1 MB TIF)</p></caption><media xlink:href="pgen.0020102.sg003.tif"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020102-sg004"><label>Figure S4</label><caption><title>Expression of mRAR&#x003b1;, mRAR&#x003b2;, and mRAR&#x003b3; in Mouse Embryos from E9 to E12.5</title><p>RAR&#x003b1; is ubiquitously expressed from E9 to E12.5 with a very high level in the anterior brain. RAR&#x003b2; is expressed in the central part of the CNS from E9 to E12.5 as well as in some parts of the head mesenchyme and in other trunk tissues. RAR&#x003b3; is mainly expressed in the forebrain, the tail, the branchial arches, and the limb buds as they develop.</p><p>(9.0 MB TIF)</p></caption><media xlink:href="pgen.0020102.sg004.tif"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020102-sg005"><label>Figure S5</label><caption><title>Expression of xRAR&#x003b1;, xRAR&#x003b2;, and xRAR&#x003b3; in <italic>Xenopus</italic> embryos</title><p>Whole mounts are oriented with anterior toward the right. Expression of RAR&#x003b1; and RAR&#x003b3; is detectable from the onset of gastrulation (stage 10) while the first signal for RAR&#x003b2; is detected at the early tailbud stage (stage 25). At mid-gastrula stage (stage 11), RAR&#x003b1; is expressed as a narrow ring around the blastopore. As gastrulation proceeds, expression intensifies and the signal around the blastopore widens preferentially on the dorsal side except in the midline, which exhibits a low level of transcripts. During neurulation (from stage 14), transcripts are found predominantly in the neurectoderm, evenly distributed along the anterior-posterior axis, with the exception of a region at the anterior end for which transcripts are largely reduced. At the tailbud stage (stage 30), RAR&#x003b1; is predominantly expressed in the spinal cord and the posterior hindbrain, in the eye and the posterior branchial arches. During gastrulation (stage 10), expression of RAR&#x003b3; is more widespread than RAR&#x003b1; expression. Transcripts are present in the mesodermal marginal zone as well as in the ectoderm. By the neurula stage (from stage 14), the staining separates into anterior and posterior domains, thus creating a gap with no RAR&#x003b3; transcripts. Expression remains localized to the posterior and anterior ends of the embryo at tailbud stages (stage 30) and is mainly restricted to the branchial arches and the tip of the tailbud. RAR&#x003b2; transcripts are detected at much lower level than RAR&#x003b1; and RAR&#x003b3; at the examined stages. The signal is restricted to the caudal part of the hindbrain and the anterior spinal cord. At the late tailbud stage (stage 32), RAR&#x003b2; is strongly expressed in the most posterior branchial arches. d, dorsal views; l, lateral views; f, frontal views.</p><p>(2.8 MB TIF)</p></caption><media xlink:href="pgen.0020102.sg005.tif"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material></sec>

In Vivo Robustness Analysis of Cell Division Cycle Genes in <named-content content-type="genus-species">Saccharomyces cerevisiae</named-content>

<p>Intracellular biochemical parameters, such as the expression level of gene products, are considered to be optimized so that a biological system, including the parameters, works effectively. Those parameters should have some permissible range so that the systems have robustness against perturbations, such as noise in gene expression. However, little is known about the permissible range in real cells because there has been no experimental technique to test it. In this study, we developed a genetic screening method, named &#x0201c;genetic tug-of-war&#x0201d; (gTOW) that evaluates upper limit copy numbers of genes in a model eukaryote <italic>Saccharomyces cerevisiae,</italic> and we applied it for 30 cell-cycle related genes (<italic>CDC</italic> genes). The experiment provided unique quantitative data that could be used to argue the system-level properties of the cell cycle such as robustness and fragility. The data were used to evaluate the current computational model, and refinements to the model were suggested.</p>

<contrib contrib-type="author"><name><surname>Moriya</surname><given-names>Hisao</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff2">2</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><contrib contrib-type="author"><name><surname>Shimizu-Yoshida</surname><given-names>Yuki</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff2">2</xref></contrib><contrib contrib-type="author"><name><surname>Kitano</surname><given-names>Hiroaki</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff2">2</xref><xref ref-type="aff" rid="aff3">3</xref></contrib>

PLoS Genetics

<sec id="s1"><title>Introduction</title><p>Intracellular biochemical parameters, such as gene expression level, are considered to have become optimized through the long history of evolution so that cells can precisely perform their biological activity. These parameters must have permissible ranges against internal perturbations, such as noise in gene expression and external perturbations such as temperature variation. On the other hand, these parameters need to be dynamically changed during cellular responses against environmental changes or the cell division cycle. Recent computational analyses using mathematical models based on molecular biological knowledge revealed characteristics of these parameters, and the robustness of biological systems against parameter perturbations has been discussed [<xref rid="pgen-0020111-b001" ref-type="bibr">1</xref>&#x02013;<xref rid="pgen-0020111-b008" ref-type="bibr">8</xref>]. However, little is known about the permissible ranges of parameters in real cells because there has been no experimental technique to comprehensively measure the limits of intracellular parameter.</p><p>To reduce the expression level of a target gene, gene knockout experiments, by which the expression level is reduced to zero, are used. For example, in model organisms such as <italic>Saccharomyces cerevisiae,</italic> comprehensive gene knockout or promoter titration analyses have been performed [<xref rid="pgen-0020111-b009" ref-type="bibr">9</xref>,<xref rid="pgen-0020111-b010" ref-type="bibr">10</xref>]. These experiments provide phenotypical information that reveals the functions of target genes. Recent synthetic knockout analyses also have provided comprehensive information on the genetic interaction of the genome [<xref rid="pgen-0020111-b011" ref-type="bibr">11</xref>&#x02013;<xref rid="pgen-0020111-b013" ref-type="bibr">13</xref>]. However, such experiments do not provide quantitative information associated with the limit of expression of the target gene in order to function. On the other hand, to increase the expression level of a target gene, promoter-swapping experiments, in which the target gene's promoter is changed into a strong promoter, are used. For example, in <italic>S. cerevisiae,</italic> the <italic>GAL1</italic> promoter, which can induce strong gene expression in galactose medium, is commonly used. This method also has provided much useful information for predicting the functions of target genes, as well as genetic interactions between target genes [<xref rid="pgen-0020111-b014" ref-type="bibr">14</xref>&#x02013;<xref rid="pgen-0020111-b017" ref-type="bibr">17</xref>]. However, it is also difficult to determine the upper limit of the expression of the target gene because this method ignores the native expression level and regulation of the target gene.</p><p>In this study, we attempted to estimate the upper limit of the gene expression level of each target gene by increasing the copy number of the gene. We used each target gene with its native regulatory DNA elements (promoter and terminator) as a unit so that the increased copy number of the gene can be determined quantitatively, and the gene expression level is expected to increase according to the copy number. We applied the properties of 2-micron-based plasmid with the <italic>leu2d</italic> marker gene, whose copy number increases more than 100 under selectable conditions. If the target gene cloned into the plasmid has an upper limit of less than 100, the plasmid copy number under the selectable condition is expected to become close to the upper limit of the target gene. We named this method &#x0201c;genetic tug-of-war&#x0201d; (gTOW).</p><p>The cell division cycle is an essential process for cells, and the process has been studied most extensively at the molecular level in <named-content content-type="genus-species">S. cerevisiae</named-content>. Many regulatory mechanisms of the cell division cycle in <named-content content-type="genus-species">S. cerevisiae</named-content> are conserved among most eukaryotic cells [<xref rid="pgen-0020111-b018" ref-type="bibr">18</xref>]. Recently, Chen et al. developed a comprehensive computer model of the cell division cycle in <named-content content-type="genus-species">S. cerevisiae</named-content> [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>]. This model represents more than 100 experimentally tested phenotypes of mutants and represents and predicts some quantitative behaviors of the system [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>&#x02013;<xref rid="pgen-0020111-b021" ref-type="bibr">21</xref>]. More than 70% of the parameters in the model have a permissible range of both a 10-fold increase and decrease to maintain the cell division cycle [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>].</p><p>In this study, we applied the gTOW method to evaluate the upper limit dosage of 30 cell division cycle&#x02013;related genes (<italic>CDC</italic> genes). The upper limit data obtained were compared with other systematic quantitative and qualitative datasets to date and used to explore the relationship between the limit and the system-level property of the cell cycle. Using predictions provided by the computer model as a reference, we discussed further directions for experimental and computational studies.</p></sec><sec id="s2"><title>Results</title><sec id="s2a"><title>Principles of the gTOW Method</title><p>To determine the upper limit copy number of target genes, we used the genetic properties of 2-micron plasmid vectors with <italic>leu2d</italic>. Plasmid vectors derived from the 2-micron circle, which is a naturally observed selfish DNA in <italic>S. cerevisiae,</italic> have copy numbers of 10 to 40 per cell with large variations from cell to cell [<xref rid="pgen-0020111-b022" ref-type="bibr">22</xref>]. <italic>leu2d</italic> is an allele of a leucine biosynthesis gene <italic>LEU2,</italic> with a very weak complementation activity because it has a large deletion in its promoter. When <italic>leu2</italic> deletion yeast cells transformed with a 2-micron plasmid with <italic>leu2d</italic> are cultured under leucine&#x02212; condition, the cells with higher plasmid copy numbers grow faster, and the cells with a copy number of more than 100 per cell are eventually concentrated [<xref rid="pgen-0020111-b022" ref-type="bibr">22</xref>]. This strong genetic selection bias was used to increase the copy number of each target gene cloned in a 2-micron plasmid with <italic>leu2d</italic> plasmid for genetic tug-of-war (pTOW<italic>-target</italic>) (<xref ref-type="fig" rid="pgen-0020111-g001">Figure 1</xref>A, red arrowhead). On the other hand, if the target gene inhibits growth when it becomes more than a certain copy number (i.e., the gene has its upper limits), the cells with plasmid copy number lower than the limit grow faster. Thus, the target gene becomes another genetic selection bias toward decreasing the plasmid copy number (<xref ref-type="fig" rid="pgen-0020111-g001">Figure 1</xref>A, blue arrowhead). The high copy selection bias due to <italic>leu2d</italic> in the leucine&#x02212; condition is always constant. In contrast, the leucine&#x02212; low copy selection bias due to the target gene is dependent on its upper limit, as in the case of Gene A and Gene B in <xref ref-type="fig" rid="pgen-0020111-g001">Figure 1</xref>A. As a result of this tug-of-war between these two selection biases, cells with optimized plasmid copy number, which is expected to be close to the upper limit copy number of the target gene, are concentrated (<xref ref-type="fig" rid="pgen-0020111-g001">Figure 1</xref>A, black arrowhead and filled circles). It should be noted that the pTOW-<italic>target</italic> has another marker <italic>URA3</italic> with complete activity, so the plasmid can be constructed and maintained in the leucine+ uracil&#x02212; condition where the strong high copy bias is free. In this condition, however, there is still a weak bias toward increasing the plasmid copy number (up to about 40) due to the nature of 2-micron plasmid (see below).</p><fig id="pgen-0020111-g001" position="float"><label>Figure 1</label><caption><title>gTOW Is a Genetic Selection Method That Determines the Upper Limit Copy Number of Target Genes</title><p>(A) Principle of the gTOW method. The cells with plasmid copy number close to the upper limit of each target gene are concentrated because of the genetic tug-of-war that emerges due to the high copy selection bias due to <italic>leu2d</italic> and the low copy selection bias due to the target gene. See text for details.</p><p>(B) A scatter plot shows the correlation of maximum growth rates and plasmid copy numbers of each <italic>CDC</italic> gene determined in the gTOW experiment. The data used in the graphs are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st001">Tables S1</xref> and <xref ref-type="supplementary-material" rid="pgen-0020111-st002">S2</xref>.</p></caption><graphic xlink:href="pgen.0020111.g001"/></fig></sec><sec id="s2b"><title>Evaluation of Upper Limit Copy Numbers of <italic>CDC</italic> Genes by the gTOW Method</title><p>As a test case, we analyzed the 30 cell division cycle&#x02013;related <italic>(CDC)</italic> genes listed in <xref ref-type="table" rid="pgen-0020111-t001">Table 1</xref> with this method. These genes are involved in the regulation of cyclin-dependent kinase (CDK) activity through the cell cycle [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>,<xref rid="pgen-0020111-b023" ref-type="bibr">23</xref>]. We cloned each <italic>CDC</italic> gene with its native regulatory DNA elements, such as promoter and terminator, so that the regulation of the gene expression can be comparable to the chromosomal copy. Because these elements have not been fully determined so far, we cloned each gene with upstream and downstream DNA sequences up to the neighboring open reading frames (ORFs) into the pTOW-<italic>target.</italic> We measured the maximum growth rate of yeast BY4741 cells [<xref rid="pgen-0020111-b024" ref-type="bibr">24</xref>] transformed with each of these pTOW-<italic>CDCgene</italic> plasmids in the leucine&#x02212; condition. We then determined the plasmid copy number (i.e., the gene copy number on the plasmid) within the cells cultured for 50 h as follows: The total DNA extracted from the cells was tested by two quantitative PCR with two primer pairs that amplify fragments of <italic>LEU2</italic> on the plasmid and <italic>LEU3</italic> on the chromosome, respectively. Then the ratio of the amount of <italic>LEU2</italic> to that of <italic>LEU3</italic> was calculated. Thus, the plasmid copy number is per haploid genome, and the actual copy number of the target gene is the plasmid copy number plus one because there is one extra copy on the chromosome.</p><table-wrap id="pgen-0020111-t001" content-type="2col" position="float"><label>Table 1</label><caption><p>Cell Division Cycle&#x02013;Related <italic>(CDC)</italic> Genes and Their Plasmids Used in This Study</p></caption><graphic xlink:href="pgen.0020111.t001"/></table-wrap><p> <xref ref-type="fig" rid="pgen-0020111-g001">Figure 1</xref>B is a scatter plot showing the relationship between the growth rate and the copy number of the 30 <italic>CDC</italic> genes determined in the gTOW experiment. The copy numbers and growth rates of genes with copy numbers of less than 60 showed a linear correlation. This indicates that the obtained copy numbers of these genes were determined by gTOW as shown in <xref ref-type="fig" rid="pgen-0020111-g001">Figure 1</xref>A (a hypothetical mechanism for this linearity is shown in <xref ref-type="supplementary-material" rid="pgen-0020111-sg001">Figure S1</xref>). Genes with more than 60, but less than 80 plasmid copies, did not show obvious growth retardation, but the copy numbers are probably determined by gTOW as well (see below). <xref ref-type="fig" rid="pgen-0020111-g002">Figure 2</xref>A shows the plasmid copy numbers of 30 <italic>CDC</italic> genes obtained in the gTOW experiment. They were diverse, from less than one to more than 100. The plasmid vector without a target gene showed about 160 copies per haploid genome. Genes with a copy number close to the vector probably do not reach the upper limit. Some cells with genes of low copy numbers showed abnormal morphologies such as cell elongation. Those morphologies were similar to the ones which have been observed when <italic>CDC</italic> genes are overexpressed (unpublished data), supporting the idea that the proteins encoded on the target genes were overexpressed and determined the copy number in the gTOW experiment.</p><fig id="pgen-0020111-g002" position="float"><label>Figure 2</label><caption><title>Upper Limit Copy Numbers of <italic>CDC</italic> Genes</title><p>Copy number of the plasmid with each of the 30 <italic>CDC</italic> genes was determined in the gTOW experiment in leucine&#x02212; (A) and uracil&#x02212; (B) conditions. *The copy numbers of <italic>CLB3</italic> and <italic>CLB2</italic> were determined in the conditions with 40 &#x003bc;g/ml and 20 &#x003bc;g/ml leucine, respectively. The data used in the graphs are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st002">Table S2</xref>.</p></caption><graphic xlink:href="pgen.0020111.g002"/></fig><p>The cells with very low copy plasmid numbers (&#x0003c;35) hardly grew in the leucine&#x02212; condition, probably because the number of <italic>leu2d</italic> is too low to support the <italic>leu2</italic> deletion. When they were cultivated for a very long time, revertant cells were sometimes observed. In this case, we further evaluated the upper limit by adding a low concentration of leucine into the growth medium in order to reduce the bias toward increasing the plasmid copy number (<xref ref-type="fig" rid="pgen-0020111-g003">Figure 3</xref>). Along with the increase of leucine concentration in the medium, growth of the cells with the vector alone progressively increased, and the plasmid copy number in the cells decreased from about 160 to 40 copies per haploid genome (vector in <xref ref-type="fig" rid="pgen-0020111-g003">Figure 3</xref>). We tested some genes with low plasmid copy numbers in the leucine&#x02212; condition. Genes such as <italic>CLB2, CLB3, CLB5, MCM1, SIC1,</italic> and <italic>SWE1</italic> showed a dramatic switch-like decrease of growth below certain leucine concentrations (<xref ref-type="fig" rid="pgen-0020111-g003">Figure 3</xref>). This switch-like decrease may be due to the nature of the regulation of these genes with positive feedbacks (see below). In these cases, we may have to regard the copy number in the leucine concentration just before the dramatic growth retardation as the upper limit.</p><fig id="pgen-0020111-g003" position="float"><label>Figure 3</label><caption><title>Determination of Upper Limit Gene Copy Number with Leucine Supplementation</title><p>Growth and copy number of plasmids in the gTOW experiment with various leucine concentrations are shown. The data used in the graphs are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st002">Table S2</xref>.</p></caption><graphic xlink:href="pgen.0020111.g003"/></fig><p>As noted above, the plasmid copy numbers in the uracil&#x02212; (leucine+) condition also seem to be determined by weak gTOW. In this condition, growth retardation due to maintaining the plasmid was not observed except <italic>CDC14</italic> (see below). The plasmid copy numbers, however, were rather firmly determined depending on the target genes cloned (<xref ref-type="fig" rid="pgen-0020111-g002">Figure 2</xref>B), and the trend was related with the copy number trend in the leucine&#x02212; condition (<xref ref-type="supplementary-material" rid="pgen-0020111-sg002">Figure S2</xref>). This indicates that there is a weak bias toward increasing the plasmid copy number due to the 2-micron plasmid, and the bias can be used to evaluate the upper limit of target genes under mild conditions. On the other hand, some genes such as <italic>BUB2, CDC15,</italic> and <italic>MIH1</italic> interestingly showed significantly higher copy number than the vector alone (<italic>p</italic> &#x0003c; 0.02, two-tailed Student's <italic>t</italic>-test). Among 30 <italic>CDC</italic> genes, only <italic>CDC14</italic> showed severe growth retardation even in the uracil&#x02212; condition (<italic>CDC14</italic> in <xref ref-type="fig" rid="pgen-0020111-g003">Figure 3</xref>). To our surprise, the copy numbers of <italic>CDC14</italic> through any leucine concentration were about 1 (<italic>CDC14</italic> in <xref ref-type="fig" rid="pgen-0020111-g003">Figure 3</xref>), and the cells even in the uracil&#x02212; condition were severely elongated (unpublished data). This very low upper limit counteracting the weak bias of the 2-micron plasmid probably caused the growth retardation even in the uracil&#x02212; condition.</p></sec><sec id="s2c"><title>Protein Expressed from the Target Gene Determines the Plasmid Copy Number in the gTOW Experiment</title><p>We cannot exclude the possibility that the plasmid copy number in the gTOW experiment is determined because of the effect of DNA or mRNA of the target gene, because the copy number of the plasmid DNA itself is increased in this experiment. We therefore disrupted the ORFs of target genes by inserting an adenine just after the start codon to introduce frame shift mutations. Thus, the DNA sequences of the plasmids with wild-type and the frame shift mutant were exactly the same except for one nucleotide. As shown in <xref ref-type="fig" rid="pgen-0020111-g004">Figure 4</xref>A, the growth rates and the plasmid copy numbers with the frame shift mutants were increased toward the levels of the vector alone. All frame shift mutants showed more than 80 copies per haploid genome, indicating that the upper limits of genes with copy numbers of less than 80 in the leucine&#x02212; condition were evaluated by the gTOW experiment due to the effect of gene products (proteins) expressed from the target genes. As mentioned above, within the genes with higher plasmid copy numbers than the vector alone in the uracil&#x02212; condition, <italic>BUB2</italic> and <italic>MIH1</italic> showed reduced copy numbers with frame shift mutation (<xref ref-type="fig" rid="pgen-0020111-g004">Figure 4</xref>B). These genes probably work positively for rapid cell growth, most likely by ignoring some checkpoints in the cell cycle.</p><fig id="pgen-0020111-g004" position="float"><label>Figure 4</label><caption><title>Control Experiments with Frame Shift Mutants in the gTOW Experiment</title><p>(A) A scatter plot showing the correlation of maximum growth rates and plasmid copy numbers with wild-type <italic>CDC</italic> genes and their frame shift mutants determined in the gTOW experiment.</p><p>(B) Plasmid copy numbers with wild-type and frame shift mutants in the gTOW experiment in the uracil&#x02212; condition. The data used in the graphs are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st002">Table S2</xref>.</p></caption><graphic xlink:href="pgen.0020111.g004"/></fig></sec><sec id="s2d"><title>The Data Obtained in the gTOW Experiment Provide Unique Information</title><p>We next compared the data obtained in the gTOW experiment with other quantitative and qualitative datasets obtained in systematic analyses. First of all, the essentialities of genes [<xref rid="pgen-0020111-b009" ref-type="bibr">9</xref>] did not show any significant relationship with the copy number in the gTOW experiment (Spearman's <italic>r</italic> = 0.08, <italic>p</italic> &#x0003e; 0.05) (<xref ref-type="fig" rid="pgen-0020111-g005">Figure 5</xref>A). Neither the fitness of deletion mutants of nonessential genes [<xref rid="pgen-0020111-b009" ref-type="bibr">9</xref>], nor haploid insufficiency [<xref rid="pgen-0020111-b025" ref-type="bibr">25</xref>], showed any correlation with the copy number (unpublished data). The endogenous level of proteins in wild-type cells, as determined by Ghaemmaghami et al. [<xref rid="pgen-0020111-b026" ref-type="bibr">26</xref>] using TAP-tagged proteins, did not show any significant correlation with the maximum tolerated plasmid copy number in our gTOW assay (Spearman's <italic>r</italic> = &#x02212;0.36, <italic>p</italic> &#x0003e; 0.05) (<xref ref-type="fig" rid="pgen-0020111-g005">Figure 5</xref>B, see also <xref ref-type="supplementary-material" rid="pgen-0020111-sg003">Figure S3</xref>). This indicates that the copy numbers in the gTOW experiment are determined not by the non-specific effect of overexpressed proteins to perturb general cellular functions, such as protein expression and protein degrading system, but by the specific effects of the function of each protein.</p><fig id="pgen-0020111-g005" position="float"><label>Figure 5</label><caption><title>Comparisons of the Data Obtained in the gTOW Experiment with Other Qualitative and Quantitative Datasets</title><p>(A) The plasmid copy numbers in the gTOW experiment with essential genes and nonessential genes [<xref rid="pgen-0020111-b009" ref-type="bibr">9</xref>,<xref rid="pgen-0020111-b041" ref-type="bibr">41</xref>].</p><p>(B) A scatter plot between protein abundance obtained by Ghaemmaghami et al. [<xref rid="pgen-0020111-b026" ref-type="bibr">26</xref>] and the plasmid copy number in the gTOW experiment.</p><p>(C) A scatter plot of growth inhibitions between the <italic>GAL</italic> experiment and gTOW experiment with the 30 <italic>CDC</italic> genes.</p><p>(D) A scatter plot between protein abundance determined by Cross et al. [<xref rid="pgen-0020111-b021" ref-type="bibr">21</xref>] and the plasmid copy number of B-type cyclins in the gTOW experiment. The protein abundance is the number of copies per log-phase yeast cell of the indicated cyclin protein C-terminally tagged with protein A expressed from the endogenous promoter and chromosomal location [<xref rid="pgen-0020111-b021" ref-type="bibr">21</xref>].</p><p>(E) A scatter plot between selectable disadvantage determined by Cross et al. [<xref rid="pgen-0020111-b021" ref-type="bibr">21</xref>] and the plasmid copy number of B-type cyclins in the gTOW experiment. The selective disadvantage parameter reflects differences in doubling time between wild-type and mutant cells. The data used in the graphs are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st001">Tables S1</xref>, <xref ref-type="supplementary-material" rid="pgen-0020111-st002">S2</xref>, and <xref ref-type="supplementary-material" rid="pgen-0020111-st003">S3</xref>.</p></caption><graphic xlink:href="pgen.0020111.g005"/></fig><p>Next, we made a comparison with the standard overexpression system using the <italic>GAL1</italic> promoter [<xref rid="pgen-0020111-b027" ref-type="bibr">27</xref>]. To make the conditions the same, we constructed a series of plasmids in which each promoter on pTOW-<italic>CDCgene</italic> was replaced by the <italic>GAL1</italic> promoter (<xref ref-type="table" rid="pgen-0020111-t001">Table 1</xref>, pGAL-<italic>CDCgene</italic>). We then measured and calculated the maximum growth rate of the cells with the plasmid. As shown in <xref ref-type="fig" rid="pgen-0020111-g005">Figure 5</xref>C, in overall, both data showed significant correlation (Spearman's <italic>r</italic> = 0.82, <italic>p</italic> &#x0003c; 0.01). However, some genes such as <italic>CLB3, CLB5, CLB6, NET1,</italic> and <italic>PDS1</italic> showed different degrees of inhibition between the two experiments. This is probably due to the difference of the nature of the expression system in both experiments. Thus, we found that the gTOW experiment provided unique quantitative data.</p></sec><sec id="s2e"><title>B-Type Cyclins</title><p>B-type cyclins (Clbs) were encoded in six genes <italic>(CLB1&#x02013;6),</italic> which are constituted of three paralogous gene pairs <italic>(CLB1</italic>/<italic>CLB2, CLB3</italic>/<italic>CLB4,</italic> and <italic>CLB5</italic>/<italic>CLB6)</italic> [<xref rid="pgen-0020111-b028" ref-type="bibr">28</xref>]. One gene <italic>(CLB2, CLB3,</italic> and <italic>CLB5)</italic> or each paralogous pair is expressed stronger and has a more major function on cellular fitness. Moreover, it is reported that the fitness of deletion mutant cells of each gene and the intracellular protein level from them have a correlation [<xref rid="pgen-0020111-b021" ref-type="bibr">21</xref>]. As shown in <xref ref-type="fig" rid="pgen-0020111-g005">Figure 5</xref>D and <xref ref-type="fig" rid="pgen-0020111-g005">5</xref>E, <italic>CLBs,</italic> in particular, showed a close correlation between their copy numbers in the gTOW experiment with their intracellular protein abundance (Spearman's <italic>r</italic> = &#x02212;0.84, <italic>p</italic> &#x0003c; 0.05) and selectable disadvantage (i.e., fitness of the deletion mutant) (Spearman's <italic>r</italic> = &#x02212;0.77, <italic>p</italic> &#x0003c; 0.05). <italic>CLBs</italic> with major function (i.e., strongly expressed) had low copy numbers in the gTOW experiment, indicating that they have relatively narrow parameter ranges. Cross et al. recently measured the upper limit of <italic>Clb2</italic> and reported it to be less than 13-fold that of the wild-type protein amount [<xref rid="pgen-0020111-b020" ref-type="bibr">20</xref>]. In the leucine&#x02212; condition the copy number of <italic>CLB2</italic> in the gTOW experiment was about 30 (<italic>CLB2</italic> in <xref ref-type="fig" rid="pgen-0020111-g003">Figure 3</xref>). In this condition, however, the growth of cells with the plasmid was strongly inhibited. The copy number in the condition just before the strong growth inhibition (i.e., 40 mg/ml leucine) was about 12, which is almost consistent with their finding (<italic>CLB2</italic> in <xref ref-type="fig" rid="pgen-0020111-g003">Figure 3</xref>). We also measured the Clb2 protein amount in this condition and the amount was about 12-fold that of the wild-type protein amount <bold>(</bold> <xref ref-type="fig" rid="pgen-0020111-g007">Figure 7</xref> <bold>,</bold> see below).</p></sec><sec id="s2f"><title>Relationship between Upper Limit Gene Dosage and System Level Property of the Cell-Cycle System</title><p>To reveal the relationship between the copy numbers obtained in the gTOW experiment and properties of the cell-cycle system, we drew a map presenting current knowledge of molecular interactions in the cell cycle in <named-content content-type="genus-species">S. cerevisiae</named-content> and represented the copy number data on it (<xref ref-type="supplementary-material" rid="pgen-0020111-sg004">Figure S4</xref>). Interestingly, six out of seven genes with the lowest copy numbers were involved in the direct regulation of B-type CDK activity. They were the major B-type cyclin paralogous genes described above <italic>(CLB2, CLB3,</italic> and <italic>CLB5), SIC1</italic> that encodes a stoichiometric inhibitor of B-type CDK [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>], and <italic>SWE1</italic> that encodes an inhibitory kinase of B-type CDK [<xref rid="pgen-0020111-b023" ref-type="bibr">23</xref>]. Moreover, it is known that they construct a subsystem with three positive feedback loops that regulate B-type CDK activity (<xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>A), and these positive feedbacks make bi-stable B-type CDK states [<xref rid="pgen-0020111-b023" ref-type="bibr">23</xref>,<xref rid="pgen-0020111-b029" ref-type="bibr">29</xref>]. The dynamic alteration between the two B-type CDK states is thought to be the core process required for robust oscillation of the eukaryotic cell cycle [<xref rid="pgen-0020111-b030" ref-type="bibr">30</xref>]. Thus, it was first observed that the subsystem designed to dynamically change parameters conversely had a low limit of permissible parameter range. In contrast, G1 cyclins <italic>(CLN1</italic>, <italic>CLN2,</italic> and <italic>CLN3), CDC20,</italic> and <italic>CDH1,</italic> which alter the stable states, had higher copy numbers (&#x0003e;50).</p><fig id="pgen-0020111-g006" position="float"><label>Figure 6</label><caption><title>System Level Analysis of the Copy Number in the gTOW Experiment</title><p>(A) Genes with low copy numbers in the gTOW experiment construct a subsystem that contains three positive feedback loops which regulate B-type cyclin activity [<xref rid="pgen-0020111-b023" ref-type="bibr">23</xref>,<xref rid="pgen-0020111-b029" ref-type="bibr">29</xref>]. This subsystem potentially makes two stable steady-states of B-type cyclin activity. Those states correspond to G1 phase and G2-S-M phase, respectively, which are altered by Cln-CDK and Cdc20-APC activity in the normal cell cycle [<xref rid="pgen-0020111-b029" ref-type="bibr">29</xref>].</p><p>(B) Comparison of the upper limit gene copy numbers obtained in the gTOW experiment and a computational cell cycle model. The upper limit gene copy numbers in vivo are considered as the plasmid copy numbers plus 1 in the gTOW experiment. Maximum fold variable change in xlog(2) toward the outside of the circle up to 256-fold is shown. The data used in the graphs are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st004">Table S4</xref>.</p><p>(C) Upper limit copy numbers of stoichiometric partners in predictions of the computer model and gTOW experiment. The data used in the graph are the same as in <xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>B.</p><p>(D) Simplified diagrams showing stoichiometric partners, where the enzymes are regulated by their inhibitors by 1:1 stoichiometric molecular interaction. The diagrams are shown as implemented in the computer model by Chen et al. [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>].</p></caption><graphic xlink:href="pgen.0020111.g006"/></fig><fig id="pgen-0020111-g007" position="float"><label>Figure 7</label><caption><title>Quantification of the Protein Level Expressed in the gTOW Experiment</title><p>(A) Typical example of the Western blots in Class I genes. Circled samples were used for the measurement of each protein bands.</p><p>(B) Typical example of the Western blots in Class II genes. Squared samples were used for the measurement of each protein bands.</p><p>(C) Scatter plot between the protein overexpression level and the gene copy number (plasmid copy number plus 1) in the gTOW experiment. The data were determined in the conditions with 100 &#x003bc;g/ml <italic>(CDC14),</italic> 20 &#x003bc;g/ml <italic>(CLB2, CLB5,</italic> and <italic>SWE1),</italic> and 30 &#x003bc;g/ml <italic>(CLB3)</italic> leucine conditions. The data used in the graphs are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st005">Table S5</xref>.</p></caption><graphic xlink:href="pgen.0020111.g007"/></fig></sec><sec id="s2g"><title>Comparison of the Copy Numbers in the gTOW Experiment and a Computational Cell-Cycle Model</title><p>Chen et al. constructed a computational model in which about 30 <italic>CDC</italic> genes in <named-content content-type="genus-species">S. cerevisiae</named-content> regulated CDK activity [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>]. The model reconstitutes 120 phenotypes out of 131 reported phenotypes and predicts quantitative behaviors of the components [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>]. We compared the prediction of upper limit copy number of <italic>CDC</italic> genes in this model with the ones obtained in the gTOW experiment. For the upper limits of <italic>CDC</italic> genes in vivo, we used the plasmid copy numbers in the gTOW experiment plus one (i.e., chromosomal copy). As shown in <xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>B, the model generally had much lower limits than in vivo results. Moreover, the model did not reproduce the system-level characteristics, that components, which directly regulate B-type cyclins (i.e., <italic>CLB2, CLB5,</italic> and <italic>MCM1</italic>), have lower limits. In fact, in the model their upper limits were rather higher than the others, although the absolute values themselves showed rather good agreement between the model prediction and in vivo data (<italic>CLB2, CLB5,</italic> and <italic>MCM1</italic> in <xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>B).</p><p>In the model, genes with the lowest upper limits being less than a 2-fold increase were involved in regulations by stoichiometric interactions; i.e., enzymes and their stoichiometric inhibitors (hereafter we call them &#x0201c;stoichiometric partners&#x0201d;), such as Cdc14 (protein phosphatase) versus Net1 and Esp1 (protease) versus Pds1 (<xref ref-type="supplementary-material" rid="pgen-0020111-sg006">Figure S7</xref>C and <xref ref-type="supplementary-material" rid="pgen-0020111-sg006">S7</xref>D). In the model, their regulations are easily disrupted when any of the stoichiometric partners is in relative excess more than the other, because they are regulated under the balance of 1:1 molecular interactions. Most of them did not have these extreme low limits in vivo (<xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>C), suggesting that there are additional regulations that are not implemented into the model. However, the surprisingly low limit in <italic>CDC14</italic> in vivo was almost perfectly consistent with the model (<xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>C). This probably indicates that Cdc14 activity is only regulated through the stoichiometric 1:1 interaction with Net1 as it is in the model.</p></sec><sec id="s2h"><title>Estimation of the Overexpression of Cdc Proteins in the gTOW Experiment</title><p>In the gTOW experiment, the target gene product is supposed to be overexpressed according to the gene dosage increase. However, if transcription factors are diluted or there is a feedback regulation, it is possible that the gene copy number and the protein expression level are not linearly correlated. We thus tried to measure the Cdc proteins in the cells in the gTOW experiment with quantitative Western blot analysis. First we tried to use epitope-tagged proteins commonly used for protein detections, such as TAP-tag, myc-tag, and HA-tag [<xref rid="pgen-0020111-b026" ref-type="bibr">26</xref>]. However, the copy numbers of pTOW-<italic>target,</italic> containing <italic>CDC</italic> genes with epitope tags, were very different from the natural one without any tag; moreover, there was no general trend depending on the tags and protein species (unpublished data). This is probably because the tags perturbed the natural activities of the Cdc proteins. Thus, it was concluded that tagged proteins are not suitable for the perturbation experiment such as gTOW.</p><p>As an alternative method, we used specific antibodies against the target Cdc proteins provided by Santa Cruz Biotechnology, Incorporated. We used 36 antibodies against 27 Cdc proteins tested in the gTOW experiment in this report (<xref ref-type="table" rid="pgen-0020111-t002">Table 2</xref>). Among them, 12 antibodies could detect endogenous target Cdc proteins in Western blot analysis. The typical examples were Clb2 and Swi6 shown in <xref ref-type="fig" rid="pgen-0020111-g007">Figure 7</xref>A (the others were shown in <xref ref-type="supplementary-material" rid="pgen-0020111-sg006">Figure S6</xref>A). We classified these antibodies and detected proteins as Class I. The overexpression of each Class I target protein in the gTOW experiment was measured as a fold increase over the endogenous target protein (<xref ref-type="supplementary-material" rid="pgen-0020111-st005">Table S5</xref>). The other seven antibodies could only detect the overexpressed target proteins in the gTOW experiment. The typical examples were Cdc28 and Cln2 shown in <xref ref-type="fig" rid="pgen-0020111-g007">Figure 7</xref>B (the others were shown in <xref ref-type="supplementary-material" rid="pgen-0020111-sg006">Figure S6</xref>B). We classified these antibodies and detected proteins as Class II. The overexpression of each Class II target protein in the gTOW experiment was estimated as a least-fold increase over the endogenous target protein according to the serial dilutions of the protein sample (<xref ref-type="supplementary-material" rid="pgen-0020111-st005">Table S5</xref>). Thus, the estimated overexpression of each Class II protein should be less than the real one. The others (17 antibodies) could not detect the target protein in our system. We classified these antibodies as Class III.</p><table-wrap id="pgen-0020111-t002" content-type="2col" position="float"><label>Table 2</label><caption><p>Antibodies Used for the Detection of Cdc Proteins.</p></caption><graphic xlink:href="pgen.0020111.t002"/></table-wrap><p> <xref ref-type="fig" rid="pgen-0020111-g007">Figure 7</xref>C is a scatter plot showing the relationship between the protein overexpression and the gene copy number of the <italic>CDC</italic> genes in the gTOW experiment. The overexpression of Class I protein, except Swi6, and the least overexpression of some Class II proteins (Cdc28, Lte1, Tem1, and Net1) showed a significant correlation with their gene copy numbers (Pearson's correlation coefficient, <italic>r</italic> = 0.94). Although Swi6 is overexpressed in the gTOW experiment (8.48-fold), the level was apparently inconsistent with the gene copy number (149 copies). We thus confirmed that in all cases we tested, the target Cdc proteins were overexpressed in the gTOW experiment, and in most cases the levels were correlated with the their gene copy number.</p></sec></sec><sec id="s3"><title>Discussion</title><p>In this study, we reported a genetic screening method that evaluates the upper limit copy number of target genes. In this method, we used a gene with its native promoter as a unit to evaluate the upper limit copy number to inhibit cellular growth, so that we could quantitatively and directly compare the upper limits among various genes.</p><p>Principally, the gTOW experiment causes overexpression of target genes. In <italic>S. cerevisiae,</italic> the <italic>GAL1</italic> promoter system is the common way to overexpress target genes [<xref rid="pgen-0020111-b027" ref-type="bibr">27</xref>]. We compared the growth inhibition in the gTOW experiment and <italic>GAL1</italic> promoter expression system, but the data did not show complete positive correlations (<xref ref-type="fig" rid="pgen-0020111-g005">Figure 5</xref>C). The differences between the two experimental results are explained by the native expression level and the expression regulation of each target gene. If the native expression level of the target gene is high, the expression from <italic>GAL1</italic> promoter does not cause so much &#x0201c;overexpression,&#x0201d; but increasing the copy number in the gTOW experiment does cause overexpression. For example, <italic>CLB3</italic> and <italic>CLB5</italic> did not show strong growth inhibition in the <italic>GAL</italic> experiment but did in the gTOW experiment (<xref ref-type="fig" rid="pgen-0020111-g005">Figure 5</xref>C); probably because they have higher native expression levels [<xref rid="pgen-0020111-b021" ref-type="bibr">21</xref>] (<xref ref-type="fig" rid="pgen-0020111-g005">Figure 5</xref>D). <italic>CLB6</italic> and <italic>PDS1</italic> are the opposite case [<xref rid="pgen-0020111-b021" ref-type="bibr">21</xref>,<xref rid="pgen-0020111-b026" ref-type="bibr">26</xref>]. In addition, if there are transcriptional and translational regulations, such as periodic expression during cell cycle and feedback regulation within the target gene, the results of both experiments will be different. Recently, Sopko et al. reported a comprehensive overexpression analysis using the <italic>GAL1</italic> promoter [<xref rid="pgen-0020111-b017" ref-type="bibr">17</xref>]. Their results and the results obtained in our <italic>GAL1</italic> experiment showed little accordance in growth inhibition (unpublished data). One of the reasons for this may be the existence of GST-tag in their experiment. In fact, we observed that the copy numbers in the gTOW experiments were perturbed by the commonly used epitope tags (unpublished data). Thus, when determining the quantitative effect of a target protein, a tagged protein is not preferred. Thus, the gTOW experiment enabled systematic evaluation of the upper limit of gene expression, about which little has been known hitherto.</p><p>The copy numbers of the 30 <italic>CDC</italic> genes determined in the gTOW experiment were very diverse, ranging from 1 to more than 100 (<xref ref-type="fig" rid="pgen-0020111-g002">Figure 2</xref>A). The data revealed some interesting properties of the cell division cycle in <italic>S. cerevisiae,</italic> which have been very difficult to clarify. Six out of the top seven genes with the lowest copy numbers constitute a subsystem that regulates B-type CDK activity, which was a core process in the cell cycle with very dynamic properties (<xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>A). We speculate that because the parameter range should be tuned up to be narrow in a system with very dynamic properties, the system should show high fragility against perturbations that change the quantity of parameters as a trade-off of the dynamics. In other words, the gTOW experiment was very effective to reveal dynamic subsystems that are fragile to changes in parameter.</p><p>To our surprise, <italic>CDC14</italic> has a very low upper limit of less than two copies per haploid genome (i.e., just one extra copy other than the chromosomal one). Interestingly, this extreme low limit was almost consistent with that predicted by the computer model developed by Chen et al. (<xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>C), which could be explained by the 1:1 stoichiometric inhibition by Net1 [<xref rid="pgen-0020111-b031" ref-type="bibr">31</xref>,<xref rid="pgen-0020111-b032" ref-type="bibr">32</xref>]. However, other stoichiometric partners, Esp1 and Pds1, did not show such extreme low limits, although the model also predicted very low limits (<xref ref-type="fig" rid="pgen-0020111-g006">Figure 6</xref>C). The discrepancy is probably explained by the fact that Esp1 needs to be recruited into the nucleus by Pds1 for its full function [<xref rid="pgen-0020111-b033" ref-type="bibr">33</xref>], the regulation of which is not yet implemented into the model. If there is such regulatory mechanism in a system, the system should be rather robust even when the stoichiometric balance is perturbed. We do not yet understand why the <italic>CDC14</italic> regulation evolved to be extremely fragile against the amount of change, but it might be a trade-off of some properties of the subsystem that <italic>CDC14</italic> is involved.</p><p>We used the data obtained in the gTOW experiment to evaluate a computer model. Generally, because intracellular biochemical parameters are very difficult to determine, it is difficult to evaluate the parameters in computer models. Since permissible ranges of parameters in a model are the integrative result of the network structure and parameters in the model, the parameter ranges are a very useful measure to evaluate the model's correctness and to suggest the direction to improve it [<xref rid="pgen-0020111-b006" ref-type="bibr">6</xref>]. The models by Chen et al. showed much fragility relative to the gTOW experimental data. We suggest two major issues to be improved in the model: one is stoichiometric partners as mentioned above and the other is paralogous gene pairs. The model implemented only one of each paralogous gene pairs (i.e., <italic>CLB2</italic> of <italic>CLB1</italic>/<italic>CLB2, CLB5</italic> of <italic>CLB5</italic>/<italic>CLB6,</italic> and <italic>CLN2</italic> of <italic>CLN1</italic>/<italic>CLN2</italic>), but each of the paralogous gene pairs had very different copy numbers (i.e., upper limits) between them in the gTOW experiment (<xref ref-type="supplementary-material" rid="pgen-0020111-sg005">Figure S5</xref>). The issue of how paralogous gene pairs are involved in cellular robustness is still being argued [<xref rid="pgen-0020111-b034" ref-type="bibr">34</xref>&#x02013;<xref rid="pgen-0020111-b036" ref-type="bibr">36</xref>] and it will be very interesting to test how robust the model becomes when the paralogous genes exist.</p><p>The gTOW method may also be used for genetic screening of positive growth regulators. Under mild copy number&#x02013;increasing bias (i.e., uracil&#x02212; condition), <italic>MIH1</italic> and <italic>BUB2</italic> had significantly higher copy numbers than the vector (<xref ref-type="fig" rid="pgen-0020111-g004">Figure 4</xref>B). Mih1 is known to dephosphorylate inhibitory phosphorylation of B-type CDK in Tyr-19 residue [<xref rid="pgen-0020111-b023" ref-type="bibr">23</xref>]. In <italic>S. cerevisiae,</italic> phosphorylation in Tyr-19 is involved in the morphological checkpoint [<xref rid="pgen-0020111-b023" ref-type="bibr">23</xref>]. High copy number <italic>MIH1</italic> may cause faster growth ignoring the morphological checkpoint. Interestingly, overexpression of Cdc25, the human homolog of Mih1, is known to be closely related to cancer development [<xref rid="pgen-0020111-b037" ref-type="bibr">37</xref>], part of which might be related to factors in cancer development as in the case of <italic>MIH1</italic>. Positive growth factors, which can be identified by gTOW experiments, under mild bias, potentially contain factors related to cancer development such as <italic>MIH1</italic>.</p><p>We estimated the protein overexpression of about 19 Cdc proteins out of 30 tested in the gTOW experiment and confirmed that most of the Cdc proteins tested were overexpressed with a good agreement with their gene copy numbers (<xref ref-type="fig" rid="pgen-0020111-g007">Figure 7</xref>C). The overexpression of 12 Cdc proteins was appropriately measured with endogenous protein level (referred to as Class I proteins in this study, <xref ref-type="supplementary-material" rid="pgen-0020111-sg006">Figure S6</xref>A and <xref ref-type="supplementary-material" rid="pgen-0020111-st005">Table S5</xref>). These will be good resources for further precise quantitative analysis, such as the protein expression in synchronized cells, or single cell variation in the gTOW experiment. Among Class I proteins, only Swi6 showed apparent discrepancy between protein overexpression level and gene copy number in the gTOW experiment (<xref ref-type="fig" rid="pgen-0020111-g007">Figure 7</xref>C); the mechanism how this occurs is also an interesting future issue. Current epitope tags used for protein detection were not preferable for the perturbation analysis such as gTOW (unpublished data), but specific antibodies were not comprehensive or qualitative enough; we thus need more comprehensive and qualitative technology to detect proteins with lowest perturbations.</p><p>In conclusion, using the gTOW method, we obtained upper limit gene copy numbers, about which little has been known at a system-wide level. This sort of quantitative data represent how intracellular parameters are set up in a certain biological system, and thus represent the robustness and fragility of the system against internal perturbations, which have been very difficult to assess with experimental data. In addition, as shown here, the data can be used to evaluate computer models and to improve them. The gTOW method can also be applied to genome-wide analysis of upper limit gene copy numbers, other than those of <italic>CDC</italic> genes, as well as to profiling of quantitative genetic interactions in mutant strains.</p></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Yeast strains and growth conditions.</title><p>A yeast strain BY4741 <italic>(MAT</italic>a, <italic>his3&#x00394;1, leu2&#x00394;0, met15&#x00394;0,</italic> and <italic>ura3&#x00394;0)</italic> [<xref rid="pgen-0020111-b024" ref-type="bibr">24</xref>] was used in this study. Yeast cells were cultured in SC media [<xref rid="pgen-0020111-b038" ref-type="bibr">38</xref>] supplemented with indicated amino acids (uracil, leucine). 2% glucose was used as a carbon source, except in the <italic>GAL</italic> experiments.</p></sec><sec id="s4b"><title>Plasmid constructions.</title><p>The plasmids constructed and used in this study are listed in <xref ref-type="table" rid="pgen-0020111-t001">Table 1</xref>. PCR was done with KOD-Plus high fidelity DNA polymerase (TOYOBO, Japan) according to the manufacturer's protocol. pSBI40 is a pYEX4T-1 derivative [<xref rid="pgen-0020111-b039" ref-type="bibr">39</xref>] (the full nucleotide sequence is available upon request). pSBI104 used in the <italic>GAL1</italic> promoter-driven overexpression experiments was made by replacing <italic>CUP1</italic> promoter and GST in pSBI40 by gap-repair [<xref rid="pgen-0020111-b040" ref-type="bibr">40</xref>] with PCR fragment containing <italic>GAL1</italic> promoter amplified from genomic DNA using primer OSBI0246 and OSBI0247 (primer sequences are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st006">Table S6</xref>B). For the <italic>gTOW,</italic> according to the <italic>Saccharomyces</italic> Genome Database (SGD) [<xref rid="pgen-0020111-b041" ref-type="bibr">41</xref>], DNA fragments containing target <italic>CDC</italic> genes (listed in <xref ref-type="table" rid="pgen-0020111-t001">Table 1</xref>), with upstream and downstream sequences up to their neighboring genes, were amplified from genomic DNA of BY4741 by PCR using primer sets (&#x0201c;up primer&#x0201d; and &#x0201c;down primer&#x0201d; for each gene, listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st006">Table S6</xref>A), then cloned into pSBI40 by the gap-repair method in BY4741. The constructed plasmids were named pTOW-<italic>target</italic>. For overexpression experiments from the <italic>GAL1</italic> promoter, DNA fragments containing target cell-cycle genes from ATG to their downstream end were amplified from the genomic DNA by PCR using primer sets (&#x0201c;<italic>GAL</italic> primer&#x0201d; and &#x0201c;down primer&#x0201d; for each gene, listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st006">Table S6</xref>A), then cloned into pSBI104 by the gap-repair method in BY4741. The constructed plasmids were named pGAL-<italic>target</italic>. To make frame shift mutation, adenine was inserted just after the ATG codon of each gene except <italic>CDC14</italic> (adenine was inserted into + 23 nt from the ATG codon), <italic>CDH1</italic> (cytosine was inserted +128 nt from the ATG codon) and <italic>PDS1</italic> (adenine was inserted +7 nt from the ATG codon); two PCR fragments amplified from each tug-of-war plasmid construct were amplified using: (1) &#x0201c;frame shift forward primer&#x0201d; for each gene and OSBI159 and (2) &#x0201c;frame shift reverse primer&#x0201d; for each gene and OSBI160; then they were combined and cloned into pSBI40 by gap-repair (primer sequences are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st006">Tables S6</xref>A and <xref ref-type="supplementary-material" rid="pgen-0020111-st006">S6</xref>C). The constructed plasmids were named pTOW-<italic>targetfs</italic>. Plasmids with more than two independent isolates from each plasmid construction were recovered from yeast, and the structure was checked by restriction digestion and partial nucleotide sequencing.</p></sec><sec id="s4c"><title>gTOW procedure.</title><p>BY4741 cells with pTOW-<italic>target</italic> containing each target gene (listed in <xref ref-type="table" rid="pgen-0020111-t001">Table 1</xref>) were grown in SC without uracil, and then they were transferred into SC without leucine and uracil. The optical density of 595 nm of the culture was monitored using a microplate reader (Model X680 Bio-Rad, Hercules, California, United States) at 30 &#x000b0;C without agitation every 30 min for 50 h. Mean doubling time (min) at maximum growth rate (min^&#x02212;1) during the 150-min interval from at least three independent experiments was calculated. For the <italic>GAL</italic> experiment, BY4741 cells with pGAL-<italic>target</italic> containing each target gene (listed in <xref ref-type="table" rid="pgen-0020111-t001">Table 1</xref>) were grown in SC without uracil, before being transferred to SC without uracil with 2% galactose.</p><p>The plasmid dosage in a yeast cell was determined by two kinetic PCRs using total DNA extracted from yeast cells as a template. Yeast cells collected from 200 &#x003bc;l of saturated culture were suspended in lysis solution (10 mM Na-phosphate [pH 7.5], 1.2 M sorbitol, and 2.5 mg/ml Zymolyase 100T) (Seikagaku, Japan) and incubated for 10 min at 37 &#x000b0;C to digest the cell wall. Then the cell suspension was treated at 94 &#x000b0;C for 15 min, at &#x02212;80&#x000b0; C for 5 min, then at 94 &#x000b0;C for 15 min. The cell suspension was chilled and centrifuged. Supernatant (containing total DNA) was used for the following two kinetic PCRs: For the kinetic PCR, we used LightCycler FastStart DNA Master^PLUS SYBER Green I (Roche, Germany) with LightCycler 2.0 instrument (Roche), according to the manufacturer's protocol. Supernatant (2 &#x003bc;l) was mixed with each reaction mix (18 &#x003bc;l) containing 0.5 &#x003bc;M of <italic>LEU2</italic> (<italic>LEU2</italic>-F and <italic>LEU2</italic>-R) and <italic>LEU3</italic> primer sets (<italic>LEU3</italic>-F and <italic>LEU3</italic>-R); primer sequences are listed in <xref ref-type="supplementary-material" rid="pgen-0020111-st006">Table S6</xref>D. <italic>LEU2</italic> and <italic>LEU3</italic> primer sets were used to amplify and quantify <italic>LEU2</italic> genes from the plasmid and <italic>LEU3</italic> gene from the genome, respectively. By comparing the relative quantity of <italic>LEU2</italic> gene and <italic>LEU3</italic> gene, the copy number of the plasmid per haploid genome (BY4741 is a haploid strain) was estimated. The calculation is: plasmid copy number = 2^{(I.P._<italic>LEU3</italic> - I.P_<italic>LEU2</italic>)} <italic>,</italic> where I.P._<italic>LEU3</italic> and I.P._<italic>LEU2</italic> are the PCR cycle numbers at inflection points of the PCR amplification curves of the <italic>LEU3</italic> and <italic>LEU2</italic> genes, respectively. Strictly speaking, the number determined by this procedure is not equivalent to the copy number per cell, because in G₂-M phase cells, there are two copies of genome per cell. Thus, the copy number is per haploid genome. We calculated the gene ratio of wild-type S288C strain (with <italic>LEU2</italic> and <italic>LEU3</italic> on the chromosome) with eight independent experiments (1.07 &#x000b1; 0.16). The host strain BY4741 gave 5 &#x000d7; 10^&#x02212;4, confirming that BY4741 is a yeast strain with <italic>leu2</italic> deletion <italic>(leu2&#x00394;0)</italic> [<xref rid="pgen-0020111-b024" ref-type="bibr">24</xref>]. For each experiment in this study, we determined the copy number of more than three independent isolates from each plasmid construction, and calculated the mean value with standard deviation.</p></sec><sec id="s4d"><title>Quantification of Cdc protein overexpression.</title><p>Cells with vector pSBI40 and pTOW-<italic>target</italic> were cultivated in 2 ml SC without uracil and 2 ml SC without leucine (otherwise stated) for two overnights, then 4 ml of the fresh medium were added to the culture which was then cultivated for 4 more h to refresh the cells. DNA was extracted from the cells and the plasmid copy number was determined. Protein extraction was performed as described [<xref rid="pgen-0020111-b042" ref-type="bibr">42</xref>]. Briefly, cells from each 2-ml culture were collected and suspended in 400 &#x003bc;l of 0.2 N NaOH; after 5 min of incubation at room temperature, cells were collected and re-suspended in 100 &#x003bc;l of SDS-sample buffer [<xref rid="pgen-0020111-b042" ref-type="bibr">42</xref>], then heated at 100 &#x000b0;C for 5 min. Cell debris was removed by centrifugation, then the supernatant with indicated dilution was separated in 10% SDS-PAGE for the standard Western blot procedure.</p><p>In immunoblotting, each target Cdc protein was detected using its specific antibody (listed in <xref ref-type="table" rid="pgen-0020111-t002">Table 2</xref>, obtained from Santa Cruz, California, United States) with 1/200 dilution, and anti-Goat IgA peroxydase conjugate (A5420, Sigma, St. Louis, Missouri, United States) with 1/5000 dilution. Hexokinase was detected using anti-hexokinase antibody (100-4159, Sigma) with 1/1000 dilution, and anti-rabbit IgG, IgM, IgA HRP conjugate (SAB1003, Open Biosystems, United States) with 1/5000 dilution. Detection of immune complex was done with SuperSignal West Dura Extended Duration Substrate (34075, Pierce Biotechnology, Rockford, Illinois, United States). Intensities of corresponding protein bands were measured using an LAS-3000 mini image analyzer (Fuji Film, Japan), and the data within the linear detection range among the serial dilution of the samples were used.</p><p>If the endogenous Cdc protein was detected (we classified the antibody and the detected protein as Class I), the Cdc protein overexpression was quantified as the fold increase of the target protein over the endogenous protein amount (i.e., vector alone), which was normalized using hexokinase as a standard of total protein amount. The calculations are: Cdc overexpression in uracil&#x02212; = <italic>(Cdc_{target_ura}</italic> / <italic>Cdc_{vec_ura})</italic> / <italic>(Hxk_{target_ura}</italic> / <italic>Hxk_{vec_ura}),</italic> Cdc overexpression in leucine&#x02212; = <italic>(Cdc_{target_leu}</italic> / <italic>Cdc_{vec_ura})</italic> / <italic>(Hxk_{target_leu}</italic> / <italic>Hxk_{vec_ura}),</italic> where <italic>Cdc</italic> and <italic>Hxk</italic> mean the intensity of the target protein and hexokinase bands in the Western blot of the samples from indicated plasmid and culture conditions described in subscript (i.e., <italic>target;</italic> pTOW-target, <italic>vec:</italic> vector, <italic>ura:</italic> uracil&#x02212; condition, <italic>leu:</italic> leucine&#x02212; condition). If the endogenous Cdc protein was not detected, but the protein overexpressed from the pTOW-<italic>target</italic> was detected (we classified the antibody and the detected protein as Class II), we estimated the least-fold increase of the target protein using the number of the highest dilution in which the Cdc protein was detectable which was normalized using hexokinase. The calculations are: Least Cdc protein overexpression in uracil&#x02212; = <italic>MD</italic> / <italic>(Hxk_{target_ura}</italic> / <italic>Hxk_{vec_ura}),</italic> Least Cdc protein overexpression in leu2&#x02212; = <italic>MD</italic> &#x000d7; <italic>(Cdc_{target_leu}</italic> / <italic>Cdc_{target_ura})</italic> / <italic>(Hxk_{target_leu}</italic> / <italic>Hxk_{vec_ura}),</italic> where <italic>MD</italic> is the maximum dilution number of the sample from the <italic>target_ura</italic> condition, in which the target Cdc protein band in Western blot could be detectable. At least two independent experiments were done for each target protein, and representative data were shown.</p></sec><sec id="s4e"><title>Computation.</title><p>Computational prediction that systematically searches upper limit parameter values in the cell-cycle model has been described [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>]. We used the Matlab code of the model with an algorithm that surveys and records the maximum fold increase of a certain parameter to maintain the normal cell cycle up to 256-fold [<xref rid="pgen-0020111-b019" ref-type="bibr">19</xref>]. To emulate the increase of the copy number of a target gene, we used the following way: If a target gene has only a single parameter for the gene expression, we increased it. If a target gene has two or more parameters for the gene expression, we increased them all together. If a target gene does not have a gene expression parameter but has the protein amount itself, then we simply increased the amount. Parameters tested are shown in <xref ref-type="supplementary-material" rid="pgen-0020111-st004">Table S4</xref>. Some gene groups described in the model are listed as one gene because of redundant function, i.e., <italic>CLB1</italic>/<italic>CLB2, CLB5</italic>/<italic>CLB6, CLN1</italic>/<italic>CLN2,</italic> and <italic>MBP1</italic>/<italic>SWI4</italic>/<italic>SWI6;</italic> whereas we analyzed these groups as independent genes (<xref ref-type="supplementary-material" rid="pgen-0020111-st004">Table S4</xref>). Computer simulations were done using Matlab version 7.0.4. The Matlab codes are provided in Text <xref ref-type="supplementary-material" rid="pgen-0020111-sd001">S1</xref>&#x02013;<xref ref-type="supplementary-material" rid="pgen-0020111-sd003">S3</xref>.</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting Information</title><supplementary-material content-type="local-data" id="pgen-0020111-sg001"><label>Figure S1</label><caption><title>A Possible Mechanism for the Linearity of the Growth Rate Copy Number Correlation in the gTOW Experiment</title><p>(41 KB PDF)</p></caption><media xlink:href="pgen.0020111.sg001.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sg002"><label>Figure S2</label><caption><title>A Scatter Plot between the Plasmid Copy Number in Uracil and Leucine Conditions in the gTOW Experiment</title><p>(16 KB PDF)</p></caption><media xlink:href="pgen.0020111.sg002.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sg003"><label>Figure S3</label><caption><title>A Scatter Plot between Protein Abundance and the Plasmid Copy Number in the gTOW Experiment</title><p>(12 KB PDF)</p></caption><media xlink:href="pgen.0020111.sg003.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sg004"><label>Figure S4</label><caption><title>Molecular Interaction Map of <italic>CDC</italic> Genes</title><p>(181 KB PDF)</p></caption><media xlink:href="pgen.0020111.sg004.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sg005"><label>Figure S5</label><caption><title>Copy Number of the Plasmid with Paralogous Gene Pairs Determined in the gTOW Experiment</title><p>(13 KB PDF)</p></caption><media xlink:href="pgen.0020111.sg005.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sg006"><label>Figure S6</label><caption><title>Quantification of the Protein Level Expressed in the gTOW Experiment</title><p>(68 KB PDF)</p></caption><media xlink:href="pgen.0020111.sg006.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-st001"><label>Table S1</label><caption><title>Doubling Time Determined in <italic>gTOW</italic> Experiments</title><p>(19 KB PDF)</p></caption><media xlink:href="pgen.0020111.st001.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-st002"><label>Table S2</label><caption><title>Plasmid Copy Number Determined in <italic>gTOW</italic> Experiments</title><p>(18 KB PDF)</p></caption><media xlink:href="pgen.0020111.st002.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-st003"><label>Table S3</label><caption><title>Doubling Time of Yeast Strains with <italic>GAL1-CDCgene</italic> Plasmids</title><p>(8 KB PDF)</p></caption><media xlink:href="pgen.0020111.st003.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-st004"><label>Table S4</label><caption><title>Parameters Increased In Silico Upper Limit Gene Dosage Search</title><p>(8 KB PDF)</p></caption><media xlink:href="pgen.0020111.st004.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-st005"><label>Table S5</label><caption><title>Quantification of Proteins Overexpressed in the gTOW</title><p>(7 KB PDF)</p></caption><media xlink:href="pgen.0020111.st005.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-st006"><label>Table S6</label><caption><title>Primers Used to Construct Plasmids Containing Cell-Cycle Related Genes</title><p>(17 KB PDF)</p></caption><media xlink:href="pgen.0020111.st006.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sd001"><label>Text S1</label><caption><title>Instruction for the Matlab Scripts</title><p>(27 KB DOC)</p></caption><media xlink:href="pgen.0020111.sd001.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sd002"><label>Text S2</label><caption><title>Matlab Script (1/2) for Parameter Analysis</title><p>(8 KB DOC)</p></caption><media xlink:href="pgen.0020111.sd002.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pgen-0020111-sd003"><label>Text S3</label><caption><title>Matlab Script (2/2) for Parameter Analysis</title><p>(32 KB DOC)</p></caption><media xlink:href="pgen.0020111.sd003.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material></sec>

<italic>Drosophila</italic> CENP-A Mutations Cause a BubR1- Dependent Early Mitotic Delay without Normal Localization of Kinetochore Components

<p>The centromere/kinetochore complex plays an essential role in cell and organismal viability by ensuring chromosome movements during mitosis and meiosis. The kinetochore also mediates the spindle attachment checkpoint (SAC), which delays anaphase initiation until all chromosomes have achieved bipolar attachment of kinetochores to the mitotic spindle. CENP-A proteins are centromere-specific chromatin components that provide both a structural and a functional foundation for kinetochore formation. Here we show that cells in <italic>Drosophila</italic> embryos homozygous for null mutations in CENP-A (CID) display an early mitotic delay. This mitotic delay is not suppressed by inactivation of the DNA damage checkpoint and is unlikely to be the result of DNA damage. Surprisingly, mutation of the SAC component BUBR1 partially suppresses this mitotic delay. Furthermore, <italic>cid</italic> mutants retain an intact SAC response to spindle disruption despite the inability of many kinetochore proteins, including SAC components, to target to kinetochores. We propose that SAC components are able to monitor spindle assembly and inhibit cell cycle progression in the absence of sustained kinetochore localization.</p>

<contrib contrib-type="author"><name><surname>Blower</surname><given-names>Michael D</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Daigle</surname><given-names>Tanya</given-names></name><xref ref-type="aff" rid="aff2">2</xref></contrib><contrib contrib-type="author"><name><surname>Kaufman</surname><given-names>Thom</given-names></name><xref ref-type="aff" rid="aff3">3</xref></contrib><contrib contrib-type="author"><name><surname>Karpen</surname><given-names>Gary H</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff4">4</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib>

PLoS Genetics

<sec id="s1"><title>Introduction</title><p>Proper kinetochore assembly and function is essential for the faithful transmission of chromosomes during both mitosis and meiosis. One critical function of the kinetochore is to serve as the site of the mitotic spindle attachment checkpoint (SAC), which monitors kinetochore microtubule attachment prior to anaphase onset [<xref rid="pgen-0020110-b001" ref-type="bibr">1</xref>,<xref rid="pgen-0020110-b002" ref-type="bibr">2</xref>]. It is hypothesized that the SAC monitors the attachment of kinetochore microtubules during prometaphase and metaphase, and inhibits anaphase progression until all chromosomes have achieved bipolar spindle attachment. The current model for SAC function suggests that unattached kinetochores recruit checkpoint proteins (such as components of the MAD, BUB, and ZW10/ROD protein complexes), and that these proteins are modified by unattached kinetochores to generate a diffusible signal that delays the onset of anaphase. It has recently been demonstrated that defects in SAC function result in organismal lethality and dominant haploinsufficiency defects. Haploinsufficiency for mouse MAD2, BUB3, or RAE1 results in elevated rates of chromosome missegregation, defects in SAC function, and a predisposition to cancer [<xref rid="pgen-0020110-b003" ref-type="bibr">3</xref>&#x02013;<xref rid="pgen-0020110-b005" ref-type="bibr">5</xref>], demonstrating the fundamental importance of this checkpoint.</p><p>The role of kinetochore localization of SAC components in the generation of the anaphase delay [<xref rid="pgen-0020110-b006" ref-type="bibr">6</xref>] has come into question recently with the published disruption of several inner kinetochore proteins. Disruption of Nuf2 or Ndc80/Hec1 in human cells results in a mitotic arrest, despite the fact that several outer kinetochore components (including Mad2) are unable to sustain kinetochore localization [<xref rid="pgen-0020110-b007" ref-type="bibr">7</xref>,<xref rid="pgen-0020110-b008" ref-type="bibr">8</xref>]. Similarly, disruption of human or chicken CENP-H or CENP-I (the homolog of <named-content content-type="genus-species">Schizosaccharomyces pombe</named-content> Mis6) arrests cells in mitosis for hours, despite the mislocalization of a variety of outer kinetochore proteins, including some SAC components [<xref rid="pgen-0020110-b009" ref-type="bibr">9</xref>&#x02013;<xref rid="pgen-0020110-b011" ref-type="bibr">11</xref>]. In contrast, disruptions of components of the SAC result in premature entry into anaphase, rather than mitotic arrest or delay. These results suggest that continuous kinetochore localization of SAC components may not be necessary to signal the cell to delay anaphase onset. However, incomplete depletion of the inner kinetochore proteins could produce partially functional kinetochores in which some outer proteins localized properly, while others were mislocalized [<xref rid="pgen-0020110-b008" ref-type="bibr">8</xref>,<xref rid="pgen-0020110-b011" ref-type="bibr">11</xref>,<xref rid="pgen-0020110-b012" ref-type="bibr">12</xref>]. A partially dysfunctional kinetochore could recruit sufficient amounts of SAC proteins to generate an effective checkpoint signal, consistent with a requirement for SAC component recruitment to kinetochores.</p><p>A key component of the inner kinetochore is the centromere-specific histone H3-like protein CENP-A [<xref rid="pgen-0020110-b013" ref-type="bibr">13</xref>]. Several recent studies have demonstrated that CENP-A proteins are present in all eukaryotes, and that these proteins are essential for both cell and organismal viability [<xref rid="pgen-0020110-b014" ref-type="bibr">14</xref>&#x02013;<xref rid="pgen-0020110-b020" ref-type="bibr">20</xref>]. CENP-A proteins replace both copies of histone H3 in centromeric nucleosomes, and physically and genetically interact with all other core histones [<xref rid="pgen-0020110-b014" ref-type="bibr">14</xref>,<xref rid="pgen-0020110-b021" ref-type="bibr">21</xref>&#x02013;<xref rid="pgen-0020110-b023" ref-type="bibr">23</xref>]. CENP-A proteins are at or near the top of the kinetochore assembly pathway, and are required for the localization of nearly all other kinetochore proteins examined to date, including all tested SAC components [<xref rid="pgen-0020110-b013" ref-type="bibr">13</xref>]. Therefore, cells lacking CENP-A would be expected to contain chromosomes with severely compromised kinetochores, which would be incapable of generating the SAC signal.</p><p>Here we report that mutations in the <italic>Drosophila</italic> CENP-A family member (CID, for &#x0201c;Centromere Identifier&#x0201d;) [<xref rid="pgen-0020110-b024" ref-type="bibr">24</xref>] result in an early mitotic delay. Furthermore, <italic>cid</italic> mutants have an intact SAC response to microtubule disruption despite the absence of kinetochore localization of SAC components ROD and BUBR1. We present data that suggest that the DNA damage/repair checkpoint is not responsible for the CID-mediated early mitotic delay. In contrast, the mitotic delay of <italic>cid</italic> mutants is partially suppressed by mutation of the SAC component <italic>bubr1.</italic> We discuss models for the role of SAC proteins in monitoring aspects of kinetochore assembly early in mitosis.</p></sec><sec id="s2"><title>Results</title><sec id="s2a"><title>CID Null Mutants Display Embryonic Lethality</title><p>In a previous study, we found that anti-CID antibody injections into syncitial embryos resulted in phenotypes expected for kinetochore disruption (failure to congress in prometaphase, metaphase arrest, and anaphase segregation defects), but also produced unusual phenotypes (interphase and prophase arrests). However, it was unclear if these phenotypes were the consequence of loss of CID function, an artifact of antibody binding to CID, or a consequence of the specialized nature of the syncitial nuclear divisions. Therefore, we examined the phenotypic consequences of <italic>cid</italic> null mutations in <italic>Drosophila</italic> embryos. The alleles examined were T11&#x02013;2 (Q51 to stop), T12&#x02013;1 (Q83 to stop), T21&#x02013;3 (Q94 to stop), and T22&#x02013;4 (Q102 to stop) (J. Cecil and T. Kaufman, unpublished observations). All of these alleles are lethal when homozygous, in <italic>trans</italic>-heterozygous combinations, and over a deficiency for the region (unpublished data); thus, the <italic>cid</italic> gene is essential for <italic>Drosophila</italic> development.</p><p>To examine the phenotypic consequences of <italic>cid</italic> disruption, crosses were made between parents heterozygous for a <italic>cid</italic> mutation and a balancer that contained an ElaV-LacZ fusion construct, which is expressed in the developing nervous system [<xref rid="pgen-0020110-b025" ref-type="bibr">25</xref>]. We collected embryos from these crosses and stained them for CID, LacZ, histone H3 phosphorylation at serine 10 (PH3) [<xref rid="pgen-0020110-b026" ref-type="bibr">26</xref>], and DNA (DAPI). <italic>cid</italic> null and heterozygous embryos were unambiguously distinguished by the absence or presence (respectively) of ElaV-LacZ expression. <italic>cid</italic> null mutant embryos died around stage 15 of embryogenesis, and displayed a phenotypic series that correlated with the temporal disappearance of maternal CID protein and the absence of newly synthesized zygotic protein. At embryonic stages 9&#x02013;10, <italic>cid</italic> null embryos displayed lagging chromosomes during anaphase and unresolved chromatin bridges during telophase, which were not observed in heterozygous controls (<xref ref-type="fig" rid="pgen-0020110-g001">Figure 1</xref>A and <xref ref-type="fig" rid="pgen-0020110-g001">1</xref>B). These phenotypes are the result of partial loss of CID protein; staining with anti-CID antibodies demonstrated that maternally derived CID is still present in stage 9&#x02013;10 embryos, albeit at reduced levels (<xref ref-type="fig" rid="pgen-0020110-g001">Figure 1</xref>A and <xref ref-type="fig" rid="pgen-0020110-g001">1</xref>B). Lagging chromosomes and chromatin bridges are entirely consistent with the phenotypes we observed after partial disruption of CID by RNAi or antibody injection [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>,<xref rid="pgen-0020110-b027" ref-type="bibr">27</xref>].</p><fig id="pgen-0020110-g001" position="float"><label>Figure 1</label><caption><title> <italic>cid</italic> Null Embryos Exhibit Multiple Mitotic Phenotypes.</title><p>CID, PH3, and DAPI staining of <italic>cid/CyO</italic> and <italic>cid/cid</italic> embryos at different stages of development were monitored to evaluate mitotic progression and segregation defects.</p><p>(A) Heterozygous <italic>(cid/CyO)</italic> stage 9&#x02013;10 embryos displayed no mitotic defects and robust CID staining at kinetochores (inset).</p><p>(B) <italic>cid</italic> null animals (<italic>trans</italic>-heterozygous for different <italic>cid</italic> alleles, see <xref ref-type="sec" rid="s4">Materials and Methods</xref>) exhibited lagging chromosomes during anaphase. Some CID staining was still visible at this stage, demonstrating that these phenotypes resulted from partial loss of CID function, due to the presence of maternal CID protein.</p><p>(C) <italic>cid</italic>/<italic>CyO</italic> stage 14&#x02013;15 embryos show normal mitotic progression and normal CID staining at kinetochores (inset).</p><p>(D) <italic>cid</italic> null animals exhibited an elevated mitotic index, lower nuclear density, and little detectable CID staining in some cells at stage 14&#x02013;15. The strong depletion of CID staining suggests that these phenotypes are the result of complete loss of zygotic <italic>cid</italic> function. <italic>cid</italic> null animals have a large number of presumably polyploidy cells (inset) suggesting high levels of aneuploidy due to repeated failures in cell division. Scale bar indicates 15 &#x003bc;m.</p></caption><graphic xlink:href="pgen.0020110.g001"/></fig><p>In later stages of development (stages 13&#x02013;15), high levels of CID staining were observed in heterozygous siblings, whereas in <italic>cid</italic> homozygous mutant embryos, most cells in mitotically active tissues had no visible CID signal (<xref ref-type="fig" rid="pgen-0020110-g001">Figure 1</xref>C and <xref ref-type="fig" rid="pgen-0020110-g001">1</xref>D). The PROD protein binds a satellite DNA near the Chromosomes 2 and 3 centromeres [<xref rid="pgen-0020110-b028" ref-type="bibr">28</xref>], and its localization is not dependent on the presence of CID [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>]. Comparison of the levels of CID and PROD staining in homozygous mutant and heterozygous control embryos suggest that approximately 90%&#x02013;100% of CID was depleted in stage 15 <italic>cid</italic> mutants (<xref ref-type="supplementary-material" rid="pgen-0020110-sg001">Figure S1</xref>, see <xref ref-type="sec" rid="s4">Materials and Methods</xref>). Thus, some cells retain small amounts of maternal CID, and these alleles behave as nulls with respect to functional zygotic protein, as predicted from the early stop codons present in the mutations.</p><p>Homozygous <italic>cid</italic> null embryos displayed few defects associated with gross morphological patterning or development (unpublished data). However, defects were associated with the organization of the developing nervous tissue, consistent with the fact that few other cell types are actively dividing in later stage embryos, and that the most severe defects are also associated with the nervous tissue in other mitotic mutants that die during embryogenesis [<xref rid="pgen-0020110-b029" ref-type="bibr">29</xref>]. These later stage, terminal embryos displayed a high degree of disorganization of the developing nervous tissue, with obvious micronuclei, large presumably polyploid nuclei, very few true metaphase plates, and few anaphases and telophases. The overall nuclear density was much lower in the <italic>cid</italic> mutants (~1/2 of heterozygous controls), which is also consistent with the aneuploidy that results from failures in chromosome segregation and cell division. These phenotypic characteristics are very similar to a recently described mutation in <italic>Drosophila</italic> CENP-C [<xref rid="pgen-0020110-b030" ref-type="bibr">30</xref>], suggesting that these defects result from disruption of the inner kinetochore.</p></sec><sec id="s2b"><title>CID Disruption Results in an Early Mitotic Delay</title><p>The appearance of H3 phospho-serine 10 (PH3), destruction of the mitotic cyclins, and mitotic spindle morphology can be used to discriminate different stages of G2 and mitosis (<xref ref-type="fig" rid="pgen-0020110-g002">Figure 2</xref>A). Cyclins A and B begin to accumulate in S and G2 phases [<xref rid="pgen-0020110-b031" ref-type="bibr">31</xref>], and PH3 begins to appear in late G2, and is used as a general marker for mitotic index [<xref rid="pgen-0020110-b032" ref-type="bibr">32</xref>]. Subsequently, cyclin A destruction is observed during prometaphase, cyclin B destruction occurs at the metaphase to anaphase (M:A) transition, and PH3 staining is gradually lost from chromosomes at the end of telophase.</p><fig id="pgen-0020110-g002" position="float"><label>Figure 2</label><caption><title> <italic>cid</italic> Null Mutants Exhibit a G2/Prophase Delay.</title><p>Cell cycle progression was monitored in <italic>cid/CyO</italic> and <italic>cid/cid</italic> embryos by staining for PH3, cyclin A, cyclin B, and tubulin.</p><p>(A) Schematic diagram of the appearance and destruction of various cell cycle regulatory factors and markers. ANA, anaphase; META, metaphase; PRO, prophase; PRO-META, prometaphase; TELO, telophase.</p><p>(B) <italic>cid/cid</italic> animals displayed an elevated mitotic index, and an increased number of cells in prophase and prometaphase, compared to <italic>cid/CyO</italic> controls. A, anaphase; M, metaphase; MI, mitotic index; P, prophase; PM, prometaphase; T, telophase.</p><p>(C) <italic>cid/cid</italic> animals had a 2-fold higher number of cyclin A and B positive cells than <italic>cid/cid</italic> controls. Scale bars indicate 15 &#x003bc;m.</p></caption><graphic xlink:href="pgen.0020110.g002"/></fig><p>To determine the effects of <italic>cid</italic> depletion on cell cycle progression, the number of cells in specific stages of mitosis was determined by staining homozygous and heterozygous mutant stage 15 embryos for tubulin, cyclin A, cyclin B, and PH3. Three observations demonstrated that <italic>cid</italic> mutants were delayed early in mitosis, predominantly in prophase/prometaphase. First, <italic>cid</italic> mutants displayed a 2.4-fold higher mitotic index (<italic>p</italic> &#x0003c; 0.01) and a 2-fold higher number of cells positive for cyclin A (<italic>p</italic> &#x02264; 0.01) and cyclin B (<italic>p</italic> &#x02264; 0.01), in comparison to heterozygous control siblings (<xref ref-type="fig" rid="pgen-0020110-g002">Figure 2</xref>B and <xref ref-type="fig" rid="pgen-0020110-g002">2</xref>C). Second, <italic>cid</italic> mutants showed a marked increase in the number of cells in prophase and prometaphase, as judged by chromosome and spindle morphology (<xref ref-type="fig" rid="pgen-0020110-g002">Figure 2</xref>B). Third, very few cells progressed to anaphase in <italic>cid</italic> mutants, suggesting that <italic>cid</italic> mutants were delayed prior to the metaphase&#x02013;anaphase transition (<xref ref-type="fig" rid="pgen-0020110-g002">Figure 2</xref>B). Thus, complete depletion of CID in embryos results in a mitotic delay, predominantly in prophase and prometaphase.</p></sec><sec id="s2c"><title>Inactivation of the DNA Damage Checkpoint Does Not Abrogate the <italic>cid-</italic>Mediated Mitotic Delay</title><p>The mitotic delay observed in homozygous <italic>cid</italic> mutant embryos suggested that <italic>cid</italic> depletion and failure to form a kinetochore activated a cell cycle checkpoint. A recent study in <italic>Xenopus</italic> suggested that DNA damage and repair may be involved in CENP-A assembly at centromeres [<xref rid="pgen-0020110-b033" ref-type="bibr">33</xref>]. Therefore, incomplete kinetochore chromatin assembly or chromosome segregation errors caused by <italic>cid</italic> mutation could result in DNA damage and activation of the DNA damage checkpoint, which would mediate the early mitotic cell cycle delay. To determine whether DNA damage phenocopies the <italic>cid</italic> null mutations, we treated <italic>cid</italic> mutant and heterozygous embryos with doxorubicin, a topoisomerase II inhibitor known to generate dsDNA breaks [<xref rid="pgen-0020110-b034" ref-type="bibr">34</xref>]. We found that doxorubicin treatment dramatically decreased the mitotic index of <italic>cid</italic> heterozygous embryos (<xref ref-type="fig" rid="pgen-0020110-g003">Figure 3</xref>A&#x02013;<xref ref-type="fig" rid="pgen-0020110-g003">3</xref>C), consistent with previous studies of the effects of DNA damage on cell cycle progression [<xref rid="pgen-0020110-b035" ref-type="bibr">35</xref>]. Doxorubicin treatment had little effect on the mitotic index of <italic>cid</italic> homozygous mutant embryos, which likely reflects the fact that these cells were already delayed in mitosis at the time of drug addition. Importantly, the decreased mitotic index in <italic>cid</italic> heterozygotes demonstrates that a DNA damage-induced cell cycle delay results in a fundamentally different (opposite) phenotype from the increased mitotic index observed in untreated <italic>cid</italic> null embryos.</p><fig id="pgen-0020110-g003" position="float"><label>Figure 3</label><caption><title>DNA Damage Is Not Responsible for the <italic>cid-</italic>Mediated Mitotic Delay</title><p>The effect of DNA damage on cell cycle progression was determined by treating stage 15 <italic>cid</italic> null and heterozygous embryos with the topoisomerase II inhibitor doxorubicin.</p><p>(A&#x02013;C) <italic>cid/CyO</italic> cells dramatically decreased entry into mitosis in response to DNA damage, whereas <italic>cid</italic>/<italic>cid</italic> cells were unaffected by doxorubicin (Dox.) treatment.</p><p>(D&#x02013;F) The MEI-41/ATR kinase was inhibited by treating <italic>cid</italic> homozygous and heterozygous embryos with 2 mM caffeine. Inactivation of the DNA damage checkpoint did not suppress the <italic>cid-</italic>mediated mitotic delay, as the mitotic index of <italic>cid</italic> mutants remained double that of controls, with the majority of the mitotic cells delayed in prophase or prometaphase.</p><p>A, anaphase; M, metaphase; MI, mitotic index; P, prophase; PM, prometaphase; T, telophase.</p><p>Scale bars indicate 15 &#x003bc;m.</p></caption><graphic xlink:href="pgen.0020110.g003"/></fig><p>We also performed the reciprocal experiment, to determine if an intact DNA damage checkpoint was necessary for the <italic>cid</italic>-mediated early mitotic delay. A central component of the DNA damage response, the MEI-41 ATR kinase [<xref rid="pgen-0020110-b035" ref-type="bibr">35</xref>], was inhibited by treating <italic>cid</italic> mutant and heterozygous embryos with 2 mM caffeine [<xref rid="pgen-0020110-b034" ref-type="bibr">34</xref>]. We found that caffeine treatment of <italic>Drosophila</italic> embryos phenocopied <italic>mei-41</italic> and <italic>grapes</italic> maternal affect mutations, and is likely to completely inactivate the DNA damage checkpoint (unpublished data). Inactivation of MEI-41 by caffeine treatment did not suppress the <italic>cid-</italic>mediated mitotic delay. The mitotic index of <italic>cid</italic> mutants remained nearly twice that of heterozygous controls, and most of the mitotic cells were found in prophase or prometaphase, with very few cells progressing to later stages of mitosis (<xref ref-type="fig" rid="pgen-0020110-g003">Figure 3</xref>D&#x02013;<xref ref-type="fig" rid="pgen-0020110-g003">3</xref>F). These results demonstrate that inactivation of the DNA damage checkpoint does not abrogate the <italic>cid</italic>-mediated mitotic delay, and confirms that this delay is not the result of DNA damage induced by <italic>cid</italic> mutations.</p></sec><sec id="s2d"><title> <italic>cid</italic> Mutant Cells Have an Intact SAC Response to Microtubule Disruption</title><p>The other major cell cycle checkpoint that could be responsible for the cell cycle delay observed in <italic>cid</italic> mutants is the SAC, which monitors kinetochore microtubule attachment and regulates the metaphase to anaphase transition. However, the <italic>cid</italic>-mediated mitotic delay appears temporally and phenotypically distinct from SAC-mediated cell cycle effects. When the SAC is activated (e.g., by the addition of microtubule polymerization inhibitors), cyclin A is degraded, but not cyclin B [<xref rid="pgen-0020110-b031" ref-type="bibr">31</xref>,<xref rid="pgen-0020110-b036" ref-type="bibr">36</xref>] (compare to <xref ref-type="fig" rid="pgen-0020110-g002">Figure 2</xref>), and cells arrest in prometaphase/metaphase with condensed but unaligned chromosomes. Finally, the absence of normal kinetochore formation in all cases in which CENP-A proteins have been depleted or mutated [<xref rid="pgen-0020110-b015" ref-type="bibr">15</xref>&#x02013;<xref rid="pgen-0020110-b020" ref-type="bibr">20</xref>,<xref rid="pgen-0020110-b037" ref-type="bibr">37</xref>] suggests that <italic>cid</italic> null mutant cells should not have an intact SAC, and that SAC components should not play a role in the <italic>cid</italic>-mediated mitotic delay.</p><p>To directly test for the presence of an SAC response to microtubule disruption in <italic>cid</italic> mutant animals, we treated stage 15 <italic>cid</italic> homozygous and heterozygous animals with the microtubule depolymerizing agent colcemid. We found that both homozygous and heterozygous cells were delayed in response to colcemid treatment (<xref ref-type="fig" rid="pgen-0020110-g004">Figure 4</xref>), consistent with the observation that the <named-content content-type="genus-species">Saccharomyces cerevisiae</named-content> CENP-A homolog Cse4 is not required for SAC function [<xref rid="pgen-0020110-b038" ref-type="bibr">38</xref>]. First, both genotypes displayed a nearly 2-fold increase in mitotic index after 1 h of treatment (<xref ref-type="fig" rid="pgen-0020110-g004">Figure 4</xref>B). Second, the increased mitotic index was accompanied by a large increase in the number of cells accumulated in prometaphase in both <italic>cid</italic> null and heterozygous animals. We conclude that <italic>cid</italic> mutant cells retain an intact SAC response to microtubule disruption; thus, SAC components could play a role in the <italic>cid</italic>-mediated early mitotic delay. In addition, the fact that some <italic>cid</italic>/<italic>cid</italic> cells accumulated in prometaphase/metaphase after colcemid treatment indicates that cells can eventually overcome the prophase delay, and that prometaphase is likely to be the terminal arrest point, similar to an SAC-mediated cell cycle arrest.</p><fig id="pgen-0020110-g004" position="float"><label>Figure 4</label><caption><title> <italic>cid</italic> Mutants Retain an Intact SAC Response to Microtubule Depolymerization</title><p> <italic>cid</italic> null and heterozygous embryos were treated with colcemid to determine if they have an intact SAC response to spindle disruption. Both <italic>cid/CyO</italic> and <italic>cid/cid</italic> cells were able to delay the cell cycle in response to spindle disruption (A), as evidenced by an approximately 2-fold increase in mitotic index and an accumulation of cells in prometaphase (B).</p><p>A, anaphase; M, metaphase; MI, mitotic index; P, prophase; PM, prometaphase; T, telophase.</p><p>Scale bars indicate 15 &#x003bc;m.</p></caption><graphic xlink:href="pgen.0020110.g004"/></fig></sec><sec id="s2e"><title>Mutation of the SAC Component <italic>bubr1</italic> Partially Suppresses the <italic>cid-</italic>Mediated Mitotic Delay</title><p>To directly examine the role of the SAC in the <italic>cid</italic>-mediated mitotic delay, we determined if a mutation that inactivates the SAC can restore normal cell cycle progression. <italic>cid bubr1</italic> double mutants were generated, and homozygous and heterozygous double mutant embryos were monitored for cell cycle progression by staining for PH3. Surprisingly, <italic>bubr1</italic> mutations partially suppressed most of the cell cycle phenotypes associated with <italic>cid</italic> mutation (<xref ref-type="fig" rid="pgen-0020110-g005">Figure 5</xref>A and <xref ref-type="fig" rid="pgen-0020110-g005">5</xref>B). The mitotic index in <italic>cid bubr1</italic> double mutants was nearly the same as observed in heterozygous controls (1.1-fold, <italic>p</italic> = 0.6, <italic>cid bubr1/cid bubr1</italic> compared to <italic>cid bubr1/CyO</italic>), compared to the 2.4-fold increase observed for <italic>cid/cid</italic> mutants over controls (see above). The number of cells delayed in prophase and prometaphase also decreased dramatically in <italic>cid bubr1</italic> double mutants and was comparable to heterozygous controls, whereas the number of cells in anaphase showed a corresponding increase and was greater that controls. Note that <italic>cid bubr1</italic> double mutants had a mitotic index nearly double that of <italic>cid</italic> single mutants, for reasons that are unclear at this time. We eliminated bias that could arise from this difference by only comparing homozygous double mutants to heterozygous double mutants (see <xref ref-type="sec" rid="s4">Materials and Methods</xref> for a more detailed discussion). We conclude that inactivation of a component of the SAC relieves the <italic>cid</italic>-mediated mitotic delay, suggesting that at least one component of the SAC is involved in delaying cell cycle progression in the absence of CID.</p><fig id="pgen-0020110-g005" position="float"><label>Figure 5</label><caption><title>A <italic>bubr1</italic> Mutation Partially Suppresses the <italic>cid</italic>-Mediated Mitotic Delay</title><p>(A) <italic>cid bubr1</italic> double mutants were examined for mitotic progression by staining for PH3 and DAPI. <italic>cid bubr1</italic> mutants show an increased nuclear density and number of anaphases compared to <italic>cid</italic> single mutants. Scale bar indicates 15 &#x003bc;m.</p><p>(B) <italic>bubr1</italic> suppressed the high mitotic index and high number of cells delayed in prophase and prometaphase observed in <italic>cid</italic> single mutants (compare to ratios in <xref ref-type="fig" rid="pgen-0020110-g002">Figure 2</xref>B). A, anaphase; M, metaphase; MI, mitotic index; P, prophase; PM, prometaphase; T, telophase.</p></caption><graphic xlink:href="pgen.0020110.g005"/></fig></sec><sec id="s2f"><title>CENP-C and the SAC Components BUBR1 and ROD Are Unable to Localize to Kinetochores in <italic>cid</italic> Mutants</title><p>It has been proposed that the APC (anaphase-promoting complex) inhibitory signal is generated by the rapid turnover of SAC proteins at unattached kinetochores [<xref rid="pgen-0020110-b039" ref-type="bibr">39</xref>&#x02013;<xref rid="pgen-0020110-b043" ref-type="bibr">43</xref>]. We previously demonstrated that all tested outer kinetochore proteins (ROD, BUBR1, Cenp-meta, and POLO) fail to localize to kinetochores in CID-depleted tissue culture cells and CID antibody-injected embryos [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>]. CENP-A disruptions in <italic>Caenorhabditis elegans,</italic> mouse, and human cells also result in failure to properly localize kinetochore components, including SAC proteins [<xref rid="pgen-0020110-b015" ref-type="bibr">15</xref>,<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>&#x02013;<xref rid="pgen-0020110-b019" ref-type="bibr">19</xref>,<xref rid="pgen-0020110-b037" ref-type="bibr">37</xref>]. Disruption of kinetochore formation and SAC protein localization in <italic>cid/cid</italic> embryos was determined by staining for inner and outer kinetochore proteins. We found that the inner kinetochore protein CENP-C [<xref rid="pgen-0020110-b030" ref-type="bibr">30</xref>] was absent in most <italic>cid/cid</italic> cells, and occasionally was mislocalized in a diffuse pattern throughout the cell, consistent with studies in other organisms and with a severe disruption of kinetochore assembly (<xref ref-type="fig" rid="pgen-0020110-g006">Figure 6</xref>). Consistent with these results, the SAC components ROD and BUBR1 were unable to localize to kinetochores in stage 15 <italic>cid</italic> null animals, whereas BUBR1 and ROD were localized to kinetochores during all stages of mitosis in heterozygous controls (<xref ref-type="fig" rid="pgen-0020110-g006">Figure 6</xref>). We conclude that <italic>cid</italic> null mutations delay cells in early mitosis in the absence of sustained kinetochore localization of essential components of the SAC, despite the requirement for at least one of these SAC components (BUBR1).</p><fig id="pgen-0020110-g006" position="float"><label>Figure 6</label><caption><title>Inner and Outer Kinetochore Protein Localizations Are Disrupted in <italic>cid</italic> Mutant Embryos</title><p>Kinetochore localization of CENP-C, ROD, and BUBR1 were determined in stage 15 embryos. In <italic>cid/CyO</italic> control embryos (left), all three proteins were localized to the centromere/kinetochore during interphase (CENP-C) or the early stages of mitosis (ROD and BUBR1). All three proteins were absent from centromeres/kinetochores in <italic>cid/cid</italic> animals (right); in some cases, CENP-C was mislocalized in a diffuse pattern. Scale bars indicate 5 &#x003bc;m.</p></caption><graphic xlink:href="pgen.0020110.g006"/></fig></sec></sec><sec id="s3"><title>Discussion</title><p>We have shown that null mutations in the <italic>Drosophila</italic> member of the CENP-A protein family result in embryonic lethality after depletion of maternal CID protein. CID-depleted embryonic cells display an early mitotic delay, consistent with cell cycle defects observed after CID antibody injection [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>], suggesting the involvement and activation of a cell cycle checkpoint. This result is similar to a recent knockout of CENP-A in chicken DT-40 cells and a CENP-C mutation in <italic>Drosophila,</italic> both of which resulted in a mitotic delay in the absence of kinetochore assembly [<xref rid="pgen-0020110-b020" ref-type="bibr">20</xref>,<xref rid="pgen-0020110-b030" ref-type="bibr">30</xref>]. However, these studies did not determine when the delay occurred in mitosis, whether the mitotic delays involved the SAC or the DNA repair checkpoint, or whether similar responses to CENP-A depletion occurred in animals.</p><p>We addressed the possible involvement of two known cell cycle checkpoints in the CID-mediated early mitotic delay, specifically the DNA damage and SACs. A recent study suggested that DNA damage and repair may be involved in CENP-A assembly in <italic>Xenopus</italic> [<xref rid="pgen-0020110-b033" ref-type="bibr">33</xref>], raising the possibility that elimination of CID alters centromeric chromatin, resulting in DNA damage at the centromere. We addressed this hypothesis in two complementary ways. First, we compared the <italic>cid</italic> mutant mitotic delay phenotypes to the behavior of cells after inducing general DNA damage with doxorubicin. Induction of DNA damage resulted in a reduced mitotic index, not the increased mitotic index observed in the <italic>cid</italic> mutant embryos. Second, we disrupted the DNA damage checkpoint in <italic>cid</italic> mutant embryos using caffeine treatment, which inhibits MEI-41 (ATR), an essential component of the DNA damage response [<xref rid="pgen-0020110-b034" ref-type="bibr">34</xref>,<xref rid="pgen-0020110-b035" ref-type="bibr">35</xref>]. Caffeine treatment did not abrogate the <italic>cid</italic>-mediated mitotic delay. We conclude that DNA damage does not appear to be the signal that induces the <italic>cid</italic>-mediated early mitotic delay, and that this delay does not require an intact DNA damage checkpoint.</p><p>These results led us to address the possible involvement of the SAC in the CID-mediated early mitotic delay in animals. The SAC monitors microtubule attachments to the kinetochore; if normal bipolar attachments are not formed, activation of the SAC blocks entry into anaphase, resulting in a prometaphase/metaphase arrest [<xref rid="pgen-0020110-b034" ref-type="bibr">34</xref>,<xref rid="pgen-0020110-b035" ref-type="bibr">35</xref>]. The fact that the CID-mediated delay occurred earlier in mitosis than expected for activation of the SAC suggested that this checkpoint might not be involved. However, we observed that <italic>cid</italic> null mutant cells retained an intact SAC response to microtubule disruption by colcemid, which is similar to the response of Cse4 mutants in <italic>Sa. cerevisiae</italic> [<xref rid="pgen-0020110-b038" ref-type="bibr">38</xref>]. In addition, mutating an essential SAC component (BUBR1) resulted in abrogation of the CID-mediated delay. Previous studies suggested that kinetochore localization of SAC proteins (e.g. MAD2, BUBR1, ROD, and CENP-E) is absolutely required for SAC function. Nevertheless, we observed that BUBR1 and ROD, and the inner kinetochore protein CENP-C, lacked kinetochore localization in <italic>cid</italic> mutant embryos. These results suggest that the CID-mediated early mitotic delay involves the SAC, and that BUBR1 is serving a kinetochore-independent role in delaying mitotic progression, as suggested by recent studies in human and yeast cells [<xref rid="pgen-0020110-b012" ref-type="bibr">12</xref>,<xref rid="pgen-0020110-b044" ref-type="bibr">44</xref>,<xref rid="pgen-0020110-b045" ref-type="bibr">45</xref>].</p><sec id="s3a"><title>Why Do <italic>cid</italic> Mutants Display an Early Mitotic Delay That Is BUBR1-Dependent?</title><p>Based on the previous observation of interphase/prophase arrest after CID antibody injection into embryos, we proposed that cells monitor kinetochore assembly early in mitosis, in addition to monitoring the presence of bipolar attachments later in mitosis [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>]. It is also possible that <italic>cid</italic> null kinetochores may be able to recruit normal levels of SAC components at early stages of mitosis, but are unable to retain functional levels later in mitosis, as observed for disruption of human Hec1 and Nuf2 and DT-40 CENP-A [<xref rid="pgen-0020110-b007" ref-type="bibr">7</xref>,<xref rid="pgen-0020110-b020" ref-type="bibr">20</xref>].</p><p>Alternatively, mitotic arrest may occur in the absence of kinetochore localization of SAC components. Consistent with this hypothesis, the <italic>cid-</italic>mediated early mitotic delay requires at least one SAC component (BUBR1), yet occurs without sustained kinetochore localization of multiple, essential components of the SAC (reported here and in [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>]). The finding that defects in kinetochore assembly lead to a BUBR1-dependent early mitotic delay is supported by several recent studies. Disruption of chicken CENP-A, CENP-H, or CENP-I, all inner kinetochore proteins, delays cells in mitosis for hours [<xref rid="pgen-0020110-b009" ref-type="bibr">9</xref>&#x02013;<xref rid="pgen-0020110-b011" ref-type="bibr">11</xref>,<xref rid="pgen-0020110-b020" ref-type="bibr">20</xref>]. These results suggest that the SAC is able to respond to multiple types of signals and inhibit cell cycle progression.</p><p>How could SAC components contribute to cell cycle delay early in mitosis, prior to their well-established role in monitoring bipolar attachments in prometaphase/metaphase? Loss of CENP-A proteins blocks kinetochore assembly, which may generate &#x0201c;free&#x0201d; (non-kinetochore localized) SAC complexes capable of inhibiting mitotic progression (<xref ref-type="fig" rid="pgen-0020110-g007">Figure 7</xref>). Since the active inhibitory complex for the SAC is present throughout the cell cycle [<xref rid="pgen-0020110-b046" ref-type="bibr">46</xref>], the complete absence of kinetochore assembly, or the presence of &#x0201c;free&#x0201d; SAC components, could block cells early in mitosis by chronically activating the SAC. It has recently been shown that both BUBR1 and MAD2 function in a kinetochore-independent manner to regulate the length of mitosis, in addition to monitoring kinetochore-microtubule attachments [<xref rid="pgen-0020110-b012" ref-type="bibr">12</xref>,<xref rid="pgen-0020110-b045" ref-type="bibr">45</xref>]. Furthermore, recent studies in <italic>Drosophila</italic> have revealed a role for Bub3 in G2 and early mitosis in promoting the accumulation of mitotic cyclins [<xref rid="pgen-0020110-b047" ref-type="bibr">47</xref>], suggesting that components can ensure normal mitotic progression by inhibiting the APC in a kinetochore-independent manner. This interpretation is also consistent with recent studies that demonstrate that SAC proteins play multiple roles in cell cycle regulation [<xref rid="pgen-0020110-b048" ref-type="bibr">48</xref>&#x02013;<xref rid="pgen-0020110-b050" ref-type="bibr">50</xref>]. For example, mutations in <italic>Drosophila bubr1</italic> have been shown to bypass the SAC, and are also able to suppress mutations that activate both the DNA damage and SAC in early embryos [<xref rid="pgen-0020110-b048" ref-type="bibr">48</xref>,<xref rid="pgen-0020110-b051" ref-type="bibr">51</xref>]. Furthermore, it has recently been demonstrated that SAC components are responsible for mediating a mitotic arrest in response to DNA damage in vertebrate cells [<xref rid="pgen-0020110-b034" ref-type="bibr">34</xref>], and the mitotic arrest in response to spindle malorientation in <italic>Sc. pombe</italic> [<xref rid="pgen-0020110-b052" ref-type="bibr">52</xref>]. These results strengthen the conclusion that the SAC can respond to more than bipolar kinetochore microtubule attachment, and suggest multiple roles for SAC components in cell cycle regulation. Therefore, the most likely explanation for the <italic>cid-</italic>mediated mitotic delay is that inhibitory SAC complexes can be formed in the absence of kinetochore localization (<xref ref-type="fig" rid="pgen-0020110-g007">Figure 7</xref>).</p><fig id="pgen-0020110-g007" position="float"><label>Figure 7</label><caption><title>SAC Components Affect Cell Cycle Progression in the Absence of Kinetochore Localization</title><p>In normal cells, CENP-A chromatin assembly is followed by the recruitment of inner and outer kinetochore proteins [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>]. We propose that until kinetochore assembly is complete, free SAC components may be responsible for cell cycle inhibition (early activation of the SAC). Upon completion of kinetochore assembly, SAC components delay anaphase until all chromosomes have achieved bipolar spindle attachment. In <italic>cid</italic> null mutants, both inner and outer kinetochore proteins are free, resulting in a SAC-dependent early mitotic delay that does not depend on localization of SAC components to kinetochores.</p></caption><graphic xlink:href="pgen.0020110.g007"/></fig><p>The role of the kinetochore in cell cycle progression and the functions of SAC components are clearly more complex than previously thought. Future studies should focus on identifying the components and mechanisms responsible for the <italic>cid-</italic>mediated mitotic delay, and determining if this complex is identical to the standard SAC inhibitory complex.</p></sec></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Cytology.</title><p> <italic>cid</italic> mutant embryos were collected from interallelic crosses and stained as described using either a formaldehyde or MeOH:EGTA fixation. <italic>Trans</italic>-heterozygous combinations of the different <italic>cid</italic> alleles were generated in order to eliminate phenotypic effects of other lethal mutations present on each of the <italic>cid</italic> mutant chromosomes (unpublished data). All of the data presented were obtained for crosses between <italic>cid^11&#x02013;2</italic> and <italic>cid^22&#x02013;4</italic>, although crosses between other alleles produced identical phenotypes. Antibodies used were cyclin A [<xref rid="pgen-0020110-b031" ref-type="bibr">31</xref>], cyclin B [<xref rid="pgen-0020110-b031" ref-type="bibr">31</xref>], LacZ (Sigma, St. Louis, Missouri, United States), tubulin (Sigma), ROD [<xref rid="pgen-0020110-b053" ref-type="bibr">53</xref>], BUBR1 [<xref rid="pgen-0020110-b051" ref-type="bibr">51</xref>], and CID [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>]. For quantitation of mitotic index and cyclin abundance, all cells within the developing central nervous system were counted from at least five mutant and five control embryos. The ratios presented are the number of PH3- or cyclin-positive cells divided by total cells, in order to normalize for the lower nuclear density present in <italic>cid</italic> mutant embryos. Quantification of the stages of mitosis was performed by costaining embryos for PH3 and tubulin. The distinction between prophase and prometaphase was made as follows: Prophase was classified as chromosomes with incomplete condensation (i.e., round PH3+ nucleus), in which no individual chromosomes or chromosome arms were visible and DNA was not obviously aligning at the metaphase plate. Prophase tubulin staining showed bright centrosomal signals with little or no obvious microtubules interacting with the chromosomes. Prometaphase was classified as chromosomes with complete condensation (i.e., clearly visible individual chromosomes and chromosome arms) in which the chromosomes were clearly in the process of aligning at the metaphase plate. Tubulin staining showed a focused bipolar microtubule array that was clearly interacting with the chromosomes.</p><p>For quantification of CID levels in mitotically active cells in mutant and control embryos (<xref ref-type="supplementary-material" rid="pgen-0020110-sg001">Figure S1</xref>), the sum of pixel values for both CID and PROD immunofluorescence from five to seven embryos was obtained using the two-dimensional polygon finding tool in softWoRx (Applied Precision, Issaquah, Washington, United States). The pixel values were summed and presented as a ratio of CID:PROD, to provide a rough estimate of the amount of CID depletion in each embryo. Based on these ratios, 90%&#x02013;100% of CID was depleted in stage 15 <italic>cid/cid</italic> mutants, relative to heterozygous controls, suggesting retention of a small amount of maternal protein in some cells. For all quantitations, standard deviations were calculated per embryo, and data were compared using the Student <italic>t</italic> test. Note that the amount of CID depletion in <italic>cid</italic> homozygotes is likely to be an underestimate (up to 2-fold) with respect to wild-type embryos, since <italic>cid</italic> mutant heterozygotes were used as the quantitation controls.</p><p>All images were acquired using a DeltaVision workstation (Applied Precision) and analyzed using softWoRx software, as described previously [<xref rid="pgen-0020110-b017" ref-type="bibr">17</xref>].</p></sec><sec id="s4b"><title>Drug treatments.</title><p> <italic>cid</italic> mutant and heterozygous embryos were bleach dechorionated and incubated in a 1:1 mixture of Schneider's medium (+10% heat-inactivated FBS) and octane as described in [<xref rid="pgen-0020110-b054" ref-type="bibr">54</xref>]. Colcemid was used at a concentration of 3 &#x003bc;g/ml for 1 h, caffeine was used at a concentration of 2 mM for 2 h, and doxorubicin was used at a concentration of 2 &#x003bc;M for 2 h. After drug treatment, embryos were fixed using formaldehyde and processed for immunofluorescence as described above.</p></sec><sec id="s4c"><title>Genetics.</title><p>The <italic>bubr1</italic> allele used was k03113, and was obtained from the Bloomington Stock Center (Bloomington, Indiana, United States). <italic>cid bubr1</italic> double mutants were generated by recombination using standard methods.</p><p>Mutations in <italic>cid</italic> (<italic>centromere identifier</italic>/CG13329) were recovered in genetic screens designed to isolate new mutant alleles of <italic>cnn (centrosomin)</italic> [<xref rid="pgen-0020110-b055" ref-type="bibr">55</xref>]. The <italic>cid</italic> locus is tightly linked to <italic>cnn</italic> in the 50A region of the right arm of the second chromosome in <named-content content-type="genus-species">D. melanogaster</named-content>. The genes in this genomic region, proximal to distal, are <italic>cnn</italic> (<italic>centrosomin</italic>/CG4832), <italic>Cbs</italic> (<italic>centrosomin's beautiful sister</italic>/CG4840), <italic>arr</italic> (<italic>arrow</italic>/CG5912), <italic>cbc</italic> (<italic>crowded by cid</italic>/CG5970), <italic>cid</italic> (<italic>centromere identifier</italic>/CG13329), <italic>bbc</italic> (<italic>b-b in a boxcar</italic>/CG6016), and <italic>drk</italic> (<italic>downstream of receptor kinase/</italic>CG6033). The initial characterization of this region included screening of cDNA libraries and expressed sequence tag (EST) collections to produce transcript profiles for each of these genes, and saturation mutagenesis screens to recover recessive lethal and sterile mutations. Breakpoint-associated mutations, principally deletions, and complementation analyses were used to localize each of the newly recovered mutations to the individual molecularly defined and computationally identified transcription units. Using primers designed from genomic and cDNA sequences the mutant alleles of each locus were sequenced and the genetic localization of the complementation groups confirmed. This screen resulted in the recovery of the four alleles of <italic>cid</italic> reported in this paper: <italic>cid^{t11&#x02013;2}</italic> Q51 to stop, <italic>cid^{t12&#x02013;1}</italic> Q83 to stop, <italic>cid^{t21&#x02013;3}</italic> Q94 to stop, and <italic>cid^{t22&#x02013;4}</italic> Q102 to stop.</p></sec><sec id="s4d"><title>Examination of <italic>cid bubr1</italic> heterozygous mutants.</title><p>During the course of scoring the mitotic parameters of <italic>cid bubr1</italic> double mutants, we noticed that <italic>cidbubr1</italic> single mutants had a mitotic index nearly twice as high as <italic>cid</italic> single mutants alone, which prompted us to investigate these heterozygous mutants further for possible haploinsufficiency effects. We found no incidence of chromosome segregation defects in <italic>cid bubr1/CyO</italic> embryos despite their elevated mitotic index. We also examined mitotic tissue of <italic>cid bubr1/CyO</italic> third instar larval brains because this tissue allows a more precise karyotypic analysis and could reveal subtle defects not seen in embryonic tissue. We found that <italic>cid bubr1/CyO</italic> animals had a higher mitotic index than <italic>cid/CyO</italic> animals (1.10 [<italic>n</italic> = 315 fields] vs. 0.76 [<italic>n</italic> = 400 fields]), yet we did not find any evidence for aneuploidy or mitotic defects in any of the mitotic figures examined. We then determined whether <italic>cid bubr1</italic> and <italic>cid</italic> heterozygous animals had a normal response to colcemid treatment by incubating brains with colcemid for 1 h. We found that <italic>cid bubr1</italic> and <italic>cid</italic> heterozygous mutants had a normal response to colcemid treatment (<italic>cid/CyO</italic> mitotic index increased from 0.86 to 2.17 [<italic>n</italic> = 551 fields], and <italic>cid bubr1/CyO</italic> mitotic index increased from 1.10 to 2.37 [<italic>n</italic> = 264 fields]). From this data we conclude that <italic>cid</italic> and <italic>cid bubr1</italic> mutants do not have a haploinsufficient effect on mitosis, and that there are likely to be other factors in the genetic background that lead to the increased mitotic index of <italic>cid bubr1</italic> double mutants. To avoid interpretation artifacts that might be caused by this difference, in all cases we only compared data from <italic>cid bubr1</italic> homozygotes to <italic>cid bubr1</italic> heterozygotes, and <italic>cid</italic> homozygotes to <italic>cid</italic> heterozygotes.</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting Information</title><supplementary-material content-type="local-data" id="pgen-0020110-sg001"><label>Figure S1</label><caption><title>Quantification of the Amount of CID Depletion in <italic>cid</italic> Mutant Embryos</title><p>Stage 15 <italic>cid/cid</italic> and <italic>cid/CyO</italic> embryos were stained for CID and PROD. PROD was present in a punctate pattern in both genotypes. Estimation of the amount of CID depletion was preformed by comparing the ratio of total CID staining to PROD staining in five different mutant and heterozygous embryos. From this analysis we estimate that 90%&#x02013;100% of CID protein is depleted in mutant embryos.</p><p>(4.0 MB PDF)</p></caption><media xlink:href="pgen.0020110.sg001.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material></sec>

