Datasets:

alon-albalak

/

licensed_pubmed

like 0

PMC3052194

Introduction ============ Pneumoperitoneum is the result of a gastrointestinal (GI) tract perforation in more than 90% of cases \[[@B1]\]. Perforation of the stomach or duodenum caused by peptic ulcer is considered the most common cause of pneumoperitoneum. Pneumoperitoneum can also be the result of a diverticular rupture or of an abdominal trauma \[[@B1]\]. It commonly presents with signs and symptoms of peritonitis, and subphrenic free gas in an upright chest radiograph is the most common radiologic finding. In most cases, pneumoperitoneum requires urgent surgical exploration and intervention \[[@B1]\]. However, sometimes pneumoperitoneum not associated with a perforated viscus can occur; this is called spontaneous pneumoperitoneum (SP) or \"non-surgical\" pneumoperitoneum. SP is associated with intrathoracic, intraabdominal, gynecologic, iatrogenic or other miscellaneous causes \[[@B1]\]. Although it is not usually complicated with peritonitis, SP is characterized by a benign course and can be managed conservatively \[[@B1]-[@B4]\]. Idiopathic SP is an even more rare condition for which no clear etiology has been established because both perforation of an intraabdominal viscus and other known causes of free intraperitoneal gas have been excluded \[[@B1],[@B5]-[@B7]\]. Idiopathic pneumoperitoneum is usually diagnosed after negative laparotomy results. SP poses significant management dilemmas for surgeons, especially when signs of peritonitis are absent or when the cause is unknown before laparotomy. Case presentation ================= A 69-year-old Greek female patient presented at our emergency department (ED) with a two-hour history of abdominal pain and vomiting. Her medical history was unremarkable except for previous cholecystectomy and appendectomy. The patient did not take any medications, and she was not a smoker or an alcohol consumer. She looked ill with a blood pressure of 130/85 mm/Hg, a pulse rate of 90 beats/min, respirations of 25 breaths/min and a temperature of 38.5°C. A thorough physical examination revealed diffuse abdominal tenderness on deep palpation without any other signs of peritonitis. The laboratory examination was unremarkable except for polymorphonuclear leucocytosis (white blood cell \[WBC\] count, 15 × 10^3^/μL; neutrophils, 86%) and an elevated C-reactive protein (14 mg/dL; reference range, 0-5). An upright chest radiograph demonstrated free subdiaphragmatic air bilaterally (Figure [1](#F1){ref-type="fig"}), which seemed to be increasing during air insufflation in the stomach via a nasogastric tube (Figure [2](#F2){ref-type="fig"}). Abdominal ultrasound examination was unremarkable. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Upright posteroanterior chest radiograph**. There is free subdiaphragmatic air bilaterally that is more clearly noted on the right side (*white arrows*). :::  ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Upright posteroanterior chest radiograph after insufflating air into the stomach**. The free subdiaphragmatic air has slightly increased in size bilaterally compared with Figure 1 (*white arrows*). :::  ::: An emergency laparotomy was performed for a suspected perforation in the upper GI tract. A few adhesions caused by previous cholecystectomy and appendicectomy were observed without any signs of peritoneal irritation or peritoneal fluid. The stomach and duodenum were fully mobilized, and the lesser sac was explored, but no evidence of perforation was found in the distal esophagus, stomach or duodenum. The small bowel and colon were also examined, but no leakage was observed. Subsequently, dilution of methylene blue in normal saline was instilled into the stomach through the nasogastric tube, but no obvious leakage was noted. Afterward, the abdominal cavity was filled with 2000 cc of normal saline, and air was again infused through the nasogastric tube into the stomach, but no air leakage from the upper GI tract was noted. Finally, because no cause of the pneumoperitoneum had been found, the operation was completed by placing a double-lumen drain. The postoperative course was uneventful, and the patient showed a significant and prompt recovery. The subdiaphragmatic air disappeared six days postoperatively (Figure [3](#F3){ref-type="fig"}). The patient was discharged home on the seventh postoperative day. One month later, esophagogastroduodenoscopy, colonoscopy and abdominal computed tomography (CT) were performed, but no pathology was detected. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### **Upright posteroanterior chest radiograph just before the patient\'s discharge**. No subdiaphragmatic free air is noted bilaterally. :::  ::: Discussion ========== SP is associated with intrathoracic, intraabdominal, gynecologic, iatrogenic and other miscellaneous causes \[[@B1],[@B2]\]. SP has been attributed to several thoracic causes, such as traumas (including barotraumas), pneumothorax and bronchoperitoneal fistulas \[[@B1]\]. SP can be accompanied by pneumomediastinum or pneumopericardium, especially in patients who are on mechanical aspiration and positive end-expiration pressure \[[@B1]\]. In extremely rare cases, scuba diving and pulmonary sepsis can cause SP. Pneumatosis cystoides intestinalis is the most common abdominal cause of nonsurgical pneumoperitoneum \[[@B1]\]. Emphysematous cholecystitis, spontaneous bacterial peritonitis, ruptured hepatic abscess and perforated pyometra in women are rare causes of SP \[[@B1]\]. In women, pneumoperitoneum after rough sexual intercourse or after Jacuzzi usage has also been reported because the air can also be transmitted to the peritoneal cavity through the vagina and saplings \[[@B1]\]. Laparoscopic or endoscopic procedures (colonoscopy) may cause iatrogenic SP \[[@B1]\]. The cause of pneumoperitoneum and the clinical signs determine its mode of treatment, surgical or not. When signs and symptoms of \"acute abdomen\" are present, surgical management is mandatory, but in cases of nonsurgical pneumoperitoneum with mild symptoms and without any signs of peritonitis, conservative treatment is indicated \[[@B2]\]. A detailed history and physical examination can be very helpful in distinguishing surgical from nonsurgical pneumoperitoneum, thus avoiding unnecessary laparotomies \[[@B2]\]. Moreover, radiographic imaging before and after air insufflation into the gastric lumen via a nasogastric tube (pneumogastrogram) is an easy and safe method, which can enhance or confirm the diagnosis of a visceral perforation in the upper GI tract \[[@B8]\]. Plain chest or abdominal radiography is the most common imaging examination for the diagnosis of even very small amounts of intraperitoneal free air in the ED setting \[[@B9]\], but abdominal CT is a more sensitive method of diagnosing pneumoperitoneum and identifying the cause of \"acute abdomen\" \[[@B10],[@B11]\]. Moreover, modern technology with multidetector CT is highly accurate for predicting the site of GI tract perforations \[[@B12],[@B13]\]. It has been proposed that in some cases with idiopathic pneumoperitoneum, a subclinical small visceral perforation may have occurred, permitting only the leakage of air and not of bowel contents \[[@B1]\]. Finally, in other cases, other unknown factors may be the cause of idiopathic pneumoperitoneum \[[@B1]\]. We report the case of a patient who underwent an urgent but nondiagnostic exploratory laparotomy, although she had compelling evidence for a surgical pneumoperitoneum. A minority of pneumoperitoneum cases are considered idiopathic, but many of them undergo surgical exploration \[[@B2]\]. van Gelder *et al.*\[[@B5]\] reported six patients with pneumoperitoneum and clinical signs of acute abdomen who underwent exploratory laparotomy, which did not reveal any intraabdominal pathology. Chandler *et a*l. \[[@B14]\] reported a laparotomy rate of 28% on nonsurgical pneumoperitoneum. In a review, Mularski *et al.*\[[@B15]\] found 196 reported cases of nonsurgical pneumoperitoneum, of which 45 underwent surgical exploration without evidence of perforated viscus. Furthermore, Mularski *et al.*\[[@B15]\] reported that 11 of 36 (31%) miscellaneous or idiopathic cases of nonsurgical PP underwent surgical exploration. Currently, laparoscopic exploration instead of laparotomy can be the operation of choice in cases of pneumoperitoneum because it can both determine and treat the cause, offering all the advantages of minimally invasive surgery. Conclusion ========== A thorough history and physical examination combined with the appropriate laboratory tests and radiologic techniques are useful tools in identifying patients with nonsurgical pneumoperitoneum and avoiding unnecessary operations. List of abbreviations ===================== CT: computed tomography; ED: emergency department; GI: gastrointestinal; SP: spontaneous pneumoperitoneum; WBC: white blood cell. Consent ======= Written informed consent was obtained from the patient for the publication of this case report and the accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= MP participated in the patient\'s treatment, had the idea for the case report, contributed to the first draft and performed all of the revisions. PZ collected the patient\'s data, participated in the first draft and performed all of the revisions. AO participated in the imaging diagnosis of the case and contributed to the writing of the paper. MK participated in the patient\'s treatment and contributed to the writing of the paper. GK contributed to the writing of the paper. CS participated in the patient\'s treatment and participated in the final revision. All authors read and approved the final manuscript.