Operational Challenges in Large Clinical Trials: Examples and Lessons Learned from the Gambia Pneumococcal Vaccine Trial

Could not extract abstract

<contrib contrib-type="author"><name><surname>Cutts</surname><given-names>Felicity T</given-names></name><xref ref-type="corresp" rid="cor1">*</xref><xref ref-type="author-notes" rid="n105"/></contrib><contrib contrib-type="author"><name><surname>Enwere</surname><given-names>Godwin</given-names></name><xref ref-type="author-notes" rid="n105"/></contrib><contrib contrib-type="author"><name><surname>Zaman</surname><given-names>Syed M. A</given-names></name><xref ref-type="author-notes" rid="n105"/></contrib><contrib contrib-type="author"><name><surname>Yallop</surname><given-names>Fred G</given-names></name><xref ref-type="author-notes" rid="n105"/></contrib>

PLoS Clinical Trials

<p>The requirements for Good Clinical Practice (GCP) in clinical trials are well documented [<xref rid="pctr-0010016-b001" ref-type="bibr">1</xref>], and ethical issues are hotly debated [<xref rid="pctr-0010016-b002" ref-type="bibr">2</xref>]. Operational aspects of trials, however, have received far less attention [<xref rid="pctr-0010016-b003" ref-type="bibr">3</xref>], perhaps due to restrictions on journal space for detailing methods. In low resource settings, however, large trials often face many logistical and organizational obstacles, and thus the practical difficulties in running a trial to GCP standards should not be ignored. Here, we describe the main operational challenges to a randomized, double-blind, placebo-controlled trial of the safety and efficacy of pneumococcal conjugate vaccine among over 17,000 infants in the Gambia. The trial began in August 2000, and after the magnitude of the challenges were recognized, a new senior principal investigator (FTC) and project manager (FGY) were recruited, taking up post in June 2001. We summarize here the major lessons learnt in trial implementation in a resource-poor setting.</p><sec id="s2"><title>Overview of Trial Preparation and Study Methods</title><p>The trial was conducted in Upper and Central River Divisions of the Gambia, covering an area of about 5,000 km², bisected by the river Gambia. Government mother-child health (MCH) services are provided at 15 fixed facilities, the two largest being Bansang hospital in the Central River Division and Basse health centre in the Upper River Division, and about 100 additional outreach sites. The Medical Research Council (MRC) main station is in Fajara, 380 km from Basse. The journey between them took 5 hours in 2000 when the trial began, and, due to deteriorating road conditions, 9&#x02013;10 hours by 2004 when it ended.</p><p>Preparatory studies took over 12 years, and included phase I [<xref rid="pctr-0010016-b004" ref-type="bibr">4</xref>] and phase II vaccine trials [<xref rid="pctr-0010016-b005" ref-type="bibr">5</xref>], as well as baseline studies of rates of disease and mortality [<xref rid="pctr-0010016-b006" ref-type="bibr">6</xref>,<xref rid="pctr-0010016-b007" ref-type="bibr">7</xref>], and selection of the best design for the phase III trial [<xref rid="pctr-0010016-b008" ref-type="bibr">8</xref>,<xref rid="pctr-0010016-b009" ref-type="bibr">9</xref>]. Toward the end of this period, the study area was mapped and a household numbering system was devised. Communities and families were informed about the forthcoming trial, through drama performances in large villages, radio spots, distribution of flyers, and community meetings. Unfortunately, we did not evaluate which method was most effective for enhancing community support and understanding of the trial.</p><p>The trial was conducted as a partnership between MRC and the Gambia Government (GG), children being recruited and vaccinated at GG MCH clinics and referred through these clinics for investigations. If MCH clinics or outreach visits were cancelled, this would jeopardise both public health services and the research. The trial funders and MRC therefore invested greatly in infrastructure (capital equipment and buildings, laboratories, radiology facilities, cold chain, and transport) for both Basse field station and GG health services.</p><p>The study methods, which followed detailed standard operating procedures (SOPs), have already been described in detail [<xref rid="pctr-0010016-b010" ref-type="bibr">10</xref>]. The primary endpoint for the trial was initially all-cause mortality but was later changed to radiologically confirmed pneumonia, and secondary endpoints were culture-confirmed invasive pneumococcal disease and hospital admissions. Surveillance for radiological pneumonia, invasive disease, and serious adverse events took place 7 days a week at Basse and Bansang health facilities, with referral of children from outlying clinics over dirt roads and river crossings, and each child was visited at home every 3 months for demographic surveillance.</p></sec><sec id="s3"><title>The Importance of a Quality Management Plan</title><p>The need for a quality management plan is recognized as part of GCP, and we developed a priority list of indicators for quality assurance (<xref ref-type="table" rid="pctr-0010016-t001">Table 1</xref>). We aimed for 100% compliance with indicators relating to following SOPs for recruitment, storage of vaccines and placebo, administration of either vaccine or placebo (which were identical in appearance and provided in numerically coded vials) correctly according to the code indicated in the randomization scheme, reporting and follow-up of serious adverse events, and investigation of children. A single error in any of these areas was discussed immediately with the relevant staff. For other indicators (e.g., dropout between first and third doses of vaccines; intervals between doses; correlation between nurses and doctors' clinical findings), we aimed for at least 95% compliance and reviewed performance at weekly staff meetings.</p><table-wrap id="pctr-0010016-t001" content-type="2col" position="float"><label>Table 1</label><caption><p>Major Indicators Used for Quality Assurance</p></caption><graphic xlink:href="pctr.0010016.t001"/></table-wrap></sec><sec id="s4"><title>The Importance of On-Site Supervision and Continuous Feedback of Adherence to GCP</title><p>We monitored quality by continuous on-site observation. We used TempTale cold chain monitors and maximum-minimum thermometers to monitor vaccine storage. All completed case report forms (CRFs) were checked by field supervisors working with the epidemiologist and principal investigator (PI), with a spirit of competition as to ability to spot mistakes. We queried the data system regularly to identify the number of errors made by each person, and we conducted random spot-checks of field work and home visits. The initial procedure for supervisory rounds stipulated that supervisors visit field workers at weekends (when no recruitment or vaccination was in progress) to collect the week's CRFs, and included few details on checking actual field activities. We changed this so that supervision took place during clinic hours, and trained supervisors to observe practices and complete checklists that included the key quality indicators. To help field workers spot their own mistakes, we conducted refresher training every 2&#x02013;3 months, with written tests on the SOPs, and on dummy completed CRFs on which deliberate mistakes had been made (e.g., putting a date of vaccination before a date of birth). We thus identified staff who had difficulty in noticing range and consistency checks, and gave them further on-the-job training. To improve the clinical classification of the sick child, paediatricians worked with small groups of three to four nurses at a time, each completing a one-page &#x0201c;quality control&#x0201d; form to record respiratory rates and the presence/absence of lower chest wall indrawing simultaneously and compare findings. We organized external quality control for the key endpoints, all positive pneumococcal cultures being confirmed in expert laboratories and a random sample of radiographs being read by a World Health Organization panel of radiologists.</p><p>Feedback through on-the-spot discussion of supervisors' findings, weekly staff meetings, and monthly written reports to all staff and external collaborators was done with the aim of ensuring that everyone understood the need to follow SOPs and meet GCP requirements. We incorporated relevant indicators into staff appraisals to further demonstrate the importance of quality assurance. We found that continuous checking and feedback was required throughout the study in order to prevent standards from dropping. For example, contamination of blood cultures was high at the beginning of the trial but was greatly reduced when we initiated weekly reports from the laboratory to the PI on the number of contaminated samples, and memos from the PI to the nurses who had taken those samples supplemented by discussion at clinical staff meetings each week. If we made this feedback less frequently, contamination rates tended to rise again.</p><p>Independent external trial monitors visited every 4 months to review regulatory and ethical aspects and check source documents [<xref rid="pctr-0010016-b001" ref-type="bibr">1</xref>]. We made it very obvious that all staff involved in the trial, from PI to field worker, were monitored. This helped to create a &#x0201c;culture of checking&#x0201d; that changed perceptions of monitoring from a threat to a management tool and overcame initial resistance to the periodic tests.</p></sec><sec id="s5"><title>The Importance of Documenting Roles and Responsibilities of Collaborating Groups</title><p>Five interlocking groups of personnel were involved in the trial: over 100 GG MCH staff; 150 full-time trial &#x0201c;clinical&#x0201d; staff (doctors and nurses working in three shifts to provide 24-hour cover and field workers doing home visits to the &#x0003e;17,000 children); 12 data staff; over 60 trial &#x0201c;support&#x0201d; staff (drivers, mechanics, administrators, clerks, cleaners, cooks, etc.), and external collaborating scientists and administrative staff based in MRC Fajara and overseas. Multiple coordination mechanisms were needed, ranging from frequent informal contacts in-person at the local level and by telephone/email internationally, to formal, minuted working groups and steering committees. Close collaboration between the trial senior management and GG health teams was important to avoid cancellation of outreach clinics that might otherwise have occurred due, for example, to shortage of transport or conflicting activities. We felt that collaboration would have been further enhanced had memoranda of understanding been written before the trial began; in their absence, it was sometimes difficult to demarcate the requirements for the research project and the wide-ranging needs of the public health system.</p></sec><sec id="s6"><title>The Importance of Transparent Human Resource Management Procedures</title><p>Human resource training and management took up a large proportion of time of both the PI (who focussed on quality assurance, management of clinical staff, and collaboration with external scientists) and project manager (who managed all support staff and liaised with external administrative and management personnel).</p><p>MRC's written, transparent policies on recruitment, career development, appraisal systems, disciplinary procedures, leave entitlements, health and safety, etc., helped to guide supervisors and managers (for example, by showing what procedures need to be documented in order for appropriate decisions to be made) and to set limits for negotiations between staff and managers. We revised job descriptions to ensure that they were realistic, clear, and achievable. For example, initially some job descriptions for different posts overlapped, creating the potential either for unhealthy competition or for each person thinking someone else was responsible for a particular task, whereas others were too broad to be feasible to complete. Other actions to improve staff performance included mentoring, rotating personnel between sites, and training. This ranged from distance-based learning undergraduate and postgraduate degree programmes and short courses inside and outside the Gambia, to literacy classes, basic computing, and in-service training and accreditation programmes for all categories of staff.</p></sec><sec id="s7"><title>The Importance of Adequate Planning for Trial Implementation</title><p>Planning for a trial requires that baseline situation assessments be conducted of the resources available for trial implementation. Given the typically long delay between planning a trial and obtaining the funding, clearances, and vaccine to begin the trial, and the time then required for recruitment and follow-up, these assessments need to predict the needs over the whole life of the trial. In our trial, adequate plans were made for purchasing cold chain equipment, but we faced substantial problems with transport and maintenance and other risks to successful completion of the trial.</p><sec id="s7a"><title/><sec id="s7a1"><title>Transport.</title><p>At the trial outset, five four-wheel drive vehicles were given to the national and divisional government health teams to assist in supervision of MCH activities. In a rapid assessment of the MCH infrastructure in the study area in mid-2001, however, we found that most clinics had severe transport problems, because support to the front-line health services had not been planned. The trial therefore assigned vehicles to clinics to take MCH teams on outreach and to refer ill children to Basse or Bansang, and assisted in the maintenance of GG as well as trial vehicles. This further stretched the ability of our maintenance team to meet demands, and increased the need for close coordination between partners to prioritize vehicles and equipment for repairs. It also highlighted the need for plans to take into account the expected duration of a trial and realistic working life of vehicles in harsh conditions, and budget for replacement costs of essential items, and for written memoranda of understanding to be developed to help to manage expectations.</p></sec><sec id="s7a2"><title>Maintenance.</title><p>Because radiological pneumonia was the primary endpoint, having access to radiology equipment was critical. The trial had refurbished radiology rooms and supplied new dryers and developing tanks, as well as bought two state-of-the-art radiology machines for Bansang hospital and Basse health centre. It proved difficult to bring engineers to maintain these in a timely way, and in 2002, we bought two portable machines for a fraction of the cost, which were more robust and produced good quality films. We maintained transport, equipment, and the all-important generators that were the source of power for the radiology equipment, laboratories, divisional cold stores, Basse health centre, and field station, and little assistance was available in this remote location. The engineering background of the project manager was therefore frequently called on, and substantial external support was required. Detailed inventories of equipment and fixed assets were developed, and use of stocks (e.g., spare parts, fuel, drugs) was computerized and monitored regularly to avoid stock-outs of critical items and reduce excessive use.</p></sec><sec id="s7a3"><title>Risk management.</title><p>No matter how good the planning and management of trials, risks will remain. Some of the &#x0201c;unforeseen&#x0201d; risks are so common in low-resource settings that they should be expected. Those that we faced included currency devaluation, natural disasters (flooding of Basse town and field station) (see <xref ref-type="fig" rid="pctr-0010016-g001">Figure 1</xref>), health and safety hazards such as road traffic accidents, national campaigns (on different occasions for measles, meningitis and polio vaccination), and deterioration in access to the study site. We needed to liaise closely with national, regional, and international authorities and identify local and external sources of support to reduce disruption of activities to a minimum.</p><fig id="pctr-0010016-g001" position="float"><label>Figure 1</label><caption><title>Main Gate into the MRC Field Station during the Floods</title><p>Photo by Fred Yallop.</p></caption><graphic xlink:href="pctr.0010016.g001"/></fig></sec></sec></sec><sec id="s8"><title>Conclusions and Lessons Learned</title><p>There are many operational challenges in running large field trials and strong leadership and teamwork is essential to confront them. We summarise the main ways to improve trial management in <xref ref-type="boxed-text" rid="pctr-0010016-box001">Box 1</xref>. To train personnel and generate the team spirit to implement a large trial to GCP in difficult field conditions, a senior principal investigator needs to be permanently on-site to organize quality assurance and ensure effective communication between all the groups involved, and a full-time, senior project manager is needed to organize and oversee all the support services needed to keep the trial running. We found other factors helping to achieve a high level of motivation and team spirit were the appointment of staff with appropriate and complementary skills, substantial training, objective and transparent monitoring procedures, feedback to each staff member that their work was important for the trial and for society, and recognition of their achievements by both senior staff and external advisory committees. Training is not just important at the start of a trial, it is a continuous need, because staff change and existing staff need constant motivation and support. We recommend the frequent use of written tests to assess knowledge of SOPs and ability to spot mistakes on CRFs, in addition to observation of ongoing practices. We also recommend setting out detailed and appropriate job descriptions, and including measurable performance objectives in staff appraisals. Experienced staff should act as mentors for new staff, and staff rotation between areas of work and locations can help to avoid boredom or a feeling of isolation of staff working in the more remote locations. A comprehensive quality assurance plan is vital and a mixture of internal and external monitoring and auditing is important. Visits from external monitors should be used not only for GCP audit but also to educate all staff of the reasons for, and importance of, quality control.</p><boxed-text id="pctr-0010016-box001" content-type="1col" position="float"><sec id="sb1"><title> <bold>Box 1.</bold> Ways to Improve the Operation of Large Phase III Field Trials</title><list list-type="order"><list-item><p>All senior staff to be on-site, including PI and project manager, who must create strong team spirit.</p></list-item><list-item><p>Establish clear SOPs, define process indicators, and set up internal and external monitoring systems for clinical, data, and support procedures.</p></list-item><list-item><p>Document roles and responsibilities of collaborating groups through memoranda of understanding.</p></list-item><list-item><p>Define tasks clearly and realistically, assign responsibility and accountability, and make sure people know what to do and have the skills and support to do it.</p></list-item><list-item><p>Make plans based on situation assessments conducted by a team including experienced managers, including risk management plans.</p></list-item><list-item><p>Identify the appropriate resources for the trial and for partners involved in implementing the trial.</p></list-item><list-item><p>Take into account the operational life of transport and equipment in the relevant field conditions and budget for replacement costs of essential items.</p></list-item><list-item><p>Make adequate arrangements for maintenance, including skilled staff, workshops with appropriate space and safety arrangements, and timely external support as required.</p></list-item><list-item><p>Monitor everything closely, anticipate problems, and react early.</p></list-item><list-item><p>Check, check, check, check, check&#x02026;and check again.</p></list-item></list></sec></boxed-text><p>Multiple stakeholders are likely to be involved and may pool resources to run large trials. Memoranda of understanding should be written at the outset, to delineate clearly the roles and responsibilities of different partners, as well as the resources provided to or by each partner and their disposition at the end of the trial. The resources available at the study site should be reviewed carefully to determine the need for external specialist advice and identify appropriate sources for this. Aspects likely to require external support when a trial is run in a remote setting include power supplies, communications, data management (we will describe this in a separate report), cold chain, health and safety, and security.</p><p>The scientific objectives of a trial need to be translated into a detailed project development plan that outlines the key steps, activities, milestones, and critical path to ensure that trial implementation proceeds as planned. Tools such as MS Project are useful to track the multiple interlocking activities and ensure that milestones are reached on time. Training on effective project planning and evaluation in biomedical research and the use of such tools are now available via the Special Programme for Research and Training in Tropical Diseases (e.g., see <ext-link ext-link-type="uri" xlink:href="www.who.int/tdr/publications/publications/training_manual.htm">www.who.int/tdr/publications/publications/training_manual.htm</ext-link>).</p><p>Existing clinical trial and data management guidelines should be expanded to incorporate monitoring of these areas and of accounting and budgetary management and control procedures. The management challenges in implementing large trials in resource-poor settings should not be underestimated and budgets must include adequate investment to meet these challenges.&#x02002;</p></sec>

Intronic Alternative Splicing Regulators Identified by Comparative Genomics in Nematodes

<p>Many alternative splicing events are regulated by pentameric and hexameric intronic sequences that serve as binding sites for splicing regulatory factors. We hypothesized that intronic elements that regulate alternative splicing are under selective pressure for evolutionary conservation. Using a Wobble Aware Bulk Aligner genomic alignment of <named-content content-type="genus-species">Caenorhabditis elegans</named-content> and <italic>Caenorhabditis briggsae,</italic> we identified 147 alternatively spliced cassette exons that exhibit short regions of high nucleotide conservation in the introns flanking the alternative exon. In vivo experiments on the alternatively spliced <italic>let-2</italic> gene confirm that these conserved regions can be important for alternative splicing regulation. Conserved intronic element sequences were collected into a dataset and the occurrence of each pentamer and hexamer motif was counted. We compared the frequency of pentamers and hexamers in the conserved intronic elements to a dataset of all <named-content content-type="genus-species">C. elegans</named-content> intron sequences in order to identify short intronic motifs that are more likely to be associated with alternative splicing. High-scoring motifs were examined for upstream or downstream preferences in introns surrounding alternative exons. Many of the high- scoring nematode pentamer and hexamer motifs correspond to known mammalian splicing regulatory sequences, such as (T)GCATG, indicating that the mechanism of alternative splicing regulation is well conserved in metazoans. A comparison of the analysis of the conserved intronic elements, and analysis of the entire introns flanking these same exons, reveals that focusing on intronic conservation can increase the sensitivity of detecting putative splicing regulatory motifs. This approach also identified novel sequences whose role in splicing is under investigation and has allowed us to take a step forward in defining a catalog of splicing regulatory elements for an organism. In vivo experiments confirm that one novel high-scoring sequence from our analysis, (T)CTATC, is important for alternative splicing regulation of the <italic>unc-52</italic> gene.</p>

<contrib contrib-type="author"><name><surname>Kabat</surname><given-names>Jennifer L</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>Barberan-Soler</surname><given-names>Sergio</given-names></name><xref ref-type="aff" rid="aff1">1</xref></contrib><contrib contrib-type="author"><name><surname>McKenna</surname><given-names>Paul</given-names></name><xref ref-type="aff" rid="aff2">2</xref><xref ref-type="author-notes" rid="n105">&#x000a4;a</xref></contrib><contrib contrib-type="author"><name><surname>Clawson</surname><given-names>Hiram</given-names></name><xref ref-type="aff" rid="aff2">2</xref></contrib><contrib contrib-type="author"><name><surname>Farrer</surname><given-names>Tracy</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="author-notes" rid="n106">&#x000a4;b</xref></contrib><contrib contrib-type="author"><name><surname>Zahler</surname><given-names>Alan M</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib>

PLoS Computational Biology

<sec id="s1"><title>Introduction</title><p>One of the interesting lessons learned from the analysis of the human genome is that we may possess fewer than 25,000 genes [<xref rid="pcbi-0020086-b001" ref-type="bibr">1</xref>]. One mechanism to dramatically increase the complexity of the human proteome from this lower-than-expected number of genes is to allow some genes to encode multiple proteins. This process can be accomplished by alternative precursor messenger RNA (pre-mRNA) splicing. Studies that use expressed sequence tag (EST) alignments to identify alternatively spliced genes have led researchers to predict that up to 60% of human genes are alternatively spliced [<xref rid="pcbi-0020086-b002" ref-type="bibr">2</xref>&#x02013;<xref rid="pcbi-0020086-b005" ref-type="bibr">5</xref>]. Alternative splicing events can be regulated in tissue-specific, developmental, and hormone-responsive manners, providing additional mechanisms for the regulation of gene expression [<xref rid="pcbi-0020086-b006" ref-type="bibr">6</xref>,<xref rid="pcbi-0020086-b007" ref-type="bibr">7</xref>]. Understanding alternative splicing and its regulation is a key component to understanding metazoan genomes.</p><p>The current models for alternative splicing regulation are based on the interactions of intronic or exonic RNA sequences, known as <italic>cis</italic> elements, with splicing regulatory proteins known as <italic>trans</italic>-acting splicing factors [<xref rid="pcbi-0020086-b008" ref-type="bibr">8</xref>]. The binding of splicing factors to the pre-mRNA regulates the ability of the spliceosomal machinery to recognize and promote alternative splicing. The role of intronic elements in regulating splicing is well established and has been shown to work in a combinatorial fashion based on the <italic>trans</italic>-acting factors that are present. For example, the inclusion of the 18-nucleotide-long, neural-specific N1 exon of the human <italic>c-SRC</italic> gene is regulated by the downstream control sequence found in the intron downstream of the N1 exon. This sequence serves as a recruitment site for both constitutive and neuronal cell-specific splicing factors such as nPTB, FOX-1, and FOX-2 [<xref rid="pcbi-0020086-b009" ref-type="bibr">9</xref>&#x02013;<xref rid="pcbi-0020086-b012" ref-type="bibr">12</xref>]. The vertebrate RNA-binding protein FOX-1 can also regulate muscle-specific alternative splicing through interactions with the RNA sequence GCAUG [<xref rid="pcbi-0020086-b013" ref-type="bibr">13</xref>], and repeats of this sequence have been shown to be important for alternative splicing regulation of the fibronectin exon EIIIB and the rat calcitonin/CGRP exon 4 [<xref rid="pcbi-0020086-b014" ref-type="bibr">14</xref>,<xref rid="pcbi-0020086-b015" ref-type="bibr">15</xref>]. Many other examples of complex and combinatorial regulation of alternative splicing through intronic <italic>cis</italic> elements have been demonstrated, and combinatorial interactions between proteins such as Nova-1, polypyrimidine tract binding protein (PTB), and ETR-3, with specific <italic>cis</italic> sequences, are important for alternative splicing regulation [<xref rid="pcbi-0020086-b016" ref-type="bibr">16</xref>&#x02013;<xref rid="pcbi-0020086-b020" ref-type="bibr">20</xref>].</p><p>Intronic sequences are non-coding, and therefore they should have less evolutionary selective pressure to maintain their sequence. An exception to this should be intronic sequences that regulate alternative splicing. In an analysis of alternatively spliced human cassette exons, it was found that on average, approximately 100 nucleotides of intron sequence, flanking either side of the exon, tend to be highly conserved between the mouse and human genomes, with 88% identity in the upstream sequences and 80% identity in the downstream sequences [<xref rid="pcbi-0020086-b021" ref-type="bibr">21</xref>]. Some clues to potential splicing regulatory motifs arise from these studies. For example, Sorek and Ast found that the sequence TGCATG was the second most common hexamer in the first 100 nucleotides downstream of alternatively spliced exons, appearing in 18% of these intronic regions [<xref rid="pcbi-0020086-b021" ref-type="bibr">21</xref>]. Another study of aligned mouse/human alternative exons found that GCATG is the most overrepresented pentamer in the proximal conserved region of the intron downstream of alternative exons [<xref rid="pcbi-0020086-b022" ref-type="bibr">22</xref>]. A third study found that TGCATG is frequently located in introns flanking brain-enriched alternative exons, and its presence and spacing are highly conserved in these genes from fish to man [<xref rid="pcbi-0020086-b023" ref-type="bibr">23</xref>].</p><p>Using the nematode <named-content content-type="genus-species">Caenorhabditis elegans</named-content> as a model system, we have been working to take advantage of comparative genomics to identify <italic>cis</italic>-acting regulators of alternative splicing. The <named-content content-type="genus-species">C. elegans</named-content> gene structure, splicing machinery, and gene expression regulation is similar to that of other higher eukaryotes, with the exception that the average intron size is smaller. Our lab has previously developed methods for identification of alternatively spliced genes in <named-content content-type="genus-species">C. elegans</named-content> by aligning the genome sequence with ESTs and mRNA sequence [<xref rid="pcbi-0020086-b024" ref-type="bibr">24</xref>]. We developed an algorithm, the Wobble Aware Bulk Aligner (WABA), for creating interspecies genome alignments between <named-content content-type="genus-species">C. elegans</named-content> and the related roundworm, <named-content content-type="genus-species">Caenorhabditis briggsae</named-content> [<xref rid="pcbi-0020086-b025" ref-type="bibr">25</xref>]. WABA employs a hidden Markov model (HMM) to identify aligned regions as coding, high homology, low homology, and no homology. It also factors the AT richness of <named-content content-type="genus-species">C. elegans</named-content> introns into its calculations when it defines an intronic region as &#x0201c;high&#x0201d; homology [<xref rid="pcbi-0020086-b025" ref-type="bibr">25</xref>]. <named-content content-type="genus-species">C. briggsae</named-content> and <named-content content-type="genus-species">C. elegans</named-content> diverged approximately 100 million years ago, yet are indistinguishable by eye [<xref rid="pcbi-0020086-b026" ref-type="bibr">26</xref>]. Alignment of these two genomes revealed that exonic sequences are highly conserved between these species, but intronic and intergenic sequences are rarely conserved [<xref rid="pcbi-0020086-b025" ref-type="bibr">25</xref>]. We found that these rare, high homology sequences in introns are more likely to occur in the introns flanking alternatively spliced exons than in total introns [<xref rid="pcbi-0020086-b025" ref-type="bibr">25</xref>]. We hypothesized that these conserved intronic regions were <italic>cis</italic>-acting regulatory elements for alternative splicing. This nematode alignment, with relatively limited regions of high homology, provides the possibility for more specific pinpointing of intronic splicing regulatory elements than the much longer 100-nucleotide-long conserved regions flanking alternative exons in mouse/human alignments [<xref rid="pcbi-0020086-b021" ref-type="bibr">21</xref>].</p><p>In this paper, we present the analysis of conserved regions of introns flanking alternatively spliced exons in <named-content content-type="genus-species">C. elegans</named-content> and correlate these conserved regions with alternative splicing regulation. We collected these conserved sequence regions into a database and searched for overrepresented pentamers and hexamers relative to a total intron database, similar to the method used by Burge's group to identify exonic splicing enhancers [<xref rid="pcbi-0020086-b027" ref-type="bibr">27</xref>]. This allowed us to create a table of potential intronic alternative splicing <italic>cis</italic>-regulatory motifs. Since many RNA recognition motif&#x02013;containing splicing factors recognize specific sequences on the order of 4&#x02013;6 nucleotides in length [<xref rid="pcbi-0020086-b011" ref-type="bibr">11</xref>,<xref rid="pcbi-0020086-b013" ref-type="bibr">13</xref>,<xref rid="pcbi-0020086-b018" ref-type="bibr">18</xref>,<xref rid="pcbi-0020086-b028" ref-type="bibr">28</xref>&#x02013;<xref rid="pcbi-0020086-b032" ref-type="bibr">32</xref>], the high-scoring motifs in this catalog may represent specific binding sites for particular splicing factors. Several of our highest scoring motifs in this analysis correlate with known vertebrate splicing regulatory elements, for example, (T)GCATG [<xref rid="pcbi-0020086-b023" ref-type="bibr">23</xref>], but several have not been previously identified. A number of candidates identified by this method were tested in an in vivo splicing reporter system in <named-content content-type="genus-species">C. elegans</named-content>. We have used this analysis to identify and confirm a new, highly conserved, alternative splicing regulatory element, (T)CTATC. We show that this sequence works in coordination with GCATG to regulate the alternative splicing of the <italic>unc-52</italic> gene.</p></sec><sec id="s2"><title>Results</title><sec id="s2a"><title>Identification of Highly Conserved Regions in Introns Flanking Alternatively Spliced Exons</title><p>In order to identify alternatively spliced cassette exons in the <named-content content-type="genus-species">C. elegans</named-content> genome, we used the Intronerator [<xref rid="pcbi-0020086-b024" ref-type="bibr">24</xref>] to generate an initial set of 1,471 putative alternatively spliced genes. We did this by aligning over 200,000 ESTs and mRNAs to the <named-content content-type="genus-species">C. elegans</named-content> genome and identifying regions where the alignments are consistent with more than one way to process a gene. The program could not distinguish between alternative splicing and alternative promoters, so we analyzed each of these alignments individually to verify alternative splicing. We found 449 examples of genes with strong cDNA evidence for alternative cassette exons, and 454 examples of genes with alternative promoters, which usually had unique first exons that are spliced to common second exons. Of the genes with alternative splicing, 162 also contained alternative promoter usage. The remaining genes in the initial set of putative alternatively spliced genes were mostly due to unusual ESTs in the database that did not fit a gene model. We also saw evidence of ESTs that showed internal deletions at short direct repeats. To the program, these indicated the potential for intron removal, but upon further inspection, these did not meet the criteria of introns (start with the dinucleotides GT or GC and end with AG) and were likely the result of cloning artifacts of the ESTs.</p><p>Intronic regions that are highly conserved between <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> are rarely identified by WABA [<xref rid="pcbi-0020086-b025" ref-type="bibr">25</xref>]. While analyzing our alternative cassette exon database, we were struck by the fact that we could identify many examples of WABA-defined high homology sequences in introns flanking the alternatively spliced cassette exons, suggesting strong evolutionary conservation. For 142 of the 449 alternatively spliced genes, WABA identified a highly conserved sequence in flanking introns upstream or downstream of 147 alternative cassette exons. <xref ref-type="fig" rid="pcbi-0020086-g001">Figure 1</xref> shows several examples of screen shots from the Intronerator browser of cDNA-confirmed <named-content content-type="genus-species">C. elegans</named-content> alternatively spliced isoforms (in blue) and a graphical representation of the WABA alignments between <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> (in purple). Dark purple regions, indicating high homology, can be seen in these introns either upstream or downstream of the alternative exon. Sometimes a single conserved element is identified by WABA, while for other cases multiple regions in the introns are conserved.</p><fig id="pcbi-0020086-g001" position="float"><label>Figure 1</label><caption><title>Images from the Intronerator Genome Browser Showing Alternatively Spliced Genes</title><p>Gene isoforms predicted by the Wormbase Consortium are shown in blue, and WABA homology alignments for <named-content content-type="genus-species">C. briggsae</named-content> to this region of the <named-content content-type="genus-species">C. elegans</named-content> genome are shown in purple. Dark purple indicates a region of WABA high homology, light purple corresponds to low homology, and white indicates no homology between species. Regions of alternatively spliced genes: (A) W01F3.1, (B) ZC477.9, (C) ZK637.8, (D) H24G06.1, and (E) C11D2.6 are shown.</p></caption><graphic xlink:href="pcbi.0020086.g001"/></fig><p>An important question is whether these 147 <named-content content-type="genus-species">C. elegans</named-content> alternative cassette exons are also alternatively spliced in <italic>C. briggsae.</italic> Due to a lack of <named-content content-type="genus-species">C. briggsae</named-content> transcript data, we have no direct evidence for alternative splicing of these exons. Therefore, we examined the <named-content content-type="genus-species">C. briggsae</named-content> homologs of these alternative exons for features of functional exons. A previous alignment of 8 MB of the <named-content content-type="genus-species">C. briggsae</named-content> genome with the complete <named-content content-type="genus-species">C. elegans</named-content> genome found that in coding exons there is 79.3% overall nucleotide identity [<xref rid="pcbi-0020086-b025" ref-type="bibr">25</xref>]. Consistent with these findings, cross-species alignment of these 147 alternative cassette exons revealed an average nucleotide identity between <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> of 81.8%. Amino acid identity of the open reading frame of these exons was 85.4%. Maintenance of the reading frame for these exons also appears to be well conserved: 57.0% of the 147 exons are the exact same size, 34.3% differ by a multiple of three nucleotides, and only 8.8% differ by a non-multiple of three nucleotides (unpublished data). The high conservation of nucleotide sequence, amino acid sequence, and maintenance of open reading frame is consistent with these being coding exons in <named-content content-type="genus-species">C. briggsae</named-content>.</p><p>One feature of alternative exons is that they are often flanked by weak splice sites which allow for splicing regulation (reviewed in [<xref rid="pcbi-0020086-b033" ref-type="bibr">33</xref>]). If these exons are alternatively spliced in <named-content content-type="genus-species">C. briggsae</named-content> we would expect that they would be flanked by 3&#x02032; and 5&#x02032; splice sites of similar strength. We used the UCSC Genome Table Browser (<ext-link ext-link-type="uri" xlink:href="http://www.genome.ucsc.edu">http://www.genome.ucsc.edu</ext-link>) in an attempt to automate rapid alignment of the splice junctions for these 147 exons in <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content>. For the majority of these alternative exons, this automated method was successful in identifying <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> 5&#x02032; and 3&#x02032; splice sites for each exon, and we used these data to compare the similarity of splice sites between the two species. For 5&#x02032; splice sites we examined the last three nucleotides of the exon and first six nucleotides of the intron, as these base-pair with U1 small nuclear RNA during initial 5&#x02032; splice-site recognition[<xref rid="pcbi-0020086-b031" ref-type="bibr">31</xref>]. We found that these nine nucleotides were completely identical for 47.1% of the alternative exons and differed by only one nucleotide in 27.1% of the exons. For the 3&#x02032; splice sites, we examined the conservation of the last six nucleotides of the intron and the first nucleotide of the exon, as these are a binding site for the <named-content content-type="genus-species">C. elegans</named-content> homolog of the heterodimeric U2 auxiliary factor (U2AF) involved in initial 3&#x02032; splice-site recognition [<xref rid="pcbi-0020086-b034" ref-type="bibr">34</xref>]. For 55.4% of the alternative exons examined, these seven nucleotides were completely identical, and for an additional 30.1% they differed by only one nucleotide. This strong conservation of splice-site sequence strength, along with regions of conservation in flanking introns, suggests that these exons may be substrates for alternative splicing in <named-content content-type="genus-species">C. briggsae</named-content>.</p></sec><sec id="s2b"><title>The Role of Conserved Elements in Alternative Splicing Regulation</title><p>There are many examples of regions of introns flanking alternatively spliced exons that function as <italic>cis</italic>-acting regulators of alternative splicing. We hypothesized that the reason that these high homology WABA-identified intronic elements have been conserved over 100 million years of evolution is that they may regulate alternative splicing. In order to test this hypothesis, we looked at the <italic>C. elegans/C. briggsae</italic> alignment of the alternatively spliced region of the <named-content content-type="genus-species">C. elegans</named-content> alpha(2) type IV collagen gene <italic>let-2. Let-2</italic> has two mutually exclusive alternative exons, exons 9 and 10, and their splicing pattern is evolutionarily conserved as far back as the distantly related parasitic nematode, <named-content content-type="genus-species">Ascaris suum</named-content> [<xref rid="pcbi-0020086-b035" ref-type="bibr">35</xref>]. Messages incorporate either exon 9 or exon 10 in a developmentally regulated manner; embryos predominantly use exon 9, adults predominantly use exon 10, and there is a gradual shift in the usage of these two exons during the larval stages [<xref rid="pcbi-0020086-b036" ref-type="bibr">36</xref>]. The WABA alignment of this 400-base intron identifies four conserved regions in the intron between exons 10 and 11 (<xref ref-type="fig" rid="pcbi-0020086-g002">Figure 2</xref>). To test the role of these conserved intronic elements in alternative splicing regulation, we employed a splicing reporter construct system containing the alternatively spliced region between exons 8 and 11 that mimics the developmental control of alternative splicing for this region when transformed into <named-content content-type="genus-species">C. elegans</named-content> [<xref rid="pcbi-0020086-b037" ref-type="bibr">37</xref>]. We mutated the first conserved element in this intron and monitored the developmental regulation of this splicing. Deletion of this conserved element, in which 28 of 34 bases have been conserved between <named-content content-type="genus-species">C. elegans</named-content> and <italic>C. briggsae,</italic> results in a major reduction in the usage of exon 10 in L4 animals (unpublished data). The identical effect on splicing of an even smaller deletion within this element, del1.2, is shown in <xref ref-type="fig" rid="pcbi-0020086-g002">Figure 2</xref>. For the L4 time point, the developmental stage at which we should detect maximal exon 10 usage, only minimal splicing of this exon is detected. Since mutation of this element results in only minimal exon 10 inclusion, we hypothesize that either adults produce a splicing factor that interacts with this element to promote exon 10 splicing or that embryos produce a splicing repressor that inhibits exon 10 splicing. In the past, researchers have identified splicing regulatory elements by creating a series of mutations across an intron. In this computational approach, the interspecies genome alignments led us directly to a small element required for the developmentally regulated switch in this splicing.</p><fig id="pcbi-0020086-g002" position="float"><label>Figure 2</label><caption><title>An Intronic Element Regulates <italic>let-2</italic> Alternative Splicing</title><p>The top of this figure shows the alignment of <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> sequences for the alternatively spliced region of <italic>let-2</italic>. Blue tracks indicate the splicing for the embryonic (top) and adult (bottom) isoforms. The purple track indicates homology between the <named-content content-type="genus-species">C. briggsae</named-content> and <named-content content-type="genus-species">C. elegans</named-content> genomes as determined by WABA. Dark purple tracks indicate regions of strong homology (&#x0003e;70%). The sequence of the first conserved element of intron 10 is shown. The box indicates the part deleted and replaced with the sequence GAA in the del1.2 splicing reporter construct. The lower left part of the figure shows the results of ³²P RT-PCR reactions with primers specific for the splicing reporter. Products for exon 9- or exon 10-containing messages are indicated. In embryos, only usage of exon 9 is detected for either reporter. At the L4 stage, 34% of the wild-type reporter messages contain exon 10 while del1.2 mutant messages contain only a trace amount of this exon.</p></caption><graphic xlink:href="pcbi.0020086.g002"/></fig></sec><sec id="s2c"><title>Analysis of a Database of Conserved Intronic Elements Flanking Alternatively Spliced Exons</title><p>As demonstrated in the previous section, evolutionarily conserved elements in introns flanking alternatively spliced exons can be important for alternative splicing. We used a computational approach to identify sequences that are more likely to occur in these conserved elements than in total intron sequences. As described above, we have identified 142 alternatively spliced genes in which the introns flanking an alternatively spliced cassette exon contain high homology regions with <named-content content-type="genus-species">C. briggsae</named-content> as defined by WABA. We extracted the highly conserved <named-content content-type="genus-species">C. elegans</named-content> sequences from these introns and put them into a database. The first or last seven nucleotides of introns were excluded from this dataset as these contain conserved signals for the constitutive splicing machinery [<xref rid="pcbi-0020086-b038" ref-type="bibr">38</xref>]. This dataset contained 537 conserved elements of average length, 38.5 bases (minimum length, seven bases, longest, 231 bases, and median length, 28 bases) for a total of 20,675 bases. See <xref ref-type="supplementary-material" rid="pcbi-0020086-st001">Table S1</xref> for a list of these elements and the alternatively spliced genes from which they were derived. We also generated a control dataset of all introns annotated in Wormbase genome sequence release WS 120.</p><p>Because many alternative splicing factors have binding preferences for relatively short sequence motifs [<xref rid="pcbi-0020086-b011" ref-type="bibr">11</xref>,<xref rid="pcbi-0020086-b013" ref-type="bibr">13</xref>,<xref rid="pcbi-0020086-b018" ref-type="bibr">18</xref>,<xref rid="pcbi-0020086-b028" ref-type="bibr">28</xref>&#x02013;<xref rid="pcbi-0020086-b032" ref-type="bibr">32</xref>], we decided to search the conserved intronic element dataset for short sequence motifs that appear more frequently here than in total introns. In order to identify sequences that are more prevalent in the conserved intronic regions bordering alternative exons, we counted the number of occurrences of every possible hexamer or pentamer motif in both our conserved intron element and total intron datasets. For each motif in the conserved element dataset, we determined the observed frequency, which was the count of each motif divided by the total number of possible motifs of that length in that dataset. We also determined an expected frequency based on the number of counts of each motif in the total intron dataset divided by the total number of possible motifs in that dataset. We then calculated an observed over expected (obs/exp) ratio for each motif, which was the observed frequency of each motif in the conserved element dataset divided by the expected frequency of each motif based on the total intron dataset. <xref ref-type="table" rid="pcbi-0020086-t001">Table 1</xref> shows the 40 top scoring pentamer and hexamer motifs in the conserved element dataset as determined by the obs/exp ratio value.</p><table-wrap id="pcbi-0020086-t001" content-type="1col" position="float"><label>Table 1</label><caption><p>Pentamer and Hexamer Sequence Motifs That Occur More Frequently in Conserved Intronic Elements Flanking Alternatively Spliced Exons than in the Total Intron Dataset</p></caption><graphic xlink:href="pcbi.0020086.t001"/></table-wrap><p>Several previously identified mammalian splicing regulatory sequences are present in our lists of high-scoring hexamers and pentamers. This is encouraging because many of the known mammalian splicing factors have <named-content content-type="genus-species">C. elegans</named-content> homologs. For example, the sequence TGCATG and its subset pentamer GCATG have been shown to be important intronic regulators of splicing for the mammalian fibronectin and calcitionin/CGRP genes [<xref rid="pcbi-0020086-b014" ref-type="bibr">14</xref>,<xref rid="pcbi-0020086-b015" ref-type="bibr">15</xref>,<xref rid="pcbi-0020086-b039" ref-type="bibr">39</xref>]. The mammalian FOX-1 protein selected the GCAUG sequence in a SELEX experiment, and it has been shown that the FOX-1 protein alters alternative splicing of several GCAUG-containing genes in transient transfection assays [<xref rid="pcbi-0020086-b013" ref-type="bibr">13</xref>]. FOX-2, a homolog of FOX-1, has also been shown to regulate splicing of transcripts containing UGCAUG in neuronal cell culture [<xref rid="pcbi-0020086-b012" ref-type="bibr">12</xref>]. FOX-1 was originally identified in <named-content content-type="genus-species">C. elegans</named-content> as a splicing factor of the transcript for the sex determination gene, <italic>xol-1</italic> [<xref rid="pcbi-0020086-b040" ref-type="bibr">40</xref>]. Another well studied mammalian motif is that of intronic GU dinucleotide repeats, which have been shown to regulate splicing of the human <italic>CFTR</italic> gene [<xref rid="pcbi-0020086-b041" ref-type="bibr">41</xref>] and serve as a binding site for the ETR-3 splicing regulatory factor [<xref rid="pcbi-0020086-b042" ref-type="bibr">42</xref>]. CU dinucleotide repeats serve as a binding site for the PTB family members in the intronic regions that regulate inclusion of the N1 exon of <italic>c-src</italic> [<xref rid="pcbi-0020086-b043" ref-type="bibr">43</xref>]. While targets for <named-content content-type="genus-species">C. elegans</named-content> alternative splicing factors might be inferred from their homology to mammalian proteins, very little experimental evidence exists for nematode splicing-factor binding sites. The <named-content content-type="genus-species">C. elegans</named-content> muscle-specific splicing factor SUP-12 regulates <italic>unc-60</italic> alternative splicing. This protein interacts with a GU-rich region, similar to the GU-rich motifs identified in <xref ref-type="table" rid="pcbi-0020086-t001">Table 1</xref>, of an <italic>unc-60</italic> intron [<xref rid="pcbi-0020086-b044" ref-type="bibr">44</xref>]. (For a review of other mammalian splicing factors with <named-content content-type="genus-species">C. elegans</named-content> homologs see [<xref rid="pcbi-0020086-b033" ref-type="bibr">33</xref>].) While many of the high-scoring conserved motifs that we identified match known mammalian splicing regulatory sequences, we have also identified new motifs that may also be involved in pre-mRNA splicing regulation in mammals as well as nematodes.</p><p>One important aspect of <named-content content-type="genus-species">C. elegans</named-content> introns is that they are shorter than introns in vertebrates. Half of <named-content content-type="genus-species">C. elegans</named-content> introns are below 60 nucleotides in length, too short to be spliced in vertebrates [<xref rid="pcbi-0020086-b038" ref-type="bibr">38</xref>]. However, many <named-content content-type="genus-species">C. elegans</named-content> introns are long and resemble vertebrate introns. In our alternatively spliced intron dataset, the median intron size is 263 bases (shortest is 43, longest is 10,719, and average is 561), indicating that these regulated introns generally belong in the larger intron class. We asked whether a similar analysis of the pentamers and hexamers found in the full introns flanking alternatively spliced exons would yield similar high-scoring motifs as our analysis of the conserved sequence subset of these introns. We did a pentamer and hexamer motif count of the full introns flanking the 147 alternatively spliced exons from which we previously extracted evolutionarily conserved elements. We counted the occurrence of every pentamer and hexamer in these full-length alternative introns. We ranked these motifs by the obs/exp ratio when we compared the occurrence of these motifs to their prevalence in our total intron dataset. The top 40 pentamer and hexamer motifs ranked by their obs/exp ratios are shown in <xref ref-type="table" rid="pcbi-0020086-t002">Table 2</xref>. Many of the top scoring pentamers and hexamers that are overrepresented in conserved elements in introns flanking alternatively spliced exons are also overrepresented in total introns flanking alternatively spliced exons (compare <xref ref-type="table" rid="pcbi-0020086-t001">Table 1</xref> with <xref ref-type="table" rid="pcbi-0020086-t002">Table 2</xref>). However, the obs/exp scores for the motifs in the alternative splicing introns is 2- to 3-fold lower than in the conserved element dataset extracted from these introns. For example, TGCATG has an observed to expected ratio of 3.35 in the whole introns that flank alternative exons but 8.87 for the conserved elements within those introns. TCTATC has a ratio of 2.55 in the total alternatively spliced intron analysis and 6.96 in the conserved element dataset analysis. The same holds true for high-scoring pentamers. CTATC goes from 1.85 for the observed to expected ratio in the introns flanking alternative exon dataset up to 4.23 in the conserved element dataset. From this analysis, the sequences of the introns flanking alternatively spliced exons can yield some data about potential pentamers and hexamers involved in alternative splicing regulation. However, limiting this analysis to WABA-conserved regions between <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> can substantially improve the signal to help us identify splicing regulatory elements.</p><table-wrap id="pcbi-0020086-t002" content-type="1col" position="float"><label>Table 2</label><caption><p>Pentamer and Hexamer Sequence Motifs That Occur More Frequently in the Introns Flanking Alternatively Spliced Exons than in Total Introns</p></caption><graphic xlink:href="pcbi.0020086.t002"/></table-wrap><p>We also used the database of introns flanking alternatively spliced exons to identify potential differences in the appearance of pentamer and hexamer motifs in the introns upstream or downstream of alternative exons. We grouped the introns upstream of the alternative exons and the introns downstream of the alternative exons separately into datasets and did motif counts and determined the obs/exp ratio for each when compared to the total intron dataset. These results are shown in <xref ref-type="table" rid="pcbi-0020086-t003">Tables 3</xref> and <xref ref-type="table" rid="pcbi-0020086-t004">4</xref>, with the top scoring pentamers and hexamers in the introns upstream and downstream of alternative exons ranked by their obs/exp ratios. Some of the high- scoring motifs show very little preference for the upstream or downstream intron. For example, high-scoring hexamer TGCATG has an obs/exp ratio of 3.90 in the downstream intron and 3.06 in the upstream intron. Others show a strong preference. CTCTCT and TCTCTC, likely binding sites for PTB, have obs/exp ratios of 5.15 and 4.84, respectively, in upstream introns, but 1.36 and 1.30 in downstream introns. TCTATC, which will be analyzed below, also has a preference for upstream introns, with an obs/exp ratio of 3.25 in upstream introns and 1.79 in downstream introns. Conversely, CCAACC has a strong preference for downstream introns, with an obs/exp ratio of 3.17 in the downstream intron and 1.29 in the upstream intron. In addition to identifying potential binding sites for splicing regulatory factors, these differences in appearance in introns upstream or downstream of alternatively spliced exons may hold some keys to function.</p><table-wrap id="pcbi-0020086-t003" content-type="2col" position="float"><label>Table 3</label><caption><p>Comparison of the Occurrence of Pentamer Motifs in Introns Upstream versus Downstream of Alternative Cassette Exons</p></caption><graphic xlink:href="pcbi.0020086.t003"/></table-wrap><table-wrap id="pcbi-0020086-t004" content-type="2col" position="float"><label>Table 4</label><caption><p>Comparison of the Occurrence of Hexamer Motifs in Introns Upstream versus Downstream of Alternative Cassette Exons</p></caption><graphic xlink:href="pcbi.0020086.t004"/></table-wrap><p>To determine whether these pentamers and hexamers are found in introns flanking all alternative cassette exons or are specific for those introns with WABA-conserved elements, another dataset was constructed from the introns flanking the alternative cassette exons of 307 genes lacking WABA-defined evolutionarily conserved intronic elements. A motif search similar to that described in the previous paragraph was conducted by dividing the new, non-conserved intron dataset into two sets: introns upstream of alternative exons and those downstream. Motifs were counted for each of the sets and compared to the total intron dataset to generate an obs/exp ratio and corresponding ranking for each element. It can be seen from <xref ref-type="table" rid="pcbi-0020086-t004">Table 4</xref> that the obs/exp scores of all of the top 30 hexamer motifs identified in conserved introns drop in the analysis of non-conserved introns, suggesting the motifs are indeed specific to introns containing conserved regions. For example, our top ranking hexamer from the conserved intron analysis, TGCATG, has an obs/exp score of 3.9 in downstream introns and 3.06 upstream. This same motif only scores 1.75 in downstream and 1.93 in upstream in non-conserved alternative introns (<xref ref-type="table" rid="pcbi-0020086-t004">Table 4</xref>). For CTCTCT and TCTCTC, our two top hexamer upstream motifs in introns with WABA-conserved elements, the obs/exp ratio drops over 3-fold in this analysis, and they rank as 579 and 688, respectively, in the upstream non-conserved intron set (<xref ref-type="table" rid="pcbi-0020086-t004">Table 4</xref>). Conversely, top scoring motifs from the non-conserved intron database are not highly represented in the conserved intron dataset. For example, GGCCAC, with an obs/exp ratio of 4.36 (<xref ref-type="supplementary-material" rid="pcbi-0020086-st002">Table S2</xref>) was the top scoring motif in the upstream non-conserved dataset. Strikingly, this same motif has an obs/exp score of 0.11 in total introns that contain WABA-conserved elements upstream of an alternatively spliced exon (unpublished data). This dramatically different representation of top scoring pentamers and hexamers in introns containing WABA-conserved elements versus those that do not suggests that the presence or absence of intronic WABA-defined conservation may define two distinct classes of alternatively spliced exons with distinct splicing regulatory mechanisms.</p><p>To ensure that the motifs identified in the conserved element dataset were truly overrepresented in this region and not a consequence of WABA creating bias towards GC-rich regions in the AT-rich nematode genomes, we analyzed the base composition of our datasets. The GC content of our WABA-defined conserved element dataset was 39%, the total intron dataset that contained these conserved elements was 34%, and the total intron reference dataset was 32%. While this slight variation in GC content might suggest that the WABA HMM may have created a bias, two arguments can be made against this. The first is that the high-scoring motifs in the WABA-conserved region were also highly represented in the total introns containing these motifs (<xref ref-type="table" rid="pcbi-0020086-t002">Table 2</xref>), consistent with these results not being derived from bias in GC content. The second is that GC enrichment bias is not observed in many of our highest scoring motifs. For example, TCTATC has a GC content of 33%, yet is highly enriched in conserved elements.</p></sec><sec id="s2d"><title>The Candidate Motifs (T)GCATG and (T)CTATC Are Required for <italic>unc-52</italic> Alternative Splicing Regulation</title><p>The <italic>C. elegans unc-52</italic> gene is a homolog of the mammalian extracellular matrix protein perlecan and contains 37 exons. Exons 16, 17, and 18 are alternatively spliced in a complex and regulated manner. These three exons each encode an Ig protein motif and can be included or skipped in any number of combinations in the final <italic>unc-52</italic> mRNA transcript [<xref rid="pcbi-0020086-b045" ref-type="bibr">45</xref>]. RT-PCR analysis shows <italic>unc-52</italic> is also alternatively spliced in <italic>C. briggsae,</italic> suggesting the splicing pattern may be controlled by similar methods (unpublished data). The use of a subset of these alternatively spliced forms is controlled by the splicing regulatory protein, MEC-8 [<xref rid="pcbi-0020086-b046" ref-type="bibr">46</xref>,<xref rid="pcbi-0020086-b047" ref-type="bibr">47</xref>]. The genetic regulation of <italic>unc-52</italic> alternative splicing makes this gene an attractive model for studying splicing regulation. Most importantly for this analysis, the introns flanking either side of alternative exon 16 contain regions of high nucleotide conservation as identified by WABA. A number of the top scoring pentamers and hexamers from our conserved element motif analysis are found in these conserved regions (<xref ref-type="fig" rid="pcbi-0020086-g003">Figure 3</xref>A).</p><fig id="pcbi-0020086-g003" position="float"><label>Figure 3</label><caption><title>Conserved Intronic Elements of <italic>unc-52</italic> Contain Putative Regulators of Alternative Splicing</title><p>(A) <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> sequence alignment is shown for the alternatively spliced portion of <italic>unc-52</italic>. Not all spliced isoforms are predicted by Wormbase software (blue); see <xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>A for observed alternative splicing patterns<italic>.</italic> PhastCons sequence alignment is shown with WABA-designated conservation in bold. Upper line of sequence is <italic>C. elegans; C. briggsae</italic> is below. High-scoring conserved motifs identified in our pentamer/hexamer analysis of conserved intronic elements flanking alternatively spliced exons, GCATG, TCTATC, CTATCC, CTATC, and TGCAC are underlined.</p><p>(B) Diagram of alternative splicing reporter constructs for testing putative <italic>cis</italic>-regulatory splicing motifs. Part of exon 15 through part of exon 19 of <italic>unc-52</italic> was cloned into a GFP/lacZ fusion vector with an <italic>unc-54</italic> promoter and nuclear localization sequence suitable for expression in <italic>C. elegans.</italic> Site-directed mutagenesis of the wild-type substrate was performed in order to test putative <italic>cis</italic>-splicing regulatory elements<italic>.</italic> A table of the splicing reporter constructs and their alterations is shown. Asterisks denote highly conserved intronic nucleotides deleted by site-directed mutagenesis. To maintain the intron length, yet remove motifs in question, a reporter was also made in which native sequence was replaced with the reverse complement sequence (shown in bold) and a HindIII site (italics) for diagnostic purposes.</p></caption><graphic xlink:href="pcbi.0020086.g003"/></fig><fig id="pcbi-0020086-g004" position="float"><label>Figure 4</label><caption><title>Deletion or Mutation of Top Scoring Pentamers or Hexamers Alters <italic>unc-52</italic> Alternative Splicing</title><p>(A) Autoradiogram of <italic>unc-52</italic> reporter gene alternative splicing. Left side shows ³²P RT-PCR analysis (with BamHI digest) performed on strains carrying each of the reporter constructs described in <xref ref-type="fig" rid="pcbi-0020086-g003">Figure 3</xref>B. The reporter is indicated at top of gel. Table on right shows the relative percentage of each alternatively spliced isoform produced from the indicated in vivo splicing reporter constructs. Arrows point to the corresponding band on the gel. Asterisk denotes a non-specific band resulting from RT-PCR.</p><p>(B) Graphical summary of <italic>unc-52</italic> in vivo splicing reporter assays. Horizontal axis indicates spliced mRNA isoform. Vertical axis represents the relative percentage of each isoform. Color key on figure indicates which bars correspond to which strain. Standard deviation and mean values were calculated based on a minimum of three independent RNA extractions and subsequent RT-PCRs for each reporter.</p></caption><graphic xlink:href="pcbi.0020086.g004"/></fig><p>To test the role of these motifs in alternative splicing of <italic>unc-52,</italic> we created a muscle-expressed alternative splicing reporter transgene derived from this region (<xref ref-type="fig" rid="pcbi-0020086-g003">Figure 3</xref>B). Expression studies using the native <italic>unc-52</italic> promoter to drive GFP expression have indicated that this extracellular matrix protein is expressed in the muscle and hypodermis [<xref rid="pcbi-0020086-b047" ref-type="bibr">47</xref>,<xref rid="pcbi-0020086-b048" ref-type="bibr">48</xref>]. However, studies of the <italic>unc-52</italic> splicing regulator <italic>mec-8</italic> have suggested that <italic>mec-8</italic> function on <italic>unc-52</italic> alternative splicing is mostly focused in hypodermal tissues [<xref rid="pcbi-0020086-b047" ref-type="bibr">47</xref>]. Our intention is to mimic the native gene's splicing as closely as possible, so we tested whether the alternative splicing of our muscle-specific <italic>unc-52</italic> reporter could be regulated by <italic>mec-8</italic>. RT-PCR analysis of our wild-type <italic>unc-52</italic> alternative splicing reporter in <italic>mec-8(+)</italic> (wild-type) and <italic>mec-8(e398)</italic> (mutant) backgrounds indicates a dramatic difference in splicing (<xref ref-type="supplementary-material" rid="pcbi-0020086-st003">Table S3</xref>). When the reporter is spliced in a <italic>mec-8(e398)</italic> mutant that lacks its RNA-binding domain, an increase in exon 18-containing transcripts is observed, as well as a 75% drop in abundance of 15-16-19 and 15&#x02013;19 isoforms. This closely mimics the effects of <italic>mec-8(e398)</italic> on native <italic>unc-52</italic> splicing [<xref rid="pcbi-0020086-b046" ref-type="bibr">46</xref>]. These results demonstrate that the tissue-specific expression of our <italic>unc-52</italic> reporter is an adequate representation of the alternative splicing of the native gene, and that it is also under the control of the splicing factor MEC-8.</p><p>In order to test the role of evolutionarily conserved intronic elements and motifs in the regulation of <italic>unc-52</italic> alternative splicing, we created <italic>unc-52</italic> alternative splicing reporter constructs with different combinations of deletions in these sequences and monitored their in vivo splicing by RT-PCR. Due to the similar size of exons 17 and 18, a restriction digest with BamHI was performed on the RT-PCR products. Spliced isoforms containing exon 18 would be cleaved by BamHI and thus run at smaller sizes on the gel (<xref ref-type="fig" rid="pcbi-0020086-g003">Figures 3</xref>B and <xref ref-type="fig" rid="pcbi-0020086-g004">4</xref>A). A summary of these results is shown in <xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>. All quantitation is based on the analysis of a minimum of three different RNA isolations from the indicated strains. In general, splicing reporters with deletions of the conserved intronic sequences flanking either side of exon 16 seemed to have the largest effect on three specific isoforms of mature <italic>unc-52</italic> mRNA, those containing exons 15-16-19, 15-16-18&#x02013;19, and 15-18-19 (<xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>B).</p><p>Our pentamer and hexamer motif analysis of conserved intronic elements flanking alternatively spliced exons identified GCATG and TGCATG as high-scoring motifs likely to be found in introns flanking either side of an alternative exon. Deletion of the GCATG upstream of alternative exon 16 from the <italic>unc-52</italic> splicing reporter resulted in nearly a doubling of the proportion of reporter transcripts containing exon 16 and a drop in the proportion using 15-18-19. This suggests that GCATG may play a role in repressing the inclusion of exon 16 (<xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>B).</p><p>A conserved intronic element just downstream of exon 16 contains another of our top scoring motifs, (T)CTATC. A deletion of this TCTATC had the opposite effect of deleting the upstream intronic GCATG; the proportion of exon 16-containing transcripts decreased while the amount of 15-18-19 increased. A larger deletion of this downstream motif (Deletion B) or a replacement of the motif with the reverse complement sequence (Mutation B) amplified this effect as summarized in <xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>B. Specifically, an approximately 75% reduction in exon 16- containing transcripts relative to wild-type can be seen when this conserved region is altered in the Mutation B construct. A similar drop of almost 85% is seen when this conserved element is deleted in the Deletion B reporter construct. This indicates that TCTATC is part of a regulatory element that enhances the inclusion of exon 16, and its loss by mutation leads to a decrease in exon 16 inclusion.</p><p>Perhaps most interesting was the effect produced by the double deletion of two conserved motifs flanking exon 16: the upstream GCATG and the downstream TCTATC. Individual deletions of these two motifs suggested they work in opposition to include or exclude exon 16. However, the double mutant causes a shift in splice pattern similar to, but much more dramatic, than that seen in the &#x00394;GCATG construct (<xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>B). The double mutant had such a dramatic increase in 15-16-19 levels that it became the new predominant isoform, making up 61% of the mRNAs. This result suggests that the conserved motifs on either side of exon 16 are responsible for the inclusion of this exon and that they work in collaboration and as part of a larger regulatory network to produce a multilayered regulation of <italic>unc-52</italic> splicing. In general, changes to the evolutionarily conserved intronic elements flanking exon 16 did not alter the relative amounts of the 15-17-19, 15-17-18&#x02013;19, and 15&#x02013;19 isoforms dramatically. These elements flanking exon 16, for the most part, appear to regulate the decision to form either the 15-16-19 isoform or the 15-18-19 isoform (<xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>B).</p></sec></sec><sec id="s3"><title>Discussion</title><p>Our comparative genomic analysis of <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> allowed us to identify 142 alternatively spliced genes that exhibit regions of high nucleotide conservation in introns flanking alternatively spliced exons. These conserved regions were then analyzed for pentamer and hexamer motifs present at a statistically higher level than found in total introns. Many of the high scorers on these lists matched known mammalian splicing regulatory elements. This indicates that this approach can find alternative splicing regulatory sequences, and it is consistent with our observations that there are <named-content content-type="genus-species">C. elegans</named-content> homologs for the major mammalian splicing regulatory factors (unpublished data). In addition to finding known splicing regulatory sequences, this approach identified potentially novel splicing regulatory sequences. We have confirmed that the sequence (T)CTATC is important for alternative splicing regulation of the <italic>unc-52</italic> gene.</p><p>The limited sequence conservation between <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> in introns flanking alternatively spliced exons contrasts with that seen in mammalian interspecies genome alignments. In mouse/human alignments of alternatively spliced human cassette exons, it was found that the 100 nucleotides of intron sequence flanking either side of alternative cassette exons tend to be highly conserved [<xref rid="pcbi-0020086-b021" ref-type="bibr">21</xref>,<xref rid="pcbi-0020086-b022" ref-type="bibr">22</xref>]. It is unclear from these mouse/human analyses which specific portion(s) of the introns immediately flanking these alternatively spliced exons are essential for alternative splicing regulation. Our analysis indicates that the WABA-identified regions of intronic homology flanking alternatively spliced exons are likely to be important for alternative splicing regulation. Therefore, the <italic>C. elegans/C. briggsae</italic> evolutionary distance, along with the sensitivity of WABA, provides for a more simplified method for pinpointing splicing regulatory sequences. One of the interesting questions resulting from our identification of intronic alternative splicing regulatory motifs in <named-content content-type="genus-species">C. elegans</named-content> is whether these will also function to regulate alternative splicing in mammals. It has been suggested that there is a difference in frequency of usage of constitutive intronic splicing enhancers between mammals and fish, which can explain why fish introns are not always spliced efficiently in mammalian cells [<xref rid="pcbi-0020086-b049" ref-type="bibr">49</xref>]. No similar study for alternative splicing regulatory elements has yet been done. Our data on <named-content content-type="genus-species">C. elegans</named-content> alternative splicing regulatory elements suggest that there are many similarities in alternative splicing regulation across metazoans. This is seen in the fact that many of the motifs we identified in worms are known to regulate alternative splicing in mammals. Whether novel elements we have identified, such as TCTATC, can also function in mammals has yet to be determined.</p><p>The identification of (T)GCATG as a top scoring alternative splicing regulatory motif demonstrated the usefulness of this approach, as many examples of this motif as a splicing regulatory element are present in the literature [<xref rid="pcbi-0020086-b014" ref-type="bibr">14</xref>,<xref rid="pcbi-0020086-b015" ref-type="bibr">15</xref>,<xref rid="pcbi-0020086-b050" ref-type="bibr">50</xref>]. (T)GCATG is statistically overrepresented in introns downstream of alternative exons conserved between human and mouse genomes [<xref rid="pcbi-0020086-b021" ref-type="bibr">21</xref>,<xref rid="pcbi-0020086-b022" ref-type="bibr">22</xref>] and may have a spatially conserved role in directing splice-site choice [<xref rid="pcbi-0020086-b023" ref-type="bibr">23</xref>]. On the other hand, in vitro studies have demonstrated that the splicing factor FOX-1 affects splicing of transcripts with GCATG either upstream or downstream of an alternative exon [<xref rid="pcbi-0020086-b013" ref-type="bibr">13</xref>]. One interesting finding of our <named-content content-type="genus-species">C. elegans</named-content> work is that GCATG in introns flanking alternatively spliced exons shows no preference for being upstream or downstream of alternative exons. This may be evidence that while the motifs that regulate alternative splicing between mammals and nematodes may be the same, they may be used in different ways to promote splicing.</p><p>A recent analysis of the <named-content content-type="genus-species">C. elegans</named-content> genome used a support vector machine to identify predictive features of <named-content content-type="genus-species">C. elegans</named-content> alternative exons [<xref rid="pcbi-0020086-b051" ref-type="bibr">51</xref>]. One of the predictive features of an alternatively spliced exon that they identified is the presence of several different hexamers in the introns surrounding alternative exons. In their supplemental materials (<ext-link ext-link-type="uri" xlink:href="http://www2.fml.tuebingen.mpg.de/raetsch/projects/RASE">http://www2.fml.tuebingen.mpg.de/raetsch/projects/RASE</ext-link>) they list the top 12 predictive hexamers in the introns upstream and downstream of the alternative exon. Comparison of their high-scoring results with our conserved element analysis indicates that many of the top scorers in both approaches are identical (for example, TGCATG, CTAACC, GTGTGT, CTCTCT, and TCTATC). That both their support vector machine and our comparative genomics methods yielded similar results helps to confirm the validity of both methods.</p><p>Comparing the motif searches from different classes of intronic sequence (total introns, WABA-conserved elements, entire conserved introns flanking alternative exons, and non-conserved introns flanking alternative exons) brought about a number of insights concerning the presence of certain motifs in introns. First, it can be noted that many of these sequences are still found, although at lower frequency, in total introns as well as those that are alternatively spliced. They may therefore, in a method analogous to SR protein-binding sites in exons, play a role in constitutive as well as alternative splicing (SR proteins reviewed in [<xref rid="pcbi-0020086-b052" ref-type="bibr">52</xref>]). Second, the comparison of motif analyses of introns lacking conserved elements and those containing WABA conservation reveals two dramatically different lists of overrepresented motifs. This suggests that there may in fact, be two separate classes of cassette exon splicing regulation, and WABA can help us distinguish between them.</p><p>Our splicing reporter system allowed us to directly test the role of the conserved elements GCATG and TCTATC flanking alternative exon 16 of the <italic>unc-52</italic> gene in vivo. Deletion of either of these elements individually led to measurable but opposing effects on the inclusion of this exon (<xref ref-type="fig" rid="pcbi-0020086-g004">Figure 4</xref>). These first results suggested the two <italic>cis</italic> elements might be counter-balancing each other to regulate splicing in this region. However, the creation of a double deletion of these two elements demonstrated that this was a naive assumption. The splicing program exhibited by the double deletion construct was similar to that seen from the &#x00394;GCATG construct: an upregulation of the exon 16-containing transcripts. However, the proportional increase of these isoforms was far more dramatic in the double than the single deletion. The 15-16-19 isoform became the predominant spliced transcript, as opposed to 15&#x02013;19 seen for all other splicing reporter constructs. This striking result is a good reminder that although these motifs can work independently and their deletion may produce measurable effects on splicing, they may work in a combinatorial way in vivo, and the effects of multiple deletions of important elements may be hard to predict. This type of combinatorial multifactor splicing regulatory mechanism has been described for several vertebrate genes including <italic>c-src</italic> [<xref rid="pcbi-0020086-b011" ref-type="bibr">11</xref>] and cardiac troponin T [<xref rid="pcbi-0020086-b017" ref-type="bibr">17</xref>]. While our comparative genomics approach gives us a method of identifying putative <italic>cis</italic>-regulatory elements of alternative splicing and our in vivo reporter assay provides a method for directly testing candidates, we still need more experimental information before we can predict the combinatorial effects of these elements on splicing.</p><p>Our comparative genomics analysis has not only provided a method for identifying intronic <italic>cis-</italic>regulatory elements, but also a starting point to investigate the mechanisms by which they control splicing. Consistent with current models of alternative splicing, <italic>cis</italic> elements are predicted to be binding sites for <italic>trans</italic> factors. Our identification of the novel splicing regulatory sequence (T)CTATC will lead us to subsequent experiments to identify its potential protein partner. The lists we generated of potential splicing regulatory elements include many sequences that have yet to be tested for a role in alternative splicing regulation. Experiments still need to be performed, similar to those done on <italic>unc-52</italic> and <italic>let-2,</italic> in which these elements are tested for a functional role in alternative splicing regulation. Due to the rapid release of new alignment programs and genomes, our list of conserved pentamers and hexamers will likely see some adjustments before all potential regulatory motifs have been tested. WABA has a limitation of needing to see relatively large sequence alignments in order to call a region as high homology (the minimum WABA high homology sequence run we detected was seven consecutive identical nucleotides). WABA high homology regions in introns flanking alternatively spliced exons were only detected for approximately 25% of alternative cassette exons in our dataset. The newer alignment algorithm PhastCons [<xref rid="pcbi-0020086-b053" ref-type="bibr">53</xref>], which uses an HMM to align multiple genomes using a smaller window size than that of WABA, can allow us to more accurately pinpoint smaller regions of conserved nucleotides within introns flanking additional alternatively spliced exons (unpublished data). The release of additional nematode genomes such as <named-content content-type="genus-species">Caenorhabditis remanei</named-content> should provide us with the prospect of creating a three-way nematode genomic alignment, allowing us even more accuracy in our regulatory motif predictions. By systematically testing the pentamers and hexamers on our list of conserved motifs, we may be able to confirm more alternative splicing regulatory motifs. A compilation of these results will provide us with a better comprehension of the rules governing this complex and essential process.</p></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Database construction.</title><p>A control dataset of introns was obtained by downloading from the UCSC Genome Browser (<ext-link ext-link-type="uri" xlink:href="http://www.genome.ucsc.edu">http://www.genome.ucsc.edu</ext-link>) all of the introns in the <named-content content-type="genus-species">C. elegans</named-content> genome as annotated in Wormbase release WS120. From the sequences of these 118,492 introns, we removed the first and last seven nucleotides to avoid constitutive splicing signals. We then counted the number of occurrences of every possible hexamer or pentamer motif in the database by using the EMBOSS Compseq program [<xref rid="pcbi-0020086-b054" ref-type="bibr">54</xref>]. For each motif in the database, we assigned a frequency score that was the number of occurrences of each motif divided by the total number of possible occurrences of motifs of that length in the dataset. This score for each possible motif in the control set was used in our analysis of both the introns that flank alternatively spliced exons and the conserved elements in these introns as our expected score. In order to obtain the obs/exp ratio for each motif in a test dataset, we divided the observed frequency score for each motif in that dataset by the expected frequency score for each motif based on the total intron control dataset.</p></sec><sec id="s4b"><title>Analysis of alternative exons.</title><p> <named-content content-type="genus-species">C. elegans</named-content> and <named-content content-type="genus-species">C. briggsae</named-content> alignments for the 147 alternative spliced exons with intronic conservation were obtained from the UCSC Genome Table Browser conservation track [<xref rid="pcbi-0020086-b055" ref-type="bibr">55</xref>]. The mean nucleotide percentage identity was computed from this alignment along with the number of insertions and deletions. In order to obtain the mean amino acid identity, all the 147 exons were compared to the complete <named-content content-type="genus-species">C. briggsae</named-content> genome (WormBase cb25.agp8) using TBLASTX [<xref rid="pcbi-0020086-b056" ref-type="bibr">56</xref>], and the mean amino acid identity was then calculated from these alignments.</p></sec><sec id="s4c"><title>Generation of <italic>let-2</italic> splicing reporter constructs.</title><p>We have previously described the development of an in vivo splicing reporter for the alternatively spliced region of the <italic>let-2</italic> gene fused to GFP and expressed in muscle cells [<xref rid="pcbi-0020086-b037" ref-type="bibr">37</xref>]. Site-directed mutagenesis of the wild-type reporter for <italic>let-2</italic> alternative splicing from that paper was used to delete the core of the first conserved element, indicated in <xref ref-type="fig" rid="pcbi-0020086-g002">Figure 2</xref>, and replace it with GAA. Methods for detecting alternative splicing of this transcript with ³²P-labeled primers using RT-PCR, denaturing polyacrylamide gel electrophoresis, and autoradiography have been previously described [<xref rid="pcbi-0020086-b037" ref-type="bibr">37</xref>].</p></sec><sec id="s4d"><title>Generation of <italic>unc-52</italic> splicing reporter constructs.</title><p>The exon 15&#x02013;19 region of <italic>unc-52</italic> was PCR-amplified from wild-type (N2) <named-content content-type="genus-species">C. elegans</named-content> genomic DNA using primers 5&#x02032;-GGAATTCGATGAGTACATCTGTATCGC and 5&#x02032;-GGAATTCACATCTGAACTGATGTCGCTC for cloning into the vector pPD96.02, developed by Andrew Fire's lab. This vector contains the <italic>unc-54</italic> body wall muscle-specific promoter driving expression of a green fluorescent protein (GFP)/&#x003b2; galactosidase (lacZ) fusion protein with an N-terminal nuclear localization signal derived from SV40 T antigen. The final 127 nucleotides of exon 15 through the first 173 nucleotides of exon 19 of <italic>unc-52</italic> and all the genomic sequence in between were cloned into a unique EcoRI site, 16 codons before the end of the lacZ open reading frame in the final exon of the fusion protein. The Kunkel method of site-directed mutagenesis was used to create the larger deletion and mutations for the plasmid constructs Mutation B and Deletion B [<xref rid="pcbi-0020086-b057" ref-type="bibr">57</xref>]. Pentamer and hexamer deletion constructs of the conserved <italic>unc-52</italic> intron sequence were created using the Quickchange site-directed mutagenesis kit (Stratagene, La Jolla, California, United States) according to the manufacturer's instructions. Animals carrying these constructs as extra-chromosomal arrays were generated by standard injection/transformation [<xref rid="pcbi-0020086-b058" ref-type="bibr">58</xref>]. Transformed N2 animals were identified by GFP-expression in nuclei of body wall muscle cells.</p></sec><sec id="s4e"><title>Growth of <named-content content-type="genus-species">C. elegans</named-content> strains and isolation of RNA.</title><p> <named-content content-type="genus-species">C. elegans</named-content> strains were grown on NGM agar plates using standard methods [<xref rid="pcbi-0020086-b059" ref-type="bibr">59</xref>]. RNA was extracted from worms as previously described [<xref rid="pcbi-0020086-b060" ref-type="bibr">60</xref>]. Strain CB938 carrying the <italic>mec-8(e398)</italic> mutant allele was obtained from the <named-content content-type="genus-species">C. elegans</named-content> Genetics Center, a National Institutes of Health funded Center for Research Resources at the University of Minnesota.</p></sec><sec id="s4f"><title>Production of cDNAs.</title><p>cDNAs of the <italic>unc-52</italic> splicing reporters were made in 25-&#x003bc;l reaction mixtures. The annealing step contained 3 &#x003bc;g total RNA and 25-pmol oligodeoxynucleotide primer complementary to the lacZ construct (5&#x02032;- GTTGAAGAGTAATTGGACTTA-3&#x02032; for <named-content content-type="genus-species">C. elegans</named-content> and 5&#x02032;- AACTGGTGTCGCTCTCCT-3&#x02032; for <named-content content-type="genus-species">C. briggsae</named-content>) in 17-&#x003bc;l final volume. These were heated to 94 &#x000b0;C for 2 min and cooled to room temperature for 10 min. The rest of the reverse transcription reaction components were added to reach a final volume of 25 &#x003bc;l. The final reaction mixtures consisted of 1 mM each of dATP, dCTP, dGTP, and dTTP; 1 U RNA Guard (Promega, Madison, Wisconsin, United States); 1 &#x000d7; AMV RT buffer (Promega); and 10 U AMV reverse transcriptase (Promega). Reactions were incubated at 37 &#x000b0;C for 1.5 h and stored at &#x02212;20 &#x000b0;C.</p></sec><sec id="s4g"><title>Polymerase chain reaction.</title><p>1.0 &#x003bc;l of the cDNA reaction mixture was used as the template in 25-&#x003bc;l PCR reaction mixtures. 1.0 pmoles of the same oligonucleotide used in the reverse transcription reactions was added to PCR reaction mixtures along with ³²P-labeled 5&#x02032; lacZ vector-specific primer (5&#x02032;-CTGGAGCCCGTCAGTATCGGC-3&#x02032; for <named-content content-type="genus-species">C. elegans</named-content> and 5&#x02032;-ATTCGTTGCTGGGTCCCAGG-3&#x02032; for <named-content content-type="genus-species">C. briggsae</named-content>). The reaction mixtures also contained 1 &#x000d7; PCR buffer, 0.25 mM of each of the four dNTPs, and <italic>Taq</italic> DNA polymerase. The 5&#x02032; oligo was labeled with (&#x003b3;-³²P) ATP by T4 polynucleotide kinase. Reaction mixtures were incubated for 25 cycles at 94 &#x000b0;C for 1.0 min, 59 &#x000b0;C for 1.0 min, and 72 &#x000b0;C for 1.0 min. 2.0 &#x003bc;l of PCR product were digested with the restriction enzyme BamHI. There is a unique BamH1 site in <italic>unc-52</italic> exon 18, and this digestion step allows us to distinguish the different alternatively spliced isoforms more clearly on the gel. Digested PCR products were separated on 40-cm long, 0.4-mm thick, 5% or 6% polyacrylamide urea gels in TBE buffer. Gels were dried onto filter paper. These were then visualized using a Molecular Dynamics PhosphorImager (Sunnyvale, California, United States). Relative splice-site usage was quantified using ImageQuant software as previously described [<xref rid="pcbi-0020086-b037" ref-type="bibr">37</xref>].</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting Information</title><supplementary-material content-type="local-data" id="pcbi-0020086-st001"><label>Table S1</label><caption><title>Complete Database of Highly Conserved Intronic Elements Used in Motif Analysis</title><p>Column (A) indicates <named-content content-type="genus-species">C. elegans</named-content> gene name. Columns (B&#x02013;AI) list the WABA high homology sequences found in <named-content content-type="genus-species">C. elegans</named-content> introns flanking alternative exons of the corresponding gene.</p><p>(51 KB XLS)</p></caption><media xlink:href="pcbi.0020086.st001.xls"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020086-st002"><label>Table S2</label><caption><title>Complete Table Detailing Ranks of Pentamer and Hexamer Motifs in Introns Lacking Evolutionary Conservation Surrounding 307 Alternatively Spliced Exons</title><p>On both worksheets: Column (A) represents rank of motif from search based on obs/exp frequency in introns lacking evolutionary conservation downstream of alternative introns. Column (B) denotes motif and (C) contains the number of times the motif was observed in the dataset. (D) The observed frequency of the motif in the dataset. (E) Contains the expected frequency of motif based on total intron dataset. Column (F) is the ratio of observed frequency to expected frequency. Columns (H&#x02013;M) are data from the same analysis in the same order for upstream introns.</p><p>(1.2 MB XLS)</p></caption><media xlink:href="pcbi.0020086.st002.xls"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020086-st003"><label>Table S3</label><caption><title>Each Row in the Table Represents a Different Spliced Isoform of the <italic>unc-52</italic> Reporter Constructs, and Each Column Denotes the Reporter Being Monitored</title><p>The first six columns show the relative percentage of alternative splice-site usage of the specified reporter in a <italic>mec-8(+)</italic> background as quantified from ³²P RT-PCR<italic>.</italic> The seventh column shows results of the wild-type sequence reporter that has been crossed into <italic>mec-8(e398)</italic> mutant animals.</p><p>(185 KB TIF)</p></caption><media xlink:href="pcbi.0020086.st003.tif"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material></sec>

Evolutionary and Physiological Importance of Hub Proteins

<p>It has been claimed that proteins with more interaction partners (hubs) are both physiologically more important (i.e., less dispensable) and, owing to an assumed high density of binding sites, slow evolving. Not all analyses, however, support these results, probably because of biased and less-than reliable global protein interaction data. Here we provide the first examination of these issues using a comprehensive literature-curated dataset of well-substantiated protein interactions in <italic>Saccharomyces cerevisiae.</italic> Whereas use of less reliable yeast two-hybrid data alone can reject the possibility that local connectivity correlates with measures of dispensability, in higher quality datasets a relatively robust correlation is observed. In contrast, local connectivity does not correlate with the rate of protein evolution even in reliable datasets. This perhaps surprising lack of correlation with evolutionary rate appears in part to arise from the fact that hub proteins do not have a higher density of residues associated with binding. However, hub proteins do have at least one other set of unusual features, namely rapid turnover and regulation, as manifest in high mRNA decay rates and a large number of phosphorylation sites. This, we suggest, is an adaptation to minimize unwanted activation of pathways that might be mediated by adventitious binding to hubs, were they to actively persist longer than required at any given time point. We conclude that hub proteins are more important for cellular growth rate and under tight regulation but are not slow evolving.</p>

<contrib contrib-type="author"><name><surname>Batada</surname><given-names>Nizar N</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><contrib contrib-type="author"><name><surname>Hurst</surname><given-names>Laurence D</given-names></name><xref ref-type="aff" rid="aff2">2</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><contrib contrib-type="author"><name><surname>Tyers</surname><given-names>Mike</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff3">3</xref></contrib>

PLoS Computational Biology

<sec id="s1"><title>Introduction</title><p>Protein-interaction networks may be scale-free networks [but see 1]. Unlike random networks in which the number of connections between entities follows a Poisson distribution, in a scale-free network the distribution of the number of connections follows a power law, such that a few members called hubs have very large numbers of connections. The properties of these hub proteins are of particular interest, not least because they may be good targets for antimicrobial agents. How, if at all, are the hub proteins different from other proteins? Intuitively, one might expect many differences. As some classes of protein&#x02013;protein interaction sites are slow evolving [<xref rid="pcbi-0020088-b002" ref-type="bibr">2</xref>], proteins with many partners might be expected to be slower evolving, as claimed [<xref rid="pcbi-0020088-b003" ref-type="bibr">3</xref>&#x02013;<xref rid="pcbi-0020088-b006" ref-type="bibr">6</xref>]. Likewise, it seems intuitive that hub proteins may be more likely to be essential (i.e., knockout-inviable), also as claimed [<xref rid="pcbi-0020088-b007" ref-type="bibr">7</xref>&#x02013;<xref rid="pcbi-0020088-b010" ref-type="bibr">10</xref>]. Similarly, we might expect there to be a correlation between growth rate of cells lacking a given protein and the number of partners of that protein.</p><p>To this list we should like to add a further possibility, namely that to minimize potentially hazardous cross talk, temporal control of the abundance and activity of hubs needs to be regulated tightly to enable continuous equilibration with binding partners and curb excessive flux through certain pathways. As such, we expect the mRNAs might be adapted to be quickly removed when synthesis of the protein is no longer needed, and the hub proteins themselves should be regulated by phosphorylation more than the average protein.</p><p>While the above features may appear intuitively reasonable, it is far from clear that any of the prior claims are robust. The main problem is the source of protein&#x02013;protein interaction data. The large datasets applied to these problems have been derived from high-throughput experiments, which, in the case of protein&#x02013;protein interactions are known to have both high false-positive [<xref rid="pcbi-0020088-b011" ref-type="bibr">11</xref>,<xref rid="pcbi-0020088-b012" ref-type="bibr">12</xref>] and high false-negative rates [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>]. More generally, use of yeast two-hybrid data often fail to replicate results derived from alternative sources [see e.g., 14&#x02013;16]. As regards rates of protein evolution, expression level is by far the strongest predictor of rates of evolution [<xref rid="pcbi-0020088-b017" ref-type="bibr">17</xref>,<xref rid="pcbi-0020088-b018" ref-type="bibr">18</xref>]. There remains debate as to whether, when controlling for rates of expression, the more connected proteins have rates of evolution any different from the average [<xref rid="pcbi-0020088-b003" ref-type="bibr">3</xref>,<xref rid="pcbi-0020088-b015" ref-type="bibr">15</xref>,<xref rid="pcbi-0020088-b019" ref-type="bibr">19</xref>,<xref rid="pcbi-0020088-b020" ref-type="bibr">20</xref>]. Similarly, while some studies suggest that protein&#x02013;protein interaction networks are scale-free and hubs tend more commonly to be the essential parts of the network [<xref rid="pcbi-0020088-b007" ref-type="bibr">7</xref>&#x02013;<xref rid="pcbi-0020088-b010" ref-type="bibr">10</xref>], both the scale-free nature of yeast protein&#x02013;protein interaction network [<xref rid="pcbi-0020088-b001" ref-type="bibr">1</xref>] and the relationship between dispensability and position in the network [<xref rid="pcbi-0020088-b014" ref-type="bibr">14</xref>] may also be artifacts of biased data.</p><p>Here we re-evaluate these issues, taking advantage of a recent effort that assembled a set of 11,334 interactions obtained by systematically curating the past 30 years of <named-content content-type="genus-species">Saccharomyces cerevisiae</named-content> primary literature [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>]. This we refer to as the literature-curated protein interaction (LC-PI) dataset. Specifically, we examine three issues: are more highly connected proteins more &#x0201c;important,&#x0201d; do they evolve at lower rates, and are they more tightly regulated? Our premise is that interactions reported in focused primary papers are inherently more reliable than high-throughput interactions, not least because low-scale experiments are done by experts in the field and the validity of interactions are scrutinized by peer review; moreover, multiple contextual information and other self-consistency checks and validations normally support the demonstration of these physical interactions, thereby reducing the false-positive error rate. For any correlation to be real it should then be transparent in this dataset.</p><sec id="s1a"><title>Comparing Literature-Curated and High-Throughput Data</title><p>The LC-PI and high-throughput protein interaction (HTP-PI) datasets are in many regards rather different. While, for example, the two are supposed to be measuring the same attribute (i.e., the identity and number of different proteins a given protein interacts with), the correlation between the two sets as regards number of interactants of each protein, although naturally highly significant, is relatively modest (Spearman rank correlation <italic>r</italic> = 0.37). The two datasets also disagree substantially on which proteins are highly connected. Comparing, for example, the top 10% by connectivity in the two datasets we find only 29% in common (see <xref ref-type="supplementary-material" rid="pcbi-0020088-sd001">Dataset S1</xref> for LC hubs and <xref ref-type="supplementary-material" rid="pcbi-0020088-sd002">Dataset S2</xref> for HTP-PI hubs). Restricting to the top 5%, the figure drops to 24%. Moreover, among the bait proteins common to both sets, ~70% of interactions reported in the literature were absent in HTP-PI [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>].</p><p>We can also ask if known biases affect both sets equally. Some affinity methods (e.g., tandem affinity purification) will preferentially capture interactions for highly expressed proteins. As these dominate the HTP datasets, it is no surprise that there exists a strong positive correlation between abundance or rate of expression (measured by codon adaptation index [CAI]) and connectivity (abundance: connectivity, <italic>r</italic> = 0.19, <italic>n</italic> = 3,001, <italic>p</italic> &#x0003c; 0.0001; CAI: connectivity, <italic>r</italic> = 0.19, <italic>n</italic> = 4,169, <italic>p</italic> &#x0003c; 0.0001). In the LC dataset, in contrast, this effect is much diminished (abundance: connectivity, <italic>r</italic> = 0.046, <italic>n</italic> = 2,464, <italic>p</italic> = 0.02; CAI: connectivity, <italic>r</italic> = 0.037, <italic>n</italic> = 4,169, <italic>p</italic> = 0.034). Indeed, if we look at essential singleton genes alone, for example, the effect goes away in the LC set (abundance: connectivity, <italic>r</italic> = &#x02212;0.006, <italic>n</italic> = 185, <italic>p</italic> = 0.93; CAI: connectivity, <italic>r</italic> = &#x02212;0.002, <italic>n</italic> = 242, <italic>p</italic> = 0.98), while remaining relatively robust in the HTP set (abundance: connectivity, <italic>r</italic> = 0.2, <italic>n</italic> = 204, <italic>p</italic> = 0.004; CAI: connectivity, <italic>r</italic> = 0.12, <italic>n</italic> = 260, <italic>p</italic> = 0.05). As a correlation between abundance and connectivity is an expected artifact of some affinity methods, the weakening of this signal provides some reassurance that the LC dataset is less biased. On an anecdotal level, consider, for example, the chaperone Hsp70, which binds to and facilitates the correct folding of many proteins. The Hsp70 family members Ssa1/Yal005c and Ssa3/Ybl075c have very high connectivity in the LC dataset (44 and 18 interactions, respectively) but not in HTP dataset (two interactions for each).</p><p>In other regards, the two networks show similar properties. For example, the LC-PI dataset shows a scale-free connectivity distribution [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>], just as HTP datasets do [<xref rid="pcbi-0020088-b021" ref-type="bibr">21</xref>&#x02013;<xref rid="pcbi-0020088-b024" ref-type="bibr">24</xref>]. While LC interactions are more likely to be real, we note that there are, nonetheless, inevitable selection biases present in curation data. Although the LC dataset represent over half the genes in yeast, it tends to favor more important or conserved proteins [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>], these in general being more closely scrutinized in biology. Given the potentially confounding differences between different datasets, we next examine how robust various conclusions drawn to date from different data sources might be.</p></sec></sec><sec id="s2"><title>Results</title><sec id="s2a"><title>Highly Connected Proteins Are Less Dispensable</title><p>Recently there has been dispute as to whether more highly connected proteins are more likely to be essential. Notably, Coulomb et al. [<xref rid="pcbi-0020088-b014" ref-type="bibr">14</xref>] have suggested that prior claims for such a relationship [<xref rid="pcbi-0020088-b007" ref-type="bibr">7</xref>&#x02013;<xref rid="pcbi-0020088-b010" ref-type="bibr">10</xref>] may be dataset artifacts. Intuitively, one would expect that loss of a highly connected protein would be more detrimental to the cell than the loss of lowly connected proteins; as in the former case, more processes would be affected than in the later case. Are then highly connected proteins more likely to be essential and for those that are nonessential do the relevant knockout strains show lower growth rates? To address these issues we made use of systematic knockout studies. As previously noted, use of laboratory differential growth rate due to gene-knockout as an estimate of dispensability may not be without faults: many proteins are likely to perform functions that are important in the environment relevant to yeast evolution, but superfluous in the laboratory conditions in which growth rates are measured [<xref rid="pcbi-0020088-b025" ref-type="bibr">25</xref>]. With this caveat in mind, we define the importance of a gene by the fitness reduction caused by the deletion of the gene in standard laboratory conditions.</p></sec><sec id="s2b"><title>For Nonessential Genes, Connectivity Correlates with Knockout Growth Rate</title><p>In both the LC and HTP datasets the more connected a protein the lower the growth rate of the knockout strain, assuming the strain is capable of growth (<xref ref-type="fig" rid="pcbi-0020088-g001">Figure 1</xref>). This effect is also manifest for singleton genes (<xref ref-type="fig" rid="pcbi-0020088-g001">Figure 1</xref> and <xref ref-type="table" rid="pcbi-0020088-t001">Table 1</xref>). However, if expression level for the analysis of singleton genes in the LC dataset is accounted for by Spearman partial correlation [<xref rid="pcbi-0020088-b026" ref-type="bibr">26</xref>], the connectivity effect becomes non-significant when controlled by protein abundance but not when controlled by CAI [<xref rid="pcbi-0020088-b027" ref-type="bibr">27</xref>]. This weakening of the result when employing protein abundance as the controlling variable is not, however, likely to be owing to the effects of the covariate, as among the nonessential genes, connectivity and expression parameters do not significantly correlate in the LC dataset (unpublished data). Instead, this increase in <italic>p-</italic>value appears to be owing to a reduced sample size in the covariate-controlled analyses. Two pieces of evidence support this suggestion. First, the value of <italic>r</italic> remains largely unaffected by covariate control. Second, we performed simulations to mimic the effects of reducing the sample size. If there were <italic>N</italic> data points in the non-covariate controlled analysis and <italic>M</italic> in the covariate controlled tests (i.e., <italic>N</italic> &#x0003e; <italic>M</italic>), we randomly selected <italic>M</italic> from <italic>N,</italic> performed a Spearman rank correlation, and asked how often the observed <italic>p-</italic>value was the same or greater than that observed in the covariate-controlled analysis. In both datasets of growth rate data, we failed to reject the null hypothesis that the increase in <italic>p-</italic>value was owing to a reduction in sample size (unpublished data).</p><fig id="pcbi-0020088-g001" position="float"><label>Figure 1</label><caption><title>Relationship between Connectivity and Dispensability</title><p>Scatter plots of growth defect upon homozygous deletion of a gene and the natural log of connectivity of the gene in the interaction network. Growth rate data are from Steinmetz et al. [<xref rid="pcbi-0020088-b045" ref-type="bibr">45</xref>].</p></caption><graphic xlink:href="pcbi.0020088.g001"/></fig><table-wrap id="pcbi-0020088-t001" content-type="1col" position="float"><label>Table 1</label><caption><p>Correlations between Connectivity and Fitness</p></caption><graphic xlink:href="pcbi.0020088.t001"/></table-wrap><p>Might the above results be dataset artifacts? The LC data are, for example, enriched for functionally important proteins that make up the translational machinery. The ribosome and other components of the translational machinery are highly expressed and central for almost all the cellular functions. Removing translation-associated proteins did not change the conclusion (<xref ref-type="supplementary-material" rid="pcbi-0020088-st001">Table S1</xref>A) but instead increased the correlation strength. The stronger correlation for non-ribosomal data can be explained because ribosomal proteins have an average connectivity of 4.8 while the mean connectivity of all proteins is approximately 7. Thus, loss of these lowly connected, but functionally important proteins increases the strength of the correlation.</p><p>The LC data contain interactions assessed using various different experimental methods [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>]; however, affinity purification and two-hybrid based methods account for most of the interactions. These two protocols tend to capture different sorts of proteins. Notably, for both sets of fitness data, the mean fitness of knockouts for proteins in the yeast two-hybrid data were on average higher than for proteins in affinity purification data (<italic>p</italic> &#x0003c; 10^&#x02212;10 for each fitness dataset). This may well reflect the greater cellular requirement for proteins in stable complexes, which are more often being captured by affinity methods. For instance, 196 out of 272 translation-related proteins are present in the affinity data, but only 52 out of 272 of these proteins are present in the two-hybrid data. It is then relevant to ask whether this biased sampling might affect conclusions. To this end, we consider only those proteins from the LC set that are found in both the yeast two-hybrid and affinity capture assays. The negative correlation between connectivity and fitness remains robust (<xref ref-type="supplementary-material" rid="pcbi-0020088-st001">Table S1</xref>B). Indeed, in this instance the correlation increases from an <italic>r²</italic> of 1%&#x02013;2% to approximately 10%.</p><p>All the above results suggest that the more connected a protein might be, the greater the impact on fitness when deleted, assuming the knockout can grow. However, to this conclusion we add one note of caution. We can ask about two sub datasets of the LC set: those being the proteins found only by the yeast two-hybrid method and those found only by the affinity purification method. Surprisingly, in both there is a lack of correlation (<xref ref-type="supplementary-material" rid="pcbi-0020088-st001">Table S1</xref>A). One possible reason for the appearance of correlation upon merging of the two types of data is that the mean fitness and mean connectivity of the two sets of data are different. As noted above, the mean fitness of knockouts for proteins in the yeast two-hybrid data was on average higher than for proteins in affinity purification data. Likewise, for proteins only in yeast two-hybrid set, the mean connectivity is 2.16, while for proteins only in the affinity set the mean is 5.2 (<italic>p</italic> &#x0003c; 0.0001, Mann-Whitney U test). Hence, we have one cluster of proteins of high average fitness and low average connectivity and another of lower average fitness and higher connectivity. Merging two such datasets would lead to a negative correlation while none need be seen in either sub datasets. To ask whether the correlation is then real or an artifact of merging potentially biased data, we analyzed a high confidence set of interactions based on the LC set but requiring multi-validation. For those proteins found in the affinity-derived set alone in the high confidence set, we again find a robust negative correlation (<italic>r</italic> = &#x02212;0.24 and <italic>p</italic> &#x0003c; 0.001). We suggest therefore that the effect is real but easily lost when the data are noisy.</p></sec><sec id="s2c"><title>Essential Genes Are More Highly Connected</title><p>As previously reported [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>] in both the HTP and the LC datasets, essential genes have on average more partners, i.e., higher connectivity (Mann-Whitney U test comparing essential and nonessential: LC-PI data mean natural log connectivity essential = 1.945 &#x000b1; 0.03, for nonessentials mean = 1.062 &#x000b1; 0.02, <italic>p</italic> &#x0003c; 0.0001; for HTP-PI data, mean essentials = 1.55 &#x000b1; 0.04, for nonessentials, mean = 0.96 &#x000b1; 0.02, <italic>p</italic> &#x0003c; 0.0001). This is also true if we analyze singleton genes alone (Mann Whitney U test, <italic>p</italic> &#x0003c; 0.0001 for both datasets). Removal of ribosomal proteins does not alter these conclusions (unpublished data). Controlling for protein abundance, the essential genes have higher connectivity than nonessential genes (ANCOVA of log abundance versus natural log of connectivity between essentials and nonessentials, for LC-PI set, <italic>F</italic> = 40.7, <italic>p</italic> &#x0003c; 0.0001, for HTP-PI set, <italic>F</italic> = 14.4, <italic>p</italic> = 0.002).</p><p>Curiously, if we restrict analysis to proteins found just in the yeast two-hybrid analyses, there is only a weak tendency for there to be a difference between the essentials and the nonessentials in their connectivity (Essentials mean connectivity: 2.333 &#x000b1; 0.267, <italic>n</italic> = 42; Nonessentials mean connectivity: 2.12 &#x000b1; 0.096, <italic>n</italic> = 573; Mann Whitney U test, <italic>p</italic> = 0.056). In contrast, in the proteins identified by the affinity capture methods alone, the result remains highly robust (essentials mean connectivity: 7.85 &#x000b1; 0.554, <italic>n</italic> = 301, nonessentials mean connectivity: 3.65 &#x000b1; 0.18, <italic>n</italic> = 754; Mann Whitney U test, <italic>p</italic> &#x0003c; 0.0001). Control for abundance or expression measured by CAI does not alter the later conclusion (ANCOVA: <italic>p</italic> &#x0003c; 0.0001). This suggests that the inability of Coloumb et al. [<xref rid="pcbi-0020088-b014" ref-type="bibr">14</xref>] to detect an effect was a consequence of employing yeast two-hybrid data. This would make sense as yeast two-hybrid method is likely to miss many true interactions and so underestimate the number of interactants of the more highly connected proteins. As, moreover, the difference between essentials and nonessentials is seen in all other datasets, we suggest that the connectivity effect is real and is not owing to covariance with expression rate.</p></sec><sec id="s2d"><title>Hubs Do Not Evolve Slower than Non-Hubs</title><p>The claim that highly connected proteins evolve slower than others could be in large part an artifact of HTP datasets [<xref rid="pcbi-0020088-b015" ref-type="bibr">15</xref>], as these typically report more interactions for proteins that are also highly expressed [<xref rid="pcbi-0020088-b011" ref-type="bibr">11</xref>]; moreover, expression rate is a robust predictor of rates of evolution [<xref rid="pcbi-0020088-b017" ref-type="bibr">17</xref>]. As highly connected proteins are more likely to be essentials and essentials may evolve slower than nonessentials, we therefore analyze the essential and nonessential genes separately so as to account for any differences between these two in mean rate of protein evolution. One reason many proteins may be nonessential is because they have a duplicate gene (or genes) present in the same genome. As duplicated genes may themselves have unusual rates of evolution [<xref rid="pcbi-0020088-b028" ref-type="bibr">28</xref>] (possibly owing to relaxation of functional constraints or to positive selection promoting diversification), it is most valuable to ask whether any of the above results are robust to analysis of singleton genes. For analysis of both duplicates and the dataset en mass see <xref ref-type="supplementary-material" rid="pcbi-0020088-sg001">Figures S1</xref> and <xref ref-type="supplementary-material" rid="pcbi-0020088-sg002">S2</xref>.</p><p>Of 3,289 proteins in the LC-PI dataset, evolutionary rate data, i.e., rate of non-synonymous substitution per site, were obtained for 306 singleton nonessentials. Spearman rank correlation, a non-parametric measure that is robust to outliers, between the connectivity and evolutionary rate for these proteins, was not significant (<italic>r =</italic> 0.03<italic>, p</italic> = 0.58) (<xref ref-type="fig" rid="pcbi-0020088-g002">Figure 2</xref> and <xref ref-type="table" rid="pcbi-0020088-t002">Table 2</xref>A). The HTP-PI network had 4,474 proteins; evolutionary rate data were obtained for 469 singleton nonessentials (<xref ref-type="table" rid="pcbi-0020088-t002">Table 2</xref>A). As originally claimed [<xref rid="pcbi-0020088-b003" ref-type="bibr">3</xref>,<xref rid="pcbi-0020088-b004" ref-type="bibr">4</xref>], a significant negative correlation between the connectivity and evolutionary rate was observed even after accounting for expression level (<xref ref-type="table" rid="pcbi-0020088-t002">Table 2</xref>A and <xref ref-type="fig" rid="pcbi-0020088-g002">Figure 2</xref>). Because large fractions of protein complexes are essential, restriction of our analysis to nonessentials weakens but does not eliminate the previous correlation between connectivity and evolutionary rate in the HTP-PI network. Similarly, the other two HTP-based networks (<xref ref-type="table" rid="pcbi-0020088-t002">Table 2</xref>A), including the recent data by Gavin et al. [<xref rid="pcbi-0020088-b029" ref-type="bibr">29</xref>] (for both full and a reduced version which controlled for false- positives due to sticky proteins by removing preys that occurred in more than 100 purifications), and another high confidence network, called filtered yeast interactome (FYI), generated by intersecting interactions from various sources [<xref rid="pcbi-0020088-b030" ref-type="bibr">30</xref>], showed negative correlation before accounting for expression level. However, in the Gavin (2006) dataset, control for expression level, measured either as protein abundance or by the CAI, removed any significance. The same lack of significance in Gavin (2006) data, after control for abundance, and in the LC data, before and after covariate control, is also seen for singleton essential genes (<xref ref-type="table" rid="pcbi-0020088-t002">Table 2</xref>B). To account for potential biases in the LC data we repeated the analysis on the LC data without ribosomal proteins and on the subset of LC data that were either two-hybrid based&#x02013;interactions or affinity purification-based interactions. For all three cases, the null hypothesis of no correlation could not be rejected (<xref ref-type="supplementary-material" rid="pcbi-0020088-st002">Table S2</xref>). The lack of correlation is robust to use of Ka/Ks rather than just the protein evolutionary rate (<xref ref-type="supplementary-material" rid="pcbi-0020088-st003">Table S3</xref>).</p><fig id="pcbi-0020088-g002" position="float"><label>Figure 2</label><caption><title>Relationship between Connectivity and Evolutionary Rate</title><p>Scatter plots of natural log of connectivity and evolutionary rate of <named-content content-type="genus-species">S. cerevisiae</named-content> proteins. Shown here are the plots for singleton genes. For analysis of all genes and those with duplicates see <xref ref-type="supplementary-material" rid="pcbi-0020088-sg001">Figures S1</xref> and <xref ref-type="supplementary-material" rid="pcbi-0020088-sg002">S2</xref>.</p></caption><graphic xlink:href="pcbi.0020088.g002"/></fig><table-wrap id="pcbi-0020088-t002" content-type="1col" position="float"><label>Table 2</label><caption><p>Correlation between Connectivity and Evolutionary Rates</p></caption><graphic xlink:href="pcbi.0020088.t002"/></table-wrap><p>The HTP-PI network may exhibit a negative correlation between connectivity and evolutionary rate because HTP methods preferentially detect interactions of abundant proteins [<xref rid="pcbi-0020088-b015" ref-type="bibr">15</xref>,<xref rid="pcbi-0020088-b019" ref-type="bibr">19</xref>] that evolve slower than lowly expressed genes [<xref rid="pcbi-0020088-b017" ref-type="bibr">17</xref>]. Network data, such as the FYI network (generated by taking intersections of various datasets), while a valid means to enrich for real interactions, further bias HTP data toward highly abundant proteins because interactions between such proteins are more likely to be replicated in HTP datasets. In contrast, interactions that are well established in the primary literature are often not re-published per se, and are therefore not validated in a formal sense, despite being highly controlled and reliable. In addition, HTP approaches tend to recover interactions among proteins that are part of multi-protein complexes, which also evolve more slowly [<xref rid="pcbi-0020088-b031" ref-type="bibr">31</xref>]. Indeed, 74% of HTP-PI proteins, for which evolutionary rate data are available, are part of a dedicated protein complex, compared to 64% of LC-PI proteins. The observed negative correlation in HTP datasets also derives from the fact that the average connectivity of subunits of protein complexes is almost two times higher than that of other proteins, in both the LC-PI and HTP-PI networks (<italic>p</italic> &#x0003c; 10^&#x02212;100, Spearman).</p><p>In conclusion, while we recover the prior strong negative correlation between connectivity and evolutionary rate in various HTP datasets, we find no such correlation in the LC dataset. We conclude that the prior claim that hub proteins are intrinsically slower evolving derives from the nature of the datasets used. This conclusion, however, raises a further problem, namely why do not all highly connected proteins have lower rates of evolution? It is known that sites of non-temporary interaction between proteins are slow evolving [<xref rid="pcbi-0020088-b002" ref-type="bibr">2</xref>]. However, do more highly connected proteins necessarily have more such sites or more importantly, a higher density of such sites? To address this issue we obtained the identity of binding residues of yeast proteins for which structures have been solved from the PRISM database [<xref rid="pcbi-0020088-b032" ref-type="bibr">32</xref>,<xref rid="pcbi-0020088-b033" ref-type="bibr">33</xref>] (see <xref ref-type="sec" rid="s4">Materials and Methods</xref>). Against expectations, the correlation between connectivity and fraction residues associated with binding is negative, albeit only weakly so (for LC data <italic>r =</italic> &#x02212;0.46, <italic>p =</italic> 0.02 and for HTP data <italic>r =</italic> &#x02212;0.3, <italic>p =</italic> 0.02; both Spearman rank correlation). This suggests that the underlying assumption that highly connected proteins have a higher density of binding sites does not hold and could explain why hub proteins evolve no faster than average. In part, the negative correlation arises from the fact that proteins with higher connectivity also are longer, although this effect is weak (Spearman rank correlation between protein length and connectivity: for LC-PI data <italic>r</italic> = 0.07, <italic>p</italic> &#x0003c; 0.0001, <italic>n</italic> = 3,256; for HTP-PI data <italic>r</italic> = 0.066, <italic>p</italic> &#x0003c; 0.0001, <italic>n</italic> = 4,142).</p></sec><sec id="s2e"><title>Activity and Lifetime of Protein Hubs Are Under Tight Regulation</title><p>The LC-PI network exhibits a highly interconnected topology that links a large fraction of the proteome. To minimize potentially hazardous cross talk, temporal control on the abundance and activity of hubs may be needed to ensure continuous equilibration with binding partners and curb excessive flux through certain pathways. Consistent with this expectation, we find that mRNAs of highly connected proteins have shorter half-lives (<italic>r</italic> = &#x02212;0.2, <italic>p</italic> = 1 &#x000d7; 10^&#x02212;22, Spearman). This negative correlation is true even after partialing-out abundance (<italic>r</italic> = &#x02212;0.17, <italic>p</italic> = 3 &#x000d7; 10^&#x02212;10, Spearman) (abundance is negatively correlated with half-life) and even when considering only nonessentials (<italic>r</italic> = &#x02212;0.14, <italic>p</italic> = 2 &#x000d7; 10^&#x02212;11, Spearman) (essentials have shorter half-life than nonessentials). As a proxy for protein regulation at a post-translational level, we analyzed the phosphorylation status of highly connected proteins. Phosphorylation often dictates regulation through protein interactions and/or protein degradation by the ubiquitin-proteasome system [<xref rid="pcbi-0020088-b034" ref-type="bibr">34</xref>,<xref rid="pcbi-0020088-b035" ref-type="bibr">35</xref>]. We find that more connected proteins are more likely to be phosphorylated (<italic>r</italic> = 0.06, <italic>p</italic> = 9 &#x000d7; 10^&#x02212;4, Spearman). Thus, one trade-off for broad specificity [<xref rid="pcbi-0020088-b036" ref-type="bibr">36</xref>] may be the energetic cost of &#x0201c;just-in-time&#x0201d; synthesis, which helps prevent entrainment of hubs by a select set of interaction partners.</p></sec></sec><sec id="s3"><title>Discussion</title><p>Analysis of LC-PI data suggests that prior analyses of the relationship between connectivity and rate of evolution were a product of the biases inherent in HTP data. In contrast, the relationship between dispensability and connectivity seems more robust. In particular, the claim that more highly connected proteins tend to be more likely to be essential is reported in all datasets, although this is only a trend when the data are derived exclusively from yeast two-hybrid assays. Similarly, the yeast two-hybrid data fail to support the idea that within the class of nonessential genes there is a correlation between growth rate and connectivity. Otherwise, we find reasonable support for this possibility, although in some instances the correlation is weak.</p><p>That highly connected proteins do not evolve slowly is intriguing given the simple intuition that the two attributes should covary. However, the logic of this intuition relies on the two interrelated assumptions: that sites of mutual protein&#x02013;protein binding should be slow evolving and that proteins with numerous interactions should have a higher density of binding sites as they have more partners. A third implicit assumption is that the proportion of sequence defined by binding sites is large enough to impact on rates of evolution given high variation in rates of evolution outside of pairing sites. The first assumption is partially upheld: sites of non-temporary binding are indeed slow evolving [<xref rid="pcbi-0020088-b002" ref-type="bibr">2</xref>]. Given this, why are not more highly connected proteins slower evolving? A key alternative possibility is that highly connected proteins tend to re-use the same binding site when interacting with multiple different proteins. Our analysis of well-described binding domains suggests this to be so and moreover, that genes with multiple interactants tend to be longer proteins; hence, for a given number of binding sites, the density of the sites will be lower in the longer proteins. However, the data for estimating the density of sites are limited and may also be biased, most especially toward proteins in obligate complexes. Nonetheless, the unexpected lack of a positive correlation between proportion of sequence involved in binding and connectivity provides evidence to support the conclusion that connectivity is not related to rate of protein evolution.</p><p>The finding that hub proteins tend to be encoded by mRNAs with rapid turnover rates provides an alternative explanation for the lack of correlation between connectivity and rate of evolution. From our data it would appear that many hub proteins are adapted to rapid turnover and/or regulation as they have both short half-lives and more phosphorylation sites. The short half-life in particular suggests that many hubs are not part of long-term stable interactions and instead form dynamic complexes that are readily removed from the system (or inactivated by phosphorylation) once a given task at a given time is performed. Many dynamic interactions likely occur via weak interactions at less conserved and hence less constrained binding surfaces, such as in the instance of kinase-substrate interactions. However, we note that dynamic interactions can also be mediated via conserved binding pockets that bury large surface areas of the interacting partners. In the latter instance, subunits marked for rapid turnover by phosphorylation, which often directs ubiquitin conjugation and subsequent degradation, can be readily stripped from the rest of the complex by the 26S proteasome and rapidly degraded [<xref rid="pcbi-0020088-b037" ref-type="bibr">37</xref>,<xref rid="pcbi-0020088-b038" ref-type="bibr">38</xref>]. In this manner, the same dedicated binding site may be used to link to many different interaction partners. A definitive test of the &#x0201c;just-in-time&#x0201d; attribute of hub interactions will require measurement of protein half-lives on a proteome-wide scale and systematic determination of the structural basis for transient protein interactions.</p></sec><sec id="s4"><title>Materials and Methods</title><sec id="s4a"><title>Interaction networks.</title><p>The HTP dataset was created by union of four HTP studies [<xref rid="pcbi-0020088-b021" ref-type="bibr">21</xref>&#x02013;<xref rid="pcbi-0020088-b024" ref-type="bibr">24</xref>]. The LC-PI dataset [<xref rid="pcbi-0020088-b013" ref-type="bibr">13</xref>] was obtained from the BioGRID database (<ext-link ext-link-type="uri" xlink:href="http://www.thebiogrid.org">http://www.thebiogrid.org</ext-link>), and the FYI dataset was obtained from Han et al. [<xref rid="pcbi-0020088-b030" ref-type="bibr">30</xref>]. A dataset corresponding to ~2,000 purified stable yeast protein complexes was created by connecting each bait to its prey in a spoke model [<xref rid="pcbi-0020088-b029" ref-type="bibr">29</xref>]. This latter analysis was also repeated with a reduced dataset that controlled for promiscuous interactions by removing preys that occurred in more than 100 purifications. Descriptive statistics of these networks are shown in <xref ref-type="table" rid="pcbi-0020088-t003">Table 3</xref>.</p><table-wrap id="pcbi-0020088-t003" content-type="1col" position="float"><label>Table 3</label><caption><p>Descriptive Statistics of Data Used in the Analysis</p></caption><graphic xlink:href="pcbi.0020088.t003"/></table-wrap></sec><sec id="s4b"><title>Miscellaneous data.</title><p>Evolutionary rate data (non-synonymous substitution per site and synonymous substitution per site) for the four-way <italic>stricto sensu</italic> species alignments for 3,036 genes [<xref rid="pcbi-0020088-b039" ref-type="bibr">39</xref>], proteome-wide protein abundance [<xref rid="pcbi-0020088-b040" ref-type="bibr">40</xref>], and genome-wide yeast mRNA half-life [<xref rid="pcbi-0020088-b041" ref-type="bibr">41</xref>] datasets were as described. CAI values were calculated as described [<xref rid="pcbi-0020088-b027" ref-type="bibr">27</xref>] using 30 most highly expressed yeast genes [<xref rid="pcbi-0020088-b042" ref-type="bibr">42</xref>]. Predicted phosphorylation data were obtained from Scansite database at: <ext-link ext-link-type="uri" xlink:href="http://scansite.mit.edu">http://scansite.mit.edu</ext-link> [<xref rid="pcbi-0020088-b043" ref-type="bibr">43</xref>] using a high stringency cut-off. Scansite identifies short protein sequence motifs that are recognized by modular signaling domains that are phosphorylated by protein Ser/Thr- or Tyr-kinases, or that mediate specific interactions with protein or phospholipid ligands. GO component complex membership information was obtained from the SGD database (<ext-link ext-link-type="uri" xlink:href="http://www.yeastgenome.org">http://www.yeastgenome.org</ext-link>) [<xref rid="pcbi-0020088-b044" ref-type="bibr">44</xref>]. To identify singleton genes, a yeast versus yeast BLASTP was done with E-value threshold of 0.1 and percent identity cut-off of 20; those proteins which did not return a hit using these parameters were considered to be singletons. This procedure resulted in 1,558 singletons. Proteins that were part of the &#x0201c;core&#x0201d; minus the &#x0201c;attachments&#x0201d; [<xref rid="pcbi-0020088-b029" ref-type="bibr">29</xref>] were considered to be stable complex subunits. Homozygous single-gene deletion fitness data in rich media (YPD) were as described [<xref rid="pcbi-0020088-b045" ref-type="bibr">45</xref>,<xref rid="pcbi-0020088-b046" ref-type="bibr">46</xref>]. Average of the two-replicate growth rate fitness measure of the homozygous deletion in rich media (YPD) was taken from the latter dataset.</p></sec><sec id="s4c"><title>Protein interaction residues.</title><p>Identity of binding residues data was obtained from the PRISM database [<xref rid="pcbi-0020088-b032" ref-type="bibr">32</xref>]. Identity of residues taking part in protein&#x02013;protein interactions were predicted using a computational approach that used structure information (mostly on multi-protein complexes) from the Protein Data Bank (PDB) [<xref rid="pcbi-0020088-b033" ref-type="bibr">33</xref>]. Briefly, interfaces were defined as the set of residues that made non-covalent contacts with residues on other chains or those that were in the vicinity of these contacting residues. Two residues from the opposite chains were marked as interacting if there was at least a pair of atoms, one from each residue, at a distance smaller than the sum of their van der Waals radii. Of the 166 <named-content content-type="genus-species">S. cerevisiae</named-content> proteins, for which interaction residue information was available, only those proteins that had a unique yeast ORF name were retained. Number of binding residues in the pair-wise domain interfaces was normalized by protein length to obtain the fraction of the sequence involved in binding. For all data used in the analysis see <xref ref-type="supplementary-material" rid="pcbi-0020088-sd001">Dataset S1</xref>.</p></sec><sec id="s4d"><title>Statistical methods.</title><p>The partial Spearman correlation between two variables, controlling for a third variable, was computed using the standard formula [<xref rid="pcbi-0020088-b026" ref-type="bibr">26</xref>]</p></sec></sec><sec sec-type="supplementary-material" id="s5"><title>Supporting information</title><supplementary-material content-type="local-data" id="pcbi-0020088-sd001"><label>Dataset S1</label><caption><title>Data Used in This Analysis</title><p>Columns: Gene, ESS? (Essential genes, 1 nonessential 0), y2h_or_aff (method used to find interaction in LC data, 1 = y2h only, 2 = affinity only, 3 = both, 0 = neither), k_lc (number of interactants in LC dataset, k_htp (number of interactants in HTP dataset), Abd (protein abundance), CAI, Rib (ribosomal, 1 = yes, 0 = no), fr_bind (fraction of residues involved in binding), mRNA_1/2 (mRNA half-life), phosphorylation (number of phosphorylation sites).</p><p>(980 KB XLS)</p></caption><media xlink:href="pcbi.0020088.sd001.xls"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020088-sd002"><label>Dataset S2</label><caption><title>Top 10% of Genes by Connectivity in the LC Dataset</title><p>(55 KB TXT)</p></caption><media xlink:href="pcbi.0020088.sd002.txt"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020088-sd003"><label>Dataset S3</label><caption><title>Top 10% of Genes by Connectivity in the HTP Dataset</title><p>(50 KB TXT)</p></caption><media xlink:href="pcbi.0020088.sd003.txt"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020088-sg001"><label>Figure S1</label><caption><title>The Relationship between Natural Log of Connectivity and Rate of Protein Evolution for All Genes</title><p>NB the slopes on the regression lines are not significantly different from zero for both LC datasets, but are highly significant (<italic>p</italic> &#x0003c; 0.0001) for the HTP data. For the essential genes in the LC set the non-parametric correlation is weakly significant (<italic>r</italic> = &#x02212;0.12; <italic>p</italic> &#x0003c; 0.01) but sensitive to control for protein abundance.</p><p>(239 KB PDF)</p></caption><media xlink:href="pcbi.0020088.sg001.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020088-sg002"><label>Figure S2</label><caption><title>The Relationship between Natural Log of Connectivity and Rate of Protein Evolution for Genes with Duplicates</title><p>NB the slopes on the regression lines are not significantly different from zero for both LC datasets, but are highly significant for the HTP data. For the essential genes in the LC set the non-parameteric correlation is weakly significant (<italic>r</italic> = &#x02212;0.12; <italic>p</italic> &#x0003c; 0.05) but sensitive to control for protein abundance.</p><p>(141 KB PDF)</p></caption><media xlink:href="pcbi.0020088.sg002.pdf"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020088-st001"><label>Table S1</label><caption><title>Degree versus Fitness for Subsets of LC data (1A) and Degree versus Fitness for Proteins Found in Both Y2h and Affinity Purification Methods (1B)</title><p>(28 KB DOC)</p></caption><media xlink:href="pcbi.0020088.st001.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020088-st002"><label>Table S2</label><caption><title>Degree versus Evolutionary Rate for Subsets of LC Data (Using dN)</title><p>(25 KB DOC)</p></caption><media xlink:href="pcbi.0020088.st002.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material><supplementary-material content-type="local-data" id="pcbi-0020088-st003"><label>Table S3</label><caption><title>Degree versus Evolutionary Rate for Subsets of LC Data (Using dN/dS)</title><p>(25 KB DOC)</p></caption><media xlink:href="pcbi.0020088.st003.doc"><caption><p>Click here for additional data file.</p></caption></media></supplementary-material></sec>