PubMed Central

2024-06-05T04:04:19.021364

2011-2-27

PMC3052195

Background ========== Endometriosis is a common chronic gynecological disease characterized by the presence of endometrial gland and stroma outside the uterine cavity, affecting approximately 10% of reproductive age women \[[@B1],[@B2]\]. The common clinical symptoms include pelvic pain, heavy menstrual bleeding, pelvic adhesion, bloating and fatigue. Notably, the prevalence of endometriosis is 0.5% to 5% in fertile women and 25% to 40% in infertile women \[[@B3]\], suggesting infertility as one possible consequence of endometriosis. To date, the implantation theory is widely accepted, stating that endometrial tissues pass through the fallopian tube, then attach and grow on pelvic tissue. However, this hypothesis cannot explain the existence of endometriosis outside the pelvis or how endometriosis progresses and invades other tissues. Additional factors such as genetic or immune differences have been suggested as possible contributors to trigger the formation of endometriosis \[[@B4]-[@B6]\]. Family history and genome-wide linkage studies also support genetic predisposition during the development of endometriosis \[[@B7]-[@B10]\]. These studies provide molecular evidence demonstrating endometriosis as a genetic disease, and it is desirable to explore more genetic variations associated with endometriosis. Similarly to malignant diseases, extensive growth of endometrial cells on the peritoneal surface and invasion of the pelvic organ are very common during the development of endometriosis. This process is frequently associated with several mechanisms involved in angiogenesis and cellular adhesion. In fact, women who have endometriosis appear to be more at risk of developing several different kinds of ovarian cancers \[[@B8],[@B11]-[@B13]\]. An epidemiological study showed that the prevalence rates of endometriosis in patients with endometrioid and clear cell ovarian carcinoma are 19 and 35.9, respectively \[[@B14]\]. These findings suggest that endometriosis and certain types of ovarian cancer may share several common genetic alterations during pathogenesis. Genes that regulate cell mobility and invasion in ovarian cancers are therefore possible candidates for playing roles in endometriosis. Mucins are high-molecular-weight glycoproteins with the function of protecting and lubricating the epithelial surface of respiratory, gastrointestinal and reproductive tracts \[[@B15]\]. Among the mucin proteins, mucin 4 (MUC4) and mucin 1 (MUC1) are the major ones expressed in the endometrial epithelium \[[@B16],[@B17]\]. In cancer studies, these two mucins have been shown to be aberrantly expressed in various malignancies and have been validated as novel targets for cancer diagnosis and therapy \[[@B18]-[@B20]\]. Distinct from MUC1, the extracellular domain of MUC4 can interact with human epidermal growth factor receptor 2 (HER2) on the cell surface and modulate downstream cell growth signaling by stabilizing and/or enhancing the activity of cell growth receptor complexes \[[@B18],[@B21],[@B22]\]. Consequently, changes of cytoarchitectures and cellular signaling may lead to the increase of cell mobility and tumor cell invasion. The above findings provide us with clues to hypothesize that genetic variations in the extracellular domain of MUC4, especially those resulting in amino acid substitutions, may play roles involved in the development of endometriosis. With endometriosis as a possible cause of infertility in women, we also would like to study the association of *MUC4*single-nucleotide polymorphisms (SNPs) with susceptibility to endometriosis-related infertility. Methods ======= Study population ---------------- A total of 140 individuals who underwent surgery for benign diseases and pathology-proven endometriosis were identified at China Medical University Hospital from 1998 to 2008 and were enrolled in this study. In general, these patients were diagnosed with ovarian cysts on the basis of sonography and had several clinical symptoms related to endometriosis, including dysmenorrhea, lower abdominal pain, infertility or abnormal menstruation. Study patients who failed to have pathology-proven endometriosis were excluded from this study. For the control group, blood samples of 150 healthy women were selected from a pool of individuals who received regular heath checkups at the same hospital and were identified as normal on the basis of the examinations conducted. A total of 142 controls were frequency-matched on the basis of age profile with the study patients (Additional file [1](#S1){ref-type="supplementary-material"}, Table S1). Controls who showed one of the endometriosis-associated symptoms, even though the results of their health checkups were normal, were excluded from this study. This study was approved by the institutional review board at China Medical University, with informed consent obtained from each patient. Clinical stages and association study ------------------------------------- Clinical information on patients was collected from clinical notes, including clinical stage, lesion size, location, drug treatment and fertility (Additional file [1](#S1){ref-type="supplementary-material"}, Table S1). The definition of endometriosis staging was based