Wiggle&#x02014;Predicting Functionally Flexible Regions from Primary Sequence

<p>The Wiggle series are support vector machine&#x02013;based predictors that identify regions of functional flexibility using only protein sequence information. Functionally flexible regions are defined as regions that can adopt different conformational states and are assumed to be necessary for bioactivity. Many advances have been made in understanding the relationship between protein sequence and structure. This work contributes to those efforts by making strides to understand the relationship between protein sequence and flexibility. A coarse-grained protein dynamic modeling approach was used to generate the dataset required for support vector machine training. We define our regions of interest based on the participation of residues in correlated large-scale fluctuations. Even with this structure-based approach to computationally define regions of functional flexibility, predictors successfully extract sequence-flexibility relationships that have been experimentally confirmed to be functionally important. Thus, a sequence-based tool to identify flexible regions important for protein function has been created. The ability to identify functional flexibility using a sequence based approach complements structure-based definitions and will be especially useful for the large majority of proteins with unknown structures. The methodology offers promise to identify structural genomics targets amenable to crystallization and the possibility to engineer more flexible or rigid regions within proteins to modify their bioactivity.</p>

<contrib contrib-type="author"><name><surname>Gu</surname><given-names>Jenny</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff2">2</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><contrib contrib-type="author"><name><surname>Gribskov</surname><given-names>Michael</given-names></name><xref ref-type="aff" rid="aff3">3</xref></contrib><contrib contrib-type="author"><name><surname>Bourne</surname><given-names>Philip E</given-names></name><xref ref-type="aff" rid="aff1">1</xref><xref ref-type="aff" rid="aff2">2</xref></contrib>

PLoS Computational Biology

<sec id="s1"><title>Introduction</title><p>Protein structures are not rigid bodies, as suggested by time-independent solid-state crystal structures. Rather, proteins are selected by nature to balance between stability and flexibility in order to traverse the funnels of the protein energy landscape that characterize the conformational states needed to achieve a specific bioactivity. In part because of the way protein structures are traditionally represented and visualized in the crystallographic structure, the dynamics of protein motion is poorly conveyed and often neglected as the protein is treated as a static entity, although intuitively we know otherwise. Furthermore, protein sequence-structure relationships have been heavily focused on creating the most stable structure that may not necessarily be optimal for the execution or regulation of protein function. If the sequence is deterministic of the adopted protein fold, then the flexibility and dynamics of proteins should also be encoded by the sequence. Support for this notion comes from the previous demonstration that large amplitude fluctuations are mostly related to the overall protein shape [<xref rid="pcbi-0020090-b001" ref-type="bibr">1</xref>,<xref rid="pcbi-0020090-b002" ref-type="bibr">2</xref>], which in turn is defined by the sequence. In this work we develop a computational methodology that takes a small, but significant, step in understanding sequence-flexibility relationships important for protein function.</p><p>The flexibility of proteins is a necessary property to allow for conformational changes observed in allosteric interactions. The classic definition of allostery is the regulation of enzymes through the binding of effector molecules. This definition is now expanded to define allostery as the consequence of the redistribution of conformational states in the protein in response to a given external stimulus [<xref rid="pcbi-0020090-b003" ref-type="bibr">3</xref>]. We are particularly interested in the contribution of entropy as an allosteric mechanism used by proteins to allow for these conformational shifts to occur [<xref rid="pcbi-0020090-b004" ref-type="bibr">4</xref>,<xref rid="pcbi-0020090-b005" ref-type="bibr">5</xref>] and how this feature may be encoded at the protein sequence level.</p><p>The Cooper-Dryden model of allostery is a theory that addresses the contribution of entropy to the allosteric free energy. In extreme cases, this theory suggests that allostery can be achieved in the absence of structural change by simply shifting the internal vibrational modes when reacting to an external stimulus such as ligand binding [<xref rid="pcbi-0020090-b006" ref-type="bibr">6</xref>]. Associated with this model is the idea of remote entropy compensation, a scenario where a local entropy decrease in one area of a protein is compensated by an increase in entropy in another area. These regions can be located distantly from each other, thereby making the entropy compensation a long-range effect.</p><p>Entropy compensation has been observed using both computational and experimental approaches in many different proteins. Here we consider five examples to make the point. First, molecular dynamic (MD) simulations of lysozyme show differences between the dynamics of substrate bound and free states. When lysozyme is in complex with the substrate, a distant loop (residues 67 to 88) increases in fluctuation to compensate for the decreasing fluctuation observed for the substrate-contacting loop (residues 101 to 107) [<xref rid="pcbi-0020090-b007" ref-type="bibr">7</xref>]. Second, global structural changes resulting from changes in local fluctuation induced by proton binding are observed in staphylococcal nucleases [<xref rid="pcbi-0020090-b008" ref-type="bibr">8</xref>]. Third, spectroscopic experiments on the Tet repressor examined the fluorescence anisotropy decay of tryptophans introduced into a functionally important loop located distantly from the site of substrate binding. An increase in fluctuation was observed in this loop when anhydrotetracycline was bound to the Tet repressor [<xref rid="pcbi-0020090-b009" ref-type="bibr">9</xref>]. Fourth, entropy compensation can be inferred from comparing X-ray structures of adenylate kinase in different conformations [<xref rid="pcbi-0020090-b010" ref-type="bibr">10</xref>]. Fluctuations localized at the nucleoside monophosphate binding and LID domains, the substrate binding interface, show an inverse relationship with the fluctuations of loops &#x003b1;4-&#x003b2;3 and &#x003b1;5-&#x003b2;4 that are located distantly. Finally, mutational studies show long-range dynamic perturbations in eglin C detected by NMR. This protein is considered to be classically nonallosteric, an example where distant fluctuations can be affected by changes in sequence [<xref rid="pcbi-0020090-b011" ref-type="bibr">11</xref>], a point we come back to subsequently.</p><p>In each of these cases, local regions of protein structure serve to accommodate the redistribution of vibrational modes and provide an energy reserve of allosteric free energy as proposed by the Cooper-Dryden model. The relaxation and tensing of regions of local structure is a transition from an ordered to disordered state and vice versa. Such local regions include hinges, recognition loops, and certain catalytic loops whose vibrational states change in the presence of an external stimulus such as substrate binding. While structurally dissimilar, hinges, recognition loops, and catalytic loops all exhibit characteristic fluctuations that differ from the mean fluctuation. Hinges are relatively immobile at the hinge point compared to surrounding fluctuations about the hinge, whereas recognition loops and, in certain examples, catalytic loops show minimal fluctuations at the extremities and maximal fluctuations at the center of the loop. We attempt to identify these regions based on the scale and cooperativeness of fluctuations that often define protein function and refer to them as <italic>functionally flexible regions</italic> (FFRs).</p><p>In this paper, we begin with a structure-based definition of an FFR to obtain our training dataset and describe prediction tools created as a result to identify these regions using only protein sequence information. With the growth of protein structures fueled by structural genomics [<xref rid="pcbi-0020090-b012" ref-type="bibr">12</xref>], it is possible to generate a training dataset to begin efforts to understand relationships between protein sequence and functional flexibility (FF). First, we devised a method using protein dynamics modeling to identify FFRs using only a single protein conformation. Then we use machine learning techniques to identify protein regions with the measured amount of flexibility needed for bioactivity without using structural information. Overlaps are expected with existing disorder and order predictions since FFRs can exist in both states. Disordered predictors are trained to predict regions of high flexibility based on temperature factor information or regions of the protein with no electron density. Sequence analyses of predicted disordered regions have revealed the existence of different types of disorder [<xref rid="pcbi-0020090-b013" ref-type="bibr">13</xref>] which may include FFRs. FFRs can also adopt ordered structures when triggered to do so under the right conditions.</p><p>The long-term goal of work such as this is to provide a generalized relationship between sequence and FF for all proteins. An immediate benefit would be in facilitating the structure solution process such that proteins less tractable for crystallization could be identified. Further, by our definition, FFRs border on forming an ordered structure; therefore, if such regions can be identified, it may be possible to introduce a few mutations to stabilize local regions that are not located on the ends of the polypeptide chain. This strategy has been utilized to successfully create a soluble analog of erythropoietin [<xref rid="pcbi-0020090-b014" ref-type="bibr">14</xref>]. We also hope to contribute to the field of de novo protein design with the understanding of the relationship between protein sequence and dynamics. Recently, a three-dimensional structure unseen in nature with a root-mean-square deviation of 1.2 &#x000c5; from the design model was engineered [<xref rid="pcbi-0020090-b015" ref-type="bibr">15</xref>]. Conceivably, understanding sequence-flexibility relationships would be useful in guiding the engineering necessary to introduce the flexibility required for bioactivity in these newly designed proteins. Furthermore, as in the example of eglin C, there are flexibility modulating regions that cannot be obviously identified with structural inspection but possibly detected with improved understanding of sequence-flexibility relationships.</p></sec><sec id="s2"><title>Results/Discussion</title><sec id="s2a"><title>Case Studies of FFRs: Identification with an FF Score</title><p>We define FFRs to have the property of coordinated participation in large amplitude fluctuations that are different from the mean vibrational fluctuation of the protein. The Gaussian network model (GNM) [<xref rid="pcbi-0020090-b016" ref-type="bibr">16</xref>], a coarse-grained protein dynamics modeling approach, was chosen to obtain fluctuation mode information needed to identify regions of interest because it is a computationally practical alternative to an all-atom MD simulation yet provides a good approximation of near-native protein fluctuation at longer time scales [<xref rid="pcbi-0020090-b017" ref-type="bibr">17</xref>]. Classic all-atom MD provides accurate, detailed descriptions of molecular motion. However, simulations are limited by computational demands to a few tens of nanoseconds. GNM is able to address large-scale fluctuations that extend beyond the time scale of MD simulation, a capability important for some types of molecular recognition and allosteric rearrangements occurring at time scales of microseconds and longer. While GNM provides only an approximation, several studies comparing coarse-grained approaches to MD have shown that it is an accurate and efficient alternative [<xref rid="pcbi-0020090-b016" ref-type="bibr">16</xref>&#x02013;<xref rid="pcbi-0020090-b020" ref-type="bibr">20</xref>].</p><p>There are two reasons for using protein dynamic modeling results instead of experimental temperature factors to define our target regions. First, by using protein dynamic modeling simulation, we are able to investigate protein flexibility with the added dimensionality of having functional importance. Second, by using modes of motion to define our target regions, we are able to focus specifically on large-amplitude fluctuations without including contributions from higher frequency fluctuations. These two features are the distinguishing qualities that set our predictors apart from other disorder predictors. The advantages of using this approach will be highlighted in the subsequent discussion and reflected in comparisons made to other disorder predictors.</p><p>To identify FFRs, we focus on the first two vibrational modes of protein fluctuation because these modes have been shown to sufficiently describe important contributions to global fluctuations necessary for protein function [<xref rid="pcbi-0020090-b021" ref-type="bibr">21</xref>&#x02013;<xref rid="pcbi-0020090-b024" ref-type="bibr">24</xref>]. Flexible regions with important functional roles can be discriminated by considering fluctuations associated with correlated motion [<xref rid="pcbi-0020090-b025" ref-type="bibr">25</xref>]. Information regarding coordinated motion for each residue can be obtained with the GNM from the cross-correlation matrix. While the definition we present here is conservative since important transitions known for protein function have been observed in other modes, it provides an initial training set that allows a support vector machine (SVM) to model the sequence subspace that encodes flexible regions with functional importance. Furthermore, with this approach we are able to identify FFRs using only a single protein conformation, making it possible to quickly generate the training dataset needed to build a prediction tool. Eliminating the need to extrapolate motion between two protein conformers allows us to expand the size of our training set.</p><p>Correlation values were used to weight mode information to create an FF score and empirically define a threshold to objectively identify FFRs (see <xref ref-type="sec" rid="s3">Materials and Methods</xref>). The FF scores are then normalized such that the mean value is 0 and the standard deviation is 1 in order to establish a standard threshold for all proteins in the training set. The threshold is established based on the hypothesis that fluctuations of functional importance will deviate from the mean fluctuation observed for the entire protein. Therefore, we consider residues with a normalized FF score greater than 1.5 standard deviations from the mean fluctuation to exhibit flexibility of functional importance.</p><p>The FF score is used for definition purposes only. With this definition procedure we are able to obtain an objectively defined dataset needed for SVM training. The dominant motions in the lowest amplitude modes correspond to rigid domain motions [<xref rid="pcbi-0020090-b026" ref-type="bibr">26</xref>,<xref rid="pcbi-0020090-b027" ref-type="bibr">27</xref>]. The normalization procedure above, scaled with the correlation value, identified the extreme fluctuations within the rigid domains. Positive FF scores indicate large fluctuations relative to the intrinsic fluctuation state of the entire protein, whereas negative values indicate smaller than average fluctuations. Regions such as recognition and activation loops will fall on the extreme positive end of this FF score spectrum. Although low values of extracted GNM modes correspond to stable regions with negligible fluctuation, extreme negative FF scores correspond to hinge regions&#x02014;the rigid domain fluctuations, modeled by the GNM, will be moving with respect to the hinge itself. Therefore, the hinge will appear to be immobile with the observed fluctuation falling below the overall mean fluctuation of the protein. Based on the examples provided subsequently, we show that this operational definition of FFRs sufficiently defines biologically confirmed flexible regions.</p><p>The FF score was first tested on HIV protease. While the recognition loop (residues 36 to 42) is identified without incorporating correlated movement information to weight normalized GNM fluctuations, the flap region (residues 46 to 56) important for dimerization was not identified because fluctuation is suppressed in the dimerized state (<xref ref-type="fig" rid="pcbi-0020090-g001">Figure 1</xref>). We subsequently show that incorporating information regarding correlated residue movements in a weighting scheme to rescale the GNM mode (see <xref ref-type="sec" rid="s3">Materials and Methods</xref>) improved the identification of FFRs. The biological function of a protein is often achieved through coordinated movements; thus, the FF score uses values extracted from the cross-correlation matrix to weight residue participation in correlated fluctuations. Furthermore, fluctuations that are biologically important for protein function are often defined by their correlated nature [<xref rid="pcbi-0020090-b025" ref-type="bibr">25</xref>]. As a result of this weighting scheme, residues with little participation in correlated movements are rescaled to have lower FF scores, whereas those with high correlation to other residues will have higher scores. Correlated and anticorrelated fluctuations are accounted for by summing the square of maximum and minimum correlation values, which are then used to scale the weighted average of the two slowest modes (see <xref ref-type="sec" rid="s3">Materials and Methods</xref>). Using this weighting scheme in the FF score, we are able to improve our definition of FFRs. For HIV protease, the weighted FF score enabled us to detect the flap region and correctly categorize it to be functionally important (<xref ref-type="fig" rid="pcbi-0020090-g001">Figure 1</xref>C and <xref ref-type="fig" rid="pcbi-0020090-g001">1</xref>D).</p><fig id="pcbi-0020090-g001" position="float"><label>Figure 1</label><caption><title>Defining FFRs in HIV Protease Using the Derived FF Score</title><p>(A) Comparison of temperature factor (dashed line) and weighted average of the two slowest modes (solid line) obtained with GNM. The HIV protease is modeled as a dimer; however, the plot shows results for a single chain.</p><p>(B) Gradient plot ranging from correlated (red) to anticorrelated (blue) movement for each residue in the dimer.</p><p>(C) Comparison of normalized scores for unweighted (dashed line) and correlation-weighted (solid line) modes for a single chain. Correlation-weighted modes define the FF score. Regions are identified as FFR when values exceed thresholds (red lines) greater than 1.5 and less than &#x02212;1.5. The flap region (residues 46 to 56) exceeds the threshold after including correlated movement information (solid line).</p><p>(D) Structural mapping of FF score with gradient from negative (blue) to positive (red), (PDB ID: 1HIV).</p></caption><graphic xlink:href="pcbi.0020090.g001"/></fig><p>Improvements in defining FFRs using the FF score were also observed for calmodulin and bovine pancreatic trypsin inhibitor (BPTI) (<xref ref-type="fig" rid="pcbi-0020090-g002">Figure 2</xref>). Calmodulin is a signaling protein consisting of an alpha helical hinge between two globular domains. While the two globular domains have been found to be structurally similar to each other, they differ dynamically [<xref rid="pcbi-0020090-b028" ref-type="bibr">28</xref>&#x02013;<xref rid="pcbi-0020090-b030" ref-type="bibr">30</xref>]. These differences are also observed in the GNM modeling result that shows the N-terminal domain to be more flexible than the C-terminal domain. However, it is the interconnecting helix, containing eight turns, that has been observed to undergo the largest structural change upon calcium and substrate binding [<xref rid="pcbi-0020090-b031" ref-type="bibr">31</xref>,<xref rid="pcbi-0020090-b032" ref-type="bibr">32</xref>]. When bound to a peptide, a kink is introduced in this alpha helical hinge leading to a collapse that forms two perpendicular alpha helices while the globular domains wrap around the peptide. FF scores less than &#x02212;1.5 identify this hinge region between the two globular domains despite having an ordered alpha helical structure.</p><fig id="pcbi-0020090-g002" position="float"><label>Figure 2</label><caption><title>FF Score Identifies FFR in Bovine Pancreatic Inhibitor and Calmodulin</title><p>Comparison of unweighted (dashed line) and weighted (solid line) FF scores for BPTI ([top], PDB ID: 5PTI) and calmodulin (bottom, PDB ID: 1CLL). FF scores are mapped with the same gradient coloring from negative (blue) to positive (red) as the scale shown in <xref ref-type="fig" rid="pcbi-0020090-g001">Figure 1</xref>D. Both recognition loops (loop 1: residues 11 to 19; loop 2: residues 35 to 42) are identified in BPTI by the FF score, whereas loop 2 is not identified with the unweighted mode. For calmodulin, the FF score allows us to identify the central hinge for this protein (residues 68 to 91 shown in blue because it exceeds the negative threshold of less than &#x02212;1.5). This central helix, containing eight turns, is known to collapse when bound to calcium and substrate.</p></caption><graphic xlink:href="pcbi.0020090.g002"/></fig><p>The binding affinity of BPTI is influenced by mutations in the active loops (residues 11 to 19 and 35 to 42) that are inserted into the active site of the proteolytic enzymes. Mutations Y35G [<xref rid="pcbi-0020090-b033" ref-type="bibr">33</xref>] and G37A [<xref rid="pcbi-0020090-b034" ref-type="bibr">34</xref>] lead to an observed increase in the fluctuation of these loops and reduce the binding affinity for trypsin compared to the native form. Structurally, the monomer G37A mutant adopts a near wild-type conformation based on comparison of nuclear Overhauser effects in NMR structures, whereas the structure of the Y35G mutant showed a 6-&#x000c5; root-mean-square deviation from the native structure. Nevertheless, both mutant proteins showed a native conformation when bound to trypsin. In this example, we stress that while both proteins continue to adopt wild-type conformation according to experimental studies, their dynamics and stability varied substantially. The regions that were impacted the most by these mutations have been identified by our definition. Residues in these loop regions are defined to be FFRs with FF scores exceeding the threshold of 1.5. While this threshold is arbitrary and defined empirically, it provides a consistent definition which we can use as targets for training our predictors to identify sequence patterns that correspond to these regions.</p></sec><sec id="s2b"><title>Features of FFRs</title><p>Based on the FF score, each residue in a nonredundant training set was classified as FFR or non-FFR. Residues were separated into a binary classification with FFRs assigned a value of 1 and non-FFRs were assigned a value of &#x02212;1. Examining the distribution of residues in the two classes shows that an FFR averages 9 &#x000b1; 11 residues in length and comprises about 20% of all residues. Residues identified as hinges comprise about 0.75% of all residues in the training set. The average maximum length of an FFR for each protein increases with increasing protein length (<xref ref-type="fig" rid="pcbi-0020090-g003">Figure 3</xref>A). This increase in length may be associated with longer flexible regions forming linker regions between multiple domains.</p><fig id="pcbi-0020090-g003" position="float"><label>Figure 3</label><caption><title>Preliminary Analysis of FFR as Identified by FF Score</title><p>(A) The average of all maximal FFR lengths plotted against overall protein length.</p><p>(B) The number of different sequence patterns observed for a given window size. Shown are the pattern counts for regions classified as FFR (dash line), non-FFR (thin line), and irrespective of classification (thick line). FFR regions sample a smaller sequence space compared to non-FFR regions. Patterns overlapping boundaries of FFR and non-FFR are excluded from these counts.</p></caption><graphic xlink:href="pcbi.0020090.g003"/></fig><p>We examined the classification preference for each amino acid and secondary structure type using the same assignment values (1 and &#x02212;1) (<xref ref-type="table" rid="pcbi-0020090-t001">Table 1</xref>). Residues in beta strands generally make up protein cores and are less likely to constitute FFRs than helices or loops, a trend observed in the data (<xref ref-type="table" rid="pcbi-0020090-t001">Table 1</xref>). Charged residues show stronger preferences to be in FFRs than non-FFRs (<xref ref-type="table" rid="pcbi-0020090-t002">Table 2</xref>). Glycine was not among the top ranking residues, ranking even lower than proline. This is expected since the conformational flexibility and nonrestrictive properties of glycine make this residue very adaptable. Moreover, glycine is found both at the surface and in the hydrophobic core with no strong preference for either. Proline is known to be a helix breaker due to the conformational restraints of the covalent bond between the side chain and backbone. This conformational limitation means that proline is more likely to be found in loops and hence have a higher FF score. As expected, hydrophobic residues tend to be found in regions of less flexibility since they are packed into the hydrophobic core. Cysteines are frequently involved in disulfide formation and were found among the least common residues in FFRs. The large standard deviations found for these classification preference indicate that neither secondary structure nor amino acid residue properties alone are sufficient to serve as the distinguishing factor for classification of FFRs.</p><table-wrap id="pcbi-0020090-t001" content-type="1col" position="float"><label>Table 1</label><caption><p>FFR Classification Preference for Secondary Structures</p></caption><graphic xlink:href="pcbi.0020090.t001"/></table-wrap><table-wrap id="pcbi-0020090-t002" content-type="2col" position="float"><label>Table 2</label><caption><p>FFR Classification Preference for Amino Acids</p></caption><graphic xlink:href="pcbi.0020090.t002"/></table-wrap></sec><sec id="s2c"><title>Accessing Sequence Pattern Preferences for FFRs</title><p>Window scanning for particular patterns reveals that FFRs occupy a smaller sequence space than their non-FFR counterparts (<xref ref-type="fig" rid="pcbi-0020090-g003">Figure 3</xref>B). For a window size of 2, all possible amino acid pairs are sampled by both FFRs and non-FFRs. The majority of triplets continue to be sampled for a window size of 3. Differences in pattern sampling become more evident for window sizes 4 and larger, indicating sequence preferences for FFRs and non-FFRs.</p><p>Certain tripeptide sequences can be overrepresented in FFRs when compared to non-FFRs. We attempt to identify these tripeptides by using a modified bootstrapping approach to calculate <italic>Z-</italic>scores and <italic>p</italic>-values for association with FFRs (see <xref ref-type="sec" rid="s3">Materials and Methods</xref>). For a window size of 3, a total of 8,000 tripeptide sequence patterns are possible. There were 7,982 patterns observed in the training set, with 7,261 patterns in FFR regions and 7,967 in non-FFR regions. The modified bootstrap sampling with 10,000 repetitions for the respective subset size showed 429 patterns in the FFR pool to be statistically associated with that category using a <italic>p</italic>-value threshold of 0.05. These patterns are either underrepresented or overrepresented in FFRs compared to the null FFR model, making it a distinctive set to help identify these regions. While the statistical associations are weak, using these values as additional input features improved the prediction performance of SVMs.</p><p>Results from this analysis suggest that there are sequence patterns associated with these regions that may be detected using machine learning techniques and these findings have been instrumental in improving the prediction quality of our SVM-based predictors when incorporated. The rationale behind the modified bootstrapping was to identify tripeptide sequence patterns associated with FFRs and to use this information to help SVMs distinguish between FFRs and non-FFRs. This finding of context dependence supports previous work that has shown that the Flory isolated-pair hypothesis does not hold true [<xref rid="pcbi-0020090-b035" ref-type="bibr">35</xref>]. This hypothesis states that the backbone conformation of residues is influenced by the nearest-neighbor residues rather than being independent of their conformations.</p></sec><sec id="s2d"><title>SVM Architecture and Training</title><p>While many successful structure predictors use multiple sequence alignments or position specific scoring matrices, we chose to use hidden Markov models (HMMs) because they additionally capture insertion and deletion probabilities that may occur within the sequence [<xref rid="pcbi-0020090-b036" ref-type="bibr">36</xref>,<xref rid="pcbi-0020090-b037" ref-type="bibr">37</xref>]. As such these probabilities capture information regarding the conservation of sequence length that can be particularly important for identifying active sites or recognition loops limited to certain lengths. A total of 29 transition and match states were used as input features to the SVM (see <xref ref-type="sec" rid="s3">Materials and Methods</xref>).</p><p>Exploring the performances of various SVM architectures have shown that a two-layered architecture yields the best performing predictor to identify residues in FFRs. The first-layer SVM makes an initial classification based on sequence and evolutionary information contained in the HMM states. The second-layer SVM serves to smooth the prediction from the first-layer SVM and uses results obtained from the modified bootstrap analysis to make better predictions. Incorporating information regarding tripeptide classification preferences was instrumental to improving the performance of our final predictor despite having a weak statistical value. Compared to a predictor that does not include tripeptide classification preferences, the performance of the SVM showed an additional 5% increase in accuracy and precision with an additional 3% improvement in recall.</p><p>The predictive performance of the SVMs was found to be a function of protein length. High false-positive rates were observed for shorter proteins (<xref ref-type="fig" rid="pcbi-0020090-g004">Figure 4</xref>A). This high error rate may be a result of original misclassification by the FF scores. For shorter protein segments, flexible regions are more likely to be assigned as non-FFRs because the dynamics of the segments will be modeled in a complex as opposed to a free monomer. Stated another way, it may be difficult to say whether these segments are intrinsically flexible in the apo form since they are always found with their binding partners. In total, complexed proteins compose 43.4% of the training set; 49.8% of proteins smaller than 200 residues are in complexes as compared to 35% found for larger proteins. Moreover, smaller proteins in crystal structures may be truncations or mimics of a flexible loop from a larger protein, leading to the misclassification of an FFR as a rigid segment even though the region may be flexible biologically.</p><fig id="pcbi-0020090-g004" position="float"><label>Figure 4</label><caption><title>Predictor Performance Is a Function of Protein Length</title><p>(A) Sequence effect on false-positive (thick line) and false-negative (thin line) error rate. Shorter sequences tend to have higher false positive identification of FFRs when trained on a nonpartitioned dataset.</p><p>(B) Comparison of SVM prediction results trained on a nonpartitioned dataset (dashed lines) and a partitioned dataset containing proteins up to 200 residues (solid lines). Improvements were seen in both the false-positive (black) and -negative (red) rates.</p><p>(C) Comparison of SVM prediction results trained on a nonpartitioned dataset (dashed lines) and a partitioned dataset containing proteins larger than 200 residues (solid lines). Minor improvements were observed in false-positive (black) and -negative (red) rates.</p></caption><graphic xlink:href="pcbi.0020090.g004"/></fig><p>To account for protein length, the original training set was partitioned into two sets: A, 760 proteins up to 200 residues in length; and B, 574 proteins longer than 200 residues. SVMs trained on the partitioned training sets both showed an improvement in performance (<xref ref-type="fig" rid="pcbi-0020090-g004">Figure 4</xref>B). Training on subset A showed an overall improvement of 12% in recall and 7.8% in precision for a total accuracy of 76.46%, precision of 48.99%, and recall of 78.27%, whereas training on subset B showed only a slight improvement over training on all proteins (<xref ref-type="fig" rid="pcbi-0020090-g004">Figure 4</xref>C) with an accuracy of 66.01%, precision of 37.11%, and recall of 70.49%.</p><p>Our final predictors, Wiggle and Wiggle200, use the radial basis kernel function in the first layer and a linear kernel in the second layer. Wiggle is the product of training on all proteins and Wiggle200 was trained on subset A containing proteins up to 200 residues. Since minor improvements were observed for the predictor trained on the subset containing larger proteins, we use Wiggle to conduct our predictions. In the following discussion, we will first revisit the dependency of the predictors on protein size in regard to domain boundary detection. Then we will discuss the performance of the predictors on three examples with experimentally verified FFRs.</p></sec><sec id="s2e"><title>Domain Boundary Identification</title><p>Flexible linkers between domains, sometimes acting as a hinge, are examples of FFRs and we evaluate the performance of Wiggle and Wiggle200 in the detection of these regions. We use a comprehensive domain boundary benchmark set (BENCH) that was curated to reflect the consensus of experts (CATH, SCOP, and authors of the protein structures) (T. Holland, S. Veretnik, I. N. Shindyalov, and P. E. Bourne, unpublished data). Because the boundary is defined between two residue positions, we expand the definition up to a window size of 15 residues, with the boundary in the center, to evaluate the performance of the predictors. We also partitioned BENCH based on protein size into BENCHA (200 residues or fewer) and BENCHB (more than 200 residues).</p><p>The general trend in predictor performance for Wiggle and Wiggle200 observed for all datasets (BENCH, BENCHA, BENCHB) is that precision increases with the size of domain boundary expansion, whereas recall increases up to window size 5 and begins to decline afterward (<xref ref-type="fig" rid="pcbi-0020090-g005">Figure 5</xref>). The overall accuracy is observed to decrease with a difference of about 2% for all datasets. For this reason, we will focus our performance comparison between the two predictors on a window size 5.</p><fig id="pcbi-0020090-g005" position="float"><label>Figure 5</label><caption><title>Predictor Performance in Identifying Domain Boundaries</title><p>Wiggle predictors were evaluated for domain boundary predictions on (A) a benchmark dataset containing domain boundary consensus between experts (BENCH), (B) a partitioned BENCH with proteins up to and including 200 residues (BENCHA), and (C) a partitioned BENCH with proteins longer than 200 residues. Definitions of domain boundaries were expanded up to a window size of 15 (win15) with the boundary in the center.</p></caption><graphic xlink:href="pcbi.0020090.g005"/></fig><p>For BENCH, we find that Wiggle outperforms Wiggle200 in recall by an additional +12.99% with little improvement in precision (+0.31%) and a decrease in accuracy (&#x02212;6.44%). Wiggle identifies domain boundaries in BENCH at an accuracy of 62.55% with a precision of 6% and recall of 54.15%. We are not surprised to see a poor precision value since both predictors will identify other flexible regions that are not linkers between domains. However, the results here show that our predictors are identifying linkers between domain boundaries, for example, possibly serving a functional purpose as a hinge.</p><p>For the partitioned benchmark set (BENCHA and BENCHB), we find that Wiggle again outperforms Wiggle200 in domain boundary recall with an additional +14.34% and +12.51%, respectively. Again, minor improvements were observed in precision (BENCHA: +0.13%, BENCHB: +0.31%) and a slight decrease in accuracy (BENCHA: &#x02212;7.68%, BENCHB: &#x02212;6.19%) was observed for Wiggle compared to Wiggle200. For the partitioned datasets, BENCHA and BENCHB, Wiggle predicts boundaries at (BENCHA: 59.08%; BENCHB: 63.24%) accuracy, (BENCHA: 9.39%; BENCHB: 5.21%) precision, and (BENCHA: 60.66%; BENCHB: 51.85%) recall, respectively. This clearly indicates that Wiggle, trained on the entire training dataset which includes larger multidomain proteins, has picked up sequence patterns associated with linker regions and is the better predictor for domain boundaries compared to Wiggle200.</p></sec><sec id="s2f"><title>SVM Performance on Experimentally Verified FFRs</title><p>Although the GNM provides a fast approach to identifying FFRs, there are limitations to the model. Dynamic modeling results are largely dependent on protein conformation, particularly that defined by bound and unbound conformations as discussed earlier for the observed higher false-positive error rate for smaller proteins. Therefore, the FF score does not always correctly define the regions of interest. We examined a few case studies where residues were largely misclassified by the FF score and compared the results to our SVM predictions. While it is ideal to have a precisely classified training dataset, we concluded that the classification made by the FF score provides a sufficient training set for the SVM to detect correct signals in sequence patterns for FFRs. In short, SVMs are powerful enough to generalize the relationship between protein sequence and FFRs as illustrated in the following examples.</p><sec id="s2f1"><title>Arc repressor.</title><p>The arc repressor is stable as a dimer, unfolded as a monomer [<xref rid="pcbi-0020090-b038" ref-type="bibr">38</xref>&#x02013;<xref rid="pcbi-0020090-b041" ref-type="bibr">41</xref>], and bound to DNA as a tetramer [<xref rid="pcbi-0020090-b041" ref-type="bibr">41</xref>,<xref rid="pcbi-0020090-b042" ref-type="bibr">42</xref>]. Extensive mutagenesis has been conducted to identify residue contributions to activity and stability [<xref rid="pcbi-0020090-b043" ref-type="bibr">43</xref>]. The beta strand near the N-terminus, the site of DNA interaction, is the least tolerant to substitution when selected for activity, but mutations have minimal effects when selected for stability. The loop between the two alpha helices (residues 28 to 34) was found to be intolerant to substitution under both circumstances. Based on these mutagenesis studies, these are some of the target regions we wish to identify using our sequence-based predictors.</p><p>Structurally, several flexible regions having important roles for protein function have been detected in the arc repressor using various experimental techniques. Despite being highly disordered in solution, according to an NMR structure determination [<xref rid="pcbi-0020090-b044" ref-type="bibr">44</xref>], the N-terminus of the repressor (residues 1 to 9) is important for specific operator binding [<xref rid="pcbi-0020090-b045" ref-type="bibr">45</xref>]. The last three residues of the C-terminus are also found to be disordered in solution [<xref rid="pcbi-0020090-b044" ref-type="bibr">44</xref>], while remaining residues at this terminus have been found to contain important contacts for tetramerization [<xref rid="pcbi-0020090-b046" ref-type="bibr">46</xref>]. Hydrogen exchange experiments show the exchange rates for the two alpha helices are concentration dependent and suggest that the protein exists as a molten globule in a monomeric state [<xref rid="pcbi-0020090-b038" ref-type="bibr">38</xref>]. In order to make all the DNA contacts observed in operator binding [<xref rid="pcbi-0020090-b047" ref-type="bibr">47</xref>], four molecules of arc repressors are needed, suggesting the existence of a tetrameric state. To shift from a monomeric to dimeric and finally tetrameric state requires considerable accommodation for conformational change. The flexibility of this protein required to accommodate these domain arrangements is not evident from the crystallographic or NMR structures alone.</p><p>Wiggle identifies residues 5 to 8, 23 to 35, 38, and 40 to 53 as FFRs, and Wiggle200 identifies residues 5, 23 to 29, and 43 to 53. The FF score only identifies residues 45 to 53 located at the C-terminus (<xref ref-type="fig" rid="pcbi-0020090-g006">Figure 6</xref>A). Residues 5 to 8, identified by Wiggle, correspond to the residues experimentally defined as important for DNA recognition at the N-terminus, while residues 23 to 35 and 38 correspond to the substitution-intolerant loop linking the two alpha helices [<xref rid="pcbi-0020090-b043" ref-type="bibr">43</xref>,<xref rid="pcbi-0020090-b046" ref-type="bibr">46</xref>] (<xref ref-type="fig" rid="pcbi-0020090-g006">Figure 6</xref>B).</p><fig id="pcbi-0020090-g006" position="float"><label>Figure 6</label><caption><title>Performance of Wiggle Predictors on Arc Repressor</title><p>(A) The dimer conformation of the Arc repressor was used to model global fluctuation. Using the FFR definition, the plot for a single chain is shown on the left with structural mapping of values onto a dimer on the right. FF scores are mapped with the gradient code from negative (blue) to positive (red). Only the C-terminal tail exceeds threshold lines (red) and is defined as an FFR while the rest of the protein is not. (PDB ID: 1BAZ)</p><p>(B) The hinge between the two helices is identified by predictors as well as N-terminal residues important for DNA recognition. Predictions from Wiggle (solid line) are mapped in green on the structure and Wiggle200 (dashed line) are mapped in orange.</p></caption><graphic xlink:href="pcbi.0020090.g006"/></fig></sec><sec id="s2f2"><title>PVUII endonuclease.</title><p>PVUII endonucleases (156 amino acids) are homodimerizing proteins that catalyze highly specific DNA cleavage. No regions of flexibility were identified with the FF score (<xref ref-type="fig" rid="pcbi-0020090-g007">Figure 7</xref>A). Wiggle identified residues 2 to 10, 26 to 31, 33 to 38, 65 to 68, 116 to 118, 121, 132 to 138, and 146 to 157 as FFRs, and Wiggle200 identified residues 2 to 8, 33, 34, 36, 53 to 58, 60, 61, 94 to 96, 117 to 120, and 150 to 157. Both predictors identified the loop involved in minor groove recognition (residues 26 to 36) [<xref rid="pcbi-0020090-b048" ref-type="bibr">48</xref>], Mg⁺⁺ ion coordination (residues 58, 67, 68, 82, and 94) [<xref rid="pcbi-0020090-b049" ref-type="bibr">49</xref>], and catalytic activity (residue 34) [<xref rid="pcbi-0020090-b048" ref-type="bibr">48</xref>,<xref rid="pcbi-0020090-b050" ref-type="bibr">50</xref>] (<xref ref-type="fig" rid="pcbi-0020090-g007">Figure 7</xref>B).</p><fig id="pcbi-0020090-g007" position="float"><label>Figure 7</label><caption><title>Wiggle Predictors Identify Important FFR in PVUII Endonuclease</title><p>(A) Plot of FF scores and mapping of values in a gradient code from negative (blue) to positive (red) onto the structure of PVUII endonuclease in complex with DNA (yellow). The following structural features are labeled: (1) minor groove binding loop, (2) catalytic loop, (3) potential hinge for DNA binding, (4) tyrosine 94 for Mg⁺⁺ ion coordination, and (5) major groove binding loop. (PDB ID: 3PVI).</p><p>(B) Wiggle predictions (solid line) are mapped in green and Wiggle200 predictions (dashed line) are mapped in orange onto the structure.</p></caption><graphic xlink:href="pcbi.0020090.g007"/></fig><p>Y94 coordinates Mg⁺⁺ ions needed for endonuclease activity in this restriction enzyme [<xref rid="pcbi-0020090-b049" ref-type="bibr">49</xref>]. Despite the availability of numerous crystal structures for this protein, no electron density was observed for Y94 until the enzyme was cocrystallized with Mg⁺⁺ [<xref rid="pcbi-0020090-b049" ref-type="bibr">49</xref>] ion, a necessary cofactor for protein function. This is indicative of the need for FF to facilitate metal ion binding, a result supported here. Structural inspection suggests that the other identified residues, unconfirmed in the literature, fall into regions that may serve as hinges for the major groove DNA recognition domain. This region could serve as a possible target for experimental studies to understand the dynamics of this protein.</p></sec><sec id="s2f3"><title>Erythropoietin.</title><p>FFRs identified in erythropoietin contain examples where local flexible regions are stabilized by mutations or glycosylations, both of which are sequence modifications that result in a shift from a disordered to ordered state. No regions of flexibility were identified using the FF score (<xref ref-type="fig" rid="pcbi-0020090-g008">Figure 8</xref>A) in this protein which functions in initiating differentiation and proliferation of progenitor cells into red blood cells. The system modeled by the GNM is a bound unit of erythropoietin to the corresponding receptor (not displayed in <xref ref-type="fig" rid="pcbi-0020090-g008">Figure 8</xref>). As a result, the fluctuation of erythropoietin appears to be diminished.</p><fig id="pcbi-0020090-g008" position="float"><label>Figure 8</label><caption><title>Wiggle Predictors Identify Regions Corresponding to Glycosylation Sites on Erythropoietin</title><p>(A) FF score plotted against residue number with thresholds shown in red. Erythropoietin is modeled by the GNM in the complexed form with the corresponding receptor (not shown). All residues have below mean fluctuation (colored blue), but none of the residues are defined as FFRs since they do not exceed the definition threshold. The four glycosylation sites (S126 and lysine substituted K24, K38, and K83) along with G151 are labeled. (PDB ID: 1EER)</p><p>(B) FFRs correspond to positive values as predicted by Wiggle (solid line) and Wiggle200 (dashed line) which are structurally mapped onto erythropoetin (green and orange, respectively). Not all loops are identified by the predictors to be functionally flexible, thus showing that discrimination is not based on structural features.</p></caption><graphic xlink:href="pcbi.0020090.g008"/></fig><p>Overlaps were found between prediction results (Wiggle: residues 1, 16 to 40, 85 to 89, 113 to 121, 123, 124, 149 to 155, and 160 to 166; Wiggle200: residues 19 to 40, 50 to 57, 86 to 90, 92, 111 to 124, 126 to 128, 139, 150 to 152, 154, 155, 157, and 162 to 166) and correspond to mutations introduced for the creation of a soluble analog [<xref rid="pcbi-0020090-b051" ref-type="bibr">51</xref>] to obtain a crystal structure. All five mutations (N24K, N38K, N83K, P121N, and P122S) reside in, or are immediately adjacent to, positively classified regions (<xref ref-type="fig" rid="pcbi-0020090-g008">Figure 8</xref>B). These mutations include lysine substitutions made at <italic>N</italic>-linked glycosylation sites and prolines removed from the CD loop which contained conformational heterogeneity. Wiggle identified the CD loop and all glycosylation sites as FFRs, with the exception of one glycosylation site where the adjacent region is predicted. Additionally, a kink introduced by G151 was also identified as an FFR.</p><p>Glycosylation of erythropoietin is necessary for its biosynthesis and bioactivity and plays a critical role in its stability [<xref rid="pcbi-0020090-b052" ref-type="bibr">52</xref>,<xref rid="pcbi-0020090-b053" ref-type="bibr">53</xref>]. Removal of all carbohydrates results in aggregation of the protein [<xref rid="pcbi-0020090-b054" ref-type="bibr">54</xref>] and can be made soluble in vitro with mutations N24L, N38L, and N83L [<xref rid="pcbi-0020090-b014" ref-type="bibr">14</xref>]. However, these mutations do not prevent the formation of insoluble aggregates or rapid degradation in vivo [<xref rid="pcbi-0020090-b052" ref-type="bibr">52</xref>]. Carbohydrates increase the half-life of this hormone, but binding affinity is negatively impacted by 20-fold [<xref rid="pcbi-0020090-b055" ref-type="bibr">55</xref>].</p><p>G151 plays an important structural role by introducing a kink in the &#x003b1;D helix. This enables K152 to come in contact with residues in the protein core to form one of the two interaction sites for erythropoietin receptors. Alanine replacement in either position 151 or 152 resulted in a substantial loss of bioactivity [<xref rid="pcbi-0020090-b056" ref-type="bibr">56</xref>]. Both of these positions were identified by Wiggle predictors as FFRs. Binding of erythropoietin to its receptor leads to a slight increase in alpha helical content [<xref rid="pcbi-0020090-b057" ref-type="bibr">57</xref>]. NMR and X-ray structures of erythropoietin in the unbound and bound state, respectively, showed the formation of a small alpha helix occurring in the highly flexible CD loop and a less pronounced kink observed at G151 in the receptor bound X-ray structure [<xref rid="pcbi-0020090-b058" ref-type="bibr">58</xref>]. These are examples of two regions where the structural changes resulting from binding interactions to the receptor correspond to local flexible regions allowing these changes to occur.</p><p>Mutagenesis performed to identify erythropoietin receptor binding sites revealed four regions (residues 11 to 15, 44 to 51, 100 to 108, and 147 to 151) important for the activation of receptor signaling [<xref rid="pcbi-0020090-b059" ref-type="bibr">59</xref>]. With the exception of residues 149 to 152, functionally flexible predictions were made outside of these binding hot spots. This shows that our predictors are not trained to predict binding sites, but rather regions where flexibility is important for bioactivity or accommodating different conformational states.</p></sec></sec><sec id="s2g"><title>Comparison to Protein Disorder Predictions</title><p>Several protein disorder predictors were compared to Wiggle and Wiggle200 predictions (<xref ref-type="fig" rid="pcbi-0020090-g009">Figure 9</xref>) to illustrate that these predictors identify different targets. Disorder predictors differ widely in their approaches, but targets are generally based on high temperature factors or missing residues in crystal structures. PONDR [<xref rid="pcbi-0020090-b060" ref-type="bibr">60</xref>] is a disorder predictor trained on fractional composition and hydropathy. DISOPRED [<xref rid="pcbi-0020090-b061" ref-type="bibr">61</xref>] uses the PSI-blast matrix as input to an SVM to detect disorder, while DisEMBL [<xref rid="pcbi-0020090-b062" ref-type="bibr">62</xref>] is a neural network trained for the predictions of coils, hot coils, and disorder. RONN [<xref rid="pcbi-0020090-b063" ref-type="bibr">63</xref>] uses a bio-basis function neural network to take advantage of information embedded in homologous proteins. GlobPlot [<xref rid="pcbi-0020090-b064" ref-type="bibr">64</xref>] and FoldIndex [<xref rid="pcbi-0020090-b065" ref-type="bibr">65</xref>] are simpler algorithms that, respectively, use running propensity for protein disorder and an index that classifies residues based on hydrophobicity and net charge. IUPRED [<xref rid="pcbi-0020090-b066" ref-type="bibr">66</xref>] uses concepts of pair-wise interaction potentials observed in globular proteins to make assignments for each residue. Finally, NORSP [<xref rid="pcbi-0020090-b067" ref-type="bibr">67</xref>] assesses regions based on low confidence predictions for secondary structural elements.</p><fig id="pcbi-0020090-g009" position="float"><label>Figure 9</label><caption><title>Comparison of Wiggle Predictors to Structural Disorder Predictors</title><p>Comparison of prediction results from Wiggle (red) to various disorder predictors (blue).</p></caption><graphic xlink:href="pcbi.0020090.g009"/></fig><p>Some overlaps are expected with disorder predictions because FFRs may be disordered depending on the conformational state of the protein. Otherwise, we expect little correlation since disorder predictors generally aim to identify structural disorder and regions with a low propensity to form an ordered unit. Potential functional roles were not considered in their design, although these regions are suggested to be important for protein-protein recognition after examining positively classified sequences [<xref rid="pcbi-0020090-b068" ref-type="bibr">68</xref>,<xref rid="pcbi-0020090-b069" ref-type="bibr">69</xref>]. With the exception of the arc repressor where predictor results exhibited significant overlap, Wiggle and Wiggle200 have been found to target regions that were not otherwise identified by disorder predictors.</p><p>For arc repressor (1BAZ), disorder predictors positively classified terminal ends, although some failed to identify it altogether. The hinge region connecting the two helices is not fully identified by most disorder predictors. While Wiggle predictors did not identify all residues involved in recognition at the major groove for PVUII endonuclease (3PVI), it identified the minor groove recognition loop, catalytic loop, and magnesium ion coordinating residues. Current disorder predicting tools failed to identify these regions. Disorder predictors that successfully identified at least one of these regions are based on an index separating hydrophobicity and net charge (FoldIndex and GlobPlot) or the use of homology information (RONN).</p><p>Most disorder predictors failed to identify all glycosylation sites on erythropoietin (1EER) with the exception of DisEMBL, having the most overlap in predictions with Wiggle. The structure of erythropoietin is entirely helical, and DisEMBL has been designed to predict coils with high B factors. The glycine kink was also missed by most disorder predictors except for DisEMBL and FoldIndex.</p><p>We also compare the performance of predictors in identifying FFRs as defined by the FF score (<xref ref-type="table" rid="pcbi-0020090-t003">Table 3</xref>). Two test sets were used: TESTALL and TEST200 containing randomly selected chains from the training dataset for all proteins and proteins up to 200 residues long, respectively. These test sets were used during one of the cross-validation runs from which the Wiggle predictors were created; therefore, the performance results reflect unseen cases for Wiggle. The results show that DISOPRED was able to identify FFRs with the highest accuracy for both test sets (TESTALL: 78.48%, TEST200: 75.20%). However, DISOPRED failed to identify FFRs as indicated by the poor recall (TESTALL: 11.54%, TEST200: 12.89%). The predictor is therefore poor at identifying FFRs by identifying most residues to be a non-FFRs despite having a high precision. We observed earlier that the residue pool is disproportionate with the FF score identifying about 20% of the residues to be located in an FFR.</p><table-wrap id="pcbi-0020090-t003" content-type="2col" position="float"><label>Table 3</label><caption><p>Comparison of Predictors Using TEST200 and TESTALL</p></caption><graphic xlink:href="pcbi.0020090.t003"/></table-wrap><p>We report the performance of Wiggle on TESTALL and Wiggle200 on TEST200. Wiggle predictors outperformed the other disorder predictors in overall performance for both test sets when comparing precision and recall values (<xref ref-type="table" rid="pcbi-0020090-t003">Table 3</xref>). These results are expected since the predictors were all trained to identify a different target property of proteins. Our predictors were designed to identify regions of flexibility with functional importance unlike the other predictors that target highly disordered regions. The comparison of predictors is an important demonstration to illustrate that the target regions identified are different. This comparison is not intended to measure or make an assessment regarding the ability of Wiggle predictors to identify protein disorder. That our test cases are actually solved structures implies some level of order for the regions to be identified.</p></sec><sec id="s2h"><title>Conclusion</title><p>The motivation for this work is to advance our understanding of protein sequence and FF through easily applied in silico methods. Protein fold and disorder properties are encoded in the amino acid sequence. We believe that functionally important protein flexibility is also encoded in the primary sequence and have successfully created tools to identify these regions. We created two predictors; one specialized for proteins shorter than 200 residues and another for all proteins regardless of size. Between the two predictors, we correctly identified flexible regions of functional importance in several test cases where structure-based classification had difficulties. Our targets include hinges, recognition loops, and localized regions that may serve to accommodate entropy dislocation necessary for allostery.</p><p>We focused on regional motion important for protein function based on residue participation in correlated low-frequency fluctuations that correspond to large global changes as modeled by the GNM. Our predictors differ from other predictors by including an additional functional consideration in our targets used for training our SVMs. Secondary structure predictors are trained against well-ordered regions of proteins to identify regular secondary structural elements and disorder predictors have been trained using various definitions that include regions missing electron density in X-ray structures or have high temperature factors. Both focus on a subset of sequence space important for structural features but do not address patterns involved in modulated protein flexibility that switch between ordered and disordered states.</p><p>With the Wiggle predictors, we were able to show detection of domain boundary and experimentally confirmed FFR in specific examples. Comparison to disorder predictors shows that, while there are expected overlaps, different regions are identified. The difference between predictors is that Wiggle predictors are trained to select for residues participating in the two largest modes of global motion, whereas disorder predictors were trained on the propensity to form ordered structures or lack thereof.</p><p>While false prediction error rates are approximately 30%, this may largely be attributed to the difficulties of defining our regions of interest with misclassifications occurring in both directions when using the FF score. SVMs trained on partitioned datasets showed improved performance, suggesting that the characteristics of FFRs are related to protein size. The Wiggle predictors are especially useful for proteins where no structural data are available. Localizing regions of FF in the absence of structural information will help identify mutational hot spots that may modulate bioactivity and these regions can be targeted in protein engineering experiments. The identification of FFRs by sequence-based methods complements and reduces the limitations in structure-based definitions of flexible regions.</p></sec></sec><sec id="s3"><title>Materials and Methods</title><sec id="s3a"><title>Training set.</title><p>A nonredundant training set of protein chains with percent sequence identity of less than or equal to 10%, resolution better than 2.0 &#x000c5;, and an R-factor less than 0.30 were retrieved from the PDB [<xref rid="pcbi-0020090-b070" ref-type="bibr">70</xref>] using PISCES [<xref rid="pcbi-0020090-b071" ref-type="bibr">71</xref>]. We further ensure nonredundancy by checking for distant protein homologs within the retrieved dataset using PSI-BLAST [<xref rid="pcbi-0020090-b072" ref-type="bibr">72</xref>]. Each protein in the dataset was used as a query to search against a sequence database clustered with CD-HIT [<xref rid="pcbi-0020090-b073" ref-type="bibr">73</xref>&#x02013;<xref rid="pcbi-0020090-b075" ref-type="bibr">75</xref>] at 90% identity. Distant homologs within the dataset (111 pairs) were eliminated if the sequence was retrieved by PSI-BLAST.</p><p>The final training set contained 1,277 sequences with 56.6% of the chains existing in the monomeric state. Multiple copies of a protein found in the asymmetric unit were eliminated. Complexes were manually inspected using the protein quaternary structure file server (PQS) [<xref rid="pcbi-0020090-b076" ref-type="bibr">76</xref>] and literature confirmation sought for biological relevance. If the complexes were not found in nature, they were removed from the training dataset. The training set was then partitioned into two subsets based on protein length and used to train specialized SVMs. Subset A contained 720 proteins of length less than or equal to 200 amino acids; subset B contained 557 proteins of length greater than 200 amino acids.</p></sec><sec id="s3b"><title>HMMs.</title><p>SAM-2tk [<xref rid="pcbi-0020090-b037" ref-type="bibr">37</xref>] was used to build HMMs for all sequences in the training datasets. Homologs for each sequence in the training set were retrieved from a sequence database clustered at 65% identity with CD-HIT [<xref rid="pcbi-0020090-b073" ref-type="bibr">73</xref>]. Clustering affects the probability states in the HMM; it was therefore important to check that patterns detected by prediction methods were not eliminated as a result. We tested the impact of CD-HIT on secondary structure predictions and found slight improvements in prediction quality (data not shown). Therefore, for reasons of increased remote homolog detection, reduced computational search time, and improved secondary structure prediction, the clustered sequence database was used in building HMMs using a target entropy weighting of 1.0 bit per column.</p></sec><sec id="s3c"><title>GNM.</title><p>The GNM [<xref rid="pcbi-0020090-b016" ref-type="bibr">16</xref>,<xref rid="pcbi-0020090-b077" ref-type="bibr">77</xref>] combines the simplicity of the elastic theory applied to random polymer network [<xref rid="pcbi-0020090-b078" ref-type="bibr">78</xref>] and the success of using a single-parameter potential [<xref rid="pcbi-0020090-b079" ref-type="bibr">79</xref>] to model protein dynamics based on coordinates of the C_&#x003b1; atoms serving as nodes. The connectivity within the protein structure is represented as a Kirchhoff matrix &#x00393; where R is the distance between the C_&#x003b1; atoms of residues <italic>i</italic> and <italic>j</italic> with r_c denoting the distance radius threshold (7 &#x000c5;).</p><disp-formula id="pcbi-0020090-e001"><graphic xlink:href="pcbi.0020090.e001.jpg" position="anchor" mimetype="image"/></disp-formula><p>The equilibrium-correlated fluctuations between two sites can be obtained by finding the inverse of the Kirchhoff matrix and is represented as: <disp-formula id="pcbi-0020090-e002"><graphic xlink:href="pcbi.0020090.e002.jpg" position="anchor" mimetype="image"/></disp-formula>where <italic>k_b</italic> is the Boltzmann constant, <italic>T</italic> is the absolute temperature, and &#x003b3; is a single-parameter harmonic potential that accounts for the fluctuations of a residue about a mean axis. </p><p>Cross-correlated fluctuations between residues <italic>i</italic> and <italic>j</italic> are defined as:</p><disp-formula id="pcbi-0020090-e003"><graphic xlink:href="pcbi.0020090.e003.jpg" position="anchor" mimetype="image"/></disp-formula><p>Participation in correlated movements was used to define flexible regions that are functionally important. Readers are referred to the original papers for details.</p></sec><sec id="s3d"><title>Definition of FFRs.</title><p>Operationally, FFRs are defined using normalized FF scores. For each residue <italic>i,</italic> the maximum and minimum values, corresponding to residues <italic>m</italic> and <italic>n,</italic> respectively, are extracted from the cross-correlation matrix <italic>C.</italic> These values, <italic>C(i,m)</italic> and <italic>C(i,n),</italic> are used to scale the weighted average of the top two modes <italic>j</italic> of protein fluctuation where &#x003bc; is the eigenmode and &#x003bb; is the corresponding eigenvalue.</p><disp-formula id="pcbi-0020090-e004"><graphic xlink:href="pcbi.0020090.e004.jpg" position="anchor" mimetype="image"/></disp-formula><p>FF scores are normalized for each protein after removing outliers using a median-based approach [<xref rid="pcbi-0020090-b080" ref-type="bibr">80</xref>]. To distinguish outliers, the median of the absolute difference <italic>(mad),</italic> taken between FF scores and the median of FF scores <italic>(m₁),</italic> for the protein is first calculated. Each residue is then assigned an <italic>M</italic> value to identify and exclude outliers, defined by <italic>M</italic> &#x0003e; 3.5 and <italic>M</italic> &#x0003c; &#x02212;3.5, prior to the calculation of the mean and standard deviation for normalization. For large sample sizes, the expected value of <italic>mad</italic> is 0.6745&#x003c3;. <disp-formula id="pcbi-0020090-e005"><graphic xlink:href="pcbi.0020090.e005.jpg" position="anchor" mimetype="image"/></disp-formula> <disp-formula id="pcbi-0020090-e006"><graphic xlink:href="pcbi.0020090.e006.jpg" position="anchor" mimetype="image"/></disp-formula>The calculated mean and standard deviation, obtained after exclusion of outliers, were used to normalize FF scores to a mean of 0 and standard deviation of 1. This normalization process rescales the protein fluctuation such that the mean fluctuation values are centered about the value 0. </p><disp-formula id="pcbi-0020090-e007"><graphic xlink:href="pcbi.0020090.e007.jpg" position="anchor" mimetype="image"/></disp-formula><p>FFRs are defined to contain amino acids with <italic>FF_norm</italic> &#x0003e; 1.5 or <italic>FF_norm</italic> &#x0003c; &#x02212;1.5. This threshold is chosen empirically based on the assumption that fluctuations differing from the mean fluctuation of the entire modeled system will be important for protein functionality.</p></sec><sec id="s3e"><title>Bootstrapping for sequence preferences.</title><p>A modified bootstrap approach was used to identify sequence preferences for FFRs defined by the FF score. The aim of this analysis is to use these findings as additional input features for SVM-based classification. Protein sequences in the dataset were window scanned to pool triplets found in the training set. These pooled triplets were analyzed to identify sequence pattern distributions most correlated with FFR and non-FFR classifications. Two null models were created, one for FFRs and another for non-FFRs, by randomly selecting from the pooled triplets with replacement. Samples were drawn to be the same size as observed for FFR and non-FFR classes. <italic>Z</italic>-scores and <italic>p</italic>-values were calculated using the generated null model distribution for each observed triplet in their respective category. These classification preferences were included as additional input features to help the SVMs identify FFRs.</p></sec><sec id="s3f"><title>SVMs.</title><p>All training schemes were performed with 5-fold cross-validation using SVM<italic>light</italic> [<xref rid="pcbi-0020090-b081" ref-type="bibr">81</xref>]. Positively categorized residues were matched by one randomly selected negative residue to create a 1:1 ratio during training. The linear kernel model was initially used to conduct performance comparisons between different SVM architectures. This kernel was chosen because the need for parameter optimization is eliminated, thus providing a faster alternative for preliminary comparisons. Performances of SVMs were evaluated based on accuracy, precision, and recall where the ratio of relative true-positive (TP), true-negative (TN), false-positive (FP), and false-negative (FN) is examined. Unlike the training phase, no residues were excluded during performance evaluations of SVM performance.</p><disp-formula id="pcbi-0020090-e008"><graphic xlink:href="pcbi.0020090.e008.jpg" position="anchor" mimetype="image"/></disp-formula><disp-formula id="pcbi-0020090-e009"><graphic xlink:href="pcbi.0020090.e009.jpg" position="anchor" mimetype="image"/></disp-formula><disp-formula id="pcbi-0020090-e010"><graphic xlink:href="pcbi.0020090.e010.jpg" position="anchor" mimetype="image"/></disp-formula><p>The predictor architecture for both Wiggle and Wiggle200 contains two layers. Input features for the first layer SVM include the nine HMM transition states and 20 match states. In HMM models, the match state probabilities give the probability of observing an amino acid at a particular position. The transition state probability is the probability of changing from one state (deletion, insertion, or match) to another from the previous state. For a window size of 9, a total of 261 (9 &#x000d7; 29) input features were used for each residue. Values are set to 0 when the window extends beyond terminal ends.</p><p>The prediction results from this first layer SVM is then included along with calculated <italic>Z</italic>-scores and <italic>p</italic>-values obtained for triplets from the modified bootstrap analysis as input features into a second-layer SVM. We find that using the radial basis kernel function to model input features for the first-layer SVM (&#x003b3; = 0.25, C = 2) and the linear kernel function for the second-layer SVM to yield the best performing predictors.</p><p>With this two-layer architecture and optimized parameters, two different predictors were developed defined by their training sets. Wiggle was trained on the entire training set, while Wiggle200 is a more specialized predictor trained on proteins up to 200 amino acids in length.</p></sec><sec id="s3g"><title>Assessment of domain boundary predictions.</title><p>Wiggle prediction results were compared to a benchmark dataset (BENCH) reflecting the consensus of domain boundaries among CATH, SCOP, and authors of the three-dimensional structures (T. Holland, S. Veretnik, I. N. Shindyalov, and P. E. Bourne, unpublished data).</p><p>This dataset contains 312 chains, of which 66% are multidomain proteins, covering 30 distinct architectures and 211 distinct topologies as defined by CATH.</p><p>The prediction performance was measured based on accuracy, precision, and recall values. Domain boundaries in the dataset were defined between two adjacent positions. We therefore investigated the performance of predictors for a variety of window sizes, up to 15 residues, with the boundary resting in the middle of the expanse. Performance evaluations were also tested on a partitioned benchmark set based on protein sizes up to 200 residues (BENCHA) and longer (BENCHB).</p></sec><sec id="s3h"><title>Comparison of disorder predictors.</title><p>To compare residue classification of Wiggle predictors to different disorder predictors for the three specific protein comparisons, we set VSL1 version of PONDR to predict with a 10% false-positive rate, and DisEMBL to predict hot coils defined as coils with high B factors. Recommended defaults for a window size of 9 when requested were used for remaining predictors.</p><p>We also compare the performances of disorder predictors with two different test sets (TEST200 and TESTALL) containing randomly selected chains used during the training of Wiggle predictors. TEST200 contains 144 chains up to 200 residues and TESTALL contains 256 chains regardless of length. For disorder predictors, we used the same default values and settings as the specific case example comparisons with the exception of PONDR. The default predictor for PONDR (VLXT) was used to accommodate larger proteins in the test sets. Wiggle was used for TESTALL and Wiggle200 for TEST200.</p></sec></sec>

Comprehensive thermodynamic analysis of 3&#x02032; double-nucleotide overhangs neighboring Watson&#x02013;Crick terminal base pairs

<p>Thermodynamic parameters are reported for duplex formation of 48 self-complementary RNA duplexes containing Watson&#x02013;Crick terminal base pairs (GC, AU and UA) with all 16 possible 3&#x02032; double-nucleotide overhangs; mimicking the structures of short interfering RNAs (siRNA) and microRNAs (miRNA). Based on nearest-neighbor analysis, the addition of a second dangling nucleotide to a single 3&#x02032; dangling nucleotide increases stability of duplex formation up to 0.8 kcal/mol in a sequence dependent manner. Results from this study in conjunction with data from a previous study [A. S. O'Toole, S. Miller and M. J. Serra (2005) <italic>RNA</italic>, <bold>11</bold>, 512.] allows for the development of a refined nearest-neighbor model to predict the influence of 3&#x02032; double-nucleotide overhangs on the stability of duplex formation. The model improves the prediction of free energy and melting temperature when tested against five oligomers with various core duplex sequences. Phylogenetic analysis of naturally occurring miRNAs was performed to support our results. Selection of the effector miR strand of the mature miRNA duplex appears to be dependent upon the identity of the 3&#x02032; double-nucleotide overhang. Thermodynamic parameters for 3&#x02032; single terminal overhangs adjacent to a UA pair are also presented.</p>

<contrib contrib-type="author"><name><surname>O'Toole</surname><given-names>Amanda S.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Miller</surname><given-names>Stacy</given-names></name></contrib><contrib contrib-type="author"><name><surname>Haines</surname><given-names>Nathan</given-names></name></contrib><contrib contrib-type="author"><name><surname>Zink</surname><given-names>M. Coleen</given-names></name></contrib><contrib contrib-type="author"><name><surname>Serra</surname><given-names>Martin J.</given-names></name><xref ref-type="corresp" rid="cor1">*</xref></contrib><aff><institution>Department of Chemistry, Allegheny College</institution><addr-line>520 N. Main Street, Meadville, PA 16335, USA</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p>Single and double-nucleotide overhangs on the 3&#x02032; end of RNA duplexes have previously been shown to contribute to the stability of a duplex in a sequence dependent manner (<xref ref-type="bibr" rid="b1">1</xref>&#x02013;<xref ref-type="bibr" rid="b5">5</xref>). The increase in duplex stability by 3&#x02032; dangling nucleotides is attributed to stacking interactions that dangling bases form with neighboring closing base pairs in the duplex as a result of A-form helical geometry. Base identity of 3&#x02032; double-nucleotide overhangs is critical in determining thermodynamic stability of duplexes.</p><p>Understanding the thermostability of RNA duplexes with 3&#x02032; double-nucleotide overhangs is essential for understanding some of the major biological roles of this structure. The 3&#x02032; double-nucleotide overhang in the sequences used in this study mimic the structure of short interfering (siRNAs) and micro RNAs (miRNAs) (<xref ref-type="bibr" rid="b6">6</xref>&#x02013;<xref ref-type="bibr" rid="b10">10</xref>). siRNAs are formed when the ribonuclease III enzyme, dicer, processes long non-coding double-stranded RNA (dsRNA) in the cell into &#x0223c;21 nt oligonucleotides containing a 19mer duplex with double-nucleotide overhangs on the 3&#x02032; ends (<xref ref-type="bibr" rid="b11">11</xref>&#x02013;<xref ref-type="bibr" rid="b15">15</xref>). One strand of the siRNA duplex, called the &#x02018;guide&#x02019; strand is then incorporated into the RNA Induced Silencing Complex (RISC) and target mRNA complementary in sequence to the siRNA, while the other strand, the &#x02018;passenger&#x02019; strand is degraded. miRNAs are endogenous to the cell and occur when a large RNA hairpin referred to as a pri-miRNA is processed into a smaller RNA hairpin pre-miRNA. This structure is further processed by the dicer enzyme into a mature miRNA sequence containing a miR/miR* duplex &#x0223c;19 bp with 2 nt overhangs on the 3&#x02032; ends. The guide, or miR, strand of the mature miRNAs are also loaded into the RNAi effector complex RISC to target complementary mRNAs, while the miR* strand is degraded. Once incorporated into RISC, siRNAs and miRNAs pair with mRNA of complementary sequence and either induce cleavage of the target mRNA or translational repression of the mRNA message (<xref ref-type="bibr" rid="b16">16</xref>). Many groups have shown that both siRNAs and mature miRNAs are incorporated into RISC in an asymmetric manner which is dependent on the stability of the base pairing at the 3&#x02032; ends of each strand in the duplex (<xref ref-type="bibr" rid="b17">17</xref>,<xref ref-type="bibr" rid="b18">18</xref>). Until now there has not been an accurate method for calculating the contribution of 3&#x02032; double-nucleotide overhangs on thermodynamic stability of duplex formation and therefore no reliable way of predicting which strand of the siRNA or the miRNA will be loaded into RISC as the effector strand in RNAi.</p><p>We have shown previously for duplexes containing a CG closing base pair, a double-nucleotide overhang on the 3&#x02032; end of the duplex where the addition of a second dangling nucleotide to a single purine dangling nucleotide can enhance stabilization of the duplex. However, if the first dangling nucleotide is a pyrimidine then the addition of a second dangling nucleotide will not provide any additional stabilization of the duplex O'Toole <italic>et al</italic>. (<xref ref-type="bibr" rid="b5">5</xref>). Here we have included in our study all possible 3&#x02032; double-nucleotide dangling ends using Watson&#x02013;Crick bases on core duplexes containing the remaining three possible orientations of Watson&#x02013;Crick terminal base pairs: GC, AU and UA. The thermodynamic parameters obtained from this study along with those from our previous work have allowed us to develop a model for improved prediction of stability of a duplex with a 3&#x02032; double-nucleotide dangling end.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><sec><title>RNA synthesis and purification</title><p>All oligomers were synthesized on CPG solid supports (Applied Biosystems 392 DNA/RNA Synthesizer) with phosphoramidites with the 2&#x02032; hydroxyl protected as the <italic>tert</italic>-butyl dimethylsilyl ether from Glen Reseach (Sterling VA). Oligomers underwent ammonia and fluoride deprotection, and the crude sample was purified using preparative TLC (<italic>n</italic>-propanol:ammonium hydroxide:water, 55:35:10) and Sep-Pak C18 (Waters) chromatography as previously described (<xref ref-type="bibr" rid="b19">19</xref>). Sample purity was determined through analytical TLC or HPLC (C-18), and was &#x0003e;95%.</p></sec><sec><title>Melting curve and data analysis</title><p>Optical melting experiments were performed using a Beckman DU 640 Spectrophotometer and High Performance Temperature Controller at 280 nm. Absorbance changes for oligomers in 1 M NaCl melt buffer (1 M NaCl, 0.01 M cacodylic acid, 0.001 M ethylenediamine tetraacetic acid, pH 7.0) were recorded as function of temperature from 90 to 5&#x000b0;C at a rate of 1&#x000b0;C/min as described previously (<xref ref-type="bibr" rid="b19">19</xref>). The experiment was repeated at 10 varying sample concentrations to give at least a 50-fold concentration range (10 &#x003bc;M&#x02013;1 mM) for each sample. Absorbance versus temperature profiles were fit to a two-state model with sloping base lines using a non-linear least squares program (<xref ref-type="bibr" rid="b20">20</xref>). Thermodynamics parameters for the oligomers were determined from both the average of the individual melt curves and plots of the reciprocal melting temperature (<inline-formula><mml:math id="M1"><mml:mrow><mml:msubsup><mml:mi>T</mml:mi><mml:mtext>M</mml:mtext><mml:mrow><mml:mo>&#x02212;</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msubsup></mml:mrow></mml:math></inline-formula>) versus ln(<italic>C</italic>_t) for self-complementary sequences or (<inline-formula><mml:math id="M2"><mml:mrow><mml:msubsup><mml:mi>T</mml:mi><mml:mtext>M</mml:mtext><mml:mrow><mml:mo>&#x02212;</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msubsup></mml:mrow></mml:math></inline-formula>) versus ln(<italic>C</italic>_t/4) for non-self-complementary sequences. Parameters derived from the two methods agreed within 10%, consistent with the two-state model (<xref ref-type="bibr" rid="b21">21</xref>).</p></sec><sec><title>Phylogenetic analysis</title><p>A total of 1290 experimentally validated miRNA sequences from miRBase release 8.0 (<ext-link ext-link-type="uri" xlink:href="http://microrna.sanger.ac.uk/sequences/"/>; February 2006) were &#x02018;conceptually diced&#x02019; using an algorithm which &#x02018;dices&#x02019; the pre-miRNA sequences based on a 19 nt region of base pairing with a 3&#x02032; double-nucleotide overhangs. These mature miRNA sequences were then analyzed for the frequency of Watson&#x02013;Crick closing base pairs with 3&#x02032; double-nucleotide overhangs. Of the 1290 sequences 1009 fit our criteria. The occurrence of each of the possible combinations of Watson&#x02013;Crick closing base pairs with double-nucleotide overhang on both the miR and miR* strands of the mature miRNA sequences was determined.</p></sec></sec><sec><title>RESULTS AND DISCUSSION</title><p>For duplexes containing terminal CG base pairs, stability of duplex formation has been shown to be influenced by the second nucleotide of a 3&#x02032; double-nucleotide overhang (<xref ref-type="bibr" rid="b5">5</xref>). The additional stability of the duplex varies based on the identity of the nucleotides in the overhang; in the order pur&#x02013;pyr &#x0003e; pur&#x02013;pur &#x0003e; pyr&#x02013;pyr = pyr&#x02013;pur = 0. To explore the generality of the trend observed previously, we examined the influence of 3&#x02032; double-nucleotide overhangs on duplexes containing all possible Watson&#x02013;Crick closing base pairs.</p><sec><title>Thermodynamic data</title><p>The measured thermodynamic parameters for all 48 of the 3&#x02032; double-nucleotide terminal overhangs are presented in <xref ref-type="table" rid="tbl1">Table 1</xref>. Thermodynamic parameters were determined using both melt curve analysis and the <italic>T</italic>_M dependence models (<xref ref-type="bibr" rid="b20">20</xref>). Data from both models agreed within 10% for all sequences, consistent with the two-state model with the exception of (CCGGCG)₂. The average deviations in thermodynamic parameter values are &#x000b1;5.6, 6.3 and 2.0% for &#x00394;<italic>H</italic>^o, &#x00394;<italic>S</italic>^o and <inline-formula><mml:math id="M3"><mml:mrow><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mtext>37</mml:mtext></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup></mml:mrow></mml:math></inline-formula>, respectively.</p><p>We have shown previously that 3&#x02032; double-nucleotide dangling ends increase the stability of RNA duplexes relative to the duplexes with only a single 3&#x02032; dangling nucleotide when the first-nucleotide overhang is a purine residue. For the oligomers in this study, (<xref ref-type="table" rid="tbl1">Table 1</xref>), the second 3&#x02032; dangling nucleotide changes the <italic>T</italic>_M of the helix by &#x02212;3.2 to &#x02212;9.2&#x000b0;C (average increase 2.4&#x000b0;C) for a 10^&#x02212;4 M solution. Additional stabilization of the duplex by the second dangling nucleotide is sequence dependent. The average increase in <italic>T</italic>_M for helices where the first dangling nucleotide is a purine is 3.6&#x000b0;C, while helices where the first dangling nucleotide is a pyrimidine, the average increase is 1.3&#x000b0;C (2.6&#x000b0;C for cytodine and 0.0&#x000b0;C for uridine). These increases are within the range seen previously for the addition of a second dangling nucleotide on the 3&#x02032; end of a duplex with a CG closing base pair (<xref ref-type="bibr" rid="b5">5</xref>).</p></sec><sec><title>Free energy parameters for 3&#x02032; terminal dangling ends on U/A base pair</title><p>Thermodynamic values have been previously determined for all the duplexes in this study containing all possible single 3&#x02032; dangling nucleotides (<xref ref-type="bibr" rid="b1">1</xref>,<xref ref-type="bibr" rid="b3">3</xref>), except for those with the (AGCGCU)₂ core. We chose this core because the previously measured single 3&#x02032; dangling nucleotides on UA closing base pairs (<xref ref-type="bibr" rid="b3">3</xref>) had been measured on duplexes with a core of (AUGCAU)₂. However, the terminal 2 bp, AU base pair neighboring a terminal UA base pair, are likely to fray affecting the interaction of the terminal base pair with the dangling ends. The core sequence used in this study (AGCGCU)₂, with a CG base pair neighboring a single terminal UA base pair is less likely to fray. In order to determine the stability contributed to the UA oligomer by the 3&#x02032; dangling nucleotide, differences in the thermodynamic values between the sequences studied and the corresponding core sequence were also determined for all of the oligomers tested using equations similar to <xref ref-type="disp-formula" rid="e1">Equation 1</xref>. Thermodynamic values for duplex formation of the core sequence are &#x02212;50.3 kcal/mol, &#x02212;136.2 eu and &#x02212;8.0 kcal/mol for &#x00394;<italic>H</italic>^o, &#x00394;<italic>S</italic>^o and <inline-formula><mml:math id="M4"><mml:mrow><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mtext>37</mml:mtext></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup></mml:mrow></mml:math></inline-formula>, respectively (<xref ref-type="bibr" rid="b22">22</xref>). <disp-formula id="e1"><label>1</label><mml:math id="M5"><mml:mrow><mml:mi>&#x00394;</mml:mi><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mn>37</mml:mn></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup><mml:msup><mml:mtext>3</mml:mtext><mml:mo>&#x02032;</mml:mo></mml:msup><mml:mtext>X</mml:mtext><mml:mo>=</mml:mo><mml:mfrac><mml:mn>1</mml:mn><mml:mn>2</mml:mn></mml:mfrac><mml:mrow><mml:mo>[</mml:mo><mml:mrow><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mn>37</mml:mn></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup><mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mtext>AGCGCUX</mml:mtext><mml:mo>)</mml:mo></mml:mrow><mml:mn>2</mml:mn></mml:msub><mml:mo>&#x02212;</mml:mo><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mn>37</mml:mn></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup><mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mtext>AGCGCU</mml:mtext><mml:mo>)</mml:mo></mml:mrow><mml:mn>2</mml:mn></mml:msub></mml:mrow><mml:mo>]</mml:mo></mml:mrow><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula> These results are summarized in <xref ref-type="table" rid="tbl2">Table 2</xref>.</p><p>The effect of terminal fraying is demonstrated by a decreased thermodynamic stabilization with the 3&#x02032; terminal overhang on the (AUGCAU)₂ core compared to stabilization attributed to 3&#x02032; overhangs on other terminal base pairs [(<xref ref-type="bibr" rid="b3">3</xref>) and <xref ref-type="table" rid="tbl2">Table 2</xref>]. This is particularly evident for the 3&#x02032; terminal pyrimidine overhangs where the additional stabilization on the (AUGCAU)₂ core was found to be only 0.1 kcal/mol (<xref ref-type="bibr" rid="b3">3</xref>). Our measured values for the additional duplex stabilization afforded by the 3&#x02032; terminal pyrimidine nucleotides on the (AGCGCU)₂ core is 0.4 kcal/mol (<xref ref-type="table" rid="tbl2">Table 2</xref>). While our values and the previously measured values are within experimental error of each other, our values are in better accordance with the stabilization found for 3&#x02032; terminal pyrimidine nucleotides with other terminal base pairs (<xref ref-type="bibr" rid="b23">23</xref>).</p></sec><sec><title>Nearest-neighbor analysis and free energy parameters for 3&#x02032; dangling double-nucleotide overhangs</title><p>In order to determine the stability contributed to the oligomer by the second 3&#x02032; dangling nucleotide, differences in the thermodynamic values between the sequences studied and the corresponding sequence containing only the first overhanging nucleotide were also determined for all of the oligomers tested using equations similar to <xref ref-type="disp-formula" rid="e2">Equation 2</xref>. <disp-formula id="e2"><label>2</label><mml:math id="M6"><mml:mrow><mml:mi>&#x00394;</mml:mi><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mn>37</mml:mn></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup><mml:msup><mml:mn>3</mml:mn><mml:mo>&#x02032;</mml:mo></mml:msup><mml:mtext>XY</mml:mtext><mml:mo>=</mml:mo><mml:mfrac><mml:mn>1</mml:mn><mml:mn>2</mml:mn></mml:mfrac><mml:mrow><mml:mo>[</mml:mo><mml:mrow><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mn>37</mml:mn></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup><mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mtext>CCGGXY</mml:mtext><mml:mo>)</mml:mo></mml:mrow><mml:mn>2</mml:mn></mml:msub><mml:mo>&#x02212;</mml:mo><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mi>G</mml:mi><mml:mrow><mml:mn>37</mml:mn></mml:mrow><mml:mtext>o</mml:mtext></mml:msubsup><mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mtext>CCGGX</mml:mtext><mml:mo>)</mml:mo></mml:mrow><mml:mn>2</mml:mn></mml:msub></mml:mrow><mml:mo>]</mml:mo></mml:mrow><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula> These results are presented in <xref ref-type="table" rid="tbl3">Table 3</xref>.</p><p>Previously, the influence of the second 3&#x02032; dangling nucleotide on the stability of the duplex, (GGCC)₂ was grouped into three categories based upon the identity of 2 nt in the overhang. If the first 3&#x02032; nucleotide was a pyrimidine, no additional stabilization is observed upon the addition of the second 3&#x02032; dangling nucleotide. When the first nucleotide was a purine, the stacking of the second nucleotide made a significant contribution to the stability of the duplex. Sequences that contained a 3&#x02032; double-nucleotide dangling end with a 3&#x02032;-pur-pyr-3&#x02032; had greater stability (0.5 kcal/mol on average) than those sequences containing an overhang with 3&#x02032;-pur-pur-3&#x02032; (0.3 kcal/mol on average).</p><p>Analysis of all 64 of the 3&#x02032; double dangling nucleotides shows that the influence of the second 3&#x02032; dangling nucleotide does depend upon the identity of the terminal base pair (<xref ref-type="table" rid="tbl3">Table 3</xref>). This is most evident in considering the 3&#x02032; double-purine overhangs; duplexes where the purine of the closing base pair neighbors the double-nucleotide overhang [e.g. (CCGGXY)₂] no additional stabilization by the second nucleotide is observed, however a significant contribution to the stability of the duplex by the second dangling nucleotide is observed when the 3&#x02032; overhang neighbors the pyrimidine of the closing base pair [e.g. (AGCGCUXY)₂]. For pur&#x02013;pyr double-nucleotide overhangs, the orientation of the terminal base pair does not affect the additional stabilization caused by the second 3&#x02032; dangling nucleotide. The additional stabilization of the second nucleotide was not significantly different for the pur&#x02013;pyr double-nucleotide overhang adjacent to a pyrimidine than it was when adjacent to a purine therefore, the thermodynamic values for these sequences were averaged together to arrive at the stabilization of a duplex attributed to the second 3&#x02032; dangling nucleotide. As observed previously (<xref ref-type="bibr" rid="b5">5</xref>), if the first nucleotide of the 3&#x02032; dangling end is a pyrimidine, no additional stabilization is observed, irrespective of the terminal base pair. These results lead to an improvement in the model to predict RNA secondary structure stability and are presented in <xref ref-type="table" rid="tbl4">Table 4</xref>.</p><p>To test the generality of conclusions from this work, thermodynamic parameters were also measured for five test sequences with 3&#x02032; double-nucleotide overhangs on different core sequences than those used to develop the model. The measured and predicted (both with and without the influence of the second 3&#x02032; dangling end) thermodynamic values for the five sequences are presented in <xref ref-type="table" rid="tbl5">Table 5</xref>. Four of the test sequences have only one 3&#x02032; double-nucleotide overhang that increases the stability, therefore, the predicted stability (<inline-formula><mml:math id="M7"><mml:mi>&#x00394;</mml:mi><mml:msubsup><mml:mtext>G</mml:mtext><mml:mn>37</mml:mn><mml:mtext>o</mml:mtext></mml:msubsup></mml:math></inline-formula>) for inclusion of the 3&#x02032; double-nucleotide dangling end increase by 0.5 kcal/mol. For the fifth test sequence, both 3&#x02032; double-nucleotide dangling ends contribute to the duplex stability and therefore, the predicted stability increases by 1.0 kcal/mol. The inclusion of the stabilization caused by the 3&#x02032; double-nucleotide overhangs improves the prediction of the thermodynamic stability for the duplexes in <xref ref-type="table" rid="tbl5">Table 5</xref>. For example, the average difference between the measured free energy and the predicted value is 0.6 kcal/mol when the contribution of the second 3&#x02032; two nucleotide dangling end is not included in the prediction, and improves to 0.3 kcal/mol when the contribution is included. In a similar fashion, the average difference in the prediction of the melting temperature also improves from 3.3 to 1.6&#x000b0;C with the contribution of the second 3&#x02032; double-nucleotide overhang taken into account. This is most strikingly seen for the last duplex in <xref ref-type="table" rid="tbl5">Table 5</xref> where both 3&#x02032; double-nucleotide overhangs contribute to the duplex stability. The inclusion of the contribution of the 3&#x02032; double-nucleotide overhangs at both ends, improves the prediction of the thermal stability by 1.0 kcal/mol and the melting temperature by &#x0003e;6&#x000b0;C.</p><p>Phylogenetic analysis of experimentally determined miRNA sequences from miRNA database revealed a total of 1290 strand sequences; 1009 of these sequences were terminated with a Watson&#x02013;Crick closing base pair and had 3&#x02032; terminal double-nucleotide overhang consisting of two of the four Watson&#x02013;Crick bases. The miR strand of all the 1009 miRNAs that we used in this study were experimentally cloned and sequenced, however since the miR* of miRNA duplexes is degraded after the duplex is unwound and the miR strand is incorporated into RISC, the miR* sequences in the database have been predicted based upon the cloned and sequenced miR strand. Frequency of appearance of each of the 64 possible combinations of Watson&#x02013;Crick closing base pairs neighboring 2 nt 3&#x02032; overhangs were determined for both miR and predicted miR* strands. The sequences were divided into categories and analyzed based upon their stability contribution of the 3&#x02032; double-nucleotide overhang on duplex stability; results of this search are presented in <xref ref-type="table" rid="tbl6">Table 6</xref>. Interestingly, the distribution of sequences into the miR or miR* strand was found to be related to the stability contribution of the 3&#x02032; terminal double-nucleotide overhang. For the double-nucleotide overhangs that do not contribute additional stability to the duplex, there is nearly an equal number of sequences found in both the miR and miR* strands (49 and 51%, respectively). For the double-nucleotide overhangs that do contribute to the stability of the duplex (red in the table), &#x0003e;75% were observed on the miR strand while only 25% were on the miR* strand. The identity of the closing base pair does influence the utilization of stabilizing double-nucleotide overhangs. For strands with either a CG closing base pair or a GC closing base pair (G/C base pair) on the end, the 3&#x02032; double-nucleotide overhang is almost exclusively (&#x0003e;90%) stabilizing; while for duplexes which end with either an AU closing base pair or a UA closing base pair (A/U), there is nearly a 3:1 ratio of non-stabilizing 3&#x02032; double-nucleotide overhangs. It is interesting that the stability of the more stable terminus, G/C, is augmented by the additional stabilization of the 3&#x02032; double-nucleotide overhang, while the opposite is seen for the less stable terminus, A/U. The 3&#x02032; double-nucleotide overhangs may therefore have been selected to enhance the distinction between the two ends of the miRNA duplex and aid in guide and passenger strand selection and loading into the RNAi effector complex, RISC; a critical step in the RNAi pathway.</p></sec></sec>

SAP30L interacts with members of the Sin3A corepressor complex and targets Sin3A to the nucleolus

<p>Histone acetylation plays a key role in the regulation of gene expression. The chromatin structure and accessibility of genes to transcription factors is regulated by enzymes that acetylate and deacetylate histones. The Sin3A corepressor complex recruits histone deacetylases and in many cases represses transcription. Here, we report that SAP30L, a close homolog of Sin3-associated protein 30 (SAP30), interacts with several components of the Sin3A corepressor complex. We show that it binds to the PAH3/HID (Paired Amphipathic Helix 3/Histone deacetylase Interacting Domain) region of mouse Sin3A with residues 120&#x02013;140 in the C-terminal part of the protein. We provide evidence that SAP30L induces transcriptional repression, possibly via recruitment of Sin3A and histone deacetylases. Finally, we characterize a functional nucleolar localization signal in SAP30L and show that SAP30L and SAP30 are able to target Sin3A to the nucleolus.</p>

<contrib contrib-type="author"><name><surname>Viiri</surname><given-names>K. M.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Korkeam&#x000e4;ki</surname><given-names>H.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Kukkonen</surname><given-names>M. K.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Nieminen</surname><given-names>L. K.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Lindfors</surname><given-names>K.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Peterson</surname><given-names>P.</given-names></name><xref rid="au2" ref-type="aff">2</xref></contrib><contrib contrib-type="author"><name><surname>M&#x000e4;ki</surname><given-names>M.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Kainulainen</surname><given-names>H.</given-names></name><xref rid="au3" ref-type="aff">3</xref><xref rid="au4" ref-type="aff">4</xref></contrib><contrib contrib-type="author"><name><surname>Lohi</surname><given-names>O.</given-names></name><xref rid="au1" ref-type="aff">1</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><aff id="au1">¹<institution>Paediatric Research Centre, University of Tampere Medical School and Tampere University Hospital</institution><addr-line>Tampere, Finland</addr-line></aff><aff id="au2">²<institution>Molecular Pathology, University of Tartu</institution><addr-line>Tartu, Estonia</addr-line></aff><aff id="au3">³<institution>Institute of Medical Technology and Tampere University Hospital</institution><addr-line>Tampere, Finland</addr-line></aff><aff id="au4">⁴<institution>Department of Biology of Physical Activity, University of Jyv&#x000e4;skyl&#x000e4;</institution><addr-line>Finland</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p>It is well established that gene expression is influenced by chromatin structure. The compacted chromatin is a sterically hindered environment for transcription factors to bind and assemble the transcription initiation complex, and is subject to active remodeling. Histone acetylation and DNA demethylation are perceived as prerequisites for the &#x02018;open state&#x02019; of chromatin, enabling transcription initiation. On the other hand, histone deacetylation and DNA methylation convert chromatin to a &#x02018;closed state&#x02019;, leading to the silencing of gene transcription. Recently, it has become evident that protein complexes that regulate histone acetylation, chromatin remodeling and DNA methylation work in concert (<xref ref-type="bibr" rid="b1">1</xref>,<xref ref-type="bibr" rid="b2">2</xref>) and at least in the ribosomal DNA locus (rDNA) these epigenetic events occur in this particular hierarchical and temporal order (<xref ref-type="bibr" rid="b3">3</xref>). The Sin3A-HDAC corepressor complex consists of multiple proteins and regulates gene expression by deacetylating histones. Sin3A itself functions as a scaffold protein that mediates various protein&#x02013;protein interactions (<xref ref-type="bibr" rid="b4">4</xref>). HDAC 1 and HDAC 2, class I histone deacetylases, the histone binding proteins RbAp46 and RbAp48, SAP18, SAP30, SDS3, SAP180 and SAP130 are recognized components of the &#x02018;core&#x02019; Sin3A-HDAC corepressor complex (<xref ref-type="bibr" rid="b5">5</xref>&#x02013;<xref ref-type="bibr" rid="b8">8</xref>). Of these, SAP30 (Sin3A-Associated Protein 30) is a specific component of the Sin3A-complex since it is lacking in other HDAC 1/2-containing complexes such as the NuRD complex (<xref ref-type="bibr" rid="b9">9</xref>). SAP30 is not required for intrinsic repression activity of the Sin3A complex but is involved in Sin3A-mediated NCoR-repression by facilitating and stabilizing the interaction between these two corepressor proteins (<xref ref-type="bibr" rid="b10">10</xref>). In fact, many studies suggest that SAP30 functions as a bridging and stabilizing molecule between the Sin3A complex and other corepressors such as CIR (<xref ref-type="bibr" rid="b11">11</xref>) or DNA-binding transcription factors like YY1 (<xref ref-type="bibr" rid="b12">12</xref>). In yeast, the DNA-binding repressor <italic>UME6</italic> targets the <italic>SIN3&#x02013;RPD3</italic> complex (Sin3A-HDAC 1 homolog in <italic>Saccharomyces cerevisiae</italic>) to its target sequence in the promoter and causes highly localized histone deacetylation, occurring over a range of only one to two nucleosomes (<xref ref-type="bibr" rid="b13">13</xref>).</p><p>In contrast to yeast, which has only one SAP30 homolog, mammals have two proteins, SAP30 and SAP30L (L for like), which share 70% sequence identity. They are both widely expressed in human tissues, with the most prominent expression being in tissues of hematopoietic origin (<xref ref-type="bibr" rid="b14">14</xref>). In this article, we have begun to characterize the function of the mammalian SAP30L protein (<xref ref-type="bibr" rid="b15">15</xref>). We report that SAP30L is able to self-associate and interact with Sin3A. Like SAP30, it has transcriptional repression capability and is able to associate with several class I histone deacetylases. Furthermore, we have identified a novel and functional nucleolar localization signal (NoLS) in SAP30L, and show that Sin3A is targeted to the nucleolus by SAP30L and SAP30 (herein referred to as SAP proteins). Our results show that SAP30L is able to associate with several partners of Sin3A-HDAC complex and suggest that this complex may play a role in the nucleolus.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><sec><title>Cloning and constructs</title><p>Human SAP30 cDNA was obtained from IMAGE clone 4074154. Human SAP30L cDNA has been described previously (<xref ref-type="bibr" rid="b15">15</xref>). Full length and deletion mutants of these proteins were cloned into pcDNA3.1&#x02013;myc-his vector (Invitrogen) for mammalian transfection experiments, and into pGEX-4T1 vector (Amersham Biosciences) for production of GST-fusion proteins in bacteria. Point mutations were created using the QuikChange^&#x000ae; Site Directed Mutagenesis Kit (Stratagene) according to manufacturer's instructions. GAL4-DBD fusion proteins were created in the GAL4-DBD-vector (Stratagene). Luciferase reporter vectors under the control of TK(6) and 14D(10) promoters harboring 4 or 5xGal4 sites were generously provided by D. Reinberg (NJ, USA) and D. Ayer (Salt Lake City, USA), respectively. pCS2+MT-mSin3A plasmid (a generous gift from D. Ayer, Salt Lake City, USA) was used as a PCR-template for mSin3A constructs created and cloned in pcDNA3.1-Myc-His-vector. For <italic>in vitro</italic> transcription and translation experiments, a stop-codon was introduced into the mSin3A constructs to remove myc-his-tag. Flag-epitope tagged HDAC1, HDAC2, RbAp46, RbAp48 and YY1 (<xref ref-type="bibr" rid="b12">12</xref>) cDNAs were obtained from W. Yu (Taipei, Taiwan). HDAC3 cDNA was obtained from U. Mahlknecht (Heidelberg, Germany) and used as a template for PCR-cloning into pcDNA3.1 vector with myc-his-tag. The precise coordinates of the constructs will be supplied on request. The authenticity of the constructs was confirmed by sequencing.</p></sec><sec><title>Cell culture, transfections</title><p>Human embryonal kidney epithelial cells (HEK293T) were cultured in DMEM (Gibco) supplemented with penicillin&#x02013;streptomycin antibiotics, 5% fetal bovine serum, 1 mM sodium pyruvate and 50 &#x000b5;g/ml of uridine. For mammalian transfection experiments, &#x0223c;2 &#x000d7; 10⁴ cells were seeded into 1 cm² surface area of tissue culture dishes. DNA was transfected with FuGENE 6 reagent (Roche) according to the manufacturer's protocol for 18&#x02013;30 h. Thereafter, cells were either lysed in Laemmli solution/lysis buffer (see below) or fixed with 4% paraformaldehyde for the immunostaining experiments.</p></sec><sec><title>GST pull-downs</title><p>GST-SAP30 and GST-SAP30L fusion proteins were produced in <italic>Escherichia coli</italic> (BL-21 strain) and purified with Glutathione Sepharose 4B beads (Amersham Biosciences) according to manufacturer's instructions. The gel profile of the GST-fusion proteins is shown in Supplementary Figure 3. <italic>In vitro</italic> transcription and translation was carried out with TnT^&#x000ae; Quick Coupled Transcription/Translation System (Promega) according to the manufacturer's protocols. For GST pull-downs, &#x0223c;1 &#x000b5;g of GST or GST fusion proteins coupled to beads were incubated with 3&#x02013;36 &#x000b5;l ³⁵S-labeled <italic>in vitro</italic> translated proteins in binding buffer [1&#x000d7; phosphate-buffered saline (PBS) (137 mM NaCl), 0.1% Igepal-CA630 and freshly added protease inhibitors (Roche)] in end over end rotation overnight at 4&#x000b0;C. The beads were washed six times with the binding buffer containing 200 mM NaCl. GST pull-downs from the HEK293T nuclear lysates were done in a similar manner. Nuclei from the HEK293T cells were isolated as described previously (<xref ref-type="bibr" rid="b10">10</xref>).</p></sec><sec><title>Immunoprecipitation</title><p>For the immunoprecipitation experiments, HEK293T whole cell lysates were prepared by lysing cells in RIPA lysis buffer [1&#x000d7; PBS (137 mM NaCl), 1% Igepal-CA630, 0.5% sodium deoxycholate and 0.1% SDS] with freshly added protease inhibitors (Roche). Lysates were passed several times through a 21-gauge needle to sheer DNA, incubated for 30 min on ice and centrifuged in 12&#x02009;000 <italic>g</italic> for 20 min at 4&#x000b0;C. Supernatants were collected. Immunoprecipitations were carried out in end over end rotation overnight at 4&#x000b0;C with agarose-conjugated antibodies against c-myc (9E10; sc-40AC) or His (H-3; sc-8036AC) (Santa Cruz Biotechnology, Inc.), and they were washed six to eight times with RIPA lysis buffer containing 500 mM NaCl and 0.5% Igepal-CA630.</p></sec><sec><title>Western blotting</title><p>For SDS&#x02013;PAGE, lysed cells or protein samples were boiled in Laemmli buffer and resolved on Novex^&#x000ae; pre-cast gels (Invitrogen). Proteins were transferred to a nitrocellulose membrane (Amersham Biosciences) and blotted with the primary antibodies indicated and HRP-conjugated secondary antibodies. Detection was performed with the ECL Plus Western Blotting Detection System (Amersham Biosciences). The primary antibodies used were c-myc (sc-40), Sin3A (sc-767), HDAC 1 (sc-7872), HDAC 2 (sc-7899) and actin (sc-8432) from Santa Cruz, and GFP (33&#x02013;2600) from Zymed. Anti-rabbit and anti-mouse HRP-conjugated secondary antibodies were from DAKO (p0217 and p0260, respectively). Band intensities were quantified using ImageQuant&#x02122;TL &#x02013;program (Amersham Biosciences)</p></sec><sec><title>Immunocytochemistry</title><p>HEK293T cells were fixed with 4% paraformaldehyde in PBS [1&#x000d7; PBS (137 mM NaCl)] for 20 min and then washed with PBS and permeabilized for 10 min with 0.2% Triton X-100 in PBS. Unspecific binding of the antibodies was blocked by 1% BSA in PBS for 60 min before incubation of the cells with primary antibody at 1:200 dilutions for 60 min at 37&#x000b0;C. After washes with PBS, the cells were incubated with secondary antibody at 1:1000 [Alexa^&#x000ae; Fluorophor conjugated anti-mouse (A11031) or anti-rabbit (A11034) IgG], washed and mounted on a DAPI-mount (VectaShield^&#x000ae;). The primary antibodies used were NPM (32-5200, Zymed), Flag (F-3165, Sigma), His (46-0693, Invitrogen), c-myc (sc-40; Santa Cruz), c-myc (sc-789; Santa Cruz) and Sin3A (sc-994; Santa Cruz). Slides were analyzed and photographed with a confocal microscope (Ultraview Confocal Imaging System, Perkin Elmer Life Sciences Inc., Boston, MA).</p></sec><sec><title>HDAC activity and repression analysis</title><p>Histone deacetylase activity was measured using Fluor de Lys AK-500-kit (Biomol) according to manufacturer's protocols. In order to explore the role of class III HDAC enzymes, NAD⁺ coenzyme (N1511, Sigma-Aldrich) was added at 200 &#x000b5;M to reactions. HDAC inhibitor Trichostatin A (TSA) was added in control reactions at 1 &#x000b5;M concentration. Fluorescence was measured at 460 nm with VICTOR² 1420 multilabel counter (Wallac, Perkin Elmer, Life Sciences).</p><p>For the repression analysis, HEK293T cells were transfected with Gal4DBD-SAP30, Gal4DBD-SAP30L or Gal4DBD-SAP30L mutants along with 5xGal4-TK-LUC or 5xGal4-14D-LUC luciferase reporter plasmids as indicated. The results were normalized by the activity of the &#x003b2;-gal produced (<xref ref-type="bibr" rid="b16">16</xref>) from the cotransfected pcDNA3.1-LacZ (Invitrogen). Twenty-fourhour post-transfection cells were split into two dishes, and treated with either TSA at 200 nM or DMSO for 24 h. Thereafter, cells were harvested and luciferase activity was measured using Luciferase Assay System (Promega). Measurements were done in duplicate from two independent experiments and the range is reported.</p></sec></sec><sec><title>RESULTS</title><sec><title>SAP30L interacts with Sin3A</title><p>SAP30 is a well-characterized binding partner for Sin3A (<xref ref-type="bibr" rid="b5">5</xref>,<xref ref-type="bibr" rid="b10">10</xref>). Owing to its similarity to SAP30L, we examined if SAP30L interacts with Sin3A. As shown in <xref ref-type="fig" rid="fig1">Figure 1a</xref>, myc-tagged mouse Sin3A co-immunoprecipitated with myc-his-tagged SAP30L in transiently transfected HEK293T cells. In this experiment, we used myc-his-tagged SAP30 and an empty myc-his vector as positive and negative controls, respectively. Consistent with previous studies, Sin3A co-immunoprecipitated with SAP30, whereas the control experiment with the vector alone remained negative, confirming the specificity of the interactions. In a reciprocal co-immunoprecipitation experiment, green fluorescent protein (GFP)-tagged SAP30L co-immunoprecipitated with the myc-tagged Sin3A while GFP alone was unable to co-immunoprecipitate with Sin3A (<xref ref-type="fig" rid="fig1">Figure 1b</xref>). These results confirm that the interactions are independent of the tag used. In GST pull-down experiments with nuclear lysates of HEK293T cells, SAP30L associated with endogenous human Sin3A similar to SAP30 (<xref ref-type="fig" rid="fig1">Figure 1c</xref>).</p><p>Next, we mapped the domains responsible for the interaction between SAP30L and Sin3A. We used C-terminally truncated versions of SAP30L (constructs are shown in <xref ref-type="fig" rid="fig2">Figure 2b</xref>) and cotransfected them with mouse Sin3A. SAP30L 1&#x02013;140 truncation mutant co-immunoprecipitated Sin3A whereas SAP30L 1&#x02013;120 construct failed to associate with Sin3A, suggesting that the region between residues 120 and 140 of SAP30L is critical for the interaction (<xref ref-type="fig" rid="fig1">Figure 1d</xref>). This finding is similar to SAP30 interaction with Sin3A, where the interaction domain resides in the C-terminus between residues 130 and 167 of SAP30 (<xref ref-type="bibr" rid="b10">10</xref>), a region sharing eight analogous residues with the SAP30L 1&#x02013;140 truncation mutant (see sequence alignment in Figure 5d). We next created three deletion mutants of Sin3A (amino acids 1&#x02013;200, 1&#x02013;400 and 1&#x02013;855) in order to map SAP30L interaction domain(s) in Sin3A. Pull-down studies with <italic>in vitro</italic> transcribed and translated Sin3A proteins revealed that the interaction with the GST-SAP30L requires PAH3/HID region of Sin3A protein (<xref ref-type="fig" rid="fig1">Figure 1e</xref>). This again resembles the interaction of SAP30 with Sin3A, which has been reported to require the PAH3 region of Sin3A (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b10">10</xref>). These results also suggest that the interaction between SAP30L and Sin3A is direct.</p></sec><sec><title>SAP30L is able to self-associate</title><p>We next investigated whether SAP proteins are able to interact with each other. As shown in <xref ref-type="fig" rid="fig2">Figure 2a</xref>, myc-his-tagged SAP30L associates with gfp-tagged SAP30L <italic>in vivo</italic> but not with gfp-tagged SAP30. On the other hand, myc-his-tagged SAP30 could not associate with gfp-tagged SAP30 or SAP30L. These results demonstrate that SAP30L is able to self-associate.</p><p>To characterize more closely the domain(s) in SAP30L needed for self-association, we carried out pull-down experiments with <italic>in vitro</italic> translated SAP30L proteins. As <xref ref-type="fig" rid="fig2">Figure 2b</xref> shows, SAP30L 1&#x02013;120 was unable to associate with full-length SAP30L. Two other SAP30L C-terminal deletion mutants (1&#x02013;140 and 1&#x02013;160) showed markedly impaired self-association capability compared to full-length SAP30L, suggesting that the entire C-terminus of SAP30L is needed for efficient binding. Consistent with this, we found that mutating the nucleolar localization signal, which is composed of amino acids 120&#x02013;127 (SAP30L 8A-mutant, see Figure 5), had no effect on self-association. Intriguingly, truncating 60 residues from the N-terminus of SAP30L increased interaction over 2-fold (<xref ref-type="fig" rid="fig2">Figure 2b</xref>). These results suggest that the ability of SAP30L to self-associate is dependent on an intact C-terminus and that deletion of the N-terminus increases this capability, possibly through conformational changes in the protein. Furthermore, nucleolar targeting of SAP30L (see Figure 5) is independent of its self-association, since the mutant incapable of nucleolar localization can self-associate with an affinity comparable with that of wild-type SAP30L.</p></sec><sec><title>SAP30L associates with histone deacetylases and represses transcription</title><p>SAP30 associates with HDAC activity and HDAC 1 and HDAC 2 proteins (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b10">10</xref>,<xref ref-type="bibr" rid="b12">12</xref>). Therefore, we asked if SAP30L also associates with HDACs. First, we carried out pull-down experiments with GST, GST-SAP30L and GST-SAP30 proteins from HEK293T nuclear lysates and measured associating HDAC activity. GST-SAP30L pulled down HDAC activity comparable with GST-SAP30 (<xref ref-type="fig" rid="fig3">Figure 3a</xref>), and this activity was sensitive to TSA, an inhibitor of HDACs. Addition of NAD⁺, which is an essential cofactor for the activity of class III HDACs, did not increase HDAC activity, suggesting that class III HDACs (<xref ref-type="bibr" rid="b17">17</xref>) do not contribute to HDAC activity associated with SAP proteins in this assay. An intact C-terminus of SAP30L was necessary for HDAC activity as shown by a series of mutants of SAP30L (<xref ref-type="fig" rid="fig3">Figure 3b</xref>). Further pull-down experiments demonstrated that GST-SAP30 and GST-SAP30L interacted with class I HDACs 1&#x02013;3 (<xref ref-type="fig" rid="fig3">Figure 3c</xref>), whereas they failed to interact with a class II histone deacetylase, HDAC 4 (data not shown).</p><p>The association of SAP30L with a functional Sin3A complex was examined using Gal4DBD fusions with a series of SAP30L constructs and SAP30. SAP30 has previously been reported to be able to repress transcription (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b10">10</xref>). Wild-type Gal4SAP30L repressed transcription of a reporter containing 14D promoter and Gal4 binding sites (5xGal14D-LUC) dramatically compared to Gal4 alone (&#x0223c;23-fold) and 1.6-fold compared to Gal4SAP30 (<xref ref-type="fig" rid="fig4">Figure 4a</xref>). Again, an intact C-terminus of SAP30L was needed for full repression capability although SAP30L 1&#x02013;140 possessed moderate repression activity (<xref ref-type="fig" rid="fig4">Figure 4b</xref>). However, cotransfection of Gal4SAP30L or Gal4SAP30 with either myc-tagged SAP30L, SAP30 or Sin3A did not yield any additive repression effect (<xref ref-type="fig" rid="fig4">Figure 4c</xref>), suggesting that the amount of these binding partners was not rate-limiting. TSA treatment greatly diminished the repression activity of SAP proteins (except for the SAP30L 1&#x02013;140 truncation), suggesting that HDAC activity plays an important role in mediating the repression capability (<xref ref-type="fig" rid="fig4">Figure 4a and b</xref>). Another reporter vector with TK promoter produced identical results (Supplementary Figure 1). Taken together, these findings suggest that SAP30L represses transcription, and that this repression involves the recruitment of Sin3A and histone deacetylases.</p></sec><sec><title>SAP30L has a functional NoLS and localizes to the nucleolus</title><p>By transfection experiments, SAP30L was previously shown to localize to the nucleus of cells and a functional nuclear localization signal (NLS) was identified (<xref ref-type="bibr" rid="b15">15</xref>). We decided to examine more closely the subcellular localization of SAP30L using tagged wt and mutant SAP30L proteins transiently transfected into a variety of cell lines. GFP- or myc-his-tagged wt SAP30L was found in the nucleus of studied cell lines (MCF-7, COS-7, IMR-90, T84, Daudi, HEK293T), and a prominent, patchy staining pattern resembling that of the nucleolus was observed in the nucleus of many cells. To confirm this, we performed colocalization experiments with a nucleolar marker, nucleophosmin (NPM or B23) (<xref ref-type="bibr" rid="b18">18</xref>). <xref ref-type="fig" rid="fig5">Figure 5a</xref> demonstrates that there is a marked colocalization between the two proteins in HEK293 cells and that SAP30L partly localizes to the nucleolus. Many nuclear and nucleolar proteins like HSP70, EBNA-5 (<xref ref-type="bibr" rid="b19">19</xref>), p53 and MDM2 (<xref ref-type="bibr" rid="b20">20</xref>) are known to accumulate in the nucleolus under proteotoxic stress caused by proteasome inhibitor MG132. Therefore, we tested whether MG132 affects the subnuclear localization of SAP30L and found that MG132 caused further accumulation of SAP30L in the nucleolus (<xref ref-type="fig" rid="fig5">Figure 5a</xref>). We decided to take advantage of this effect in our later experiments mapping the NoLS in SAP30L (see below). GFP-tagged SAP30L showed similar strong nucleolar accumulation under MG132 treatment (Supplementary Figure 2) whereas GFP alone did not relocalize (data not shown) indicating that tags do not contribute to the results. SAP30 behaved similarly, although more slowly (Supplementary Figure 2): SAP30L accumulated into the nucleolus within 4 h whereas for SAP30 the accumulation took 6 h (data not shown).</p><p>Confocal microscopy with the C-terminally truncated versions of SAP30L protein (SAP30L 1&#x02013;160, SAP30L 1&#x02013;140 and SAP30L 1&#x02013;120) revealed that the largest C-terminal deletion mutant (1&#x02013;120) caused significant mislocalization of the protein to the cytoplasm and complete disappearance of the nucleolar localization (<xref ref-type="fig" rid="fig5">Figure 5b</xref>). This suggested the presence of a NoLS in the region between residues 120 and 140 of SAP30L. Closer examination of the sequence of SAP30L showed that this region harbors a stretch of basic residues consistent with a proposed NoLS consensus sequence (R/K-R/K-x-R/K) [<xref ref-type="fig" rid="fig5">Figure 5d</xref> and Ref. (<xref ref-type="bibr" rid="b21">21</xref>)]. In order to assess the role of this region in the nucleolar targeting of SAP30L, we constructed SAP30L 1&#x02013;121, SAP30L 1&#x02013;127 and SAP30L 1&#x02013;131 mutants (<xref ref-type="fig" rid="fig5">Figure 5c</xref>). In contrast to 1&#x02013;121 truncation, SAP30L 1&#x02013;127 deletion mutant accumulated in the nucleolus under MG132 treatment, showing that the critical region responsible for nucleolar localization resides between the residues 120&#x02013;127 of SAP30L (<xref ref-type="fig" rid="fig5">Figure 5c</xref>). Next, we created two mutants with either three or four basic residues mutated to alanines in this region of SAP30L (₁₂₀RRYKR₁₂₄ &#x02192; <bold>AA</bold>Y<bold>A</bold>R or <bold>AA</bold>Y<bold>AA</bold>). These mutations reduced nucleolar accumulation of SAP30L, but failed to abolish it under MG132 treatment (data not shown). However, a larger mutation in the region (SAP30L 8A-mutant: ₁₂₀RRYKRHYK₁₂₇ &#x02192; <bold>AAAAAAAA</bold>) completely abolished nucleolar localization of SAP30L, demonstrating that these eight residues are responsible for correct localization of SAP30L to the nucleolus (<xref ref-type="fig" rid="fig5">Figure 5c and d</xref>).</p><p>Previously, GFP-tagged SAP30L was reported to contain a functional NLS between residues 87 and 91 (<xref ref-type="bibr" rid="b15">15</xref>). We recreated this NLS mutant in a wt SAP30L construct with myc-his-tag (₈₇KRKRK₉₁ &#x02192; K<bold>AAA</bold>K). This mutant partly localized to the cytoplasm but still showed some nucleolar localization (<xref ref-type="fig" rid="fig5">Figure 5b</xref>), suggesting that NLS signal in SAP30L is functional only in nuclear targeting of the protein. When both signals were mutated simultaneously, nuclear localization of SAP30L was significantly impaired (data not shown), indicating that the NoLS signal also contributes to the nuclear localization of SAP30L. We found an additional signal in the N-terminus of SAP30L (₅₈KKLK₆₁), which is also consistent with the NoLS proposed by Horke <italic>et al.</italic> (<xref ref-type="bibr" rid="b21">21</xref>). However, N-terminally deleted SAP30L protein (SAP30L 61&#x02013;183) showed strong nucleolar localization of SAP30L, indicating that it is not a functional nucleolar localization signal (<xref ref-type="fig" rid="fig5">Figure 5b</xref>).</p><p>Next, we investigated whether MG132 treatment causes relocalization to the nucleolus of other members of the Sin3A corepressor complex. A fraction of the endogenous Sin3A pool responded to MG132 treatment in a manner similar to the SAP proteins by accumulating in the nucleolus (Supplementary Figure 2). Consistent with other studies, HDAC1 and HDAC2 enzymes were detectable in the nucleolus (<xref ref-type="bibr" rid="b22">22</xref>) although there was no marked relocalization after MG132 treatment. Importantly, MG132 did not alter the subcellular localization of RbAp46, RbAp48 and YY1 proteins (Supplementary Figure 2). These results suggest that separate Sin3A complexes are present in cells, and that SAP proteins together with Sin3A and HDAC 1/2 enzymes belong to one such subcomplex, possibly within the nucleolus. This is consistent with a study reporting at least three separate Sin3A complexes with unique protein compositions (<xref ref-type="bibr" rid="b23">23</xref>).</p></sec><sec><title>SAP proteins target SIN3A to the nucleolus</title><p>In quiescent cells, Sin3A is known to localize in the perinucleolar sites where early DNA replication origins are situated (<xref ref-type="bibr" rid="b24">24</xref>). Since Sin3A does not have any apparent NoLS sequence, we asked whether SAP30 and SAP30L proteins are able to target Sin3A to the nucleolus. To examine this, we cotransfected Sin3A with either SAP30L, SAP30L 1&#x02013;120 or SAP30 in HEK293T cells. In confocal microscopy, wt SAP30L and SAP30 proteins dramatically increased the number of Sin3A-positive nucleoli (<xref ref-type="fig" rid="fig6">Figure 6a and b</xref>). Importantly, C-terminally deleted SAP30L (1&#x02013;120), which does not associate with Sin3A, failed to target Sin3A to the nucleoli. The results were similar in the presence or absence of MG132, although in MG132-treated cells, there was a low but constant level of Sin3A visible in the nucleolus (see Supplementary Figure 2). We also scored Sin3A-positive nucleoli, and found that 42 and 7% of the SAP30L- and SAP30-transfected cells, respectively, were positive whereas none of the control vector-transfected cells were (<xref ref-type="fig" rid="fig6">Figure 6b</xref>). The transfected SAP30L and SAP30 proteins were also able to relocate endogenous Sin3A to the nucleolus (data not shown). These results indicate that Sin3A can be targeted to the nucleolus by SAP proteins. It is noteworthy that SAP30L is more efficient than SAP30 in nucleolar targeting, consistent with its more prominent localization within the nucleolus.</p></sec><sec><title>The turnover of SAP30L is regulated by its C-terminus</title><p>Transfected C-terminally deleted SAP30L (1&#x02013;120) protein was expressed over 35 times more efficiently than the wt SAP30L protein (<xref ref-type="fig" rid="fig7">Figure 7</xref>). The expression levels of two other C-terminally deleted SAP30L proteins (1&#x02013;140 and 1&#x02013;160) declined progressively: SAP30L 1&#x02013;140 was expressed &#x0223c;20 and SAP30L 1&#x02013;160 &#x0223c;16 times more efficiently than the wt SAP30L protein. Furthermore, N-terminally deleted SAP30L (61&#x02013;183), which localizes intensively within the nucleolus (<xref ref-type="fig" rid="fig5">Figure 5b</xref>), was expressed at very low levels, i.e. five times less than the wt SAP30L protein. In other words, nucleolar localization correlated inversely with protein expression levels. In these experiments, transfection efficiencies were normalized to cotransfected lacZ protein and endogenous actin was used as a loading control. HEK293T cells treated for 10 h with MG132 stabilized ectopically expressed wt SAP30L by 10-fold compared to control cells (DMSO-treated cells). Stabilization caused by MG132 was dependent of the residues between 120 and 140 of SAP30L since mutant lacking these residues (compare SAP30L 1&#x02013;120 with SAP30L 1&#x02013;140 and 1&#x02013;160 mutants) did not show any stabilization after MG132 treatment (<xref ref-type="fig" rid="fig7">Figure 7</xref>). Since MG132 inhibits proteasomes, and ubiquitination of proteins often marks them for degradation (<xref ref-type="bibr" rid="b25">25</xref>), we reasoned that SAP30L could be ubiquitinated. However, we failed to detect any ubiquitination of SAP30L even after MG132 treatment (data not shown). This may imply that the stabilization of SAP30L is secondary to the inhibition of degradation of other protein(s). Treating cells with cycloheximide, which ceases protein translation, further demonstrated that mutants lacking the C-terminus of SAP30L, and particularly the mutant lacking residues 120&#x02013;140, have extended turnover compared to other SAP30L proteins (<xref ref-type="fig" rid="fig7">Figure 7</xref>). Others have previously reported that the turnover of endogenous SAP30 protein is normally short, only 2 h in HeLa cells (<xref ref-type="bibr" rid="b10">10</xref>).</p></sec></sec><sec><title>DISCUSSION</title><p>We report here a novel component of the Sin3A corepressor complex, SAP30L. It binds to the PAH3/HID region of mouse Sin3A with residues 120&#x02013;140, a region harboring several residues that are also conserved in SAP30. We provide evidence that SAP30L induces transcriptional repression, possibly via the recruitment of Sin3A and histone deacetylases. We have also identified a region in SAP30L with a stretch of basic residues representing a functional NoLS signal (<xref ref-type="bibr" rid="b21">21</xref>). Moreover, both SAP proteins are capable of targeting Sin3A to the nucleolus.</p><p>SAP30L and SAP30 are both transcribed from independent genomic loci (5q33.2 for SAP30L and 4q34.1 for SAP30). These two chromosomes are known to share chromosome-duplication blocks (<xref ref-type="bibr" rid="b26">26</xref>) and, in fact, macroscale analysis of gene composition of distal arms of 4q and 5q chromosomes suggests that a gene duplication event may have occurred during evolution. Closer inspection of the genomic sequences and phylogenetic footprinting analysis (Consite website, data not shown) of SAP genes reveal that they have different sets of conserved transcription factor binding sites on the promoters. Thus, different promoter sequences could allow specific transcription factors to regulate their expression in a timely and tissue-specific manner. Accordingly, previous studies (<xref ref-type="bibr" rid="b10">10</xref>,<xref ref-type="bibr" rid="b15">15</xref>) and gene expression databases (<xref ref-type="bibr" rid="b14">14</xref>) show that they are both ubiquitously expressed but with differences in expression pattern in, for example, testis, placenta and kidney. In our luciferase reporter analysis, SAP30L had &#x0223c;1.6-fold higher repression capacity than SAP30. If also true <italic>in vivo</italic>, use of a specific SAP protein could be a way to fine-scale the repression efficiency of the Sin3A corepressor complex. Various SAP proteins could also be used in specific Sin3A-subcomplexes or recruited in response to varying demands of repression activity during the progression of cell cycle or in specific cell types.</p><p>Our results show that both SAP proteins localize partly within the nucleolus. The nucleolus is the most prominent specialized organelle inside the nucleus. Its principal function is the transcription and processing of rRNA and the assembly of ribosomes, although other functions, such as ribonucleoprotein (RNP) assembly, cell cycle control, mRNA maturation, stress response and protein sequestration, have recently been attributed to the nucleolus (<xref ref-type="bibr" rid="b27">27</xref>,<xref ref-type="bibr" rid="b28">28</xref>). In <italic>S.cerevisiae</italic>, <italic>SIN3A</italic>, <italic>SAP30</italic> and <italic>RPD3</italic> have been shown to affect the transcription of the mating-type, telomeric and rDNA loci. Interestingly, deletion of any of these genes enhances silencing of RNA polymerase II-transcribed reporter genes inserted into the above-mentioned three loci (<xref ref-type="bibr" rid="b29">29</xref>). Similarly, a genetic screen for genes involved in rDNA silencing in <italic>S.cerevisiae</italic> identified mutations in <italic>SIN3A</italic>, <italic>SAP30</italic> and <italic>RPD3</italic> genes (<xref ref-type="bibr" rid="b30">30</xref>). An alternate function for the Sin3A complex in yeast was suggested by Meskauskas <italic>et al.</italic> (<xref ref-type="bibr" rid="b31">31</xref>), who showed that the main components of this complex participate not in the transcription, but in the early processing of rRNA. In the light of these reports, it is not surprising that we found also mammalian SAP proteins in the nucleolus. Furthermore, we were able to identify a NoLS signal in both SAP30L and SAP30. NoLS is generally regarded as more of a protein&#x02013;protein interaction domain than as a specific localization or targeting signal (<xref ref-type="bibr" rid="b32">32</xref>). It is thought that NoLS mediates interactions and thereby retains proteins in the nucleolus (<xref ref-type="bibr" rid="b33">33</xref>). In addition, our results show that Sin3A can be targeted to the nucleolus by SAP proteins. Therefore, it can be postulated that the SAP proteins interact with component(s) of the nucleolus and, by this means, recruit the Sin3A corepressor complex into the nucleolus for the regulation of rDNA transcription and/or rRNA processing. In future experiments, it will be essential to examine further the functional consequences of this nucleolar localization and recruitment.</p></sec><sec><title>SUPPLEMENTARY DATA</title><p>Supplementary Data are available at NAR Online.</p></sec>

Mutagenic nucleotide incorporation and hindered translocation by a food carcinogen C8-dG adduct in <italic>Sulfolobus solfataricus</italic> P2 DNA polymerase IV (Dpo4): modeling and dynamics studies

<p>Bulky carcinogen-DNA adducts commonly cause replicative polymerases to stall, leading to a switch to bypass polymerases. We have investigated nucleotide incorporation opposite the major adduct of 2-amino-1-methyl-6-phenylimidazo[4,5-<italic>b</italic>]pyridine (PhIP) in the DinB family polymerase, Dpo4, using molecular modeling and molecular dynamics (MD) simulations. PhIP, the most prevalent heterocyclic aromatic amine formed by cooking of proteinaceous food, is mutagenic in mammalian cells and is implicated in mammary and colon tumors. Our results show that the dG-C8-PhIP adduct can be accommodated in the spacious major groove Dpo4 open pocket, with Dpo4 capable of incorporating dCTP, dTTP or dATP opposite the adduct reasonably well. However, the PhIP ring system on the minor groove side would seriously disturb the active site, regardless of the presence and identity of dNTP. Furthermore, the simulations indicate that dATP and dTTP are better incorporated in the damaged system than in their respective mismatched but unmodified controls, suggesting that the PhIP adduct enhances incorporation of these mismatches. Finally, bulky C8-dG adducts, situated in the major groove, are likely to impede translocation in this polymerase (Rechkoblit <italic>et al.</italic> (2006), <italic>PLoS Biol</italic>., <bold>4</bold>, e11). However, <italic>N</italic>²-dG adducts, which can reside on the minor groove side, appear to cause less hindrance when in this position.</p>

<contrib contrib-type="author"><name><surname>Zhang</surname><given-names>Ling</given-names></name></contrib><contrib contrib-type="author"><name><surname>Rechkoblit</surname><given-names>Olga</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Wang</surname><given-names>Lihua</given-names></name><xref rid="au2" ref-type="aff">2</xref></contrib><contrib contrib-type="author"><name><surname>Patel</surname><given-names>Dinshaw J.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Shapiro</surname><given-names>Robert</given-names></name></contrib><contrib contrib-type="author"><name><surname>Broyde</surname><given-names>Suse</given-names></name><xref rid="au2" ref-type="aff">2</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><aff><institution>Department of Chemistry, New York University</institution><addr-line>New York, NY, USA</addr-line></aff><aff id="au1">¹<institution>Structural Biology Program, Memorial Sloan-Kettering Cancer Center</institution><addr-line>New York, NY, USA</addr-line></aff><aff id="au2">²<institution>Department of Biology, New York University</institution><addr-line>New York, NY, USA</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p>DNA polymerases of the Y-family have in recent years been shown to play a predominant role in synthesis past DNA bulky lesions, such as those derived from polycyclic aromatic chemicals present in tobacco smoke, automobile exhaust, and broiled meat and fish (<xref ref-type="bibr" rid="b1">1</xref>&#x02013;<xref ref-type="bibr" rid="b5">5</xref>). High fidelity replicative DNA polymerases are usually impeded by such damage, leading to a switch to one or more low fidelity bypass polymerases for translesion synthesis (<xref ref-type="bibr" rid="b6">6</xref>&#x02013;<xref ref-type="bibr" rid="b11">11</xref>).</p><p>The model Y-family polymerase, DNA polymerase IV (Dpo4), is from the archaeon bacterium <italic>Sulfolobus solfataricus</italic> P2. It is a member of the DinB family (<xref ref-type="bibr" rid="b12">12</xref>), of which human pol &#x003ba; is also a member (<xref ref-type="bibr" rid="b13">13</xref>). It has been extensively investigated by X-ray crystallography in binary complexes with primer/template DNA and in ternary complexes in the presence of dNTP, both with and without DNA damage (<xref ref-type="bibr" rid="b14">14</xref>&#x02013;<xref ref-type="bibr" rid="b23">23</xref>). These structures have revealed a spacious, water accessible active site that is capable of accommodating two templating bases, in contrast to high fidelity replicative polymerases whose ternary complexes show tight fit of the nascent base pair with exclusion of solvent (<xref ref-type="bibr" rid="b24">24</xref>). Dpo4 has an open pocket on the major groove side of the template, as well as a smaller open space on the minor groove side; this is in contrast to replicative polymerases, which only have an open pocket on the major groove side, while the minor groove side is packed with protein&#x02013;DNA interactions critical for polymerase fidelity (<xref ref-type="bibr" rid="b25">25</xref>&#x02013;<xref ref-type="bibr" rid="b29">29</xref>). In addition like other Y-family polymerases, Dpo4 has a unique little finger domain, also called wrist or polymerase associated domain (PAD) at the C-terminus (<xref ref-type="bibr" rid="b15">15</xref>,<xref ref-type="bibr" rid="b30">30</xref>,<xref ref-type="bibr" rid="b31">31</xref>). The flexibility of this little finger domain is believed to play an important role in accommodating specific types of DNA lesions (<xref ref-type="bibr" rid="b14">14</xref>,<xref ref-type="bibr" rid="b15">15</xref>,<xref ref-type="bibr" rid="b21">21</xref>). Crystal structures of Dpo4 binary and ternary complexes also reveal that the little finger domain plays a key role in translocation (<xref ref-type="bibr" rid="b23">23</xref>).</p><p>Recently, it has been suggested that the DinB family polymerases may be specifically suited for bypass of <italic>N</italic>²-dG adducts (<xref ref-type="bibr" rid="b32">32</xref>). The presence of a conserved &#x02018;steric gate&#x02019;, usually phenylalanine or tyrosine, is shown to be crucial in bypass of <italic>N</italic>²-dG minor groove adducts (<xref ref-type="bibr" rid="b32">32</xref>). However, the DinB polymerases may be less well suited for bypassing C8-dG bulky adducts. Such adducts would normally be expected, from their position of substitution on guanine, to reside in the major groove of double-stranded DNA. However, rotation of the modified guanine from <italic>anti</italic> to <italic>syn</italic> would place the adduct in the minor groove area, in a position roughly similar to an <italic>N</italic>²-substituted guanine. Despite this formal possibility, experimental primer extension data for two C8-dG adducts derived from 2-acetylaminofluorene (AAF), namely <italic>N</italic>-(deoxyguanosin-8-yl)-2-acetylaminofluorene (dG-C8-AAF) with Dpo4 (<xref ref-type="bibr" rid="b12">12</xref>) and <italic>N</italic>-(deoxyguanosin-8-yl)-2-aminofluorene (dG-C8-AF) with <italic>Escherichia coli</italic> DinB and human pol &#x003ba;&#x00394;c (33), indicates that these adducts cause polymerase stalling or blockage, with only small amounts of primer extension beyond the lesion site. A recent molecular dynamics (MD) study from our group has provided structural rationale for the case of dG-C8-AAF in Dpo4 (<xref ref-type="bibr" rid="b34">34</xref>). However, this adduct contains an acetyl group, which adds to the steric hindrance in C8 adducts.</p><p>Here, we investigate a C8-dG adduct (<xref ref-type="fig" rid="fig1">Figure 1a</xref>) derived from the most prevalent heterocyclic aromatic amine formed by cooking proteinaceous food, 2-amino-1-methyl-6-phenylimidazo[4,5-<italic>b</italic>]pyridine (PhIP) (<xref ref-type="bibr" rid="b35">35</xref>&#x02013;<xref ref-type="bibr" rid="b43">43</xref>). This substance causes predominantly G to T transversions in mammalian cells, with some G to A transitions, and a few deletions (<xref ref-type="bibr" rid="b44">44</xref>&#x02013;<xref ref-type="bibr" rid="b52">52</xref>). In a previous molecular modeling and MD study of this adduct in the Pol &#x003b1; family replicative polymerase RB69, we found considerable active site distortion caused by this lesion at the templating base or in the first double-stranded extension position, when partnered by C or A (<xref ref-type="bibr" rid="b53">53</xref>). These results suggested that the replicative polymerase would be stalled by the lesion, affording opportunity for switch to one or more bypass polymerases. We now use this dG-C8-PhIP adduct, <italic>N</italic>²-(2&#x02032;-deoxyguanosin-8-yl)-PhIP, as a model to study the structural feasibility of accommodating a C8-dG lesion lacking the acetyl group in Dpo4, employing Dpo4 type I (<xref ref-type="bibr" rid="b16">16</xref>) and type II (<xref ref-type="bibr" rid="b15">15</xref>) crystal structures as initial models (<xref ref-type="fig" rid="fig1">Figure 1</xref>).</p><p>Our results show that this dG-C8 adduct, like dG-C8-AAF, if accommodated on the Dpo4 minor groove side pocket, would cause serious distortions to the active site region. However, the PhIP ring system can be accommodated in the spacious major groove pocket of this polymerase, and Dpo4 appears capable of incorporating dCTP, dTTP or dATP reasonably well. In addition, our results show that the PhIP-modified lesion produces less distorted structures for <italic>anti</italic>-G*<italic>&#x000b7;anti</italic>-dTTP and <italic>anti</italic>-G*<italic>&#x000b7;syn</italic>-dATP than for their respective unmodified controls (<italic>anti</italic>-G<italic>&#x000b7;anti</italic>-dTTP and <italic>anti</italic>-G<italic>&#x000b7;syn</italic>-dATP). These findings may be relevant to observed mutagenic behavior of PhIP in inducing G to T transversions and G to A transitions in mammalian systems (<xref ref-type="bibr" rid="b44">44</xref>&#x02013;<xref ref-type="bibr" rid="b52">52</xref>), if the human DinB polymerase pol &#x003ba; is involved in the error-prone incorporation opposite the lesion, and its structural properties prove to be similar to those of its prokaryotic homolog Dpo4. Finally, based on Dpo4 binary and ternary complex structures (<xref ref-type="bibr" rid="b23">23</xref>), we suggest that translocation may be seriously inhibited by bulky dG-C8 adducts positioned on the major groove side of the DNA duplex region, while <italic>N</italic>²-dG adducts residing on the minor groove side (<xref ref-type="bibr" rid="b54">54</xref>) would be less hindering.</p></sec><sec><title>COMPUTATIONAL METHODS</title><sec><title>Molecular modeling</title><p><italic>G&#x000b7;dNTP initial models.</italic> The crystal structure of Dpo4 polymerase with 10<italic>R</italic>(+)-<italic>cis-anti</italic>-benzo[<italic>a</italic>]pyrene(BP)-<italic>N</italic>⁶-dA located on the major groove side of the modified DNA (BP-2 complex) (<xref ref-type="bibr" rid="b16">16</xref>) was used to obtain initial structures for the G&#x000b7;dNTP models (PDB ID: 1S0M) (<xref ref-type="bibr" rid="b55">55</xref>). Of the many type I structures of Dpo4, the BP-2 complex has the most reaction-ready active site, since it contains two metal ions in the active site and the 3&#x02032; hydroxyl group at the primer end. This structure was further remodeled, as described in previous work from our group (<xref ref-type="bibr" rid="b34">34</xref>), to achieve ideal Mg²⁺ coordination (Supplementary Table S1) and an O3&#x02032;-P&#x003b1; distance of 3.1 &#x0212b;, in the reaction-ready range.</p><p>The DNA sequence was then adjusted to match the sequence in the adenomatous polyposis coli (<italic>Apc</italic>) gene mutational hotspot for PhIP, codon 635 (<xref ref-type="fig" rid="fig1">Figure 1b</xref>), with the guanine selected for PhIP modification situated at the templating position opposite the incoming dNTP. The initial models for dynamics were then constructed by locating structures with minimal crowding between the PhIP moiety and the polymerase, by rotating &#x003b1;&#x02032; and &#x003b2;&#x02032; at 10&#x000b0; intervals, in combination (<xref ref-type="fig" rid="fig1">Figure 1a</xref>). &#x003b3;&#x02032; was initiated at 26&#x000b0; as in the NMR solution structure (<xref ref-type="bibr" rid="b56">56</xref>). Both <italic>anti</italic> and <italic>syn</italic> conformations of the glycosidic torsion &#x003c7; were investigated, with &#x003c7; adjusted to achieve optimal hydrogen bonding and stacking in the nascent base pair for each &#x003c7; domain. The torsion angles in the initial models are summarized in Supplementary Table S2.</p><p>For the G*&#x000b7;dATP mismatch, three starting models were obtained. Two of them featured the PhIP-modified guanine (G*) <italic>anti</italic> and the PhIP rings on the major groove side of the template, with the incoming nucleotide dATP either <italic>anti</italic> or <italic>syn</italic>. A <italic>syn</italic>-dATP was investigated since it has been observed opposite a lesion in a Dpo4 crystal structure (<xref ref-type="bibr" rid="b17">17</xref>). A high resolution NMR solution structure containing a dG-C8-PhIP in an 11mer duplex shows a base-displaced intercalation model (<xref ref-type="bibr" rid="b56">56</xref>). In this conformation, the modified guanine is <italic>syn</italic> and displaced into the major groove, while the PhIP rings intercalate within the DNA duplex. However, the PhIP rings in such a base-displaced intercalation conformation would occupy the position of the incoming nucleotide in Dpo4. A PhIP-modified <italic>syn</italic>-guanine model could be built with the PhIP rings on the minor groove side of the dNTP, which does allow the accommodation of the dNTP at the active site. With a <italic>syn</italic>-G* as the template, only <italic>anti</italic>-dATP is favorable, since the bases in the <italic>syn</italic>-G*&#x000b7;<italic>syn</italic>-dATP pair are too far apart for hydrogen bonding. Hydrogen bonding in the nascent base pair is an important consideration in building the initial models, since Dpo4 as well as other DinB family polymerases, such as Dbh and pol &#x003ba; rely more on hydrogen bonding for catalytic efficiency than the high fidelity polymerases (<xref ref-type="bibr" rid="b57">57</xref>,<xref ref-type="bibr" rid="b58">58</xref>).</p><p>In the G*&#x000b7;dCTP/dTTP/dGTP series, the G*s were all constructed in the same <italic>anti</italic> conformation as in the G*&#x000b7;dATP mismatch. A <italic>syn</italic>-G* was eliminated based on results for the <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP simulation; these show that a <italic>syn</italic>-G* generated much more disturbance to the polymerase active site than an <italic>anti</italic> one, as described in Results, and these disturbances are essentially independent of the dNTP. In the G*&#x000b7;dCTP model, the incoming dCTP is <italic>anti</italic>, in order to form Watson&#x02013;Crick hydrogen bonds with the <italic>anti</italic>-G*. In the G*&#x000b7;dTTP model, the dTTP is <italic>anti</italic> to achieve a wobble pair with the <italic>anti</italic>-G*. A wobble paired T&#x000b7;dGTP has been observed in a Dpo4 ternary complex (<xref ref-type="bibr" rid="b18">18</xref>). A <italic>syn</italic>-dTTP can form no hydrogen bonds with <italic>anti</italic>-G* and <italic>syn</italic> pyrimidines have scarcely been observed (<xref ref-type="bibr" rid="b59">59</xref>). In the G*&#x000b7;dGTP model, a <italic>syn</italic>-dGTP was employed, since an <italic>anti</italic>-dGTP could not form hydrogen bonds with <italic>anti</italic>-G* in the polymerase, and would collide with either the <italic>anti</italic>-G* or the minor groove side protein residues.</p><p>As an unmodified control for all the simulations, an unmodified <italic>anti</italic>-G opposite a Watson&#x02013;Crick paired <italic>anti</italic>-dCTP was used. Additional controls containing an unmodified <italic>anti</italic>-G opposite <italic>anti</italic>-dATP for the G*&#x000b7;dATP mismatch, <italic>anti</italic>-dTTP for the G*&#x000b7;dTTP mismatch and <italic>syn</italic>-dGTP for the G*&#x000b7;dGTP mismatch were also obtained. Stereoviews of the active sites of the initial models are shown in Supplementary Figure S1.</p><p><italic>&#x02212;1 deletion initial models.</italic> The crystal structure of the Dpo4 type II ternary complex (PDB ID: 1JXL) (<xref ref-type="bibr" rid="b15">15</xref>) was used to obtain the starting models for the &#x02212;1 deletion structures (<xref ref-type="fig" rid="fig1">Figure 1</xref>). In this crystal structure, the coordinates of the first base at the 5&#x02032; end of the template could not be resolved, probably due to its flexible conformation outside the polymerase. We therefore modeled it into the structure, using the conformation of an analogous terminal base in the Dpo4 type I structure (PDB ID: 1JX4). Hydrogen atoms absent in the crystal structures were added by the AMBER suite. The dideoxy group at the 3&#x02032; end of the primer was replaced by a hydroxyl group. The Ca²⁺ ion, residing in the position of the nucleotide binding metal ion, was replaced by a Mg²⁺ ion, and repositioned for proper octahedral coordination with water molecules and amino acid residues (Supplementary Table S1). The DNA sequence was also remodeled to match the sequence in the <italic>Apc</italic> gene mutational hotspot codon 635 (<xref ref-type="fig" rid="fig1">Figure 1c</xref>). The PhIP moiety was linked to the unpartnered guanine in the active site, while the incoming dCTP paired with the guanine on the 5&#x02032; side of the adduct. Starting models were obtained using the same approach as for G*&#x000b7;dNTP models, by rotating torsion angles &#x003b1;&#x02032; and &#x003b2;&#x02032; at 10&#x000b0; intervals, in combination. Structures were selected for <italic>anti</italic> and <italic>syn</italic> domains of &#x003c7; based on minimal steric close contacts and optimal stacking. The torsion angles in these starting models are summarized in Supplementary Table S2 and stereoviews of the active sites of the initial models are shown in Supplementary Figure S1.</p></sec><sec><title>Force field parameterization</title><p>Parameters for the <italic>anti</italic>-dG-C8-PhIP adduct and the incoming <italic>anti</italic>-dCTP, <italic>anti</italic>-dATP, <italic>syn</italic>-dATP and <italic>anti</italic>-dTTP were the same as in earlier work (<xref ref-type="bibr" rid="b53">53</xref>,<xref ref-type="bibr" rid="b54">54</xref>). The parameters for the <italic>syn</italic>-dG-C8-PhIP adduct and the <italic>syn</italic>-dGTP were obtained using the same method described previously (<xref ref-type="bibr" rid="b53">53</xref>). Supplementary Tables S3 and S4 shows the AMBER atom type, connection type and partial charge assignment for these cases.</p></sec><sec><title>MD simulation and data analyses</title><p>Simulations were carried out using the SANDER module of the AMBER 6.0 MD software package (University of California, San Francisco), the Cornell <italic>et al.</italic> force field (<xref ref-type="bibr" rid="b60">60</xref>) and the parm99 parameter set (<xref ref-type="bibr" rid="b61">61</xref>). Electrostatic interactions were approximated by the particle mesh Ewald method (<xref ref-type="bibr" rid="b62">62</xref>), and a 10 &#x0212b; cutoff was applied to Lennard&#x02013;Jones interactions. All bonds involving hydrogen atoms were constrained by the SHAKE algorithm (<xref ref-type="bibr" rid="b63">63</xref>) with a tolerance of 0.0005 &#x0212b;, and a 2 fs time-step was used in the dynamics simulation. Periodic boundary conditions were applied, and all MD simulations were carried out under constant temperature, 300 K, with a temperature coupling parameter of 4.0 ps and at constant atmospheric pressure with a 1.0 ps coupling parameter. The translational motion of the center of mass was removed every 1 ps. No obvious overall rotation of the system was observed during the simulation; thus, energy leakage from internal motion to global rotation through the &#x02018;flying ice cube effect&#x02019; did not happen here (<xref ref-type="bibr" rid="b64">64</xref>).</p><p>Fifteen Na⁺ ions were added to each system for neutralization using the LEaP module of AMBER, followed by 600 steps of steepest descent (SD) and 400 steps of conjugate gradient (CG) to relax the added Na⁺ ions and crystal waters. Then &#x0223c;15&#x02009;000 TIP3P water molecules were added to solvate each system creating a rectangular periodic box containing a total of &#x0223c;50&#x02009;000 atoms. The systems were further minimized and heated up to 300 K. The equilibration protocol was conducted in the same manner as in earlier work (<xref ref-type="bibr" rid="b53">53</xref>) and the details are given in Supplementary Data. Production MD simulation was carried out for 3 ns.</p><p>The PTRAJ, ANAL and CARNAL modules of AMBER were employed to analyze the trajectories. The root-mean-square deviations (RMSDs) of each system were calculated relative to the starting structures as shown in Supplementary Figure S2. The overall structure and the active site region of the protein&#x02013;DNA complexes became reasonably stable after &#x0223c;500 ps production MD simulation in each system. Therefore, the last 2.5 ns of MD simulation were used for structural analyses. Stereoviews of unmodified and modified systems after the total 3 ns of production MD simulations are shown in Supplementary Figure S3, and the active sites of these systems are shown in <xref ref-type="fig" rid="fig2">Figures 2</xref> and <xref ref-type="fig" rid="fig3">3</xref>, and Supplementary Figure S4.</p><p>Molecular modeling was carried out with Insight II 97.0 (Accelrys, Inc., a subsidiary of Pharmacopeia, Inc.). Figures of structures were prepared with PyMOL (DeLano Scientific, LLC.).</p></sec></sec><sec><title>RESULTS</title><sec><title>Initial models</title><p>Our initial models (see Computational Methods) investigated both <italic>anti</italic> and <italic>syn</italic> conformations of the adduct in the active site of Dpo4, partnered with dATP or unpaired in the &#x02212;1 deletion simulations. In addition, <italic>anti</italic>-G* opposite <italic>anti</italic>-dCTP, <italic>anti</italic>-dTTP or <italic>syn</italic>-dGTP were investigated using results from dATP to guide selection of the starting structures. Production MD simulations were carried out for 3 ns and the ensembles of structures derived from the last 2.5 ns trajectory were analyzed.</p></sec><sec><title>Structural analyses</title><p>A number of structural criteria were evaluated in each modified system and compared to the undamaged control simulation. These include: (i) the number and occupancy of the hydrogen bonds in the nascent base pair (Supplementary Figure S5 and Supplementary Tables S5 and S10); (ii) the distance between C1&#x02032; of the template and C1&#x02032; of the incoming nucleotide (dNTP), normally near 10.8 &#x0212b; in a Watson&#x02013;Crick pair (<xref ref-type="bibr" rid="b65">65</xref>) (Supplementary Table S7); (iii) stacking interactions between the nascent base pair and the primer-terminus base pair (<xref ref-type="fig" rid="fig2">Figures 2</xref> and <xref ref-type="fig" rid="fig3">3</xref>, and Supplementary Tables S8 and S9); (iv) the frequency of sampling a near reaction-ready distance (3.1 to 3.5 &#x0212b;) between O3&#x02032; of the primer-terminus and P&#x003b1; of the dNTP (Supplementary Table S7); (v) angle O3&#x02032; (primer 3&#x02032; end)-P&#x003b1;(dNTP)-O3&#x003b1;(dNTP), ideally 180&#x000b0; for in-line attack of O3&#x02032; on P&#x003b1; (<xref ref-type="bibr" rid="b66">66</xref>) (Supplementary Table S7); (vi) the chelation of Mg²⁺ ions (<xref ref-type="fig" rid="fig2">Figures 2</xref> and <xref ref-type="fig" rid="fig3">3</xref>); (vii) the distance between the two Mg²⁺ ions (Supplementary Table S7) and (viii) hydrogen bonds between the nascent base pair and the polymerase. These are summarized in <xref ref-type="table" rid="tbl1">Tables 1</xref> and <xref ref-type="table" rid="tbl2">2</xref>, using a distortion scoring function to evaluate the lesion impact on the polymerase active site (<xref ref-type="bibr" rid="b34">34</xref>). While we analyze the structural features individually, we realize there is interdependence among them; how they affect each other remains unclear but their composite quality sheds light on a given model's structural feasibility. While scoring has subjective criteria elements, it reflects our current stage for evaluating polymerase active site distortions. We hypothesize that the more distorted the active site is, the less efficient the nucleotidyl transfer reaction will be.</p></sec><sec><title>G&#x000b7;dNTP models</title><p><italic>Unmodified control of anti-G&#x000b7;anti-dCTP.</italic> In the unmodified control simulation, <italic>anti</italic>-G Watson&#x02013;Crick pairs with an incoming <italic>anti</italic>-dCTP. Views of this complex and its active site after 3 ns production MD simulation are shown in Supplementary Figures S3a and S4a, and <xref ref-type="fig" rid="fig2">Figure 2a</xref>. The active site remains essentially undisturbed throughout the simulation. The Watson&#x02013;Crick hydrogen bonds in the nascent base pair have occupancies &#x0003e;90% (Supplementary Table S5); the C1&#x02032;&#x02013;C1&#x02032; distance in the nascent base pair remains normal (10.9 &#x000b1; 0.1 &#x0212b;); the chelation of the two Mg²⁺ ions is preserved (<xref ref-type="fig" rid="fig2">Figure 2a</xref>); the distance between the two Mg²⁺ has an average value of 3.7 &#x000b1; 0.1 &#x0212b;; the P&#x003b1;&#x02013;O3&#x02032; distance samples the near reaction-ready range frequently, during 78.7% of the time; and the in-line attack angle is 167.1 &#x000b1; 5.1&#x000b0;, close to the ideal value of 180&#x000b0;. These active site features are similar in polymerase ternary complex crystal structures (<xref ref-type="bibr" rid="b67">67</xref>,<xref ref-type="bibr" rid="b68">68</xref>). In the active site of Dpo4 polymerase, the flat face of the nascent base pair is topped by protein residues with unusually small and hydrophobic side chains: Val32, Ala42, Ala44, Ala57 and Gly58 (<xref ref-type="fig" rid="fig2">Figure 2a</xref>). These residues form the &#x02018;<italic>ceiling</italic>&#x02019; of the dNTP binding pocket (<xref ref-type="bibr" rid="b25">25</xref>). The &#x02018;<italic>floor</italic>&#x02019; of this pocket is considered to be the flat face of the base pair at the primer terminus (<xref ref-type="bibr" rid="b25">25</xref>). The geometry of the dNTP binding pocket remains normal here (<xref ref-type="fig" rid="fig2">Figure 2a</xref>). A highly conserved aromatic residue in the DinB family (Tyr12 in Dpo4, Phe12 in Dbh and Phe13 in DinB) is tightly packed on the sugar ring of the dNTP. This has the effect of placing the dNTP properly for catalysis and acting as a &#x02018;steric gate&#x02019; to prevent binding of ribo-NTPs through steric exclusion of a 2&#x02032;-OH (<xref ref-type="bibr" rid="b32">32</xref>,<xref ref-type="bibr" rid="b69">69</xref>). In our unmodified control simulation, the proper position of Tyr12 is preserved. Polymerase residues Tyr10, Phe11, Tyr12, Ser34, Thr45, Arg51, Lys159, Thr250 and Arg331 form hydrogen bonds with the nascent base pair in the unmodified control simulation (Supplementary Table S6).</p><p><italic>Syn-G*&#x000b7;anti-dATP minor groove.</italic> The <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP minor groove model is highly disturbed. With G* <italic>syn</italic>, the PhIP rings are on the minor groove side and inserted between the nascent base pair and the base pair at the primer terminus (<xref ref-type="fig" rid="fig2">Figure 2f</xref>). The location of the PhIP rings in the <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP simulation disrupts the flat face of the primer-terminus base pair, and also causes poor stacking interactions between the nascent and the primer-terminus base pairs, as shown by energetic consideration: a partial energetic assessment of differential stacking interactions can be obtained from van der Waals interactions between bases for the same sequences. In the <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP simulation, this interaction between the nascent and the primer-end base pairs is only &#x02212;2.3 &#x000b1; 0.5 kcal/mol, significantly weaker than &#x02212;16.8 &#x000b1; 0.9 kcal/mol in the <italic>anti</italic>-G*&#x000b7;<italic>syn</italic>-dATP, or &#x02212;15.8 &#x000b1; 0.9 kcal/mol in the <italic>anti</italic>-G*&#x000b7;<italic>anti</italic>-dATP case. The <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP also manifests an enlarged C1&#x02032;&#x02013;C1&#x02032; distance (12.2 &#x000b1; 0.4 &#x0212b;), and a less frequently sampled near reaction-ready P&#x003b1;-O3&#x02032; distance (59.8%). The hydrogen bond in the nascent base pair in the <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP initial model, between O6 of the <italic>syn</italic>-G* and N6 of the <italic>anti</italic>-dATP, is ruptured during equilibration and remains broken throughout the simulation. To achieve the hydrogen bond in the initial model, the PhIP moiety was positioned in a crowded region near the ceiling of the dNTP binding pocket, which proved unfavorable (Supplementary Figure S1f). These disturbances, triggered by the <italic>syn</italic>-G* adduct, are essentially independent of the specific dNTP.</p><p><italic>Anti-G*&#x000b7;anti-dATP and anti-G*&#x000b7;syn-dATP major groove.</italic> The <italic>anti</italic>-G*&#x000b7;<italic>anti</italic>-dATP and <italic>anti</italic>-G*&#x000b7;<italic>syn</italic>-dATP major groove models are more disturbed than the unmodified control, but less than the <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP minor groove model. In the <italic>anti</italic>-G*&#x000b7;<italic>syn</italic>-dATP complex, the phenyl ring of the PhIP is directed towards the 3&#x02032; end of the template, along the template backbone (<xref ref-type="fig" rid="fig2">Figure 2e</xref>), while in the <italic>anti</italic>-G*&#x000b7;<italic>anti</italic>-dATP complex, it is oriented toward the 5&#x02032; end of the template (<xref ref-type="fig" rid="fig2">Figure 2d</xref>). A &#x0223c;180&#x000b0; difference in &#x003b1;&#x02032; is responsible for this difference in orientation (Supplementary Figure S6). With an incoming <italic>syn</italic>-dATP, one full and one bifurcated hydrogen bond are formed in the nascent base pair (Supplementary Figure S5 and Supplementary Table S5), with a close to normal C1&#x02032;&#x02013;C1&#x02032; distance (11.1 &#x000b1; 0.2 &#x0212b;). On the other hand, with the incoming nucleotide <italic>anti</italic>, the C1&#x02032;&#x02013;C1&#x02032; distance is significantly enlarged (12.7 &#x000b1; 0.2 &#x0212b;), and the full hydrogen bond between O6 of G* and N6 of <italic>anti</italic>-dATP has an occupancy 20% lower than for <italic>syn</italic>-dATP (Supplementary Table S5). A bifurcated hydrogen bond is formed between the little finger residue Arg332 and the PhIP rings only in the <italic>anti</italic>-G*&#x000b7;<italic>syn</italic>-dATP simulation (Supplementary Table S6). The frequencies of sampling the near reaction-ready P&#x003b1;-O3&#x02032; distance and the geometries of the binding pocket are not dramatically different in these two major groove complexes (Supplementary Table S7 and <xref ref-type="fig" rid="fig2">Figure 2</xref>).</p><p><italic>Anti-G*&#x000b7;anti-dTTP and anti-G*&#x000b7;syn-dGTP major groove.</italic> The <italic>anti</italic>-G*&#x000b7;<italic>anti</italic>-dTTP major groove model is little distorted, while the <italic>anti</italic>-G*&#x000b7;<italic>syn</italic>-dGTP major groove structure is very distorted. At this stage, the <italic>syn</italic>-G* was eliminated due to poor accommodation of the minor groove positioned PhIP regardless of dNTP, as revealed in the <italic>syn</italic>-G*&#x000b7;<italic>anti</italic>-dATP simulation (<xref ref-type="fig" rid="fig2">Figure 2f</xref> and <xref ref-type="table" rid="tbl1">Table 1</xref>). Therefore, only <italic>anti</italic>-G* was employed to build the models for G*&#x000b7;dTTP and G*&#x000b7;dGTP. With the incoming dTTP <italic>anti</italic>, the <italic>anti</italic>-G* forms a wobble pair with the <italic>anti</italic>-dTTP (Supplementary Figure S5). During the <italic>anti</italic>-G*&#x000b7;<italic>anti</italic>-dTTP simulation, the two hydrogen bonds in the wobble nascent base pair are highly occupied (&#x0003e;94% of the time); the C1&#x02032;&#x02013;C1&#x02032; distance has a normal average value of 10.8 &#x000b1; 0.2 &#x0212b;; the P&#x003b1;-O3&#x02032; distance remains in the near reaction-ready range almost all the time (93.0%), and the nascent base pair is well stacked with the base pair at the primer-terminus. In the G*&#x000b7;G mismatch model, the incoming dGTP is <italic>syn</italic> and forms two hydrogen bonds with the <italic>anti</italic>-G* (Supplementary Figure S5 and Supplementary Table S5). However, the C1&#x02032;&#x02013;C1&#x02032; distance is enlarged to 11.9 &#x000b1; 0.3 &#x0212b;, and a somewhat less frequently sampled near reaction-ready P&#x003b1;-O3&#x02032; distance is noted (68.1%).</p><p><italic>Anti-G*&#x000b7;anti-dCTP major groove.</italic> With <italic>anti</italic>-G* opposite a Watson&#x02013;Crick paired dCTP in the active site, the structural features are comparable to those of the unmodified control. In this <italic>anti</italic>-G*&#x000b7;<italic>anti</italic>-dCTP simulation, the Watson&#x02013;Crick hydrogen bonds in the nascent base pair are highly occupied (&#x0003e;95% of the time); the C1&#x02032;&#x02013;C1&#x02032; distance has a normal average value of 10.9 &#x000b1; 0.9 &#x0212b;; the nascent base pair stacks well with the base pair at the primer-terminus, and the near reaction-ready P&#x003b1;-O3&#x02032; distance is frequently sampled (82.4% of the time). However, two protein&#x02013;DNA interactions have lower hydrogen bond occupancies in comparison to the unmodified control (Supplementary Table S6).</p><p><italic>Controls of anti-G&#x000b7;syn-dATP, anti-G&#x000b7;anti-dTTP and anti-G&#x000b7;syn-dGTP.</italic> These control models containing G&#x000b7;dATP, G&#x000b7;dTTP and G&#x000b7;dGTP mismatches without PhIP modification show distortions at the active sites in comparison to the Watson&#x02013;Crick G&#x000b7;dCTP paired unmodified control (<xref ref-type="table" rid="tbl1">Table 1</xref>). In the control of <italic>anti</italic>-G<italic>&#x000b7;syn</italic>-dATP, although the C1&#x02032;&#x02013;C1&#x02032; distance has a close to normal value of 11.2 &#x000b1; 0.2 &#x0212b; (Supplementary Table S7), the full and the bifurcated hydrogen bonds in the nascent base pair have lower occupancies (&#x0003c;90% of the time, Supplementary Table S5), and the near reaction-ready P&#x003b1;-O3&#x02032; distance is less frequently sampled (61.4% of the time, Supplementary Table S7). In the control of <italic>anti</italic>-G<italic>&#x000b7;anti</italic>-dTTP, the C1&#x02032;&#x02013;C1&#x02032; distance has a normal value of 10.8 &#x000b1; 0.2 &#x0212b; (Supplementary Table S7). However, one of the two hydrogen bonds in the nascent pair is less frequently occupied (80.2% of the time, Supplementary Table S5), and the P&#x003b1;-O3&#x02032; distance sampled the near reaction-ready range during only 59.7% of the time (Supplementary Table S5). In the control of <italic>anti</italic>-G<italic>&#x000b7;syn</italic>-dGTP, the C1&#x02032;&#x02013;C1&#x02032; distance is enlarged to 11.9 &#x000b1; 0.3 &#x0212b; and one of the two hydrogen bonds in the nascent base pair is poorly occupied (only 46.6% of the time, Supplementary Table S5). However, the near reaction-ready P&#x003b1;-O3&#x02032; distance is frequently sampled (76.7% of the time, Supplementary Table S5).</p></sec><sec><title>&#x02212;1 Deletion models</title><p><italic>Anti-G*, syn-G* and unmodified control.</italic> In these models, the damaged template guanine has been skipped and the incoming dCTP pairs with the guanine on the 5&#x02032; side of the adduct (<xref ref-type="fig" rid="fig1">Figure 1c</xref>). In comparison to the unmodified &#x02212;1 control system, the two adduct systems, the <italic>anti</italic>-G* &#x02212;1 deletion and <italic>syn</italic>-G* &#x02212;1 deletion are significantly distorted. In the <italic>anti</italic>-G* &#x02212;1 deletion model, the PhIP moiety is positioned on the major groove side, near-perpendicular to the template strand (<xref ref-type="fig" rid="fig3">Figure 3b</xref>); in the <italic>syn</italic>-G* &#x02212;1 deletion model, the PhIP moiety is positioned on the minor groove side (<xref ref-type="fig" rid="fig3">Figure 3c</xref>). Views of these ternary complexes and their active sites after 3 ns production MD simulations are shown in Supplementary Figures S3 and S4, and <xref ref-type="fig" rid="fig3">Figure 3</xref>. For the <italic>syn</italic>-G* &#x02212;1 deletion system, the PhIP rings are reoriented through rotation of &#x003b1;&#x02032; and &#x003b2;&#x02032; after about 320 ps of MD (Supplementary Figure S7) to avoid crowding between the distal phenyl ring and Lys78. This rearrangement results in a favorable environment around the PhIP ring system. The phenyl ring is in van der Waals contact with hydrophobic residue Ile104; the imidazo ring N3 and N4 atoms have favorable electrostatic interactions with the amino group of Lys78 (actual hydrogen bonds are formed &#x0223c;9% of the time). As shown in Supplementary Figure S8, the PhIP moiety is pocketed by the protein residues, the nascent base pair and the primer-terminus base pair.</p><p>Watson&#x02013;Crick hydrogen bonds in the nascent base pair are preserved in all three simulations (Supplementary Table S10). The C1&#x02032;&#x02013;C1&#x02032; distances are also normal in all cases (Supplementary Table S7), as is the chelation of the two Mg²⁺ ions (<xref ref-type="fig" rid="fig3">Figure 3</xref>). Compared to the unmodified control, the primer end is repositioned towards the 5&#x02032; direction and away from the active site in the <italic>anti</italic>-G* and <italic>syn</italic>-G*adduct simulations (Supplementary Figure S9); this is more prominent in the <italic>syn</italic>-G* &#x02212;1 deletion simulation, where the terminal primer base is shifted by about 3 &#x0212b;. The backbone of the primer strand is concomitantly shifted, accompanied by small rotation of the thumb domain, which maintains contacts with the relocated backbone phosphate groups of the primer strand. In order to retain the B-DNA geometry, the backbone of the template strand is also adjusted, followed by small rotation of the little finger domain in order to maintain contacts with the same phosphate groups of the template strand. A similar relocation of the primer terminal base was observed in a Dpo4 crystal structure containing the 10<italic>R</italic>(+)-<italic>cis-anti</italic>-[BP]-<italic>N</italic>⁶-dA adduct, with BP aromatic rings intercalated between the nascent and the primer terminal base pair (<xref ref-type="bibr" rid="b16">16</xref>). Here, the primer-terminus was positioned &#x0003e;10 &#x0212b; away from the incoming dNTP. In both the <italic>anti</italic>-G* and <italic>syn</italic>-G* &#x02212;1 deletion simulations, the dCTP loses its proper stacking contact with the &#x02018;steric gate&#x02019; Tyr12 from the finger domain, and its hydrogen bonding with the finger residue Arg51. The separation of the dCTP from the finger domain could be a result of hydrogen bonds formed between the dCTP and the adduct (Supplementary Table S11). The P&#x003b1;-O3&#x02032; distance is also affected by this primer side relocation. In the unmodified control system for the &#x02212;1 deletion, the near reaction-ready P&#x003b1;-O3&#x02032; distance is sampled during the whole simulation (51.4%, Supplementary Figure S10k). In the <italic>anti</italic>-G* &#x02212;1 deletion model, this distance is only achieved in the 380&#x02013;480 ps time frame, and comes close to 4 &#x0212b; occasionally after this time range (Supplementary Figure S10l). However, in the <italic>syn</italic>-G* &#x02212;1 deletion case, the distance has an average value of 7.5 &#x000b1; 0.7 &#x0212b; throughout the simulation, and does not approach closer than 5.5 &#x0212b; (Supplementary Figure S10m). In the <italic>syn</italic>-G* &#x02212;1 deletion system, the stacking between the PhIP-modified <italic>syn</italic>-G and its 3&#x02032;-neighbour is also disrupted (<xref ref-type="fig" rid="fig3">Figure 3c</xref>). Thus, the &#x02212;1 deletion simulations suggest various distortions likely to hinder nucleotide incorporation.</p></sec></sec><sec><title>DISCUSSION</title><p>Our simulations suggest that regardless of the incoming dNTP or whether the damaged base has a partner, the dG-C8-PhIP adduct is not likely to be accommodated in the minor groove side pocket of the Dpo4 DinB family polymerase (<xref ref-type="table" rid="tbl1">Tables 1</xref> and <xref ref-type="table" rid="tbl2">2</xref>). When situated in this position, with the G* in the <italic>syn</italic> conformation, the bulky aromatic rings disrupt the geometry of the active site, particularly on the primer side (<xref ref-type="fig" rid="fig2">Figure 2f</xref> and <xref ref-type="fig" rid="fig3">Figure 3c</xref>). With the G* <italic>anti</italic>, however, the aromatic rings fit well on the major groove side. With dCTP opposite the <italic>anti</italic>-G*, Watson&#x02013;Crick pairing is preserved and simulations show only modest distortions compared to the unmodified control (Supplementary Table S5 and <xref ref-type="table" rid="tbl1">Table 1</xref>). An incoming dTTP affords a well-formed wobble pair with only small distortions. In the case of the dATP, greater distortions are observed, with <italic>syn</italic>-dATP providing the more favorable structure, which contains hydrogen bonds employing the Hoogsteen edge of the dATP. Incoming dGTP was most distorted because even the <italic>syn</italic> conformation for dGTP caused enlargement of the nascent base pair. Furthermore, in the case of structures with no partner opposite the lesion, the &#x02212;1 deletion models, the PhIP aromatic rings can again reside in the large major groove side open space; however, the active site region is notably distorted, and the approach of the primer-terminus to the &#x003b1;-phosphate of dNTP is inhibited by the PhIP moiety, particularly by its protruding methyl group (Supplementary Figure S9 and Supplementary Table S7).</p><p>In contrast, an earlier study of the 10<italic>S</italic>(+)-<italic>trans-anti</italic>-[BP]-<italic>N</italic>²-dG showed that this adduct can be accommodated reasonably well on the minor groove side of the Dpo4, irrespective of incoming dNTP, with only modest enzyme perturbation including opening of the little finger and some small rearrangement of active site region residues (<xref ref-type="bibr" rid="b54">54</xref>). In this case, the modified dG adopts the <italic>anti</italic> conformation. The major groove position with the <italic>N</italic>²-dG adduct in <italic>syn</italic> conformation also provides a good accommodation for the BP ring system. Experimental kinetic studies revealed promiscuous nucleotide incorporation for this case. Running-start primer extension experiments indicated that the damage can be bypassed to a significant extent (<xref ref-type="bibr" rid="b54">54</xref>).</p><p><xref ref-type="fig" rid="fig4">Figure 4</xref> illustrates the differential accommodation of the C8 and <italic>N</italic>²-dG adducts in the Dpo4 minor groove, showing why the C8 adduct is not well positioned there, while the <italic>N</italic>² adduct is, regardless of dNTP (see Supplementary Figure S11 for stereoview). Specifically, this <italic>N</italic>² adduct ring system is directed 5&#x02032; along the template strand in our models; however, the C8 adduct rings are oriented 3&#x02032; along the template strand, protrude to the primer side and thereby disrupt the active site directly where the nucleotidyl transfer reaction is to take place. Perhaps C8-dG bulky adducts, such as dG-C8-PhIP and dG-C8-AAF might less readily allow nucleotide incorporation by Dpo4, in part because of their poor accommodation in the enzyme minor groove side pocket; thus, only one site is available to harbor the lesion, while <italic>N</italic>² adducts, such as 10<italic>S</italic>(+)-<italic>trans-anti</italic>-[BP]-<italic>N</italic>²-dG can reside in either major or minor groove side pockets.</p><p>Additionally, our structures suggest that translocation may be difficult for such major groove positioned multi-ringed adducts. Specifically, a recent crystal structure of Dpo4 has suggested that translocation starts with movement of the little finger during nucleotide binding, followed by thumb movement during the chemical reaction (<xref ref-type="bibr" rid="b23">23</xref>). The replication cycle involves a screw-like counterclockwise rotation/translocation of the polymerase along the DNA helix axis (viewed in the 5&#x02032; to 3&#x02032; direction of the template strand). This threading of the DNA through the polymerase places the next templating base in the active site, awaiting entry of its complementary dNTP. Prior study of the dG-C8-AAF adduct suggested that the preferred 3&#x02032;-directed orientation of the fluorenyl ring system along the template strand, in the major groove, would impede the rotation of the little finger (<xref ref-type="bibr" rid="b34">34</xref>); this impediment is not due to the acetyl group. The current study of the dG-C8-PhIP adduct suggests a similar impact on translocation. As shown in <xref ref-type="fig" rid="fig5">Figure 5a and b</xref>, in the insertion position the adduct in the <italic>anti</italic> conformation is enveloped by the little finger without collision (see Supplementary Figure S12 for stereoview). However, once in the post-insertion site, the adduct would be in collision with the little finger, indicating that translocation to the post-insertion site would be difficult. Very large rearrangement of the PhIP adduct would be required to move it to another orientation that allows translocation. Additional bypass polymerases may be required to further extend beyond the lesion (<xref ref-type="bibr" rid="b4">4</xref>). Furthermore, it appears that <italic>N</italic>² adducts, such as 10<italic>S</italic>(+)-<italic>trans-anti</italic>-[BP]-<italic>N</italic>²-dG, would be less likely to impede translocation by colliding with the little finger, when in the <italic>anti</italic> conformation and placed in the minor groove side polymerase pocket (<xref ref-type="fig" rid="fig5">Figure 5c</xref> and d). The number of rings in the adduct together with their orientation will determine the extent of these effects on translocation and this remains to be elucidated.</p><p>Primer extension studies of dG-<italic>N</italic>²-AAF and dG-C8-AAF with hpol &#x003ba; have shown that the frequency of bypass of dG-<italic>N</italic>²-AAF is at least 4 orders of magnitude higher than for dG-C8-AAF, and it was proposed that the observed difference may be due to the different linkage site (<xref ref-type="bibr" rid="b70">70</xref>). This is consistent with our suggested poorer accommodation on the minor groove side of C8 than <italic>N</italic>² adducts. However, we do not know exactly how structurally similar pol &#x003ba; and Dpo4 ternary complexes are, since to date, only a crystal structure of the apo-pol &#x003ba; catalytic core is available (<xref ref-type="bibr" rid="b71">71</xref>).</p><p>Furthermore, our results show that the PhIP-modified lesion produces less distorted structures for G*&#x000b7;dTTP and G*&#x000b7;dATP (<italic>anti</italic>-G*<italic>&#x000b7;anti</italic>-dTTP and <italic>anti</italic>-G*<italic>&#x000b7;syn</italic>-dATP) than for their respective unmodified controls (<italic>anti</italic>-G<italic>&#x000b7;anti</italic>-dTTP and <italic>anti</italic>-G<italic>&#x000b7;syn</italic>-dATP) (<xref ref-type="table" rid="tbl1">Table 1</xref>). The structural origin of this interesting phenomenon is at least partly in the formation of hydrogen bonds between the PhIP rings and the little finger residue Arg332, which stabilize the <italic>anti</italic> conformation in the adduct (<xref ref-type="fig" rid="fig2">Figure 2</xref> and Supplementary Table S6). These findings would be relevant to observed mutagenic behavior of PhIP in inducing G to T transversions and G to A transitions in mammalian systems (<xref ref-type="bibr" rid="b44">44</xref>&#x02013;<xref ref-type="bibr" rid="b52">52</xref>), if error-prone incorporation opposite the lesion involved the human DinB polymerase pol &#x003ba;, and its structural properties prove similar to those of its prokaryotic homolog Dpo4.</p><p>In conclusion, our modeling and MD simulations for dG-C8-PhIP suggest that the adduct would increase the infidelity of Dpo4 and hinder translocation by the enzyme. We hope that the hypotheses resulting from our modeling studies will provide useful suggestions for future experimental investigations.</p></sec><sec><title>SUPPLEMENTARY DATA</title><p>Supplementary Data are available at NAR online.</p></sec>

Recombination R-triplex: H-bonds contribution to stability as revealed with minor base substitutions for adenine

<p>Several cellular processes involve alignment of three nucleic acids strands, in which the third strand (DNA or RNA) is identical and in a parallel orientation to one of the DNA duplex strands. Earlier, using 2-aminopurine as a fluorescent reporter base, we demonstrated that a self-folding oligonucleotide forms a recombination-like structure consistent with the R-triplex. Here, we extended this approach, placing the reporter 2-aminopurine either in the 5&#x02032;- or 3&#x02032;-strand. We obtained direct evidence that the 3&#x02032;-strand forms a stable duplex with the complementary central strand, while the 5&#x02032;-strand participates in non-Watson&#x02013;Crick interactions. Substituting 2,6-diaminopurine or 7-deazaadenine for adenine, we tested and confirmed the proposed hydrogen bonding scheme of the A*(T&#x000b7;A) R-type triplet. The adenine substitutions expected to provide additional H-bonds led to triplex structures with increased stability, whereas the substitutions consistent with a decrease in the number of H-bonds destabilized the triplex. The triplex formation enthalpies and free energies exhibited linear dependences on the number of H-bonds predicted from the A*(T&#x000b7;A) triplet scheme. The enthalpy of the 10 nt long intramolecular triplex of &#x02212;100 kJ&#x000b7;mol^&#x02212;1 demonstrates that the R-triplex is relatively unstable and thus an ideal candidate for a transient intermediate in homologous recombination, t-loop formation at the mammalian telomere ends, and short RNA invasion into a duplex. On the other hand, the impact of a single H-bond, 18 kJ&#x000b7;mol^&#x02212;1, is high compared with the overall triplex formation enthalpy. The observed energy advantage of a &#x02018;correct&#x02019; base in the third strand opposite the Watson&#x02013;Crick base pair may be a powerful mechanism for securing selectivity of recognition between the single strand and the duplex.</p>

<contrib contrib-type="author"><name><surname>Shchyolkina</surname><given-names>Anna K.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Kaluzhny</surname><given-names>Dmitry N.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Arndt-Jovin</surname><given-names>Donna J.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Jovin</surname><given-names>Thomas M.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Zhurkin</surname><given-names>Victor B.</given-names></name><xref rid="au2" ref-type="aff">2</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><aff><institution>Engelhardt Institute of Molecular Biology, Russian Academy of Sciences</institution><addr-line>119991 Moscow, Russia</addr-line></aff><aff id="au1">¹<institution>Department of Molecular Biology, Max Planck Institute for Biophysical Chemistry</institution><addr-line>D-37070 Goettingen, Germany</addr-line></aff><aff id="au2">²<institution>Laboratory of Cell Biology, National Cancer Institute</institution><addr-line>NIH, Bethesda, MD 20892, USA</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p>The alignment of three strands of nucleic acids, in which the third strand is identical and oriented parallel to one of the DNA duplex strands, occurs in recombination (<xref ref-type="bibr" rid="b1">1</xref>), at telomeres during t-loop formation (<xref ref-type="bibr" rid="b2">2</xref>), in DNA rearrangements caused by replication of mitochondrial DNA (<xref ref-type="bibr" rid="b3">3</xref>), and has been invoked in initiation of heterochromatin by small hairpin RNAs (<xref ref-type="bibr" rid="b4">4</xref>). In principle, such an alignment is capable of forming the R-triplex structure (recombination triplex DNA) as well as a strand exchange structure (D-loop) resulting from the third strand invasion into the duplex. An R-triplex (or R-form DNA) structure was predicted theoretically (<xref ref-type="bibr" rid="b5">5</xref>) to consist of isomorphic R-triplets; i.e. the putative R-triplex can accommodate any arbitrary sequence. R-type hydrogen bond schemes for the G*(C&#x000b7;G) and C*(G&#x000b7;C) triplets have been visualized by X-ray analysis (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b7">7</xref>), while the T*(A&#x000b7;T) and A*(T&#x000b7;A) triplets were verified by FTIR (<xref ref-type="bibr" rid="b8">8</xref>,<xref ref-type="bibr" rid="b9">9</xref>). Data supporting the existence of the R-triplex in protein-free systems derived from UV and fluorescence thermal denaturation curves, chemical probing, as well as gel shift and FRET assays (<xref ref-type="bibr" rid="b10">10</xref>&#x02013;<xref ref-type="bibr" rid="b13">13</xref>) have been published by our laboratories.</p><p>Special oligonucleotide constructs (RCW fold) consisting of three strands connected with nucleotide linkers (<xref ref-type="fig" rid="fig1">Figure 1</xref>) in which the 5&#x02032;-terminal strand is denoted as R (Recombination), the central strand as C (Crick, &#x02018;complementary&#x02019;) and the 3&#x02032;-strand as W (Watson) have been used to study the alignment of the strands (<xref ref-type="bibr" rid="b13">13</xref>). The two linker loops differ in their conformational flexibility, the GAA loop being extremely stable, while the TTTT loop is less rigid. Importantly, such an RCW fold can accommodate any nucleotide sequence with the two identical R- and W-strands and a complementary C-strand.</p><p>We demonstrated previously that the fluorescent base analog, 2-aminopurine (2AP), can be substituted for adenine in the A*(T&#x000b7;A) triplet, forming the 2AP*(T&#x000b7;A) triplet, stereochemically consistent with the R-form (<xref ref-type="fig" rid="fig2">Figure 2</xref>). The temperature-dependent cooperative dissociation of the R-strand from the duplex was detected by two techniques: (i) fluorescence of the 2AP which monitored disruption of the individual A*(T&#x000b7;A) triplet and (ii) conventional UV absorbance at 260 nm reflecting melting of the third strand as a whole. Both methods revealed the same temperature-dependent cooperative dissociation of the R-strand from the duplex part (<xref ref-type="bibr" rid="b13">13</xref>), thus demonstrating that the fluorescence of the 2AP reported faithfully the thermodynamic parameters of the RCW triplex fold.</p><p>In the present work, we extended this approach by placing the fluorescent 2AP and other substitutions for adenine, site-specifically, either in the 5&#x02032;- or 3&#x02032;-strand. These substitutions allowed us (i) to monitor the difference in conformational behavior between the two identical R- and W-strands in the RCW fold and (ii) to obtain quantitative data on the energetics of the R-triplex formation. In particular, we were interested in deducing the relative impact of the base&#x02013;base interactions in the duplex part of the triplex, and between the duplex and the third strand, i.e. issues directly related to the fidelity of nucleic acid recognition.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><sec><title>Oligonucleotides</title><p>R<italic>^a</italic>CW, RCW<italic>^a</italic>, R<italic>^add</italic>CW, R<italic>^a</italic>CW<italic>⁷</italic>, R<italic>^a</italic>CW<italic>⁷⁷</italic>, R<italic>^aaa</italic>CW<italic>⁷⁷⁷</italic> and CW<italic>^a</italic> were synthesized and purified by high-performance liquid chromatography by Midland Certified Reagent Co. Inc. (TX); R<italic>^add</italic>CW<italic>^ddd</italic> and RCW<italic>^a</italic> were synthesized and purified in PAAG by Syntol (Moscow) (for designations and folding schemes see <xref ref-type="fig" rid="fig1">Figure 1</xref>). The hydrogen bonding schemes for the A*(T&#x000b7;A) triplet and the triplets with 2AP, 2,6-diaminopurine (DAP) and 7-deazaadenine (7DAA) substitutions tested in the study are given in <xref ref-type="fig" rid="fig2">Figure 2</xref>. Samples contained 0.9&#x02013;1.4 &#x000b5;M oligonucleotides, 0.5 M LiCl and 10 mM Tris&#x02013;HCl buffer, pH 7.6.</p></sec><sec><title>Fluorescence measurements</title><p>The temperature dependence of the fluorescence emission of oligonucleotides containing a single 2AP substitution were recorded with a Cary Eclipse fluorescence spectrophotometer (Varian) in a thermostated cuvette at the constant heating of 0.5&#x000b0;C/min. The excitation wavelength was 310 nm, and the maximum of the emission was 370 nm. Samples contained 1 &#x000b5;M oligonucleotides, 0.5 M LiCl and 10 mM Tris&#x02013;HCl buffer, pH 7.6.</p></sec><sec><title>UV thermal denaturation profiles</title><p>UV thermal denaturation curves were recorded at 260 nm with a Cary 100 Scan UV-visible spectrophotometer (Varian) with a constant heating gradient of 0.5 or 0.2&#x000b0;C/min.</p></sec><sec><title>Thermodynamic analysis of the intramolecular triplex formation</title><p>The detailed analysis of the thermal denaturation profiles, recorded by fluorescence of the 2AP reporter, has been described elsewhere (<xref ref-type="bibr" rid="b13">13</xref>). In essence, we analyzed the intramolecular binding of the dangling third strand to its double-helical CW part by fitting a theoretical curve to the experimental fluorescence melting curve. From these data we deduced the basic thermodynamic parameters of the intramolecular transition (denoted below as the &#x02018;triple helix formation&#x02019;).</p></sec></sec><sec><title>RESULTS AND DISCUSSION</title><sec><title>Probing conformation of the two homologous strands with 2AP</title><p>Substitution of 2AP for adenine constitutes an ideal structural probe for testing the RCW fold; the probe neither destabilizes nor distorts the 3D structure (<xref ref-type="bibr" rid="b13">13</xref>). Using 2AP as a reporter base in the 5&#x02032;-strand, we demonstrated previously (<xref ref-type="bibr" rid="b13">13</xref>) that the R<italic>^a</italic>CW oligonucleotide forms a sequence-specific structure, whose conformational equilibrium is shifted toward the R-type triplex. However, in these studies, we presented no direct evidence regarding conformation of the 3&#x02032;-strand.</p><p>To detect the difference in conformation between the two identical R- and W-strands in the RCW fold, we substituted 2AP for adenine in the Watson 3&#x02032;-strand (<xref ref-type="fig" rid="fig1">Figure 1</xref>, RCW<italic>^a</italic> oligonucleotide) and compared the temperature dependence of the fluorescence intensity for RCW<italic>^a</italic> with those for CW<italic>^a</italic> and R<italic>^a</italic>CW (<xref ref-type="fig" rid="fig3">Figure 3A</xref>). The enhancement of R<italic>^a</italic>CW fluorescence (circles) in the temperature range from 5 to 40&#x000b0;C reflected increasing exposure of 2AP to solvent due to melting of the triplex, namely a progressive loss of the 2AP stacking with the adjacent bases. Above 35&#x02013;40&#x000b0;C, the R<italic>^a</italic>CW melting profile followed that of the single R2AP strand (<xref ref-type="bibr" rid="b13">13</xref>). The temperature profile for the RCW<italic>^a</italic> oligonucleotide was entirely different. In this case, the fluorescence monotonically decreased (<xref ref-type="fig" rid="fig3">Figure 3A</xref>, squares), similar to the profile for the double-stranded hairpin CW<italic>^a</italic> (<xref ref-type="fig" rid="fig3">Figure 3A</xref>, diamonds). These results, and the fact that the RCW<italic>^a</italic> and CW<italic>^a</italic> oligonucleotides fluoresce similarly at equal concentrations, demonstrated that the fluorophore is in the same state in the RCW<italic>^a</italic> and CW<italic>^a</italic> folds, showing a temperature dependence of 2AP fluorescence that is typical for a DNA duplex. Thus, we conclude that the Watson 3&#x02032;-strand retains base pairing with the complementary Crick strand under our experimental conditions.</p><p>The specific features of the melting profiles of 2AP fluorescence are likely related to a relatively hydrophobic environment with considerable base stacking inside the double helix (<xref ref-type="bibr" rid="b14">14</xref>) resulting in a markedly quenched fluorescence of the 2AP compared with that of 2AP in the third strand of the triplex. Further quenching of fluorescence of 2AP located within the double helix was probably promoted with increasing temperature through the non-radiative relaxation of excited states (<xref ref-type="bibr" rid="b15">15</xref>). A dynamic invasion of the 5&#x02032;-strand in the Watson&#x02013;Crick duplex displacing the identical 3&#x02032;-strand appears improbable, as it would have caused a detectable increase in the 2AP fluorescence in RCW<italic>^a</italic> compared with CW<italic>^a</italic>. In other words, these data provide direct evidence against a dynamic mixture of two different duplex folds, i.e. branch migration structure, whereby the complementary C-strand is partially paired with both the 5&#x02032;- and the 3&#x02032;-strand.</p><p>Interpretation of the data is based on the assumption that the oligonucleotides form intramolecular folds rather than intermolecular associates. The nucleotide sequence, salt conditions and the procedure for the sample preparation have been selected to avoid formation of intermolecular species (<xref ref-type="bibr" rid="b13">13</xref>). The intramolecular character of the R<italic>^a</italic>CW folding has been demonstrated by evaluating the oligonucleotide hydrodynamic volume under these experimental conditions (<xref ref-type="bibr" rid="b13">13</xref>). Here, we present additional, independent evidence for the intramolecular folding of R<italic>^a</italic>CW by comparing melting curves for two concentrations of the oligonucleotide at 1 and 50 &#x000b5;M (<xref ref-type="fig" rid="fig3">Figure 3B</xref>). The identity of two melting transitions for the concentrations differing by 50-fold unambiguously corroborates the intramolecular folding of the R<italic>^a</italic>CW oligonucleotide.</p></sec><sec><title>2,6-Diaminopurine substitutions for adenines in the homologous R- and W-strands</title><p>To gain further information about the RCW fold structure, we made substitutions of adenines by DAP in the Watson and/or R-strands. The oligonucleotide R<italic>^add</italic>CW contained two diaminopurines in the R-strand (<xref ref-type="fig" rid="fig1">Figure 1</xref>) in addition to the 2AP as in R<italic>^a</italic>CW. Potentially, these additional substitutions would lead to formation of two DAP*(T&#x000b7;A) triplets (<xref ref-type="fig" rid="fig2">Figure 2</xref>), thereby stabilizing the RCW triplex through the additional H-bond in these triplets. Alternatively, the complementary C-strand may &#x02018;change&#x02019; partners and form a canonical duplex with the 5&#x02032;-R-strand (strand invasion), since DAP is known to pair with thymine forming three hydrogen bonds (<xref ref-type="bibr" rid="b16">16</xref>). We measured the temperature-dependent 2AP fluorescence of R<italic>^add</italic>CW to distinguish between these two possibilities.</p><p>The temperature-dependent fluorescence curve of the R<italic>^add</italic>CW oligonucleotide (<xref ref-type="fig" rid="fig3">Figure 3A</xref>, triangles) gave a similar profile to that of CW<italic>^a</italic> (<xref ref-type="fig" rid="fig3">Figure 3A</xref>, diamonds) and RCW<italic>^a</italic> (<xref ref-type="fig" rid="fig3">Figure 3A</xref>, squares), but different from that of R<italic>^a</italic>CW (<xref ref-type="fig" rid="fig3">Figure 3A</xref>, circles). Based on these data, we conclude that the 5&#x02032;-strand forms a duplex with the C-strand, which involves two stable T&#x000b7;DAP base pairs. Therefore, introduction of two DAP molecules in the 5&#x02032;-strand strongly affects the mode of RCW folding and promotes strand exchange as opposed to an R-triplex. Note that the rearrangement of the RCW fold observed here upon the A substitutions by DAP in the R-strand is topologically equivalent to the strand exchange promoted by RecA protein (<xref ref-type="bibr" rid="b1">1</xref>,<xref ref-type="bibr" rid="b5">5</xref>).</p><p>In oligonucleotide R<italic>^add</italic>CW<italic>^ddd</italic>, DAP was substituted for the adenines in both the 5&#x02032;-R- and 3&#x02032;-W-strands except for the reporter 2AP in the R-strand (<xref ref-type="fig" rid="fig1">Figure 1</xref>). According to the triplet scheme of <xref ref-type="fig" rid="fig2">Figure 2</xref>, we expected the two DAP*(T&#x000b7;DAP) triplets to stabilize the RCW fold by providing additional hydrogen bonds. Indeed, the R<italic>^add</italic>CW<italic>^ddd</italic> construct has a higher stability as follows from the fact that the temperature-dependent fluorescence profile is shifted rightward (<xref ref-type="fig" rid="fig4">Figure 4</xref>, circles).</p><p>Although DAP absorbs weakly at 310 nm, it was necessary to assess the potential contribution of the five DAP residues in the R<italic>^add</italic>CW<italic>^ddd</italic> construct to the R<italic>^add</italic>CW<italic>^ddd</italic> emission upon excitation at this wavelength. We compared the R<italic>^add</italic>CW<italic>^ddd</italic> fluorescence melting curve with the transition registered by UV absorption at 260 nm (<xref ref-type="fig" rid="fig4">Figure 4</xref>, inset, circles and solid line, respectively). The two curves coincide below 40&#x000b0;C, i.e. in the region where the UV melting reflects mainly the thermal dissociation of the third stand from the more stable duplex part of the R<italic>^add</italic>CW<italic>^ddd</italic>. We conclude that the contribution of DAP fluorescence to the melting profiles of the oligonucleotide R<italic>^add</italic>CW<italic>^ddd</italic> (as well as R<italic>^add</italic>CW, also containing the DAP bases) upon excitation at 310 nm is negligible.</p><p>Thermodynamic parameters for this construct as well as for R<italic>^a</italic>CW, determined with our fitting procedure are listed in <xref ref-type="table" rid="tbl1">Table 1</xref>. The R<italic>^add</italic>CW<italic>^ddd</italic> triplex melting temperature increased by 6&#x000b0;C, resulting in gain in <italic>&#x00394;H</italic> of &#x0223c;27 kJ&#x000b7;mol^&#x02212;1 and an increase in <italic>&#x00394;G</italic> of &#x0223c;4 kJ&#x000b7;mol^&#x02212;1 at 0&#x000b0;C (<xref ref-type="table" rid="tbl1">Table 1</xref>). Thus, the R<italic>^add</italic>CW<italic>^ddd</italic> construct has the highest stability among the R-triplexes of mixed nucleotide sequences studied so far (<xref ref-type="bibr" rid="b10">10</xref>,<xref ref-type="bibr" rid="b11">11</xref>,<xref ref-type="bibr" rid="b13">13</xref>,<xref ref-type="bibr" rid="b17">17</xref>). The increased stability of the DAP*(T&#x000b7;DAP) triplets compared with the A*(T&#x000b7;A) triplets is consistent with the proposed H-bonding scheme for the R-triplet (<xref ref-type="bibr" rid="b5">5</xref>) (<xref ref-type="fig" rid="fig2">Figure 2</xref>).</p></sec><sec><title>Testing the H-bonding scheme of the A*(T&#x000b7;A) triplet with 7-deazaadenine substitutions</title><p>To further test the impact of hydrogen bonds on the stability of the R-triplex, we constructed oligonucleotides R<italic>^a</italic>CW<italic>⁷</italic>, R<italic>^a</italic>CW<italic>⁷⁷</italic> and R<italic>^aaa</italic>CW<italic>⁷⁷⁷</italic> with the 7-deazaadenine (7DAA) substituted for adenine in the Watson strand (<xref ref-type="fig" rid="fig1">Figure 1</xref>). In contrast to DAP, the 7DAA substitutions are expected to destabilize the RCW triplex (<xref ref-type="bibr" rid="b18">18</xref>). The 7DAA has a CH group in the ring position 7 instead of the negatively charged nitrogen N7. Hence, 7DAA cannot be an acceptor of a proton from the 2AP amino group and instead, the CH group would produce a steric clash with the amino group, thereby destabilizing the 2AP*(T&#x000b7;7DAA) triplet shown in <xref ref-type="fig" rid="fig2">Figure 2</xref>. Similarly, if an adenine is positioned in the 5&#x02032;-R-strand opposite to 7DAA (see R<italic>^a</italic>CW<italic>⁷⁷</italic> in <xref ref-type="fig" rid="fig1">Figure 1</xref>), then the A&#x02013;7DAA interaction is also expected to be unfavorable. In this case, the CH&#x000b7;&#x000b7;&#x000b7;HC electrostatic interaction (<xref ref-type="bibr" rid="b18">18</xref>) would be a major repulsion factor (<xref ref-type="fig" rid="fig2">Figure 2</xref>).</p><p>Thermal denaturation data for the R<italic>^a</italic>CW<italic>⁷</italic> triplex are shown in <xref ref-type="fig" rid="fig4">Figure 4</xref> (triangles). R<italic>^a</italic>CW<italic>⁷</italic> melted earlier than the R<italic>^add</italic>CW<italic>^ddd</italic> triplex by 7&#x000b0;C. Furthermore, the slopes of the melting curves were visibly different, corresponding to a significantly lower formation enthalpy and free energy for the R<italic>^a</italic>CW<italic>⁷</italic> triplex (<xref ref-type="table" rid="tbl1">Table 1</xref>). Comparison of the thermodynamic parameters for the R<italic>^a</italic>CW<italic>⁷</italic> triplex and the &#x02018;reference&#x02019; R<italic>^a</italic>CW fold indicates that the R<italic>^a</italic>CW<italic>⁷</italic> is demonstrably less stable than the R<italic>^a</italic>CW (<xref ref-type="table" rid="tbl1">Table 1</xref>), although the former differs from the latter by only one hydrogen bond (<xref ref-type="fig" rid="fig1">Figures 1</xref> and <xref ref-type="fig" rid="fig2">2</xref>).</p><p>The A*(T&#x000b7;A) triplet (<xref ref-type="fig" rid="fig2">Figure 2</xref>) is presumed to form a &#x02018;weak&#x02019; CH&#x000b7;&#x000b7;&#x000b7;N hydrogen bond [reviewed in (<xref ref-type="bibr" rid="b19">19</xref>)]. To test the contribution of such a bond on R-triplex stability we constructed the oligonucleotide R<italic>^a</italic>CW<italic>⁷⁷</italic>, containing two putative A*(T&#x000b7;7DAA) triplets in addition to a &#x02018;standard&#x02019; reporting 2AP*(T&#x000b7;A) triplet (<xref ref-type="fig" rid="fig1">Figure 1</xref>). This fold lacks two CH&#x000b7;&#x000b7;&#x000b7;N hydrogen bonds compared to the reference fold R<italic>^a</italic>CW. The impact of this substitution resulted in a decrease in the <italic>T</italic>_m by 3&#x000b0;C with an associated &#x00394;&#x00394;<italic>H</italic> = 12 &#x000b1; 2 kJ&#x000b7;mol^&#x02212;1 and &#x00394;&#x00394;<italic>G</italic>(0&#x000b0;C) = 1.2 &#x000b1; 0.5 kJ&#x000b7;mol^&#x02212;1 (<xref ref-type="table" rid="tbl1">Table 1</xref> and Supplementary Figure S1). Thus, the &#x02018;weak&#x02019; CH&#x000b7;&#x000b7;&#x000b7;N hydrogen bond contributes noticeably to the R-triplex stability.</p><p>The oligonucleotide R<italic>^aaa</italic>CW<italic>⁷⁷⁷</italic> contains three 2AP*(T&#x000b7;7DAA) triplets, and therefore is expected to lose three hydrogen bonds compared to R<italic>^a</italic>CW (<xref ref-type="fig" rid="fig1">Figures 1</xref> and <xref ref-type="fig" rid="fig2">2</xref>). Fluorescence is not a method of choice for detection of the denaturation profile of R<italic>^aaa</italic>CW<italic>⁷⁷⁷</italic> since the 2AP fluorescence emission would be affected by energy transfer between the three closely positioned 2-aminopurines. For this reason, the thermal denaturation profile of R<italic>^aaa</italic>CW<italic>⁷⁷⁷</italic> was determined by the UV absorption at 260 nm (Supplementary Figure S2; the derived thermodynamic parameters are presented in <xref ref-type="table" rid="tbl1">Table 1</xref>). As expected, the stability of the R<italic>^aaa</italic>CW<italic>⁷⁷⁷</italic> triplex was exceptionally low, showing a <italic>T</italic>_m 6&#x000b0;C lower than that of R<italic>^a</italic>CW with lowered transition enthalpy, entropy and free energy values.</p></sec><sec><title>Additive effect of the hydrogen bonds on the R-triplex stability</title><p>A graphical representation (<xref ref-type="fig" rid="fig5">Figure 5</xref>) of the thermodynamic data (<xref ref-type="table" rid="tbl1">Table 1</xref>) accentuates the energy impact of hydrogen bonds on the R-triplex stability. Both the formation <italic>&#x00394;H</italic>s and <italic>&#x00394;G</italic>s plotted against the predicted number of hydrogen bonds in the R-triplex (relative to the R<italic>^a</italic>CW fold) may be fitted with straight lines within the limits of experimental errors. Such a linear dependence implies that the global triplex conformation at low temperature is similar for all of the oligonucleotides; i.e. in the simplest case, the overall stability of the intramolecular DNA fold would be proportional to the number of stabilizing hydrogen bonds, as has been postulated previously for double- and H-form triple-stranded DNA structures (<xref ref-type="bibr" rid="b20">20</xref>,<xref ref-type="bibr" rid="b21">21</xref>).</p><p>One may deduce the contribution of a single H-bond in the R-triplex from these data. A mismatch that removes a single H-bond from the triplex reduces the formation enthalpy by &#x00394;&#x00394;<italic>H</italic> = 18 &#x000b1; 1.0 kJ&#x000b7;mol^&#x02212;1 and the formation free energy by &#x00394;&#x00394;<italic>G</italic>(0&#x000b0;C) = 1.8 &#x000b1; 0.2 kJ&#x000b7;mol^&#x02212;1 (<xref ref-type="fig" rid="fig5">Figure 5</xref>). As a result, the loss of only three H-bonds in the R<italic>^aaa</italic>CW<italic>⁷⁷⁷</italic> dramatically destabilizes the triplex and leads to a drop in the absolute value of formation enthalpy from 99 &#x000b1; 3 to 45 &#x000b1; 3 kJ&#x000b7;mol^&#x02212;1(54%) and the free energy from 7.1 &#x000b1; 0.5 to 2.3 &#x000b1; 0.7 kJ&#x000b7;mol^&#x02212;1 (67%). The plots presented in <xref ref-type="fig" rid="fig5">Figure 5</xref> predict that mismatches removing 4&#x02013;5 hydrogen bonds (two &#x02018;wrong&#x02019; bases) from a potential 10 nt long R-triplex would practically abolish triplex formation under our experimental conditions and prevent recognition of the duplex by the third strand. For various DNA duplexes, the absolute values of &#x00394;&#x00394;<italic>G</italic>(0&#x000b0;C) per H-bond estimated from the data given in Ref. (<xref ref-type="bibr" rid="b20">20</xref>) vary from 2 to 10 kJ&#x000b7;mol^&#x02212;1. An additional hydrogen bond in the C⁺*(G&#x000b7;C) Hoogsten triplet yielded an additional &#x00394;&#x00394;<italic>G</italic>(25&#x000b0;C) = 5 kJ&#x000b7;mol^&#x02212;1 (<xref ref-type="bibr" rid="b22">22</xref>,<xref ref-type="bibr" rid="b23">23</xref>). Thus, the R-triplex stability estimated here at &#x00394;&#x00394;<italic>G</italic>(0&#x000b0;C) = 1.8 &#x000b1; 0.2 kJ&#x000b7;mol^&#x02212;1 corresponds to the weakest hydrogen bonds observed in nucleic acid base pairs (<xref ref-type="bibr" rid="b24">24</xref>).</p></sec><sec><title>A relatively high contribution of hydrogen bonds to the R-triplex stability</title><p>Unlike the absolute average contribution of H-bonds to the overall DNA structures stability, the relative impact of hydrogen bonds in R-triplex differs significantly from those in duplex and conventional triplex DNA; i.e. the contribution of a single hydrogen bond to the R-triplex formation enthalpy &#x00394;&#x00394;<italic>H</italic>/&#x00394;<italic>H</italic> = 18 kJ&#x000b7;mol^&#x02212;1/99 kJ&#x000b7;mol^&#x02212;1 &#x02248; 0.17 is quite substantial (<xref ref-type="table" rid="tbl1">Table 1</xref>). The relative free energy contribution of an H-bond to the R-triplex stability is even greater, &#x00394;&#x00394;<italic>G</italic>(0&#x000b0;C)/&#x00394;<italic>G</italic>(0&#x000b0;C) = 1.8 kJ&#x000b7;mol^&#x02212;1/7.1 kJ&#x000b7;mol^&#x02212;1 &#x02248; 0.25 (<xref ref-type="table" rid="tbl1">Table 1</xref>). In contrast, the relative impact of an additional hydrogen bond to the free energy as well as to formation enthalpy of the H-form triplex for an 11 nt long pyrimidine sequence is 0.09 (<xref ref-type="bibr" rid="b23">23</xref>). These values are close to the corresponding estimations for DNA duplexes (<xref ref-type="bibr" rid="b21">21</xref>,<xref ref-type="bibr" rid="b24">24</xref>,<xref ref-type="bibr" rid="b25">25</xref>).</p><p>The origin of this difference between the R-triplex and the H-triplex is the appreciably lower overall stability of the R-triplex, whose formation enthalpy is only &#x02212;10 kJ&#x000b7;mol^&#x02212;1 per base contact (<xref ref-type="table" rid="tbl1">Table 1</xref> and <xref ref-type="fig" rid="fig5">Figure 5</xref>), whereas the average formation enthalpy for the H-triplex (<xref ref-type="bibr" rid="b26">26</xref>&#x02013;<xref ref-type="bibr" rid="b28">28</xref>) and DNA duplex (<xref ref-type="bibr" rid="b29">29</xref>&#x02013;<xref ref-type="bibr" rid="b31">31</xref>) are about &#x02212;30 kJ&#x000b7;mol^&#x02212;1. The low stability of the R-triplex is related to its structural features: (i) poor base stacking along the R-strand (<xref ref-type="bibr" rid="b5">5</xref>) and (ii) major &#x02018;sub-grooves&#x02019; geometries, that may be less favorable for interactions with ions and water molecules compared with the H-triplex (<xref ref-type="bibr" rid="b32">32</xref>). At the same time, the R-triplex stability is extremely sensitive to the sequence, as stated previously.</p></sec></sec><sec><title>CONCLUSIONS</title><p><list list-type="roman-lower"><list-item><p>The formation of the R-triplex exclusive of an alternative &#x02018;branch migration&#x02019; structure has been confirmed directly by strand-specific labeling using the fluorescence base analog 2-aminopurine (2AP).</p></list-item><list-item><p>Strand exchange could be promoted by substitution of two adenines in the third R-strand by 2,6-diaminopurine (DAP) due to a greater stability of the DAP&#x000b7;T base pair in comparison with the A&#x000b7;T base pair.</p></list-item><list-item><p>The hydrogen bonding scheme of the previously proposed (<xref ref-type="bibr" rid="b5">5</xref>) A*(T&#x000b7;A) R-type triplet was strongly supported by the results of adenine substitutions that increased (2,6-diaminopurine) or decreased (7-deazaadenine) the number of hydrogen bonds in the fold.</p></list-item><list-item><p>The relative impact of hydrogen bonds on the overall stability of the R-triplex was significant (compared to the stability of a DNA duplex or an H-triplex).</p></list-item><list-item><p>The large impact of a &#x02018;correct&#x02019; base in the third strand opposite the Watson&#x02013;Crick base pair on the thermodynamic properties of the R-triplex may be a powerful mechanism for ensuring proper recognition of the nucleic acid duplex by the single strand.</p></list-item></list></p></sec><sec><title>SUPPLEMENTARY DATA</title><p>Supplementary Data are available at NAR Online.</p></sec> <sec sec-type="supplementary-material"> <title>Supplementary Material</title> <supplementary-material id="PMC_1" content-type="local-data"> <caption> <title>[Supplementary Material]</title> </caption> <media mimetype="text" mime-subtype="html" xlink:href="nar_34_11_3239__index.html"/> <media xlink:role="associated-file" mimetype="application" mime-subtype="pdf" xlink:href="nar_34_11_3239__1.pdf"/> </supplementary-material> </sec>