on the classification of the American Society for Reproductive Medicine: Stage 1, minimal; Stage 2, mild; Stage 3, moderate; and Stage 4, severe \[[@B23]\]. Genomic DNA extraction and genotyping of SNPs in *MUC4* ------------------------------------------------------- Genomic DNA was extracted from peripheral blood leukocytes according to standard protocols (Genomic DNA kit; Qiagen, Valencia, CA, USA). DNA fragments containing the target SNP sites were amplified by polymerase chain reaction (PCR) assay using the TaqMan SNP Genotyping Assay System from Applied Biosystems, Inc. (Carlsbad, CA, USA). The probe search and design are available on the Applied Biosystems, Inc. website \[[@B24]\]. Additional file [1](#S1){ref-type="supplementary-material"}, Table S2, lists probe identifications for the six SNPs tested. PCR amplification conditions consisted of initial denaturation at 95°C for 5 minutes followed by 40 cycles at 95°C for 10 seconds, 56°C for 10 seconds and 72°C for 20 seconds, with one additional cycle at 72°C for 5 minutes. Genetic variations were detected by reading the fluorescence signals of PCR products. A positive signal indicates a perfect match between the probe and the tested DNA, thus identifying the allele types. Ten percent of study participants were randomly chosen and genotyped in duplicate to confirm the concordance of the genotyping results. In our study, the call rates for these SNP probes were above 94% (Additional file [1](#S1){ref-type="supplementary-material"}, Table S3). Statistical analysis -------------------- The allelic frequency and genotype frequency distributions for the six polymorphisms of patients with endometriosis and the controls were performed by χ^2^analysis using SPSS software (version 10.0; SPSS Inc., Chicago, IL, USA). An unordered, 2 *df*two-sided test was used for the statistical analyses of our genotyping results. *P*\< 0.05 was considered statistically significant. Allelic and genotypic frequencies are expressed as percentages of total alleles and genotypes. Odds ratios (ORs) were calculated from allelic and genotype frequencies with 95% confident intervals (95% CIs). The major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. Adherence to the Hardy-Weinberg equilibrium (HWE) constant was tested using a χ^2^test with 1 *df*. To study the association of the six SNPs with clinical stages and reproductive ability, Fisher\'s exact tests instead of χ^2^tests were used because of the small number of participants tested. The haplotypes of each individual were determined using the Bayesian statistical method available in the free download software program PHASE 2.1 \[[@B25]\]. This approach incorporates *a priori*expectations of haplotypic structure based on population genetics and coalescence theory. Lewontin\'s D\' (\|D\'\|) and the linkage disequilibrium (LD) coefficient *r*^2^were determined between selected pairs of biallelic loci \[[@B26]\]. Haploview version 3.2 software (Whitehead Institute for Biomedical Research, Cambridge, MA, USA) was used to examine the structure of the LD block \[[@B27]\]. This program uses two-marker expectation maximization to estimate the maximum likelihood values of the four gamete frequencies from which the D\' and log of odds (LOD) values are derived. The genetic effects of the inferred haplotypes were analyzed in the same way as the analysis of polymorphisms. The reported haplotype percentages are estimated on the basis of allele frequencies and LD. The *P*values are based on a comparison of a given haplotype with all other haplotypes combined. Functional analyses and secondary structure predictions of MUC4 protein ----------------------------------------------------------------------- Functional characterization and annotation of MUC4 were performed by aligning the sequence with functional motifs and/or signatures in the PROSITE protein domain database \[[@B28]\]. To predict the secondary structure of the MUC4 sequence, the Chou and Fasman method was used \[[@B29]\]. An improved method was applied to increase the accuracy of the predictions by locating nucleation regions with refined wavelet transformation technology and by calculating propensity factors with larger data sets \[[@B30]\]. The program gives the propensity of each residue to be a part of an α-helix, a β-strand or a loop. We considered propensities *P*~α~\> 1.03 as significant for helices and propensities *P*~β~\> 1.05 as significant for strands. Predicted regions with fewer than four contiguous residues were not considered secondary structure units. For a region with both helix and strand tendencies, the secondary structure conformed with higher propensity: *P*~α~\>*P*~β~or *P*~β~\>*P*~α~is predicted. To plot hydrophobicity and surface probability, the Kyte and Doolittle method \[[@B31]\] and Emini *et al*. surface accessibility prediction (SAP) \[[@B32]\] were used, respectively. We slid a window along the MUC4 sequence to assign a \"hydrophobicity\" or \"surface probability\" value to each amino acid. The values were summed in the window, and the results were plotted. Results ======= *MUC4*gene polymorphisms and endometriosis ------------------------------------------ Six SNPs in the extracellular domain of the *MUC4*gene with a frequency greater than 20% in Chinese Han