Information-driven protein&#x02013;DNA docking using HADDOCK: it is a matter of flexibility

<p>Intrinsic flexibility of DNA has hampered the development of efficient protein&#x02212;DNA docking methods. In this study we extend HADDOCK (High Ambiguity Driven DOCKing) [C. Dominguez, R. Boelens and A. M. J. J. Bonvin (2003) <italic>J. Am. Chem. Soc.</italic> <bold>125</bold>, 1731&#x02013;1737] to explicitly deal with DNA flexibility. HADDOCK uses non-structural experimental data to drive the docking during a rigid-body energy minimization, and semi-flexible and water refinement stages. The latter allow for flexibility of all DNA nucleotides and the residues of the protein at the predicted interface. We evaluated our approach on the monomeric repressor&#x02212;DNA complexes formed by bacteriophage 434 Cro, the <italic>Escherichia coli</italic> Lac headpiece and bacteriophage P22 Arc. Starting from unbound proteins and canonical B-DNA we correctly predict the correct spatial disposition of the complexes and the specific conformation of the DNA in the published complexes. This information is subsequently used to generate a library of pre-bent and twisted DNA structures that served as input for a second docking round. The resulting top ranking solutions exhibit high similarity to the published complexes in terms of root mean square deviations, intermolecular contacts and DNA conformation. Our two-stage docking method is thus able to successfully predict protein&#x02212;DNA complexes from unbound constituents using non-structural experimental data to drive the docking.</p>

<contrib contrib-type="author"><name><surname>van Dijk</surname><given-names>Marc</given-names></name></contrib><contrib contrib-type="author"><name><surname>van Dijk</surname><given-names>Aalt D. J.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Hsu</surname><given-names>Victor</given-names></name></contrib><contrib contrib-type="author"><name><surname>Boelens</surname><given-names>Rolf</given-names></name></contrib><contrib contrib-type="author"><name><surname>Bonvin</surname><given-names>Alexandre M. J. J.</given-names></name><xref ref-type="corresp" rid="cor1">*</xref></contrib><aff><institution>NMR Spectroscopy Research Group, Bijvoet Center for Biomolecular Research, Faculty of Sciences, Utrecht University</institution><addr-line>The Netherlands</addr-line></aff><aff id="au1">¹<institution>Department of Biochemistry and Biophysics, Oregon State University</institution><addr-line>Corvallis, USA</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p>Computational docking has proven to be a valuable tool in the study of biomolecular complexes (<xref ref-type="bibr" rid="b1">1</xref>,<xref ref-type="bibr" rid="b2">2</xref>). In particular, the field of &#x02018;<italic>ab initio</italic>&#x02019; protein&#x02212;protein docking has made considerable progress as illustrated by recent results from the community-wide CAPRI experiment [critical assessment of predicted interactions (<xref ref-type="bibr" rid="b3">3</xref>,<xref ref-type="bibr" rid="b4">4</xref>)]. However, where this field has in many ways matured, the development of docking methods to model protein&#x02212;DNA interactions has lagged behind. These play an important role in recognition and gene expression (<xref ref-type="bibr" rid="b5">5</xref>). Powerful protein&#x02212;DNA docking methods would thus be of great benefit for their study. However, two particular problems have hampered the development of efficient docking methods: the sparsity of the information to define the DNA-binding interface and the inherent flexibility of DNA. For protein&#x02212;protein docking there is often enough information available (e.g. from sequence, conservation or biological knowledge) to identify the interaction surfaces of the docking partners. This information can be used to drive the docking (<xref ref-type="bibr" rid="b6">6</xref>) and limit the conformational space to be searched. Identification of the interaction surface on DNA is less straightforward than on proteins. There is still no general recognition code and the global conformation of the DNA can play an important role in modulating the eventual interaction surface (<xref ref-type="bibr" rid="b7">7</xref>). DNA indeed often exhibits large conformational changes upon binding to a protein, which can greatly alter the shape of the interaction surface. Owing to this, the total conformational space that needs to be searched in order to find favourable conformations becomes even larger. Flexibility in DNA can be separated into global and local components (<xref ref-type="bibr" rid="b8">8</xref>). Global flexibility is constrained to two primary motions: bending and twisting. It results from a combination of conformational changes in the flexible base pairs and sugar-phosphate backbone. Allowing for global and local flexibility in DNA during docking while maintaining the relevant conformation is a major challenge in protein&#x02212;DNA docking.</p><p>In the last few years several methods have been developed to solve one or both of these problems, each with varying degrees of success. The program FTDOCK (<xref ref-type="bibr" rid="b9">9</xref>) has been used to perform a large search through conformational space by rotating and translating the protein along the DNA while evaluating shape and electrostatic complementarity; an approximation of flexibility was achieved by allowing some degree of overlap between protein and DNA in the scoring. In another approach, a library of pre-bent DNA structures was used to minimize the search through DNA conformational space (<xref ref-type="bibr" rid="b10">10</xref>); a selection was made based on structures that could be electrostatically preorientated in the potential of the protein and these were rotated and translated with respect to the protein. To account for some degree of local flexibility protein side chains and DNA base pairs were allowed to move in two separate refinement stages. Knegtel <italic>et al</italic>. (<xref ref-type="bibr" rid="b11">11</xref>) developed MONTY which uses a Monte Carlo search allowing for flexibility in both protein and DNA and experimentally determined contacts to drive the docking. The initial position of the protein in the predicted complex should, however, not deviate too much from that of the actual complex; small deviations in the position of the protein with respect to the interaction interface of the DNA resulted in DNA curling around the protein. Tzou and Hwang (<xref ref-type="bibr" rid="b12">12</xref>) modelled the CAP-DNA and Rep-DNA systems from the repressors in their bound conformation and canonical B-DNA in a series of molecular mechanics and dynamics simulations using distance restraints derived from a statistical analysis of homologous protein&#x02013;DNA complexes. This method successfully introduced DNA bending and local opening of the major groove. All of these docking procedures were able to make predictions that were representative of the published complexes in terms of spatial disposition. Only a few methods allowed for flexibility of the DNA and protein side chains during the docking. They, however, required extensive knowledge to position the two components relative to each other (<xref ref-type="bibr" rid="b12">12</xref>) and problems were encountered in the absence of such information (<xref ref-type="bibr" rid="b11">11</xref>).</p><p>Here we demonstrate that both global and local DNA flexibility can successfully be accounted for in protein&#x02212;DNA modelling using HADDOCK (High Ambiguity Driven DOCKing) (<xref ref-type="bibr" rid="b13">13</xref>), a computational docking approach developed in our group. HADDOCK makes use of available experimental and bioinformatics data to drive the docking process (<xref ref-type="bibr" rid="b14">14</xref>). Its successful use in NMR-based structure calculations of protein&#x02013;DNA and protein&#x02013;RNA complexes has been shown previously (<xref ref-type="bibr" rid="b15">15</xref>&#x02013;<xref ref-type="bibr" rid="b18">18</xref>). Global and local DNA flexibility is introduced in the docking by allowing the DNA sugar-phosphate backbone and DNA base pairs to sample conformations during a semi-flexible refinement stage and by starting the docking from a library of pre-generated DNA structures representing various degrees of conformational flexibility. The latter allows for the sampling of a larger conformational space. Flexibility in the protein is introduced as described previously (<xref ref-type="bibr" rid="b13">13</xref>), first along the side chains at the interface and then for both backbone and side chains. We demonstrate here the feasibility of this approach for three repressor complexes in their monomeric form: Cro from bacteriophage 434 (<xref ref-type="bibr" rid="b19">19</xref>), the Lac headpiece of <italic>Escherichia coli</italic> (<xref ref-type="bibr" rid="b20">20</xref>) and Arc from bacteriophage P22 (<xref ref-type="bibr" rid="b21">21</xref>). The first two recognize the DNA major groove via an &#x003b1;-helix/turn/&#x003b1;-helix motif and the last one via a two-stranded antiparallel &#x003b2;-sheet motif. To drive the docking we make use of mutation data, sequence/structure conservation, DNA footprinting and ethylation interference data. We show that our approach is successful in predicting protein&#x02013;DNA complexes from unbound constituents by accounting for both global and local DNA flexibility during the docking.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><sec><title>Initial structures for protein and DNA</title><p>The coordinate files of all proteins and protein&#x02013;DNA complexes were obtained from the RCSB Protein Data Bank (PDB) (<xref ref-type="bibr" rid="b22">22</xref>). The PDB entry codes of the respective complexes and their unbound components are as follows: 3CRO, crystal structures of the Cro/O1R complex (<xref ref-type="bibr" rid="b19">19</xref>); 1ZUG, NMR ensemble of the unbound Cro monomer (<xref ref-type="bibr" rid="b23">23</xref>); 1LCC, NMR ensemble of the Lac/O1 complex (<xref ref-type="bibr" rid="b24">24</xref>); 1LQC, NMR ensemble of the unbound monomer of the Lac headpiece (<xref ref-type="bibr" rid="b25">25</xref>); 1BDT, crystal structure of the Arc/DNA complex (<xref ref-type="bibr" rid="b26">26</xref>) and 1ARQ, NMR ensemble of the unbound Arc monomer (<xref ref-type="bibr" rid="b27">27</xref>). The monomeric reference structures for Cro and Arc (right halfside) were extracted from the dimeric PDB structures.</p><p>Models of canonical B-DNA were constructed with the nucleic acid analysis and rebuilding program 3DNA (<xref ref-type="bibr" rid="b28">28</xref>), using the fiber models provided by Chandrasekaran and Arnott (<xref ref-type="bibr" rid="b29">29</xref>). All hydrogens were added according to the standard assigning scheme of CNS followed by a short energy minimization step during the initiation stage in HADDOCK. Base pair and base pair step parameters of the resulting type BII B-DNA starting structures are shown in <xref ref-type="table" rid="tbl1">Table 1</xref>. The DNA backbone torsion angles are &#x003b1; = 309&#x000b0;, &#x003b2; = 159&#x000b0;, &#x003b3; = 37&#x000b0;, &#x003b4; = 146&#x000b0;, &#x0025b; = 218&#x000b0;, &#x003b6; = 191&#x000b0;, &#x003c7; = 260&#x000b0; and the sugar pseudo-rotation phase angle (<italic>P</italic>) = 155&#x000b0;, the sugar pucker was thus in the C2&#x02032;-endo conformation.</p><p>Custom DNA libraries for the three operator sequences were generated by manipulation of the base pair step parameters of their respective B-DNA structures using 3DNA. The introduction of curvature was accomplished by changing the value of roll using the following equation (<xref ref-type="bibr" rid="b30">30</xref>): <disp-formula><mml:math id="M1"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mi>n</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mi>k</mml:mi><mml:mtext>&#x000a0;</mml:mtext><mml:mi mathvariant="italic">cos</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mi>T</mml:mi><mml:mo>*</mml:mo><mml:mi>n</mml:mi><mml:mo>&#x003b8;</mml:mo></mml:mrow><mml:mo>)</mml:mo><mml:mo>,</mml:mo></mml:mrow></mml:mrow></mml:math></disp-formula> where <italic>R_n</italic> is the roll value for each base pair step in one helical turn, <italic>n</italic> is the number of base pair steps in one helical turn, <italic>k</italic> is the average curvature for each base pair step in one helical turn and <italic>T</italic> is the value for twist. The direction of the curvature in Cartesian space can be controlled by changing the phase (&#x003b8;) of the cosine function. The positive linear relationship between the value of the slide parameter and the width of the major groove was used to adjust the major groove width.</p></sec><sec><title>Restraints used in the docking</title><p><italic>Ambiguous Interaction Restraints (AIR)</italic> (<xref ref-type="table" rid="tbl2">Table 2</xref>): All active residues have a relative solvent accessibility &#x0003e;50% as calculated with NACCESS (<xref ref-type="bibr" rid="b31">31</xref>). Residues located in the predicted interaction interface or in a continuous stretch of residues near the predicted interaction interface for which no information is available were defined as passive. AIRs for the protein were defined based on sequence conservation [HSSP (<xref ref-type="bibr" rid="b32">32</xref>)] and mutation data. For the DNA only active residues were defined. The recognition sequences of the operators have been determined using DNA-footprinting methods before the experimental structures of the actual complexes became available. This information was used in our docking procedure to define interaction restraints. For those bases shown to be involved in specific interactions with the repressor, only atoms able to interact by hydrogen-bond or non-bonded interactions were defined. Based on ethylation interference experiments, only the oxygen atoms of phosphate groups shown to interact with the repressor were defined as active.</p><p><italic>DNA restraints</italic>: In order to preserve the helical conformation of DNA the following restraints were defined: planarity restraints for the purine and pyrimidine rings were introduced, and the sugar pucker was restrained to the C2&#x02032;-endo conformation. Watson&#x02212;Crick base pairs were defined and hydrogen bond lengths of the input structure (either the initial starting DNA conformation or the conformation obtained after semi-flexible refinement prior to water refinement) were measured and restricted to &#x000b1;0.05 &#x0212b;. In a similar way the dihedral angles of the sugar-phosphate backbone of the input structure (inp) were measured and used as restraints. (Restricted to &#x003b1; = &#x003b1;_inp &#x000b1; 10&#x000b0;, &#x003b2; = &#x003b2;_inp &#x000b1; 40&#x000b0;, &#x003b3; = &#x003b3;_inp &#x000b1; 20&#x000b0;, &#x003b4; = &#x003b4;_inp &#x000b1; 50&#x000b0;, &#x0025b; = &#x0025b;_inp &#x000b1; 10&#x000b0; and &#x003b6; = &#x003b6;_inp &#x000b1; 50&#x000b0;).</p></sec><sec><title>Docking protocol</title><p>Our docking protocol consists of (i) rigid-body docking, (ii) semi-flexible refinement stage and (iii) final refinement in explicit solvent.</p><p><italic>Rigid-body docking</italic>. A total of 100 structures were generated for each protein&#x02212;DNA combination from the ensembles of starting structures. Each docking attempt was performed 10 times and the solution with the lowest HADDOCK score was kept. For each protein we used an ensemble of 10 NMR structures; thus 1000 rigid-body docking solutions were generated for each of the three canonical B-DNA docking runs and 5000 structures were generated for each of the DNA library docking runs (5 different pre-bent and twisted DNA structures and 10 protein structures resulting in 50 different combinations). For the docking of the protein and DNA in their bound conformation a total of 1000 structures were generated. Systematic sampling of 180&#x000b0; rotated solutions was used in the rigid-body docking stage to minimize the occurrence of false positives (principles described in Results). This basically doubled the number of docking trials bringing the total to 20 000 and 100 000 evaluations for docking from canonical B-DNA and DNA libraries, respectively.</p><p><italic>Semi-flexible refinement</italic>. Of all structures generated in the rigid-body docking stage the best 20% based on the HADDOCK score were further refined in the semi-flexible refinement stage consisting of three parts: rigid-body torsion angle dynamics (500 MD steps at 2000 K and 500 MD cooling steps to 500 K with a 8 fs time step), semi-flexible simulated annealing stage (1000 MD steps from 1000 to 50 K with 4 fs time steps) with the side chains of the protein residues at the interface and the complete DNA (excluding terminal base pairs) allowed to move and a final semi-flexible simulated annealing stage (1000 MD steps from 300 to 50 K with 2 fs time steps) with both side chains and backbone of the protein residues at the interface and the complete DNA (excluding terminal base pairs) allowed to move.</p><p><italic>Water refinement</italic>. This final stage consists of a gentle refinement (100 MD heating steps at 100, 200 and 300 K followed by 750 sampling steps at 300 K and 500 MD cooling steps at 300, 200 and 100 K all with 2 fs time steps) in an 8 &#x000c5; shell of TIP3P water molecules (<xref ref-type="bibr" rid="b33">33</xref>).</p><p>Semi-flexible segments for the proteins were defined as residues 7&#x02013;20, 24&#x02013;37 for Cro, residues 6&#x02013;30, 50&#x02013;56 for Lac and residues 1&#x02013;17, 54&#x02013;70 for Arc. In all cases the complete DNA, excluding the terminal base pairs, were defined as semi-flexible.</p></sec><sec><title>Scoring</title><p>A HADDOCK score is defined to rank the structures after each docking stage. It is a weighted sum of intermolecular electrostatic (Elec), van der Waals (vdW), desolvation (Dsolv) and AIR energies, and a buried surface area (BSA) term: rigid-body score = 1.0 * Elec + 1.0 * vdW &#x02212; 0.05 * BSA + 1.0 * Dsolv + 1.0 * AIR, final score = 1.0 * Elec + 1.0 * vdW + 1.0 * Dsolv + 1.0 * AIR. A cluster analysis was performed on the final docking solutions using a minimum cluster size of 4. The cut-off for clustering was manually determined for each docking run. The root mean square deviation (r.m.s.d.) matrix was calculated over the backbone atoms of the interface residues of the DNA after fitting on the interface residues of the protein. Final structures within a cluster were selected according to their summed base pair and base pair step deformation energies and the conformation of the helix (classified as B-DNA). Deformation energies were calculated with an extension script of 3DNA (provided by Marc Parisien, University of Montreal, Canada) using the statistical population preferences as determined by Olson <italic>et al</italic>. (<xref ref-type="bibr" rid="b34">34</xref>) and Lankas <italic>et al</italic>. (<xref ref-type="bibr" rid="b35">35</xref>).</p><p>Default HADDOCK (version 2.0_devel) parameters were used except for the dielectric constant (epsilon) that was set to 78 for the vacuum part of the protocol. To speed up calculations, non-polar hydrogens were omitted. Inter- and intramolecular energies were evaluated using full electrostatic and van der Waals energy terms with an 8.5 &#x0212b; distance cut-off. OPLSX non-bonded parameters from the parallhdg5.3.pro parameter file (<xref ref-type="bibr" rid="b36">36</xref>) were used for the protein. Topology and linkage parameter files for the DNA were taken from the CNS (<xref ref-type="bibr" rid="b37">37</xref>) distribution (dna-rna-allatom.top and dna-rna-allatom.param respectively). The HADDOCK package is freely available to academic users (<ext-link ext-link-type="uri" xlink:href="http://www.nmr.chem.uu.nl/haddock"/>).</p></sec><sec><title>Analysis</title><p>The r.m.s.d. values of the complexes were calculated using ProFit (A. C. R. Martin, <ext-link ext-link-type="uri" xlink:href="www.bioinf.org.uk/software/profit"/>) All heavy atoms were used to calculate the r.m.s.d. of the total complex, of the DNA and of the interface. The interface was composed of residues 15&#x02013;44/3&#x02013;7, 31&#x02013;37 of Cro/O1R; 6&#x02013;32/4&#x02013;10, 13&#x02013;19 of Lac/O1; and 8&#x02013;36, 61&#x02013;89/1&#x02013;9, 13&#x02013;21 of Arc/repressor. The backbone r.m.s.d. was calculated using all P and C&#x003b1; atoms of the complex. Residues in the flexible termini of the protein (having either high B-factors in the X-ray structures or poorly defined in the NMR reference structures) were left out of the calculation. Intermolecular contacts were evaluated using LIGPLOT (<xref ref-type="bibr" rid="b38">38</xref>) using a 5 &#x0212b; cut-off. The fraction of native contacts (Fnat) is defined as the number of native intermolecular contacts on a nucleotide-residue basis (hydrogen-bonded and non-bonded) identified in a docking solution divided by the total number of contacts in the reference structure. Values for base pair and base pair step parameters as well as torsion angles for the sugar-phosphate backbone and the sugar pucker were obtained using the program 3DNA (<xref ref-type="bibr" rid="b28">28</xref>). The overall bend-angle of the DNA was calculated using CURVES (<xref ref-type="bibr" rid="b39">39</xref>).</p></sec><sec><title>Hardware</title><p>HADDOCK docking runs were performed on a Transtec (Transtec AG, Tubingen, Germany) computer cluster operating with 32, 2.0 GHz, 64 bit Opteron processors. As a measure of CPU requirements, one complete run starting with 1000 structures in the rigid-body docking stage could be performed in &#x0223c;2 h on 32 processors.</p></sec></sec><sec><title>RESULTS</title><sec><title>Bound rigid-body docking</title><p>The use of readily available biochemical and/or biophysical information can alleviate the lack of a general recognition code for protein&#x02013;DNA interactions. HADDOCK uses this information encoded as AIRs (<xref ref-type="bibr" rid="b13">13</xref>) to drive the docking; this reduces the necessary search through interaction space and increases the fraction of unique solutions. In the definition of AIRs we distinguish between active and passive residues. Active residues are defined as those important for the interaction based on conservation [HSSP (<xref ref-type="bibr" rid="b32">32</xref>)], mutation or ethylation interference data or any other appropriate experimental data. Passive residues are defined as the solvent-accessible neighbours of active residues (<xref ref-type="table" rid="tbl2">Table 2</xref>).</p><p>We first evaluated the use of AIRs in protein&#x02212;DNA docking for the three selected complexes by bound docking (i.e. the reconstruction of the complexes from their separate components). Since the molecules are already in their bound conformation no flexible segments were defined and only rigid-body docking was performed. The best docking solutions for each of the Lac, Arc and Cro repressor in complex with their operators exhibit high similarity with the published complexes based on r.m.s.d. values and intermolecular contacts (<xref ref-type="table" rid="tbl3">Table 3</xref>); all base-specific intermolecular contacts are recovered.</p><p>In the biologically relevant complexes the repressors are bound as dimers that are symmetrically oriented on the two recognition sites of the operator. In this study we use the repressors in their monomeric form (in this form the Arc repressor is a symmetrical dimer). Symmetry in the AIR set and in the shape of the protein&#x02013;DNA interaction surface can result in false positives: these are structures with a favourable HADDOCK score (weighted sum of several energy terms, see Materials and Methods) but with one of the two components rotated 180&#x000b0; with respect to the published complex. To minimize the occurrence of false positives 180&#x000b0; rotated solutions were systematically sampled during the rigid-body docking stage. For this, a 180&#x000b0; rotation around a vector defined by the centres of masses of the interfaces of the protein and DNA was applied and the resulting conformation again minimized. The solution with the lowest HADDOCK score was kept. Using this approach the amount of false positives after the rigid-body docking stage was reduced from &#x0223c;70 to &#x0223c;40%. In subsequent unbound docking runs including flexibility we selected the best 20% of all solutions from the rigid-body docking stage based on their HADDOCK score. Owing to the sampling of 180&#x000b0; rotations this subset contained no false positives for the Cro and Lac repressor/operator complexes (<xref ref-type="fig" rid="fig1">Figure 1</xref>). Because of the intrinsic symmetry of the Arc repressor, 180&#x000b0; rotated symmetrical solutions are similar and can thus not be distinguished. Therefore the problem of rotational false positives does not apply to the Arc repressor. In unbound docking false positives were obtained that correspond to shifted false positives. These are solutions in which the repressor is shifted 1 or 2 bp upstream or downstream of the true interaction surface on the DNA (<xref ref-type="fig" rid="fig1">Figure 1</xref>).</p></sec><sec><title>Unbound semi-flexible docking to B-form DNA</title><p>We used the AIR sets to dock an ensemble of NMR structures of the unbound repressors to canonical B-DNA (chosen for it's biological relevance). In contrast to the previous bound docking runs in which only rigid-body docking was performed, we now included flexibility in a semi-flexible refinement stage: side chains and backbone of the protein at the predicted interface and the entire DNA were allowed to sample additional conformations. A set of restraints was imposed on the DNA that allowed for local flexibility but preserved the overall helical conformation (Materials and Methods). The final refined structures were clustered based on their pairwise r.m.s.d. matrix. The best cluster was selected based on the HADDOCK score.</p><p>The solutions in the selected clusters appeared to be very similar with respect to the protein and the spatial disposition of the complex but less similar on the level of the DNA conformation. An analysis of the base pair and base pair step parameters of the DNA in the selected clusters revealed a higher variation in buckle, propeller, roll and tilt than in other parameters (<xref ref-type="table" rid="tbl1">Table 1</xref>). Previous studies have also observed a larger variation for these parameters in both free and bound DNA when it is bending and twisting (<xref ref-type="bibr" rid="b5">5</xref>,<xref ref-type="bibr" rid="b8">8</xref>,<xref ref-type="bibr" rid="b34">34</xref>,<xref ref-type="bibr" rid="b40">40</xref>,<xref ref-type="bibr" rid="b41">41</xref>). This is not surprising as buckle, propeller, roll and tilt parameters are less restricted by Watson&#x02212;Crick hydrogen bonds and the conformation of the sugar-phosphate backbone, than is the case with the other parameters. However, their large variation occasionally resulted in an overall loss of B-DNA conformation in the docking solutions as assessed by 3DNA (<xref ref-type="bibr" rid="b28">28</xref>). These solutions, however, did not have worse HADDOCK scores than solutions with a smaller variation in the noted parameters. They could, however, in most cases be distinguished by their higher DNA deformation energy. For this we calculated the combined base pair and base pair step deformation energy for every solution in the selected cluster and ranked them according to this energy term (Materials and Methods). The ranked solutions were checked on having a general B-DNA conformation and the best five were selected. This procedure proved successful in selecting solutions that are in better agreement to the published complexes in terms of r.m.s.d. values (<xref ref-type="fig" rid="fig1">Figure 1</xref>).</p><p>To assess the effect of flexibility on the docking we compared the top ranking solutions after the semi-flexible refinement stage with their initial conformation after rigid-body docking: the results show a clear improvement in r.m.s.d. from the published structure of the complex and fraction of native contacts (<xref ref-type="table" rid="tbl3">Table 3</xref>). Analysis of the DNA revealed that the backbone torsion angles were all located in the most populated regions as derived from a statistical analysis of non-complexed DNA structures (<xref ref-type="bibr" rid="b42">42</xref>,<xref ref-type="bibr" rid="b43">43</xref>) (data not shown). Base pair buckle, propeller, tilt and roll parameters, which are at the origin of overall DNA bending and twisting, showed larger differences between the published complexes and the rigid-body docking solutions than after introduction of flexibility (<xref ref-type="table" rid="tbl1">Table 1</xref>). Base pair opening, stagger, stretch and shear parameters and base pair step twist, slide and shift parameters showed little differences. In all three complexes the DNA is slightly bent towards the protein. In this respect tilt rotation is reported to be both statistically and energetically less favourable than roll rotation (<xref ref-type="bibr" rid="b44">44</xref>&#x02013;<xref ref-type="bibr" rid="b47">47</xref>). This relationship is observed in the published complexes and the top ranking docking solutions as they show smaller variations in tilt than in roll. Statistical analysis of crystal structures has revealed that a positive change in roll is often accompanied with unwinding and negative slide (<xref ref-type="bibr" rid="b46">46</xref>,<xref ref-type="bibr" rid="b48">48</xref>,<xref ref-type="bibr" rid="b49">49</xref>). In our best solutions we also witness that roll is negatively correlated with both twist and slide (<xref ref-type="table" rid="tbl1">Table 1</xref>); more precisely, twist values &#x0003c;36&#x000b0; are often accompanied with negative sliding in bent DNA. This relation is observed at the interface of the top ranking docking solutions (the central 4 bp steps in <xref ref-type="fig" rid="fig2">Figure 2D&#x02013;I</xref>). On a global level the distribution of major groove widths over the different base pair steps followed a trend similar to the published complexes (<xref ref-type="fig" rid="fig2">Figure 2A&#x02013;C</xref>).</p></sec><sec><title>Unbound docking from custom-build DNA libraries</title><p>The results above show that the introduction of flexibility results in the prediction of a more native-like complex in comparison with rigid-body docking. To account for even larger DNA conformational changes we explored the possibility of using a library of pre-bent and twisted DNA structures as input structures for the docking procedure. Although the DNA in the best clusters of the flexible docking runs starting from canonical B-DNA showed variation on a local level (e.g. buckle, propeller, roll and tilt parameters) the global conformation of all solutions was quite similar. Analysis of the resulting DNA conformations provided information in the form of bend angles and the width of the major groove, which was used to construct custom DNA libraries. For the Cro/O1R complex the major groove width increased from 11.6 &#x0212b; (canonical B-DNA) to 12.5 &#x000b1; 0.5 &#x0212b; and the DNA adopted a curve towards the protein of 9.4 &#x000b1; 3.6&#x000b0;. For the Lac and Arc repressors in complex with their operator similar events occur, resulting in major groove widths of 11.3 &#x000b1; 0.4 and 12.2 &#x000b1; 0.8 &#x0212b; and curves towards the protein of 11.3 &#x000b1; 3.8 and 12.9 &#x000b1; 5.2&#x000b0;, respectively. Based on this information we constructed for each operator five DNA structures that sample values around the averaged major grooves widths and bend angles from the previous docking runs. Docking from these libraries using the flexible protocol described above resulted in solutions with twist and slide parameters as well as major groove widths in better agreement with those of the published complexes (<xref ref-type="fig" rid="fig2">Figure 2</xref>). The overall results (<xref ref-type="table" rid="tbl3">Table 3</xref>) demonstrate that the use of a custom library of pre-bent and twisted structures improves the prediction structures of the complexes as assessed by r.m.s.d. values, intermolecular contacts and DNA conformation. Only for the Arc repressor/operator complex did the use of a custom DNA library not result in a significant improvement compared to canonical B-DNA docking. The best docking solutions superimposed onto their reference structure are presented in <xref ref-type="fig" rid="fig3">Figure 3</xref>.</p></sec></sec><sec><title>DISCUSSION</title><p>Our modelling of protein&#x02212;DNA complexes is based on AIRs to drive the docking process. These are essential in successfully positioning the protein at the interface of the DNA and, together with flexibility, influence DNA bending in the semi-flexible refinement stage. We used a limited number of easily obtainable experimental data to define the restraints. These were nevertheless sufficient to accurately predict the conformation of the DNA in the complex when starting from canonical B-DNA. This information subsequently allowed us to refine our models by performing docking from a custom-built DNA library instead of canonical B-DNA. This two-stage docking approach significantly improves the conformation of the DNA in the resulting complexes; the protein, however, is less affected and its conformation remains close to the conformation of the respective starting unbound structures.</p><p>In this study we did not investigate the effects of a variable number or type of restraints on the docking results. From analogous protein&#x02212;protein docking studies it is known that the amount of or the ambiguity in the data can influence the reproducibility of the docking. HADDOCK allows the random deletion of a fraction of the restraints for each docking trial to account for errors in their definition, an approach that has proved successful in the past (<xref ref-type="bibr" rid="b14">14</xref>). This option was not used in this study. The AIRs were defined with an upper distance limit of 2.0 &#x0212b; that can affect the packing of the docking solutions. For the Lac/O1 and Arc/operator complexes the BSA was comparable to that of the reference (1496 &#x000b1; 103 &#x0212b; versus 1560 &#x0212b; and 1990 &#x000b1; 155 &#x0212b; versus 2072 &#x0212b;, respectively). For the Cro/O1R complex the BSA of the top ranking solutions was larger than that of the reference (1694 &#x000b1; 52 &#x0212b; versus 1453 &#x0212b;). The tighter packing might contribute to the significant increase in the fraction of native contacts (<xref ref-type="table" rid="tbl3">Table 3</xref>) for the Cro/O1R complex, with respect to the other two test systems.</p><p>We have demonstrated that the use of readily available non-structural experimental data and the incorporation of DNA flexibility during the docking significantly improve repressor&#x02212;DNA complex prediction in comparison to rigid-body docking. The method successfully predicted global conformational changes taking place in the DNA upon complexation. The information extracted from these results is sufficient to refine the models by starting a second docking round from custom-built DNA libraries of pre-bent and twisted structures.</p><p>The flexible protein&#x02212;DNA docking approach described in this paper has biological implications since it can benefit studies of protein&#x02013;DNA interactions at several levels. It can be used to generate models of protein&#x02013;DNA complexes when the structure of the unbound protein is known and suitable experimental data are available. It is also applicable to study the effects of mutations or different operator sequences on complex formation. In addition, it can assist in experimental structural studies: it can, for example, speed up structure determination of protein&#x02013;DNA complexes by NMR by providing initial models to guide the tedious NMR analysis and assignment process. In summary, by allowing the inclusion of a large variety of experimental and/or bioinformatics data, together with a flexible description of the DNA, the proposed docking approach should be a useful tool in structural studies of protein&#x02013;DNA and even protein&#x02013;RNA interactions provided suitable RNA models are available for the latter.</p></sec>

Metabolic regulation of <italic>ApoB</italic> mRNA editing is associated with phosphorylation of APOBEC-1 complementation factor

<p>Apolipoprotein B (<italic>apoB</italic>) mRNA editing is a nuclear event that minimally requires the RNA substrate, APOBEC-1 and APOBEC-1 Complementation Factor (ACF). The co-localization of these macro-molecules within the nucleus and the modulation of hepatic <italic>apoB</italic> mRNA editing activity have been described following a variety of metabolic perturbations, but the mechanism that regulates editosome assembly is unknown. APOBEC-1 was effectively co-immunoprecipitated with ACF from nuclear, but not cytoplasmic extracts. Moreover, alkaline phosphatase treatment of nuclear extracts reduced the amount of APOBEC-1 co-immunoprecipitated with ACF and inhibited <italic>in vitro</italic> editing activity. Ethanol stimulated <italic>apoB</italic> mRNA editing was associated with a 2- to 3-fold increase in ACF phosphorylation relative to that in control primary hepatocytes. Significantly, phosphorylated ACF was restricted to nuclear extracts where it co-sedimented with 27S editing competent complexes. Two-dimensional phosphoamino acid analysis of ACF immunopurified from hepatocyte nuclear extracts demonstrated phosphorylation of serine residues that was increased by ethanol treatment. Inhibition of protein phosphatase I, but not PPIIA or IIB, stimulated <italic>apoB</italic> mRNA editing activity coincident with enhanced ACF phosphorylation <italic>in vivo</italic>. These data demonstrate that ACF is a metabolically regulated phosphoprotein and suggest that this post-translational modification increases hepatic <italic>apoB</italic> mRNA editing activity by enhancing ACF nuclear localization/retention, facilitating the interaction of ACF with APOBEC-1 and thereby increasing the probability of editosome assembly and activity.</p>

<contrib contrib-type="author"><name><surname>Lehmann</surname><given-names>David M.</given-names></name><xref rid="au3" ref-type="aff">3</xref><xref rid="au4" ref-type="aff">4</xref></contrib><contrib contrib-type="author"><name><surname>Galloway</surname><given-names>Chad A.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Sowden</surname><given-names>Mark P.</given-names></name><xref rid="au1" ref-type="aff">1</xref><xref rid="au2" ref-type="aff">2</xref></contrib><contrib contrib-type="author"><name><surname>Smith</surname><given-names>Harold C.</given-names></name><xref rid="au1" ref-type="aff">1</xref><xref rid="au2" ref-type="aff">2</xref><xref rid="au3" ref-type="aff">3</xref><xref rid="au4" ref-type="aff">4</xref><xref rid="au5" ref-type="aff">5</xref><xref ref-type="corresp" rid="cor1">*</xref></contrib><aff id="au1">¹<institution>Department of Biochemistry and Biophysics, University of Rochester</institution><addr-line>Rochester, NY 14642, USA</addr-line></aff><aff id="au2">²<institution>Department of Pathology and Laboratory Medicine, University of Rochester</institution><addr-line>Rochester, NY 14642, USA</addr-line></aff><aff id="au3">³<institution>Department of Toxicology, University of Rochester</institution><addr-line>Rochester, NY 14642, USA</addr-line></aff><aff id="au4">⁴<institution>The Environmental Health Sciences Center, University of Rochester</institution><addr-line>Rochester, NY 14642, USA</addr-line></aff><aff id="au5">⁵<institution>James P. Wilmot Cancer Center, University of Rochester</institution><addr-line>Rochester, NY 14642, USA</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p><italic>ApoB</italic> mRNA editing involves the site-specific deamination of cytidine 6666 to uridine within a glutamine codon (CAA) thereby creating an in-frame translation stop codon (<xref ref-type="bibr" rid="b1">1</xref>). Consequently, two apoB protein variants are expressed, full-length apoB100 and the truncated protein apoB48, both of which participate in lipid transport, but with markedly different roles as atherogenic risk factors (<xref ref-type="bibr" rid="b1">1</xref>). Minimally, <italic>apoB</italic> mRNA editing requires the cytidine deaminase APOBEC-1 as a homodimer (<xref ref-type="bibr" rid="b2">2</xref>&#x02013;<xref ref-type="bibr" rid="b5">5</xref>), APOBEC-1 Complementation Factor (ACF) (<xref ref-type="bibr" rid="b6">6</xref>&#x02013;<xref ref-type="bibr" rid="b9">9</xref>) and the tripartite editing motif within the mRNA substrate (<xref ref-type="bibr" rid="b10">10</xref>&#x02013;<xref ref-type="bibr" rid="b12">12</xref>). ACF is the mooring sequence-specific RNA binding protein that directs site-specific editing (<xref ref-type="bibr" rid="b6">6</xref>&#x02013;<xref ref-type="bibr" rid="b9">9</xref>,<xref ref-type="bibr" rid="b13">13</xref>).</p><p>Limited tissue expression of APOBEC-1 and <italic>apoB</italic> mRNA restricts editing in humans to the small intestine (&#x02265;85% editing), but <italic>apoB</italic> mRNA editing also occurs in the liver of several species (<xref ref-type="bibr" rid="b3">3</xref>,<xref ref-type="bibr" rid="b14">14</xref>&#x02013;<xref ref-type="bibr" rid="b16">16</xref>). Hepatic editing is modulated by fasting and refeeding in part due to an insulin-dependent increase in APOBEC-1 expression (<xref ref-type="bibr" rid="b17">17</xref>). Hepatic editing is also regulated independently of changes in APOBEC-1 expression levels by developmental, hormonal and nutritional perturbations (<xref ref-type="bibr" rid="b17">17</xref>&#x02013;<xref ref-type="bibr" rid="b23">23</xref>). The mechanism for this form of editing activity regulation has not been defined, but involves the nuclear trafficking of editing factors (<xref ref-type="bibr" rid="b24">24</xref>&#x02013;<xref ref-type="bibr" rid="b27">27</xref>).</p><p><italic>ApoB</italic> mRNA editing occurs primarily on spliced and polyadenylated RNA in the nucleus (<xref ref-type="bibr" rid="b2">2</xref>,<xref ref-type="bibr" rid="b24">24</xref>,<xref ref-type="bibr" rid="b25">25</xref>,<xref ref-type="bibr" rid="b28">28</xref>&#x02013;<xref ref-type="bibr" rid="b30">30</xref>). Despite this, APOBEC-1 and ACF are distributed in both the cytoplasm and nucleus (<xref ref-type="bibr" rid="b24">24</xref>,<xref ref-type="bibr" rid="b26">26</xref>,<xref ref-type="bibr" rid="b29">29</xref>&#x02013;<xref ref-type="bibr" rid="b31">31</xref>). In nuclear extracts, APOBEC-1 and ACF co-sedimented in 27S, editing-competent complexes, but as inactive 60S complexes in cytoplasmic extracts (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b24">24</xref>). Under <italic>in vitro</italic> editing conditions, 60S complexes reorganized to active 27S complexes on reporter RNAs (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b24">24</xref>). Furthermore, localization studies demonstrated that ACF and APOBEC-1 traffick between the cytoplasm and the nucleus (<xref ref-type="bibr" rid="b26">26</xref>,<xref ref-type="bibr" rid="b27">27</xref>). In support of trafficking as a regulatory mechanism, ethanol, insulin and thyroid hormone stimulation of hepatocyte editing activity were associated with an an increase in nuclear localization of ACF (<xref ref-type="bibr" rid="b24">24</xref>,<xref ref-type="bibr" rid="b29">29</xref>,<xref ref-type="bibr" rid="b32">32</xref>).</p><p>Induction of editing by ethanol occurred within minutes (<xref ref-type="bibr" rid="b21">21</xref>,<xref ref-type="bibr" rid="b23">23</xref>,<xref ref-type="bibr" rid="b29">29</xref>), and withdrawal of the stimulus both reduced the abundance of ACF in the nucleus and suppressed editing activity (<xref ref-type="bibr" rid="b23">23</xref>,<xref ref-type="bibr" rid="b24">24</xref>). Ethanol induced editing is not dependent on <italic>de novo</italic> protein or RNA synthesis (<xref ref-type="bibr" rid="b33">33</xref>) suggesting that modulation of pre-existing editing factors is sufficient to support enhanced editing activity. These observations substantiated the possibility that cytoplasmic 60S complexes may serve as a reservoir of editing factors necessary for rapid assembly of nuclear 27S editosomes. Protein phosphorylation is one of the most common protein modifications known and its importance in the regulation of protein activity has been well documented (<xref ref-type="bibr" rid="b34">34</xref>). In fact phosphorylation has been implicated as having a role in <italic>apoB</italic> mRNA editing although its mechanism remains unclear (<xref ref-type="bibr" rid="b35">35</xref>).</p><p>We show that ACF was phosphorylated on one or more serine residues, and that ethanol and insulin induction of <italic>apoB</italic> mRNA editing was accompanied by phosphorylation of ACF. PhosphoACF was only detected in the nucleus, and was selectively recovered with active 27S editosomes. Although ACF and APOBEC-1 are both present in the cytoplasm, APOBEC-1 co-immunoprecipitated with ACF only from nuclear extracts. Recovery of ACF/APOBEC-1 complexes and <italic>apoB</italic> mRNA editing activity were dependent on protein phosphorylation. Protein phosphatase inhibitor studies suggest that protein phosphatase 1 is involved in regulating editing activity, ACF phosphorylation and ACF subcellular distribution. The significance of ACF phosphorylation for ACF trafficking to the nucleus, association with APOBEC-1 and assembly into 27S editosomes and the regulation of editing efficiency is discussed.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><sec><title>Animal care, primary hepatocyte isolation and hepatoma cell culture</title><p>Male Sprague-Dawley rats (275&#x02013;325 g BW/Charles River Laboratories, Wilmington, MA) were housed under 12 h light/dark cycles and fed normal rat chow (Purina, St. Louis, MO) <italic>ad libitum</italic> and euthanized between 9 and 10 a.m. Primary hepatocytes were isolated (<xref ref-type="bibr" rid="b23">23</xref>) and plated onto BIOCOAT type I collagen coated dishes (Becton Dickinson Labware, Franklin Lakes, NJ) in Waymouth's 752/1 media (Sigma Chemical Co., St Louis, MO) containing 0.1 nM porcine insulin (Sigma) for 12&#x02013;16 h prior to the onset of each experiment.</p><p>McArdle RH7777 cells (ATCC Manassas, VA) stably expressing HA epitope-tagged APOBEC-1 (<xref ref-type="bibr" rid="b36">36</xref>) were treated for 4 h with 0.9% ethanol and fractionated into nuclear extracts (<xref ref-type="bibr" rid="b23">23</xref>).</p></sec><sec><title><italic>In vivo</italic> phosphorylation of ACF</title><p><italic>In vivo</italic> ³²P labeling was performed by intraperitoneal injection of rats with 12.5 mCi of orthophosphoric acid (10 mCi/ml ³²PO₄; NEN, Boston, MA) buffered with 50 mM HEPES, pH 7.0 and 150 mM NaCl. After 4 h, rats were sacrificed and hepatic cytoplasmic and nuclear extracts prepared. Primary hepatocyte cultures in 60 mm dishes were incubated in phosphate-free Minimum Essential Eagle Media (Sigma) containing 0.1 nM porcine insulin for 6 h and subsequently in fresh media containing 10 nM insulin (equivalent to that seen in post-prandial serum) or 0.45&#x02013;0.9% ethanol (AAPER Alcohol and Chemical Co., Shelbyville, KY) (<xref ref-type="bibr" rid="b23">23</xref>,<xref ref-type="bibr" rid="b29">29</xref>) plus 2.0 mCi orthophosphoric acid. Cultures were labeled for 4 h prior to extract preparation.</p></sec><sec><title>Subcellular extract preparations</title><p>Livers were perfused <italic>in situ</italic> with 0.25 M sucrose, 50 mM Tris, pH 7.0 and 5 mM MgCl₂ and protease inhibitors (Roche, Indianapolis, IN) (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b24">24</xref>). Cytoplasmic and nuclear extracts (<xref ref-type="bibr" rid="b37">37</xref>) were supplemented with 10 mM NaF and fractionated through glycerol gradients.</p><p>Cultured primary hepatocytes and hepatoma cell lines were rinsed with 1&#x000d7; Tris-buffered saline (TBS; 10 mM Tris and 150 mM NaCl, pH 7.5) and scraped into TBS containing proteinase/phosphatase inhibitors (Roche). Nuclear and cytoplasmic extracts were prepared using the NE-PER kit (Pierce, Rockford, IL) supplemented with 10 mM NaF and proteinase inhibitors.</p></sec><sec><title>Glycerol gradient fractionation</title><p>Preparative (35 ml) 10&#x02013;50% glycerol gradients containing 10 mM NaF were loaded with 20 mg or 2 mg of cytoplasmic or nuclear S100 extract respectively and sedimented at 100&#x02009;000&#x000d7; <italic>g</italic> for 10 h at 7&#x000b0;C. The distribution of 11S, 27S and 60S complexes (fractions 3/4, 5/6 and 7/8) was determined by the sedimentation of bovine serum albumin (6S), catalase (11S) and 60S spliceosomes (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b24">24</xref>). Gradient fractions were treated with DNase I and RNaseT1 to ensure the removal of DNA or RNA prior to the analysis of ACF ³²P-labeling.</p></sec><sec><title><italic>In vitro</italic> editing reactions and RNA binding assays</title><p><italic>In vitro</italic> editing reactions and quantitative poisoned-primer extension assays were carried out using 80 &#x003bc;g of extract proteins as described previously (<xref ref-type="bibr" rid="b11">11</xref>,<xref ref-type="bibr" rid="b37">37</xref>). <italic>ApoB</italic> RNA binding assays were performed by subjecting nuclear extract (80 &#x003bc;g) to <italic>in vitro</italic> editosome assembly and ultraviolet cross-linking (<xref ref-type="bibr" rid="b37">37</xref>).</p></sec><sec><title>Protein phosphatase inhibitor studies</title><p>Primary hepatocytes were treated for 6 h with cantharidin, endothall or okadaic acid at concentrations encompassing their respective <italic>in vivo</italic> IC₅₀ values as described by the manufacturer (Calbiochem, La Jolla, CA) and their suggested references. The effects of the compounds on <italic>apoB</italic> mRNA editing were evaluated by RT&#x02013;PCR and poisoned-primer extension analysis (<xref ref-type="bibr" rid="b38">38</xref>). The effect of 470 nM cantharidin on ACF subcellular distribution was investigated by treating primary hepatocyte cultures for 4 h followed by nuclear and cytoplasmic extract preparation.</p><p>To evaluate the effect of phosphatase inhibition on ACF phosphorylation, cultures were pre-incubated for 2 h with 470 nM cantharadin in phosphate-free Waymouth&#x02019;s media and subsequently supplemented with 0.5 mCi ³²PO₄ and incubated for an additional 4 h. Cultures were harvested and subcellular extracts prepared.</p></sec><sec><title>Immunological techniques</title><p>Rabbit polyclonal peptide-specific antibodies were raised against ACF N-terminal (NT) sequence (NHKSGDGLSGTQKE) and C-terminal (CT) sequence (HTLQTLGIPTEGGD) (<xref ref-type="bibr" rid="b24">24</xref>) and affinity purified with the corresponding peptides (Bethyl Laboratories, Inc., TX). For immunoprecipitation analyses all radiolabeled extracts were adjusted to 5 mM MgCl₂ and digested with 100 U of DNase I (Promega, Madison, WI), RNase T1 (Roche) and RNase A (Sigma) for 1 h on ice to remove radiolabeled nucleic acid. Where applicable, extracts were incubated overnight with ACF CT antibody at 4&#x000b0;C and subsequently reacted with Protein A-agarose (Oncogene Research Products, Boston, MA) pre-washed with 1&#x000d7; TBS/10 mM NaF. The immuno-absorbed material was washed three times with 1&#x000d7; TBS/10 mM NaF, three times with 1&#x000d7; TBS/1 M NaCl (<xref ref-type="bibr" rid="b36">36</xref>) and finally three times with 1&#x000d7; TBS. Immuno-absorbed material to be treated with alkaline phosphatase was washed with 50 mM Tris, pH 8.4, 1 mM MgCl₂, 0.1 mM ZnCl₂, 25% glycerol and then incubated with 5 U CIAP for 1 h at 30&#x000b0;C and then washed six times with 1&#x000d7; TBS. ACF-antibody complexes were eluted with 3 M sodium thiocyanate, acetone precipitated and analyzed by 10.5% SDS&#x02013;PAGE followed by autoradiography and/or western blotting with ACF NT antibody. For ACF immunoprecipitation from gradient fractions an equal volume of pooled gradient fractions from sedimentation zones of interest were reacted with sub-saturating amounts of ACF CT antibody to ensure that the recovery of phosphoACF was not simply a reflection of ACF abundance in each zone.</p></sec><sec><title>Two-dimensional gel electrophoresis</title><p>Extracts isolated from hepatocytes were analyzed for ACF charge isoforms using the Protean isoelectric focusing (IEF) system (Bio-Rad Laboratories, Hercules, CA). Immobilized pH gradient (IPG) strips (Bio-Rad Laboratories, pH range 3&#x02013;9.3) were hydrated with 150 &#x003bc;g nuclear or 340 &#x003bc;g cytoplasmic extract in two-dimensional (2D) loading buffer (7.5 M urea, 1.0 M thiourea, 1% CHAPS, 58 mM DTT and 0.2% biolytes) and electrophoresed to equilibrium.</p><p>After completion of IEF the IPG strip was equilibrated in SDS buffer for 30 min and then in iodoacetamide buffer for additional 30 min according to the manufacturer&#x02019;s recommendations. Proteins were resolved through a 10.5% Criterion gel (Bio-Rad Laboratories), transferred to nitrocellulose and reacted with ACF NT antibody.</p></sec><sec><title>Two-dimensional phosphoamino acid analysis</title><p>ACF was immunopurified using the ACF CT antibody from 27S enriched nuclear extracts of ³²P-labeled primary hepatocytes cultured in basal media (0.1 nM insulin) or in basal media containing 0.9% ethanol. Immunoprecipitates were resolved by SDS&#x02013;PAGE, blotted onto PVDF membrane (Bio-Rad, CA) and ACF was identified by autoradiography, excised and acid hydrolyzed in 5.7 N HCl at 110&#x000b0;C for 1 h (<xref ref-type="bibr" rid="b39">39</xref>). Lyophilized hydrolysates were spiked with unlabeled phosphoserine, phosphothreonine and phosphotyrosine (Sigma) and resolved on thin layer chromatography plates (Merck, Germany) by 2D electrophoresis using an HTLE 7000 peptide mapping system (CBS Scientific Co. Del Mar, CA) (<xref ref-type="bibr" rid="b39">39</xref>). The migration of the unlabeled standards and ACF radiolabeled amino acid(s) were visualized by ninhydrin staining and PhosphorImager Scanning densitometry, respectively.</p></sec></sec><sec><title>RESULTS</title><p>The minimal, functional editosome is composed of ACF, a homodimer of APOBEC-1 and the <italic>apoB</italic> RNA substrate (<xref ref-type="bibr" rid="b40">40</xref>,<xref ref-type="bibr" rid="b41">41</xref>). ACF and APOBEC-1 are distributed in the cytoplasm and nucleus of editing competent cells where they co-localize in macromolecular complexes of 60S and 27S, respectively (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b24">24</xref>). However, <italic>apoB</italic> mRNA editing is only associated with nuclear 27S complexes (<xref ref-type="bibr" rid="b24">24</xref>,<xref ref-type="bibr" rid="b29">29</xref>). Given that endogenous levels of APOBEC-1 are below the detection limits of currently available antibodies (<xref ref-type="bibr" rid="b36">36</xref>), the interaction of ACF with APOBEC-1 in the nucleus and cytoplasm was analyzed by co-immunoprecipitation from extracts prepared from McArdle cells that stably express HA-tagged APOBEC-1 (<xref ref-type="bibr" rid="b25">25</xref>,<xref ref-type="bibr" rid="b36">36</xref>). As anticipated, ACF and APOBEC-1 were abundant in both cytoplasmic and nuclear starting material (<xref ref-type="fig" rid="fig1">Figure 1</xref>). However, APOBEC-1 was efficiently co-immunoprecipitated with ACF only from nuclear extracts.</p><p>Given the selective recovery of APOBEC-1 with nuclear ACF (<xref ref-type="fig" rid="fig1">Figure 1</xref>), we investigated whether the interaction between ACF and APOBEC-1 in nuclear extracts was mediated by post-translational modifications such as protein phosphorylation. Treatment of nuclear extracts with alkaline phosphatase (10 U CIAP) resulted in a 3-fold reduction in HA-tagged APOBEC-1 recovered with immunoprecipitated ACF (<xref ref-type="fig" rid="fig2">Figure 2A</xref>).</p><p>As an interaction between ACF and APOBEC-1 is critical for editing activity (<xref ref-type="bibr" rid="b42">42</xref>) we investigated the effect of alkaline phosphatase treatment on <italic>in vitro</italic> editing activity of hepatocyte extracts. <italic>ApoB</italic> mRNA editing activity in CIAP-treated nuclear extract was significantly inhibited by 2.5-fold relative to control extracts (<xref ref-type="fig" rid="fig2">Figure 2B</xref>). The observed reduction in editing activity could be attributed to the failure of ACF to interact with APOBEC-1 or with the RNA substrate. In contrast to the interaction with APOBEC-1, CIAP treatment (1, 5 or 10 U) did not significantly increase ultraviolet light cross-linking of ACF to <italic>apoB</italic> mRNA (<xref ref-type="fig" rid="fig2">Figure 2C</xref>). These data suggest that the reduction in editing activity in CIAP-treated extracts was most probably the result of suppression of interactions between ACF and APOBEC-1 rather than an alteration in the binding affinity of ACF to <italic>apoB</italic> mRNA.</p><sec><title>ACF is a phosphoprotein</title><p>Previous studies using site-directed mutagenesis of APOBEC-1 and overexpression of protein kinase C_&#x003b8; implicated phosphorylation of APOBEC-1 as a mechanism for activating <italic>apoB</italic> mRNA editing (<xref ref-type="bibr" rid="b35">35</xref>). Although our aforementioned data are consistent with this possibility, the low level of endogenous APOBEC-1 expression has prohibited <italic>in vivo</italic> validation of this finding. Furthermore, <italic>in vivo</italic> studies carried out in our laboratory using exogenous rat APOBEC-1 were unable to detect APOBEC-1 phosphorylation (data not shown) and the phosphorylation sites suggested by the authors are not conserved between rat and human (<xref ref-type="bibr" rid="b35">35</xref>).</p><p>To evaluate whether endogenous ACF is phosphorylated <italic>in vivo</italic>, rats were radiolabeled for 4 h via an intraperitoneal injection of orthophosphoric acid in HEPES-buffered saline. Following extensive digestion of hepatic nuclear extracts with DNase I and RNase T1 to remove radiolabeled nucleic acids, ACF was immunoprecipitated with the CT antibody, resolved by SDS&#x02013;PAGE, transferred to nitrocellulose and analyzed by autoradiography as well as immunoblotting with ACF NT antibody (Figure 4A). A single band was detected by autoradiography that super-imposed with ACF-specific immunoblot reactivity. Greater than 90% of the ³²P-label was removed by CIAP treatment (Figure 4B), consistent with the post-translational phosphorylation of protein. Cytoplasmic ACF did not become radiolabeled (Figure 4A). The amino acid sequence of rat ACF is predicted to contain 23 serine/threonine and 7 tyrosine high probability sites of phosphorylation (<ext-link ext-link-type="uri" xlink:href="http://www.cbs.dtu.dk/services.NetPhosK"/> and <ext-link ext-link-type="uri" xlink:href="http://expasy.org/tools/scanprosite"/>). To investigate the complexity of physiologically relevant ACF phosphorylation sites, liver extracts prepared from control rats were resolved by equilibrium 2D gel electrophoresis and immunoblotted with ACF NT antibody (<xref ref-type="bibr" rid="b24">24</xref>). The predicted isoelectric point (pI) of ACF is 8.8 (<ext-link ext-link-type="uri" xlink:href="http://www.scripps.edu/~cdputnam/protcalc.html"/>) and the covalent addition of phosphate would be expected to cause an acidic shift of the pI (<xref ref-type="bibr" rid="b43">43</xref>). Nuclear ACF was detected as two predominant charge isoforms (<xref ref-type="fig" rid="fig3">Figure 3B</xref>). The major isoform migrated at pI 8.8 and is likely unmodified ACF or ACF containing both acidic and basic modifications. A second, less abundant, isoform migrated with a pI of 8.3 and is consistent with phosphorylation of ACF. To verify that the observed charge heterogeneity was due to protein phosphorylation, nuclear extracts were treated with CIAP which resulted in an almost complete loss of the acidic isoform (pI 8.3) and concomitant increase in the pI 8.8 isoform (<xref ref-type="fig" rid="fig3">Figure 3B</xref>), which is consistent with the removal of 2&#x02013;3 phosphates (0.2&#x02013;0.3 pH units per phosphate) (<xref ref-type="bibr" rid="b43">43</xref>). Cytoplasmic ACF migrated as a single isoform at pI 8.8, which did not shift upon phosphatase treatment (data not shown), further confirming that cytoplasmic ACF was not phosphorylated under our assay conditions.</p><p>To determine whether serine, threonine and/or tyrosine residues were the target of phosphorylation, primary hepatocytes were labeled to high specific activity with ³²PO₄ and ACF was immunoprecipitated from nuclear extracts. 2D thin layer electrophoresis of acid hydrolyzed ACF demonstrated that phosphorylation occurred on serine residues (<xref ref-type="fig" rid="fig3">Figure 3C</xref>, left panel). Although this technique is not intended for quantitation, it is interesting that the amount of phosphoserine detected increased (1.5-fold) in ACF immunopurified from an equivalent amount of nuclear extract isolated from primary hepatocytes incubated with 0.9% ethanol during the labeling period (<xref ref-type="fig" rid="fig3">Figure 3C</xref>, right panel). The 1.5-fold change in phosphoserine abundance likely represents a minimum as the radioactivity in partially hydrolyzed peptides (migrating along the first dimension) suggest greater ³²P incorporation (see Materials and Methods). These data suggest that threonine and tyrosine residues are not phosphorylated in rat ACF or that their phosphorylation exhibits a slow rate of turnover, preventing them from incorporating ³²P label during the experiment.</p></sec><sec><title>Phosphorylated ACF is only recovered with nuclear editosomes</title><p>If phosphorylated ACF is relevant to editing activity, it should be associated with nuclear 27S editosomes. To evaluate this, cytoplasmic and nuclear extracts from radiolabeled primary hepatocyte cultures were sedimented through 10&#x02013;50% glycerol gradients. ACF was immunoprecipitated with sub-saturating quantities of the ACF CT antibody from pooled fractions corresponding to (i) 11S (pre-editosomal ACF and APOBEC-1), (ii) the 27S editosome and (iii) 60S and greater. The ACF immunoprecipitates were resolved by SDS&#x02013;PAGE, transferred to nitrocellulose, autoradiographed and subjected to PhosphorImager scanning to detect and quantify radiolabeled proteins prior to immunoblotting with the ACF NT antibody. Consistent with prior analyses (<xref ref-type="bibr" rid="b24">24</xref>), ACF was recovered by immunopurification from all gradient fractions (<xref ref-type="fig" rid="fig4">Figure 4A</xref>). Although ACF was widely distributed in all nuclear fractions, phosphoACF was restricted to fractions containing the 27S editosome. Editing activity, indicative of the assembly of APOBEC-1 with ACF has only been observed in 27S gradient fractions (<xref ref-type="bibr" rid="b6">6</xref>). Considering data presented in <xref ref-type="fig" rid="fig1">Figures 1</xref>, <xref ref-type="fig" rid="fig2">2A and B</xref>, the selective recovery of phosphorylated ACF in the nuclear 27S editosome fraction is highly suggestive. Thus, a correlation linking phosphoACF to the physiologically relevant 27S editing complexes can be made.</p></sec><sec><title>Phosphorylation of ACF is metabolically regulated</title><p><italic>ApoB</italic> mRNA editing is regulated by a variety of hormonal and dietary factors [reviewed in (<xref ref-type="bibr" rid="b44">44</xref>)]. To determine if ACF phosphorylation is correlated with metabolic regulation of editing rat primary hepatocytes were labeled with ³²PO₄ for 4 h in the presence of either 0.45% ethanol or 10 nM insulin. A 3.5-fold increase of phosphoACF was associated with nuclear 27S editosomes isolated from ethanol-treated hepatocytes (<xref ref-type="fig" rid="fig4">Figure 4B</xref>). These data are consistent with the increase in serine phosphorylation observed following ethanol treatment (<xref ref-type="fig" rid="fig3">Figure 3C</xref>). Moreover, phosphoACF was not detected in any cytoplasmic fractions despite a 10-fold higher protein load compared with nuclear extracts onto the gradients and a prolonged autoradiographic exposure (<xref ref-type="fig" rid="fig4">Figure 4B</xref>).</p><p>To determine if enhanced ACF phosphorylation is a more general mechanism associated with editing induction, we investigated the effect of insulin treatment on primary hepatocytes. Insulin, like ethanol, stimulates <italic>apoB</italic> mRNA editing (<xref ref-type="bibr" rid="b17">17</xref>,<xref ref-type="bibr" rid="b24">24</xref>) and the nuclear accumulation of ACF (<xref ref-type="bibr" rid="b24">24</xref>). Consistent with data from ethanol-treated primary hepatocytes, addition of 10 nM insulin during the 4 h labeling period resulted in a 2.5-fold increase in the recovery of phosphorylated ACF in nuclear 27S editosomes (<xref ref-type="fig" rid="fig4">Figure 4C</xref>). No phosphoACF was detected in any cytoplasmic fractions of insulin stimulated hepatocytes. These data demonstrate that ACF phosphorylation is a general characteristic associated with modulation of <italic>apoB</italic> mRNA editing activity.</p></sec><sec><title><italic>ApoB</italic> mRNA editing and ACF phosphorylation can be modulated by protein phosphatase inhibitors</title><p>If ACF phosphorylation is integral to the regulation of <italic>apoB</italic> mRNA editing activity, and ACF must become dephosphorylated in order to traffick to the cytoplasm, we reasoned that inhibition of the appropriate protein phosphatase would result in a nuclear accumulation of phosporylated ACF and stimulate <italic>apoB</italic> mRNA editing. To evaluate this, hepatocytes were treated with a series of protein phosphatase inhibitors and editing activity determined. Protein phosphatase inhibitors were selected such that we could identify or rule-out a role for a specific class of protein phosphatases. Cantharidin is a protein phosphatase inhibitor with markedly different inhibitory concentrations for PP2A and PP1 (40 and 473 nM, respectively) (<xref ref-type="bibr" rid="b45">45</xref>,<xref ref-type="bibr" rid="b46">46</xref>). When tested at 470 nM, the IC₅₀ for PP1, we detected a reproducible, but not statistically significant increase in editing (<xref ref-type="fig" rid="fig5">Figure 5A</xref>). Additional experiments using 4.7 &#x003bc;M, a concentration anticipated to inhibit &#x0003e;90% PP1 activity, editing increased from 66 to &#x0003e;90% (<xref ref-type="fig" rid="fig5">Figure 5A</xref>). In addition, treatment of hepatocytes with both cantharidin and ethanol simultaneously, conditions proposed to stimulate both ACF nuclear localization and phosphorylation as well as to inhibit dephosphorylation resulted in the largest stimulation of editing activity.</p><p>To further support the role of PP1 in editing regulation, two additional protein phosphatase inhibitors, okadaic acid (<xref ref-type="bibr" rid="b45">45</xref>) and endothall (<xref ref-type="bibr" rid="b47">47</xref>) were tested. Editing was enhanced from 66 to &#x0003e;90% (<italic>P</italic> &#x02264; 0.01) in hepatocytes treated with okadaic acid at concentrations 10-times the IC₅₀ for PP1 (<xref ref-type="table" rid="tbl1">Table 1</xref>). Similarly, treatment with endothall at a concentration 10-times the IC₅₀ for PP1 stimulated editing to statistically significant levels (<italic>P</italic> &#x02264; 0.01) (<xref ref-type="table" rid="tbl1">Table 1</xref>).</p><p>Since calcium levels have been implicated in the regulation of editing (<xref ref-type="bibr" rid="b18">18</xref>), we investigated the effect of inhibition of the calcium-sensitive protein phosphatase 2B (PP2B) (<xref ref-type="bibr" rid="b48">48</xref>) on editing. Treatment of hepatocytes with cyclosporin A (<xref ref-type="bibr" rid="b49">49</xref>) and cypermethrin (<xref ref-type="bibr" rid="b50">50</xref>) did not affect significantly <italic>apoB</italic> mRNA editing (<xref ref-type="table" rid="tbl1">Table 1</xref>). In addition to ruling out PP2B, these data also demonstrate that enhanced editing in the presence of protein phosphatase inhibitors is not a non-specific effect owing to small molecule inhibitors; strengthening our position that PP1 is involved in editing regulation.</p><p>In order to correlate the observed effects on editing with changes in ACF phosphorylation, ACF was immunoprecipitated from extracts isolated from hepatocytes treated with increasing concentrations of cantharidin (<xref ref-type="fig" rid="fig5">Figure 5B</xref>). ACF phosphorylation was increased in hepatocytes treated with 470 nM cantharidin. Maximal ACF phosphorylation was observed when hepatocytes were treated with both ethanol and cantharidin, reflecting the editing activity data described in <xref ref-type="fig" rid="fig5">Figure 5A</xref>. Taken together, these data demonstrate that the cellular effects of cantharidin that lead to enhanced editing activity are associated with increased ACF phosphorylation.</p><p>Previous reports demonstrated accumulation of ACF in the nucleus of hepatocytes treated with ethanol and insulin and return of ACF to the cytoplasm upon removal of stimuli (<xref ref-type="bibr" rid="b24">24</xref>). To evaluate the effect of PP1 inhibition on the nuclear retention of ACF, hepatocytes were incubated with 470 nM cantharidin for 4 h, fractionated into nuclear and cytoplasmic extracts, resolved by SDS&#x02013;PAGE and immunoblotted for ACF. Subsequently, the blots were re-probed for actin and Histone H1 to verify cell fractionation quality and to serve as normalization standards for protein loading (<xref ref-type="fig" rid="fig6">Figure 6</xref>). The normalized relative abundance of nuclear to cytoplasmic ACF was 1:1 in control cells. Upon incubation with cantharidin a 5-fold increase in nuclear ACF was observed, consistent with our hypothesis that inhibition of ACF dephosphorylation prevents its nuclear export.</p></sec></sec><sec><title>DISCUSSION</title><p>Hepatic <italic>apoB</italic> mRNA editing is a regulated, nuclear process that requires the assembly of multi-protein editosomes. Reconstitution assays have identified the essential protein factors as the cytidine deaminase APOBEC-1 and the auxiliary factor ACF. Data presented in this report demonstrate that ACF is a phosphoprotein and that phosphoACF is restricted to nuclear 27S editosomes. ACF was phosphorylated on one or more serine residues under basal media conditions and the proportion of total cellular ACF that became phosphorylated increased upon ethanol or insulin treatment, stimuli both known to enhance editing activity. This suggests that ACF phosphorylation is metabolically regulated and a part of the mechanism for activating <italic>apoB</italic> mRNA editing regardless of whether APOBEC-1 expression is increased [insulin treatment, (<xref ref-type="bibr" rid="b17">17</xref>,<xref ref-type="bibr" rid="b22">22</xref>)] or not [ethanol treatment, (<xref ref-type="bibr" rid="b21">21</xref>)].</p><p>Recombinant APOBEC-1 and ACF purified from <italic>Escherichia coli</italic>, baculovirus or prepared from <italic>in vitro</italic> translation extracts can edit <italic>apoB</italic> RNA <italic>in vitro</italic> suggesting that phosphorylation is not essential for editosome assembly and editing activity. However, we argue that the efficiency of recombinant APOBEC-1 alone with recombinant ACF is entirely dependent on input protein concentration and the reaction has a very poor catalytic turnover, capable of only attomolar RNA substrate editing per hour (<xref ref-type="bibr" rid="b3">3</xref>,<xref ref-type="bibr" rid="b31">31</xref>,<xref ref-type="bibr" rid="b51">51</xref>). This is in sharp contrast to the highly efficient endogenous hepatic or intestinal cell editing activity. In fact, editing activity <italic>in vivo</italic> is regulated in a species- and tissue-specific manner, and inducible during development and in response to metabolic and hormonal perturbations (<xref ref-type="bibr" rid="b14">14</xref>,<xref ref-type="bibr" rid="b16">16</xref>) [and reviewed in (<xref ref-type="bibr" rid="b1">1</xref>)]. In this context, our data suggested that phosphorylation of ACF optimized and/or stabilized the functional interaction with APOBEC-1 in the nucleus leading to efficient editing activity. This mechanism explained how <italic>apoB</italic> mRNA editing activity can be activated metabolically or during development using pre-existing editing factors.</p><p>Under basal conditions, inhibition of PP1 activity resulted in the nuclear retention and increased recovery of phosphoACF and increased <italic>apoB</italic> mRNA editing activity, suggesting that ACF phosphorylation/dephosphorylation contributes to the modulation of editosome assembly and editing activity. Given that not all nuclear ACF is phosphorylated (<xref ref-type="fig" rid="fig3">Figures 3B</xref> and <xref ref-type="fig" rid="fig4">4</xref>) and that not all ACF is assembled in 27S editosomes [<xref ref-type="fig" rid="fig4">Figure 4</xref> and (<xref ref-type="bibr" rid="b12">12</xref>,<xref ref-type="bibr" rid="b24">24</xref>)] our data suggest that at any given time not all of the cellular ACF is involved in editing. This implies that there is a pool of ACF that can be used to rapidly modulate editosome assembly upon metabolic or hormonal stimuli or that ACF has additional roles in the cell.</p><p>In addition to editosome structure and function, ACF plays an important role in the cellular regulation of <italic>apoB</italic> mRNA editing through its trafficking activity between the cytoplasm and the nucleus (<xref ref-type="bibr" rid="b24">24</xref>&#x02013;<xref ref-type="bibr" rid="b27">27</xref>,<xref ref-type="bibr" rid="b30">30</xref>). Although the site of <italic>apoB</italic> mRNA editing is within the cell nucleus (<xref ref-type="bibr" rid="b2">2</xref>,<xref ref-type="bibr" rid="b29">29</xref>) and takes place during or immediately after pre-mRNA splicing (<xref ref-type="bibr" rid="b2">2</xref>,<xref ref-type="bibr" rid="b28">28</xref>,<xref ref-type="bibr" rid="b38">38</xref>), APOBEC-1 and ACF are distributed in both the nucleus and cytoplasm (<xref ref-type="bibr" rid="b9">9</xref>,<xref ref-type="bibr" rid="b24">24</xref>,<xref ref-type="bibr" rid="b26">26</xref>,<xref ref-type="bibr" rid="b27">27</xref>,<xref ref-type="bibr" rid="b30">30</xref>). The data presented here suggest that nuclear retention/import of ACF was increased in ethanol or insulin treated hepatocytes through ACF phosphorylation. The mechanism for regulating APOBEC-1 and ACF trafficking are unknown, and the dependence of each protein&#x02019;s trafficking on ACF&#x02013;APOBEC-1 complex formation is controversial (<xref ref-type="bibr" rid="b25">25</xref>&#x02013;<xref ref-type="bibr" rid="b27">27</xref>,<xref ref-type="bibr" rid="b30">30</xref>). Data from our laboratory suggest that APOBEC-1 has strong cytoplasmic retention signals, and that its nuclear import is mediated by interactions with ACF (<xref ref-type="bibr" rid="b25">25</xref>,<xref ref-type="bibr" rid="b30">30</xref>). We report that ACF and APOBEC-1 are present in both the cytoplasm and nucleus of editing competent cells, but that they only co-immunopurify from nuclear extracts. Our data suggest that nuclear retention/import of ACF is increased in ethanol or insulin treated hepatocytes through modulation of ACF phosphorylation state. We propose a model in which phosphorylation of ACF results in its nuclear accumulation and enhances or stabilizes APOBEC-1 nuclear retention and ACF binding, leading to increased editing activity. In support of this model, phosphatase treatment of cell extracts was associated with reduced co-immunoprecipitation of APOBEC-1 with ACF and reduced editing activity.</p><p>A small proportion of nuclear ACF was phosphorylated in non-stimulated hepatocytes (0.1 nM insulin) (<xref ref-type="fig" rid="fig3">Figures. 3B and C</xref>) which increased several-fold upon insulin and ethanol stimulation (<xref ref-type="fig" rid="fig3">Figures 3B and C</xref>). These data suggested that a low level turnover of ACF phosphorylation maybe required to maintain basal <italic>apoB</italic> mRNA editing activity. Significantly, PP1 inhibition stimulated editing activity even under basal media conditions. The turnover of editing complexes has been suggested from studies that demonstrated nucleocytoplasmic shuttling of ACF and APOBEC-1 (<xref ref-type="bibr" rid="b26">26</xref>,<xref ref-type="bibr" rid="b27">27</xref>,<xref ref-type="bibr" rid="b30">30</xref>,<xref ref-type="bibr" rid="b36">36</xref>) and from <italic>in vitro</italic> studies of editosome assembly (<xref ref-type="bibr" rid="b12">12</xref>). Addition of ethanol (or its catabolite, acetaldehyde) or chemicals affecting protein kinases and phosphatases to nuclear extracts did not affect <italic>in vitro</italic> editing activity or ACF phosphorylation (data not shown). These data indicate that intact cell signal transduction cascades are required for the regulation of ACF phosphorylation and <italic>apoB</italic> mRNA editing. The identification of PP1 as a candidate phosphatase involved in regulating phosphate turnover on ACF is relevant, as high levels of PP1 are present in rat hepatocyte nuclei (<xref ref-type="bibr" rid="b52">52</xref>) of which 90% was associated with chromatin (<xref ref-type="bibr" rid="b53">53</xref>). Nuclear ACF is also associated with chromatin (<xref ref-type="bibr" rid="b24">24</xref>) placing it theoretically within the general domain of nuclear PP1. The ability of phosphoACF to serve as substrate for PP1 remains to be formally addressed.</p><p>The lack of labeled ACF in the cytoplasm also suggests that ACF is dephosphorylated prior to or during nuclear export. These data suggest an interesting hypothesis: if phosphorylation of ACF is restricted to the nucleus and associated with enhanced editing activity, then dephosphorylation of ACF might regulate its nuclear export. Given that ACF binds to both unedited and edited <italic>apoB</italic> mRNA (<xref ref-type="bibr" rid="b6">6</xref>) and that dephosphorylated ACF binds to <italic>apoB</italic> RNA. ACF is likely to remain bound to <italic>apoB</italic> mRNA and co-export to the cytoplasm following <italic>apoB</italic> mRNA editing and ACF dephosphorylation. In this scenario, the regulation of ACF dephosphorylation would modulate <italic>apoB</italic> mRNA export to the cytoplasm in addition to protecting edited <italic>apoB</italic> mRNA from nonsense mediated decay (NMD) (<xref ref-type="bibr" rid="b31">31</xref>).</p><p>Under basal conditions, 2D gel electrophoresis analyses suggested that ACF contained 2&#x02013;3 phosphates. 2D phosphoamino acid analyses indicated serines were the residues phosphorylated following metabolic stimulation. ACF phosphorylation and whether these sites are the same as the &#x02018;basal&#x02019; phosphorylation sites or are additional sites of phosphorylation remains to be determined. Although parallels were observed when comparing ethanol with insulin editing induction (i.e. ACF accumulation in the nucleus and ACF hyperphosphorylation), we cannot assume that the same sites of ACF phosphorylation are involved or that the phosphorylation state of additional editing factors is not affected. In fact, previous studies involving alanine and aspartic acid site-specific mutagenesis of predicted serine phosphorylation sites suggested that APOBEC-1 may have two sites of phosphorylation which had opposing effects of editing activity (<xref ref-type="bibr" rid="b35">35</xref>). APOBEC-1 has never been validated as a phosphoprotein through metabolic labeling studies similar to those reported here because the expression of endogenous APOBEC-1 is prohibitively low. We evaluated phosphorylation of APOBEC-1 overexpressed in a stable McArdle cell line under basal and ethanol stimulated conditions and found no evidence for radiolabeling during a 4 h incubation period (data not shown). If APOBEC-1 is a phosphoprotein, the sites of phosphorylation may not be subjected to acute regulation and in fact Ser47 and Ser72, which were proposed to be phosphorylated in human APOBEC-1, are not conserved in rat APOBEC-1. The expression level of, and availability of high titre antibodies against, two APOBEC-1 homologs namely activation induced deaminase (AID) and APOBEC-3G have made equivalent studies possible and both proteins were identified as phosphoproteins (<xref ref-type="bibr" rid="b54">54</xref>,<xref ref-type="bibr" rid="b55">55</xref>). Consequently, phosphorylation of APOBEC-1 and its role in regulating editing activity remains a formal possibility.</p><p>In conclusion, regulation of hepatic <italic>apoB</italic> mRNA editing by ethanol and insulin promotes serine phosphorylation of ACF and its localization to active nuclear 27S editosomes. The data support a role for the metabolic regulation of ACF phosphorylation that promotes its interaction with APOBEC-1, in editosome assembly, as well as ACF nuclear retention/import. Thus, phosphorylation of ACF adds a new level of understanding of the control mechanisms cells use to modulate <italic>apoB</italic> mRNA editing in the context of current models of editosome composition and assembly.</p></sec>

Detecting non-orthology in the COGs database and other approaches grouping orthologs using genome-specific best hits

<p>Correct orthology assignment is a critical prerequisite of numerous comparative genomics procedures, such as function prediction, construction of phylogenetic species trees and genome rearrangement analysis. We present an algorithm for the detection of non-orthologs that arise by mistake in current orthology classification methods based on genome-specific best hits, such as the COGs database. The algorithm works with pairwise distance estimates, rather than computationally expensive and error-prone tree-building methods. The accuracy of the algorithm is evaluated through verification of the distribution of predicted cases, case-by-case phylogenetic analysis and comparisons with predictions from other projects using independent methods. Our results show that a very significant fraction of the COG groups include non-orthologs: using conservative parameters, the algorithm detects non-orthology in a third of all COG groups. Consequently, sequence analysis sensitive to correct orthology assignments will greatly benefit from these findings.</p>

<contrib contrib-type="author"><name><surname>Dessimoz</surname><given-names>Christophe</given-names></name><xref ref-type="corresp" rid="cor1">*</xref></contrib><contrib contrib-type="author"><name><surname>Boeckmann</surname><given-names>Brigitte</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Roth</surname><given-names>Alexander C. J.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Gonnet</surname><given-names>Gaston H.</given-names></name></contrib><aff><institution>ETH Zurich, Institute of Computational Science</institution><addr-line>CH-8092 Z&#x000fc;rich</addr-line></aff><aff id="au1">¹<institution>Swiss Institute of Bioinformatics, CMU</institution><addr-line>Michel-Servet 1, CH-1211 Gen&#x000e8;ve</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p>The identification of orthologous genes is a central problem in bioinformatics. Orthologs are genes that evolve from a common ancestor through speciation events, as opposed to paralogs, that result from gene duplication (<xref ref-type="bibr" rid="b1">1</xref>). Discriminating orthologs from paralogs is an important, but non-trivial task. It is important, because function conservation is considerably higher among orthologs (<xref ref-type="bibr" rid="b2">2</xref>), and also because only orthologs reflect the history of their species (<xref ref-type="bibr" rid="b1">1</xref>), meaning that phylogeny inferences must be based on orthologs. It is non-trivial because this distinction requires precise estimates of evolutionary distances from data that are often noisy. Other complications include gene deletion, variations in evolutionary rates, lateral gene transfer (LGT), or simply the fact that orthology and paralogy are non-transitive relations, meaning that the relation of every pair of genes must be analyzed separately.</p><p>So far, several projects have addressed this problem systematically. Of those, the COGs database (<xref ref-type="bibr" rid="b3">3</xref>,<xref ref-type="bibr" rid="b4">4</xref>) is by far the best established, probably due to its early inception, its wide scope, its reasonable performance and its presence on the NCBI website. The significance of COG in the community is reflected by hundreds of references in scientific articles. Even more importantly, most current initiatives for the identification of orthologs use ideas derived from the methodology of COG, in particular the idea of genome-specific best hit (<xref ref-type="bibr" rid="b5">5</xref>&#x02013;<xref ref-type="bibr" rid="b7">7</xref>). Of all those projects depending either on the methods or results from COG, few question the accuracy of them.</p><p>In its last accessible release (2003), the COGs database groups 138&#x02009;458 proteins from 66 prokaryotes into 4873 groups that consist of orthologs and in-paralogs. The term in-paralog was coined by Remm and coworkers (<xref ref-type="bibr" rid="b6">6</xref>) and describes in this context paralogs inside the same species (&#x02018;trivial paralogs&#x02019;), as opposed to out-paralogs that result from a duplication event prior to the last speciation event. [Strictly speaking, in/out-paralogy is a relation defined over two sequences and a speciation event of reference. When that event is omitted, it is here the last speciation event that is implied.] The inclusion of in-paralogs is usually justified by the fact that such sequences are orthologous to every other sequence within their group. Consequently, the relation of every pair of sequences inside the same COG is unambiguous: pairs of sequences from the same species are paralogs, otherwise, they are expected to be orthologous. The construction of COG groups is based on the fact that orthologous genes almost always have a higher level of sequence conservation than paralogs. Hence, genome-specific best hits (&#x02018;BeTs&#x02019;) are likely to be formed between orthologs. Yet, if the corresponding ortholog is missing, a BeT might link paralogous sequences. That problem is partly taken care of by COG's approach: BeTs are only grouped when they form triangles, and triangles are merged only when they have a common side. However, if more than one species have lost the corresponding ortholog, the construction over triangles will not suffice to prevent paralogs from being clustered together. This scenario is far from being unlikely, because losses occurring before speciation events get replicated, and therefore the problem becomes very significant as more species and strains are included for analysis. In fact, simple situations, such as the one illustrated on <xref ref-type="fig" rid="fig1">Figure 1</xref> are sufficient to have paralogs clustered together. It is then up to the human curation step at the end of the COG building process (<xref ref-type="bibr" rid="b3">3</xref>) to resolve all such cases.</p><p>The difficulty caused by a single missing ortholog can be easily avoided by requiring that all BeTs be symmetrical, which is what most other projects do. However, if the corresponding ortholog is missing in both genomes, even a symmetrical BeT will link paralogs. Therefore, BeTs, even symmetrical, are not necessarily linking orthologs.</p><p>This problem could be solved through phylogenetic analysis of the relevant gene families, in particular tree reconciliation (<xref ref-type="bibr" rid="b8">8</xref>), but this procedure is not yet practical in large-scale, automated contexts (<xref ref-type="bibr" rid="b2">2</xref>). In the following, we present an algorithm that detects non-orthology without the need of gene tree construction, then report its application on the last version of the COGs database. The algorithm was developed in the context of our own orthology classification project OMA (<xref ref-type="bibr" rid="b9">9</xref>), in which it is used to verify every predicted orthologous relation.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><p>The algorithm presented here is designed to detect non-trivial paralogous relations within groups of orthologs such as COG groups. Knowing that a paralogous relation within a group is likely to be caused by the loss of the corresponding ortholog in both species, the algorithm looks for a third-party species, which we call the &#x02018;witness of non-orthology&#x02019;, in which both corresponding orthologs are present (<xref ref-type="fig" rid="fig2">Figure 2</xref>). Under the assumptions of good and complete data, and similar evolutionary rates among orthologs, such a situation is characterized by the following three requirements on the evolutionary distances: (i) In <italic>Z, z</italic>₃ is the closest protein to <italic>x</italic>₁ and <italic>z</italic>₄ is the closest protein to <italic>y</italic>₂. (ii) The pair (<italic>x</italic>₁, <italic>z</italic>₃) must be significantly closer than (<italic>x</italic>₁, <italic>z</italic>₄), and conversely, (<italic>y</italic>₂, <italic>z</italic>₄), must be significantly closer than (<italic>y</italic>₂, <italic>z</italic>₃), That excludes cases where <italic>z</italic>₃ and <italic>z</italic>₄ are in-paralogs (<xref ref-type="fig" rid="fig3">Figure 3</xref>, left), because for in-paralogs to fulfill those conditions, convergent evolution at the sequence level would be required, a phenomenon that is so unlikely that we ignore it (<xref ref-type="bibr" rid="b10">10</xref>). (iii) The distance between (<italic>x</italic>₁, <italic>z</italic>₄), must be similar to (<italic>y</italic>₂, <italic>z</italic>₃). That excludes cases where <italic>X</italic> (respectively <italic>Y</italic>) speciated before the duplication event, in which case <italic>x</italic>₁ (respectively <italic>y</italic>₂) is orthologous to all three other genes (<xref ref-type="fig" rid="fig3">Figure 3</xref>, right).</p><p>We finish this overview of the algorithm by considering the impact of LGT and gene fusion/fission. Clearly, the algorithm presented here was not designed to detect LGT events between <italic>x</italic>₁ and <italic>y</italic>₂, an interesting problem in itself that remains largely unsolved. More importantly here, an LGT in a third-party species <italic>Z</italic> can lead to a situation where <italic>Z</italic> wrongly appears to be witness of non-orthology: consider three orthologous proteins <italic>x</italic>₁, <italic>y</italic>₂ and <italic>z</italic>₃ in three species <italic>X</italic>, <italic>Y</italic> and <italic>Z</italic>. At some point, <italic>Z</italic> acquires through LGT a member of that orthologous family, which we now refer to as <italic>z</italic>₄. <italic>Z</italic> keeps both copies <italic>z</italic>₃ and <italic>z</italic>₄. Furthermore, <italic>Z</italic> happens to be closer to <italic>X</italic> than <italic>Y</italic>, while the donor of <italic>z</italic>₄ is closer to <italic>Y</italic> than <italic>X</italic>. This situation leads to a misclassification by our algorithm. Although such cases cannot be ruled out, we did not encounter any among the numerous case-by-case analysis performed on the results. It could be that orthologous gene displacement of <italic>z</italic>₃ by <italic>z</italic>₄ through homologous recombination is a much more likely scenario, and besides, the frequency of LGT appears to be higher among closely related species (<xref ref-type="bibr" rid="b11">11</xref>). As for gene fusion or gene fission, the units for amino acid sequence analysis are no longer proteins but domains. Even though the analysis of homologous domains from distinct proteins is scientifically meaningful, our analysis remains at the level of entire proteins to simplify matters.</p><p>Note that the complications caused by LGT events and, probably to a lesser extent, by gene fusion/fission are not specific to our method and pose challenges to other approaches as well, in particular tree reconciliation.</p><sec><title>Input data</title><p>The algorithm uses two inputs: the COGs database and pairwise sequence alignments between all proteins involved in the analysis. As introduced above, the orthology of two sequences is verified through an exhaustive search of the corresponding sequences in complete, third-party genome. Therefore, a large number of genomes is desirable. However, since the relation between every pair of sequence is needed, such searches require the computation of a very large number of pairwise alignments. For practical reasons, all results presented here use results from the Smith&#x02013;Waterman (<xref ref-type="bibr" rid="b12">12</xref>) all-against-all protein alignments precomputed in the scope of the OMA project (<xref ref-type="bibr" rid="b9">9</xref>).</p><p>For each alignment, a PAM distance estimate and the corresponding variance is computed using maximum likelihood and numeric integration (<xref ref-type="bibr" rid="b13">13</xref>,<xref ref-type="bibr" rid="b14">14</xref>).</p></sec><sec><title>Comparison of evolutionary distances</title><p>The algorithm uses evolutionary distances to detect paralogs. However, the distances estimates are subject to perturbation, which must be taken into account when comparing them. Therefore, assuming that errors are normally distributed, the difference &#x00394;(<italic>d</italic>₁, <italic>d</italic>₂) of two distances <italic>d</italic>₁, <italic>d</italic>₂ has expected value: <disp-formula><mml:math id="M1"><mml:mrow><mml:mi>E</mml:mi><mml:mo>[</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>]</mml:mo><mml:mo>=</mml:mo><mml:mi>E</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>&#x02212;</mml:mo><mml:mi>E</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:math></disp-formula> with variance <disp-formula><mml:math id="M2"><mml:mrow><mml:msup><mml:mi>&#x003c3;</mml:mi><mml:mn>2</mml:mn></mml:msup><mml:mo>[</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>]</mml:mo><mml:mo>=</mml:mo><mml:msup><mml:mi>&#x003c3;</mml:mi><mml:mn>2</mml:mn></mml:msup><mml:mo>(</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>+</mml:mo><mml:msup><mml:mi>&#x003c3;</mml:mi><mml:mn>2</mml:mn></mml:msup><mml:mo>(</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>&#x02212;</mml:mo><mml:mn>2</mml:mn><mml:mtext>Cov</mml:mtext><mml:mo>(</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:math></disp-formula> If the two distances are independent, the covariance term disappears and the variance of the difference can be obtained directly from the individual variances. But more often than not, <italic>d</italic>₁ and <italic>d</italic>₂ involve a common protein and are therefore not independent, meaning that not taking the covariance into account overestimates the error. We have developed a method to approximate the covariance of two evolutionary distances, which will be the subject of a separate article.</p></sec><sec><title>Algorithm</title><p>The algorithm goes through each COG group, and verifies inside each of them that every two genes <italic>x</italic>₁, <italic>y</italic>₂ coming from different species have a significant alignment, and are indeed orthologs. Alignments are considered significant if the score is above 130 (47 bits, which typically corresponds to an <italic>E</italic>-value around 2e&#x02212;6) and the length of the alignment not &#x0003c;50% of the smallest sequence. The verification of orthology is performed through the search, in each third-party genome <italic>Z</italic>, of two genes <italic>z</italic>₃ and <italic>z</italic>₄ that fulfill the three conditions (i&#x02013;iii) presented at the beginning of this section: <disp-formula id="e1"><label>1</label><mml:math id="M3"><mml:mtable columnalign="left"><mml:mtr><mml:mtd><mml:mo>&#x02200;</mml:mo><mml:msub><mml:mi>z</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>&#x02260;</mml:mo><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>:</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>&#x0003c;</mml:mo><mml:mi>k</mml:mi><mml:mo>&#x022c5;</mml:mo><mml:mi>&#x003c3;</mml:mi><mml:mo>[</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>]</mml:mo></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mo>&#x02200;</mml:mo><mml:msub><mml:mi>z</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>&#x02260;</mml:mo><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>:</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>&#x0003c;</mml:mo><mml:mi>k</mml:mi><mml:mo>&#x022c5;</mml:mo><mml:mi>&#x003c3;</mml:mi><mml:mo>[</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>]</mml:mo></mml:mtd></mml:mtr></mml:mtable></mml:math></disp-formula> <disp-formula id="e2"><label>2</label><mml:math id="M4"><mml:mtable columnalign="left"><mml:mtr><mml:mtd><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>&#x0003e;</mml:mo><mml:mi>k</mml:mi><mml:mo>&#x022c5;</mml:mo><mml:mi>&#x003c3;</mml:mi><mml:mo>[</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>]</mml:mo></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>&#x0003e;</mml:mo><mml:mi>k</mml:mi><mml:mo>&#x022c5;</mml:mo><mml:mi>&#x003c3;</mml:mi><mml:mo>[</mml:mo><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>]</mml:mo></mml:mtd></mml:mtr></mml:mtable></mml:math></disp-formula> <disp-formula id="e3"><label>3</label><mml:math id="M5"><mml:mrow><mml:mrow><mml:mo>&#x02223;</mml:mo><mml:mrow><mml:mi>&#x00394;</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>)</mml:mo></mml:mrow><mml:mo>&#x02223;</mml:mo></mml:mrow><mml:mo>&#x0003c;</mml:mo><mml:mi>k</mml:mi><mml:mo>&#x022c5;</mml:mo><mml:msqrt><mml:mrow><mml:mo>[</mml:mo><mml:msup><mml:mi>&#x003c3;</mml:mi><mml:mn>2</mml:mn></mml:msup><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>4</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>+</mml:mo><mml:msup><mml:mi>&#x003c3;</mml:mi><mml:mn>2</mml:mn></mml:msup><mml:mo>(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>z</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo>)</mml:mo><mml:mo>]</mml:mo></mml:mrow></mml:msqrt><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula> where <italic>k</italic> is the confidence level, which we set to 1.96. If the quartet (<italic>x</italic>₁, <italic>y</italic>₂, <italic>z</italic>₃, <italic>z</italic>₄), fulfills all three conditions, there is enough evidence to consider <italic>x</italic>₁, <italic>y</italic>₂ paralogs. The algorithm was implemented in the programming environment Darwin (<xref ref-type="bibr" rid="b15">15</xref>).</p><p>A note about parameter choice. As mentioned previously, the classification of protein pairs in orthologs and non-orthologs can be very difficult or even impossible, especially when a speciation event immediately follows a duplication event, or in the situation of frequent gene gain and gene loss, as it is observed in certain groups of proteins, such as metabolic enzymes. Here, the choice of <italic>k</italic> = 1.96 standard deviations was established empirically such that the false-positive rate (orthologs misclassified as non-orthologs) is much smaller than the false-negatives rate (missed non-orthologs). In other words, we expect that our algorithm reports only clear-cut cases of paralogy.</p></sec><sec><title>Phylogenetic analysis</title><p>To verify individual cases reported by the algorithm, phylogenetic trees were constructed using independent, common software packages, as follows: sequences were aligned using Muscle (<xref ref-type="bibr" rid="b16">16</xref>) and ClustalW (<xref ref-type="bibr" rid="b17">17</xref>). Whenever they differed, the one that seemed more likely was selected. Short sequences, suspicious regions and most gap-containing columns removed. Distance matrices (JTT, gamma) generated with protdist (<xref ref-type="bibr" rid="b18">18</xref>) were used to construct phylogenetic trees using neighbor (<xref ref-type="bibr" rid="b18">18</xref>). Clusters of interest were selected for detailed analysis. Alignments of the selected data were performed using Tcoffee (<xref ref-type="bibr" rid="b19">19</xref>) and the result subsequently modified as described above, and considering the Tcoffee CORE (consistency of overall residue evaluation) values for the alignment. Information on the stability of the tree topology was assessed building an extended majority rule consensus tree using consense (<xref ref-type="bibr" rid="b18">18</xref>) from BIONJ (<xref ref-type="bibr" rid="b20">20</xref>) searches performed on 1000 bootstrap replicates, which were constructed with seqboot (<xref ref-type="bibr" rid="b18">18</xref>). Protein trees of the data subset were constructed using the Bayesian tree-building method MrBayes (<xref ref-type="bibr" rid="b21">21</xref>) (JTT; invgamma-4; 1&#x02009;000&#x02009;000 generations). The trees were rooted using an outgroup whenever a suitable ancient paralog could be found. Note that since the analysis attempts at clustering homologs into clans, and not at predicting their hierarchical order, placement of the root is not critical here.</p></sec><sec><title>Validation</title><p>The performances of the algorithm were evaluated using the HAMAP database (<xref ref-type="bibr" rid="b22">22</xref>), a collection of orthologous microbial protein families generated manually by expert curators in the Swiss&#x02013;Prot group. The database was retrieved on November 23, 2005. Proteins from the 99 most represented species also present in our OMA project were used in the analysis: of all 29 245 proteins, there were 21&#x02009;831 proteins (75.6%), grouped in 1189 orthologous families. That yielded 309&#x02009;829 pairwise relations to be verified by our procedure.</p><p>The algorithm classified 279&#x02009;568 (90.2%) relations as orthologous and 9420 (3.0%) as paralogous. The remaining 20&#x02009;841 (6.7%) relations had alignments below our significance threshold and could therefore not be processed. The accuracy of the algorithm, in particular its very low false-positive rate was confirmed by following observations:</p><p>First, paralogy is often reflected by different Swiss&#x02013;Prot ID names (e.g. GREA/GREB) (<xref ref-type="bibr" rid="b23">23</xref>). From the 9420 predicted paralogs, only 2728 (29.0%) of them have identical ID names. Second, the distribution of the paralogs among HAMAP families was investigated: all 9420 cases of paralogy found by the algorithm are concentrated in only 150 (12.6%) of the 1189 HAMAP families. This is consistent with the fact that the inclusion of just one paralogous protein into an orthologous family is likely to result in several paralogous relations inside that family. And indeed, in all except 8 of these 150 families, more than one paralogous pair was detected. Third, these 8 improbable cases were inspected individually using phylogenetic analysis, which confirmed that they are bona fide paralogs (possibly xenologs). Fourth, the predicted cases of paralogy were compared to the gene trees over HAMAP families built by the group of Laurent Duret (<ext-link ext-link-type="uri" xlink:href="http://pbil.univ-lyon1.fr/help/HAMAP.html"/>), in a similar way as HOBACGEN (<xref ref-type="bibr" rid="b24">24</xref>). 7217 predicted cases could be mapped to those trees. In 6418 (88.9%) instances, paralogy was confirmed by the trees, a remarkably high level of consistency considering that the two methods are very different. As for the conflicting 799 cases, which are distributed among 51 families, we believe that most of them are caused by inaccuracies on the gene trees, which are constructed using a variant of Neighbor Joining on observed divergence, a rather crude measure of evolutionary distance.</p></sec></sec><sec><title>RESULTS AND DISCUSSION</title><p>The algorithm was run on the current release of the COGs database (<xref ref-type="bibr" rid="b4">4</xref>) (<ext-link ext-link-type="uri" xlink:href="http://www.biomedcentral.com/1471&#x02013;2105/4/41"/>). We used the precomputed all-against-all results from 107 complete genomes, of which 52 are represented in COGs, whereas the remaining 55 genomes were only used as potential witnesses of non-orthology. [The complete list is available in the Supplementary Data.] From all 4654 COGs, there is a total of 5&#x02009;537&#x02009;713 pairwise relations. Pairs between proteins from the same species (484 043) were not considered further. Additionaly, 2&#x02009;733&#x02009;371 relations involve at least one protein from a species outside our set of 107 genomes. Consequently, the following results were obtained through the verification of 2&#x02009;320&#x02009;199 relations, 45.9% of all potential orthologous relations.</p><p>The results are presented in <xref ref-type="table" rid="tbl1">Table 1</xref>. Surprisingly, 44% of the relations had alignment scores below our significance threshold of 130, which corresponds to an <italic>E</italic>-value of about 2e&#x02212;6, and could therefore not be verified. This implies that an important fraction of relations within COGs cannot be, on the basis of pairwise alignments, reliably considered homologous.</p><p>The other result is the significant proportion of non-orthologous relations found by the algorithm, more than a quarter of the pairs that could be verified. They are distributed among about a third of all COGs. The list of such groups, along with all detected non-orthology cases are available in the Supplementary Data.</p><p>If we require the presence of at least two witnesses of non-orthology for a pair to be considered non-orthologous, the algorithm still finds 251 391 (19.4%) such pairs within 1146 (24.6%) COGs. When removing the sequence with the most non-orthologous relations from each COG group, the total number of non-orthologous pairs decreases by only 24&#x02009;868 (1.9%).</p><p>The majority (70%) of the groups predominantly non-orthologs are involved in metabolic processes, according to the functional description of the COGs database, although they only constitute a minority of all COGs. In contrast, groups involved in information storage and processing (8%) or cellular processing and signaling (11%) include less frequently non-orthologs. The remainder 11% are poorly characterized proteins. This result is in agreement with previous studies, which state that in prokaryotes, metabolic functions are under high evolutionary pressure from changing environments (<xref ref-type="bibr" rid="b25">25</xref>).</p><sec><title>Phylogenetic analysis of selected COG groups</title><p>The presence of non-orthology in some COG groups is hardly a surprise and was in fact recently acknowledged by Koonin, coauthor of COG, in a review article (<xref ref-type="bibr" rid="b2">2</xref>). What is surprising here is rather the extent of non-orthology detected by the algorithm. That prompted us to verify, in addition to the validation work reported in the previous section, a number of our predictions using detailed phylogenetic analysis. In this section, we report the conclusion of such analysis on three COGs, for which we could build Bayesian likelihood trees of high confidence, confirmed by consensus NJ trees with high bootstrap values. Clan assignments were made based on those trees, and considering lineage and function, whenever reliable annotations could be found. We strongly expect that pairs of proteins across clans be non-orthologous, and use these results to evaluate the accuracy of the predictions made by the algorithm.</p><p>COG0508 consists of complex-forming acyltransferases that are composed of an N-terminal biotin or lipoic acid attachment domain, a central protein&#x02013;protein interaction domain, followed by the catalytic 2-oxoacid dehydrogenases acyltransferase domain. The phylogenetic analysis of roughly half of the proteobacterial sequence data from COG0508 suggests the existence of at least four distinct subgroups (see <xref ref-type="fig" rid="fig4">Figure 4</xref>): clan 1 is formed by sequences from gammaproteobacteria, including the dihydrolipoyllysine-residue acetyltransferase component of the pyruvate dehydrogenase complex (EC 2.3.1.12) (AceF) from <italic>Escherichia coli</italic>. Clan 2 consists of proteins highly similar to the <italic>Bacillus subtilis</italic> lipoamide acyltransferase component of the branched-chain alpha-keto acid dehydrogenase complex (EC 2.3.1.168). All sequences in clan 2 are alphaproteobacterial, except for <italic>Pseudomonas aeruginosa</italic> proteins, which are found in both clan 1 and clan 2. As mentioned in section 2, such situation could arise through lateral gene transfer from an alphaproteobacteria to <italic>P.aeruginosa</italic>. If that was the case, there would be strong evidence that clans 1 and 2 should be merged. However, in the present case, it is possible to populate both clans with additional sequences from more distant species (data not shown), legitimating the separation in two clans. Additionally, the long distance between the two clans and the distinct function of at least one family member of each subgroup also supports this conclusion. Clan 3 includes the dihydrolipoyllysine-residue succinyltransferase component of 2-oxoglutarate dehydrogenase complex (EC 2.3.1.61) (SucB) of <italic>E.coli</italic>. Note that clan 3 includes two protein sequences of <italic>Rhizobium meliloti</italic>, but those are clearly ancient duplicates, and thus sequence 3b is likely to form yet a separate clan on its own. Finally clan 4 is formed by a presumably further dehydrogenase component from alphaproteobacteria. The algorithm predicted 382 cases of non-orthologous relations within the sequences considered here. An extract of the result list is given in <xref ref-type="table" rid="tbl2">Table 2</xref> (the full list of paralogy is available in the Supplementary Data). A total of 379 predictions are consistent with the clan assignment, while the remaining three predictions support the exclusion of <italic>R.meliloti</italic> 3b from clan 3. Furthermore, comparison with the clan assignment reveals that the algorithm missed 24 non-orthologous relations, which implies a false-negative rate of 6.0%.</p><p>COG0513 includes various DEAD-box containing RNA helicases. The phylogenetic analysis of the proteobacterial data from this group suggests the existence of six clans (see <xref ref-type="fig" rid="fig5">Figure 5</xref>), of which five are formed around the following proteins from <italic>E.coli</italic>: (i) the ATP-dependent RNA helicase SrmB, which is involved in an early assembly step of 50S ribosomal subunits (<xref ref-type="bibr" rid="b26">26</xref>); (ii) the cold-shock DEAD-box protein A (DeaD), required for cell division and normal cell growth at low temperature (<xref ref-type="bibr" rid="b27">27</xref>); (iii) the DEAD-box RNA helicase B (RhlB), a component of the RNA degradosome, which seems to have little activity unless being activated by the endoribonuclease RNase E (<xref ref-type="bibr" rid="b28">28</xref>); (iv) the putative RNA helicase RhlE, which has been shown to be non-essential for normal cell growth (<xref ref-type="bibr" rid="b29">29</xref>); (v) the ATP-independent RNA 3&#x02032;&#x02192;5&#x02032; helicase DbpA (<xref ref-type="bibr" rid="b30">30</xref>) and (vi) the subgroup includes RNA helicases that are conserved in some alphaproteobacteria. The algorithm predicted 408 cases of non-orthology, 88.9% of the 459 non-orthologous relations that can be deduced from the clan assignment. In this case, there was no false-positive prediction.</p><p>COG1113 consists of members of the amino acid-polyamine-organo-cation (APC) superfamily from bacteria, specifically those integral membrane proteins that are involved in the transport of amino acids in prokaryotes. The phylogenetic analysis of this group suggests the existence of various clans (see <xref ref-type="fig" rid="fig6">Figure 6</xref>), including those formed around the seven proteins found in <italic>E.coli</italic>: (i) phenylalanine-specific permease (PheP), (ii) aromatic amino acid transport protein (AroP), (iii) probable transport protein YifK, (iv) proline-specific permease (ProY), (v) <sc>d</sc>-serine/<sc>d</sc>-alanine/glycine transporter (CycA), (vi) <sc>l</sc>-asparagine permease (AnsP), (vii) GABA (4-aminobutyrate) permease (GabP). The seven clans were predicted with high probability and their clusterings confirmed by significant bootstrap values (99&#x02013;100%) except for one (92%). The analyzed dataset includes members of quite related organisms, but most clans can already be populated with further members from other species of COG1113. The algorithm predicted 257 pairs of non-orthologs, of which 254 are consistent with the phylogenetic analysis. That represents 97.7% of the 260 non-orthologous relations that can be deduced from the clan assignment. The conflicting three predictions suggest that <italic>P.aeruginosa</italic> 4a is non-orthologous to <italic>E.coli K12</italic> ProY and to <italic>E.coli H7 EDL933</italic> 4, and that <italic>P.aeruginosa</italic> 4b is non-orthologous to <italic>Yersinia pestis</italic> 4b. But here too, the extension of the phylogenetic analysis using additional sequences from the UniProtKB database supports the division of clan 4 into further subgroups (data not shown).</p></sec></sec><sec><title>CONCLUSION</title><p>We present here a new algorithm for the detection of non-orthologous relations caused by the limitations of genome-specific best hit methods, such as the COGs database. The algorithm, rather than building gene trees, a process both computationally expensive and error-prone, works with pairwise distance estimates. The accuracy of the algorithm was evaluated through verification of the distribution of predicted cases, case-by-case phylogenetic analysis and comparisons with prediction from other projects using independent methods. Using conservative parameters, the algorithm detected non-orthology in a third of the COG groups. Methods sensitive to correct orthology assignments, such as function prediction, phylogenetic trees or genome rearrangement analysis, will profit from both the algorithm and the results presented here.</p></sec><sec><title>SUPPLEMENTARY DATA</title><p>Supplementary Data are available at NAR Online.</p></sec>

Cytoplasmic expression systems triggered by mRNA yield increased gene expression in post-mitotic neurons

<p>Non-viral vectors are promising vehicles for gene therapy but delivery of plasmid DNA to post-mitotic cells is challenging as nuclear entry is particularly inefficient. We have developed and evaluated a hybrid mRNA/DNA system designed to bypass the nuclear barrier to transfection and facilitate cytoplasmic gene expression. This system, based on co-delivery of mRNA(A64) encoding for T7 RNA polymerase (T7 RNAP) with a T7-driven plasmid, produced between 10- and 2200-fold higher gene expression in primary dorsal root ganglion neuronal (DRGN) cultures isolated from Sprague&#x02013;Dawley rats compared to a cytomegalovirus (CMV)-driven plasmid, and 30-fold greater expression than the enhanced T7-based autogene plasmid pR011. Cell-free assays and <italic>in vitro</italic> transfections highlighted the versatility of this system with small quantities of T7 RNAP mRNA required to mediate expression at levels that were significantly greater than with the T7-driven plasmid alone or supplemented with T7 RNAP protein. We have also characterized a number of parameters, such as mRNA structure, intracellular stability and persistence of each nucleic acid component that represent important factors in determining the transfection efficiency of this hybrid expression system. The results from this study demonstrate that co-delivery of mRNA is a promising strategy to yield increased expression with plasmid DNA, and represents an important step towards improving the capability of non-viral vectors to mediate efficient gene transfer in cell types, such as in DRGN, where the nuclear membrane is a significant barrier to transfection.</p>

<contrib contrib-type="author"><name><surname>Farrow</surname><given-names>Paul J.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Barrett</surname><given-names>Lee B.</given-names></name><xref rid="au1" ref-type="aff">1</xref></contrib><contrib contrib-type="author"><name><surname>Stevenson</surname><given-names>Mark</given-names></name></contrib><contrib contrib-type="author"><name><surname>Fisher</surname><given-names>Kerry D.</given-names></name></contrib><contrib contrib-type="author"><name><surname>Finn</surname><given-names>Jonathan</given-names></name><xref rid="au2" ref-type="aff">2</xref></contrib><contrib contrib-type="author"><name><surname>Spice</surname><given-names>Rachel</given-names></name><xref rid="au3" ref-type="aff">3</xref></contrib><contrib contrib-type="author"><name><surname>Allan</surname><given-names>Michael A.</given-names></name><xref rid="au3" ref-type="aff">3</xref></contrib><contrib contrib-type="author"><name><surname>Berry</surname><given-names>Martin</given-names></name><xref rid="au3" ref-type="aff">3</xref></contrib><contrib contrib-type="author"><name><surname>Logan</surname><given-names>Ann</given-names></name><xref rid="au3" ref-type="aff">3</xref></contrib><contrib contrib-type="author"><name><surname>Seymour</surname><given-names>Leonard W.</given-names></name><xref ref-type="corresp" rid="cor1">*</xref></contrib><contrib contrib-type="author"><name><surname>Read</surname><given-names>Martin L.</given-names></name><xref rid="au3" ref-type="aff">3</xref></contrib><aff><institution>Department of Clinical Pharmacology, University of Oxford</institution><addr-line>Oxford OX2 6HE, UK</addr-line></aff><aff id="au1">¹<institution>Department of Anesthesia and Critical Care, Massachusetts General Hospital and Harvard Medical School</institution><addr-line>Boston, MA 02129, USA</addr-line></aff><aff id="au2">²<institution>Department of Biochemistry and Molecular Biology, University of British Columbia</institution><addr-line>Vancouver, British Columbia, Canada</addr-line></aff><aff id="au3">³<institution>Molecular Neuroscience Group, Department of Medicine, University of Birmingham</institution><addr-line>Birmingham B15 2TT, UK</addr-line></aff>

Nucleic Acids Research

<sec><title>INTRODUCTION</title><p>The concept of gene therapy is attractive and nearly two decades of research have demonstrated its potential as either a standalone treatment or an adjuvant therapy for inheritable and acquired diseases. Despite the significant advances made, there has been limited success in the clinical application of gene therapy, which is largely attributable to a lack of safe and efficient gene transfer (<xref ref-type="bibr" rid="b1">1</xref>&#x02013;<xref ref-type="bibr" rid="b3">3</xref>). Synthetic vectors based on polycations or cationic lipids are promising vectors for gene delivery as they are relatively safe and can be readily modified by the incorporation of ligands for targeting to specific cell types. However, the levels of gene expression mediated by synthetic vectors are low compared to viral vectors (<xref ref-type="bibr" rid="b4">4</xref>). A major factor restricting transgene expression is inefficient transfer of DNA from the cytoplasm to the nucleus with only 1% of polyplex DNA reaching the nucleus following cytoplasmic microinjection (<xref ref-type="bibr" rid="b5">5</xref>). The majority of cells <italic>in vivo</italic> are post-mitotic or quiescent, and transfection rates are particularly poor in these cell types as there is limited breakdown of the nuclear envelope (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b7">7</xref>). Recent improvements in non-viral vectors, such as lipid/peptide vectors (<xref ref-type="bibr" rid="b8">8</xref>) and DNA nanoparticles consisting of a single molecule of DNA (<xref ref-type="bibr" rid="b9">9</xref>), have enhanced transgene expression levels in non-dividing cells. However, it is clear that further improvements in the design of non-viral vectors are required to achieve the expression levels required for a broad range of therapeutic applications.</p><p>A widely used approach to enhance transgene expression is to modify the nucleic acid payload delivered by the non-viral vector. To promote nuclear uptake of DNA, for instance, nuclear localization sequence (NLS) peptides have been utilized to exploit intracellular transport mechanisms (<xref ref-type="bibr" rid="b10">10</xref>). However, the results to date have been largely disappointing. A 3-log increase in luciferase reporter activity was reported in one study following ligation of an oligonucleotide&#x02013;NLS conjugate to one or both ends of a linear DNA molecule, although no conclusive localization or mechanistic data was given (<xref ref-type="bibr" rid="b11">11</xref>). Whereas, we (<xref ref-type="bibr" rid="b12">12</xref>) and others (<xref ref-type="bibr" rid="b13">13</xref>) found no effect or only modest increases in gene expression following linkage of NLS peptides to DNA.</p><p>Non-viral vectors have also been used to deliver mRNA instead of plasmid DNA (<xref ref-type="bibr" rid="b14">14</xref>,<xref ref-type="bibr" rid="b15">15</xref>), or T7-based autogene plasmids utilizing bacteriophage T7 RNA polymerase (T7 RNAP) to mediate cytoplasmic expression (<xref ref-type="bibr" rid="b16">16</xref>&#x02013;<xref ref-type="bibr" rid="b18">18</xref>). The main advantage of delivering mRNA is that upon entry into the cytoplasm it is immediately translated, and so it can be used to successfully express proteins in post-mitotic, quiescent and slowly-dividing cells (<xref ref-type="bibr" rid="b14">14</xref>,<xref ref-type="bibr" rid="b19">19</xref>,<xref ref-type="bibr" rid="b20">20</xref>). The relative instability of mRNA, however, restricts the duration of gene expression and therapeutic applications have so far been limited. By comparison, the efficiency of T7-based autogene plasmids has gradually increased since their initial development over a decade ago. In most contemporary systems the cytomegalovirus (CMV) promoter has been incorporated to enable easier bacterial amplification of the plasmid. This has meant that there is now a requirement for nuclear entry to trigger the autogene, which is likely to compromise the utility of these systems in non-dividing cells (<xref ref-type="bibr" rid="b21">21</xref>,<xref ref-type="bibr" rid="b22">22</xref>).</p><p>Here, we report on the development and evaluation of a hybrid mRNA/DNA system, designed to bypass the nuclear barrier to transfection and facilitate cytoplasmic gene expression. In particular, we demonstrate that a hybrid mRNA/DNA cassette utilizing T7 RNAP mRNA represents a versatile and efficient system for mediating cytoplasmic expression in post-mitotic dorsal root ganglion neuronal (DRGN) cultures, where the nuclear membrane is a significant barrier to transfection.</p></sec><sec sec-type="materials|methods"><title>MATERIALS AND METHODS</title><sec><title>Cell lines and primary cells</title><p>Human lung carcinoma cells (A549, ATCC #CCL-185) and human prostate carcinoma cells (PC-3, ATCC #CRL-1435) were maintained in DMEM, without sodium pyruvate, with <sc>l</sc>-glutamine (PAA Laboratories, Somerset, UK) supplemented with 10% fetal calf serum (FCS, PAA Laboratories, Somerset, UK) and 1% penicillin streptomycin solution. All cells were incubated at 37&#x000b0;C in a 5% CO₂ humidified environment. DRGN cultures were prepared from adult Sprague&#x02013;Dawley rats. Briefly, spinal ganglia were removed aseptically, washed twice in Neurobasal-A media (Invitrogen, Paisley, UK) and digested in 0.125% collagenase/supplemented Neurobasal-A (2% B27 supplement, 0.5 mM <sc>l</sc>-glutamine and 0.5% gentimicin) for 2 h at 37&#x000b0;C. Disassociated spinal ganglia cells were removed from the collagenase, washed in Neurobasal-A and purified by centrifugation for 10 min at 120 g on a 15% BSA column.</p></sec><sec><title>Sources of nucleic acids</title><p>Plasmid DNA was grown in <italic>Escherichia coli</italic> and purified using endotoxin-free Qiagen maxiprep kits (Crawley, West Sussex, UK). The concentration and purity of DNA was checked on a spectrophotometer at A₂₆₀ and A₂₈₀ absorbance wavelengths. The reporter gene construct pCMVLuc1 was a gift from Dr Manfred Ogris (Munich, Germany). The design and construction of cytoplasmic expression plasmids used in this study are shown in <xref ref-type="fig" rid="fig1">Figure 1A</xref> and Supplementary Figures 1&#x02013;2. Briefly, pCMV/T7-T7pol was constructed as described by Brisson <italic>et al</italic>. (<xref ref-type="bibr" rid="b21">21</xref>); pCMV/T7-T7pol(A64) is a modified version of pCMV/T7-T7pol into which an oligonucleotide containing 64 adenylate residues was inserted into the BamHI&#x02013;NotI restriction sites downstream from the T7 RNAP gene; pT7Luc, also known as pEMCLuc&#x003b2;gAn (<xref ref-type="bibr" rid="b17">17</xref>), contains the <italic>Photinus pyralis</italic> luciferase gene flanked upstream by the T7 promoter, and pR011 contains the T7 RNAP gene driven by CMV and triple phage promoters (T7, T3 and SP6), and the luciferase gene driven by a T7 promoter (<xref ref-type="bibr" rid="b22">22</xref>).</p></sec><sec><title>Production of mRNA</title><p><italic>In vitro</italic> transcription to produce capped mRNA encoding T7 RNAP protein was performed using the Ribomax&#x02122; large scale mRNA production system (Promega, Southampton, UK), with the addition of either a standard m7G(5&#x02032;)ppp(5&#x02032;)G cap, or anti-reverse cap analogue (Ambion) substituted for a portion of the rGTP at a 4:1 ratio, as recommended by the manufacturer. A poly(A) tailing kit (Ambion) was used as indicated to add a &#x0003e;200 base poly(A) tail to cell-free transcribed mRNA prior to purification. The synthesis of mRNA encoding enhanced green fluorescent protein (EGFP) and luciferase has been described previously (<xref ref-type="bibr" rid="b20">20</xref>). Products were checked by denaturing gel electrophoresis.</p></sec><sec><title>Cell-free translation systems</title><p>The standard rabbit reticulocyte lysate (RRL) cell-free system (Promega) was used to characterize translation products from mRNA encoding T7 RNAP protein. Reactions were incubated at 30&#x000b0;C for 90 min, prior to storage at &#x02212;80&#x000b0;C and performed following the manufacturer's instructions. The cell-free RRL translation system was also used in a standard cell-free transcription/translation system to analyse expression of T7 RNAP from mRNA by subsequent expression and detection of luciferase protein from plasmid pT7Luc. T7 RNAP protein was also added with pT7Luc to the RRL cell-free system and levels of luciferase activity detected. Reaction mixtures were incubated at 30&#x000b0;C and 5 &#x003bc;l aliquots removed at specific time points and assayed for luciferase activity with results expressed in relative light units (RLU).</p></sec><sec><title>Quantitative RT&#x02013;PCR analysis of mRNA</title><p>A quantitative RT&#x02013;PCR assay was used to quantify intracellular levels of T7 RNAP mRNA at specific time points post-transfection. In brief, total RNA was extracted using the RNeasy mini kit (Qiagen) from A549 cells co-transfected with T7 RNAP mRNA (100 ng) and pT7Luc (400 ng). A total of 150 ng of total RNA was reverse transcribed to cDNA using random hexamer primers and multiscribe reverse transcriptase at 48&#x000b0;C for 30 min. Following addition of T7 forward primer (5&#x02032;-TCACGACTCCTTCGGTACCAT-3&#x02032;), T7 reverse primer (5&#x02032;-CATAGTTTCGCGCACTGCTTT-3&#x02032;), T7 probe labelled with 5&#x02032; FAM and 3&#x02032; TAMRA (5&#x02032;-CGGCTGACGCTGCGAACCTGTT-3&#x02032;) and qRT&#x02013;PCR mastermix, 40 cycles of a two step thermal cycling protocol (denature: 95&#x000b0;C, 15 s; anneal/extend: 60&#x000b0;C, 1 min) was performed using the Applied Biosystems 7000 Sequence Detection System. 18S Ribosomal RNA was detected using appropriate primers and probe (Applied Biosystems) to demonstrate that a similar amount of total RNA was used in each qRT&#x02013;PCR. The amount of T7 RNAP mRNA present in each sample was then calculated from a T7 RNAP mRNA standard curve generated by performing qRT&#x02013;PCR on lysate samples prepared from mock-transfected cells spiked with a known amount of T7 RNAP mRNA (equivalent to 0.1, 1, 10 and 100 ng mRNA per well).</p></sec><sec><title>Transfection studies</title><p>Plasmid DNA and/or mRNA were added to a polypropylene microcentrifuge tube at a final concentration of 50 &#x003bc;g/ml in 10 mM HEPES&#x02013;NaOH (pH 7.4). When mRNA was used, aliquots were thawed from &#x02212;80&#x000b0;C, heated at 80&#x000b0;C for 10 min and chilled on ice for 2 min prior to addition. DOTAP was then added to nucleic acids, unless otherwise stated, at a (w/w) ratio of five and mixed by pipetting the solution up and down &#x000d7;5 prior to use. Lipoplexes were added directly to a 48-well plate containing 3&#x02013;4 &#x000d7; 10⁴ cells per well in 120 &#x003bc;l DMEM or optimem without FCS or supplement (cells were plated 24 h before transfection). Freshly isolated DRGN cultures were seeded at a density of &#x0223c;1500 cells per well in supplemented Neurobasal-A. After 4 h, the medium was discarded and replaced with 500 &#x003bc;l per well of either DMEM containing 10% FCS, or supplemented Neurobasal-A. Cells were cultured for various times prior to analysis of reporter gene expression as indicated. Cell viability was determined using the MTS assay (Promega) after exposure to free DOTAP or lipoplexes as indicated and normalized to values obtained in their absence as described previously (<xref ref-type="bibr" rid="b23">23</xref>).</p></sec><sec><title>Assay of reporter genes</title><p>Luciferase expression following transfection was measured by a luminescence assay using cell lysates. The culture medium was discarded and cell lysates harvested after incubation of cells for 30 min at &#x02212;80&#x000b0;C in 100 &#x003bc;l lysis reagent buffer (Promega) before being thawed at RT. The lysate was mixed well by pipetting and 20 &#x003bc;l was diluted into 25&#x02013;100 &#x003bc;l of luciferase assay reagent (Promega). The luminescence was integrated over 10 s on a Victor² 1420 Multilabel Counter (Wallac, Bucks, UK) or a Lumat LB9507 (Berthold Instruments, UK) and the results expressed as RLU per mg of cell protein, determined using the Advanced Protein Assay (Cytoskeleton, Denver). Analysis of EGFP expression was carried out on a Coulter Epics XL flow cytometer. Cells were trypsinized at appropriate times after transfection, washed with phosphate-buffered saline (PBS) and then fixed in 2% paraformaldehyde. EGFP was excited using the 488 nm line of an Argon laser and emitted light collected at 520 nm (green fluorescence) and 575 nm (red fluorescence) to enable correction for autofluorescence by diagonal gating. Background fluorescence and autofluorescence were determined using mock treated cells. The software programme WinMDI was used to analyse data and expressed as the percentage of EGFP-positive cells. Statistical analysis was performed on at least three samples for each transfection by calculating the mean value, SD and using an unpaired <italic>t</italic>-test where appropriate (<ext-link ext-link-type="uri" xlink:href="www.graphpad.com/quickcalcs/ttest1.cfm"/>).</p></sec><sec><title>Western blot analysis</title><p>Cells were lysed in M-PER cell lysis buffer (100 &#x003bc;l, Perbio Science, Northumberland, UK) containing complete protease inhibitor cocktail (Roche, East Sussex, UK). The lysates were normalized for protein concentration using the Advanced Protein Assay (Cytoskeleton) and stored at &#x02212;70&#x000b0;C until used for western blot analysis. Each sample (20 &#x003bc;g total protein) was heated at 90&#x000b0;C for 5 min and separated on a 9% SDS&#x02013;polyacrylamide gel (Invitrogen). Proteins were transferred to nitrocellulose membranes overnight and blocked for 90 min at room temperature in PBS containing 5% non-fat milk. Membranes were blotted with the relevant primary antibody for 90 min, washed in PBS containing 0.1% Tween and the secondary antibody added. For detection LumiGLO^&#x000ae; chemiluminescent substrate (Upstate) was added and chemiluminescence detected using the MultiImage&#x02122; light cabinet and Fluorchem 8800 software (Alpha Innotech, CA).</p></sec></sec><sec><title>RESULTS</title><sec><title>Production and analysis of T7 RNAP mRNA</title><p>Construction of pCMV/T7-T7pol and insertion of a 64 bp oligonucleotide to prepare pCMV/T7-T7pol(A64) are outlined in <xref ref-type="fig" rid="fig1">Figure 1A</xref> and Supplementary Figure 1. We first prepared mRNA encoding T7 RNAP protein from NotI linearized pCMV/T7-T7pol and pCMV/T7-T7pol(A64) to investigate whether it can trigger expression of co-delivered plasmid DNA, and improve the efficiency of T7-based cytoplasmic plasmids. Denaturing electrophoretic analysis showed that pCMV/T7-T7pol(A64) produced mRNA of the expected size (2899 nt) that was slightly longer than the transcript derived from pCMV/T7-T7pol (2880 nt, <xref ref-type="fig" rid="fig1">Figure 1B</xref>). To confirm that transcribed mRNA encoded T7 RNAP protein, T7 RNAP mRNA containing the 64 bp poly(A) sequence was translated utilizing the RRL system. Immunoblotting for the T7 RNAP protein using a mouse anti-T7 RNAP monoclonal IgG antibody gave a band of &#x0223c;100 kDa that was similar in size to recombinant T7 RNAP protein, indicating that T7 RNAP mRNA had been translated successfully (<xref ref-type="fig" rid="fig1">Figure 1C</xref>).</p><p>A cell-free transcription/translation assay incorporating the reporter plasmid pT7Luc was performed next to confirm that T7 RNAP protein translated from mRNA was biologically functional. Incubation of pT7Luc in the presence of T7 RNAP protein resulted in luciferase expression that was dose-dependent upon the amount of T7 RNAP protein added (<xref ref-type="fig" rid="fig2">Figure 2A</xref>). The addition of 100 U of T7 RNAP protein, for instance, generated a maximum of &#x0223c;1.7 &#x000d7; 10⁴ RLU after 390 min. By comparison, plasmid pT7Luc incubated with 1 &#x003bc;g of T7 RNAP mRNA gave &#x0223c;100-fold higher level of luciferase expression (1.7 &#x000d7; 10⁶ RLU, <xref ref-type="fig" rid="fig2">Figure 2B</xref>). These results demonstrated that T7 RNAP protein translated from T7 RNAP mRNA was functional and mediated luciferase expression at levels up to 100-fold greater than using T7 RNAP protein directly. However, it is interesting to note that after shorter times T7 RNAP protein mediated higher expression levels than T7 RNAP mRNA. For example, after 30 min, expression driven by T7 RNAP protein (2.0 &#x000d7; 10³ RLU) was 10-fold greater than T7 RNAP mRNA (2.0 &#x000d7; 10² RLU, <xref ref-type="fig" rid="fig2">Figure 2</xref>). This is likely to reflect the requirement for translation of T7 RNAP mRNA to occur prior to transcription of the luciferase gene.</p></sec><sec><title>Co-delivery of mRNA encoding T7 RNAP protein enhances transgene expression</title><p>Having established the functional activity of mRNA encoding T7 RNAP in a cell-free assay, we next evaluated the efficiency of T7 RNAP mRNA to trigger luciferase expression from plasmid pT7Luc in A549 and PC-3 cells. Initial transfections showed that the level of luciferase expression increased with the dose of T7 RNAP mRNA, generating a maximum of 1.0 &#x000d7; 10⁸ RLU/mg protein in A549 cells and 1.9 &#x000d7; 10⁸ RLU/mg protein in PC-3 cells (<xref ref-type="fig" rid="fig3">Figure 3A</xref>). By comparison, only background levels of luciferase expression (&#x0223c;1.0 &#x000d7; 10⁴ RLU/mg protein in A549 cells) were detected in the absence of T7 RNAP mRNA. The amount of pT7Luc used to transfect A549 and PC-3 cells was also an important determinant of luciferase expression. Expression mediated by 125 ng of T7 RNAP mRNA increased with the dose of pT7Luc up to the maximum amount of 500 ng added to A549 (&#x0223c;9.4 &#x000d7; 10⁷ RLU/mg protein) and PC-3 (&#x0223c;2.3 &#x000d7; 10⁸ RLU/mg protein) cells (<xref ref-type="fig" rid="fig3">Figure 3B</xref>). Doubling the quantity of plasmid pT7Luc from 250 to 500 ng resulted in a 4-fold increase in expression in A549 cells and a 6-fold increase in PC-3 cells.</p><p>Transfections were next performed to directly compare the efficiency of T7 RNAP mRNA with T7 RNAP protein to mediate luciferase expression from pT7Luc. Preliminary experiments indicated that 10&#x02013;50 U of T7 RNAP protein were required to produce maximum expression levels with 500 ng of pT7Luc in A549 (&#x0223c;3.0 &#x000d7; 10⁶ RLU/mg protein) and PC-3 (&#x0223c;2.0 &#x000d7; 10⁷ RLU/mg protein) cells (<xref ref-type="fig" rid="fig3">Figure 3C</xref>). Subsequent transfections to compare T7 RNAP mRNA with T7 RNAP protein in the same experiment showed that co-delivery of 125 ng T7 RNAP mRNA with 500 ng of pT7Luc gave 5.2-fold and 6.7-fold greater luciferase expression compared to T7 RNAP protein at a dose of 100 U in A549 and PC-3 cells, respectively (Supplementary Figure 3). Taken together, these results demonstrate the superior transfection properties of cytoplasmic expression systems triggered by T7 RNAP mRNA compared to T7 RNAP protein.</p></sec><sec><title>Parameters influencing the efficiency of the T7 RNAP mRNA/pT7Luc expression system</title><p>We next examined several parameters that have been reported to affect the efficiency of nucleic acid-based expression systems, including the structure of mRNA, intracellular stability, time course of delivery and cytotoxicity (<xref ref-type="bibr" rid="b4">4</xref>,<xref ref-type="bibr" rid="b24">24</xref>,<xref ref-type="bibr" rid="b25">25</xref>). First, we compared the efficiency of T7 RNAP mRNA containing different structural components to trigger luciferase expression from co-delivered pT7Luc in A549 and PC-3 cells. <xref ref-type="fig" rid="fig4">Figure 4A</xref> shows that T7 RNAP mRNA modified with either m7G(5&#x02032;)ppp(5&#x02032;)G (Cap), or an anti-reverse cap analogue (ARCA), and a poly(A64) tail produced the highest levels of luciferase gene expression (&#x0223c;1.2 &#x000d7; 10⁷ RLU/mg protein in A549 cells). However, reporter gene expression was abolished by &#x0223c;80% corresponding to a 4.8&#x02013;5.6-fold decrease using capped T7 RNAP mRNA lacking a poly(A64) tail. The importance of the poly(A) tail was further demonstrated by a 2.2 to 3.0-fold increase in gene expression when a poly(A) tail of &#x0223c;200 residues was added to capped T7 RNAP mRNA lacking the A64 tail with a poly(A) tailing kit (<xref ref-type="fig" rid="fig4">Figure 4A</xref>).</p><p>To evaluate the intracellular stability and persistence of T7 RNAP protein and pT7Luc, we transfected A549 cells with the relevant nucleic acid components at different time points. For instance, expression levels were abolished by &#x0003e;240-fold (4.9 &#x000d7; 10⁴ RLU/mg protein) when A549 cells were transfected with 500 ng pT7Luc 12 h prior to transfection with 125 ng T7 RNAP mRNA compared to that detected when both T7 RNAP mRNA and pT7Luc were transfected together (&#x0223c;1.2 &#x000d7; 10⁷ RLU/mg protein, <xref ref-type="fig" rid="fig4">Figure 4B</xref>). In contrast, luciferase expression (7.7 &#x000d7; 10⁶ RLU/mg protein) detected when T7 RNAP mRNA transfection preceded pT7Luc transfection was still &#x0223c;65% of that when both nucleic acids were transfected together, indicating that the half-life of the T7 RNAP protein was at least 12 h (<xref ref-type="fig" rid="fig4">Figure 4B</xref>). This is in agreement with published studies that have estimated the half-life of T7 RNAP protein in 293 cells to be &#x0003e;24 h (<xref ref-type="bibr" rid="b26">26</xref>). A qRT&#x02013;PCR assay was next developed to evaluate the persistence of T7 RNAP mRNA in A549 cells and used to calculate the quantity of T7 RNAP mRNA present in transfected cultures over a time course of 24 h. <xref ref-type="fig" rid="fig4">Figure 4C</xref> shows that the signal from T7 RNAP mRNA decreased over time (exponential decay) with an estimated half-life in A549 cells of 4.79 h. Finally, cell viability assays using A549 cells showed that transfection with T7 RNAP mRNA (100 ng) complexed with DOTAP or DOTAP alone gave negligible toxicity (<xref ref-type="fig" rid="fig4">Figure 4D</xref>). Co-delivery of T7 RNAP mRNA with pT7Luc also gave minimal toxicity with between 88 and 91% cells remaining viable after 24 h, depending on the dose of plasmid DNA, compared to mock-transfected cells.</p><p>Altogether, these results characterize a number of parameters that can influence the efficiency of the T7 RNAP mRNA/pT7Luc expression system. In particular, the intracellular stability of plasmid DNA appears to be the most important determinant in limiting gene expression, and not the duration of expression or stability of the T7 RNAP protein.</p></sec><sec><title>Enhanced transfection efficiency of mRNA triggered systems in DRGN cultures</title><p>We next examined whether cytoplasmic expression systems triggered by mRNA can efficiently transfect primary DRGN cultures, where the majority of cells are likely to be post-mitotic or slowly-dividing. In preliminary experiments, fluorescence microscopy and flow cytometry were used to evaluate the ability of mRNA or plasmid DNA encoding EGFP to transfect DRGN cultures freshly isolated from adult Sprague&#x02013;Dawley rats. <xref ref-type="fig" rid="fig5">Figure 5A and B</xref> show that a high proportion (&#x0003e;50%) of phenotypically identifiable DRGN in culture were transfected with EGFP mRNA. In contrast, no EGFP-positive cells were detected using mRNA encoding for luciferase (LUC mRNA) as a control (<xref ref-type="fig" rid="fig5">Figure 5A</xref>). Furthermore, we observed &#x0003c;2% EGFP-positive cells in DRGN cultures with the CMV-driven plasmid pEGFPN1 (<xref ref-type="fig" rid="fig5">Figures 5B</xref> and <xref ref-type="fig" rid="fig6">6</xref>).</p><p>The transfection efficiency of T7 RNAP mRNA with pT7Luc was then compared against the CMV-driven plasmid pCMVLuc1 in DRGN cultures and A549 cells. In primary DRGN cultures, pCMVLuc1 was a particularly inefficient vector with luciferase expression no higher than that seen in non-transfected cells (&#x0223c;1.2 &#x000d7; 10⁴ RLU/mg protein; <xref ref-type="fig" rid="fig5">Figure 5C</xref>). By comparison, transfection of DRGN cultures with the T7 RNAP mRNA triggered system gave &#x0003e;100-fold increase in luciferase activity (1.3 &#x000d7; 10⁶ RLU/mg protein; <italic>P</italic> = 0.0185). This level of reporter gene expression was similar to that obtained using the same system in A549 cells (4.2 &#x000d7; 10⁶ RLU/mg protein). In contrast, in rapidly dividing A549 cells, where nuclear access was relatively easy the functionality of pCMVLuc1 was evident, with high luciferase activity (7.8 &#x000d7; 10⁶ RLU/mg protein) (<xref ref-type="fig" rid="fig5">Figure 5C</xref>). Further analysis using a more sensitive luciferase assay showed that transfection of DRGN cultures with the T7 RNAP mRNA triggered system gave &#x0003e;2200-fold increase in luciferase activity (2.2 &#x000d7; 10⁸ RLU/mg) compared to pCMVLuc1 (9.9 &#x000d7; 10⁴ RLU/mg, Supplementary Figure 4). Evaluation of the time course of gene expression in DRGN cultures showed significant luciferase activity with the T7 RNAP mRNA triggered system after 6 h that increased up to 24 h (<xref ref-type="fig" rid="fig5">Figure 5D</xref>). At all time points examined the T7 RNAP mRNA triggered system produced significantly greater luciferase activity than pCMVLuc1 in the range of 13- to 634-fold (<italic>P</italic> &#x0003c; 0.05, <xref ref-type="fig" rid="fig5">Figure 5D</xref>).</p><p>Previously, pCMVLuc1 has been shown to be highly active in primary rat oligodendrocytes when delivered using a particle bombardment gene delivery approach, which suggests that in this study the poor activity of the plasmid alone was due to limited access to the nucleus rather than inactivity of the CMV promoter or the pCMVLuc1 vector in rat cells (<xref ref-type="bibr" rid="b27">27</xref>). Hence, these results demonstrate that cytoplasmic expression systems triggered by mRNA are efficient in transfecting post-mitotic or slowly-dividing cells, especially in cell types that are resistant to gene transfer using conventional CMV promoter driven vectors requiring nuclear entry.</p></sec><sec><title>Improved efficiency of T7-based autogene systems with T7 RNAP mRNA</title><p>We next investigated whether the efficiency of cytoplasmic expression systems based on autogene plasmids containing the T7 RNAP gene, which enable autocatalytic amplification of the T7 RNAP protein, were also improved by co-delivery of T7 RNAP mRNA. For these experiments, we used an enhanced autogene plasmid pR011 (Supplementary Figure 2), which was recently developed by Finn <italic>et al</italic>. (<xref ref-type="bibr" rid="b22">22</xref>). Enhancements to plasmid pR011 included an IRES sequence located between the bacteriophage promoters and the T7 RNAP gene, and the incorporation of a luciferase expression cassette. Using plasmid pR011 in BHK cells, Finn <italic>et al</italic>. (<xref ref-type="bibr" rid="b22">22</xref>) have demonstrated levels of reporter gene expression that were 20-fold higher than standard CMV-based nuclear expression systems. In addition, direct evidence was given for an exponential, autocatalytic increase in gene expression using autogene-based plasmids. However, pR011 is still limited by a requirement for nuclear entry to drive initial T7 RNAP production to trigger the autogene.</p><p>The schematic in <xref ref-type="fig" rid="fig7">Figure 7A</xref> outlines the principle of co-delivering T7 RNAP mRNA to trigger expression of the luciferase reporter gene from cytoplasmic expression plasmids pR011 and pT7Luc. The ability of T7 RNAP mRNA to improve the transfection efficiency of pR011 was evident in A549 cells with a 2.7-fold (<italic>P</italic> = 0.0026) and 3.4-fold increase (<italic>P</italic> &#x0003c; 0.0001) in luciferase expression observed after 24 and 48 h, respectively, compared to pR011 alone (<xref ref-type="fig" rid="fig7">Figure 7B</xref>). Transfection of primary DRGN cultures with pR011 in the presence of T7 RNAP mRNA also increased luciferase expression after 24 h (4.8-fold) compared to pR011 alone (<italic>P</italic> = 0.0131, <xref ref-type="fig" rid="fig7">Figure 7C</xref>). By comparison, transfection of DRGN cultures with pT7Luc and T7 RNAP mRNA gave &#x0003e;30-fold higher level of luciferase expression than pR011 (<italic>P</italic> &#x0003c; 0.05, <xref ref-type="fig" rid="fig7">Figure 7C</xref>). These results demonstrate that T7 RNAP mRNA can also be used successfully to trigger T7-based autogene plasmids and increase gene expression in both rapidly dividing cell lines and primary cultures of post-mitotic DRGN. Although in DRGN cultures combining a simple T7-driven expression plasmid with T7 RNAP mRNA gave significantly greater gene expression than the T7-based autogene plasmid, either with or without T7 RNAP mRNA.</p></sec></sec><sec><title>DISCUSSION</title><p>A prerequisite for gene transfer with synthetic vectors is the efficient delivery of exogenous nucleic acids to the nucleus. This goal is challenging as nuclear entry is particularly inefficient in post-mitotic or slowly-dividing cells since there is limited breakdown of the nuclear envelope (<xref ref-type="bibr" rid="b6">6</xref>,<xref ref-type="bibr" rid="b7">7</xref>). In this study, we used a hybrid mRNA/DNA system to facilitate cytoplasmic gene expression and bypass problems associated with the nuclear barrier to transfection. In primary DRGN cultures, for example, co-delivery of T7 RNAP mRNA with pT7Luc produced between 10- and 2200-fold higher gene expression compared to the CMV-driven plasmid, pCMVLuc1 and &#x0003e;30-fold increase compared to an enhanced T7-autogene plasmid pR011. In addition, only a small quantity of mRNA was required to mediate significant gene expression. In A549 cells, for instance, co-delivery of 10 ng of T7 RNAP mRNA with pT7Luc was sufficient to mediate a 3-log increase in luciferase activity. These properties demonstrate that the combination of T7 RNAP mRNA with a T7-driven DNA cytoplasmic expression plasmid represents a powerful and versatile non-viral approach for mediating gene expression.</p><p>The co-delivery of plasmid DNA with T7 RNAP protein is one of the simplest and most widely used systems to achieve cytoplasmic gene expression. This approach, however, has been shown to be relatively inefficient even in cultured cells (<xref ref-type="bibr" rid="b28">28</xref>), with low levels of protein transduction likely to be a limiting factor. Another potential problem is immune recognition of the T7 RNAP protein that will hinder repeated administration <italic>in vivo</italic> (<xref ref-type="bibr" rid="b21">21</xref>). Therefore, the major advantages of using the hybrid mRNA/DNA system are that problems of inefficient protein transduction and immune responses associated with repeat administration of the T7 RNAP protein will be diminished as it is a non-viral formulation based solely on nucleic acids. Preliminary results from a cell-free assay demonstrated that co-incubation of T7 RNAP mRNA with pT7Luc was more efficient than using T7 RNAP protein, with up to a 100-fold increase in luciferase expression. This trend was repeated in A549 cells and primary DRGN cultures (data not shown) with up to a 6.7-fold higher level of expression achieved by co-delivery of pT7Luc with T7 RNAP mRNA rather than T7 RNAP protein. It is likely that amplification of the T7 RNAP protein by multiple rounds of translation from T7 RNAP mRNA also contributed towards mediating higher gene expression compared to using a static quantity of T7 RNAP protein.</p><p>We also used T7 RNAP mRNA instead of plasmid DNA encoding the T7 RNAP gene as it does not require nuclear uptake for expression, and translation occurs almost immediately upon entry into the cytoplasm. This is especially important in post-mitotic or slowly-dividing cells where access of plasmid DNA to the nucleus will be limited. The incorporation of mRNA into the hybrid expression system therefore enables cytoplasmic expression of co-delivered plasmid DNA in a larger proportion of cells compared to a nuclear dependent plasmid. Evidence for this comes from a number of studies where EGFP mRNA was shown to transfect a higher percentage of cells, such as primary HUVEC (<xref ref-type="bibr" rid="b20">20</xref>) and PC-3 cells (<xref ref-type="bibr" rid="b29">29</xref>), compared to an EGFP expression plasmid. We also demonstrated in this study that plasmid pEGFPN1 gave poor levels of transfection in DRGN cultures, whereas a significantly higher proportion of cells were transfected with EGFP mRNA.</p><p>T7-based autogene systems have the capacity to mediate higher levels of gene expression since, after initial triggering of the T7-autogene by T7 RNAP, further expression of T7 RNAP occurs in an autocatalytic positive feedback loop producing high levels of the protein (<xref ref-type="bibr" rid="b16">16</xref>). Subsequently, the large quantities of T7 RNAP produced can be used to mediate expression of a co-delivered therapeutic gene. The efficiency of T7-based autogene plasmids has gradually increased since their initial development. Incorporation of the CMV promoter into the autogene plasmid has enabled easier bacterial amplification of the plasmid, and offered a new triggering system, independent of protein transduction (<xref ref-type="bibr" rid="b21">21</xref>,<xref ref-type="bibr" rid="b22">22</xref>). While this approach was shown to be effective in rapidly dividing 293 cells, there is now a requirement for nuclear entry to trigger the autogene, which is likely to compromise the utility of these plasmids in non-dividing cells.</p><p>A significant finding in this study was that transfection of DRGN cultures with pT7Luc and T7 RNAP mRNA gave &#x0003e;30-fold higher gene expression than the enhanced T7-autogene plasmid pR011. We also showed in DRGN cultures that T7 RNAP mRNA can be used to enhance gene expression mediated by pR011 with up to a 4.8-fold increase in luciferase activity. By comparison, the enhancement in expression with T7 RNAP mRNA was slightly lower in A549 cells (3-fold), which may reflect the higher transfection activity of pR011 in these cells due to increased nuclear entry and subsequent triggering of the CMV promoter. The hybrid T7 RNAP mRNA/DNA system contains important features that will have contributed towards the superior transfection profile in primary DRGN cultures. First, the T7 RNAP mRNA was capped and polyadenylated, and so would have been translated more efficiently and have a longer half-life than autogene derived, IRES containing transcripts. Finn <italic>et al</italic>. (<xref ref-type="bibr" rid="b30">30</xref>) recently showed that IRES mediated translation was relatively inefficient compared to the cap dependent form, with luciferase expression from IRES containing mRNA almost 20-fold lower than from an equivalent capped transcript. Furthermore, T7 RNAP protein produced by mRNA translation drives efficient expression from a co-delivered reporter plasmid almost immediately, without the need for mRNA amplification to occur. This may provide a critical time advantage since degradation of plasmid DNA by cytoplasmic nucleases limits cytoplasmic expression (<xref ref-type="bibr" rid="b5">5</xref>,<xref ref-type="bibr" rid="b25">25</xref>).</p><p>The major factor that appeared to limit expression with the hybrid T7 RNAP mRNA/DNA system was the quantity of plasmid DNA available as a cytoplasmic transcription template, rather than T7 RNAP protein produced from T7 RNAP mRNA. Doubling the quantity of pT7Luc transfected from 250 to 500 ng, for instance, resulted in a 6-fold increase in gene expression in PC-3 cells. The inclusion of an autogene expression cassette within the formulation was not required as a small quantity of T7 RNAP mRNA was sufficient to drive optimal expression from co-transfected plasmid DNA. The challenges now lie in improving the duration of transgene expression and persistence of plasmid DNA, and we are currently exploring different strategies to address these issues, including the use of CpG-depleted plasmids to prolong expression <italic>in vivo</italic> (<xref ref-type="bibr" rid="b31">31</xref>&#x02013;<xref ref-type="bibr" rid="b34">34</xref>). It will also be necessary to determine the extent to which intracellular expression of the T7 RNAP protein leads to potential immunogenicity problems or whether the use of a nucleic acid formulation is sufficient to diminish immune responses to the T7 RNAP protein, and so maintain the viability of transfected cells without the use of immunosuppressants (<xref ref-type="bibr" rid="b35">35</xref>).</p><p>Based on our findings in primary DRGN cultures, potential therapeutic applications for this hybrid expression system include the delivery of neurotrophic factors to promote axonal regeneration, in possibly a range of neurological conditions, from spinal cord injury to chronic progressive neurodegenerative diseases. Recently, we demonstrated that an effective strategy to sustain gene delivery to axotomized neurons was to immobilize plasmid DNA in gene-activated matrices (GAM) that were placed between the proximal and distal stumps of severed rat optic nerves (<xref ref-type="bibr" rid="b36">36</xref>) or into a lesioned rat dorsal column (<xref ref-type="bibr" rid="b37">37</xref>). Hence, we envisage that GAM-mediated delivery of hybrid expression systems encoding for neurotrophic factors will represent a more effective strategy to enhance DRGN axonal regeneration <italic>in vivo</italic> than the use of CMV-driven plasmid DNA. Furthermore, the central nervous system has historically been designated as an &#x02018;immunologically privileged&#x02019; site, as it lacks normal surveillance by cells and mediators of the immune system. The central nervous system therefore represents an ideal site for delivery of non-viral vectors containing the hybrid mRNA/DNA expression system, especially when potentially immunogenic proteins are expressed.</p><p>In summary, the results from this study demonstrate that co-delivery of mRNA is a promising strategy to yield increased expression with plasmid DNA and facilitate cytoplasmic gene expression. Furthermore, this work represents an important step towards the development of non-viral formulations based on a hybrid mRNA/DNA system that should prove useful for the expression of proteins in post-mitotic or slowly-dividing cells, where the nuclear barrier represents a significant barrier to transfection.</p></sec><sec><title>SUPPLEMENTARY DATA</title><p>Supplementary Data are available at NAR Online.</p></sec>

Race, Ethnicity, and Linguistic Isolation as Determinants of Participation in Public Health Surveillance Surveys

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Link</surname><given-names>Michael W</given-names></name><degrees>PhD</degrees><aff>Centers for Disease Control and Prevention</aff><address><email>mlink@cdc.gov</email><addr-line>4770 Buford Hwy NE, Mail Stop K-66, Atlanta, GA 30341</addr-line><phone>770-488-5444</phone></address></contrib><contrib contrib-type="author"><name><surname>Mokdad</surname><given-names>Ali H</given-names></name><degrees>PhD</degrees><aff>Centers for Disease Control and Prevention, Atlanta, Ga</aff></contrib><contrib contrib-type="author"><name><surname>Stackhouse</surname><given-names>Herbert F</given-names></name><degrees>MA</degrees><aff>Centers for Disease Control and Prevention, Atlanta, Ga</aff></contrib><contrib contrib-type="author"><name><surname>Flowers</surname><given-names>Nicole T</given-names></name><degrees>MD</degrees><aff>Centers for Disease Control and Prevention, Atlanta, Ga</aff></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>Reducing racial and ethnic disparities in health is an overarching goal of <italic>Healthy People 2010</italic> (<xref rid="B1" ref-type="bibr">1</xref>). To reach this goal, however, public health officials require valid and reliable data from health surveillance to plan, implement, and evaluate programs designed to improve health conditions among racial and ethnic minority populations (<xref rid="B2" ref-type="bibr">2</xref>-<xref rid="B4" ref-type="bibr">4</xref>). For instance, health surveillance efforts have highlighted racial and ethnic disparities in health conditions such as cardiovascular disease, hypertension, diabetes, certain cancers (e.g., colon and rectal, pancreatic, stomach), and nationally notifiable diseases (e.g., chlamydia, gonorrhea, salmonelosis) and in risk factors for chronic conditions such as physical inactivity, excessive alcohol consumption, and cigarette smoking (<xref rid="B5" ref-type="bibr">5</xref>-<xref rid="B11" ref-type="bibr">11</xref>). Despite the success of many surveillance efforts, monitoring chronic disease and behavioral risk factors among minority populations remains a challenge.</p><p>The proportion of racial and ethnic minorities who participate in major health surveys is often lower than the proportion for the overall U.S. population. Some of the reasons for lower rates of participation among racial and ethnic minorities include disproportionate mistrust of government and the research community, cultural and language barriers, lower rates of literacy and health literacy (the capacity to obtain, process, and understand basic health information and services needed to make appropriate health decisions), high mobility patterns, reluctance to reveal personal information, and data-collection procedures (e.g., characteristics of the interviewers) (<xref rid="B12" ref-type="bibr">12</xref>-<xref rid="B26" ref-type="bibr">26</xref>). Because minority groups, particularly groups of lower socioeconomic status, may be underrepresented in public health statistics generated by these surveys, the health risks and health problems that they face may be inadequately described.</p><p>The potential for such problems has increased over the past several decades as the U.S. population has grown more diverse. From 1980 to 2002, according to the U.S. Census Bureau, the proportion of minorities among the civilian, noninstitutionalized population grew from 6.4% to 13.3% among Hispanics, 11.7% to 13.0% among blacks, and 1.5% to 4.4% among Asians (<xref rid="B27" ref-type="bibr">27</xref>-<xref rid="B30" ref-type="bibr">30</xref>). Additionally, as of 2002, 11.7% of U.S. residents were reported to have been born in a foreign country, with 53.3% of those saying they had been born in Latin America, 25.0% in Asia, 13.7% in Europe, and 8.0% in some other region of the world (<xref rid="B31" ref-type="bibr">31</xref>). Moreover, the various racial and ethnic groups are not distributed equally across the United States (Figures 1&#x02013;3). As a result, the potential impact of race and ethnicity on survey participation rates varies considerably among and within regions.</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure 1</label><caption><p>Percentage of Hispanic or Latino adults aged 18 years and older, United States. Source: U.S. Census 2000 (<xref rid="B32" ref-type="bibr">32</xref>).</p></caption><alt-text>Map of the United States showing the percentage distributions of Hispanic or Latino adults aged 18 years and older. The highest percentages (58.72% to 97.23%) are found primarily in the southwestern United States and southern Florida.</alt-text><graphic xlink:href="PCD31A09s01" position="float"/></fig></boxed-text><boxed-text position="float"><fig position="float" id="F2" fig-type="diagram"><label>Figure 2</label><caption><p>Percentage of black or African American adults aged 18 years and older, United States. Source: U.S. Census 2000 (<xref rid="B32" ref-type="bibr">32</xref>).</p></caption><alt-text>Map of the United States showing percentage of black or African American adults aged 18 and older, United States. The greatest percentages (48.47%-84.34%) are located primarily in the southeastern United States.</alt-text><graphic xlink:href="PCD31A09s02" position="float"/></fig></boxed-text><boxed-text position="float"><fig position="float" id="F3" fig-type="diagram"><label>Figure 3</label><caption><p>Percentage of Asian or Pacific Islander adults aged 18 years and older, United States. Source: U.S. Census 2000 (<xref rid="B32" ref-type="bibr">32</xref>).</p></caption><alt-text>Map of the United States showing percentage of Asian or Pacific Islander adults aged 18 years and older, United States. The greatest percentages (21.74%-50.20%) are located along the Pacific coast.</alt-text><graphic xlink:href="PCD31A09s03" position="float"/></fig></boxed-text><p>There has also been a corresponding growth in the percentage of U.S. residents who primarily speak a language other than English. According to the 2000 census, 47.0 million (18%) of the 262.4 million people aged 5 years and older spoke a language other than English at home (<xref rid="B33" ref-type="bibr">33</xref>). This percentage increased from 14% in 1990 and from 11% in 1980. <italic>Linguistic isolation</italic> is defined by the U.S. Census Bureau as living in a household in which all members aged 14 years and older speak a non-English language and also speak English less than "very well" (i.e., have difficulty with English) (<xref rid="B32" ref-type="bibr">32</xref>). In 2000, approximately 4.5% of the U.S. population could have been considered linguistically isolated. Among certain subpopulations, however, the percentage of people who said they spoke English less than very well was high: 51% of individuals spoke an Asian or Pacific Island language, 49% spoke Spanish, and 34% spoke another Indo-European language (<xref rid="B33" ref-type="bibr">33</xref>). <xref rid="T1" ref-type="table">Table 1</xref> shows the major languages included in each group. Again, the types of languages spoken within linguistically isolated households across the United States vary by region (Figures 4&#x02013;6). Because most health surveys are typically conducted in English only, linguistic isolation can be expected to significantly increase the level of nonresponse among people who do not speak English.</p><boxed-text position="float"><fig position="float" id="F4" fig-type="diagram"><label>Figure 4</label><caption><p>Percentage of linguistically isolated Spanish-language households, United States. Source: U.S. Census 2000 (<xref rid="B32" ref-type="bibr">32</xref>)</p></caption><alt-text>Map of the United States showing percentage of linguistically isolated Spanish-language households, United States. The greatest percentages (17.71%-33.69%) are located in the southwestern United States and southern Florida.</alt-text><graphic xlink:href="PCD31A09s04" position="float"/></fig></boxed-text><p> <bold> <xref rid="F5" ref-type="fig">Figure 5</xref>.</bold> Percentage of linguistically isolated Asian-language or Pacific Island-language households, United States. Source: U.S. Census 2000 (<xref rid="B32" ref-type="bibr">32</xref>).</p><boxed-text position="float"><fig position="float" id="F5" fig-type="diagram"><label>Figure 5</label><caption><p>Percentage of linguistically isolated Asian-language or Pacific Island-language households, United States. Source: U.S. Census 2000 (<xref rid="B32" ref-type="bibr">32</xref>).</p></caption><alt-text>Map of the United States showing percentage of linguistically isolated Asian-language or Pacific Island-language households, United States. The greatest percentages (8.75%-22.73%) are located along the Pacific coast.</alt-text><graphic xlink:href="PCD31A09s05" position="float"/></fig></boxed-text><boxed-text position="float"><fig position="float" id="F6" fig-type="diagram"><label>Figure 6</label><caption><p>Percentage of linguistically isolated Indo-European&#x02013;language households, United States. Source: U.S. Census 2000 (<xref rid="B32" ref-type="bibr">32</xref>).</p></caption><alt-text>Map of the United States showing percentage of linguistically isolated Indo-European&#x02013;language households, United States. The greatest percentages (6.73%-14.24%) are located in the upper Northeast (New Hampshire and Maine), Louisiana, North and South Dakota, and Montana.</alt-text><graphic xlink:href="PCD31A09s06" position="float"/></fig></boxed-text><p>As one of the world's largest health surveillance systems, the Behavioral Risk Factor Surveillance System (BRFSS) has been instrumental in tracking health disparities across populations in the United States (<xref rid="B34" ref-type="bibr">34</xref>). Yet over the past decade, BRFSS participation rates, like those of most other surveys, have declined sharply (<xref rid="B35" ref-type="bibr">35</xref>). As part of an effort to reverse this trend and ensure the reliability and validity of BRFSS data, we assessed the impact of race, ethnicity, and linguistic isolation on measures of survey participation. This report includes the results of that assessment as well as a discussion of potential means of improving survey participation rates among these groups, thus making population-based health surveys like the BRFSS surveys more representative of the entire population.</p></sec><sec><title>Methods</title><p>The BRFSS gathers data through computer-assisted telephone interview (CATI) surveys designed to collect uniform, state-specific data on preventive health practices and risk behaviors that are linked to the leading causes of morbidity and mortality among adults. The survey is conducted by all 50 states and the District of Columbia, as well as by Puerto Rico, Guam, and the Virgin Islands with assistance from the Centers for Disease Control and Prevention (CDC). However, the three territories are not included in the analysis presented here. Further details on the BRFSS design, methodology, and questionnaire are presented elsewhere (<xref rid="B36" ref-type="bibr">36</xref>).</p><sec><title>Measures and variables</title><p>To examine aspects of survey participation, we calculated the following six dependent measures of survey participation at the county level based on final case disposition for telephone numbers called between January 1, 2003, and December 31, 2003:</p><p>1. Resolution rate: the percentage of all sampled telephone numbers for which household status with a working telephone number has been determined</p><p>2. Screening rate: the percentage of all known households in which the presence or absence of an eligible respondent has been determined</p><p>3. Cooperation rate: the percentage of known, eligible households in which a completed or partially completed interview has been obtained</p><p>4. Response rate: the percentage of all confirmed and potentially eligible sample members for whom an interview has been completed, which we calculated using Response Rate 4 recommended by the American Association for Public Opinion Research (<xref rid="B37" ref-type="bibr">37</xref>)</p><p>5. Language-barrier rate: the percentage of all sampled households given a final disposition of <italic>language barrier.</italic> Interviewers could not communicate with household members because of the language spoken in the home (which was presumably not English or Spanish, the two languages in which the BRFSS survey is conducted)</p><p>6. Refusal rate: the percentage of all sampled households given a final disposition of <italic>refusal</italic>, indicating either the selected sample member's refusal to complete the interview or the interviewer's inability to recontact the household because of a hang-up or other refusal by someone in the household</p><p>We conducted the analysis at the county level because of a lack of available information about survey nonrespondents at the individual or household level. Counties were included in the analysis if they had 30 or more observations in the denominator of each of the six participation measures. Our use of these criteria ensured greater stability in the measures calculated and helped us compare the impact of independent variables across the six models estimated; however, it also limited the analysis to 1894 of the 3141 counties in the United States.</p><p>County-level predictor variables for race, ethnicity, and linguistic isolation were derived from 2000 U.S. census counts (<xref rid="B32" ref-type="bibr">32</xref>). We calculated the percentage of each county's population that was black or African American, Asian, and Hispanic, as well as the percentage who spoke only Spanish, only an Asian language, or only another Indo-European language. On average, the included counties had somewhat higher percentages of blacks than did the nonincluded counties (9.0% compared with 7.1%), Asians (1.2% compared with 0.4%), Asian-language&#x02013;only households (0.2% compared with 0.1%), and Indo-European&#x02013;language&#x02013;only households (0.4% compared with 0.3%) but slightly lower percentages of Hispanics (4.8% compared with 6.9%) and Spanish-language&#x02013;only households (1.0% compared with 1.5%).</p><p>We also developed several county-level control variables to account for some of the other factors that are thought to affect participation rates within certain geographic areas. Socioeconomic status, often measured through a combination of income and education levels, is an important mediator of racial and ethnic health disparities and an important predictor of survey participation (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B38" ref-type="bibr">38</xref>). Likewise, living in an urban area, being away from home frequently, and screening calls with answering machines, caller-identification devices, or similar devices have been shown to reduce respondent contactability and participation rates (<xref rid="B21" ref-type="bibr">21</xref>,<xref rid="B38" ref-type="bibr">38</xref>,<xref rid="B39" ref-type="bibr">39</xref>)We also developed several county-level control variables to account for some of the other factors that are thought to affect participation rates within certain geographic areas. Socioeconomic status, often measured through a combination of income and education levels, is an important mediator of racial and ethnic health disparities and an important predictor of survey participation (2,38). Likewise, living in an urban area, being away from home frequently, and screening calls with answering machines, caller-identification devices, or similar devices have been shown to reduce respondent contactability and participation rates (21,38,39). We used 2000 U.S. census data to develop four control variables based on 1) the percentage of households in each county with incomes of $50,000 or more, 2) the percentage of adults aged 25 and older in each county who had less than a high school education (e.g., no high school diploma or equivalency), 3) the percentage of households in each county that were in urban areas, and 4) the percentage of households in each county with heads of household who had a one-way commute to work of 30 minutes or more. We used BRFSS data to calculate a fifth control variable measuring the percentage of all calls made within a county that resulted in contact with an answering machine, privacy manager, or some other identifiable type of call-screening device.&#x000a0;</p></sec><sec><title>Statistical analysis</title><p>Because all variables in the analysis were expressed as percentages, we used ordinary least squares (OLS) regression modeling to assess the impact of race, ethnicity, and linguistic isolation on survey participation. In preliminary analyses, we found a high degree of correlation between Asian race and Asian-language isolation (r = 0.92; <italic>P</italic> &#x0003c; .001) and between Hispanic ethnicity and Spanish-language isolation (r = 0.93; <italic>P</italic> &#x0003c; .001). We used separate models to determine which variables (race, ethnicity, or linguistic isolation) were better predictors, but we found the differences between them to be marginal. Although linguistic isolation is a definite barrier to survey participation, race and ethnicity may or may not be factors; thus, the Asian-language&#x02013;only and Spanish-language&#x02013;only variables were retained in the final models, but the variables Asian race and Hispanic ethnicity were not retained. Because of the strong correlation between race, ethnicity, and language-isolation variables, however, we had difficulty determining the proportional impact of each.</p><p>We also examined the possible effects of multicollinearity in our analysis. Because we found that regressing the other predictor and control variables on urbanicity explained more than 50% of the variance in the percentage-urban variable (adjusted R² = 0.52; F = 296.7), we removed urbanicity from the final models.</p><p>The final OLS models were estimated for each of the six participation measures (rates of resolution, screening, cooperation, response, language barriers, and refusal). The dependent variables used were the county-level estimates for percentage of black adults, percentage of Spanish-language&#x02013;only households, percentage of other Indo-European-language&#x02013;only households, and percentage of Asian-language&#x02013;only households. The control variables used were the percentage of households with incomes of $50,000 or more, the percentage of adults aged 25 and older with less than a high school education, the percentage of heads of household with a one-way work commute of 30 minutes or more per day, and the percentage of BRFSS calls that reached an answering machine or call-screening device. Model selection was based on forced entry of all variables into the models rather than stepwise selection. The models were estimated using SPSS 13.0 (SPSS Inc, Chicago, Ill) with the Complex Samples module.&#x000a0;</p><p>Finally, we used the OLS coefficients from the final models and the maximum county-level population parameters to calculate the maximum impact of race, ethnicity, and linguistic isolation on the six measures of survey participation.</p></sec></sec><sec><title>Results</title><p>In general, minority race and ethnicity and linguistic isolation had significant negative correlations with survey participation rates (<xref rid="T2" ref-type="table">Table 2</xref>). The regression coefficients (&#x003b2;) in these models estimate the amount of increase or decrease in the dependent measures for every one-unit difference in the independent variables. For example, for every percentage-point increase in the black population of a county, the county-level response rate declined by 0.06%. Counties with higher percentages of black residents tended to have significantly lower rates of participation and higher refusal rates. Statistically significant (&#x003b1; = .05) negative relationships were noted between the percentage of black adults in a county and county-level resolution rates (&#x003b2; = &#x02212;0.03; <italic>P</italic> = .001), screening rates (&#x003b2; = &#x02212;0.18; <italic>P</italic> = .001), cooperation rates (&#x003b2; = &#x02212;0.13; <italic>P</italic> = .001), and response rates (&#x003b2; = &#x02212;0.06; <italic>P</italic> = .001), although a significant positive relationship was seen with refusal rates (&#x003b2; = 0.06; <italic>P</italic> = .001). The percentage of black residents in a county did not have a significant effect on the rate of nonparticipation attributed to a language barrier.</p><p>Linguistic isolation also had a negative effect on participation rates, although the magnitude of this effect differed across the three language types. Higher rates of Spanish-language isolation led to lower resolution rates (&#x003b2; = &#x02212;0.11; <italic>P</italic> = .04), screening rates (&#x003b2; = &#x02212;0.92; <italic>P</italic> = .001), cooperation rates (&#x003b2; = &#x02212;0.58; <italic>P</italic> = .001), and response rates (&#x003b2; = &#x02212;0.26; <italic>P</italic> = .001). The impact of Spanish-language isolation on response rates was more than four times the impact of the percentage of black adults in a county. Counties with higher percentages of Spanish-language&#x02013;only households also had higher percentages of nonparticipation attributed to language barriers (&#x003b2; = 0.12; <italic>P</italic> = .001). Rates of Spanish-language isolation did not, however, significantly affect the percentage of nonparticipation attributed to refusals.</p><p>The percentage of households in which only Indo-European languages were spoken did not significantly affect resolution rates, but it did have a significant negative effect on screening rates (&#x003b2; = &#x02212;1.36; <italic>P</italic> = .001), cooperation rates (&#x003b2; = &#x02212;1.39; <italic>P</italic> = .001), and response rates (&#x003b2; = &#x02212;0.64; <italic>P</italic> = .001). Counties with higher rates of Indo-European-language&#x02013;only households also had higher language-barrier and refusal rates.</p><p>In contrast, Asian-language&#x02013;isolated households had less effect on survey participation rates. Counties with higher percentages of Asian-language&#x02013;only households did have significantly lower resolution rates (&#x003b2; = &#x02212;0.48; <italic>P</italic> = .03) and screening rates (&#x003b2; = &#x02212;0.98; <italic>P</italic> = .03), but they did not have significantly lower cooperation or response rates. Similarly, although higher percentages of Asian-language isolation led to a significant increase in the language-barrier rate (&#x003b2; = 0.14; <italic>P</italic> = .001), there was a significant decrease in the refusal rate in these counties (&#x003b2; = &#x02212;1.90; <italic>P</italic> = .001).</p><p>Overall, race, ethnicity, and language-isolation models explained 27% to 31% of the variance in the screening, response, and language-barrier rates, but the models were only about half as effective in explaining variance in resolution, cooperation, and refusal rates.</p><p>Because the impact of race, ethnicity, and linguistic-isolation variables depended on the size of a county subpopulation, we calculated the maximum impact of these variables among the subset of 1894 counties examined here. <xref rid="T3" ref-type="table">Table 3</xref> shows the amount of change we might expect in the percentage of each rate in counties with the highest concentrations of black residents and language-isolated households. We calculated this expected change by multiplying the high range value for each population characteristic by its corresponding OLS coefficient from <xref rid="T2" ref-type="table">Table 2</xref>. We found, for example, that in counties in which slightly more than one fourth of the households spoke only Spanish, screening rates were approximately 25% lower than in counties with no Spanish-language&#x02013;only households. We also found that in counties in which approximately three fourths of the adult population was black, response rates were 5% lower than in counties with no black residents. Response rates were 7% lower in counties with the highest concentrations of households in which Spanish was the predominant language and no one in the household spoke English very well. Likewise, response rates were approximately 7% lower in counties with higher concentrations of households in which other Indo-European languages were spoken rather than English.</p></sec><sec><title>Discussion</title><p>Our study revealed that survey participation rates were significantly lower in areas with higher concentrations of racial and ethnic minorities and linguistically isolated households. These important findings indicate the need to ensure adequate representation of these populations in large-scale health surveys such as the BRFSS. As we examine ways of increasing BRFSS participation rates, these findings will help us to design and implement more effective means of involving these hard-to-reach populations.</p><p>One particularly disturbing finding was the significant impact of Spanish-language isolation on participation rates, given that BRFSS surveys are offered in both Spanish and English. Education is an important mediating factor in survey participation among Hispanic individuals because lower levels of literacy and health literacy have been related to a greater reluctance by Hispanic individuals to participate in health surveys (<xref rid="B26" ref-type="bibr">26</xref>,<xref rid="B40" ref-type="bibr">40</xref>). Our study shows, however, that even after controls are added for education, areas with higher concentrations of Spanish-only&#x02013;speaking households are less likely to participate in health surveys. This may be because of ineffective procedures for contacting and eliciting participation from predominantly Spanish-speaking households, lack of bilingual or Spanish-speaking interviewers, or inadequate training of Spanish-speaking interviewers. It is also likely that current Spanish-language survey translations do not adequately address the different Spanish dialects spoken in the United States, such as those spoken by individuals or families originating from Mexico, Puerto Rico, or Cuba (<xref rid="B41" ref-type="bibr">41</xref>). Moreover, it may also reflect the impact of ethnic and cultural issues. Therefore, we may have to assume that concepts and interpretation are culturally dependent (<xref rid="B42" ref-type="bibr">42</xref>,<xref rid="B43" ref-type="bibr">43</xref>).&#x000a0;We were unable, however, to disentangle the influence of language and culture.</p><p>Our findings also indicate that more needs to be done to improve participation among other minorities, such as African Americans, Asians who are isolated by language, and other linguistically isolated groups. To this end, researchers are investigating ways to address disparities in participation rates by postsurvey adjustments, culturally appropriate data-collection procedures, and multiple language use.</p><p>Standard techniques are widely used to compensate for demographic differences between a survey sample and the general population it represents (<xref rid="B21" ref-type="bibr">21</xref>). Postsurvey adjustments such as weighting and stratification represent standard practices in most major health surveys. However, these techniques are often limited to a few key demographic variables for which population estimates are available. Moreover, they may produce larger standard errors that decrease the precision of estimates.</p><p>Researchers need to develop survey designs that better address the increasingly complex racial, ethnic, and linguistic mix of the U.S. population. A U.S. Department of Health and Human Services report recommended that "culturally and linguistically appropriate interviewing techniques need to be employed at all times when conducting surveys on racial and ethnic issues" (<xref rid="B4" ref-type="bibr">4</xref>). The report further recommended that relevant cultural factors and language requirements be incorporated into survey designs when feasible. Researchers need to be cognizant of the customs, values, and beliefs of individuals in minority communities, particularly because they relate to the sharing of personal information, including health care practices and health conditions (<xref rid="B44" ref-type="bibr">44</xref>).&#x000a0;Focus groups and cognitive interviews of people from various backgrounds can help determine whether respondents will interpret and respond to survey requests and questions as intended (<xref rid="B45" ref-type="bibr">45</xref>,<xref rid="B46" ref-type="bibr">46</xref>).</p><p>Some research has shown that the race, ethnicity, and sex of an interviewer can affect a respondent's level of cooperation (<xref rid="B14" ref-type="bibr">14</xref>). Because an interviewer with a background and characteristics similar to those of a potential survey participant may not be available, it is important that interviewers be trained to understand and manage multiple culturally specific issues. This understanding requires the development and implementation of cultural-sensitivity training programs for interviewers. Culturally specific scripts could also be made available to interviewers in anticipation of challenging situations.</p><p>Researchers also need to consider increasing the number of languages in which a survey is offered, especially in communities where rates of linguistic isolation are high. Moreover, it is important to ensure that the translated questions are culturally equivalent in terms of coherence and appropriateness (<xref rid="B19" ref-type="bibr">19</xref>,<xref rid="B20" ref-type="bibr">20</xref>,<xref rid="B47" ref-type="bibr">47</xref>,<xref rid="B48" ref-type="bibr">48</xref>).</p><p>Researchers have used two approaches in addressing linguistic isolation. The first approach is to translate the questionnaire and hire interviewers who are fluent in that language. This native-language speaker approach is used by the California Health Interview Survey (CHIS), a state-based telephone survey similar in content to the BRFSS survey. The 2001 CHIS was translated into Spanish, Mandarin Chinese, Cantonese, Vietnamese, Korean, and Cambodian (Khmer). Approximately 10% of the completed interviews in the state were conducted in Spanish, and 5% were conducted in one of the Asian languages (<xref rid="B49" ref-type="bibr">49</xref>).</p><p>The second approach is to rely upon third-party interpreters to administer the survey. Some language-service providers can provide interpreters in more than 150 languages (<xref rid="B50" ref-type="bibr">50</xref>). Using a three-way telephone connection, for example, an interpreter (who has access to an English version of the questionnaire but not necessarily a version translated into the respondent's language) translates the conversation between the English-speaking interviewer and the native-language&#x02013;speaking respondent. This approach is used for the National Immunization Survey (NIS), a telephone survey that collects immunization information on children aged 19 to 35 months living in U.S. households. In 2002, interviews conducted by this method accounted for 4% to 5% of the completed interviews in areas such as Boston, Newark, New York City, and King County, Wash (<xref rid="B51" ref-type="bibr">51</xref>).</p><p>Both of these translation approaches have advantages and disadvantages. The use of native-language&#x02013;speaking interviewers helps ensure that the survey questionnaire is administered in a standardized manner but reduces the number of languages and interviewers available. In contrast, third-party interpretation allows questionnaires to be administered in many languages, but administration of the questionnaire may be less consistent. Third-party interpretation also does not allow for assessment of cultural equivalence, thereby potentially leading to measurement error. Both approaches are also relatively costly. Additionally, neither approach provides a complete solution to the problem of increasing survey participation among people isolated by language.</p><p>There are several limitations to the current study. First, sample sizes in some counties limited the analysis to 1894 of 3141 counties. Second, because information on key variables such as race, ethnicity, and language spoken in the household was not available at the individual level of nonresponding households, the analysis was conducted at an aggregate (county) level. Although aggregate-level approaches to studying racial and ethnic disparities have been encouraged by the U.S. Department of Health and Human Services when individual-level data are not available, future studies of survey participation could be strengthened by surveys that are designed to collect data on key variables from nonrespondents (<xref rid="B4" ref-type="bibr">4</xref>). Third, the high correlation between the race, ethnicity, and language variables for Asian and Hispanic individuals limited our ability to disentangle the effects on survey participation of culture and language for these two groups.</p><p>Adequately identifying racial and ethnic disparities in health care and developing effective strategies to eliminate these disparities depends on the availability of valid and reliable data. Considerations of race, ethnicity, acculturation, and language are critical to the success of such health surveillance efforts. Researchers need to infuse these elements into their study designs, data-collection protocols, and data-processing routines. Indeed, several pilot studies are now being conducted in conjunction with the BRFSS to try to address these issues. These studies are using alternative sampling frames to reach individuals who are inaccessible by landline telephones, multiple modes of survey data collection, prenotification techniques tailored to minority racial and ethnic populations, surveys in languages other than English and Spanish, and case-management techniques for preassigning likely non-English&#x02013;speaking households to bilingual interviewers. Such efforts are essential for meeting the challenges to health surveillance posed by the growing diversity of the U.S. population.</p></sec>

The Annual African American Conference on Diabetes: Evolving Program Evaluation With Evolving Program Implementation

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Houston</surname><given-names>Jacquelyn M</given-names></name><degrees>MPH, APRN, BC</degrees><aff>Centers for Disease Control and Prevention</aff><address><email>houstojm@dhec.sc.gov</email><addr-line>2600 Bull St, Columbia, SC 29201</addr-line><phone>803-545-4472</phone></address></contrib><contrib contrib-type="author" corresp="no"><name><surname>Martin</surname><given-names>Maurice</given-names></name><degrees>PhD, MEd, CHES</degrees><aff>Centers for Disease Control and Prevention, Atlanta, Ga</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Williams</surname><given-names>Joel E</given-names></name><degrees>MPH, PhD</degrees><aff>South Carolina Diabetes Prevention and Control Program, South Carolina Department of Health and Environmental Control, Columbia, SC</aff><aff>Dr Williams is currently affiliated with the Department of Psychology, University of South Carolina, Columbia, SC</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Hill</surname><given-names>Rhonda L</given-names></name><degrees>PhD, CHES</degrees><aff>South Carolina Diabetes Prevention and Control Program, South Carolina Department of Health and Environmental Control, Columbia, SC</aff></contrib>

Preventing Chronic Disease

<sec><title>Background</title><p>Diabetes is a significant public health problem that affects approximately 18 million people (<xref ref-type="bibr" rid="B1">1</xref>). The prevalence among adults is expected to double by 2025 (<xref ref-type="bibr" rid="B2">2</xref>). Diabetes is more prevalent among older people, and it disproportionately affects people in minority populations (<xref ref-type="bibr" rid="B3">3</xref>). Diabetes prevalence also differs geographically. According to 2003 Behavioral Risk Factor Surveillance System data, South Carolina has the fourth highest overall rate of diabetes among the 50 states (9.3%) (<xref ref-type="bibr" rid="B4">4</xref>) but the second highest rate among African Americans (15.5%). Nationwide, African Americans are disproportionately affected by diabetes (<xref ref-type="bibr" rid="B4">4</xref>).</p><p>Self-management is the cornerstone of diabetes care and treatment, yet most people with diabetes do not receive any formal self-management education (<xref ref-type="bibr" rid="B5">5</xref>). Diabetes education integrated into comprehensive diabetes care has effectively improved self-management and diabetes clinical outcomes (<xref ref-type="bibr" rid="B6">6</xref>). Regardless of race or ethnicity, diabetes and its complications can be controlled through early diagnosis and proper self-management (<xref ref-type="bibr" rid="B1">1</xref>). Studies show that intensive glucose control can prevent retinopathy, nephropathy, neuropathy, and microvascular complications among people with diabetes (<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>).</p><p>Because African Americans in South Carolina are disproportionately affected by diabetes, the South Carolina Diabetes Prevention and Control Program (SC DPCP) and the Diabetes Today Advisory Council (DTAC) have sponsored the African American Conference on Diabetes (AACD). Since 1997, the AACD has been convened to help educate African Americans with diabetes, their families, and their caregivers. The rigor of evaluation techniques to assess the effects of AACD has increased since 2002. This article describes the AACD's evolution and the simultaneous improvement in its evaluation.</p></sec><sec><title>Context</title><sec><title>Diabetes in South Carolina</title><p>The goal of the Centers for Disease Control and Prevention's (CDC's) National Diabetes Prevention and Control Program is to help people with diabetes have long, healthy, satisfying lives (<xref ref-type="bibr" rid="B9">9</xref>). One of the CDC's national objectives is to reduce diabetes-related disparities among high-risk populations (<xref ref-type="bibr" rid="B9">9</xref>). The shared mission of the SC DPCP and DTAC is to prevent diabetes and its complications among African Americans through diabetes education and management and to make individuals aware of community resources.</p><p>South Carolina is a rural, medically underserved state with a significant diabetes problem among African Americans. In South Carolina, 40% of African Americans live in rural areas, with 26% of them living below the poverty level (<xref ref-type="bibr" rid="B10">10</xref>).</p><p>Specialized care is primarily available in areas with larger populations but is often inaccessible for poor individuals who live in rural areas (<xref ref-type="bibr" rid="B10">10</xref>). Barriers to diabetes self-management include the lack of funds or insurance to cover the cost of ongoing care, medicines, supplies, and diabetes self-management education. Among the 12 counties with a diabetes prevalence that is higher than the state average, two of the counties do not have even one certified diabetes educator (<xref ref-type="bibr" rid="B10">10</xref>). Furthermore, six South Carolina counties have a ratio of less than one certified diabetes educator per 10,000 people (<xref ref-type="bibr" rid="B10">10</xref>).</p><p>Before 1997, SC DPCP's efforts to educate rural African American communities about diabetes through health fairs and presentations in churches had limited exposure. In 1997, the SC DPCP and DTAC hosted the first annual AACD and focused on diabetes self-management. Organizers of the AACD hoped it would be a forum to provide diabetes education and resources to African Americans with diabetes and their families and caregivers across South Carolina. Organizers selected Columbia, the state capital, as the conference site because of its central location and accessibility. Since its inception, conference attendance has continued to increase almost every year; more than 1000 people attended in 2004.</p></sec><sec><title>Evolution of the AACD</title><p>The AACD is held each November during National Diabetes Awareness Month in Columbia, SC. In 1997, the AACD's first year, the conference was sponsored by two community-based agencies. By 2004, sponsorship funding had increased through educational grants from pharmaceutical companies and start-up money from private businesses. The funding increase allowed attendance to grow steadily from 195 participants in 1997 to 1044 in 2004, with a slight decrease in 2002 (<xref rid="F1" ref-type="fig">Figure</xref>). The 2002 decrease in attendance coincided with the origination of a $5 registration fee; even this nominal fee excluded some people from participating. Since 2003, community partners have assisted&#x000a0;people who needed help with the registration cost.</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure</label><caption><p>African American Diabetes Conference attendance, 1997&#x02013;2004. The 2002 decrease in attendance was attributed to the origination of a $5 registration fee.</p></caption><alt-text>Bar chart</alt-text><alternatives><graphic xlink:href="PCD31A18s01"/><table frame="hsides" rules="groups"><thead><tr><th scope="col" align="left" valign="top" rowspan="1" colspan="1"> <bold>Year</bold> </th><th scope="col" valign="top" rowspan="1" colspan="1"> <bold>Number of Attendees</bold> </th></tr></thead><tbody><tr><td align="left" valign="top" rowspan="1" colspan="1">1997</td><td align="center" valign="top" rowspan="1" colspan="1">195</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">1998</td><td align="center" valign="top" rowspan="1" colspan="1">345</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">1999</td><td align="center" valign="top" rowspan="1" colspan="1">450</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">2000</td><td align="center" valign="top" rowspan="1" colspan="1">602</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">2001</td><td align="center" valign="top" rowspan="1" colspan="1">840</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">2002</td><td align="center" valign="top" rowspan="1" colspan="1">630</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">2003</td><td align="center" valign="top" rowspan="1" colspan="1">1010</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">2004</td><td align="center" valign="top" rowspan="1" colspan="1">1044</td></tr></tbody></table></alternatives></fig></boxed-text><p>The AACD features educational encounter sessions (EESs), which are brief, didactic, skill-building sessions led by experts. AACD sessions educate participants about diabetes care practices (e.g., visiting primary care physicians, having a hemoglobin A1c test, monitoring blood glucose levels regularly) and encourage people with diabetes to adopt diabetes self-management skills and behaviors. At these sessions, participants share and learn not only from experts but also from each other. The SC DPCP and DTAC expanded their partnerships with health professionals and community volunteers to increase the number of people who could receive the AACD's resources.&#x000a0;</p><p>In 2002, a session titled "Ask the Doctor" was incorporated into the conference. During the session, participants are allowed to question a panel of physician specialists: an internist, a dentist, an optometrist, a podiatrist, and a pharmacist. In a foot care session, a clinician examined participants' feet and then taught them how to examine their own feet. Participants were shown how to use glucometers and products that make blood glucose testing easier and less painful. In addition, they were taught the importance of taking prescribed medication, monitoring their blood glucose levels, and keeping records to share with their health care providers.</p><p>Sessions on physical activity focused on having fun while moving to music and burning calories. A nutrition session was designed to teach participants how to prepare healthy soul food that was low in sodium, fat, and sugar. Health and community organizations and exhibitors presented their resources and products.</p><p>New formats and topics for future AACD sessions evolve on the basis of participant feedback and interest. For example, the initial AACD format included morning plenary sessions. Because participants were reluctant to voice their concerns and ask questions in the plenary format, the sessions were replaced with the more intimate concurrent group EESs. Attendee feedback indicated a desire for more diverse diabetes-related topics and more EESs. For example, in 2002, a session about depression and diabetes was incorporated. Participants practiced relaxation techniques and were encouraged to talk with health care providers about their mental and emotional health. Also as a result of participant feedback, AACD offered screening stations for blood pressure, cholesterol levels, and kidney disease. The AACD is the only source of diabetes-related education and other resources for many of its participants.</p></sec></sec><sec sec-type="methods"><title>Methods</title><p>The AACD's program evaluation became more rigorous as the conference became more complex. Before 2002, only participant registration records and feedback on individual sessions were evaluated. In 2003, a participant questionnaire was added to gauge satisfaction with the AACD programming. Each ensuing year, participant feedback was used formatively to plan the AACD programming.</p><p>The evaluation was conducted in two phases: phase 1 in 2002 and phase 2 in 2004. In 2002, program planners realized that the AACD conference had matured and that impact evaluation methods should be used to 1) assess the program's effect, 2) increase the number of people affected by the conference, and 3) improve programming. We selected the focus group method to gain a more comprehensive understanding of participants' thoughts and feelings about diabetes and the AACD.</p><p>Data from focus groups can provide insight into the cultural norms that shape diabetes self-management perspectives. A focus group can also create an interactive environment that allows participants to freely discuss issues (<xref ref-type="bibr" rid="B11">11</xref>). Qualitative and quantitative methods used in tandem work well for evaluating and planning educational interventions (<xref ref-type="bibr" rid="B12">12</xref>). In 2004, the evaluation included a quantitative assessment to determine whether attending the AACD improved participants' basic understanding of diabetes and awareness of important self-management skills.</p><sec><title>Phase 1: focus groups</title><p>The CDC gave the SC DPCP a nonresearch determination for program evaluation in public health practice, so no Institutional Review Board approval was required for data collection.</p><sec><title>Recruitment</title><p>Focus group participants were selected from the AACD registration forms, which included a question about participants' previous AACD attendance. Results showed that 70 registrants had attended at least two previous conferences. These registrants were contacted by telephone to confirm their prior attendance and to determine whether they 1) wanted to participate in a focus group and 2) had been diagnosed with diabetes. The majority of registrants were willing to participate; most who were excluded did not meet the disease status requirement (i.e., did not have diabetes). Of the 70 contacted, 28 met the inclusion criteria and were invited to participate. Of the 28 potential participants, two declined to participate because they were not planning to attend the 2002 conference. Later, six additional people dropped out because they were unable to attend the AACD. The remaining 20 participants were assigned to group 1 or group 2 based on the numerical order in which their names were listed on the original list of 28 eligible registrants. Odd numbers were assigned to group 1 (n = 12) and even numbers to group 2 (n = 8).</p></sec><sec><title>Facilitation</title><p>Two independent, 1-hour focus groups were conducted at the 2002 AACD. Before each focus group, participants wrote on paper their demographic data and diabetes history. Each participant received $20 at the end of the session.</p><p>The facilitator told the participants that the focus group was being used as a program evaluation tool to improve the AACD. Participants were assured that their responses would be confidential, told that participation was voluntary, and told that their continued participation would be considered permission to report the aggregate information to stakeholders.</p><p>The facilitator was an African American certified health education specialist trained in focus group facilitation and experienced in diabetes prevention and control. The same structured discussion guide (designed to be flexible to allow probing for clarification) was used for each session to ensure that the presentations were consistent. During each session, a staff member wrote the themes of the conversation on a flip chart so that participants could review and validate their responses. Two staff members took notes on the responses and another person audiotaped the sessions.</p></sec><sec><title>Analysis</title><p>Audiotapes were transcribed verbatim, and the content was analyzed to find recurring themes. The transcription's accuracy was confirmed by comparing it with the field notes. Two individuals coded the transcribed records for themes, one of whom had no previous involvement with the evaluation or program. Both coders were experts in diabetes prevention and control and experienced in qualitative analysis. The coders discussed the themes and came to a consensus before issuing the report. Because no significant demographic differences existed between the two focus groups, we reported aggregated results.</p></sec></sec><sec><title>Phase 2: quantitative measurement of diabetes understanding</title><p>In 2004, conference planners expanded the evaluation by gathering data on the short-term effect of the AACD educational sessions on participants' understanding of diabetes and its treatment.</p><sec><title>Data collection instrument</title><p>A modified 13-item Diabetes Understanding Scale was developed from section IV of the Diabetes Care Profile (DCP), an instrument for assessing understanding of diabetes and its treatment (<xref ref-type="bibr" rid="B13">13</xref>). The scale addresses topics presented during the AACD. The scale was modified by combining the diet, exercise, and medication items into one item and eliminating a diabetes and pregnancy item because no sessions were held on this topic. Two items on prevention and treatment of high "blood sugar" and low "blood sugar" were presented as the following four items: 1) prevention of high blood sugar, 2) prevention of low blood sugar, 3) treatment of high blood sugar, 4) and treatment of low blood sugar. The item alterations allowed evaluators to consider separately changes in participant understanding of prevention and treatment for high and low blood sugar.</p></sec><sec><title>Administration</title><p>All participants at the 2004 AACD were invited to participate in the program evaluation by completing the preconference and postconference surveys in their registration packets. Unique identifiers on the forms allowed us to match the participants' preconference forms with their postconference forms. During the presession breakfast, an introduction to the evaluation process was presented. Confidentiality was assured, and participants' completed surveys were considered consent to use their data in an aggregate form.</p><p>Participants completed and submitted the preconference diabetes understanding survey and a short demographic questionnaire before the morning sessions began. After the closing session, participants completed the postconference diabetes understanding survey. To encourage participation, participants received free raffle tickets, which they could only turn in with their completed evaluation forms. Of the 1044 attendees, 628 attendees completed at least one of the surveys and provided useable data. Fewer postconference evaluation forms were completed than preconference forms because many participants left the conference before filling out the postconference survey. Using a 5-point Likert scale (with 1 = poor, and 5 = excellent), respondents ranked how well they understood specific diabetes issues. The data were entered into an EpiData version 3.02 database (EpiData Association, Odense, Denmark) and exported as a SAS file (SAS Institute Inc, Cary, NC).</p></sec><sec><title>Analysis</title><p>Survey data were analyzed using SAS version 9.1. Descriptive statistics were used to describe the respondent sample by sex, race, age, and diabetes status. The mean survey scores, standard error, and number of respondents for each subscale item preconference and postconference were calculated. Cronbach ? for internal consistency was computed for the preconference subscale items. Proc Ttest was used to examine significant differences in participants' mean survey scores, which had been matched by repeated measures at preconference and postconference (by unique identification numbers).</p></sec></sec></sec><sec><title>Consequences</title><sec><title>Phase 1: focus groups</title><p>Twenty adults, predominantly African American women, participated in two focus groups (<xref rid="T1" ref-type="table">Table 1</xref>). Participants freely engaged in discussions, sharing personal and family stories about their diabetes experiences. The devastation caused by diabetes in their families and concern for preventing diabetes and its complications among loved ones dominated the conversations. <xref rid="T2" ref-type="table">Table 2</xref> includes direct quotes for each theme.</p><sec><title>Motivation for participation</title><p>When participants were asked, "What inspired you to participate in this conference over the years?" education and learning was the most frequently reported theme among the answers. Participants also frequently cited family and communication issues &#x02014; a social support theme. They reported that the AACD provided a unique opportunity to meet and talk to others with similar problems.</p></sec><sec><title>Logistics</title><p>Participants were also asked about which components of the conference were helpful and should be continued in future conferences. Again, the themes of education, social support, and information about resources emerged in their answers. When asked about items that needed improvement, their answers focused on logistics (e.g., the setting, programming issues, exhibitors, conference amenities). Participants were most concerned about the conference center's limited space and getting the registration information early. Participants also expressed a desire to be involved in the conference planning.</p></sec><sec><title>Knowledge and behavior change</title><p>When asked what they would have done differently as a result of the conference to handle their diabetes, participants said that they were more confident in their ability to manage their diabetes because of what they learned at the conference. They were better able to use the social support systems in their communities and families, making them more effective at performing self-care tasks and more willing to seek professional health care when needed. Access to resources (human and material) was a third theme that emerged. Participants found the information about resources useful.</p></sec></sec><sec><title>Phase 2: diabetes understanding</title><p>In 2004, 36% of AACD attendees responded to all of the items on the preconference and postconference surveys. <xref rid="T3" ref-type="table">Table 3</xref> includes a description of the survey participants. <xref rid="T4" ref-type="table">Table 4</xref> presents preconference-to-postconference score changes for each item, the mean number of respondents, and the significance levels for change by item.</p><p>The mean item scores of participants who completed the preconference and postconference surveys indicate a significant increase in self-rated understanding for each item on the scale. Internal consistency analyses of the preconference data revealed that the scale was highly reliable (Cronbach &#x003b1; = 0.96).</p></sec></sec><sec><title>Interpretation</title><p>In 1997, the AACD was a small conference with fewer than 200 attendees and limited sponsorship, resources, and marketing ability. It has evolved into a program with more than 1000 attendees and greater sponsorship, resources, and marketing. As the AACD has evolved, so too has the quality of its programming, stakeholder expectations, and evaluation rigor. Qualitative and quantitative methods were used to effectively evaluate and plan this educational intervention.</p><p>Qualitative findings from focus groups suggest that participants at the AACD were motivated to attend the EESs because they received quality diabetes education, social support, and resources in an inviting, interactive environment. They felt empowered to help others manage diabetes. They also reported that previous AACD attendance improved their ability to adopt effective diabetes self-care practices.</p><p>The participants' perspectives were used in planning the logistics of each succeeding AACD. In 2004, the program site was relocated to a larger convention center to provide more space and address the logistical issues identified in 2002. To let more people know about the conference and allow more people to register early, the organizers marketed the conference through multiple media channels. The AACD continued to offer an array of topics on diabetes self-management and expanded the variety of concurrent EESs to provide participants more opportunities to learn about different topics.</p><p>The AACD has been providing diabetes education and resources to the community since 1997; however, 2004 was the first year that the cognitive impact of the AACD on participants was evaluated. Quantitative findings from 2004 suggest that the AACD conference format with EESs improved participants' self-reported understanding of diabetes self-management. The survey reliability measures were high and similar to those found in other studies (<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref>). Examination of preconference to postconference changes in self-rated understanding suggests that the AACD had a significant effect on diabetes-related understanding, at least in the short term.</p><p>Increased conference attendance and increased understanding about diabetes self-management does not necessarily lead to behavior change, which is the primary purpose of the AACD. Although focus groups suggest that the behavior change occurred among multiyear attendees with diabetes, focus groups cannot be used to determine the actual proportion of participants who changed their behavior. Understanding how the AACD benefits stakeholders and participants is important for planners of future programs, as is determining how to transform program evaluation into behavioral research. As the rigor of evaluation techniques increases, new research questions will emerge.</p><p>The evolutionary process described in this article is an example of the way that program evaluation not only improves programming but also plays a role in behavioral research. From the AACD evaluation, important research questions have already emerged:</p><list list-type="bullet"><list-item><p>How much of the target audience can we reach using the AACD format?</p></list-item><list-item><p>How can we recruit more attendees?</p></list-item><list-item><p>What is the long-term effect of the AACD on participants' diabetes-related knowledge, attitudes, and behaviors? How can we measure the effects?</p></list-item></list><p>As the next step in the evolution of the AACD's program evaluation plan, we are considering using the Behavioral Risk Factor Surveillance System diabetes module to track the behavior of a sample of individuals who attend the conference each year.</p></sec>

The Common Threads in Program Evaluation

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Shadish</surname><given-names>William R</given-names></name><degrees>PhD</degrees><role>Professor, Founding Faculty, and Chair</role><aff>School of Social Sciences, Humanities, and Arts, University of California at Merced</aff><address><email>wshadish@ucmerced.edu</email><addr-line>PO Box 2039, Merced, CA 95344</addr-line><phone>209-724-4372</phone></address></contrib>

Preventing Chronic Disease

<p>A well-known evaluator once said, "Evaluation &#x02014; more than any science &#x02014; is what people say it is, and people currently are saying it is many different things" (<xref rid="B1" ref-type="bibr">1</xref>). Ask an economist what program evaluation is, and you will get a very different answer than if you asked a psychologist; and they would both differ from what an educator might say. Indeed, the field is so large and diverse, and the use of the term <italic>program evaluation</italic> is so ubiquitous, that it is often difficult to discern any common threads. Yet common threads do exist, and I would like to point to some of them in the articles of this special issue of <italic>Preventing Chronic Disease</italic>.</p><p>Five common concerns are woven throughout the literature on program evaluation (<xref rid="B2" ref-type="bibr">2</xref>). First is a concern with <italic>how to construct valid knowledge</italic>. This concern has both a philosophical component and a methodological component; the philosophical component concerns the kinds of things we can know about programs, and the methodological component concerns the designs, measures, and analyses that we use to create and organize data. Second is a concern with <italic>how we place value</italic> on evaluation results. One often hears it said that the data speak for themselves, but that is rarely the case. This concern articulates the many theoretical and practical tools we have available to help us in this valuation. Third is a concern with <italic>how programs change</italic>. After all, program evaluation is intended to be a very practical area of study, one that aims to make a real difference in people's lives. If we do not know the leverage points for program change, we cannot apply evaluation results to gain that leverage. Fourth is <italic>how to use evaluation results</italic> in the policy process. This concern is about how to get our results to the stakeholders who influence those leverage points in a way that helps stakeholders make use of the data. Fifth and finally is the paramount concern with <italic>how to organize evaluation practice</italic>, given the implications of all the preceding issues for what the evaluator actually does in any given evaluation. This fifth concern is always a matter of tradeoffs, for one can never do everything well in a single evaluation.</p><sec><title>Knowledge Construction</title><p>In the early years of program evaluation, evaluators approached the task using the methods they learned in their primary disciplines. Psychologists tended to use experiments, educators relied heavily on measurement, and economists leaned toward sophisticated statistical analysis of observational data. Gradually, however, evaluators realized that no single method was enough. The choice of evaluation method should follow question choice, but so many different questions are important in program evaluation that no single method can answer all of them well. As examples, evaluators are asked to answer questions about the following:</p><list list-type="bullet"><list-item><p> <italic> <bold>Need.</bold> </italic> Every program targets a real or perceived need, such as the need to reduce rates of HIV infection, hypertension, or diabetes or to reduce the costs associated with the care of those conditions. Characterizing those needs requires such methods as needs assessment, stakeholder surveys, and epidemiological studies of incidence and prevalence. Jack et al (<xref rid="B3" ref-type="bibr">3</xref>) describe how the National Center for Chronic Disease Prevention and Health Promotion of the Centers for Disease Control and Prevention uses public health surveillance and epidemiologic studies to measure the need for behavioral interventions.</p></list-item><list-item><p> <italic> <bold>Program implementation.</bold> </italic> Programs can be large enterprises composed of multiple elements that come into play to different degrees over time. Creating a program infrastructure is an early task, engaging clients and providers comes a bit later, and providing follow-up care comes even later. Assessing program implementation requires such methods as inspection of records, cost analysis, creation of management information systems, and observation of client interactions with the program from intake to follow-up, with all these methods being differentially useful depending on the stage of program implementation. Besculides et al (<xref rid="B4" ref-type="bibr">4</xref>) use a combination of quantitative measures and qualitative interviews to assess successful implementation of lifestyle programs for women.</p></list-item><list-item><p> <italic> <bold>Program outcome.</bold> </italic> A key rationale for implementing programs is that they will result in beneficial outcomes. The most precise method for measuring those outcomes is often a randomized experiment. When randomization is not feasible for ethical or practical reasons, or the precision yielded by randomization is unnecessary, many kinds of nonrandomized experiments can be used, such as interrupted time series, regression discontinuity, and nonequivalent comparison group designs. Sometimes the interest is not in attributing outcomes to the intervention but in monitoring outcomes over time to see if they are approaching a defined standard, as with the study by Mukhtar et al (<xref rid="B5" ref-type="bibr">5</xref>) of progress toward meeting selected national health outcomes in diabetes.</p></list-item></list><p>No single method can provide a precise answer to all of these questions. That is why the organization of evaluation teams benefits so much from being multidisciplinary.</p><p>Not surprisingly, then, the articles in this special issue describe a wide array of evaluation methods because the authors are asking many different kinds of questions. Among the methods and tools used are focus groups (<xref rid="B4" ref-type="bibr">4</xref>,<xref rid="B6" ref-type="bibr">6</xref>), logic models (<xref rid="B7" ref-type="bibr">7</xref>), and monitoring of program outcomes (<xref rid="B5" ref-type="bibr">5</xref>). Some studies used a combination of methods, sometimes to answer more than one question and sometimes to try to capture the merits of both high bandwidth and high fidelity in one study. Besculides and colleagues (<xref rid="B4" ref-type="bibr">4</xref>) did this in their study of best practices in implementing lifestyle interventions targeting women. So did Houston et al (<xref rid="B6" ref-type="bibr">6</xref>) in their study of a 1-day lay health diabetes conference and Tucker et al (<xref rid="B7" ref-type="bibr">7</xref>) in their evaluation of the REACH 2010 initiative. The cost of such mixed-methods research is that one must take away from Peter to pay Paul; because evaluation budgets are always limited, we can devote fewer resources to doing one method well if we spread those resources over more than one method. Whether or not this tradeoff is acceptable is a decision that must be made on a case-by-case basis.</p></sec><sec><title>Valuing</title><p>Value judgments are present throughout an evaluation. One example is choosing outcome variables. Outcomes connect to values in two ways. The first is that some outcomes are measures of the need to which the program is intended to respond, so that a program is better to the extent that it ameliorates those needs. For example, if we cite the fiscal costs of chronic disease as the need that justifies a public health intervention, then that intervention ought to reduce those costs. An example is in the study by Rein et al (<xref rid="B8" ref-type="bibr">8</xref>), who argue that interventions to prevent hypertension should both improve health and reduce costs. The second way that outcomes connect to values is through stakeholder opinions. Programs have stakeholders whose opinions vary widely about what is a good outcome. For instance, one study of the outcomes of long-term care for people who were chronically mentally ill (<xref rid="B9" ref-type="bibr">9</xref>) found that patients and their families valued safe shelter, good food, and adequate medical care, but federal and academic stakeholders valued programs that helped patients move toward independent living. Especially when achieving one outcome sometimes comes at the cost of sacrificing another, stakeholders often disagree about whether an outcome is good or bad. This is one reason why so many approaches to evaluation start with identification of and contact with program stakeholders, as in Martin and Heath's (<xref rid="B10" ref-type="bibr">10</xref>) use of a six-step model that starts with engaging stakeholders.</p><p>Evaluators also deal with value judgments when they (sometimes implicitly) set standards by which they judge how much improvement is sufficient to be valuable, what is sometimes called <italic>practical significance</italic>. A common implicit standard is to compare the treatment to a control, declaring the treatment good if it improves upon what is accomplished by the control. An innovation is often thought to be especially valuable if it improves on the outcomes of a usual treatment control, as opposed to a no-treatment control. Rein et al (<xref rid="B8" ref-type="bibr">8</xref>) provide an example when they compare health outcomes under a state-funded education and direct service program with both no preventive treatment for high blood pressure and private-sector preventive treatment. Most stakeholders would argue that a public intervention that improves over private-sector treatment is more valuable than one that merely improves over no treatment at all. However, evaluators sometimes also refer to minimum absolute standards, and if a program falls below this level, it fails no matter how it performs in other respects. For example, Mukhtar et al (<xref rid="B5" ref-type="bibr">5</xref>) use the <italic>Healthy People 2010</italic> objectives for selected diabetes outcomes as a standard that must be met to reach a positive evaluation. Such absolute cutoffs tend to be rare, and even when they are available they tend to be used in combination with comparison to a control.</p><p>Finally, evaluators deal with value judgments when they synthesize the diverse results of a study to reach an overall evaluative conclusion. Because stakeholders value different outcomes differently, a single overall synthesis is often difficult to justify to all parties. For example, the tradeoff between lowering one's blood pressure and risking sexual impotence may be valued differently by the researcher and the patient, as witnessed by the number of therapies that researchers may judge successful but with which some patients refuse to comply. Consider, for instance, the evaluation of <italic>Sesame Street</italic> by Bogatz and Ball (<xref rid="B11" ref-type="bibr">11</xref>). It found that children exposed to <italic>Sesame Street</italic> learned several more letters of the alphabet per year than did control group children but that disadvantaged children learned less than advantaged children. Is that good or bad? Cook and his colleagues (<xref rid="B12" ref-type="bibr">12</xref>) argued that if you believe that such programs should be improving outcomes on average, then it is good; but if you believe such programs should be closing the gap between the most and least needy children, then it is bad. The most common safeguard is to seek diverse input from stakeholders about how they prioritize among the results and then refer to those priorities by creating multiple syntheses that reflect the major positions among stakeholders (<xref rid="B13" ref-type="bibr">13</xref>).</p></sec><sec><title>Social Programs</title><p>The role of program evaluator is not the same as the role of program developer. Indeed, some evaluators argue that the two roles are incompatible because the developer is often biased toward wanting a positive evaluation of the program (<xref rid="B14" ref-type="bibr">14</xref>). Still, many evaluators find themselves involved in program development because they often have broad experience that comes from having evaluated similar kinds of programs in the past, because they know that it often makes for a better evaluation if the evaluator can assist with program development from the start, or because their job description calls for both activities. Balamurugan and colleagues (<xref rid="B15" ref-type="bibr">15</xref>) illustrate this melding of the two roles in their article about programs for diabetes self-education management. They show how the lack of advanced evaluation planning impeded not only the evaluation but also the effectiveness of the program itself.</p><p>All evaluators benefit from knowledge of how programs come into being, change, end, and function in their environment. For example, if we aim to create sustainable public health programs, we must know the economic, social, political, and psychological factors that make programs sustainable. Similarly, if we believe that individual, family, health system, community, and societal factors all contribute to the rise of chronic disease (<xref rid="B3" ref-type="bibr">3</xref>), then we have to do research on those factors to know how to change them. For example, Besculides et al (<xref rid="B4" ref-type="bibr">4</xref>) study factors that lead to successful lifestyle interventions targeting women; such knowledge tells the evaluator where the leverage points for productive program change might be so that the evaluation can be directed toward answering questions about those points.</p><p>A general rule of thumb is that the smaller the intervention, the more easily it can be eliminated from or added to the things that service providers do &#x02014; which makes change more feasible. For example, if Houston et al's (<xref rid="B6" ref-type="bibr">6</xref>) 1-day lay health diabetes conference is effective, it is easy to disseminate it to other places; and if it is ineffective, it could be terminated and the resources moved elsewhere without too much resistance. This is far less the case for larger interventions, such as an entire clinic devoted to health promotion. Starting such clinics in other places is an expensive and time-consuming endeavor, and closing down such clinics entirely is a rare and often controversial event. Another rule of thumb is that the kinds of evaluation activities we use with new programs should be different from those we use with mature programs. For new programs, we should use evaluation activities associated with needs assessment and program implementation. For mature programs, we should emphasize outcome evaluation &#x02014; after the program has worked out the initial kinks that inevitably occur when a program begins.</p></sec><sec><title>Use of Results in Policy</title><p>Maximizing the chances that evaluation results will be used is a paramount concern in evaluation. In this special issue, use of evaluation results is discussed explicitly by Martin and Heath (<xref rid="B10" ref-type="bibr">10</xref>) as part of a six-step model of evaluation. In the early years, few evaluators thought much about whether evaluation results would be used in policy. They simply assumed that their results would be used once presented, but having evaluation results used proved complicated. First, several kinds of use can occur. These include <italic>instrumental use</italic> in which evaluation results are used to make a policy decision, <italic>conceptual use</italic> in which evaluation results may change the way stakeholders think about policy even though the results may not result in an immediate policy change, and <italic>persuasive use</italic> in which evaluation results are used to advocate for or against a policy. Instrumental use tends to occur least frequently and often involves small changes, because small changes are often more feasible than big changes. Conceptual use is ubiquitous among those who keep informed about a policy issue and can have a profound impact over time on how future generations of stakeholders shape the policy process. Persuasive use is also common from lobbyists to legislators who desire certain policies to be established and who use evaluation results to support their case.</p><p>Second, use can occur at any time after evaluation results are presented. Some use occurs immediately, but much use occurs later, sometimes decades later (<xref rid="B16" ref-type="bibr">16</xref>). The more immediate the use, the greater the likelihood that the change made is a small and incremental one. Large changes to a system take time because they involve so many ancillary changes and because changes that are not immediately feasible often become feasible later when the context has changed.</p><p>Third, use rarely happens without the evaluator doing things to make it happen. Evaluators have learned that instrumental use can be facilitated by having frequent and early contact with users, studying things the user actually controls, clarifying action implications of findings, and disseminating results in forms other than traditional research reports. Conceptual use can be facilitated by challenging fundamental ideas and assumptions and by circulating results throughout the network of people concerned with the issues in the outlets they read.</p></sec><sec><title>Evaluation Practice</title><p>All of the prior issues come together in evaluation practice, in which evaluators must decide whether to do an evaluation, what questions to ask and methods to use, how to involve stakeholders in the evaluation, what values should be represented, and how to facilitate use. After all, time and resource constraints imply that evaluators cannot, for example, ask every question, use every method, or foster every kind of use. Many evaluators therefore use a set of concepts that help them to focus their practice. For example, Jack et al (<xref rid="B3" ref-type="bibr">3</xref>) note that "<italic>decision and accountability</italic>, <italic>utilizations focused</italic>, <italic>client centered and responsive</italic>, <italic>case study</italic>, and <italic>outcomes monitoring and value added</italic> are a few examples of evaluation approaches." In addition, Lavinghouze (<xref rid="B17" ref-type="bibr">17</xref>) describes the theory-driven approach to evaluation. Such evaluation approaches are ways of helping the evaluator decide on the tradeoffs involved in conducting an evaluation.</p><p>Evaluation practice also entails often-complex structures for how the evaluation process is to be organized. Even the lone evaluator within a health center faces organizational obstacles to evaluation. As the evaluation context grows, the organizational challenges increase dramatically. Discussions of this issue figure prominently in several&#x000a0; articles in this issue of <italic>Preventing Chronic Disease</italic>, including MacDonald et al's (<xref rid="B18" ref-type="bibr">18</xref>) description of methods for coordinating national and community-level evaluation efforts in the Steps to a HealthierUS program, Balamurugan et al's (<xref rid="B15" ref-type="bibr">15</xref>) discussion of ways to organize evaluation in rural Arkansas, and Tucker et al's (<xref rid="B7" ref-type="bibr">7</xref>) analysis of how to combine local site-specific evaluations with national evaluations that are able to synthesize at least some of the evidence across sites.</p></sec><sec><title>Conclusion</title><p>Public health has a long history of involvement in program evaluation. Indeed, many evaluators have forgotten that the first textbook on program evaluation in public health, by Edward Suchman (<xref rid="B19" ref-type="bibr">19</xref>), was published in 1967. The field has made much progress since then (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B20" ref-type="bibr">20</xref>), and it is a pleasure to see this tradition continued in the articles of this issue.</p></sec>