Beijing were selected from International HapMap Project databank \[[@B33]\] (Additional file [1](#S1){ref-type="supplementary-material"}, Table S2). Genotypic analyses (Figure [1](#F1){ref-type="fig"}) indicated rs882605 as a unique SNP with a higher frequency of the TG genotype in patients than in controls (*P*= 0.04, OR = 1.97, 95% CI, 1.17 to 3.32) (Table [1](#T1){ref-type="table"}), while allele type analyses of these SNPs showed no statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. To confirm the genetic impact of SNPs on endometriosis, the top two high-risk alleles at rs882605 and rs1104760 were selected for haplotype analyses. Significantly, the frequency of haplotype T-T was found to be higher in patients than in controls (*P*= 0.0353) (Table [2](#T2){ref-type="table"}) (Additional file [2](#S2){ref-type="supplementary-material"}, Figure S1), suggesting the association of *MUC4*SNPs with endometriosis development. The genotyping results were confirmed in duplicate, and the concordance of duplicates was 97.6%. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Allelic discrimination plots of the six tested single-nucleotide polymorphisms (SNPs) in the mucin 4 (*MUC4*) gene**. The DNA samples from patients and controls were genotyped by using the TaqMan SNP Genotyping Assay System. The major (also the wild-type) alleles were detected by 6-carboxyfluorescein (FAM)-labeled probes (blue), and the minor alleles were detected by 2\'-chloro-7\'-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC)-labeled probes. The genotyping results of the six SNPs in the *MUC4*gene are presented as allelic discrimination plots. Of note, the intensity of FAM signals tended to be similar among samples in our assays, thus the dots for a wild-type genotype overlapped each other. \"X\" indicates the participant who failed to be genotyped. :::  ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Genotype and allele distributions of the six SNPs in the *MUC4*gene in Taiwanese endometriosis patients and controls^a^ ::: SNPs Genotype/allele No. (%)of patients HWE No. (%)of controls HWE *P*value OR (95% CI) ----------- ----------------- -------------------- -------- -------------------- ----- ---------- ------------- ------ ------ ------------------- rs882605 TT 9 (7.0) 0.62 10 (7.2) 0.07 0.04 1.13 (0.43 to 3.00) TG 54 (42.2) 40 (28.8) 1.97 (1.17 to 3.32)^b^ GG 65 (50.8) 89 (64.0) 1.00 Reference T 72 (28.1) 60 (21.6) 0.08 1.42 (0.96 to 2.11) G 184 (71.9) 218 (78.4) 1.00 Reference rs1104760 CC 7 (5.4) 0.94 9 (6.3) 0.09 0.30 0.96 (0.33 to 2.82) CT 47 (36.2) 39 (27.5) 1.49 (0.89 to 2.51) TT 76 (58.5) 94 (66.2) 1.00 Reference C 61 (23.5) 57 (20.1) 0.34 1.22 (0.81 to 1.84) T 199 (76.5) 227 (79.9) 1.00 Reference rs2246901 GG 8 (6.0) 0.96 10 (7.1) 0.15 0.56 0.91 (0.33 to 2.53) TG 49 (36.6) 43 (30.5) 1.30 (0.78 to 2.17) TT 77 (57.5) 88 (62.4) 1.00 reference G 65 (24.3) 63 (22.3) 0.60 1.11 (0.75 to 1.65) T 203 (75.7) 219 (77.7) 1.00 Reference rs2258447 AA 9 (6.4) 0.37 10 (7.0) 0.07 0.70 1.02 (0.38 to 2.77) AG 44 (32.1) 40 (28.2) 1.25 (0.74 to 2.11) GG 81 (61.4) 92 (64.8) 1.00 Reference A 62 (22.5) 60 (21.1) 0.57 1.12 (0.75 to 1.68) G 206 (77.5) 224 (78.9) 1.00 Reference rs2291652 CC 11 (8.7) 0.92 11 (8.4) 0.26 0.52 1.17 (0.46 to 2.96) CT 52 (40.9) 45 (34.4) 1.35 (0.81 to 2.28) TT 64 (50.4) 75 (57.3) 1.00 Reference C 74 (29.1) 67 (25.6) 0.36 1.20 (0.81 to 1.76) T 180 (70.9) 195 (74.4) 1.00 Reference rs2688513 CC 8 (6.0) 0.62 10 (7.0) 0.09 0.66 0.91 (0.33 to 2.53) TC 45 (33.8) 41 (28.9) 1.25 (0.74 to 2.10) TT 80 (60.2) 91 (64.1) 1.00 Reference C 61 (22.9) 61 (21.5) 0.68 1.09 (0.73 to 1.63) T 205 (77.1) 223 (78.5) 1.00 Reference ^a^*MUC4*, mucin 4 gene; SNP, single-nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval; HWE, *P*values of deviation from the Hardy-Weinberg equilibrium constant. Allelic frequencies were determined by χ^2^test using 2 × 2 contingency tables. Genotype frequencies were determined by χ^2^test using 2 × 3 contingency tables. ^b^*P*\< 0.05 was considered statistically significant. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Haplotype frequencies of *MUC4*polymorphisms in endometriosis patients and controls^a^ ::: rs1104760/rs882605 Patients, % Controls, % *P*value -------------------- ------------- ------------- ----------- TG 73.2 78.8 0.1202 CT 22.9 20.1 0.4134 TT 3.9 1.1 0.0353^b^ ^a^*MUC4*, mucin 4 gene. The reported haplotype percents are estimated percents based on allele frequencies and the linkage disequilibrium. The *P*values are based on a comparison of a given haplotype with all other haplotypes combined. ^b^*P*\< 0.05 was considered statistically significant. ::: Association of *MUC4*gene polymorphisms and stages -------------------------------------------------- We next asked whether *MUC4*genetic variations could possibly associate with clinical stages. Patients were divided into two groups: the mild stage group with patients at stages 1 or 2 and the advanced group with patients at stages 3 or 4. Strikingly, genotype analyses revealed strong associations of CC type at rs2688513 (*P*= 0.04) and GG type at rs2246901 (*P*= 0.03), with more advanced endometriosis at stage 3 or 4 (Table [3](#T3){ref-type="table"}). Dominant effects were found for other genetic variations at rs1104760 (CC + CT versus TT) and rs2258447 (AA + AG versus GG) during endometriosis progression. Allele type analyses suggested C allele at rs1104760, C allele at rs2688513, G allele at rs2246901 and A allele at rs2258447 as risk factors that correlated with more severe endometriosis. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Genotype and allele distributions of SNPs in the *MUC4*gene in endometriosis patients at different clinical stages^a^ ::: SNP Genotype/allele **No. (%) at mild stage**^**b**^ **No. (%) at severe stage**^**c**^ *P*value ----------- ----------------- ---------------------------------- ------------------------------------ ---------- -------- --------- rs882605 TT 0 (0.0) 7 (8.0) 0.25 TG 2 (25.0) 41 (46.6) GG 6 (75.0) 40 (45.5) TT+TG 2 (25.0) 48 (54.5) 0.11 GG 6 (75.0) 40 (45.5) T 2 (12.5) 55 (31.3) 0.12 G 14 (87.5) 121 (68.8) rs1104760 CC 0 (0.0) 5 (5.8) 0.05 CT 0 (0.0) 36 (41.9) TT 7 (100.0) 45 (52.3) CC+CT 0 (0.0) 41 (47.7) 0.0^d^ TT 7 (100.0) 45 (52.3) C 0 (0.0) 46 (26.7) 0.03^d^ T 14 (100.0) 126 (73.3) rs2688513 CC 0 (0.0) 6 (6.9) 0.04^d^ CT 0 (0.0) 34 (39.1) TT 8 (100.0) 47 (54.0) CC+CT 0 (0.0) 40 (46.0) 0.01^d^ TT 8 (100.0) 47 (54.0) C 0 (0.0) 46 (26.4) 0.02^d^ T 16 (100.0) 128 (73.6) rs2246901 GG 0 (0.0) 6 (6.8) 0.03^d^ TG 0 (0.0) 36 (40.9) TT 8 (100.0) 46 (52.3) GG+TG 0 (0.0) 42 (47.7) 0.01^d^ TT 8 (100.0) 46 (52.3) G 0 (0.0) 48 (27.3) 0.02^d^ T 16 (100.0) 128 (72.7) rs2258447 AA 0 (0.0) 7 (7.9) 0.08 AG 0 (0.0) 32 (36.0) GG 7 (100.0) 50 (56.2) AA+AG 0 (0.0) 39 (43.8) 0.02^d^ GG 7 (100.0) 50 (56.2) A 0 (0.0) 46 (25.8) 0.03^d^ G 14 (100.0) 132 (74.2) rs2291652 CC 0 (0.0) 7 (8.3) 0.61 CT 2 (33.3) 36 (42.9) TT 4 (66.7) 41 (48.8) CC+CT 2 (33.3) 43 (51.2) 0.40 TT 4 (66.7) 41 (48.8) C 2 (16.7) 50 (29.8) 0.33 T 10 (83.3) 118 (70.2) ^a^SNP, single-nucleotide polymorphism; *MUC4*, mucin 4 gene. Allelic frequencies were determined by Fisher\'s exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher\'s exact test using 2 × 3 contingency tables. ^b^Mild stage, patients at clinical stage1 or stage 2; ^c^Severe stage, patients at clinical stage 3 or stage 4; ^d^*P*\< 0.05 was considered statistically significant. ::: *MUC4*gene polymorphisms and infertility ---------------------------------------- Since endometriosis has been suspected as one potent factor leading to infertility in women \[[@B3]\], we also studied the possible linkage between *MUC4*SNPs and infertility. Though no significant difference was found in our genotype association study, our data indicated T allele at rs882605 as a protective factor that associated with reduced frequency of infertility in patients with endometriosis (Table [4](#T4){ref-type="table"}). Two other alleles, C at rs2688513 and G at rs2246901, showed similar protective effects, but the data did not reach statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. In our haplotype analyses, we thus sought to ascertain the impact of the genetic combination of these top three protective alleles. Table [5](#T5){ref-type="table"} indicates that patients with haplotype T-C-G did show a lower frequency of infertility, although the results did not show a statistically significant difference (*P*= 0.099). By contrast, haplotype G-T-T showed a strong association with infertility in patients (*P*= 0.012) (Table [5](#T5){ref-type="table"}) (Additional file [2](#S2){ref-type="supplementary-material"}, Figure S2) and could be used as a risk indicator for patients at higher risk of developing severe complications such as infertility. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Genotype and allele distributions of the six SNPs in the *MUC4*gene in endometriosis patients with different reproductive ability^a^ ::: SNP Genotype/allele No. (%) infertility No. (%) noninfertility *P*value OR (95% CI) ----------- ----------------- --------------------- ------------------------ ---------- ------------- --------- ------ ---------------- rs882605 TT 0 (0.0) 8 (8.7) 0.07 \- \- TG 5 (26.3) 42 (45.7) 0.36 (0.12 to 1.08) GG 14 (73.7) 42 (45.7) 1.00 Reference T 5 (13.2) 58 (31.5) 0.03^b^ 0.33 (0.12 to 0.89) G 33 (86.8) 126 (68.5) 1.00 Reference rs1104760 CC 0 (0.0) 6 (6.3) 0.41 \- \- CT 4 (25.0) 37 (38.5) 0.48 (0.14 to 1.60) TT 12 (75.0) 53 (55.2) 1.00 Reference C 4 (12.5) 49 (25.5) 0.12 0.42 (0.14 to 1.25) T 28 (87.5) 143 (74.5) 1.00 Reference rs2688513 CC 0 (0.0) 7 (7.4) 0.32 \- \- TC 5 (25.0) 34 (36.2) 0.52 (0.17 to 1.56) TT 15 (75.0) 53 (56.4) 1.00 Reference C 5 (12.5) 48 (25.5) 0.09 0.42 (0.15 to 1.12) T 35 (87.5) 140 (74.5) 1.00 Reference rs2246901 GG 0 (0.0) 7 (7.4) 0.25 \- \- TG 5 (25.0) 37 (38.9) 0.46 (0.15 to 1.38) TT 15 (75.0) 51 (53.7) 1.00 Reference G 5 (12.5) 51 (26.8) 0.06 0.39 (0.14 to 1.05) T 35 (87.5) 139 (73.2) 1.00 Reference rs2258447 AA 1 (5.0) 7 (7.3) 0.41 0.53 (0.05 to 5.53) AG 4 (20.0) 33 (34.4) 0.45 (0.14 to 1.48) GG 15 (75.0) 56 (58.3) 1.00 Reference A 6 (15.0) 47 (24.5) 0.22 0.54 (0.22 to 1.38) G 34 (85.0) 145 (75.5) 1.00 Reference rs2291652 CC 0 (0.0) 7 (7.8) 0.29 \- \- CT 7 (35.0) 40 (44.4) 0.58 (0.21 to 1.60) TT 13 (65.0) 43 (47.8) 1.00 Reference C 7 (17.5) 54 (30.0) 0.12 0.49 (0.21 to 1.19) T 33 (82.5) 126 (70.0) 1.00 Reference ^a^SNP, single-nucleotide polymorphism; *MUC4*, mucin 4 gene; OR, odds ratio; 95% CI, 95% confidence interval. Allelic frequencies were determined by Fisher\'s exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher\'s exact test using 2 × 3 contingency tables. ^b^*P*\< 0.05 was considered statistically significant. ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Haplotype frequencies of *MUC4*polymorphisms in endometriosis patients with infertility^a^ ::: rs882605/rs2246901/rs2688513 Infertility, % Noninfertility, % *P*value ------------------------------ ---------------- ------------------- ---------- GTT 87.2 67.4 0.012^b^ TGC 12.5 24.4 0.099 TTT 0.2 4.4 0.206 GGT 0.0 2.2 0.344 ^a^*MUC4*, mucin 4 gene. The reported haplotype percentages are estimated percentages based on allele frequencies and the linkage disequilibrium. The *P*values are based on a comparison of a given haplotype with all other haplotypes combined. ^b^*P*\< 0.05 was considered statistically significant. ::: *MUC4*gene polymorphisms and amino acid substitutions ----------------------------------------------------- Because these endometriosis-associated SNPs can cause amino acid substitutions (Additional file [1](#S1){ref-type="supplementary-material"}, Table S2), the biofunctions of MUC4 might be altered by changes in hydrophilicity and protein folding. Figure [2](#F2){ref-type="fig"} illustrates the functional domains in MUC4 protein sequence and secondary structures that contain these SNPs. Our data show that genetic variations of rs882605 and rs2688513 cause amino acid substitutions in long-loop regions (\> 10 residues) between secondary structure units (α-helices or β-strands) with high hydrophilicity and moderate surface probability (Figure [2](#F2){ref-type="fig"}). Amino acid composition analyses also revealed that these two loops (300-319 and 4,134-4,158) contain several negatively charged aspartate (D) residues and reverse turn elements, glycine (G) and proline (P), suggesting the importance of these loops in protein folding and functional regulation \[[@B34],[@B35]\]. In addition, rs2246901 locates in a type D von Willebrand factor (VWFD) domain responsible for protein-protein interaction and cell adhesion and/or migration \[[@B36],[@B37]\]. Our findings support the functional roles of *MUC4*SNPs in regulating cellular mobility and invasion of endometrial cells during endometriosis development and progression. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Functional domains in the MUC4 protein sequence and the predicted secondary structures**. Six functional domains and/or signatures (boxes) were annotated by aligning MUC4 protein sequence in PROSITE protein domain database \[[@B28]\]. The boundaries of each signature are listed. Among six SNPs tested in this study (stars), rs882605 and rs2688513 (black) were found in two different long-loop regions, 300-319 and 4,134-4,158, respectively (bold letters refer to amino acid substitution sites). The SNP rs2246901 (gray) was found in a type D von Willebrand factor (VWFD) domain. The *MUC4*reference sequence can be located at National Center for Biotechnology Information databank NP\_060876.4. :::  ::: Discussion ========== Previous studies have shown polymorphism of cytokines and adhesion molecules which were associated with the pathogenesis of endometriosis \[[@B38],[@B39]\]. To the best of our knowledge, however, no other study to date has investigated the possible association of *MUC4*and endometriosis. The purpose of this study was to evaluate whether genetic variations in *MUC4*associate with endometriosis in the Taiwanese population. Our data prove the association of *MUC4*polymorphisms with advanced stages of endometriosis and the related infertility. Since the extracellular domain of MUC4 is critical for HER2 interaction and cell invasiveness, these defined SNPs located in putative functional domains of MUC4 may play important roles during endometriosis development and progression. The development of endometriosis and certain types of ovarian cancer share several similar clinical features. For example, endometriosis could progressively invade pelvic viscera, resulting in adhesion, and could recur after medical treatment or surgery. Because of the functions involved in the acquisition of adhesion ligands or receptors and the loss of antiadhesion, proteins such as MUC4 and MUC1, the two major mucins present in endometrial epithelium, thus become suspect in endometriosis development \[[@B16],[@B17]\]. In addition to gene overexpression, genetic variations in *MUC1*have also been reported as risk factors contributing to cell mobility and the severity of cancer \[[@B40]-[@B43]\]. However, the influence of *MUC4*genetic variations on cell behavior remains unclear. In this study, Pro4135Ser (rs2688513) and Ala4693Ser (2246901) substitutions in the putative functional domains of MUC4 were found to be associated with advanced stages of endometriosis. Since MUC4 is an emerging target for ovarian cancer \[[@B18],[@B19],[@B44],[@B45]\], our study provides a new direction from which to address the roles of MUC4 in the development of gynecological disorders. To study the genetic effects of mucin proteins by SNPs, the major interest focuses on the variable number of tandem repeat (VNTR) polymorphisms, which result in different-sized gene transcripts. For examples, *MUC1*variations in the VNTR domain have been found to play roles in