Solution for Survey Discrepancies in Washington State Smoking Prevalence

Could not extract abstract

<contrib contrib-type="author"><name><surname>Boysun</surname><given-names>Michael J</given-names></name><degrees>MPH</degrees><aff>Tobacco Prevention and Control Program, Washington State Department of Health, Olympia, Wash</aff></contrib><contrib contrib-type="author"><name><surname>Maher</surname><given-names>Julie E</given-names></name><degrees>PhD</degrees><aff>Program Design and Evaluation Services, Multnomah County Health Department, Oregon Department of Human Services, Portland, Ore</aff></contrib><contrib contrib-type="author"><name><surname>Stark</surname><given-names>Michael J</given-names></name><degrees>PhD</degrees><aff>Program Design and Evaluation Services, Multnomah County Health Department, Oregon Department of Human Services, Portland, Ore</aff></contrib><contrib contrib-type="author"><name><surname>Pizacani</surname><given-names>Barbara A</given-names></name><degrees>PhD</degrees><aff>Program Design and Evaluation Services, Multnomah County Health Department, Oregon Department of Human Services, Portland, Ore</aff></contrib><contrib contrib-type="author"><name><surname>Rohde</surname><given-names>Kristen</given-names></name><degrees>MA</degrees><aff>Program Design and Evaluation Services, Multnomah County Health Department, Oregon Department of Human Services, Portland, Ore</aff></contrib><contrib contrib-type="author"><name><surname>Dilley</surname><given-names>Julia</given-names></name><degrees>MES</degrees><aff>Tobacco Prevention and Control Program, Steps to a HealthierUS Washington State, Olympia, Wash</aff></contrib><contrib contrib-type="author"><name><surname>Wynkoop Simmons</surname><given-names>Katrina</given-names></name><degrees>PhD</degrees><aff>Behavioral Risk Factor Surveillance System Coordinator, Washington State Department of Health, Olympia, Wash</aff></contrib>

Preventing Chronic Disease

<sec><title>To the Editor:</title><p>Consistent with the findings of Ramsey et al (<xref rid="B1" ref-type="bibr">1</xref>), we found that the Washington State smoking prevalence data from the Adult Tobacco Survey (ATS) were lower than the prevalence data from the Behavioral Risk Factor Surveillance System (BRFSS). In this letter, we discuss how Washington resolved this problem of disparate prevalence estimates and still obtained population-based survey data on many tobacco-related measures.</p><p>Although the BRFSS is conducted in Washington to collect data on health behaviors, including tobacco-related health behaviors (<xref rid="B2" ref-type="bibr">2</xref>), the Washington Tobacco Prevention and Control Program also conducted the ATS from 2000 through 2002 to obtain extensive information on tobacco-related knowledge, attitudes, and behaviors. Like the New Hampshire ATS and BRFSS, the Washington ATS and BRFSS were both random-digit&#x02013;dialed statewide telephone surveys of noninstitutionalized adults that used the same questions to measure tobacco prevalence, and both surveys had similar response rates (<xref rid="T1" ref-type="table">Table</xref>). However, the ATS contained strong, tobacco-specific introductory language, and the BRFSS contained general, health survey introductory language. For each year from 2000 through 2002, the ATS found a lower smoking prevalence in Washington State than did the BRFSS, and this difference became statistically significant in 2001 and 2002 (<xref rid="T1" ref-type="table">Table</xref>). The lack of a significant difference between the ATS and BRFSS findings in 2000 was likely a result of survey estimate variability related to the smaller BRFSS sample sizes. Therefore, our results support the conclusion of a California study by Cowling et al that the tobacco-specific survey introduction is associated with underreported tobacco use by some smokers (<xref rid="B3" ref-type="bibr">3</xref>). Cowling et al state: "The specificity of the introduction may cue respondents to adjust their responses (i.e., deny tobacco use) in order to shorten the length of the interview experience" and "provide a socially desirable response." Cowling et al do not provide additional information on whether the order of the smoking questions might have also contributed to this difference in prevalence, a possibility suggested by Ramsey et al (<xref rid="B1" ref-type="bibr">1</xref>).</p><p>In 2003, Washington began incorporating ATS questions into the BRFSS, partly to prevent this apparent underreporting in the ATS. Using a modular approach similar to that used for Oregon's BRFSS, we created an instrument that meets the needs of general public health surveillance tools (e.g., the BRFSS) as well as tobacco-related surveillance tools (e.g., the ATS). In our instrument, the general health survey introduction from the BRFSS is used for all survey respondents. Respondents answer the core demographics and health questionnaire of the BRFSS and then either a module of state-specific questions or a module of tobacco-specific questions, many of which are from the ATS. The average survey length of each module is about the same as the length of each module in the 2002 BRFSS survey.</p><p>Smoking prevalence based on the expanded BRFSS data for 2003 (N = 18,644) was 19.8% (95% confidence interval [CI], 19.2%&#x02013;20.6%), which was more similar to previous BRFSS prevalence estimates than previous ATS estimates (<xref rid="T1" ref-type="table">Table</xref>). This finding was reassuring and suggested that the presence of numerous tobacco-related questions on the BRFSS did not create a bias similar to that generated by a tobacco-specific introduction.</p><p>In addition to providing potentially less-biased surveillance data, the modular approach to the BRFSS provides additional benefits. First, more states conduct the BRFSS than the ATS, so more comparisons of results can be made, and unlike the ATS the BRFSS is conducted throughout the year. Second, our modular approach has tripled the size of the core BRFSS questionnaire, enabling the tobacco-control program and other programs to perform more subgroup analyses. Third, this approach facilitates examination of associations between tobacco-related measures and other health indicators. Fourth, more room on the survey is available on the nontobacco module for other programs to add questions. Finally, the modular approach streamlines the surveillance by saving staff time and minimizing the time required of participants. Using procedures and protocols developed for the BRFSS incorporates oversight expertise for both surveys into one operation.</p></sec>

A Cost Evaluation of the Georgia Stroke and Heart Attack Prevention Program

Could not extract abstract

<contrib contrib-type="author"><name><surname>Rein</surname><given-names>David B</given-names></name><degrees>PhD</degrees><aff>Research Triangle Institute International, Waltham, Mass</aff></contrib><contrib contrib-type="author" corresp="yes"><name><surname>Orenstein</surname><given-names>Diane</given-names></name><degrees>PhD</degrees><aff>Division for Heart Disease and Stroke Prevention, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention</aff><address><email>dro1@cdc.gov</email><addr-line>4770 Buford Hwy, Mail Stop K-47, Atlanta, GA 30341</addr-line><phone>770-488-8003</phone></address></contrib><contrib contrib-type="author"><name><surname>Constantine</surname><given-names>Roberta T</given-names></name><degrees>PhD</degrees><aff>Research Triangle Institute International, Waltham, Mass</aff></contrib><contrib contrib-type="author"><name><surname>Chen</surname><given-names>Hong</given-names></name><degrees>MS</degrees><aff>Research Triangle Institute International, Waltham, Mass</aff></contrib><contrib contrib-type="author"><name><surname>Jones</surname><given-names>Patricia</given-names></name><degrees>RN, CDE</degrees><aff>Georgia Division of Public Health, Chronic Disease Prevention and Health Promotion Branch, Atlanta, Ga</aff></contrib><contrib contrib-type="author"><name><surname>Brownstein</surname><given-names>J. Nell</given-names></name><degrees>PhD</degrees><aff>Division for Heart Disease and Stroke Prevention, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Ga</aff></contrib><contrib contrib-type="author"><name><surname>Farris</surname><given-names>Rosanne</given-names></name><degrees>PhD, RD</degrees><aff>Division for Heart Disease and Stroke Prevention, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Ga</aff></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>Hypertension is a leading cause of stroke, coronary artery disease, heart attack, and heart and kidney failure in the United States. Currently, 50 million Americans have hypertension and another 45 million have prehypertension (blood pressure of 120&#x02013;139 mm Hg [systolic] or 80&#x02013;89 mm Hg [diastolic]) (<xref rid="B1" ref-type="bibr">1</xref>). More than 70% of U.S. adults with hypertension do not have it under control (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B3" ref-type="bibr">3</xref>). Hypertension is particularly common among African Americans, who have a 30% higher prevalence of hypertension than whites (<xref rid="B1" ref-type="bibr">1</xref>). As might be expected, African Americans experience hypertension-related deaths at younger ages than whites and have higher rates of stroke, left ventricular hypertrophy, and heart attack (<xref rid="B3" ref-type="bibr">3</xref>). Some but not all of these differences are explained by the lower socioeconomic status (SES) of African Americans, because lower SES is also strongly related to uncontrolled blood pressure (<xref rid="B4" ref-type="bibr">4</xref>). As many as 30% of all deaths among African American men and 20% of all deaths among African American women can be attributed to high blood pressure (<xref rid="B5" ref-type="bibr">5</xref>).</p><p>Aggressive treatment of hypertension, which usually involves medication, significantly decreases the risk of coronary artery disease, congestive heart failure, stroke, and resulting disability. For example, a 12-point to 13-point reduction in blood pressure can lower the risk of heart attack by 21%, stroke by 37%, and total cardiovascular deaths by 25% (<xref rid="B6" ref-type="bibr">6</xref>). Results of recent large hypertension trials demonstrated that inexpensive thiazide-type diuretics are superior in preventing one or more major forms of cardiovascular disease&#x000a0;(<xref rid="B7" ref-type="bibr">7</xref>). Unfortunately, low-income individuals without prescription drug coverage are significantly more likely to skip doses to save money or make their hypertension medication prescriptions last longer. In one recently observed population, systemic hypertension was adequately controlled among only 38% of those who paid for their medication themselves (<xref rid="B8" ref-type="bibr">8</xref>).</p><p>The Georgia Stroke and Heart Attack Prevention Program (SHAPP) is an education and direct service program for low-income patients with hypertension. The program is based on the Chronic Care Model, a framework for identifying the essential elements of a health care system and involving patients in their own care (<xref rid="B9" ref-type="bibr">9</xref>). SHAPP patient services are provided through the county health departments and include screening, referral to physicians, diagnosis, and treatment. Treatment protocols are based on Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure recommendations (<xref rid="B10" ref-type="bibr">10</xref>). Individual patients are assigned to nurses who act as case managers. SHAPP nurses then coordinate the wide array of treatment services including physical and family history assessments, diagnostic testing, lifestyle counseling and education, medication, and patient monitoring (follow-up visits for medication, blood pressure assessment, and any needed testing) as stipulated by the SHAPP protocol. Information on patient medical and family history, physical characteristics, and risk factors is collected. Diagnostic testing is done to determine baseline and follow-up blood pressure levels. Lifestyle counseling includes advice on diet, weight, smoking cessation, alcohol consumption, and physical activity and information about signs and symptoms of stroke and heart attack. Nurses track medication side effects and monitor patients to ensure they keep clinic appointments and adhere to medication schedules. SHAPP also supplies prescription drugs at low or no cost according to a treatment protocol.</p><p>According to the Centers for Disease Control and Prevention's (CDC's) Behavioral Risk Factor Surveillance System (BRFSS), 1.7 million Georgians had hypertension in 2004. Of those, 469,800 were low income, uninsured, or underinsured and were potentially eligible for the SHAPP program. (Eligibility is based on both income and hypertension severity.) However, given the small enrollment in SHAPP (15,819 clients in fiscal year 2003), there is likely a much a greater need in Georgia for SHAPP services than the program is currently able to meet.</p><p>Reducing hypertension can lead to marked reductions (<xref rid="B10" ref-type="bibr">10</xref>) in the risk of several high-consequence adverse events such as hemorrhagic stroke (<xref rid="B11" ref-type="bibr">11</xref>), ischemic stroke, heart disease (<xref rid="B12" ref-type="bibr">12</xref>), and kidney failure (<xref rid="B13" ref-type="bibr">13</xref>). However, one recent study suggests that less intensive interventions that rely only on patients to manage their own hypertension care are relatively ineffective (<xref rid="B14" ref-type="bibr">14</xref>). Comparatively, SHAPP is a higher-intensity intervention guided by the premise that providing low-cost preventive services to medically indigent patients provides benefits to patients and savings to the state. SHAPP patients who control their blood pressure could reduce their risks for adverse events, thus prolonging and improving the quality of their lives and lowering the annual medical costs borne by the state for high-cost hospital care and procedures. Although advocates of SHAPP have long suspected that the program results in cost savings, the association among SHAPP services, patient outcomes, and medical costs has never been formally evaluated. In this article, we discuss a limited, first-time evaluation of the costs and benefits of SHAPP to determine whether this promising practice results in enhanced patient health and reduced medical costs for the state.</p></sec><sec><title>Methods</title><sec><title>Selection of districts</title><p>We collected administrative data on the costs and hypertension control outcomes of SHAPP in two Georgia health districts with high rates of blood pressure control for fiscal year 2003 (July 1, 2002&#x02013;June 30, 2003). Blood pressure <italic>control</italic> is defined by the Georgia Department of Human Resources (DHR) as a reading of less than 140/90 mm Hg, based on the average of at least two blood pressure readings taken on the most recent visit.</p><p>The selection and number of the districts studied was guided by several factors. The primary goal of the study was to examine the critical components of SHAPP in sites with high success rates so that lessons could be shared with other hypertension control programs. In addition to analyzing costs, the full study included focus groups with patients, interviews with clinic and administrative staff, and an examination of the medical records. Time and funding limitations allowed for only two districts to be studied. The two districts were selected based on their success in controlling hypertension, use of different computer systems, geographic diversity between districts, demographic diversity among counties within the districts, and the mix of patients managed solely by the health department with patients managed jointly by the health department and private physicians. Although the characteristics of SHAPP districts and clinics vary widely, the intention is for SHAPP to perform at the same level in all districts. This analysis represents the upper boundary of the potential effectiveness of the SHAPP program.</p></sec><sec><title>Impact of SHAPP compared with plausible alternatives</title><p>After examining administrative data for each district, we extrapolated the number of adverse health events &#x02014; hemorrhagic stroke (<xref rid="B11" ref-type="bibr">11</xref>), ischemic stroke, heart disease (<xref rid="B12" ref-type="bibr">12</xref>), and kidney failure (<xref rid="B13" ref-type="bibr">13</xref>) &#x02014; that would be expected given the level of blood pressure control within each district, and we assigned costs to these events. We then compared cost and health outcomes of SHAPP with two simulated plausible alternatives: 1) no care and 2) the typical treatment received in the private sector nationally (referred to as <italic>usual care</italic>). We chose no care to represent the lower boundary that Georgia's SHAPP patients would receive in the absence of the program, and we chose usual care to represent the upper boundary. If SHAPP were eliminated, its patients could be expected to receive no care (the worst-case scenario) or usual care (the best-case scenario).</p><p>Patients who received no care would be expected to have no costs related to blood pressure control, but they would also be expected to have a higher number of adverse events. Our analysis evaluated whether the number of expected adverse events prevented by SHAPP was sufficient to justify the additional cost of SHAPP preventive treatment.</p><p>Next, we compared SHAPP patients with a scenario in which patients who had characteristics identical to SHAPP patients received usual care. SHAPP patients and patients who received usual care would be expected to differ in the following ways: 1) cost of treatment, 2) level of hypertension control outcomes, and 3) the probability of receiving treatment. Nationally, only 58% of people with hypertension receive any regular preventive care, compared with 100% of SHAPP patients, assuming SHAPP patients seek care and are eligible for the program (<xref rid="B1" ref-type="bibr">1</xref>). This analysis evaluated whether SHAPP resulted in less costly treatment as well as fewer expected adverse health events than usual care. Advocates have suggested that SHAPP is both less expensive for and more effective at controlling blood pressure than care provided in private settings because SHAPP uses evidence-based protocols whereas private providers may substitute alternative protocols; SHAPP uses fewer new, more costly (and not necessarily more effective) prescription drugs; and SHAPP substitutes nurse practitioners for physicians to manage patient care. SHAPP could be more beneficial than usual care by offering full coverage to all patients who are eligible and seek treatment, by providing services at lower costs, or by achieving better outcomes.</p></sec><sec><title>SHAPP data and costs</title><p>We examined typical direct and indirect SHAPP program costs using methodology recommended by the National Panel on Cost-Effectiveness Analysis in Health and Medicine convened by the U.S. Public Health Service (<xref rid="B15" ref-type="bibr">15</xref>). Total program costs include the cost of services (e.g., diagnostic testing, patient visits), medications, and overhead costs. Overhead costs include personnel costs (e.g., salary, benefits) and operating costs (e.g., pharmacy, clinical). Total annual SHAPP clinical costs were supplied for two selected counties in District 1 and District 2. For each district, data were obtained from one county in which the study occurred and in another county selected randomly.</p><p>The Georgia DHR supplied the following data for the selected counties: 1) the number of patients treated and the percentage of patients who achieved controlled blood pressure; 2) the costs of prescription drugs, postage, and overhead; and 3) the cost of clinical services. To estimate clinical costs per patient, we divided the annual total costs per county by the total number of SHAPP patients treated in each county. We assumed that the costs identified in the selected counties were representative of the overall per-patient costs in each health district.</p><p>The percentage of SHAPP patients treated with prescription drugs was calculated by dividing the number of patients who received any prescription drug in each district by the number of patients in that district. The cost per patient for prescribed drugs was calculated by dividing the total annual prescription drug cost in each district by the number of patients who received any drugs. Postage costs per patient were calculated by dividing aggregate postage costs (to mail prescriptions to SHAPP clinics) in each district by the number of patients in each district. Government overhead costs for SHAPP statewide were provided by Georgia DHR for all SHAPP patients statewide. These costs included personnel costs (salary and benefits) and operating costs (pharmacy and clinical). This total cost was converted to a per-person cost by dividing it by the number of statewide SHAPP participants, and this cost was then applied to patients in the respective districts. SHAPP's annual treatment cost per patient was calculated as the sum of per-patient clinical, prescription drug, postage, and state overhead costs.</p></sec><sec><title>Comparison treatment costs</title><p>For comparison purposes, estimates of the number of patients receiving treatment and the expected level of hypertension control for patients who receive treatment were obtained from reported national hypertension trends derived from the third National Health and Nutrition Examination Survey (NHANES III) (<xref rid="B1" ref-type="bibr">1</xref>). Because no exact definition exists of what treatment in the private sector entails, we assumed that such treatment was based on the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) guidelines, which would be comparable to SHAPP treatment (<xref rid="B3" ref-type="bibr">3</xref>). Therefore, we defined usual care (for people who received treatment) as the average number of annual hypertension visits for SHAPP participants (3.71 visits), combined with the same drug treatment used by SHAPP patients in District 1.</p><p>The cost of usual care was estimated by multiplying the average number of annual hypertension visits for SHAPP participants (3.71 visits) by the Centers for Medicaid and Medicare Services (CMS) Medicare reimbursement rate for office visits ($80). To this we added the average per-person drug costs observed in the SHAPP program. We assumed no postage or state overhead costs for usual care. Because not all patients with hypertension receive care, we then multiplied this total per-patient cost for those who receive care by the estimated proportion of patients who receive any treatment (<xref rid="B1" ref-type="bibr">1</xref>). <xref rid="T1" ref-type="table">Table 1</xref> presents a comparison of costs and outcomes for patients in District 1 and District 2 and for patients nationally.</p><p>We noted that prescription costs per patient were higher in District 1. On further examination of the data, we found that even though drug costs per type of prescription were consistent statewide, District 1 used a greater quantity of some drugs such as hydrocortothiazide, hydralazine, and fosinopril. In contrast, per-patient clinical services were more costly in District 2. Although most reported procedure codes are the same in both districts, the data show that some differences in service use exist. Overall, patients in District 2 used more clinical services such as phone consultation, preventive counseling, and laboratory test reviews than patients in District 1. Also, the cost per procedure was routinely more expensive per unit of service in District 2 than District 1. Focus groups with patients and key informant interviews with administrators and staff in both districts indicated that each SHAPP district offers the same basic set of services. The observed differences in costs between the two districts studied, however, indicate that SHAPP implementation varies among districts; the technology used to capture and report cost and usage data may vary as well.</p></sec><sec><title>Outcomes</title><p>We defined SHAPP effectiveness as the proportion of patients with controlled blood pressure in each health district, based on statistics from Georgia DHR annual reports. Using these reported levels of blood pressure control, we then estimated the number of adverse events expected in each health district based on the results of a published statistical model by Flack and colleagues (<xref rid="B12" ref-type="bibr">12</xref>). The model was designed to estimate the annual probability of hemorrhagic and ischemic stroke and heart disease for individuals in three categories of blood pressure treatment and control: 1) treated and controlled, 2) treated but uncontrolled, and 3) untreated and uncontrolled (<xref rid="B12" ref-type="bibr">12</xref>). We used this model to make estimates because time and funding limitations prevented us from observing adverse outcomes or measuring associated costs directly. The Flack study was selected for modeling purposes because it provided the most recent and comprehensive information related to the SHAPP study. It examined the effect of inadequate blood pressure control on selected cardiovascular disease outcomes and analyzed related costs for the U.S. population with hypertension. In addition, the study developed a sophisticated model, provided incidence rates for cardiovascular disease morbidity and mortality, and integrated hypertension statistics from NHANES III and cost estimates for stroke, congestive heart failure, and myocardial infarction (<xref rid="B12" ref-type="bibr">12</xref>). NHANES III was conducted in 1999&#x02013;2000; the published results were the most recent available at the time of the Flack study.</p><p>We calculated the cost-effectiveness of SHAPP by comparing two other treatment possibilities &#x02014; no treatment and usual care &#x02014; based on expected adverse outcomes observed in the absence of a public program. The proportions of the U.S. population with treated and controlled, treated but uncontrolled, and untreated and uncontrolled blood pressure were taken from a published analysis of NHANES III surveillance data for 1999&#x02013;2000 (<xref rid="B1" ref-type="bibr">1</xref>) (<xref rid="T2" ref-type="table">Table 2</xref>). The probability of hemorrhagic stroke based on treatment and control of blood pressure was taken from the published results of a randomized controlled study of more than 45,000 participants in the Netherlands that provided population-based estimates; the goal of this study was to examine the outcomes (i.e., number of strokes) associated with insufficient treatment of hypertension (<xref rid="B11" ref-type="bibr">11</xref>). The probability of ischemic stroke was derived by applying the ratio of ischemic strokes to hemorrhagic strokes identified in the 2001 Medical Expenditure Panel Study (MEPS) (<xref rid="B16" ref-type="bibr">16</xref>) to the probabilities of hemorrhagic stoke identified in the Netherlands study. The probability of kidney failure was derived from two separate studies of hypertension-related adverse events, the second of which studied hypertension-related renal failure (<xref rid="B13" ref-type="bibr">13</xref>,<xref rid="B17" ref-type="bibr">17</xref>). Rates of heart disease and the costs of each expected adverse event were obtained from the Flack simulation model of cardiovascular disease associated with uncontrolled blood pressure (<xref rid="B12" ref-type="bibr">12</xref>). Costs for stroke and heart attack reported in the Flack study represent estimates of inpatient and outpatient costs during 1 year after the adverse event. Inpatient costs represent the majority of costs and were obtained from the National Inpatient Profile (a database derived from the National Hospital Discharge Survey); outpatient costs included typical follow-up care, medications, laboratory tests, and office visits (<xref rid="B12" ref-type="bibr">12</xref>). The cost of treating congestive heart failure was obtained from the economic burden-of-illness estimates from the American Heart Association (<xref rid="B12" ref-type="bibr">12</xref>).</p></sec></sec><sec><title>Results</title><p>The SHAPP program achieved blood pressure control rates of 68.1% in District 1 and 59.7% in District 2 (<xref rid="T1" ref-type="table">Table 1</xref>). The average control rate for all SHAPP districts is 54%, with a range of 41% to 68% (data not shown). The comparative national control rate was 53% for patients in treatment, translating to a 31% control rate for all patients nationally when accounting for individuals who do not seek treatment (<xref rid="T2" ref-type="table">Table 2</xref>) (<xref rid="B1" ref-type="bibr">1</xref>). Annual preventive treatment costs per patient were $132.36 in District 1 and $260.39 in District 2. The average number of clinical services between the two districts and the same pharmaceutical care that was used in District 1 would cost $322 per patient in the private sector. However, because only 58% of patients with hypertension nationally receive preventive treatment, the estimated national annual per-patient cost for treated patients with hypertension ($187.04) was between the annual costs per patient in the two districts.</p><p>Because SHAPP achieved higher blood pressure control rates and offered care to all patients who were eligible and sought treatment, SHAPP patients in both districts were expected to experience lower rates of hemorrhagic stroke, ischemic stroke, heart disease, and kidney failure compared with both other treatment scenarios (no care and usual care) (<xref rid="T3" ref-type="table">Table 3</xref>). Patients in District 1 were expected to experience roughly 10 fewer expected adverse events than if they had received no treatment and seven fewer than if they had received usual care. Patients in District 2 were expected to experience roughly 30 fewer expected adverse events than if they had received no treatment and 21 fewer than if they had received usual care.</p><p>The differences in the number of expected adverse outcomes translated into substantial differences in costs among the three scenarios. Total expected annual costs for SHAPP patients, including both preventive treatment and care related to expected adverse outcomes in District 1, were estimated at $289,617 for no treatment, $323,095 for usual care, and $209,800 for SHAPP treatment. In District 2, costs were estimated at $870,451 for no treatment, $971,070 for usual care, and $848,254 for SHAPP treatment (<xref rid="T4" ref-type="table">Table 4</xref>). In each county, SHAPP was the least expensive of the three treatment scenarios. Total expected costs for SHAPP patients in District 1 were 27.5% below the costs of no treatment, and 35.1% below the costs of treatment offered only through the private sector. In District 2, where treatment costs were higher, total costs of SHAPP were 2.6% below the costs of no treatment and 12.7% below the costs of treatment offered only through the private sector. When examining both districts to provide a more global picture of SHAPP, we found that implementation of the SHAPP program resulted in both lower costs and greater potential health benefits than either of the alternative treatment scenarios. SHAPP saved costs and provided greater health benefits when compared with both no treatment for hypertension and usual care (<xref rid="T5" ref-type="table">Table 5</xref>).</p><p>SHAPP costs differed between the two health districts. <xref rid="T1" ref-type="table">Table 1</xref> shows that although District 1 reported lower costs per patient overall than District 2, District 1 prescribed medications to a greater proportion (94%) of clients than District 2 (63%). District 1 also paid more for medications ($49.56 per patient) than District 2 ($15.19 per patient). In contrast, District 1 used fewer clinical services per patient (8.0) than District 2 (12.6) and paid less for the services.</p></sec><sec><title>Discussion</title><p>SHAPP is an education and direct service program that appears to save costs for the state of Georgia. SHAPP resulted in lower costs and better health outcomes than either no treatment or treatment offered at the average level expected nationally. For the two districts examined, SHAPP was found to be preferable to the other two options because it resulted in both better blood pressure control levels (which are expected to translate into fewer adverse health events), lower treatment costs for those who receive treatment, and lower overall costs per eligible patient. These results were supported by blood pressure control rates of 68.1% in District 1 and 59.7% in District 2. Both these rates exceed the national average control rate for patients in treatment (53%) and for all patients with hypertension, including the untreated (31%). The average control rate for SHAPP during this study period was 54%. Although the SHAPP districts that were evaluated had higher control rates, we believe that these rates can be reached by other districts.</p><p>SHAPP still saved costs when expected adverse outcomes were considered, although the cost savings covered a narrow range of direct medical costs. Had we included lost productivity and deaths associated with the expected adverse events that SHAPP prevented, the benefits of SHAPP compared with the benefits of other treatment options would likely appear even more substantial. This cost analysis was one element of an initial program evaluation that also included both medical record review and focus groups with patients as well as key informant interviews with administrative and clinical staff. This evaluation also identified important components that contribute to high blood pressure control rates: intense patient monitoring, follow-up, access to medication, and counseling. &#x000a0;</p><p>This study is limited by several factors. First, because of funding and time constraints, only two districts were examined. Also, districts with high blood pressure control rates were analyzed because the primary objectives of the study were to examine program components that contributed to successful blood pressure outcomes and to communicate results to other hypertension programs. Limited information on adverse events among SHAPP patients was available, and no actual outcomes in the population were observed. All expected adverse events were inferred from the medical literature using the probability of adverse events based on different levels of blood pressure control. The conclusions are accurate to the extent that populations observed in other studies reflect the characteristics of the SHAPP population. Populations in other studies used for this analysis vary in their similarity to the SHAPP population. When selecting parameter values for the model, we balanced demographic similarity with data completeness, fit of the parameter, and size of the study. For example, Klungel et al used data from a study of more than 45,000 individuals and thus had the statistical power to observe differences that would be missed by other smaller studies. In addition, Klungel et al examined precisely the measure we sought: the probability of stroke given uncontrolled and controlled hypertension.</p><p>Second, medical costs for private-sector care were estimated based on service usage observed in SHAPP. Thus, the results may overestimate the effectiveness of the SHAPP statewide. A better approach for estimation would include direct observation of the costs of preventive services in the private sector and comparison of private-sector costs with SHAPP costs. In addition, a more comprehensive approach would be to examine SHAPP on a statewide basis, including districts with varying rates of success. Future models should use more complicated simulations that vary the number of assumptions.</p><p>On the other hand, the costs assigned in this study were conservative, particularly prescription costs. A more precise estimate of private-sector costs would likely make SHAPP appear more cost-effective because SHAPP provides protocol-driven, evidence-based treatment, and most treatment is provided by nurses instead of physicians. We have no way of knowing how SHAPP patients would use services in the absence of the program, so we must rely on hypothetical scenarios for comparison. Finally, because we used data from two higher-performing SHAPP districts, this study identified the upper range of program effectiveness. However, hypertension control rates representing all SHAPP districts (ranging from 41% to 68%) exceed the rates of hypertension control observed nationally in NHANES (31%).</p><p>This study evaluated the costs and benefits of SHAPP based on observed data on program costs and outcomes and on similar data published in the medical literature translated into adverse events in other settings. These findings show that SHAPP treatment is more cost-effective than no treatment or treatment offered only through the private sector. Compared with the two other plausible scenarios tested, SHAPP resulted in the lowest medical costs and the best patient health outcomes. Given these conclusions, we hypothesize that SHAPP's full coverage of patients is preferable to both no patient care and the average amount of care expected nationally both in terms of costs and health outcomes. It is important to view these results in the context of the growing expense of health care and the importance of implementing prevention programs that are successful and reduce costs.</p><p>However, it is also important to keep the limitations of the study in mind. Promising practices at the state level too often are left unexplored because of a lack of funding or an inability to acquire data. This evaluation provides preliminary evidence of the effectiveness of an intervention based on the Chronic Care Model to control hypertension among disadvantaged individuals, but limited time and resources led to a simple simulation that used data from a limited number of locations and relied on several assumptions. We recommend more extensive and more formal evaluations of SHAPP and other hypertension interventions based on the Chronic Care Model to better understand how these interventions work and to more precisely measure program success.</p><p>Because SHAPP costs are borne by the taxpayer and at least some portion of private care is paid for by consumers, it is important to consider who would bear the costs of caring for adverse outcomes if SHAPP were eliminated. First, because patients in SHAPP are indigent, they would be far more likely to receive no care than the average amount of care received nationally if the program were eliminated. This analysis suggests that such a change would result in a substantial increase of expected adverse events and deaths. Because most SHAPP patients do not carry private insurance, the higher cost of caring for these adverse events would likely fall on already overextended public hospitals, the state Medicaid program, and federally funded indigent care programs. Thus, the elimination of SHAPP would likely result in higher costs for both the state and federal governments, making it financially prudent for Georgia to maintain the program. The higher costs of not providing care due to the occurrence of adverse events far exceed the costs of treating hypertension and preventing those events. More importantly, SHAPP provides access to vital services for indigent Georgians to address this critical, life-threatening health issue.</p></sec>

The REACH 2010 Logic Model: An Illustration of Expected Performance

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Tucker</surname><given-names>Pattie</given-names></name><degrees>DrPH, MPH</degrees><aff>Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention (CDC)</aff><address><email>ptucker1@cdc.gov</email><addr-line>4770 Buford Hwy, NE, Mail Stop K-30, Atlanta, GA 303411</addr-line><phone>770-488-5445</phone></address></contrib><contrib contrib-type="author" corresp="no"><name><surname>Liao</surname><given-names>Youlian</given-names></name><degrees>MD</degrees><aff>Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, CDC, Atlanta, Ga</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Giles</surname><given-names>Wayne H</given-names></name><degrees>MD, MS</degrees><aff>Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, CDC, Atlanta, Ga</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Liburd</surname><given-names>Leandris</given-names></name><degrees>MPH, MA</degrees><aff>Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, CDC, Atlanta, Ga</aff></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>Established in 1999, Racial and Ethnic Approaches to Community Health (REACH 2010) is the Centers for Disease Control and Prevention's (CDC's) cornerstone initiative aimed at eliminating disparities in the health status of African Americans, Alaska Natives, American Indians, Asian Americans, Hispanics, and Pacific Islanders. The CDC supports 40 REACH 2010 community coalitions in designing, implementing, and evaluating community-driven strategies to eliminate health disparities in one or more of six priority areas: breast and cervical cancer screening and management, cardiovascular disease, diabetes, HIV/AIDS, immunizations, and infant mortality (<xref ref-type="bibr" rid="B1">1</xref>). Local strategies incorporate community-based participatory approaches designed to reduce risk factors and the prevalence and impact of chronic diseases. Interventions include individual, family, provider, or community activities focused on the prevention, detection, treatment, and management of one or more of the priority areas. In addition to local evaluation plans, the CDC has a national evaluation strategy for cross-site evaluation of grantee programs to identify and assess successful community partnerships and to determine whether local choices of strategies and interventions produced desired changes in health disparities among racial and ethnic groups. The national evaluation plan has two components: process evaluation and outcome evaluation. The process evaluation collects data to gain insights about coalition characteristics and actions that affect the implementation of the local REACH 2010 program (<xref ref-type="bibr" rid="B2">2</xref>). The outcome evaluation uses surveillance data to determine the impact of local interventions that are implemented to reduce health disparities in racial and ethnic groups (<xref ref-type="bibr" rid="B3">3</xref>). In an effort to provide the REACH 2010 grantees with a clear road map of what is ahead, the CDC developed the REACH 2010 logic model to identify anticipated processes and outcomes and assist communities with evaluation.</p><p>A program logic model is defined as a picture of how an organization does its work &#x02014; the theory and assumptions underlying the program (<xref ref-type="bibr" rid="B4">4</xref>). The program logic model links outcomes (both short term and long term) with program activities or processes and the theoretical assumptions and principles of the program (<xref ref-type="bibr" rid="B4">4</xref>). The logic model helps create a shared understanding of the program's goals and methodology, relating activities to projected outcomes (<xref ref-type="bibr" rid="B4">4</xref>). The basic components of a logic model include factors (resources that potentially enable or limit program effectiveness); activities (techniques, tools, events, and actions of the planned program); outputs (the direct results of program activities); outcomes (changes in attitudes, behaviors, knowledge, skills, status, or level of functioning); impacts (organization-level, community-level, or system-level changes) (<xref ref-type="bibr" rid="B4">4</xref>); and relevant external influences (<xref ref-type="bibr" rid="B5">5</xref>). Variations of the logic model have different names, and these variations are all related to program theory (<xref ref-type="bibr" rid="B5">5</xref>). Logic models come in different shapes and sizes and may be a combination of various program logic models (<xref ref-type="bibr" rid="B6">6</xref>-<xref ref-type="bibr" rid="B8">8</xref>).</p><p>This article describes the logic model developed for REACH 2010 that visually depicts the program's theory of change (<xref ref-type="bibr" rid="B5">5</xref>) for addressing health disparities in local communities. The model is theoretically based and includes the conditions being addressed, activities used to address these conditions, and the expected outcomes of the activities (<xref ref-type="bibr" rid="B4">4</xref>). The REACH 2010 logic model is designed to test the effectiveness of multisite community-based programs in improving the health of racial and ethnic populations. The logic model provides communities with a plausible and sensible model of how the program will work to solve identified problems (<xref ref-type="bibr" rid="B5">5</xref>).</p><p>The REACH 2010 logic model illustrates how a coalition could theoretically produce the desired local health disparity reductions and impacts in racial and ethnic groups. It focuses on the logical approaches of a community coalition that organizes to learn the context of, causes of, and solutions for local health disparities and is prepared to take actions to reduce and eliminate the disparities. It also is a tool used to explain and illustrate program concepts and approaches for key stakeholders. As such, it has assisted REACH 2010 communities in identifying, documenting, and evaluating local attributes in the reduction and elimination of community health disparities.</p></sec><sec><title>Elements of the REACH 2010 Logic Model</title><p>The REACH 2010 logic model (<xref ref-type="fig" rid="F1">Figure</xref>) is a theory model that links theoretical constructs together to explain the underlying assumptions of the program. The theory model is appropriate for complex, multifaceted initiatives aimed at entire communities (e.g., community coalition partnerships addressing chronic disease prevention within the community) (<xref ref-type="bibr" rid="B8">8</xref>). The REACH 2010 logic model helps prioritize aspects of the program that are most critical for tracking and reporting, and it also helps identify data needed for monitoring and improving the program.</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure</label><caption><p>Logic model for Racial and Ethnic Approaches to Community Health (REACH 2010).</p></caption><alt-text>Logic Model</alt-text><alternatives><graphic xlink:href="PCD31A21s01"/><table frame="hsides" rules="groups"><tbody><tr><td valign="top" align="left" rowspan="1" colspan="1">This figure shows the REACH 2010 program logic model. This logic model is a theory model that links together theoretical constructs to explain the underlying assumptions of the program. In general, the figure reads from left to right and is divided into two sections, Planning Phase (on the left) and Implementation and Evaluation Phase (on the right). The Planning Phase has four boxes in it. The first one, &#x0201c;Understanding Context, Causes, and Solutions for Health Disparity,&#x0201d; is in the upper left-hand corner of the model and has an arrow pointing to a box below labeled &#x0201c;Coalition.&#x0201d; The arrow points in both directions to indicate that the coalition continues to gain understanding of the context, causes, and solutions for health disparity after it is formed or expanded. The coalition is responsible for &#x0201c;Planning and Capacity Building,&#x0201d; indicated by a box underneath &#x0201c;Coalition&#x0201d; with a two-directional arrow. The fourth box in the Planning Phase of the model is labeled &#x0201c;Community Action Plan.&#x0201d; This box has an arrow pointing to and from the &#x0201c;Coalition&#x0201d; box, indicating that the coalition produces the Community Action Plan and continues to be involved as the Community Action Plan leads to &#x0201c;Targeted REACH Action&#x0201d; and &#x0201c;Existing Activities,&#x0201d; the first box in the Implementation and Evaluation section of the model.</td></tr><tr><td valign="top" align="left" rowspan="1" colspan="1">Targeted REACH action leads to two side-by-side boxes, &#x0201c;Community and Systems Change&#x0201d; and &#x0201c;Change Agents Change.&#x0201d; Below these two boxes is &#x0201c;Widespread Change in Risk/Protective Behaviors.&#x0201d; This stage of implementation and evaluation is brought about by changes in change agents and the community. Two-directional arrows point to this box from the two &#x0201c;Change&#x0201d; boxes above it. The box below &#x0201c;Widespread Change in Risk/Protective Behaviors" is &#x0201c;Reduced Health Disparity&#x0201d;; these two boxes are also connected by a two-directional arrow. Finally, &#x0201c;Reduced Health Disparity&#x0201d; leads to &#x0201c;Other Outcomes&#x0201d; on its left. &#x0201c;Other Outcomes&#x0201d; may also result from &#x0201c;Widespread Change in Risk/Protective Behaviors&#x0201d; or from &#x0201c;Community and Systems Change.&#x0201d; Both &#x0201c;Other Outcomes&#x0201d; and &#x0201c;Reduced Health Disparity&#x0201d; are linked by an arrow back to &#x0201c;Coalition.&#x0201d;</td></tr><tr><td rowspan="1" colspan="1">The five stages in the Implementation and Evaluation Phase of the model are described as follows:</td></tr><tr><td rowspan="1" colspan="1">The five stages in the Implementation and Evaluation Phase of the model are described as follows:</td></tr><tr><td rowspan="1" colspan="1"> <bold>Stage 1: Planning and Capacity Building</bold> <break/> </td></tr><tr><td rowspan="1" colspan="1">The readiness of a coalition and its members to take action aimed at changing risk/protective behaviors and transforming community conditions and systems in a supportive context to sustain behavior changes over time.</td></tr><tr><td rowspan="1" colspan="1"> <bold>Stage 2: Targeted REACH Action</bold> <break/> </td></tr><tr><td rowspan="1" colspan="1">All intervention activities believed to bring about desired effects. Actions may include a broad range of tactics.</td></tr><tr><td rowspan="1" colspan="1"> <bold>Stage 3: Community and Systems Change and Change Among Change Agents</bold> <break/> </td></tr><tr><td rowspan="1" colspan="1">Community and Systems Change involves changing &#x0201c;risk conditions&#x0201d; by altering the environment within which individuals and groups behave. Change Among Change Agents includes documented changes in knowledge, attitudes, beliefs or behavior among influential individuals or groups with the intent of diffusing similar changes to a broader community population.</td></tr><tr><td rowspan="1" colspan="1"> <bold>Stage 4: Widespread Change in Risk or Protective Behaviors</bold> <break/> </td></tr><tr><td rowspan="1" colspan="1">The changing rates of behaviors linked to health status, either as risk or protective factors, among a significant proportion of individuals in the identified community.</td></tr><tr><td rowspan="1" colspan="1"> <bold>Stage 5: Reduced Health Disparity</bold> <break/> </td></tr><tr><td rowspan="1" colspan="1">Narrowing of gaps in health status relative to an appropriate referent.</td></tr></tbody></table></alternatives></fig></boxed-text><p>Community health initiatives are often shaped by a public health framework that uses technical assistance and evaluation to help build local capacities for addressing identified community concerns (<xref ref-type="bibr" rid="B9">9</xref>). The CDC used the <italic>espoused theory of action</italic> for its framework. Delineating an espoused theory of action involves identifying critical assumptions, conceptual gaps, and information gaps (<xref ref-type="bibr" rid="B10">10</xref>). The conceptual gaps are filled by logic, discussion, and policy analysis. The information gaps are filled by evaluation research (<xref ref-type="bibr" rid="B10">10</xref>). This approach was used to develop the REACH 2010 logic model; that is, the stakeholders developed a framework based on their perceptions of how community coalitions and their partners function (<xref ref-type="bibr" rid="B10">10</xref>) to reduce health disparities. During technical assistance workshops, the CDC introduced the REACH 2010 logic model to grantees in their 1-year planning phase and again during the second through fourth years of their cooperative agreements. The CDC discussed each component of the logic model and perceptions of events that are likely to occur in a community addressing health disparities.</p></sec><sec><title>Planning</title><p>The REACH 2010 program logic rests on several related but distinct components in two separate phases. The first phase, planning, includes the following components: 1) a community coalition forms or expands its network; 2) the coalition develops, plans, and builds capacity; 3) the coalition meets regularly to gain an understanding of the context of, causes of, and solutions for health disparity; and 4) a Community Action Plan (CAP) &#x02014; an intervention strategy designed to reduce levels of disparity within the community &#x02014; is produced. The local evaluation plan is incorporated in the CAP and often includes the contract services of an evaluator. Such expertise is sought from universities, evaluation organizations, private practice evaluators, and others. The CAP takes into account all other activities occurring in the environment that might affect the level of health disparities in a specific community. The actions taken by the REACH 2010 coalition should be aligned to coordinate with other interventions and thereby contribute to a coordinated community initiative aimed at eliminating health disparities and achieving other community outcomes.</p></sec><sec><title>Implementation and Evaluation</title><p>The second phase, implementation and evaluation, includes the following components: 1) the implementation of REACH 2010 targeted activities and the assessment and acknowledgment of existing activities that are aimed at the community; 2) the implementation of targeted actions that are thought likely to bring about changes in the community and systems or changes among change agents; 3) change in widespread risk or protective behaviors in the community of focus; 4) reductions in health disparities; 5) other or unexpected outcomes; and 6) the examination and recognition by the coalition of external influences on the community. It is important to note that in the REACH 2010 logic model, all arrows point in both directions. This aspect of the model allows for flexibility and inevitable self-correction as new conditions occur and new knowledge is attained and incorporated into the CAP. It is important to examine the external conditions under which a program is implemented and how those conditions influence outcomes (<xref ref-type="bibr" rid="B5">5</xref>).</p><p>The REACH 2010 logic model identifies key measurement and evaluation stages. The following is a description of each evaluation stage and related examples of REACH 2010 activities and evaluation measures.</p><sec><title>Stage 1: capacity building</title><p>Many REACH 2010 community-based coalitions were formed or expanded to address health issues. These coalitions are primarily driven by the residents of the community at every stage, including planning, implementation, and evaluation. In the REACH 2010 logic model, capacity building refers to the readiness or ability of the coalition and its members to take action aimed at changing risk or protective behaviors and transforming community conditions and systems so that a supportive environment exists to sustain behavior changes over time. Capacity building among REACH 2010 grantees includes establishing traditional public health partnerships among voluntary organizations (<xref ref-type="bibr" rid="B6">6</xref>), a national nurses' association (<xref ref-type="bibr" rid="B11">11</xref>), a health care delivery system (<xref ref-type="bibr" rid="B11">11</xref>), family services (<xref ref-type="bibr" rid="B12">12</xref>), clinicians (<xref ref-type="bibr" rid="B13">13</xref>), consumers (<xref ref-type="bibr" rid="B13">13</xref>), social service agencies (<xref ref-type="bibr" rid="B13">13</xref>), and the Indian Health Board (<xref ref-type="bibr" rid="B14">14</xref>), as well as nontraditional partnerships such as those between faith-based national organizations and local institutions (<xref ref-type="bibr" rid="B6">6</xref>,<xref ref-type="bibr" rid="B13">13</xref>), a quality assurance foundation (<xref ref-type="bibr" rid="B6">6</xref>), a housing authority (<xref ref-type="bibr" rid="B15">15</xref>), and a gardening group (<xref ref-type="bibr" rid="B15">15</xref>) that are supportive of improving the health and well-being of the community.</p></sec><sec><title>Stage 2: targeted actions</title><p>Targeted actions or interventions are planned, identifiable, and discrete activities included in a program to produce change in the population of focus (<xref ref-type="bibr" rid="B16">16</xref>). The REACH 2010 targeted actions are activities that make up the intervention that is believed to bring about desired effects. Interventions include activities such as distribution of the Gold Card, a patient mini-record for individuals with diabetes (<xref ref-type="bibr" rid="B17">17</xref>); use of health advocates to enroll pregnant individuals in a prenatal care system (<xref ref-type="bibr" rid="B18">18</xref>); formal continuing medical education seminars on cervical cancer (<xref ref-type="bibr" rid="B19">19</xref>); and media education campaigns (<xref ref-type="bibr" rid="B20">20</xref>).</p></sec><sec><title>Stage 3: community and systems change</title><p>Community and systems change refers to changing at-risk conditions by altering the environmental context within which individuals and groups behave, for example, by implementing a neighborhood farmers' market (<xref ref-type="bibr" rid="B15">15</xref>), establishing community walking groups (<xref ref-type="bibr" rid="B15">15</xref>), and offering immunization education in the supplemental nutrition program for Women, Infants, and Children (WIC) and child care and Head Start programs (<xref ref-type="bibr" rid="B21">21</xref>). Change among change agents refers to documented changes in knowledge, attitudes, beliefs, or behaviors among a community's influential individuals or groups with the intent of diffusing similar changes to a broader community population. The change agents might include community health advocates and advisors, lay health workers, or peer health promoters who promote community-level awareness of chronic diseases and their related risk factors (<xref ref-type="bibr" rid="B12">12</xref>); beauty and barbershop operators who help disseminate educational information throughout the community on the prevention of chronic diseases (<xref ref-type="bibr" rid="B22">22</xref>); ministers, physicians, and nurses (<xref ref-type="bibr" rid="B6">6</xref>); and parents who keep their children immunized (<xref ref-type="bibr" rid="B21">21</xref>).</p></sec><sec><title>Stage 4: widespread change in risk or protective behaviors</title><p>Widespread change in risk or protective behaviors occurs when a significant proportion of individuals in the identified community changes behaviors that are linked to health status. Examples of changes in protective behaviors in REACH 2010 communities include an increase in immunization rates among children older than 18 months who were enrolled in an intervention program, from 32% at enrollment to 74% at the 1-year follow-up (<xref ref-type="bibr" rid="B21">21</xref>), and an increase in Papanicolaou test use among Vietnamese women in an intervention group, from 62.1% to 76.9% (<xref ref-type="bibr" rid="B20">20</xref>).</p></sec><sec><title>Stage 5: health disparity reduction</title><p>Health disparity reduction occurs when there is a narrowing of the gap in health status between a racial or ethnic group and an appropriate referent group; for REACH 2010, the referent group is the general population (<xref ref-type="bibr" rid="B23">23</xref>). In addition to the examples provided in the previous paragraph, the actions of a REACH 2010 diabetes coalition have led to better health among African Americans with diabetes; between 1999 and 2002, the gap between African Americans and whites in rates of annual hemoglobin A1c testing, which is used to measure blood glucose control, was virtually eliminated in their communities (<xref ref-type="bibr" rid="B24">24</xref>).</p></sec></sec><sec><title>Measuring Performance</title><p>Measuring a program's performance is a way to address accountability or to collect information that helps stakeholders understand how the program is working (<xref ref-type="bibr" rid="B5">5</xref>), both of which contribute to more informed decision making about improvements needed to enhance the quality of the program. The logic model is a tool that can guide and assess the program implementation and program input (<xref ref-type="bibr" rid="B2">2</xref>). In the REACH 2010 logic model, evaluation stages 1 through 5 are being monitored. The connecting lines are hypothesized linkages or causal relationships that require in-depth study to determine and explain what happened (<xref ref-type="bibr" rid="B2">2</xref>). The arrows in the REACH 2010 logic model and the measurement of the linkages provide information on how community actions or activities are presumed to contribute to reducing or eliminating health disparities (<xref ref-type="bibr" rid="B2">2</xref>).</p><p>The CDC provides the REACH 2010 grantees with assistance in documenting community-level changes, both qualitative and quantitative. Qualitative evaluation emphasizes describing the process and understanding social phenomena and other data that influence the direction of the program. Process-oriented evaluation promotes better understanding of program implementation, the causal events leading to change, and the program components that most influence change (<xref ref-type="bibr" rid="B2">2</xref>). The REACH 2010 grantees document qualitative data related to the logic model stages 1 through 3 in the REACH Information Network (REACH IN) system (ORC Macro International, Calverton, Md). The REACH IN system allows grantees at the local level to perform data entry, storage, and retrieval and to produce graphs and reports. These data also are used by the CDC to monitor the types of local activities used to address health disparities.</p><p>By using epidemiological methods to establish estimates of the programs' effects, the quantitative data are systematically and uniformly collected and used to assess impact (<xref ref-type="bibr" rid="B25">25</xref>). The annual REACH Risk Factor Survey data are collected by a CDC contractor, a National Organization for Research (NORC) at the University of Chicago (<xref ref-type="bibr" rid="B23">23</xref>). The REACH 2010 Risk Factor Survey includes questions related to the respondents' health status; health care access; self-reported body measurements; tobacco use; awareness of hypertension, cholesterol, cardiovascular disease, and diabetes and diabetes care; and receipt of preventive services (<xref ref-type="bibr" rid="B23">23</xref>). The CDC analyzes these data and prepares and disseminates reports (<xref ref-type="bibr" rid="B26">26</xref>). Data files are also issued to grantees for local analysis and interpretation. REACH 2010 grantees use local evaluation plans for more detailed and precise measurement of the program, with an eye toward obtaining data that meet local requirements. The use of multiple methods of data collection, often referred to as <italic>triangulation</italic>, can strengthen the validity of findings if results produced by different methods are congruent (<xref ref-type="bibr" rid="B25">25</xref>).&#x000a0;</p></sec><sec><title>Discussion and Conclusions</title><p>The REACH 2010 logic model is a tool that illustrates the theoretical approach and perspective of 40 community-based programs addressing health disparities in racial and ethnic groups. Using the REACH 2010 logic model is beneficial because it 1) provides an opportunity to communicate and to share stakeholders' ideas and assumptions, 2) establishes a common perception or understanding of the coalitions' probable actions or experiences for addressing health disparities, 3) supports the program design and identifies linkages among program elements, 4) identifies a set of key performance measurement points and evaluation issues, and 5) assists with data collection for local use (<xref ref-type="bibr" rid="B5">5</xref>).</p><p>Measuring the achievement of the goal to reduce or eliminate disparities is an essential precept of the REACH 2010 logic model. Data collected by the REACH 2010 Risk Factor Survey and the REACH IN system will continue to be used to measure national progress toward the goal. The analysis of national data is necessary to ascertain predictors of the REACH 2010 logic model. The CDC is also relying on REACH 2010 coalitions to provide credible evidence to explain how targeted local actions contribute to changes in conditions and behaviors that support the elimination or reduction of health disparities.</p><p>The REACH 2010 logic model represents the initial theoretical framework of the program. We recognize that since 1999, additional planning and refinement of the model have occurred at the community level (<xref ref-type="bibr" rid="B4">4</xref>). As the REACH 2010 logic model process continues to unfold, the stakeholders responsible for implementing the model will be involved in its evaluation. They will assess whether the model has accomplished its goal of capturing the unique perspectives and community processes that guide community coalitions, that is, determining the way and the conditions under which local coalitions actually work to achieve their goals. Lessons learned from REACH 2010 will provide needed practice-based evidence to inform the next generation of community-based programs working to reduce health disparities.</p></sec>

Using the Office of Management and Budget (OMB) Clearance Process in Program Planning and Evaluation

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Martin</surname><given-names>Maurice</given-names></name><degrees>PhD, MEd, CHES</degrees><aff>Centers for Disease Control and Prevention (CDC), Division of Diabetes Translation (DDT)</aff><address><email>beq2@cdc.gov</email><addr-line>4770 Buford Hwy NE, Atlanta, GA 30431</addr-line><phone>770-488-5385</phone></address></contrib><contrib contrib-type="author"><name><surname>Thomas</surname><given-names>Darlene</given-names></name><aff>CDC, DDT, Atlanta, Ga</aff></contrib>

Preventing Chronic Disease

<sec><title>Program Evaluation and Data Collection</title><p>Through well-planned health promotion and education programming, federally funded projects are often leaders in the development and implementation of sound public health practices. In public health practice, program evaluations are used to 1) determine whether program objectives related to health status have been achieved, 2) improve program implementation, 3) increase community support, 4) contribute to a scientific base, 5) provide accountability to community members and other stakeholders, and 6) guide program-specific policy decisions (<xref rid="B1" ref-type="bibr">1</xref>,<xref rid="B2" ref-type="bibr">2</xref>). To monitor program processes and obtain the evidence necessary to demonstrate a program's effectiveness, some form of evaluation involving data collection is usually completed (<xref rid="B3" ref-type="bibr">3</xref>). Evaluation is considered a key component of public health programming (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B4" ref-type="bibr">4</xref>).</p><p>Evaluation methodologies vary significantly among programs, as do data-collection techniques. In some evaluations, it is appropriate to use previously collected data that were not originally intended for measuring a program's processes and outcomes. For instance, mortality or disease rates from county vital records or reportable disease records of state health departments can be used to guide certain health promotion programming decisions. However, generally, comprehensive program evaluations involve planned process measures to gauge completeness and quality of activities leading to outcomes in addition to less formal process measures that come about as adjustments to health promotion programming. More formal evaluation measures are characterized as systematic; they have long-term outcome indicator measurements that are planned before the program implementation (<xref rid="B2" ref-type="bibr">2</xref>-<xref rid="B5" ref-type="bibr">5</xref>).</p></sec><sec><title>The Paperwork Reduction Act (PRA)</title><p>Evaluations of public health programs are often conducted at taxpayers' expense because federal agencies are accountable for the quantity and quality of the information resulting from these evaluations. The U.S. public deserves this, and the U.S. Congress demands it (<xref rid="B6" ref-type="bibr">6</xref>). However, in response to constituent complaints about unnecessary and often redundant data collections, Congress passed the Paperwork Reduction Act (PRA) in 1980 to help balance the public demand for accountability and the resulting paperwork burden on the public (<xref rid="B7" ref-type="bibr">7</xref>). The law was created to ". . . ensure that federal agencies do not overburden the public with federally sponsored data collections," with <italic>burden</italic> being defined as ". . . the total time, effort, or financial resources expended by persons to generate, maintain, retain, or disclose or provide information" (<xref rid="B6" ref-type="bibr">6</xref>). In 1995, more reductions were proposed, and the PRA was amended. Congress mandated a 25% reduction in burden in the 3 years immediately following the amendment's passage (<xref rid="B8" ref-type="bibr">8</xref>,<xref rid="B9" ref-type="bibr">9</xref>).</p><p>Compliance with the PRA is required whenever a federal agency sponsors a data collection by using identical questions, using identical reporting or record-keeping requirements, or asking respondents to provide the same level of information on the same subject involving 10 or more respondents in a 12-month period (<xref rid="B7" ref-type="bibr">7</xref>,<xref rid="B10" ref-type="bibr">10</xref>). The law applies to all federal employees, contractors, people in cooperative agreements, and anyone else&#x000a0;who asks the public for information for the purpose of research, public health practice, program evaluation, or any other reason. The PRA also addresses customer satisfaction inventories, focus group inquiries, all types of surveys, telephone interviews, and electronic environments. One notable exception to the PRA is that federal employees may be solicited for information if the solicitation is in their line of work and provides information that is relevant to their work experience (<xref rid="B10" ref-type="bibr">10</xref>).</p></sec><sec><title>Office of Management and Budget (OMB) Clearance</title><p>When the information gathering intended for evaluation meets the parameters specified in the PRA, the federal agency sponsoring the data collection applies for review and approval before the data collection begins. The U.S. Office of Management and Budget (OMB) oversees all requests for review and approval under the PRA (<xref rid="B10" ref-type="bibr">10</xref>). The agency sponsoring the data collection must submit an OMB clearance package explaining and justifying the data collection. In addition, two notices with subsequent 30- and 60-day comment periods must be published in the <italic>Federal Register,</italic> a daily U.S. government publication of the National Archives and Records Administration. OMB clearance typically takes 6 to 9 months (<xref rid="B10" ref-type="bibr">10</xref>). However, some OMB clearance processes take more than 12 months, partially because of the overwhelming number of submissions and limited number of staff members (<xref rid="B9" ref-type="bibr">9</xref>). To compensate for the PRA requirements and time required to obtain OMB clearance, evaluation plans should begin months before data collection begins.</p><p>Although the time involved in obtaining OMB clearance can pose a considerable challenge to evaluators, the process can also significantly improve the quality of the data collection. Program planners and evaluators should work together to focus their efforts and use ethical data-collection methods to obtain useful and necessary data. Such cooperation improves the program results while decreasing the public's burden.</p><p>During the clearance process, OMB requires detailed descriptions of the following (<xref rid="B10" ref-type="bibr">10</xref>):</p><list list-type="order"><list-item><p>The reasons the data collection is necessary</p></list-item><list-item><p>The purpose and use of the information that will be obtained from the data collection</p></list-item><list-item><p>The use of improved information technology to reduce burden</p></list-item><list-item><p>Efforts to identify the duplication and use of similar information</p></list-item><list-item><p>The possible impact of the data collection on small businesses or other small entities (e.g., other people or groups)</p></list-item><list-item><p>The consequences of collecting the data less frequently than planned</p></list-item><list-item><p>Special circumstances relating to Title 5, Part 1320.5 ("Controlling Paperwork Burdens on the Public") of the Code of Federal Regulations (CFR) guidelines about federal information collections</p></list-item><list-item><p>Comments in response to the <italic>Federal Register</italic> notice and evidence of efforts to consult with individuals other than those in the agency collecting the information</p></list-item><list-item><p>An explanation of any payments or gifts to respondents</p></list-item><list-item><p>Confidentiality assurances provided to respondents</p></list-item><list-item><p>Justification for sensitive questions</p></list-item><list-item><p>An estimate of annualized cost and burden hours</p></list-item><list-item><p>Annualized government costs&#x000a0;</p></list-item><list-item><p>An explanation for any program changes or adjustments</p></list-item><list-item><p>A plan for tabulation and publication and a time schedule</p></list-item><list-item><p>Any reasons that an OMB expiration date might be inappropriate to post on instrumentation (i.e., all PRA surveys and materials)</p></list-item><list-item><p>Exceptions to certification for the PRA submission</p></list-item></list></sec><sec><title>Conclusion</title><p>Careful attention to the requirements for OMB clearance enhances program efficiency and improves evaluation processes and outcomes. The requirements are consistent with academic and professional recommendations that program planning be carried out with evaluation in mind; the two processes need to be simultaneous for best results (<xref rid="B3" ref-type="bibr">3</xref>,<xref rid="B11" ref-type="bibr">11</xref>).</p><p>The general public recognizes that the OMB review and approval process is a positive procedure that holds the federal government accountable for the public's burden to provide data. Federal agencies that are striving to provide evidence that tax dollars are being well spent consider OMB compliance to be a difficult but necessary obstacle to overcome. The PRA, which is enforced by the OMB, is a reasonable compromise between reducing paperwork burdens on the public and maximizing the benefits of data collection to ensure that well-planned public health programs have meaningful evaluations. Knowledge about and compliance with the PRA requirements are the essential component of the compromise, improving taxpayer satisfaction and government accountability. Fortunately, all necessary information about the PRA and OMB is available online from <ext-link xlink:href="www.whitehouse.gov/omb/inforeg/infocoll.html" ext-link-type="uri">www.whitehouse.gov/omb/inforeg/infocoll.html</ext-link>, which guides program planners and evaluators through the process.</p></sec>

Provision of School-based Preventive Oral Health Services to Medicaid Beneficiaries

Could not extract abstract

<contrib contrib-type="author"><name><surname>Redmond</surname><given-names>Anne R</given-names></name><degrees>MPH, CHES</degrees><aff>Bureau of Health, Department of Health and Human Services, Augusta, Me</aff><aff>At the time this work was completed, Ms Redmond was affiliated with the Public Health Prevention Service, Centers for Disease Control and Prevention, Atlanta, Ga.</aff></contrib><contrib contrib-type="author"><name><surname>Martin</surname><given-names>Nancy</given-names></name><degrees>MS, RDH</degrees><aff>Oral Health Program, Department of Health and Human Services, Concord, NH</aff></contrib>

Preventing Chronic Disease

<sec><title>To the Editor:</title><p>Children from families who qualify for Medicaid, although more likely to receive dental care than uninsured children (<xref rid="B1" ref-type="bibr">1</xref>), are less likely to receive dental care than children from middle-income and upper-income families (<xref rid="B2" ref-type="bibr">2</xref>). School-based or school-linked oral health programs can play a key role in facilitating regular access to preventive oral health services for children participating in the Medicaid program (<xref rid="B3" ref-type="bibr">3</xref>). The New Hampshire Department of Health and Human Services recently assessed the extent to which Medicaid participants are being served through school-based oral health programs that treat children without access to dental services.</p><sec><title>Methods</title><p>We analyzed data from all school-based oral health programs in New Hampshire that billed Medicaid for preventive oral health services during July 2000 through June 2003. The number of Medicaid children served by year and service type was examined. All children in selected grades (ranging from kindergarten through grade 12 with grades one through three consistently served) in participating schools were eligible for oral health services. Preventive services were provided only to children who had not seen a dentist in the past year and who had parental consent. Preventive services included prophylaxis, fluoride application, oral health education, and sealants, which were reimbursed at the following intervals: prophylaxis, twice per year; fluoride application, once per year; oral health education, once every 3 years; and sealants, once every 5 years. Because children were eligible for services each year, total counts might have included the same beneficiaries more than once.</p></sec><sec><title>Results</title><p>During July 2000 through June 2003, 6 (40.0%) of 15 school-based oral health programs geographically dispersed across the state billed Medicaid for preventive oral health services. There were 25,895 children eligible for oral health services at schools served by the six school-based oral health programs. Of these, 10,859 (41.9%) were screened to determine oral health needs, 2739 (10.6%) received preventive services, and 1024 (4.0%) were Medicaid beneficiaries for whom a bill was submitted (<xref rid="F1" ref-type="fig">Figure</xref>). The majority of the 1024 Medicaid beneficiaries received prophylaxis (75.7%), fluoride application (75.0%), and oral health education (67.2%) (<xref rid="T1" ref-type="table">Table</xref>); only 4.3% received sealants. Overall, New Hampshire school-based oral health programs billed Medicaid for 1024 (37.4%) of 2739 children who received preventive services from 2000&#x02013;2003.</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure</label><caption><p>Delivery of school-based oral health services to children enrolled in Medicaid, New Hampshire, July 2000 through June 2003.</p></caption><alt-text>This flowchart shows the stages in delivery of school-based oral health services to children in New Hampshire enrolled in Medicaid during July 2000 through June 2003. The figure consists of four boxes and reads from top to bottom. Each of the first three boxes has an arrow that points down to the box below. The first box is the number of children eligible for oral health services (N = 25,895). This box leads to a box labeled &#x0201c;Children screened&#x0201d; (N = 10,859). The number of children screened leads to a box labeled &#x0201c;Children receiving preventive oral health services&#x0201d; (N = 2739); and this box leads to the last box at the bottom of the figure, &#x0201c;Children receiving preventive oral services billed to Medicaid&#x0201d; (N = 1024).</alt-text><graphic xlink:href="PCD31A28s01" position="float"/></fig></boxed-text></sec><sec><title>Discussion</title><p>A substantial number of Medicaid beneficiaries received preventive oral health services through New Hampshire school-based programs. The majority of Medicaid beneficiaries received prophylaxis, oral health education, and fluoride application. A smaller proportion of beneficiaries received sealants, in part because only three of the six school-based programs included in this study were providing this service. Challenges exist in delivering preventive services, including obtaining parental consent and Medicaid billing information and meeting the state requirement for an examination by a dentist before sealant placement. School nurses often play a role in communicating with parents about school-based oral health services. Some programs have designated staff to work individually with parents to gain consent and Medicaid information. Because sealants are one of the most effective methods of preventing tooth decay (<xref rid="B3" ref-type="bibr">3</xref>,<xref rid="B4" ref-type="bibr">4</xref>) and school-based sealant programs can reduce oral health disparities among children (<xref rid="B5" ref-type="bibr">5</xref>), a 3-year statewide sealant project has been established in collaboration with the New Hampshire Dental Society and volunteer dentists to perform on-site dental examinations and deliver school-based sealant services. Children not eligible for sealants because of untreated decay are linked to area dentists for restorative care. School-based programs provide access to and provision of preventive oral health services for hard-to-reach populations, contributing to increased oral health and the prevention of tooth decay. Reimbursement from Medicaid presents an opportunity for oral health programs to leverage financial support to enhance services.</p></sec></sec>

The Challenge of Preventing Cardiovascular Disease in Tunisia

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Ghannem</surname><given-names>Hassen</given-names></name><degrees>MD, MSc</degrees><aff>Department of Epidemiology, University Hospital Farhat Hached</aff><address><email>Hassen.Ghannem@rns.tn</email><addr-line>4000 Sousse, Tunisia1</addr-line><phone>+216-73-219-496</phone></address></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>Chronic disease, particularly cardiovascular disease (CVD), is the major cause of death in most developed countries (<xref rid="B1" ref-type="bibr">1</xref>,<xref rid="B2" ref-type="bibr">2</xref>), despite the downward trend observed during the last three decades (<xref rid="B3" ref-type="bibr">3</xref>-<xref rid="B5" ref-type="bibr">5</xref>). The risk factors for chronic disease are also well known in most industrialized countries (<xref rid="B6" ref-type="bibr">6</xref>-<xref rid="B11" ref-type="bibr">11</xref>), and knowledge of risk factors has led to implementation of effective preventive programs (<xref rid="B12" ref-type="bibr">12</xref>,<xref rid="B13" ref-type="bibr">13</xref>). Although CVD is emerging in developing countries, little is known about the level of CVD risk factors in these countries (<xref rid="B14" ref-type="bibr">14</xref>,<xref rid="B15" ref-type="bibr">15</xref>). The problems of chronic disease are more serious for developing countries because many of them have not yet conquered communicable diseases, and their health systems are ill prepared to provide the costly care required for chronic diseases. Despite the new interest in and emphasis on public health and disease prevention in developing countries, it appears that the challenge of controlling CVD remains.</p><p>Tunisia is now facing the phenomenon of epidemiologic transition (<xref rid="B16" ref-type="bibr">16</xref>): total mortality is decreasing, life expectancy is increasing, and lifestyles associated with chronic disease, particularly diabetes and CVD, are being adopted (<xref rid="B17" ref-type="bibr">17</xref>,<xref rid="B18" ref-type="bibr">18</xref>). With this transition, the health care system in Tunisia is challenged with the expansion of chronic disease. Environmental and behavioral changes &#x02014; such as new dietary habits, the lack of physical activity, and the stresses of urbanization and work conditions &#x02014; can lead to the rise of CVD and its risk factors.</p><p>The major CVD risk factors &#x02014; high blood cholesterol, high blood pressure, cigarette smoking, physical inactivity, and unhealthy diet &#x02014; satisfy the public health criteria of causality (<xref rid="B19" ref-type="bibr">19</xref>). In fact, strong epidemiological evidence suggests that these risk factors explain at least 75% of new cases of coronary heart disease (CHD) each year. Available evidence supports the feasibility and effectiveness of population-wide prevention programs directed toward increasing the proportion of people at low risk for CVD. The public health effort should be directed to this population-based approach (<xref rid="B20" ref-type="bibr">20</xref>). Evidence shows that several risk factors and conditions are commonly associated with major chronic diseases. This means that integrated actions against selected risk factors (i.e., smoking, physical inactivity, and unhealthy diet) implemented within the social context can lead to the reduction of major chronic disease; the Countrywide Integrated Noncommunicable Diseases Intervention (CINDI) illustrates this idea at work (<xref rid="B21" ref-type="bibr">21</xref>).</p><p>In Tunisia, a much-needed community-based intervention program to control CVD is being planned. This program will promote healthy living, smoke-free air, healthy nutrition, regular physical activity, and supportive living and working environments. Its ultimate goal is to reduce the burden of CVD.</p></sec><sec><title>The Burden of CVD in Tunisia</title><p>In the context of epidemiologic transition and disease prevention, it is important to be able to assess the scope of existing CVD, but there is a lack of systematic monitoring of CVD morbidity and mortality in Tunisia and in most of the developing world. In Tunisia, the only available population-based epidemiologic data is a profile of CVD risk factors (<xref rid="B22" ref-type="bibr">22</xref>-<xref rid="B24" ref-type="bibr">24</xref>). In 1996, we conducted an epidemiological survey of a representative household sample (n = 957) of the adult urban population in Sousse. The objectives of the study were 1) to estimate the prevalence of the primary CVD risk factors in an urban context and 2) to collect data that would serve as a baseline for assessing future trends in risk factors. We observed a high prevalence (18.8%) of hypertension (blood pressure greater than 160/95 mm Hg); 28.8% had blood pressure greater than 140/90 mm Hg. Of the survey participants, 10.2% had a history of diabetes and 27.7% were obese (body mass index [BMI] &#x02265;30), with a significantly higher rate among women (34.4%). The rate of android obesity (i.e., distribution of fat concentrated in the center of the body) was 36.0%, and the rate of smoking was 21.5%, with a significantly higher rate among men (61.4%).</p><p>Based on this profile of CVD risk factors, Tunisia can be compared with Western societies and should consider a national strategy of primary prevention and heart-health promotion in addition to the efforts recently made in secondary prevention of some chronic diseases such as hypertension and diabetes.</p><p>Urbanization is expected to raise the level of CVD risk factors in Tunisia as a result of the adoption of new dietary habits, the lack of physical activity, and the stresses of working conditions in urban areas. Changes in dietary habits have not yet reached a critical point, however; a study published in 1999 showed that the food regimen in Tunisia was still based mainly on carbohydrates, and the percentage of lipid calories did not reach 20% of total intake (<xref rid="B25" ref-type="bibr">25</xref>).</p><p>The assessment of CVD risk factors during childhood is also important because the underlying process of developing CVD starts early in life. The assessment of CVD risk factors before adulthood could inform us about the etiology of CVD. The results would serve as basic information in a public health context, notably in the promotion of healthy lifestyles. In 1999, we conducted a study of students aged 13 to 19 years; the study was partially funded by the Tunisian Ministry of Higher Education, Scientific Research and Technology (<xref rid="B26" ref-type="bibr">26</xref>) and was based on a representative sample of 1569 youths in Sousse to assess the following CVD risk factors: hypertension, diabetes, hypercholesterolemia, dyslipoproteinemia, obesity, lack of physical activity, and smoking. The study established local reference values for the percentile distribution of CVD risk factors. The study showed that BMI, diastolic blood pressure, total cholesterol, low-density cholesterol, and high-density cholesterol were significantly higher for girls than for boys. However, boys had significantly higher levels of systolic blood pressure. In addition, 7.9% of the study population was obese; a greater percentage of girls (9.7%) was obese than boys (6.0%). Overweight was also significantly higher for girls (16.1%) than for boys (11.1%). Smoking was reported by 7.6% of the population; the percentage of boys who smoked (14.7%) was significantly higher than for girls (1.1%).</p></sec><sec><title>Effectiveness of a Community-based Intervention Program</title><p>Available evidence supports the feasibility and effectiveness of population-wide prevention directed toward increasing the proportion of people at low risk of developing CVD (<xref rid="B19" ref-type="bibr">19</xref>). CVD risk factors can be linked directly to social, economic, and environmental determinants of health. Factors that have a major impact on the development of chronic diseases include education, availability and affordability of healthy foods, access to health services, and infrastructures that support a healthy lifestyle (<xref rid="B27" ref-type="bibr">27</xref>). Advances in etiological research of CVD have resulted in numerous intervention projects and programs throughout the developed world. The scope of these activities is wide, from preventive action on a single risk factor (e.g., tobacco use, hypertension) or on a single disease (e.g., CHD) to a more comprehensive approach involving several risk factors common to several chronic diseases (<xref rid="B21" ref-type="bibr">21</xref>). These programs have clearly demonstrated the feasibility of comprehensive community-based cardiovascular programs and the need to extend intervention activities to other chronic diseases. The North Karelia project in Finland is a good example of how a demonstration project can be expanded to a national level (<xref rid="B28" ref-type="bibr">28</xref>).</p><p>For developing countries, evidence supports the benefits of lowering risk distributions. For example, the Asia Pacific Cohort Studies Collaboration indicates that a 2% reduction of mean blood pressure has the potential to prevent 1.2 million stroke deaths annually (approximately 15% of all stroke deaths) and 0.6 million coronary deaths (6% of all CHD deaths) by the year 2020 (<xref rid="B29" ref-type="bibr">29</xref>). Much more attention should also be directed to modifying the environmental determinants of physical inactivity and the resulting obesity (<xref rid="B30" ref-type="bibr">30</xref>).</p><p>One could expect a long-term positive impact of a community-based intervention program to control CVD in developing countries, based on the accumulated experiences of developed countries. The CVD epidemic is not a matter of fate in developing countries; it can be controlled or at least postponed in a country with a transitional economy, such as Tunisia. Because communicable diseases have not yet been conquered, the health care system in Tunisia is ill prepared to provide the costly care required for managing chronic diseases, but some efforts have been made recently. A national program was launched in 1993 by the Ministry of Health; this program was designed to improve the quality of care for patients with diabetes and hypertension in primary care settings. Its main objective was to standardize the management of these two CVD risk factors with a focus on general physician training and health education. This program is insufficient, however, and needs to be completed with a primary prevention approach, which is the main objective of our current initiative.</p></sec><sec><title>Steps to Launch a Community-based Intervention Program</title><p>In Tunisia, we are in the initial stages of documenting the CVD burden and identifying its risk factors. This phase should be followed by a population-based preventive intervention to control the expansion of CVD. The potential for prevention generated by a community-based intervention program is important, and so is the idea of building collaboration among different categories of professionals toward a common goal of CVD prevention. In Tunisia, the launch of such an intervention would represent a cornerstone for a chronic disease control program. To launch the intervention, we should adopt a stepwise approach, described below.</p><sec><title>Step 1: community mobilization and assessment</title><p>In this phase, we will focus on community mobilization and establish a task force of politicians, health professionals, and nongovernmental organizations committed to decreasing CVD in Sousse. This stage will also include a community assessment with the following parameters: levels of CVD risk factors; knowledge, attitudes, and behaviors related to CVD risk factors; patterns of use of primary health services; and barriers to and facilitators of healthy behavior.</p></sec><sec><title>Step 2: intervention planning and implementation&#x000a0;</title><p>First, the task force will use the community assessment data for setting priorities. Second, educational and skill-building activities will be coordinated through existing social and organizational structures such as primary health care centers, work settings, secondary schools, health clubs, and community recreational facilities.</p></sec><sec><title>Step 3: sustainability and evaluation</title><p>This phase will focus on the sustainability and evaluation of the intervention based on its feasibility and impact on CVD in Sousse. Evaluation is a continuous, systematic process of determining what has been achieved in a program and comparing these achievements with the objectives set out in the plan. Evaluation should not take place only at the end of the program; it must be considered in the planning phase of the program, as soon as program activities begin, and should continue until after the activities have ended. Components of evaluation should include the following:</p><list list-type="bullet"><list-item><p> <bold>Implementation.</bold> This component systematically examines what is delivered and how it is delivered. It does not focus on program results but provides documentation of program delivery.</p></list-item><list-item><p> <bold>Impact.</bold> This component assesses the intervention's ability to produce positive changes in levels of risk factors and other determinants of CVD (e.g., physical activity, BMI, healthy diet, knowledge, attitudes, behaviors).</p></list-item><list-item><p> <bold>Outcome.</bold> This component assesses the effectiveness of the intervention in producing long-term changes in CVD among the target population of Sousse.</p></list-item></list></sec></sec><sec><title>Integrated Program to Control CVD in Sousse</title><p>Sousse is the fourth largest city in Tunisia, with a total population of 547,000, according to the most recent census in 2004. There are 92 primary health care centers and two university hospitals. The total number of primary health care physicians is 168; 92 physicians are in the private sector and 76 are in the public sector. There are 57 secondary schools with a total of 1762 classes and 56,752 students. Sousse has 413 worksites that employ more than 50 employees each.</p><sec><title>Description of the program</title><p>The integrated program will target youths as well as adults through lifestyle-education activities within the general framework of community mobilization. Improving the preventive practices of health professionals at different levels of care will also be central to the program. The program will address the entire community (from symptom-free individuals to high-risk individuals) and propose interventions centered on promoting healthy habits (e.g., smoking abstinence, balanced nutrition, sustained physical activity) and preventing main risk factors (e.g., arterial hypertension, diabetes, hypercholesterolemia) toward the ultimate goal of reducing or delaying CVD.</p><p>The program focuses not only on interventions targeting internal factors that are controlled by each individual but also on the external environmental factors not individually controlled. The effectiveness of interventions on behavior modification (adoption of healthy habits) represents our primary challenge.</p></sec><sec><title>Program objectives</title><p>The program objectives include a 10% relative improvement in healthy habits in 5 years in the target population of Sousse, including the following:</p><list list-type="bullet"><list-item><p>10% reduction in the proportion of sedentary adults</p></list-item><list-item><p>10% reduction in the prevalence of adult smokers&#x000a0;</p></list-item><list-item><p>10% reduction in the prevalence of young smokers&#x000a0;</p></list-item><list-item><p>10% reduction in the prevalence of obese adults&#x000a0;</p></list-item><list-item><p>10% increase in the proportion of the adult population consuming 5 daily servings of fruits and vegetables</p></list-item></list><p>The program will be implemented gradually with the following objectives:</p><list list-type="bullet"><list-item><p>To implement the intervention annually among 30% of the physicians in the private sector and 30% of the physicians in the public sector</p></list-item><list-item><p>To implement the intervention annually among 30% of worksites with more than 50 employees</p></list-item><list-item><p>To implement the intervention annually among 30% of the secondary schools&#x000a0;</p></list-item></list><p>Other objectives may be added during implementation as we adapt to the process; for example, we may add other target groups such as pharmacists and nurses.</p></sec><sec><title>Program strategies</title><p>The program will emphasize an integrated approach that combines two strategies: educational actions and environmental actions.</p><p> <bold>Educational actions</bold> </p><p>The educational actions will aim to encourage physicians and other primary health care professionals to introduce a brief counseling session to patients on the healthy habits related to CVD risk factors. For physicians, workshops of continuing medical education will focus on managing CVD risk factors in primary care.</p><p> <bold>Environmental actions</bold> </p><p>The environmental actions will aim to modify the environments of work settings, schools, and the general community. Three types of environments are targeted: physical, economic, and social. At worksites, we will actively engage the support of members of management and occupational health and medicine groups. We plan to approach a new group of schools each year during the program so that by the end, we reach all schools in the region. In the community, we will focus on leisure services and sports facilities to help them develop and implement public policy to modify environments to promote the adoption of healthy habits. The promotion of healthy lifestyles in these environments is intended to create smoke-free environments, encourage smoking cessation, promote the consumption of fruits and vegetables, and encourage regular physical activity. These dimensions are summarized in the Figure.</p><boxed-text position="float"><fig position="float" id="F2" fig-type="diagram"><label>Figure</label><caption><p>Integrated program of chronic disease control, Sousse, Tunisia.</p></caption><alt-text>Flow Chart</alt-text><graphic xlink:href="PCD31A13s01" position="float"/></fig></boxed-text></sec></sec><sec><title>Conclusion</title><p>Tunisia is a country with a transitional economy that faces the challenge of an increase in morbidity associated with chronic disease. With this epidemiologic transition, the health care system must be prepared to address the growing expansion of chronic disease, particularly CVD. Many epidemiologic studies have demonstrated high levels of CVD risk factors among Tunisian adults and children. The time for prevention is now. There is evidence that several risk factors and conditions are commonly associated with major chronic diseases. This means that integrated actions against selected risk factors (smoking, physical inactivity, and unhealthy diet) implemented within the social context can lead to the reduction of major chronic disease. Prevention of chronic disease should focus on decreasing risk factors early in childhood. The community-based program in Tunisia will target access to positive healthy living, smoke-free air, healthy nutrition, regular physical activity, and supportive living and working environments. Its ultimate goal is to reduce the burden of CVD. The implementation of this program is a public health priority and will serve as a useful example for other developing countries.</p></sec>

Transforming a Master of Public Health Program to Address Public Health Practice Needs

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Woodhouse</surname><given-names>Lynn D</given-names></name><degrees>EdD, MEd, MPH, CHES</degrees><role>Professor and MPH Program Director</role><aff>East Stroudsburg University</aff><address><email>cwoodhouse@po-box.esu.edu or woodhouseld@comcast.net</email><addr-line>200 Prospect St, East Stroudsburg, PA 18301</addr-line><phone>570-422-3702</phone></address></contrib><contrib contrib-type="author"><name><surname>Cardelle</surname><given-names>Alberto C</given-names></name><degrees>PhD, MPH</degrees><role>Associate Professor, Chair, and Director</role><aff>Health Services Undergraduate Program, Health Department</aff></contrib><contrib contrib-type="author"><name><surname>Godin</surname><given-names>Steven W</given-names></name><degrees>PhD, MPH, CHES</degrees><role>Professor and Director of Community Health Undergraduate Program</role><aff>Health Department</aff></contrib><contrib contrib-type="author"><name><surname>Shive</surname><given-names>Steven E</given-names></name><degrees>PhD, MPH, MA, CHES</degrees><role>Assistant Professor</role><aff>Health Department</aff></contrib><contrib contrib-type="author"><name><surname>Williams</surname><given-names>Tonya L</given-names></name><role>Administrative Assistant</role><aff>Health Department</aff></contrib><contrib contrib-type="author"><name><surname>Bitto</surname><given-names>Adenike C</given-names></name><degrees>MDDS, DrPH, MPH, CHES</degrees><role>Associate Professor</role><aff>Health Department, East Stroudsburg University, East Stroudsburg, Pa</aff></contrib><contrib contrib-type="author"><name><surname>Brensinger</surname><given-names>Elizabeth A</given-names></name><degrees>MPH</degrees><role>Adjunct Assistant Professor, Member, East Stroudsburg University Master of Public Health Community Advisory Board</role><aff>East Stroudsburg University, East Stroudsburg, Pa, and Consultant, RedRoad Enterprises, New Tripoli, Pa</aff></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>Public health is an interdisciplinary profession undergoing dynamic but sometimes conflicting changes. The future of public health can be enhanced by emphasizing the development of practice skills (<xref rid="B1" ref-type="bibr">1</xref>-<xref rid="B5" ref-type="bibr">5</xref>).</p><p>To increase the effectiveness of the public health profession and progress toward the goal of having the healthiest possible population, various adaptable approaches for improving graduate public health training and the skills of the public health workforce continue to emerge (<xref rid="B4" ref-type="bibr">4</xref>,<xref rid="B6" ref-type="bibr">6</xref>-<xref rid="B9" ref-type="bibr">9</xref>). It is important that emerging approaches to public health training support the development of competency-based training grounded in curriculum models (<xref rid="B2" ref-type="bibr">2</xref>). It is equally important that the training programs support the broad vision of ensuring social justice and promoting the elimination of health disparities.</p><p>East Stroudsburg University (ESU) is one of 14 institutions in the Pennsylvania State System of Higher Learning. Faculty members in ESU's accredited Master of Public Health (MPH) program have a documented history of improving the quality of its curriculum (<xref rid="B7" ref-type="bibr">7</xref>). The recent program change described in this article provides a potential model for refocusing graduate public health programs on community health and highlighting community-health education and practice. Given the need for quality assurance, the emphasis on outcomes, and the competition for social jurisdiction among overlapping professions (<xref rid="B8" ref-type="bibr">8</xref>), this model of change may help other programs work toward similar goals.</p></sec><sec><title>The Change Process</title><p>To ensure that the accredited MPH program at ESU continued to meet the public health needs of our communities and region (northeastern Pennsylvania) in a rapidly changing environment, faculty members who taught required public health courses (the ESU MPH Public Health Faculty Council) in 2003 began developing an adaptable model for change. With input from students, graduates, and community public health professionals, we began a multistage, interactive, community-focused process to transform our graduate public health education training program. The new curriculum was approved in fall 2004.</p><p>The program transition was based on a feedback system that highlighted the need for change at every program level. The first goal was to ensure that all program elements incorporated the appropriate program vision, values, mission, goals, and objectives, or VVMGO. The faculty members decided that the program VVMGO should focus on social justice and community health from an ecological perspective rather than focus on promoting program success (<xref rid="B3" ref-type="bibr">3</xref>,<xref rid="B10" ref-type="bibr">10</xref>,<xref rid="B11" ref-type="bibr">11</xref>). The second goal was to incorporate opportunities for flexible assessments through continuous data input and evaluation at every level. We envisioned a process that would blend the adaptability needed to use data with our efforts to keep the program focused on community health.</p><p>The figure shows the continual feedback process. The faculty began the process in 2003 with <italic>point 1;</italic> however, other graduate programs could begin using the model at any entry point.</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure</label><caption><p>Feedback loop for the initial and continual development processes for program vision, values, mission, goals, and objectives (VVMGO); student competencies; and the curriculum learning objectives and assessments linked to student competencies. The model was adapted from iterations of planning materials used from 2003 to 2005. COL indicates Council on Linkages Between Academia and Public Health Practice.</p></caption><alt-text>Flow chart</alt-text><long-desc>This figure is a flowchart with eight boxes that are arranged in a clockwise circle and has arrows pointing from one box to the next. Each box contains one point of the program's development. In order, the boxes read as follows: Point 1, continuous and regular review of pertinent documents and national change processes (e.g., COL, Institute of Medicine report) as well as our program evaluation findings; Point 2, development or revision of program-level VVMGO (measurable objectives); validation by faculty; Point 3; development or revision of student competencies that guide planning for student-focused aspects of mission, goals, and objectives and curriculum change; validation by faculty; Point 4, proactive effort to encourage review, editing, and validation of VVMGO and student competencies by students, graduates, community advisors, employers, and stakeholders; Point 5, final and ongoing review, editing, and validation of VVMGO and competencies by faculty; Point 6, two-stage review of current curriculum&#x02019;s ability to support VVMGO and student competency development; Point 7, development or revision of evaluation model displaying the multiple levels of program and curriculum evaluation, including community stakeholder input; and point 8, Input from all levels of evaluation data; VVMGO assessments and curriculum assessments.</long-desc><graphic xlink:href="PCD31A22s01" position="float"/></fig></boxed-text><p>During the 2-year process, five national initiatives affecting the public health workforce influenced our activities: 1) The Institute of Medicine (IOM) report, &#x0201c;Who Will Keep the Public Healthy? Educating the Public Health Workforce for the 21st Century,&#x0201d; which emphasizes practice experience, the ecological model, and expanding the core curriculum of public health (<xref rid="B3" ref-type="bibr">3</xref>); 2) the impact of a more than 10-year dialogue about developing a process for credentialing the public health workforce (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B12" ref-type="bibr">12</xref>); 3) the development of multiple sets of public health competencies from many stakeholders (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B3" ref-type="bibr">3</xref>,<xref rid="B13" ref-type="bibr">13</xref>-<xref rid="B16" ref-type="bibr">16</xref>); 4) the existing graduate roles and competencies for health education (<xref rid="B16" ref-type="bibr">16</xref>, <xref rid="B17" ref-type="bibr">17</xref>); and 5) the current and future requirements for public health program accreditation, including the Council on Education for Public Health (CEPH) accreditation criteria and efforts to blend the MPH in health education concentration and the health education approval and accreditation processes at the graduate and undergraduate levels (<xref rid="B18" ref-type="bibr">18</xref>,<xref rid="B19" ref-type="bibr">19</xref>).</p><p>Four interactive stages of development emerged from an ongoing strategic planning process. The first stage, based on a synthesis of the previously mentioned materials and an examination of regional needs, involved developing a revised draft proposing new program VVMGO. Once approved by the faculty, the draft of the program VVMGO was used to guide the second stage. (A complete list of the revised ESU MPH program VVMGO is available from <ext-link xlink:href="www.esu.edu/mph" ext-link-type="uri">www.esu.edu/mph</ext-link>)</p><p>The second stage involved developing a list of student competencies that were grounded in the ecological model (<xref rid="B3" ref-type="bibr">3</xref>,<xref rid="B11" ref-type="bibr">11</xref>) and organized into 10 domains. These competencies were compiled after an interactive process involving students, graduates, community stakeholders, employers, community advisors, and faculty members (<xref rid="F1" ref-type="fig">Figure</xref>, <italic>points 1</italic> to <italic>4</italic>). (A complete list of the revised ESU MPH program student competencies is available from <ext-link xlink:href="www.esu.edu/mph" ext-link-type="uri">www.esu.edu/mph</ext-link>.) The domains and competencies that emerged are primarily a combination of competency lists from the Centers for Disease Control and Prevention (CDC) Public Health Prevention Service, the Council on Linkages Between Academia and Public Health Practice, the Columbia University School of Nursing, and the Joint Committee for the Development of Graduate-Level Preparation Standards, as well as competencies from many general or discipline-specific approaches provided in the references of the 2003 IOM report and on the CDC Web site (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B3" ref-type="bibr">3</xref>). These competency frameworks were continually synthesized, considering content knowledge and skill development required in each framework. Using a matrix or chart as a guide, the third stage included an examination of each course and required program activity to determine the relevance of selected competencies for the program curriculum. While focusing on the VVMGO, the fourth stage involved examining each relevant competency to determine whether it was a current focus of the program and if so, how it was being assessed, or whether it should become a focus and if so, how it should be emphasized.</p><p>In August and September 2003, the two written drafts were shared with students, graduates, and community public health professionals. The drafts were accompanied by a cover letter describing our interactive process and requesting input, validation, or both (<xref rid="F1" ref-type="fig">Figure</xref>, <italic>point 5</italic>). Participants ranked the importance of goals, objectives, and competencies and suggested changes, validated the draft, or both. We allowed this feedback to be anonymous (although many people signed their submissions), and the information was returned in our envelope.</p><p>Aided by mailed-in information and data from 4 years of program evaluation findings (2001&#x02013;2004), including student and graduate surveys about curriculum value and outcomes of the program, exit-interview summaries, and internship preceptor interviews, we began the process of revising the curriculum. The revision focused on the courses, student assessments, and program requirements, with the goal of ensuring that graduates who completed the program would have the identified competencies (<xref rid="F1" ref-type="fig">Figure</xref>, <italic>point 6</italic>). Learning objectives for several courses were changed, experiential learning was expanded, and new courses were added.</p></sec><sec><title>Integrated Evaluation Processes</title><p>To ensure that program changes were monitored effectively, a modified logic model was created to illustrate the relationships among the revised curriculum requirements, revised course learning objectives, and the program's VVMGO and student competencies (<xref rid="T1" ref-type="table">Table</xref>). The model was developed to help plan the processes for and implementation of the new evaluation, and its development will help us monitor the program's outcomes and allow the faculty to revise the evaluation of the program as needed (<xref rid="F1" ref-type="fig">Figure</xref>, <italic>point 7</italic>). Creating a visual representation of the interactive nature of the program helped the faculty embrace the idea that program success in all areas is necessary to enhance community health and community health practice.</p><p>As part of a CEPH reaccreditation self-study and site visits in spring and fall 2004, meetings were conducted with community advisors, stakeholders, students, graduates, and community public health professionals to obtain&#x000a0; additional input into and final validation of the proposed changes. In addition, the first round of evaluation data using the new processes enhanced this assessment (<xref rid="F1" ref-type="fig">Figure</xref>, <italic>point 8</italic>). Although the revised VVMGO, competencies, curriculum, and evaluation plan have only been implemented for a year, the evaluation &#x02014; including a revised student and graduate survey, revised outcome measures, and a greater emphasis on community stakeholder input &#x02014; has been providing preliminary information. Some successes have been revealed, as have areas that need more emphasis, such as environmental health, in which we need to expand experiences and internship opportunities.</p></sec><sec><title>Value of the Process</title><p>An important product of this holistic process of program change is the impact on the faculty council. Because the change process was grounded in strategic planning, it helped us define what the faculty and the program should be able to accomplish. It also fostered a culture &#x02014; a shared vision &#x02014; of community collaboration that guides the training for public health education practice, applied social behavioral science research, and population-based initiatives emerging from the program's students, graduates, and faculty members (<xref rid="B9" ref-type="bibr">9</xref>,<xref rid="B20" ref-type="bibr">20</xref>-<xref rid="B23" ref-type="bibr">23</xref>). This vision may be atypical for some graduate public health training institutions, but we consider it an important component of a high-quality graduate public health training experience (<xref rid="B22" ref-type="bibr">22</xref>).</p><p>Many alternative approaches can be used to ensure that the VVMGO of a graduate public health training program support and guide the students, graduates, and faculty members and facilitate community efforts to enhance the health of the public. The ESU MPH program model process was successful for the MPH program. In the future, the process will serve as a quality-control mechanism for the evolution of public health worker certification or credentialing. The next step is to use the process to ensure that the ESU undergraduate community-health and health-services programs are effectively linked to the graduate program and can meet public health workforce needs by graduating students with core public health skills, health-services skills, community-health practice and education skills, or all of these. The emerging potential for undergraduate public health program accreditation makes this step essential.</p></sec>

Response to &#x0201c;Old Black Water&#x0201d;

Could not extract abstract

<contrib contrib-type="author"><name><surname>Zanca</surname><given-names>Jane A</given-names></name><degrees>MPH, CHES</degrees><role>Technical Writer and Editor</role><aff>Office of Enterprise Communications, Centers for Disease Control and Prevention, Atlanta, Ga</aff></contrib>

Preventing Chronic Disease

<sec><title>To the Editor:</title><p>Thanks for the wonderful item on the 1927 Mississippi River flood (<xref rid="B1" ref-type="bibr">1</xref>). I grew up in New Orleans and on the Gulf Coast of Mississippi, hearing of storms and wading through floods all of my life, but no one had ever talked about the 1927 events, and this critical southern event was not taught in any history courses at my schools. Reading John Barry's <italic>Rising Tide: The Great Mississippi Flood of 1927 and How It Changed America</italic> (<xref rid="B2" ref-type="bibr">2</xref>) was an epiphany and a joyous discovery. I had always wondered what it was like at the bottom of the river; he could answer that question.</p><p>Barry also describes deliberate destruction of the levee along St. Bernard Parish to spare New Orleans from the rising tide in the 1927 flood. I had heard rumors of the destruction &#x02014; but no mention of the flood in 1927 &#x02014; for years but had never seen documentation to support it until Barry's book. The rumors reemerged in Hurricane Betsy, which drove my family out of their home in Arabi in the middle of the night with flooding from the Lake Pontchartrain end of the Industrial Canal. Many in the flooded communities below the canal firmly believed the levee had been bombed &#x02014; once again to spare New Orleans at the expense of the 9th Ward, Arabi, Chalmette, and other areas. An Arabi neighbor told me that in the decades after Hurricane Betsy, any time a storm was threatening New Orleans, vigilantes from below the Industrial Canal patrolled the levees. I can't testify to that &#x02014; but I wouldn't be surprised.</p><p> <!-- Edit 1 --> <bold>Jane A. Zanca</bold> Technical Writer and Editor, Office of Enterprise Communications, Centers for Disease Control and Prevention, Atlanta, Ga</p></sec>

The Behavioral and Clinical Effects of Therapeutic Lifestyle Change on Middle-aged Adults

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Aldana</surname><given-names>Steven G</given-names></name><degrees>PhD</degrees><aff>College of Health and Human Performance, Brigham Young University</aff><address><email>steve_aldana@byu.edu</email><addr-line>274 SFH, College of Health and Human Performance, Brigham Young University, Provo, UT 84602-2214</addr-line><phone>801-422-2145</phone></address></contrib><contrib contrib-type="author"><name><surname>Greenlaw</surname><given-names>Roger L</given-names></name><degrees>MD</degrees><aff>SwedishAmerican Center for Complementary Medicine, Rockford, Ill</aff></contrib><contrib contrib-type="author"><name><surname>Salberg</surname><given-names>Audrey</given-names></name><degrees>RN</degrees><aff>SwedishAmerican Center for Complementary Medicine, Rockford, Ill</aff></contrib><contrib contrib-type="author"><name><surname>Diehl</surname><given-names>Hans A</given-names></name><degrees>DrHSc, MPH</degrees><aff>Lifestyle Medicine Institute, Loma Linda, Calif</aff></contrib><contrib contrib-type="author"><name><surname>Merrill</surname><given-names>Ray M</given-names></name><degrees>PhD, MPH</degrees><aff>SwedishAmerican Center for Complementary Medicine, Rockford, Ill</aff></contrib><contrib contrib-type="author"><name><surname>Thomas</surname><given-names>Camille</given-names></name><degrees>MS</degrees><aff>College of Health and Human Performance, Brigham Young University, Provo, Utah</aff></contrib><contrib contrib-type="author"><name><surname>Ohmine</surname><given-names>Seiga</given-names></name><aff>Department of Molecular and Microbiology, Brigham Young University, Provo, Utah</aff></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>Chronic diseases such as cancer, cardiovascular disease, stroke, and diabetes are responsible for most deaths in the United States (<xref rid="B1" ref-type="bibr">1</xref>). Between 70% and 90% of these deaths are believed to be caused by poor nutrition, sedentary living, and tobacco use and are preventable (<xref rid="B2" ref-type="bibr">2</xref>-<xref rid="B4" ref-type="bibr">4</xref>). These lifestyle factors appear to play a prominent role in the mechanisms and processes that lead to the development of many chronic diseases. The largest reductions in chronic disease prevalence in the United States will be achieved when individuals adopt and maintain lifestyles that include a healthy diet and regular physical activity.</p><p>During the 1980s, Nathan Pritikin conducted several in-patient lifestyle-change programs that documented how a low-fat, high&#x02013;complex-carbohydrate, high-fiber diet and regular exercise could improve blood lipid levels and insulin sensitivity (<xref rid="B5" ref-type="bibr">5</xref>-<xref rid="B7" ref-type="bibr">7</xref>). Variations of this holistic approach to preventing and arresting chronic diseases have more recently been evaluated in randomized clinical trials such as the PREMIER clinical trial (<xref rid="B8" ref-type="bibr">8</xref>), the DASH dietary study (<xref rid="B9" ref-type="bibr">9</xref>), and other trials in the United States, United Kingdom, and New Zealand (<xref rid="B10" ref-type="bibr">10</xref>-<xref rid="B12" ref-type="bibr">12</xref>). Most of these trials used inpatient treatment or controlled feeding to encourage and monitor changes in diet and physical activity. All of them demonstrated reductions in cardiovascular risk factors, including obesity, blood pressure, and blood lipid levels.</p><p>The Coronary Health Improvement Project (CHIP) was created with the goal of reducing chronic diseases and improving the overall health of the public by providing a lifestyle-change program to both the community and the workplace (<xref rid="B13" ref-type="bibr">13</xref>). The CHIP is a 40-hour live-lecture educational course that highlights the importance of making better lifestyle choices for reducing chronic disease risk factors. A one-group pretest&#x02013;posttest analysis of the program revealed that after 4 weeks, participants significantly reduced their blood pressure, blood glucose, body weight, and total and low-density lipoprotein (LDL) cholesterol (<xref rid="B13" ref-type="bibr">13</xref>). This exploratory study demonstrated that the program had the potential to improve not only coronary risk factors but also the risks associated with cancer, diabetes, and the metabolic syndrome. These results were repeated in a quasi-experimental design that included results from six groups of working adults (<xref rid="B14" ref-type="bibr">14</xref>).</p><p>A large randomized clinical trial was initiated to further explore the effect of the CHIP (<xref rid="B15" ref-type="bibr">15</xref>). Six-week results from this study revealed that adults who completed the program improved their nutrition and physical activity behavior and reduced cardiovascular disease risk factors (<xref rid="B15" ref-type="bibr">15</xref>). We present the behavioral and clinical changes that participants in this therapeutic lifestyle-change program experienced after 6 months.</p></sec><sec><title>Methods</title><sec><title>Subject recruitment and study design</title><p>Recruitment was conducted by the SwedishAmerican Center for Complementary Medicine (SACCM) using targeted advertising, marketing through the SwedishAmerican Health System Centers of Excellence, CHIP alumni groups, corporate client sites, and the SwedishAmerican Health System. Recruitment efforts were aimed at adults (aged at least 18 years) in the greater Rockford, Ill, metropolitan area. To be enrolled in the study, each participant had to be willing to start participating in the program in 1 month or in 7 months. <xref rid="F1" ref-type="fig">Figure 1</xref> shows participant progress through the study. Eligible and interested participants provided informed consent. Participants were highly encouraged to participate with a spouse or significant other and were randomized as a paired unit. All other participants were randomized as individual units. The allocation sequence was created using a random number generator. Program sign-up, randomization, and group assignments were made by the study coordinator. The study was approved by the Institutional Review Board of the SwedishAmerican Health System on August 29, 2002.</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure 1</label><caption><p>Process for a therapeutic lifestyle-modification intervention with a group of community volunteers, Rockford, Ill.</p></caption><alt-text>Logic model</alt-text><long-desc>This flowchart shows how community volunteers were organized during the enrollment, allocation, follow-up, and analysis of a randomized clinical trial. During the enrollment period, 403 volunteers were assessed for eligibility. Of these, 26 were excluded. Baseline data were collected for the remaining 377 volunteers; of these, 29 refused to participate, leaving 348 participants. Of these 348 volunteers, 174 were allocated to an intervention group, and 174 were allocated to a control group. During the follow-up period, the intervention group lost 21 participants for the following reasons: 3 became ill (unrelated to study), 7 were unwilling to commit to the intervention, 7 failed to complete the follow-up, and 4 could not be located. During the same follow-up period, the control group lost 9 participants for the following reasons: 2 became ill (unrelated to the study), 1 lost interest, 2 were unwilling to commit, and 4 failed to complete follow-up. During analysis, the intervention group consisted of 153 nondropouts and 174 intent-to-treat participants, and the control group consisted of 165 nondropouts and 174 intent-to-treat participants.</long-desc><graphic xlink:href="PCD31A05s01" position="float"/></fig></boxed-text></sec><sec><title>Intervention</title><p>The intervention for this study was a live version of the CHIP (<xref rid="B13" ref-type="bibr">13</xref>). Participants met for 4 weeks &#x02014; 4 times each week for 2 hours &#x02014; to receive instruction during April 2003. The curriculum included the following topics: modern medicine and health myths, atherosclerosis, coronary risk factors, obesity, dietary fiber, dietary fat, diabetes, hypertension, cholesterol, exercise, osteoporosis, cancer, lifestyle and health, the Optimal Diet, behavioral change, and self-worth.</p><p>In conjunction with CHIP lectures, participants received a textbook and workbook that closely followed the curriculum topics and included assignments with learning objectives for every topic. Copies of these materials can be obtained from CHIP at www.chipusa.org. Assignments were designed to help participants understand and integrate the information presented. Dietitians and medical professionals spoke to the group weekly, introducing them to the latest nutritional and medical information related to the prevention of chronic diseases. Participants had access to scheduled shopping tours and cooking demonstrations given by a dietitian.</p><p>The diet guidelines approximate the recommendations previously used in the Pritikin program (<xref rid="B5" ref-type="bibr">5</xref>-<xref rid="B7" ref-type="bibr">7</xref>), and the exercise guidelines are from the Surgeon General's Report on Physical Activity and Health (<xref rid="B16" ref-type="bibr">16</xref>). Participants were encouraged to follow preset dietary and exercise goals. The dietary goal was to adopt a more plant-food&#x02013;based diet that emphasizes as-grown, unrefined food. Participants were encouraged to eat the following foods: whole grains, legumes, vegetables, and fresh fruits. In addition, the diet was low in fat (less than 20% of energy), animal protein, sugar, and salt; very low in cholesterol; and high in fiber. Concurrently, program participants were encouraged to work toward walking or exercising for at least 30 minutes each day. Participants were given a pedometer and encouraged to keep an exercise log to record the miles walked each day. In addition, at the completion of the program, participants were encouraged to join the Rockford CHIP alumni association for an annual cost of $25 for an individual or $35 for a couple. The purpose of the alumni organization was to help prevent relapse and help participants maintain their new behaviors. Alumni receive a monthly newsletter that contains news of health-promoting community events such as healthy dinners, walking groups, and support-group meetings. The alumni were encouraged to attend special lectures on healthy living and ways to avoid relapse. &#x000a0;</p><p>The primary objectives of this therapeutic lifestyle-change program were to improve cognitive understanding of the importance of healthy lifestyles, nutrition, and physical activity behavior and reduce risk factors associated with hypertension and cardiovascular disease. The cost to participate in the entire CHIP was $395 per person or $595 per couple.</p></sec><sec><title>Measures</title><p>Variables gathered included cognitive and behavioral measurements and physiologic outcomes related to chronic disease. Demographic data were collected at baseline in April 2003. Attendance at each of the classes was tracked and averaged. Participants attended an average of 89% of the classes.</p><p>The intervention was designed to assist individuals in adopting healthy eating and physical activity behaviors. To assess dietary intake, the Block 98 full-length dietary questionnaire was used (Block 98.2, Block Dietary Data Systems, Berkeley, Calif). The Block 98 questionnaire has been extensively studied and validated (<xref rid="B17" ref-type="bibr">17</xref>). The questionnaire contains self-reported data and is optically scanned and scored. The questionnaire measures the following variables (in addition to others) on a per-day basis: nutrients obtained from food; percentage of calories from fats, carbohydrates, and protein; fiber from different sources; and food group servings per day.</p><p>To ascertain energy expenditure contributed by physical activity, a 7-day self-recorded pedometer log was maintained by each participant. Participants wore the Walk4Life Model 2000 Life Stepper pedometer (Walk4Life Inc, Plainfield, Ill) on a belt at the right hip directly above the right kneecap each day for 7 days. Immediately before going to bed, participants recorded the number of steps for the day and reset the pedometer. Strike counts from pedometers are a valid and reliable method of monitoring and measuring free-living physical activity (<xref rid="B18" ref-type="bibr">18</xref>).</p><p>The primary outcome variables for this study included several chronic disease risk factors. The following data were collected from April to October 2003. Blood was drawn from participants (after a 12-hour fast) by phlebotomists from the SwedishAmerican Health System's outpatient laboratory using a vacutainer (Becton-Dickinson Vacutainer Systems, Rutherford, NJ). Samples were allowed to clot and were centrifuged. Clinical analyses were completed at the SwedishAmerican Health System laboratory. Lipid analysis followed the lipid standards provided by the Centers for Disease Control and Prevention. Glucose, total cholesterol, high-density lipoprotein (HDL), and triglyceride concentrations were determined using Beckman-Coulter LX-20 instrumentation (Beckman Coulter, Inc, Fullerton, Calif). Glucose measurements were obtained with the oxygen-rate method using a Beckman oxygen electrode; cholesterol measurements were obtained with the timed-endpoint enzymatic method using cholesterol oxidase; triglyceride measurements were obtained with the timed-endpoint enzymatic method using glycerol kinase; and HDL measurements were obtained with the homogeneous timed-endpoint method using polyanion detergent to separate HDL and non-HDL lipids. For participants with triglyceride values below 400, LDL values were calculated as follows: LDL = total cholesterol &#x02013; HDL &#x02013; (triglycerides/5) (<xref rid="B19" ref-type="bibr">19</xref>). High-sensitivity C-reactive protein (CRP) measurements were determined using a microplate protocol based on a latex-bead&#x02013;enhanced immunoturbidity assay (<xref rid="B20" ref-type="bibr">20</xref>). Glucose measurements were determined using a Kodak Ektachem (Kodak, Rochester, NY). Trained program staff took blood pressure measurements. Blood pressure was measured in participants after a 5-minute rest, using the guidelines set forth by the American Heart Association. Weight and height were measured using standard medical weight and height scales recently calibrated by the biometrics department of the SwedishAmerican Health System. Percentage of body fat was estimated with Tanita TBF-300A Body Composition Analyzer/Scale using bioelectrical impedance analysis (Tanita, Tokyo, Japan) (<xref rid="B21" ref-type="bibr">21</xref>). Body mass index (BMI) was determined using the following formula: weight (kg)/height (m²).</p></sec><sec><title>Statistical analyses</title><p>Cross-tabulations were used to perform bivariate analyses between selected variables, with statistical significance based on the chi-square test for independence. For testing differences in means, <italic>t</italic> tests were used. Because multiple pair-wise tests were performed, an adjusted &#x003b1; was used to minimize the overall probability of committing a type I error. The modified &#x003b1; is .0001, based on the Bonferroni correction, 28 pair-wise tests, and &#x003b1; = .05. This conservative &#x003b1; was used to determine significance for data in Tables <xref rid="T2" ref-type="table">2</xref> through <xref rid="T7" ref-type="table">7</xref>. Risk factor cut-points (Tables <xref rid="T6" ref-type="table">6</xref> and <xref rid="T7" ref-type="table">7</xref>) were previously established (<xref rid="B22" ref-type="bibr">22</xref>,<xref rid="B23" ref-type="bibr">23</xref>) and categorized accordingly. Results are based on the intent-to-treat method in which all participants were retained in the analyses. Where participant data were lost to follow-up, the last-test carry-forward method was applied to the participant's most recent data. The results did not differ significantly when participants lost to follow-up were dropped from the analyses. These results are not reported. Analyses were performed using SAS 9.0 (SAS Institute Inc, Cary, NC). Procedure statements used in SAS for assessing the data were PROC UNIVARIATE, PROC FREQ, PROC TTEST, and PROC GLM.</p></sec></sec><sec><title>Results</title><p>There were 318 participants who completed both baseline and 6-month evaluations. An additional 30 completed the baseline evaluation but not the 6-month evaluation. Of these lost to follow-up, 21 were in the intervention group, and 9 were in the control group (<xref rid="F1" ref-type="fig">Figure 1</xref>).</p><p>Analyses were based on 348 participants. Ages ranged from 24 to 81 years, with little difference in the mean age between intervention and control groups (50.1 years, intervention group; 50.8 years, control group, <italic>t</italic> ₃₄₆ = &#x02212;0.57, <italic>P</italic> = .57). A description of participants in both intervention and control groups is presented according to selected demographic characteristics in <xref rid="T1" ref-type="table">Table 1</xref>. There were no statistically significant differences between groups for these variables. Within each group, the majority of participants had the following characteristics: female, white, married, an annual family income of at least $60,000, and at least some college education. Of the intervention participants, 47 (27%) joined the CHIP alumni association.</p><p>Because the unit of randomization was <italic>pairs</italic> for those who participated with a partner and <italic>individuals</italic> for those who participated as individuals, comparisons were made of the effect of the program between pairs and individuals. Of the 348 randomized participants, 146 (42%) participated as pairs. There were no significant differences in the outcomes of pairs and individuals. After 6 months, participants in the intervention group experienced significant improvements in all physical activity and nutrition variables except calories from protein and whole-grain servings (<xref rid="T2" ref-type="table">Table 2</xref>). Changes in the control group were generally not statistically significant, or they were much smaller in magnitude than the changes in the intervention group. For each variable except total steps per week and daily sodium intake, the change observed in physical activity or nutrition was significantly greater for participants in the intervention group compared with the control group (<xref rid="T3" ref-type="table">Table 3</xref>). The control group consumed significantly more fat calories and fewer whole-grain servings at 6-month follow-up compared with the control group at 6-month follow-up.</p><p>After 6 months, participants in the intervention group showed significant reductions in BMI, weight, body fat, systolic and diastolic blood pressure, and resting heart rate (<xref rid="T4" ref-type="table">Table 4</xref>). The control group experienced significant improvements in systolic and diastolic blood pressure and HDL, but total cholesterol and LDL were significantly worse. For BMI, weight, and body fat, changes were significantly greater for participants in the intervention group compared with the control group (<xref rid="T5" ref-type="table">Table 5</xref>).</p><p>Mean baseline, 6-month, and change in mean scores are presented according to standard health risk cut-points for the risk factor variables according to intervention group (<xref rid="T6" ref-type="table">Table 6</xref>) and control group (<xref rid="T7" ref-type="table">Table 7</xref>). This analysis stratifies results according to risk status. Individuals with low risk would not be expected to experience large changes, but risk values considered to be high would be expected to change significantly. For the intervention group, the distributions favorably changed between baseline and 6 months for BMI, systolic blood pressure, and diastolic blood pressure. Corresponding significant change in the distribution between baseline and 6 months was observed in the control group for systolic blood pressure and diastolic blood pressure but not for BMI. Favorable changes in risk behaviors were generally higher and more likely to be significant for individuals in the intervention group than for individuals in the control group.</p><p>Whereas total cholesterol significantly increased between baseline and 6 months for participants in the control group, no significant difference was observed in the intervention group. For both intervention and control groups, total cholesterol significantly increased among participants with total cholesterol in the normal range and decreased (but not significantly) for those with cholesterol in the high-risk category. Cholesterol medication played a minimal role in the change observed in cholesterol. At baseline, there were 77 participants in the intervention group who reported using blood pressure medication. At 6 months, 60 participants (75%) indicated no change in their medication over the study period, 9 participants (11.2%) indicated a dosage increase, and 11 participants (13.8%) indicated a dosage decrease. There was not a significant difference in the use of blood pressure medication from baseline to 6 months between the intervention and control groups (&#x003c7;² ₁ = 1.14, <italic>P</italic> = .56).</p></sec><sec><title>Discussion</title><p>Therapeutic lifestyle change can result in significant improvements in nutrition and physical activity behavior and reductions in many cardiovascular disease risk factors. Six months after the intervention began, program participants continued to demonstrate dramatic improvements in nutrition and physical activity behavior. Increases in the number of servings of fruit and vegetables and whole grains, increases in physical activity, and decreases in dietary sodium are likely responsible for the improvements in both systolic and diastolic blood pressure. Intervention group participants consumed 2.3 more servings of fruit and vegetables per day at 6 months compared with baseline. In the PREMIER study (<xref rid="B8" ref-type="bibr">8</xref>), participants who completed a behavior-change program and adopted the DASH diet increased fruit and vegetable servings by 3.0 servings after 6 months. Those PREMIER program participants decreased their percentage of calories from fat by 9.5% and lost an average of 5.8 kg of body weight. This compares to a percentage fat reduction of 8.2% and a 4.5 kg weight loss for intervention participants in the present study.</p><p>At baseline, the intervention group included 77 participants who were at least diastolic prehypertensive at 6 months; this number decreased by 44% to 43 participants at 6 months (<xref rid="T6" ref-type="table">Table 6</xref>). The number of intervention-group participants who were at least systolic prehypertensive at baseline declined by 20%, from 122 participants at baseline to 98 at 6 months. The average reductions in blood pressure were greater than the reductions reported in the DASH study (<xref rid="B9" ref-type="bibr">9</xref>) and comparable with the results of the PREMIER clinical trial (<xref rid="B8" ref-type="bibr">8</xref>).</p><p>Previous reports of the CHIP intervention showed sharp improvements in blood lipid levels at 6 weeks, but most of these changes disappeared at 6 months (<xref rid="B7" ref-type="bibr">7</xref>). Other therapeutic lifestyle trials that lasted longer than 3 months and included lipid outcomes reported similar findings (<xref rid="B10" ref-type="bibr">10</xref>-<xref rid="B12" ref-type="bibr">12</xref>,<xref rid="B24" ref-type="bibr">24</xref>). In this study, dietary cholesterol among the intervention group was reduced by 122 mg/day (a 56% reduction), and dietary saturated fat was cut by half. Despite these favorable changes in dietary cholesterol precursors, a return to previous lipid levels suggests that there is a significant increase in endogenous cholesterol, most of which appears to be LDL cholesterol (<xref rid="B25" ref-type="bibr">25</xref>). It is also possible that these changes in blood lipid levels were affected by seasonal variation. Without more accurate measures of endogenous cholesterol biosynthesis, it is impossible to determine the exact cause of the cholesterol increase (<xref rid="B26" ref-type="bibr">26</xref>).</p><p>Pedometer data show that program participants increased physical activity by 30%. The average number of steps for the intervention group after 6 months did not meet the recommended 10,000 steps per day (<xref rid="B27" ref-type="bibr">27</xref>). For this predominately middle-aged and obese population, however, an increase in physical activity of 30% likely contributed to risk factor reductions. When combined with diet changes, improvement in physical activity is the likely explanation for the percentage decreases in BMI (&#x02212;5%), weight (&#x02212;5%), and percentage body fat (&#x02212;6%) among the intervention group. Improved physical activity was also associated with a significant decrease in resting heart rate, a correlated measure of cardiorespiratory fitness thought to be caused by increased heart size, blood volume, stroke volume, and cardiac output (<xref rid="B28" ref-type="bibr">28</xref>).&#x000a0;</p><p>Poor nutrition and sedentary living are associated with a constellation of risk factors, some identified in the metabolic syndrome, and all linked to common chronic diseases (<xref rid="B29" ref-type="bibr">29</xref>).&#x000a0;Improvements in nutrition and physical activity are associated with significant reductions in diabetes risk as whole body glucose tolerance improves, insulin sensitivity increases, and the amount of glucose transporter (GLUT4) increases (<xref rid="B30" ref-type="bibr">30</xref>). The number of individuals with diabetes (glucose &#x02265;126 mg/dL) in the intervention group was reduced by 19%, demonstrating that this therapeutic lifestyle-change program improves insulin sensitivity. Similar results were reported by other lifestyle trials reporting glucose findings (<xref rid="B11" ref-type="bibr">11</xref>,<xref rid="B12" ref-type="bibr">12</xref>).&#x000a0;</p><p>These improvements in behavior and risk are not unexpected because the intervention lectures were structured on the health belief and transtheoretical models. Video clips, testimonials, role playing, short presentations from physicians, social support strategies, food selection and planning activities, and other behavior-change&#x02013;driven pedagogical activities helped to encourage participants to enthusiastically evaluate personal behaviors and commit to make lifestyle changes.</p><p>Most of the participants were white and sufficiently self-motivated to volunteer to participate in the intervention. On average, participants were slightly more educated than the community average. Participants had lifestyles that permitted them to attend most, if not all, of the classes. This is evident in the high rate of attendance to this time-intensive program. These delimitations threaten the generalizability of these findings and make application of the intervention to other populations problematic. Because the participants were self-selected, the results from this intervention may represent a best-case scenario.</p><p>Despite the apparent effect of this intervention, there are some shortcomings associated with the study design. Both the physical activity and nutrition data were self-reported. For some variables, the control group also experienced significant improvement. Significant decreases were observed in the control group in percentage of fat calories and dietary-fat grams, sodium grams, and total calories as well as small increases in total steps. In addition, the control group experienced similar improvement in blood pressure compared with the intervention group. There are more than 27 restaurants in the Rockford metropolitan area that offer healthy, CHIP-recommended menu items, which could have contributed to improvements in the control group. When conducting lifestyle trials, the question of what to do with the control group is difficult to answer because there is no such thing as a lifestyle placebo. After participants were assigned to an intervention or control group, some control-group participants expressed happiness with their assignment because they had personal or work-related conflicts that would have prohibited them from participating in the intervention group. Others were disappointed in their control-group assignment but realized when they agreed to participate in the research study that there was always the chance that they would have to wait to participate in the program.</p><p>This study indicates that an intervention that uses various behavior modification tools, such as live lectures, workbooks, and professional advice, and is implemented among a group of middle-aged volunteers can result in reduced risk factors for cardiovascular disease after 6 months. Further research is needed to examine the effects of the program on other populations.</p></sec>

Process, Rationale, and Interventions of Pakistan&#x02019;s National Action Plan on Chronic Diseases

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Nishtar</surname><given-names>Sania</given-names></name><degrees>FRCP, PhD</degrees><role>President</role><aff>Heartfile</aff><address><email>sania@heartfile.org</email><addr-line>1 Park Rd; Chak Shahzad, Islamabad, Pakistan</addr-line><phone>+0092-51-224-3580</phone></address></contrib><contrib contrib-type="author"><name><surname>Mohamud Bile</surname><given-names>Khalif</given-names></name><degrees>PhD</degrees><aff>World Health Organization (WHO), Islamabad, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Ahmed</surname><given-names>Ashfaq</given-names></name><degrees>MBBS</degrees><aff>Office of International Health, Ministry of Health, Islamabad, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Faruqui</surname><given-names>Azhar M.A.</given-names></name><degrees>FRCP</degrees><aff>National Institute of Cardiovascular Diseases, Karachi, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Mirza</surname><given-names>Zafar</given-names></name><degrees>MPH</degrees><aff>Network for Consumer Protection, Islamabad, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Shera</surname><given-names>Samad</given-names></name><degrees>FRCP</degrees><aff>Diabetes Association of Karachi, Karachi, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Ghaffar</surname><given-names>Abdul</given-names></name><degrees>PhD</degrees><aff>Global Forum for Health Research, Geneva, Switzerland</aff></contrib><contrib contrib-type="author"><name><surname>Minhas</surname><given-names>Fareed A</given-names></name><degrees>FRCP</degrees><aff>WHO Collaborating Centre for Psychiatry, Rawalpindi, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Khan</surname><given-names>Aslam</given-names></name><degrees>FRCP, MRCP</degrees><aff>Military Hospital, Rawalpindi, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Jaffery</surname><given-names>Naeem A</given-names></name><degrees>FRCP</degrees><aff>Ziauddin Medical University, Karachi, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Rajput</surname><given-names>Majid</given-names></name><degrees>MPH, FCPS</degrees><aff>Office of the Director General Health, Ministry of Health, Islamabad, Pakistann</aff></contrib><contrib contrib-type="author"><name><surname>Mirza</surname><given-names>Yasir A</given-names></name><degrees>MSc</degrees><aff>Communications Department, Heartfile, Islamabad, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Aslam</surname><given-names>Mohammad</given-names></name><degrees>MSc, FCPS</degrees><aff>Communications Department, Heartfile, Islamabad, Pakistan</aff></contrib><contrib contrib-type="author"><name><surname>Rahim</surname><given-names>Ejaz</given-names></name><degrees>MA</degrees><aff>Office of the Secretary of Health, Ministry of Health, Islamabad, Pakistan</aff></contrib>

Preventing Chronic Disease

<sec><title>Background</title><p>Chronic noncommunicable diseases (NCDs) are estimated to have caused 33.4 million deaths worldwide in 2002; of these, 72% occurred in developing countries (<xref rid="B1" ref-type="bibr">1</xref>). In Pakistan, chronic diseases (cardiovascular disease, diabetes, chronic lung diseases, and cancer) are among the top 10 causes of morbidity and mortality and account for approximately 25% of total deaths (<xref rid="B2" ref-type="bibr">2</xref>,<xref rid="B3" ref-type="bibr">3</xref>). Thirty-three percent of the adult population older than 45 years has high blood pressure; 10% of the adult population older than 18 years has diabetes, and more than 54% of men use tobacco. Data from an unselected autopsy series have shown coronary artery involvement (of greater than 50% luminal diameter reduction) in more than 24% of those studied. Moreover, the coastal metropolis of Karachi, with a population of more than 15 million, reports one of the highest incidences of breast cancer for any Asian population (<xref rid="B4" ref-type="bibr">4</xref>-<xref rid="B9" ref-type="bibr">9</xref>). Over the years, Pakistan's federal and provincial Ministries of Health have been heavily burdened with reproductive-health and infectious-disease issues. However, Pakistan is undergoing an epidemiological transition and, hence, is now also focusing on chronic diseases.</p><p>Pakistan has a population of 150 million and an annual gross national product (GNP) per capita of U.S. $700; during the last 10 years, 0.6% to 0.8% of its GNP and 5.1% to 11.6% of its development budget has been spent on the health sector (<xref rid="B2" ref-type="bibr">2</xref>). Seventy percent of clinical services are delivered by private-sector health care providers, and out-of-pocket payments are the major source of health financing, despite the existence of an extensive public-sector&#x02013;owned health care system. Preventive and educational services are delivered almost exclusively by the public sector. Lately, as part of certain preventive programs (HIV and AIDS programs in particular), nongovernmental organizations (NGOs) have been delivering preventive care, albeit in a contractual role in which NGOs enter into contracts with the public sector.</p><p>As in most other developing countries, NCDs had not featured prominently on Pakistan's health agenda until 2003, when a national integrated plan for health promotion and the prevention and control of NCDs, known as the National Action Plan on NCD Prevention, Control, and Health Promotion (NAP-NCD), was initiated. NAP-NCD attempts to obviate the challenges associated with addressing chronic diseases in countries with limited resources. Initially, an agreement was developed between the Ministry of Health and Heartfile, an Islamabad, Pakistan-based nonprofit, NGO focused on chronic disease prevention and control and health promotion; a month later, the World Health Organization (WHO) was asked to join the initiative. Heartfile's role in the NAP-NCD was to advocate for increased focus on chronic disease in the national health agenda. This was the first time an NGO had participated in a national health program in more than a contractual role.</p><p>The partnership, developed on a national level, was mandated with the task of developing and implementing a strategic plan to prevent and control the rates of NCDs (<xref rid="B10" ref-type="bibr">10</xref>). The NAP-NCP was released on May 12, 2004, within a year of the agreement's signing, and as of January 2006 is in its first phase of implementation (<xref rid="B11" ref-type="bibr">11</xref>,<xref rid="B12" ref-type="bibr">12</xref>).</p></sec><sec><title>Developing the Plan</title><p>Chronic diseases generally are linked by common risk factors and include cardiovascular disease (CVD), diabetes, cancer, and chronic lung disease. However, the NAP-NCD also includes injuries and mental illness in its framework because of government requirements.</p><p>A three-stage process was used to develop the NAP-NCD: 1) planning within the disease categories, 2) setting priorities, and 3) developing an integrated approach to preventing NCDs (<xref rid="B13" ref-type="bibr">13</xref>).</p><p>Next, a situational analysis was conducted in which data on current epidemiological evidence for NCDs were gathered, existing strategies and policy measures were summarized, gaps in the system and opportunities that existed for integration with existing programs were outlined, and the potential for program implementation was analyzed. Then, a broad-based consultative process was established, which included health professionals, NGOs, professional societies, community representatives, donor and development agencies, corporations, and legislators, and priority action areas were identified. In the absence of local cost-effectiveness data, other priority-setting criteria were used, such as the extent to which an intervention was locally feasible, promoted community empowerment and participation, built on the strengths of partnerships, and contributed to capacity building and health systems strengthening. In addition, the capacity of the public health system and the ability of health care leaders to implement the NAP-NCD were also identified as important criteria.</p><p>Finally, a tool called the Integrated Framework for Action (IFA) was developed to identify action items that could be applied to all NCDs&#x000a0; (<xref rid="B14" ref-type="bibr">14</xref>). (The IFA is available from <ext-link xlink:href="http://heartfile.org/pdf/IFAPDF.htm" ext-link-type="uri">http://heartfile.org/pdf/IFAPDF.htm</ext-link>.) Additionally, the IFA included two sets of strategies &#x02014; those that were common to all NCDs (<xref rid="T1" ref-type="table">Table 1</xref>) and those specific to each NCD (<xref rid="T2" ref-type="table">Table 2</xref>). The first set of strategies includes behavioral-change communication, focusing health services on NCDs, development of institutional mechanisms, and monitoring and surveillance; the second set covers legislative or regulatory matters and research. The IFA &#x02014; which also provides guidance to administrators and health policy planners &#x02014; helps set national goals at process, output, and outcome levels; defines integrated actions to meet those goals; and allows for program assessment.</p></sec><sec><title>Components and Configuration of the NAP-NCD</title><p>The NAP-NCD prioritizes a population-based approach to chronic diseases that encompasses public education, behavioral-change communication, legislation, and regulation. These approaches have the greatest potential to reduce NCD risk and uphold the principles of WHO's "Health-for-All Policy for the 21st Century" because the high-risk approach (i.e., targeting individuals at high risk for chronic disease rather than populations as a whole) may be inaccessible to the majority of the country's underprivileged population. Thirty-two percent of Pakistan's population is below the poverty level of U.S. $1 a day (<xref rid="B15" ref-type="bibr">15</xref>). The NAP-NCD will be implemented in two phases.</p><sec><title>First phase</title><p>The first phase of implementation, which spans 3 years (May 2004 through July 2006), is jointly funded by the Ministry of Health, Heartfile, and WHO. The implementation status is reviewed for accountability and program evaluation every 3 months, and progress is posted online (<xref rid="B24" ref-type="bibr">24</xref>); the process and output indicators stipulated in the IFA are used for process evaluation.&#x000a0;The first phase of the NAP-NCD's implementation focuses on the action items summarized in Tables 1 and 2 and is organized into the following three priority areas: an integrated and sustainable population-based NCD surveillance system, an integrated behavioral-change communication strategy, and legislation in key areas.</p><p>An integrated and sustainable population-based NCD surveillance system is a prerequisite for effective planning, implementation, and evaluation of NCD prevention programs (with the exception of cancer, because a registry has to be used for its surveillance) and is regarded as an entry point for activities related to the prevention of NCDs &#x02014; an approach validated in several settings (<xref rid="B16" ref-type="bibr">16</xref>,<xref rid="B17" ref-type="bibr">17</xref>). The NAP-NCD's surveillance model includes population surveillance of primary NCD risk factors (poor diet, physical inactivity, and smoking) and combines modules on population surveillance of injuries, mental health, and stroke. In addition, the model has been adapted for program evaluation, which enables it to use indicators to track implementation processes and facilitates an assessment of how interventions work and which components are the most successful. An initial cross-sectional survey with a sample of sufficient size and the power to detect population-level changes over time of the risk factors and NCDs has been conducted (<xref rid="B18" ref-type="bibr">18</xref>,<xref rid="B19" ref-type="bibr">19</xref>).</p><p>The integrated behavioral-change communication strategy consisted of two interventions. The first intervention included a media campaign targeting 90% of the country's population. The second intervention introduced chronic disease prevention into the work plan of Lady Health Workers (LHWs) &#x02014; Pakistan's field force of more than 83,000 grassroots health caregivers.</p><p>For the media campaign, 30-second spots and 5-minute programs are being aired for 2 years during prime time on national television and radio and began in May and June 2005. One announcement focuses on creating awareness about high blood pressure by advocating opportunistic screening (i.e., using every clinical encounter "opportunity" to check the blood pressure of every patient); the other emphasizes the principles of cardiovascular disease prevention.</p><p>Until recently, LHWs were involved in delivering reproductive-health&#x02013; and communicable-disease&#x02013;related services to poor and underprivileged households in rural areas covering 50% of Pakistan's population. Heartfile had previously pilot tested an approach in which CVD prevention was introduced into its work plan in the Lodhran district as part of a CVD prevention demonstration project. Seven hundred LHWs were involved in this pilot project from 2001 through 2003 (<xref rid="B20" ref-type="bibr">20</xref>,<xref rid="B21" ref-type="bibr">21</xref>). Lessons learned from this experience have enabled the introduction of the chronic disease perspective into the training module in addition to an increased focus on chronic disease in 17 other districts as part of the NAP-NCD.</p><p>Other priority areas for the first phase of implementation include lobbying for key legislative actions, identifying research areas, building capacity within the health system, and focusing on institutional measures.</p></sec><sec><title>Second phase</title><p>The second phase of implementation will broaden the scope to include measures that focus health services on prevention; the launch of the second phase is planned for 2006. The second phase will have implications for training and capacity building of health professionals, improving basic infrastructure, and ensuring availability and access to certain drugs at all levels of health care.</p><p>Health care delivery in Pakistan is characterized by a variety of roles played by different categories of health care providers, and all will be drawn into the loop. The NAP-NCD makes recommendations to ensure physician training as a permanent function of the health care system by establishing links with provincial and district health departments; it also makes recommendations on how to form a comprehensive continuing medical education program structured around broad-based prevention-related goals and objectives to ensure ongoing training for both private-sector and pubic-sector physicians. The second phase of the NAP-NCD will also include other legislative actions.</p></sec></sec><sec><title>Merits and Limitations of the Approach</title><p>The NAP-NCD presents one approach to developing a national strategy for chronic diseases in countries with limited resources. The strategy includes integration at six distinct levels. By <italic>grouping chronic diseases</italic> and <italic>integrating actions</italic>, there is a shift from a national-level approach to an approach based on diseases, which has significant implications for maximizing health care resources. Horizontally integrating actions with existing initiatives strengthens the public health system; adopting integrated models on surveillance and behavioral-change communication in addition to focusing health services on NCDs will yield important empirical evidence for emerging chronic disease programs. Additionally, integrating health promotion and prevention within the same program achieves two objectives for two populations with common activities.&#x000a0;</p><p>The evaluation mechanism of this model, which is structured within the accountability and progress charts of the IFA, allows program assessment at a process-and-outcomes level and assessment of the level of contribution partners have made to achieve the NAP-NCD's objectives.</p><p>This program is of value to all partners. By leveraging the strengths of the nonprofit private-sector technical partner (Heartfile), the government included NCD prevention in its policies. NGOs and the civil society can contribute to achieving national goals; however, this potential remains largely untapped. This model provides a mechanism for engaging NGOs in the national decision-making process and ensures their participation both in the formulation of health policy and implementation of national plans. Though an evaluation mechanism, it also enables the assessment of each partner's contribution to achieving objectives. In this model, WHO is gaining experience with a model in which WHO resources &#x02014; which are otherwise allocated for the public sector &#x02014; support the private sector in a national model. The partnership is therefore integrated with national health priorities and complements state initiatives.</p><p>This program is one of the few examples of a public&#x02013;private partnership for chronic disease prevention, an area that has largely remained unexplored as part of global efforts to build public&#x02013;private partnerships. This program's implementation is expected to yield important information about the performance of the health system by building chronic disease partnerships in evidence-based models. As for infectious disease partnerships, ethical, methodological, accountability, sustainability, and governance issues must be considered (<xref rid="B22" ref-type="bibr">22</xref>-<xref rid="B25" ref-type="bibr">25</xref>).</p><p>This initiative also created a mechanism for visible involvement and participation of many other stakeholders in the national consultation process in addition to avenues for their participation in the process of implementation. This is important because many factors that affect NCDs are outside of the health sector domain; these include trade, agriculture, finance, education, and communication. However, there is also the need to fit this strategy within a more explicit policy framework &#x02014; one that makes it obligatory to link relevant health ministries in a manner that is mutually supportive of national NCD goals. The program needs to be supported by a clear, strong, and sustained political commitment.</p><p>The NAP-NCD can serve as both an empirical basis for an integrated approach to NCDs and an experimental basis of health sector reform in the area of public&#x02013;private collaboration; most developing countries have limited experience with each. It is also likely to yield useful lessons for ministries of health, NGOs, and multilateral agencies for establishing chronic disease programs in developing countries.</p></sec>

The Nutrition and Physical Activity Program to Prevent Obesity and Other Chronic Diseases: Monitoring Progress in Funded States

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Yee</surname><given-names>Sue Lin</given-names></name><degrees>MA, MPH</degrees><aff>Centers for Disease Control and Prevention (CDC)</aff><address><email>sby9@cdc.gov</email><addr-line>4770 Buford Hwy, NE, Mail Stop K-24, Atlanta, GA 30341</addr-line><phone>770-488-5361</phone></address></contrib><contrib contrib-type="author" corresp="no"><name><surname>Williams-Piehota</surname><given-names>Pam</given-names></name><degrees>PhD</degrees><aff>Research Triangle Institute (RTI) International, Research Triangle Park, NC</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Sorensen</surname><given-names>Asta</given-names></name><degrees>MA</degrees><role>Research Triangle Institute (RTI) International, Research Triangle Park, NC</role><aff>Health Department</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Roussel</surname><given-names>Amy</given-names></name><degrees>PhD</degrees><aff>Research Triangle Institute (RTI) International, Research Triangle Park, NC</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Hersey</surname><given-names>James</given-names></name><degrees>PhD</degrees><aff>RTI International, Washington, DC</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Hamre</surname><given-names>Robin</given-names></name><degrees>MPH, RD</degrees><aff>CDC, Atlanta, Ga</aff></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>In the past decade, the United States has experienced a dramatic increase in the prevalence of obesity and overweight. According to self-reported weights and heights, all states had obesity rates of less than 20% for adults in 1991 (<xref ref-type="bibr" rid="B1">1</xref>). In 2003, the Behavioral Risk Factor Surveillance System revealed that 31 states had adult obesity rates of 20% to 24%, and four states had obesity rates of 25% or greater (<xref ref-type="bibr" rid="B1">1</xref>). Rates of overweight among children have also increased at an alarmingly rapid pace. Results from the 1999&#x02013;2002 National Health and Nutrition Examination Survey (NHANES)&#x000a0;showed that more than 10% of children aged between 2 and 5 years were overweight from 1999 through 2002 (<xref ref-type="bibr" rid="B2">2</xref>). In addition, approximately 16% of children and adolescents aged 6 to 19 years were overweight (<xref ref-type="bibr" rid="B3">3</xref>), which is a 5 percentage point increase in prevalence from 1988 through 1994, when 11% of children and adolescents in this age group were overweight (<xref ref-type="bibr" rid="B4">4</xref>).</p><p>According to a study of national costs attributed to overweight (body mass index [BMI] = 25&#x02013;29.9) and obesity (BMI &#x02265;30), the related medical expenses accounted for 9.1% of the total U.S. medical expenditures in 1998 and may have been as high as $78.5 billion (<xref ref-type="bibr" rid="B5">5</xref>). The increasing prevalence in obesity among the U.S. population places a financial strain on individual states. For instance, a 2004 study (<xref ref-type="bibr" rid="B6">6</xref>) found that total state expenditures on obesity-related medical expenditures were approximately $75 billion, excluding costs related to absenteeism and loss of productivity (<xref rid="F1" ref-type="fig">Figure 1</xref>). The Centers for Disease Control and Prevention's (CDC's) Nutrition and Physical Activity Program to Prevent Obesity and Other Chronic Diseases currently funds 3 of the 4 states (75%) that have the highest total obesity costs in the United States and 8 of 11 states (73%) with total obesity costs greater than $2 million. The state-level estimates can help state policy makers determine how best to allocate public health resources to address obesity prevention and control in partnership with public and private stakeholders throughout their states.</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure 1</label><caption><p>Funded states and state-level estimates of annual medical expenditures attributable to obesity (2003 dollars, in millions). The states funded through 2004 are the 20 states addressed in this article. Source: Adapted from Finkelstein EA et al (<xref ref-type="bibr" rid="B5">5</xref>)</p></caption><alt-text>Map of the US</alt-text><alternatives><table frame="hsides" rules="groups"><thead><tr><td align="left" valign="top" rowspan="1" colspan="1"> <bold>State</bold> </td><td align="left" valign="top" rowspan="1" colspan="1"> <bold>$</bold> </td><td align="left" valign="top" rowspan="1" colspan="1"> <bold>State</bold> </td><td align="left" valign="top" rowspan="1" colspan="1"> <bold>$</bold> </td></tr></thead><tbody><tr><td align="left" valign="top" rowspan="1" colspan="1">Alabama</td><td align="left" valign="top" rowspan="1" colspan="1">1320</td><td align="left" valign="top" rowspan="1" colspan="1">Nebraska</td><td align="left" valign="top" rowspan="1" colspan="1">454</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Alaska</td><td align="left" valign="top" rowspan="1" colspan="1">195</td><td align="left" valign="top" rowspan="1" colspan="1">Nevada</td><td align="left" valign="top" rowspan="1" colspan="1">337</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Arizona (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">752</td><td align="left" valign="top" rowspan="1" colspan="1">New Hampshire</td><td align="left" valign="top" rowspan="1" colspan="1">302</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Arkansas (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">663</td><td align="left" valign="top" rowspan="1" colspan="1">New Jersey</td><td align="left" valign="top" rowspan="1" colspan="1">2342</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">California</td><td align="left" valign="top" rowspan="1" colspan="1">7675</td><td align="left" valign="top" rowspan="1" colspan="1">New Mexico (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">324</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Colorado (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">874</td><td align="left" valign="top" rowspan="1" colspan="1">New York (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">6080</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Connecticut</td><td align="left" valign="top" rowspan="1" colspan="1">856</td><td align="left" valign="top" rowspan="1" colspan="1">North Carolina (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">2138</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Delaware</td><td align="left" valign="top" rowspan="1" colspan="1">207</td><td align="left" valign="top" rowspan="1" colspan="1">North Dakota</td><td align="left" valign="top" rowspan="1" colspan="1">209</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Florida (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">3987</td><td align="left" valign="top" rowspan="1" colspan="1">Ohio</td><td align="left" valign="top" rowspan="1" colspan="1">3304</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Georgia (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">2133</td><td align="left" valign="top" rowspan="1" colspan="1">Oklahoma (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">854</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Hawaii</td><td align="left" valign="top" rowspan="1" colspan="1">290</td><td align="left" valign="top" rowspan="1" colspan="1">Oregon (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">781</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Idaho</td><td align="left" valign="top" rowspan="1" colspan="1">227</td><td align="left" valign="top" rowspan="1" colspan="1">Pennsylvania (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">4138</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Illinois (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">3439</td><td align="left" valign="top" rowspan="1" colspan="1">Rhode Island (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">305</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Indiana</td><td align="left" valign="top" rowspan="1" colspan="1">1637</td><td align="left" valign="top" rowspan="1" colspan="1">South Carolina (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">1060</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Iowa (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">783 </td><td align="left" valign="top" rowspan="1" colspan="1">South Dakota (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">195</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Kansas</td><td align="left" valign="top" rowspan="1" colspan="1">657</td><td align="left" valign="top" rowspan="1" colspan="1">Tennessee</td><td align="left" valign="top" rowspan="1" colspan="1">1840</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Kentucky (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">1163</td><td align="left" valign="top" rowspan="1" colspan="1">Texas (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">5340</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Louisiana</td><td align="left" valign="top" rowspan="1" colspan="1">1373</td><td align="left" valign="top" rowspan="1" colspan="1">Utah</td><td align="left" valign="top" rowspan="1" colspan="1">393</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Maine (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">357</td><td align="left" valign="top" rowspan="1" colspan="1">Vermont (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">141</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Maryland (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">1533</td><td align="left" valign="top" rowspan="1" colspan="1">Virginia</td><td align="left" valign="top" rowspan="1" colspan="1">1641</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Massachusetts (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">1822</td><td align="left" valign="top" rowspan="1" colspan="1">Washington (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">1130</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Michigan (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">2931</td><td align="left" valign="top" rowspan="1" colspan="1">Washington, DC</td><td align="left" valign="top" rowspan="1" colspan="1">372</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Minnesota</td><td align="left" valign="top" rowspan="1" colspan="1">1307</td><td align="left" valign="top" rowspan="1" colspan="1">West Virginia (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">588</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Mississippi</td><td align="left" valign="top" rowspan="1" colspan="1">757</td><td align="left" valign="top" rowspan="1" colspan="1">Wisconsin (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">1487</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Missouri (2004)</td><td align="left" valign="top" rowspan="1" colspan="1">1636</td><td align="left" valign="top" rowspan="1" colspan="1">Wyoming</td><td align="left" valign="top" rowspan="1" colspan="1">87</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Montana (2005)</td><td align="left" valign="top" rowspan="1" colspan="1">175 </td><td align="left" valign="top" rowspan="1" colspan="1">&#x000a0;</td><td align="left" valign="top" rowspan="1" colspan="1">&#x000a0;</td></tr></tbody></table><graphic xlink:href="PCD31A23s01"/></alternatives></fig></boxed-text><p>Established in 1999, the CDC Nutrition and Physical Activity Program to Prevent Obesity and Other Chronic Diseases was designed to help states prevent obesity and other chronic diseases by addressing two closely related factors &#x02014; poor nutrition and inadequate physical activity. The program incorporates five evidence-based strategies, including balancing caloric intake and expenditure, increasing physical activity, increasing the consumption of fruits and vegetables, decreasing television-viewing time, and increasing breastfeeding.</p><p>States receive funding at two different levels: capacity building and basic implementation. Capacity-building states are expected to gather data, build partnerships, and create statewide health plans, which are critical steps that must be completed before implementing nutrition and physical activity interventions. To receive basic implementation funding, capacity-building states must implement a nutrition and physical activity intervention that addresses at least two levels of the social&#x02013;ecological model. The social&#x02013;ecological model is based on the premise that changes in individual behavior will come about through a combination of societal, community, organizational, interpersonal, and individual efforts (<xref ref-type="bibr" rid="B7">7</xref>,<xref ref-type="bibr" rid="B8">8</xref>). Basic-implementation states have begun to develop new and sustainable interventions, evaluate existing interventions, support additional state and local efforts to prevent obesity and other chronic diseases, or all of these.</p><p>In 2004, 20 states received funding by the Nutrition and Physical Activity Program to Prevent Obesity and Other Chronic Diseases: 17 states each received $300,000 to $450,000 for capacity building. Three states each received $800,000 to $1.5 million for basic implementation. In this article, we present an overview of the progress of the 20 states through June 2004. (Currently, there are 28 funded states: 23 capacity-building states and 5 basic-implementation states.)</p><p>States submit semiannual progress reports to the CDC and address their program infrastructure, collaborations, implementation, and evaluation. The Division of Nutrition and Physical Activity uses the reports for program management and program improvement. This article includes information provided by the states in their December 2004 semiannual reports, which included activities from January 1 through June 30, 2004. Because states have received funds for varying lengths of time, their progress varies, with some in the planning stages for statewide obesity prevention and control programs and some implementing the interventions. Ongoing activities could include activities initiated before January 1, 2004.</p></sec><sec><title>Findings</title><p>The 20 funded states have made progress in three key areas: 1) capacity building, 2) environmental change, and 3) interventions. <italic>Capacity building</italic> includes forging partnerships and leveraging financial resources. <italic>Environmental changes</italic> are environmental modifications that create a health-promoting environment, such as public policies, legislative acts, an increase in access to healthy foods, urban planning, and other efforts. <italic>Interventions</italic> are activities developed by funded states that focus on the program's five evidence-based strategies.</p><sec><title>Capacity building</title><p>In the early stages of a program, developing capacity and infrastructure is a fundamental activity. Hiring staff members, gathering data, building partnerships, and creating statewide health plans enable states to marshal broad-based support for their programs.</p><sec><title>Developing partnerships</title><p>A key aspect of capacity building involves establishing collaborative relationships with partners from state and local governments and the private sector. States have formed numerous partnerships with governmental organizations, health care organizations (e.g., state departments of transportation, local health departments, the American Academy of Family Physicians), volunteer agencies (e.g., the YMCA, the American Heart Association), universities, organizations that address health disparities (e.g., the Indian Health Service), private companies (e.g., Nike), and other types of organizations (e.g., the National Guard). Each state reported that it had many partners, ranging from 17 to 36 partners per state (median = 26). The instructions for the progress monitoring reports specified that states list only the three most important partners for each of several types of partner organizations, so each state may have had more partners than indicated in its reports.</p><p>As part of the progress monitoring reports, the states were asked to indicate how each partner had contributed to the state plan or program during the previous 6 months. The answers indicated that every state had partners that participated in planning activities. In addition, most states (18 states, or 90%) had partners that contributed staff time, cosponsored obesity-prevention events (16 states, or 80%), and implemented interventions (14 states, or 70%). (The interventions that were implemented with partners did not necessarily meet the program's operational definition of an intervention. Furthermore, some of the interventions implemented by capacity-building states may not have qualified the state for basic-implementation funding status. For example, the interventions may have been activities in which the partner took the lead role but collaborated with the state program.) Ten (50%) of the states had partners that contributed funds.</p></sec><sec><title>Leveraging financial resources</title><p>States have been able to leverage additional federal and state program funds to increase the financial resources that support their activities. The majority of states (15, or 75%) have obtained, appropriated, or reallocated funds from outside their cooperative agreements for nutrition and physical activity programs. The states have primarily leveraged funds from state programs (10 states, or 50%) or federal programs other than the CDC (7 states, or 35%). Although less common, some states have leveraged funds from foundation grants (2 states, or 10%) and private businesses (2 states, or 10%).</p><p>The amount of funding that states leveraged varied considerably, ranging from no funding to more than $1 million. Five states (25%) had no leveraged funding, four states (20%) had less than $100,000 in leveraged funding, seven states (35%) had $100,000 to $499,000 in leveraged funding, and four states (20%) had $1 million or more in leveraged funding. Funding was acquired for planning and programs (13 states, or 65%), building infrastructure (9 states, or 45%), and evaluation and surveillance activities (8 states, or 40%).</p></sec></sec><sec><title>Environmental changes</title><p>One hallmark of the program has been the states' effectiveness in stimulating changes to physical and social environments to make them more conducive to health promotion. In the progress monitoring report, states were asked to describe the policies, legislative acts, or environmental changes that they initiated, modified, or planned as a result of the state planning process during the previous 6 months. They were instructed not to report the same environmental change in more than one of the categories. Twelve states described a policy, a legislative change, or an environmental change.</p><sec><title>Policies</title><p>Policies for promoting public health change involve organizational statements or general rules designed to facilitate healthy lifestyle choices. In other words, health promotion policies are an attempt to produce healthy behaviors that are likely to be sustained. Most states are in the initial stages of developing and implementing policies that support environmental changes related to nutrition and physical activity. As shown in <xref rid="F2" ref-type="fig">Figure 2</xref>, six states (30%) reported initiating policies related to nutrition and physical activity in the previous 6 months. Policies that promote nutrition and physical activity in schools were the most commonly reported and planned policy changes.</p><boxed-text position="float"><fig position="float" id="F2" fig-type="diagram"><label>Figure 2</label><caption><p>Percentage of states reporting environmental changes through policies, legislation, and other methods. Data are based on December 2004 progress reports from the 20 state programs and reflect environmental changes that were initiated and planned between January and June 2004. <italic>Other environmental changes</italic> are strategies other than policies and legislation, such as urban planning, that alter or control the legal, social, economic, and physical environment affecting nutrition and physical activity.</p></caption><alt-text>Bar graph</alt-text><alternatives><table frame="hsides" rules="groups"><tbody><tr><td align="left" valign="top" rowspan="1" colspan="1">Policies</td><td align="left" valign="top" rowspan="1" colspan="1">Initiated: 30% (6)</td><td align="left" valign="top" rowspan="1" colspan="1">Planned: 30% (6)</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Legislation</td><td align="left" valign="top" rowspan="1" colspan="1">Initiated: 45% (9)</td><td align="left" valign="top" rowspan="1" colspan="1">Planned: 55% (11)</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Other <break/>Environmental<break/>Changes</td><td align="left" valign="top" rowspan="1" colspan="1">Initiated: 15% (3)</td><td align="left" valign="top" rowspan="1" colspan="1">Planned: 10% (2)</td></tr></tbody></table><graphic xlink:href="PCD31A23s02"/></alternatives></fig></boxed-text></sec><sec><title>Legislative acts</title><p>Legislative acts are strategies that involve creating laws supporting the health-promoting behavior of individuals, organizations, or both. Almost half of the states (45%) reported initiating, modifying, or enforcing legislative acts related to nutrition and physical activity in the previous 6 months. Several legislative acts focused on improving nutrition and increasing physical activity in schools. For example, seven states (35%) reported legislation that would set standards for foods available and sold in schools, eliminate soft drinks from school vending machines, or require school districts to incorporate daily physical activity into their curricula. Other legislation focused on research, establishing new programs within state departments of health to study obesity-related issues, and providing insurance coverage for health services to improve nutrition and prevent obesity.</p></sec><sec><title>Other environmental changes</title><p>Other environmental changes are interventions that alter or control the legal, social, economic, and physical environment related to nutrition and physical activity. Examples include Rails-to-Trails programs, closing a dangerous street near a school, and zoning and planning for parks and other recreation areas. Three states (15%) reported initiating other environmental changes in the previous 6 months. Environmental changes primarily focused on improving access to physical activity opportunities and healthy foods through new walking trails, community gardens, changes to the school cafeteria menu, and changes in school vending machine options.</p></sec></sec><sec><title>Interventions</title><p>The program considers health promotion interventions to be a series of activities designed to change or influence existing behaviors or practices related to obesity, nutrition, and physical activity. As part of the progress monitoring report, the states were asked how many interventions they had in place at the time of the report. The states were instructed to include pilot projects, interventions with funds from the Nutrition and Physical Activity Program to Prevent Obesity and Other Chronic Diseases, and interventions based on the program's concepts. The states indicated that they were in various stages of developing and implementing interventions to prevent obesity and other chronic diseases, perhaps reflecting the varying periods of time over which the 20 states included in this article were funded. (Although all 20 states received funding by July 2003, some initially received funds through a previous cooperative agreement and continued to be funded.) Eleven of the 20 states (55%) reported having interventions in place in the past 6 months.</p><sec><title>Strategies</title><p>As mentioned previously, the CDC's Division of Nutrition and Physical Activity identified five strategies that states can use to focus their program interventions. Increased physical activity was the most frequently used strategy, followed by increasing fruit and vegetable consumption (<xref rid="F3" ref-type="fig">Figure 3</xref>). Promoting caloric balance, decreasing television-viewing time, and increasing breastfeeding were used less frequently. Most interventions (17 of 29 interventions, or 59%) incorporated multiple strategies.</p><boxed-text position="float"><fig position="float" id="F3" fig-type="diagram"><label>Figure 3</label><caption><p>Percentage of interventions incorporating key evidence-based strategies. Percentages were calculated based on 29 active interventions from January through June 2004. Because some interventions incorporated multiple strategies, totals across all columns exceed 100%</p></caption><alt-text>Bar graph</alt-text><alternatives><table frame="hsides" rules="groups"><tbody><tr><td align="left" valign="top" rowspan="1" colspan="1">Increasing Physical Activity</td><td align="left" valign="top" rowspan="1" colspan="1">83% (24)</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Increasing Fruit and Vegetable Consumption</td><td align="left" valign="top" rowspan="1" colspan="1">55% (16)</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Balancing Caloric Intake and Expenditure</td><td align="left" valign="top" rowspan="1" colspan="1">38% (11)</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Decreasing Television-Viewing Time</td><td align="left" valign="top" rowspan="1" colspan="1">38% (11)</td></tr><tr><td align="left" valign="top" rowspan="1" colspan="1">Increasing Breastfeeding</td><td align="left" valign="top" rowspan="1" colspan="1">21% (6)</td></tr></tbody></table><graphic xlink:href="PCD31A23s03"/></alternatives></fig></boxed-text></sec><sec><title>Settings</title><p>The most frequently reported intervention setting was the school setting (12 of 29 interventions, or 41%), and the second most frequently reported intervention setting was the community (7 interventions, or 24%). Childcare centers (6 interventions, or 21%) and worksites (4 interventions, or 14%) were also popular settings. A few interventions took place in a family setting (3 interventions, or 10%) and in health care or hospital settings (2 interventions, or 7%). Eleven interventions (38%) involved settings such as youth programs, early childhood education programs, a recreation center, and a religious setting. These settings reflect the target populations; the majority of states focused their interventions on children.</p></sec></sec></sec><sec><title>Implications</title><sec><title>Infrastructure</title><p>The funded states have numerous partners planning, donating staff time, implementing interventions, and cosponsoring events; half of the states have partners contributing money. The majority of the states have obtained, appropriated, or reallocated funds from outside their cooperative agreements for nutrition and physical activity programs, primarily from state and federal programs, with leveraged funding amounts ranging from no funding to more than $1 million. The majority of states leveraged money for planning and programs.</p></sec><sec><title>Environmental changes</title><p>The funded states are implementing environmental changes, most frequently through legislation. Polices and other environmental changes such as urban planning are also being used, although less frequently.</p></sec><sec><title>Interventions</title><p>More than half of the states reported having interventions in place from January through June 2004. The most frequently used strategies for an intervention were increased physical activity and increased fruit and vegetable consumption. The most frequently reported settings were school systems and communities.</p></sec></sec><sec sec-type="conclusions"><title>Conclusion</title><p>The states funded by the Nutrition and Physical Activity Program to Prevent Obesity and Other Chronic Diseases have made progress in establishing the infrastructure needed for health promotion. More than half of the states have begun implementing interventions using evidence-based strategies in various settings. Environmental modifications have the potential for creating sustainable change, so states' efforts in implementing polices and other environmental changes are particularly encouraging. The initial accomplishments of the state programs indicate that states can promote environmental and policy changes to address the challenges of obesity and other chronic diseases.</p></sec>

Using Concept Mapping to Develop a Logic Model for the Prevention Research Centers Program

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Sundra</surname><given-names>Demia L</given-names></name><degrees>MPH</degrees><aff>Centers for Disease Control and Prevention, Prevention Research Centers Program</aff><address><email>dsundra@cdc.gov</email><addr-line>4770 Buford Hwy, Mail Stop K-45, Atlanta, GA 30341</addr-line><phone>770-488-5506</phone></address></contrib><contrib contrib-type="author" corresp="no"><name><surname>Anderson</surname><given-names>Lynda A</given-names></name><degrees>PhD</degrees><aff>Centers for Disease Control and Prevention, Atlanta, Ga, and Rollins School of Public Health, Emory University, Atlanta, Ga</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Gwaltney</surname><given-names>Margaret K</given-names></name><degrees>MBA</degrees><aff>COSMOS Corporation, Bethesda, Md</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Brownson</surname><given-names>Ross C</given-names></name><degrees>PhD</degrees><aff>Prevention Research Center, Saint Louis University School of Public Health, St. Louis, Mo</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Kane</surname><given-names>Mary</given-names></name><degrees>MS</degrees><aff>Concept Systems, Inc, Ithaca, NY</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Cross</surname><given-names>Alan W</given-names></name><degrees>MD</degrees><aff>University of North Carolina at Chapel Hill, Chapel Hill, NC</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Mack</surname><given-names>Richard</given-names><suffix>Jr</suffix></name><degrees>PhD</degrees><aff>Harlem Center for Health Promotion and Disease Prevention, Columbia University, New York, NY</aff><aff>Dr Mack is now a consultant, New York, NY</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Schwartz</surname><given-names>Randy</given-names></name><degrees>MSPH</degrees><aff>American Cancer Society, New England Division, Framingham, Mass</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>Sims</surname><given-names>Tom</given-names></name><degrees>MA</degrees><aff>West Virginia Bureau for Public Health, Charleston, WV</aff></contrib><contrib contrib-type="author" corresp="no"><name><surname>White</surname><given-names>Carol R</given-names></name><degrees>MPH</degrees><aff>University of Kentucky, Lexington, Ky. Ms Gwaltney is now with Abt Associates Inc, Bethesda, Md</aff></contrib>

Preventing Chronic Disease

<sec><title>Introduction</title><p>The Centers for Disease Control and Prevention's (CDC's) <italic>Framework for Program Evaluation in Public Health</italic> provides public health practitioners and evaluators with a practical, six-step approach for effective evaluation (<xref ref-type="bibr" rid="B1">1</xref>). The framework helps public health programs address increased accountability requirements, program improvement processes, and public health decision making (<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref>). The two initial steps in the CDC's evaluation framework are describing the program and engaging stakeholders. The program description step includes developing a logic model that visually depicts the hypothesized relationships among program resources, program activities, and the results the program hopes to achieve &#x02014; in other words, the program's underlying theory of change (<xref ref-type="bibr" rid="B3">3</xref>). The CDC evaluation framework and other models recommend engaging stakeholders during the logic model development to increase the usefulness and validity of the resulting model (<xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B4">4</xref>-<xref ref-type="bibr" rid="B6">6</xref>). The logic model can then be used as the basis for future program evaluation efforts.</p><p>Examples are available of public health programs that have used participatory methods to develop logic models (<xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B7">7</xref>-<xref ref-type="bibr" rid="B9">9</xref>), but the methods used by the programs to encourage stakeholder input are not the focus of those publications. In addition, participatory methods for developing logic models have typically involved small or single-site programs or engaged a small group of program representatives. In this article, we detail the efforts of the CDC's Prevention Research Centers (PRC) Program, in which concept mapping was used to develop a national program logic model.</p><p>Concept mapping can be used to identify key elements of a program and show their relationships to one another (<xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B11">11</xref>). Several projects have used concept mapping to set priorities, plan programs, and evaluate programs (<xref ref-type="bibr" rid="B12">12</xref>-<xref ref-type="bibr" rid="B15">15</xref>). Although the methodology has been used previously to develop a logic model for a single program (<xref ref-type="bibr" rid="B16">16</xref>), we could find no published studies that used concept mapping to develop a logic model for a national program. We describe the application of concept mapping in the PRC Program, a large, multisite program with national, state, and local stakeholders distributed throughout the United States. These methods should be beneficial for individuals involved in programs that are developing logic models for evaluation planning.</p><p>The PRC Program funds 33 university-based research centers to conduct community-based participatory research and training on chronic disease and health promotion issues facing communities today (<xref ref-type="bibr" rid="B17">17</xref>). The PRC Program is the CDC's largest extramural research program and encourages academic, community, and public health collaboration in conducting prevention research and applying research in practice and policies (<xref ref-type="bibr" rid="B18">18</xref>). Stakeholders in the PRC Program include researchers in schools of public health, schools of medicine, and other academic departments; community members; community-based organizations; tribal organizations; public health practitioners in state, county, and city health departments; other government agencies; school administrators and teachers; national advocacy organizations and public health associations; the CDC; Congress; and many others. During the first year of the logic model development project (2001), the PRC Program funded 26 centers in 24 states.</p><p>To address the increased emphasis on accountability and meet the recommendations made in the 1997 Institute of Medicine (IOM) review of the PRC Program (<xref ref-type="bibr" rid="B19">19</xref>), the program's leaders decided to initiate a national evaluation strategy. Using the CDC evaluation framework as a guide (<xref ref-type="bibr" rid="B1">1</xref>), an evaluation planning project was funded, with the goal of engaging stakeholders to develop an overall program description and logic model (steps 1 and 2 of the CDC evaluation framework). An external evaluation contractor was funded to facilitate a participatory process that would ensure the key stakeholders of the PRC Program had a role in developing the logic model.</p></sec><sec sec-type="methods"><title>Methods</title><p>The national logic model was developed in three stages. First, we constructed a logic model draft using data from the concept mapping process. Second, we refined the draft through regional meetings with PRC Program stakeholders. Third, we distributed the draft and written narrative to stakeholders and obtained suggestions through a structured feedback tool designed to help revise the model.</p><p>A collaborative evaluation design team (CEDT) comprising representatives from major stakeholder groups was formed and oversaw all aspects of the project. This group included experts in community-based participatory research, public health, disease prevention, and program evaluation who worked in various settings, including universities, state health departments, voluntary health agencies, and local organizations. The CEDT assisted with the concept mapping process and development of the PRC logic model, communicated with the constituency represented by each team member, and advised the evaluation contractor and the CDC on all aspects of project implementation.</p><sec><title>Concept mapping</title><p>We used concept mapping to develop our program framework, or logic model. Concept mapping provides a visual representation of the complex relationships among ideas and results and integrates qualitative processes with quantitative methods (<xref ref-type="bibr" rid="B20">20</xref>). Unlike other qualitative methods such as focus groups, concept mapping provides a structured approach that allows participants to identify issues and participate in the actual interpretation of their group perceptions (<xref ref-type="bibr" rid="B21">21</xref>). Concept mapping also incorporates statistical tools that provide precise and credible data from qualitative information. The method was selected because it can elicit ideas from large and diverse groups about an issue or a topic within a short time and because its design enables it to overcome geographic barriers (<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B22">22</xref>).</p><p>The concept mapping process had three phases: 1) project planning, which included developing the focus prompt (i.e., the type of input desired) and identifying participants (November 2001&#x02013;January 2002); 2) idea generation and structuring (February&#x02013;March 2002), and 3) analysis and interpretation of the concept maps (April&#x02013;June 2002). During each step, we encouraged ongoing communication through committee meetings and conference calls to obtain stakeholder input and provide updates about each step of the concept mapping process.</p><sec><title>Project planning phase</title><p>The evaluation contractor collaborated with the CEDT to develop the following two prompts to elicit ideas about the purpose and function of the PRC Program, with one focusing on the national level and one on the local level:</p><list list-type="bullet"><list-item><p>To ensure national excellence in prevention, a Prevention Research Center should have the following specific characteristic or function . . .</p></list-item><list-item><p>To successfully promote health in a community, an effective Prevention Research Center should have the following specific characteristic or skill . . .</p></list-item></list><p>We compiled a list of 175 PRC Program stakeholders to participate in the concept mapping process using the nationally focused prompt. Stakeholders included representatives from national organizations, such as Chronic Disease Directors, Directors of Health Promotion and Education, Association of Schools of Public Health, and Association of Teachers of Preventive Medicine; members selected from the IOM report review committee (<xref ref-type="bibr" rid="B19">19</xref>); CDC leaders familiar with the PRC Program; CDC program staff members; the PRC national community committee, which is composed of representatives from each PRC community committee, who advise the program, facilitate training of community members, and educate about prevention research (<xref ref-type="bibr" rid="B23">23</xref>); and PRC leaders, such as principal investigators, directors, administrators, and researchers from the PRCs. PRC leaders could invite other key stakeholders such as university leaders to participate in the brainstorming process at the national level.</p><p>We generated a similar list of 165 stakeholders to participate in the concept mapping process using the locally focused prompt. Participants were identified from the following groups: PRC community committees, research participants, health department partners, and PRC leaders such as principal investigators, directors, administrators, and researchers from the PRCs.&#x000a0; Because we knew that some stakeholders might not be able to respond online or by fax, and to ensure that the community's input was obtained, we selected a community liaison in each PRC who assisted community representatives in the concept mapping process. We invited some stakeholders who had national and local perspectives on the PRC Program to respond to both focus prompts.</p><sec><title>Idea generation and structuring phase</title><p>We invited participants to submit up to 10 ideas in response to the focus prompt using a secure Web site or by mailing or faxing their ideas to the evaluation contractor. Because participants submitted their ideas anonymously, we could not calculate exact response rates or the average number of items submitted per respondent. However, based on unique identifiers, we estimated that 145 stakeholders (83%) responded to the nationally focused prompt, and 135 responded (82%) to the locally focused prompt.</p><p>Members of the CEDT reviewed the statements that had been generated for each prompt and eliminated repetitive statements, yielding 88 unique statements for the national responses and 75 unique statements for the local responses. The statements were sorted into themes (<xref ref-type="bibr" rid="B24">24</xref>). The national and local statements were then sorted independently by two subsets of participants who were selected for their familiarity with PRCs. For the national statements, 35 stakeholders were contacted, with 20 (57%) resulting participants. For the local statements, 30 stakeholders were contacted, with 17 (57%) resulting participants. The individuals were asked to sort the statements into categories, or themes, based on similarity of ideas. Participants either used the project's Web site to sort the statements into categories or manually sorted statements that had been printed on cards. Participants were asked to create their own categories; they were told that each statement could be placed into only one category, and the sorting process should result in more than one category but fewer categories than the total number of statements.</p></sec><sec><title>Analysis and interpretation phase</title><p>We used a software tool designed for multiple stakeholder input (Concept Systems, Inc, Ithaca, NY) to construct two separate concept maps (<xref ref-type="bibr" rid="B12">12</xref>). An expert in concept mapping conducted the analysis. First, a similarity matrix was constructed that represented the relative similarity of participants' sorting statements. Second, the total similarity matrix was analyzed using nonmetric multidimensional scaling analysis with a two-dimensional solution, which generated x and y coordinates in two-dimensional space for each statement based on its mathematical similarity to other statements. Configuring the multidimensional scaling of the statement points in two dimensions on a point map was the foundation for the final results. Third, statements were combined into clusters using a hierarchical cluster analysis. The results of the hierarchical cluster analysis were superimposed on the multidimensional scaling results to create a map displaying the points graphically within each group, with polygonal boundaries surrounding the points in each cluster group. A hierarchical cluster analysis yields all possible cluster solutions, from each statement in its own cluster to all statements in one cluster. A standardized, systematic process is applied to identify the most useful cluster number for each project. The appropriate number of clusters is determined by working with subject experts who consider the range of issues represented, the purpose and intended uses of the resulting map, and the observed coherence of clusters at different levels (<xref ref-type="bibr" rid="B21">21</xref>).</p><p>The CEDT reviewed the two PRC Program cluster maps and the statements associated with each cluster. The CEDT members then agreed on a theme and label for each cluster on both maps. These maps became the national- and local-level concept maps for the PRC Program.</p></sec></sec><sec><title>Developing the logic models</title><p>We developed a draft logic model diagram, showing PRC Program inputs, activities, and outcomes and incorporating data from the concept mapping process. This information was supplemented by information from program documents. We presented the draft logic models at regional meetings in May and June 2002 and distributed the logic model with a written narrative in a structured feedback tool in September 2002. We used the feedback received through these mechanisms to make final revisions.&#x000a0;The final logic model and narrative were then broadly distributed to the PRCs and other stakeholders.</p></sec></sec><sec><title>Results</title><sec><title>Concept maps</title><p>The national-level concept map had nine clusters (<xref rid="F1" ref-type="fig">Figure 1</xref>):</p><list list-type="bullet"><list-item><p>Diversity and sensitivity</p></list-item><list-item><p>Community engagement</p></list-item><list-item><p>Research methods</p></list-item><list-item><p>Research agenda</p></list-item><list-item><p>Core expertise and resources</p></list-item><list-item><p>Active dissemination</p></list-item><list-item><p>Technical assistance</p></list-item><list-item><p>Training</p></list-item><list-item><p>Relationships and recognition</p></list-item></list><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure 1</label><caption><p>National concept map showing 9 clusters and 88 statements. PRC indicates Prevention Research Center; CDC, Centers for Disease Control and Prevention.</p></caption><alt-text>Component model</alt-text><alternatives><table frame="hsides" rules="groups"><tbody><tr><td rowspan="1" colspan="1"> <bold>Diversity and sensitivity cluster: associated statements</bold> <break/>Understanding of cultural diversity and its impact on a PRC's mission; a PRC staff that reflects the diversity of the community; demonstrated ability to listen to and understand the audience; willingness to tackle tough and politically sensitive issues; demonstrated accountability to multiple stakeholders such as the community, academic institutions, and funding sources</td></tr><tr><td rowspan="1" colspan="1"> <bold>Community engagement cluster: associated statements</bold> <break/>A demonstrated view of the needs, interests, and values of the community; success in establishing community partnerships; demonstrated awareness of prevention-oriented research and service programs previously carried out in the community; mechanisms for documenting collaborations with the community; the capacity to effectively identify health promotion or prevention issues of interest to the community; a strong relationship with community leaders; visibility in the defined community; methods for regularly monitoring the changing health risks in the community; a community advisory group that is active in guiding the key functions of a PRC; ability to collaborate with communities to set research priorities that address community-identified needs; connections and strong relationships with the defined community; representatives of the defined community (e.g. ethnic community representative, individuals with HIV, voluntary health organizations, outreach workers) on the community advisory group</td></tr><tr><td rowspan="1" colspan="1"> <bold>Research methods cluster: associated statements</bold> <break/>A research focus on social, environmental, and political influences in addition to the behavior of individuals; success in conducting effective participatory research; implementation of a national needs assessment to monitor trends in chronic disease; contributions to the measurement of risk factors and disease; activities related to eliminating health disparities; an effective plan to conduct participatory research; demonstrated use of science in designing community-based health interventions; success in conducting behavioral interventions in the defined community; implementation of economic evaluations to determine the costs and benefits of prevention; methods for evaluating the level of community participation in research; ability to conduct multisite effectiveness trials of interventions that have been shown to be efficacious; demonstrated knowledge of the advantages and disadvantages of participatory methods for research and training; measures of capacity for community involvement in research</td></tr><tr><td rowspan="1" colspan="1"> <bold>Research agenda cluster: associated statements</bold> <break/>A research focus on social, environmental, and political influences in addition to the behavior of individuals; success in conducting effective participatory research; implementation of a national needs assessment to monitor trends in chronic disease; contributions to the measurement of risk factors and disease; activities related to eliminating health disparities; an effective plan to conduct participatory research; demonstrated use of science in designing community-based health interventions; success in conducting behavioral interventions in the defined community; implementation of economic evaluations to determine the costs and benefits of prevention; methods for evaluating the level of community participation in research; ability to conduct multisite effectiveness trials of interventions that have been shown to be efficacious; demonstrated knowledge of the advantages and disadvantages of participatory methods for research and training; measures of capacity for community involvement in research</td></tr><tr><td rowspan="1" colspan="1"> <bold>Core expertise and resources cluster: associated statements</bold> <break/>An emphasis on hiring faculty and staff with experience in public health practice; documented specific and measurable outcomes; PRC employees with strong interpersonal and team-building skills; documented qualitative data collection and analysis skills among PRC researchers; a PRC unit or function focused on methodology relevant to the goals of the PRC; an evaluation strategy for the center's faculty and staff; an effective plan for keeping up-to-date on the latest research findings and best practices in the area of focus; PRC leaders who possess effective administrative skills; self-evaluation using internal and external resources (e.g., a Community Advisory Group) to implement self-evaluation measures; measures of multidisciplinary expertise in PRC projects; demonstrated physical infrastructure to carry out the PRC stated objectives; a PRC unit or function focused on program evaluation; a skilled staff to assist researchers in the administrative aspects of the research</td></tr><tr><td rowspan="1" colspan="1"> <bold>Active dissemination cluster: associated statements</bold> <break/>A systematic plan for the dissemination of PRC research findings; success in promoting research findings among practitioners to influence policy development and program planning; success in promoting research findings among policy makers to influence policy development and program planning; dissemination of prevention research to practitioners, researchers, policy makers, and the public in ways that these audiences can use the information; publication of community-based research in major peer-reviewed journals; integration of the translation of research findings into practice as a part of projects; a core PRC unit or function focused on dissemination of findings related to the goals of the PRC; timely publication of research findings in peer-reviewed journals; success in translating research findings for public health settings; leadership in disseminating research findings to the local health departments and to the community in a form that is in addition to and different from publications; development and implementation of studies on dissemination research; ability to communicate research methods, issues, and results in a clear and easy-to-understand manner</td></tr><tr><td rowspan="1" colspan="1"> <bold>Technical assistance cluster: associated statements</bold> <break/>Visible contributions from the PRC among international audiences; a means of regularly sharing with other PRCs updates on various prevention research projects; success in working with state and or county health departments in developing, implementing and translating research into practice; capacity to provide technical assistance to public health organizations; willingness to provide in-kind technical assistance to public health organizations that have limited resources</td></tr><tr><td rowspan="1" colspan="1"> <bold>Training cluster: associated statements</bold> <break/>Implementation of training programs that develop the next generation of prevention science researchers; a national scholar program that involves an exchange of researchers among PRCs and/or between PRCs and the CDC; a training component intended to ensure the availability of qualified prevention scientists in the future; a core PRC unit or function focused on training related to the goals of the PRC; PRC training programs for international audiences; a program to mentor mid-level researchers for future leadership positions in prevention research</td></tr><tr><td rowspan="1" colspan="1"> <bold>Relationships and recognition cluster: associated statements</bold> <break/>A beneficial relationship with the leadership at the CDC; success in working with national organizations in strategic planning, implementation, and advocacy; capacity to collaborate with other PRCs that share common research interests; a beneficial relationship with the university administration; a training program for community audiences; visibility at the national level; prominence within the academic health center; visibility in the academic institution in which the PRC is located; collaboration among PRCs through sharing research projects and the products of research projects; strong national leadership (i.e., CDC) to influence the direction, focus, and identity of the prevention research program; the ability to connect with other academic units at the institution where the PRC is located in order to conduct research</td></tr></tbody></table><graphic xlink:href="PCD31A06s01"/></alternatives></fig></boxed-text><p>The local-level concept map had 11 clusters (<xref rid="F2" ref-type="fig">Figure 2</xref>):</p><list list-type="bullet"><list-item><p>Communication and dissemination</p></list-item><list-item><p>Outreach</p></list-item><list-item><p>Promotes community involvement</p></list-item><list-item><p>Responsive to community input</p></list-item><list-item><p>Builds community capacity</p></list-item><list-item><p>Committed community advisory group</p></list-item><list-item><p>Trust</p></list-item><list-item><p>Defining and measuring community outcomes</p></list-item><list-item><p>Training and mentoring</p></list-item><list-item><p>Human resources</p></list-item><list-item><p>Translation of research to practice</p></list-item></list><boxed-text position="float"><fig id="F2" fig-type="diagram" position="float"><label>Figure 2</label><caption><p>Local concept map showing 11 clusters and 75 statements. PRC indicates Prevention Research Center; CDC, Centers for Disease Control and Prevention.</p></caption><alt-text>Component model</alt-text><table-wrap id="d34e500" position="anchor"><table frame="hsides" rules="groups"><tbody><tr><td rowspan="1" colspan="1"> <bold>Communication and dissemination cluster: associated statements</bold> <break/>Communication with the public using multiple different media; success at disseminating products of research with other PRCs; reports research findings back to community leaders; a strategic plan for disseminating products of research; success in disseminating research results back to the community in a useful form; effective communication with multiple audiences such as public health agencies, private health agencies, community groups, and the public; documented publicity of PRC events and activities; effective communication with policy makers and law makers; success in sharing evidence-based programs in addition to conducting research; develops research reports that are used by community leaders</td></tr><tr><td rowspan="1" colspan="1"> <bold>Outreach cluster: associated statements</bold> <break/>Success in sharing the "how-to's" of community involvement; an effective system of communication among partners and the community; continuous tracking and reporting of health indicators to community health agencies; develops awareness in the community of the difference between the economic impacts of health care and the economic impacts of health promotion and disease prevention; PRC staff that have designated responsibility for community outreach.</td></tr><tr><td rowspan="1" colspan="1"> <bold>Promotes community involvement cluster: associated statements</bold> <break/>Effective leadership that models and supports sensitive approaches to community partnerships; ability to advocate for community involvement in planning, implementing, and evaluating intervention research; demonstrated desire to share resources, power, and expertise with partners; success in implementing the principles of community work and knowing how these principles differ from clinical research; ability to link community members with relevant decision makers; knowledge of the community power structure and the decision-making systems in a community; capacity to share control with partners</td></tr><tr><td rowspan="1" colspan="1"> <bold>Responsive to community input cluster: associated statements</bold> <break/>An agenda that is largely determined by the community; research that is driven primarily by community needs rather than funders; participation of community members in all aspects of research design and intervention; an ability to elicit and monitor community concerns; documented strategies for developing a sense of responsibility among the community groups who assist with research; documented input from community representatives regarding the accomplishments of the PRC; success at responding to input from the community even when it may not result in ideas that are fundable; prioritized community research needs based on community input; effective use of data from the community (e.g., needs assessment) to identify and address specific health problems; demonstrated understanding of the structure of the community; willingness to act on community recommendations; an equal partnership between the PRC and community members; demonstrated respect for contributions of all members of the community</td></tr><tr><td rowspan="1" colspan="1"> <bold>Builds community capacity cluster: associated statements</bold> <break/>Demonstrated responsiveness for improving the health of the community; assessment of a community's capacity to implement health promotion and disease prevention among its community members; documented plans to allow time for the community to build ownership; programs that target the health needs of the community served; a demonstrated genuine care and concern for the health and well being of the community; established methods to enhance a community's capacity</td></tr><tr><td rowspan="1" colspan="1"> <bold>Committed community advisory group cluster: associated statements</bold> <break/>An established community advisory group for the PRC; documented guidelines addressing the role of the community advisory group; a community advisory board with members knowledgeable about specific needs and assets of the community; a diverse community advisory group with respect to multiple criteria, such as ethnicity, organizations represented, and target health condition; a community advisory group for the core research project</td></tr><tr><td rowspan="1" colspan="1"> <bold>Trust cluster: associated statements</bold> <break/>Recognition by members of the defined community that the PRC's activities benefit that community; recognized as a resource for the community; credibility within the community the PRC serves; considered as trustworthy within the community the center serves</td></tr><tr><td rowspan="1" colspan="1"> <bold>Defining and measuring community outcomes cluster: associated statements</bold> <break/>Commitment to act on the issues identified by prior research; a broad definition of health incorporated into the PRC&#x02019;s mission; a documented focus on social issues that will have a beneficial effect on multiple disease risk problems; measurable PRC objectives; a multidisciplinary approach that goes beyond a sole focus on individual behavior changes; measures to define the community; evaluation of progress and success with new initiatives</td></tr><tr><td rowspan="1" colspan="1"> <bold>Training and mentoring cluster: associated statements</bold> <break/>Success at training students to be culturally and ethnically sensitive; ongoing training and learning opportunities for local public health practitioners; a mentorship or scholarship program that involves underrepresented groups or individuals</td></tr><tr><td rowspan="1" colspan="1"> <bold>Human resources cluster: associated statements</bold> <break/>Incentives or a reward system for faculty who undertake community-based, participatory research; adequate facility and personnel to support projects; capacity to effectively address disparities in health care and outcomes; research staff with "real-world" experience working with communities &#x02014; that is, staff who have worked outside of academia; staff with dissemination skills; a diverse team of staff and faculty; established skills in community organization and community action; a diverse PRC staff; a PRC staff that includes community members; demonstrated ability to sustain successful programs</td></tr><tr><td rowspan="1" colspan="1"> <bold>Translation of research to practice cluster: associated statements</bold> <break/>Success in implementing community-based programs that have been shown to be effective; effective translation of research findings into practice; scientists who can communicate with nonscientists; translation of research findings into practice using culturally effective methods; capability to translate research into practical strategies to change practice and policy</td></tr></tbody></table></table-wrap><graphic xlink:href="PCD31A06s02"/></fig></boxed-text></sec><sec><title>Development of the program logic model</title><p>We placed the concept map data into the appropriate columns of the logic models: program input, activity, or outcome (Tables <xref ref-type="table" rid="T1">1</xref> and <xref ref-type="table" rid="T2">2</xref>). For example, the core expertise and resources cluster from the concept map (<xref rid="F1" ref-type="fig">Figure 1</xref>) was placed in the input column of the draft national logic model (<xref rid="T1" ref-type="table">Table 1</xref>). Likewise, the community engagement cluster was placed in the activities column of the national logic model. We continued this process until all clusters from the national concept map had been categorized into the columns of the national logic model. Using the same process for the local logic model, we placed the committed community advisory board cluster from the local concept map (<xref rid="F2" ref-type="fig">Figure 2</xref>) into the input column of the local logic model (<xref rid="T2" ref-type="table">Table 2</xref>) and the trust cluster from the concept map into the outcome column of the model. The remaining cluster information from the local concept map was placed into the appropriate columns of the local logic model. We reviewed program documents, such as the IOM report (<xref ref-type="bibr" rid="B19">19</xref>), authorizing legislation (<xref ref-type="bibr" rid="B25">25</xref>), and PRC guiding principles (<xref ref-type="bibr" rid="B17">17</xref>), to identify other activities and outcomes relevant to the program. Information from these documents augmented the concept mapping data.</p><p>We presented the draft logic models at three regional meetings. The meetings were attended by 57 participants representing academic, community, and public health partners within the PRC Program. Based on comments received, we combined the two draft logic models into one logic model for the national PRC Program. Meeting participants agreed that the single PRC Program logic model should reflect the key clusters from the locally focused prompt that were not associated with the nationally focused prompt: community capacity building, trust, and translation of research to practice.</p><p>We distributed the single national logic model with a written narrative in a structured feedback tool. Representatives in 28 PRCs (rather than 26, because two additional PRCs had been funded) received the feedback tool, including members of the Chronic Disease Directors, the Directors of Health Promotion and Education, the PRC National Community Committee, and the CDC program staff. We asked each PRC to gather input from various respondents, including academic and community partners, and then provide a single response representing the individual PRC. The PRCs were asked to send their comments to the evaluation contractor; the response rate was 100%. As a result of the feedback, the logic model underwent minor revisions.</p><p>The PRC Program office at the CDC distributed the final logic model and accompanying narrative to program stakeholders and posted it on the PRC Program Web site (<ext-link ext-link-type="uri" xlink:href="http://www.cdc.gov/prc/">http://www.cdc.gov/prc/</ext-link>). We have presented the logic model at several national evaluation, public health, and health education conferences and meetings, such as the National Conference on Chronic Disease Prevention and Control and meetings of the American Public Health Association, American Evaluation Association, and Society of Public Health Educators.</p></sec></sec><sec><title>Discussion</title><p>Concept mapping can be a useful tool for constructing a logic model for a national program. We identified several benefits from our experiences with the PRC Program. First, the most obvious benefit was that the logic model was based on a set of concepts that came directly from stakeholders. The concept map and underlying statements served as the foundation for the logic model refinement process. In addition, components of the final logic model were easily linked to the original concept mapping ideas submitted by stakeholders. Second, compared with an initially proposed logic model (available upon request) developed by a few CDC staff members and select partners, the logic model based on the concept mapping data was more comprehensive and representative of the processes and outcomes involved in prevention research. For the first time, community representatives could see themselves visually represented in a program's activities and outcomes. For example, their role in establishing a research agenda is clear, as is the intended outcome of enhanced community capacity for disease prevention.</p><p>Consistent with the CDC framework for evaluation recommendations, engaging stakeholders in the development of the program logic model was worth the investment of resources (<xref ref-type="bibr" rid="B1">1</xref>). Concept mapping encouraged participants to provide their opinions about the PRC Program anonymously during the idea-generation phase. The ability to provide anonymous input was important during the early project phases because trust was being established among the various stakeholder groups. Combining concept mapping with other methods for eliciting feedback throughout the project helped address the significant numbers of stakeholders who expressed differing views or general skepticism about the process, an issue that may be inherent in any large, multisite program. Overall, open discussions, compromise among people with conflicting views, transparent use of feedback and decision making, inclusion of stakeholder perspectives, and repeated explanations of the process were important methods for keeping all participants positively engaged and supportive of the final product. Our experiences and challenges were similar to those reported in other participatory evaluation process reports (<xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B27">27</xref>) and will be presented in another article.</p><p>Concept mapping has gained acceptance by researchers in the last 15 years; in the last 5 years, its use has been facilitated by Web applications for participant data collection and analysis. In addition, online data collection methods are more cost-effective and efficient than other participatory methods involving large groups. Another benefit of using a Web-based system is that the initial maps can be presented to stakeholders quickly. In our experience, the process allowed us to gather data from stakeholders in numerous geographic areas and then present the concept maps to PRC representatives 1 month after the idea generation and structuring were completed.</p><p>Concept mapping as a tool for developing a logic model does have some shortcomings. First, a logic model derived from a concept map is based on stakeholder perspectives; it is not a tested theory of how a program functions and arrives at intended outcomes. Therefore, it may not reflect some realities of program implementation and outcomes (<xref ref-type="bibr" rid="B4">4</xref>). Future evaluation efforts in the PRC Program will clarify the concepts and logic in the national model. Second, concept mapping was a new process for most stakeholders. Many who were not familiar with qualitative methods and terminology initially struggled to understand how the concept mapping activities would result in the construction of a logic model for the program. Finally, although many diverse perspectives are represented in the findings of the concept mapping process, they should not be interpreted as representing the views of all stakeholders.</p><p>Given the challenges faced during the project, we recommend using three of the strategies we found most helpful. First, program evaluation experts should be used to obtain the information from the concept mapping statements and other program documents to construct an initial logic model. Second, stakeholders should be fully informed about the concept mapping process and given concept mapping examples (such as this article) so that they can become familiar with the use of concept mapping as a tool for logic model development. Third, concept mapping data should be supplemented with program documents and stakeholder feedback, a strategy that is consistent with recommendations for using multiple methods for developing a program theory (<xref ref-type="bibr" rid="B6">6</xref>). Future evaluation project planners should consider using electronic methods for gathering feedback, such as Web-based conferencing and telephone focus groups.</p><p>Concept mapping is a valuable method for developing a logic model, particularly for a large program with a diverse group of stakeholders. Having a national logic model has permitted the PRC Program to identify its centers' outcomes and functions. The process and final logic model has incorporated the input of the program's national and community partners, engaged stakeholders, and provided the PRC Program with a platform on which to design and implement a national evaluation strategy.</p></sec></sec>

Barriers to Diabetes Self-management Education Programs in Underserved Rural Arkansas: Implications for Program Evaluation

Could not extract abstract

<contrib contrib-type="author" corresp="yes"><name><surname>Balamurugan</surname><given-names>Appathurai</given-names></name><degrees>MD, MPH</degrees><role>Senior Epidemiologist</role><aff>Arkansas Department of Health and Human Services</aff><aff>Dr Balamarugan is also an assistant professor in the Fay W. Boozman College of Public Health at the University of Arkansas for Medical Sciences, Little Rock, Ark</aff><address><email>abalamurugan@healthyarkansas.com</email><addr-line>4815 W Markham, Slot 32, Little Rock, AR 72205</addr-line><phone>501-280-4830</phone></address></contrib><contrib contrib-type="author"><name><surname>Rivera</surname><given-names>Mark</given-names></name><degrees>PhD</degrees><role>Evaluation and Research Scientist</role><aff>Applied Behavioral Research, Epidemiology, Surveillance, and Evaluation (ABRESE), Centers for Disease Control and Prevention, Division of Diabetes Translation, Program Development Branch, Atlanta, Ga</aff></contrib><contrib contrib-type="author"><name><surname>Jack</surname><given-names>Leonard</given-names><suffix>Jr</suffix></name><degrees>PhD, MSc</degrees><role>Team Lead</role><aff>Applied Behavioral Research, Epidemiology, Surveillance, and Evaluation (ABRESE), Centers for Disease Control and Prevention, Division of Diabetes Translation, Program Development Branch, Atlanta, Ga</aff></contrib><contrib contrib-type="author"><name><surname>Morris</surname><given-names>Sharon</given-names></name><role>Project Officer</role><aff>Applied Behavioral Research, Epidemiology, Surveillance, and Evaluation (ABRESE), Centers for Disease Control and Prevention, Division of Diabetes Translation, Program Development Branch, Atlanta, Ga</aff></contrib><contrib contrib-type="author"><name><surname>Allen</surname><given-names>Kristen</given-names></name><degrees>RD, CDE</degrees><role>Nutrition Consultant</role><aff>Arkansas Diabetes Prevention and Control Program, Little Rock, Ark</aff></contrib>

Preventing Chronic Disease

<sec><title>Background</title><p>Diabetes prevalence has reached epidemic proportions in the United States. In 2002, 18 million people were estimated to have diabetes (<xref rid="B1" ref-type="bibr">1</xref>). The direct medical and indirect expenditures attributable to diabetes were estimated at $132 billion in 2002 (<xref rid="B2" ref-type="bibr">2</xref>). Future projections indicate that diabetes prevalence will continue to increase, expenditures will remain high, and diabetes will continue to be a serious health concern (<xref rid="B1" ref-type="bibr">1</xref>). Establishing the efficacy and effectiveness of disease management and education interventions that target health care providers, patients, families, and communities is critically important.</p><p>A systematic review of published studies addressing the effectiveness of population-based diabetes-related interventions recommends diabetes self-management education (DSME) (<xref rid="B3" ref-type="bibr">3</xref>). DSME empowers people to manage diabetes through education about nutrition, medication and insulin therapy, stress management, and preventive foot and eye care (<xref rid="B4" ref-type="bibr">4</xref>). DSME has been shown to be effective in community settings (<xref rid="B5" ref-type="bibr">5</xref>).</p><p>Although few studies have examined the challenges and barriers associated with establishing DSME programs in underserved areas (<xref rid="B6" ref-type="bibr">6</xref>,<xref rid="B7" ref-type="bibr">7</xref>), issues such as accessibility to quality health care in underserved areas have been well documented (<xref rid="B8" ref-type="bibr">8</xref>). Studies of barriers to quality health care have mostly addressed patient-level factors such as transportation and financial issues; system-level factors affecting program implementation in underserved rural areas are seldom mentioned. Incorporating formative evaluation during DSME program conception is one way to identify and overcome some of the barriers faced during program implementation (<xref rid="B9" ref-type="bibr">9</xref>). In this study, we discuss the barriers faced during the implementation of DSME programs in medically underserved rural areas of Arkansas. We also discuss measures taken to overcome them and lessons learned from not having an evaluation plan.</p></sec><sec><title>Context</title><p>Arkansas is a rural state, with most counties in southeast Arkansas designated by the Health Services and Resources Administration as areas with a shortage of health professionals. Diabetes prevalence in Arkansas has been higher than the national average for the past decade, with 7.9% of Arkansans aged 18 years and older diagnosed with diabetes in 2002 (<xref rid="B10" ref-type="bibr">10</xref>). Costs for diabetes-related hospitalizations in Arkansas in 2001 were estimated to be $55 million.</p><p>DSME reduces diabetes complications as well as associated costs (<xref rid="B11" ref-type="bibr">11</xref>). In 2001, only 42% of Arkansans diagnosed with diabetes had ever participated in a DSME program (<xref rid="B10" ref-type="bibr">10</xref>). This low percentage may have partly resulted from DSME programs being located primarily within central and northwestern counties of the state (<xref rid="F1" ref-type="fig">Figure 1</xref>), whereas the prevalence of diabetes is disproportionately higher in southeastern counties (i.e., counties within the Mississippi Delta region) (<xref rid="F2" ref-type="fig">Figure 2</xref>). The southeastern counties are more impoverished, more rural, and have poorer health care infrastructure than other counties. Also, most of these counties have a higher proportion of racial and ethnic minorities (up to 50%), predominantly African Americans, than the state overall (16%).</p><boxed-text position="float"><fig position="float" id="F1" fig-type="diagram"><label>Figure 1</label><caption><p>Distribution of pre-existing and newly established diabetes self-management education (DSME) programs recognized by the American Diabetes Association in Arkansas, by county.</p></caption><alt-text>Map of Arkansas with program counties highlighted</alt-text><alternatives><graphic xlink:href="PCD31A15s01" position="float"/><table frame="hsides" rules="groups"><thead><tr><th scope="col" valign="top" align="left" rowspan="1" colspan="1"> <bold>Counties with Pre-existing Programs</bold> <break/> </th><th scope="col" valign="top" align="left" rowspan="1" colspan="1"> <bold>Counties with Newly Established Programs</bold> <break/> </th></tr></thead><tbody><tr><td valign="top" align="left" rowspan="1" colspan="1"> Benton<break/> Washington<break/> Sebastian<break/> Pope<break/> Garland<break/> Clark<break/> Faulkner<break/> Pulaski<break/> Jefferson<break/> Craighead<break/> Baxter<break/> Monroe<break/> </td><td valign="top" align="left" rowspan="1" colspan="1"> Columbia<break/> Quachita<break/> Union<break/> Drew<break/> Ashley<break/> Chicot<break/> Phillips<break/> Crittenden<break/> Woodruff<break/> White<break/> Cleburne<break/> </td></tr></tbody></table></alternatives><table-wrap position="anchor" id="d95e233"/></fig></boxed-text><boxed-text position="float"><fig position="float" id="F2" fig-type="diagram"><label>Figure 2</label><caption><p>Prevalence of diabetes in Arkansas, by county, 2002. Source: Behavioral Risk Factor Surveillance System.</p></caption><alt-text>Map of Arkansas indicating the prevalance of diabetes by county</alt-text><alternatives><graphic xlink:href="PCD31A15s02" position="float"/><table frame="hsides" rules="groups"><thead><tr><th scope="col" valign="top" align="left" rowspan="1" colspan="1"> <bold>County</bold> </th><th scope="col" valign="top" rowspan="1" colspan="1"> <bold>Prevalence of Diabetes, %</bold> </th></tr></thead><tbody><tr><td valign="top" rowspan="1" colspan="1">Arkansas</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Ashley</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Baxter</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Benton</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Boone</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Bradley</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Calhoun</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Carroll</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Chicot</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Clark</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Clay</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Cleburne</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Cleveland</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Columbia</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Conway</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Craighead</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Crawford</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Crittenden</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Cross</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Dallas</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Desha</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Drew</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Faulkner</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Franklin</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Fulton</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Garland</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Grant</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Greene</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Hempstead</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Hot Spring</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Howard</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Independence</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Izard</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Jackson</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Jefferson</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Johnson</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Lafayette</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Lawrence</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Lee</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Lincoln</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Little River</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Logan</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Lonoke</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Madison</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Marion</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Miller</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Mississippi</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Monroe</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Montgomery</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Nevada</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Newton</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Perry</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Phillips</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Pike</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Poinsett</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Polk</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Pope</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Prairie</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Pulaski</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Quachita</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Randolph</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Saline</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Scott</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Searcy</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Sebastian</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Sevier</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Sharp</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">St. Francis</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Stone</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Union</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Van Buren</td><td valign="top" rowspan="1" colspan="1">10.0-10.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Washington</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">White</td><td valign="top" rowspan="1" colspan="1">4.8-7.9</td></tr><tr><td valign="top" rowspan="1" colspan="1">Woodruff</td><td valign="top" rowspan="1" colspan="1">11.0-12.6</td></tr><tr><td valign="top" rowspan="1" colspan="1">Yell</td><td valign="top" rowspan="1" colspan="1">8.0-9.9</td></tr></tbody></table></alternatives><table-wrap position="anchor" id="d95e635"/></fig></boxed-text><p>The Arkansas Diabetes Prevention and Control Program (ADPCP) assembled a coalition of public and private partners to establish DSME programs in counties with a high prevalence of diabetes. Particular attention was paid to counties with no DSME programs. The ADPCP used this opportunity to help Arkansas reach the <italic>Healthy People 2010</italic> target to provide diabetes education to 60% of people in the state diagnosed with diabetes (<xref rid="B12" ref-type="bibr">12</xref>). The ADPCP also intended to assess DSME program effectiveness in an effort to improve preventive care practices and clinical outcomes.</p></sec><sec><title>Methods</title><sec><title>The ADPCP coalition</title><p>In fall 2001, the ADPCP formed a coalition consisting of public entities including the Department of Human Services, the Arkansas Foundation for Medical Care, Health Information Design, the American Diabetes Association (ADA), and the Arkansas Minority Health Commission. The coalition also included private entities (e.g., Eli Lilly and Company). The coalition's goal was to establish 12 high-quality DSME programs in underserved rural areas with a disproportionately high prevalence of diabetes. Objectives included identifying and recruiting hospitals and clinics interested in establishing DSME programs by February 2003, assisting with the resources required to establish DSME programs, assisting the DSME programs in obtaining ADA recognition by June 2003, assisting program instructors to become certified diabetes educators, and assessing program effectiveness at the end of 1 year of recruitment of all DSME programs. The coalition members identified their roles and responsibilities and worked together in making decisions for recruiting clinics, providing assistance with resources, and evaluating the intervention.</p></sec><sec><title>Recruitment of DSME sites</title><p>The coalition identified underserved areas across the state as potential sites for DSME programs and assessed the existing infrastructure in those areas. Three certified diabetes educators were hired to assist with site recruitment and program implementation. Key hospital or clinic staff members (e.g., chief executive officers, medical directors) in underserved areas were contacted by telephone to assess their interest in establishing the DSME program. Coalition members and diabetes educators then conducted a 1-day visit with key personnel at each site expressing interest. They discussed the benefits of DSME, resources that could be provided to establish a program, details of the ADA-recognition application process, and reimbursement benefits of ADA recognition. The coalition provided sample educational tools and additional information through follow-up telephone calls. Based on these solicitations, the first 12 clinics that expressed interest were recruited to establish a DSME program. The clinics signed a memorandum of agreement for their roles and responsibilities, which included patient enrollment, patient education (DSME), and data collection, in return for the resources made available to them by the coalition. The recruitment phase began in January 2002, and 12 DSME programs were established by February 2003. DSME programs identified and enrolled people with diabetes through local physicians, pharmacies, and grocery stores.</p></sec><sec><title>Resources provided to DSME sites</title><p>Each site that established a DSME program received resources, including the ADA program manual (<italic>Life with Diabetes</italic>); a copy of&#x000a0;<italic>Core Curriculum for Diabetes Education</italic> (<xref rid="B13" ref-type="bibr">13</xref>); a license and payment of monthly fees for Dia-Trac, an online data collection system (Control Diabetes Services, Plano, Tex); a glycosylated hemoglobin (HbA1c) analyzer; professional consultation provided by three certified diabetes educators; continuing education credits provided through workshops arranged by the coalition; and reimbursement of cost for ADA recognition. The diabetes educators also assisted program staff to become certified diabetes educators. Funding was made available by Eli Lilly and Company. This financial support was exclusively intended and used for public health promotion and not to promote or influence the use of any Eli Lilly product.</p></sec><sec><title>Intervention</title><p>DSME was provided to program participants by a registered nurse and a registered dietitian who followed the ADA core curriculum (<xref rid="B13" ref-type="bibr">13</xref>). Following the ADA curriculum helped ensure provision of quality diabetes education. After a 1-hour assessment of their educational needs, participants received 10 hours of diabetes education and 3 hours of medical nutrition therapy. Diabetes education was divided into three visits: an initial visit occurring shortly after the initial education assessment, a second at 6 months, and the third 1 year after program entry.</p><p>The diabetes education for each visit was provided in a group session. The curriculum addressed 10 content areas: the diabetes disease process; nutrition; physical activity; medications; monitoring and using test results; acute complications; chronic complications; goal setting and problem solving; psychosocial adjustment; and preconception care, pregnancy, and gestational diabetes (<xref rid="B13" ref-type="bibr">13</xref>). The diabetes education sessions were tailored to fit participants' needs. During each visit, educators gathered information from participants through questionnaires, including questions on demographics, self-care skills, and preventive care practices. The program staff members entered the data from the questionnaire into the Dia-trac data collection system. Control Diabetes Services was responsible for obtaining written informed consent from all patients and protecting the confidentiality of the data. The senior epidemiologist for the Arkansas Department of Health obtained the aggregate data from Control Diabetes Services with all identities removed.</p></sec></sec><sec><title>Consequences</title><sec><title>Program participants</title><p>The number of participants enrolled in the 12 DSME programs increased from 308 in February 2003 to 734 in March 2004. Of these 734 participants, 93% had type 2 diabetes. More than 75% were aged 45 years or older; 69% were white, and 30% were African American. More than 50% did not have a college degree.</p><p>Of the 319 participants due for the 1-year visit, only 20% (65) completed the 13 hours of diabetes education. Data were collected for 43 of these 65 participants on daily blood glucose monitoring, daily foot examination, and systolic and diastolic blood pressure. HbA1c level was obtained for 27 participants. There was some evidence of improvement in daily blood glucose monitoring, daily foot examination, systolic and diastolic blood pressure, and HbA1c levels (<xref rid="T1" ref-type="table">Table 1</xref>). These changes were not statistically significant, except for daily foot examination at baseline compared with 6-month follow-up (<italic>P</italic> = .03). The average HbA1c value for participants who completed the DSME program decreased from 8.15 at baseline to 7.65 at year end, a decrease of 0.5 units.</p></sec><sec><title>Barriers to program implementation</title><p>Barriers to program implementation were frequently identified through informal discussions among coalition members and DSME program staff. Patient-level barriers were identified and reported to the coalition by the DSME program staff. The coalition held a monthly teleconference with the DSME program staff to discuss progress and barriers experienced at both program and patient levels. During these calls, approaches to overcoming some of the barriers were proposed. The coalition members worked on applying solutions to the program implementation barriers. DSME program staff members worked to address patient-level barriers within their own clinics. There were anticipated and unanticipated barriers to implementation at both the patient and program levels. <xref rid="T2" ref-type="table">Table 2</xref> provides a summary of strategies used to minimize or eliminate anticipated and unanticipated barriers.</p><p> <bold>Anticipated barriers</bold> </p><p>At the program level, anticipated barriers centered on staffing and reimbursement for DSME. To obtain ADA recognition, the program needed at least one registered nurse and one registered dietician. Arkansas is a predominantly rural state, and more than half (58%) of its population lives in areas having a shortage of health professionals. Recruiting health professionals, particularly registered dietitians, was a challenge. Some DSME programs shared a registered dietitian to fulfill the ADA requirement.</p><p>Reimbursement constraints took a number of forms. Insurance reimbursement only took place after ADA recognition of a DSME program, which did not occur until 6 months into the program. Although the coalition was not able to provide financial assistance during this period, the resources provided to DSME programs helped to overcome this barrier. A related barrier was that although there was no formal pre-existing DSME program in the participating counties, most counties included diabetes education as a subcomponent of their broader health care services. Rural health centers were not always reimbursed because diabetes education was considered a service already available. These rural health centers perceived the DSME program as contributing beyond their current services, so they applied for grants to cover program costs. This was one approach used by DSME programs to secure additional funds.</p><p>Anticipated patient-level barriers included transportation, literacy, and reimbursement. Patients with no means of transportation needed to travel long distances to reach a DSME program site. To address this barrier, some DSME programs provided transportation by hospital vehicles; others coordinated transportation through local churches. Some patients had very little formal education, which presented a substantial barrier to understanding key DSME messages. In response, program staff members assisted patients by reading the materials to them. Medicaid members were not reimbursed for diabetes education. This barrier was anticipated, but the coalition was not able to overcome it. Because Medicaid members were asked to pay for DSME at their own expense, many dropped out of the program.</p><p> <bold>Unanticipated barriers</bold> </p><p>Unanticipated barriers included a lack of consistent data collection processes among DSME sites and participant retention. DSME programs were asked to enter participant information into the data collection system regularly, but this was not consistently done. Some program staff members said they lacked the resources (people or time) for data entry. Because the coalition could not assist with data entry and staff members did not understand the significance of gaps in data collection, the problem remained unsolved. Motivation to collect data was further decreased once sites received ADA recognition.</p><p>Participant retention posed a challenge partly as a result of environmental factors associated with rural health settings. The coalition's intent was to establish DSME programs in underserved areas of high diabetes prevalence where DSME programs would not have been available otherwise. However, the participation rate fell to 34% at the 6-month visit and 20% at the end-of-year visit. Some, but not all, program staff members reminded participants of their impending visits by postcard or telephone call.</p></sec><sec><title>Evaluation results and lessons learned</title><p>The ADPCP hired an epidemiologist during fall 2002 after the program had formally begun. The epidemiologist engaged key stakeholders and DSME program staff in spring 2003 to plan and implement a program evaluation. Although integrating evaluation early in the program planning process can be very helpful, this is often not done for fear that evaluation will be seen as punitive, exclusionary, and adversarial (<xref rid="B10" ref-type="bibr">10</xref>). This was true in the present study. There was also no logic model developed by the coalition, although process measures were put in place to capture program implementation at each site. Information gleaned from these measures will be used to shape future DSME programs and to develop a DSME program logic model that may foster a clearer understanding of the barriers faced by these programs in rural Arkansas and their relationship to program outcomes.</p></sec></sec><sec><title>Interpretation</title><sec><title>Progress toward program goals and objectives</title><p>Key evaluation issues chosen to assess DSME program outcomes included whether 1) the coalition was able to establish 12 DSME programs in rural underserved counties in Arkansas, 2) the DSME program fostered progress toward achieving the <italic>Healthy People 2010</italic> target to provide diabetes education to 60% of people with diabetes, and 3) the DSME program fostered preventive care practices.</p><p>The coalition met its goal of establishing the 12 DSME programs in underserved counties. <xref rid="F1" ref-type="fig">Figure 1</xref> shows the location of pre-existing and newly established DSME programs. Of the 12 DSME programs, 11 met the minimum participation and 6-month follow-up requirements and obtained ADA recognition. By the 6-month follow-up, DSME programs were required to have 1) a minimum of 20 patients enrolled and 2) a continuous quality improvement (CQI) measure for patients. All 12 DSME programs collected HbA1c results as a CQI measure for ADA recognition. One DSME program did not obtain ADA recognition during the time frame because it had fewer than 20 patients enrolled in the program. One DSME program staff member became a certified diabetes educator, and two staff members from other DSME programs are preparing to take the certification examination.</p><p>The number of people receiving diabetes education in Arkansas more than doubled from February 2003 to March 2004. This increase highlights success in addressing diabetes education among the most hard-to-reach populations in the state. Although recruitment efforts for the DSME program had some success, the lack of a unified effort to retain participants, along with reimbursement-related barriers, may have contributed to high rates of attrition.</p><p>Key stakeholders, including coalition members and DSME program staff, were given the evaluation results. The coalition understood the weaknesses of the follow-up and realized that evaluation should have been incorporated early in program planning. If an evaluation planning process had been incorporated into the early coalition meetings, it may have led to the identification of key barriers and resulted in changes to program content, resources, and timeline. These changes may have, in turn, increased program effectiveness and usefulness of evaluation findings. Another potential limitation is that clinics self-selected to establish the DSME programs. This may limit the generalizability of findings because participating clinics may not be representative of clinics in rural Arkansas.</p></sec><sec><title>New DSME sites and program improvements</title><p>The ADPCP and its coalition members plan to implement six more DSME program sites in underserved rural Arkansas counties by spring 2006. For that purpose, coalition and DSME site staffs will incorporate evaluation planning before the new sites are fully implemented. Developing formative and impact evaluation plans prior to program implementation helps to ensure an evaluation provides useful information. Impact evaluations are used to determine the degree to which a program has led to desired changes and may also have implications for future programs. The coalition will consider which evaluation data are needed from each site to enable a comprehensive assessment of program goals for utility, feasibility, propriety, and accuracy (<xref rid="B14" ref-type="bibr">14</xref>).</p><p>Program and evaluation efforts for the six new sites will include a review of the evaluation findings of other similar DSME programs to determine how best to address attrition, data consistency, and other key barriers. For example, studies examined attrition rates for diabetes education programs that included a follow-up component (<xref rid="B6" ref-type="bibr">6</xref>); attrition rates in these studies ranged from 0% to 79%. These studies showed that attrition may be due to participant, researcher, study, or environmental factors. Attrition rates were found to decrease when participant factors such as motivation, values, and beliefs are encouraged and certain program outreach methods are used (<xref rid="B15" ref-type="bibr">15</xref>).</p><p>Establishing the 12 DSME programs in underserved rural areas of Arkansas provided important lessons about the importance of an evaluation plan. The authors view the development of an evaluation plan as a necessary and valuable initial step toward better addressing the educational needs of people diagnosed with diabetes in rural Arkansas.</p><p>Even in the face of serious resource challenges, the coalition attempted to address both anticipated and unanticipated barriers. Findings from this program evaluation will affect the establishment of future DSME sites in rural Arkansas. Particular attention will be given to an evaluation plan that embraces fiscal, human, and environmental factors that affect program planning, implementation, and sustainability. Findings from this evaluation may prove useful to others working in medically underserved rural communities throughout the United States.</p></sec></sec>