regulating *Helicobacter pylori*binding to gastric cells \[[@B46]\]. Other studies have also concluded that VNTR polymorphisms can influence T-antigen presentation and the local immune responses, which consequently have potential effects on gastric cancer development \[[@B47],[@B48]\]. With regard to *MUC4*, a high degree of polymorphism in the VNTR domain was observed in human tissues, including the endometrial epithelium \[[@B49],[@B50]\]. However, the different-sized *MUC4*transcripts did not show association with embryo implantation or cancer development. By contrast, MUC4 can promote cell proliferation and antiapoptotic effects in cancer cells by interacting with HER2 on the cell surface \[[@B18],[@B19],[@B22]\], suggesting the potency of functional domains in the extracellular domain of MUC4. In this study, two SNPs (rs2688513 and rs2246901) that locate in a putative functional loop and the VWFD protein binding domain, respectively, were found to be associated with advanced stages of endometriosis. Further study may clarify whether these amino acid substitutions could change the interaction with HER2 and/or play crucial roles in regulating the cellular activity of the spread of endometriosis. Endometriosis could cause pelvic adhesion and tubal occlusion that may lead to infertility. However, among patients with endometriosis-related infertility, 50% to 60% of them were diagnosed with minimal or mild endometriosis \[[@B3]\]. Impaired folliculogenesis, bad oocyte quality and impaired implantation of embryo are therefore considered to be the possible mechanisms for endometriosis-related infertility. Changes in cytokines and growth factors in endometrium, follicular fluid and peritoneal fluid have been suggested as the key players for inducing the above-mentioned phenomena \[[@B5]\]. Recently, several studies have shown that MUC4 could promote cell migration, change the endometrial environment and create weak points in the epithelium, thus facilitating the failure of embryo implantation \[[@B50],[@B51]\]. The Carraway *et al*. \[[@B52]\] study also showed that embryo implantation was associated with downregulation of MUC4 expression in an animal model. In this study, women with a T allele in rs882605 had a lower risk of endometriosis-related infertility, whereas rs2688513 and rs2246901 SNPs did not show any association with the reproductive ability of patients. The rs882605 SNP locates in a putative functional loop within the VNTR domain of MUC4 that may control T-cell antigen presentation and the local immune responses. Our findings may support the view that the regulation of local immunity, rather than uncontrolled cell proliferation, in the endometrium may play a more important role in the development of endometriosis-related infertility. Our study shows that the T/G genotype at rs882605, as compared with the T/T or G/G genotypes, is unique in patients with endometriosis. So far, our data cannot provide sufficient information to explain why the T/T genotype does not show higher risk of endometriosis than T/G. One reason could be the relatively small study group, while other possibilities could exist. For example, the SNPs analyzed are in tight LD with other unknown allele variants which have an opposite effect. In this case, only individuals with the T/G heterozygous genotype could be observed. Because this is a hospital-based study with a modest sample size, enrollment of a larger cohort based on a population approach could help to elucidate the functional role of MUC4 in endometriosis and the related infertility. Conclusions =========== In this study, we observed an association of *MUC4*polymorphism with endometriosis development and endometriosis-related infertility in a Taiwanese population. However, the true mechanism of how MUC4 modulates the pathogenesis of endometriosis and infertility was not clearly understood. In addition, the risk SNPs pose for endometriosis stages and infertility differ, suggesting dissimilar molecular mechanisms for these clinical features. More detailed studies are needed to investigate the biochemical pathways regulated by MUC4 during the development of endometriosis. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= CYYC collected samples, carried out the clinical association study and drafted the manuscript. HWC participated in sample collection and the clinical association study. CMC carried out sample preparation and SNP analyses. CYL participated in SNP analyses. CPC participated in the clinical association study. CHL carried out statistical analyses. WYL carried out sample pretreatment and RNA extraction. JJCS carried out the experimental design and drafted the manuscript. FJT carried out the experimental design and revised the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1741-7015/9/19/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional file 1 **Supplementary Tables S1 to S3**. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 2 **Supplementary Figures S1 and S2**. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ The authors appreciate technical assistance from Carmen Chan at China Medical University Hospital. This study was funded by grant DMR-98-073 from China Medical University Hospital in Taichung, Taiwan.

PubMed Central

2024-06-05T04:04:19.022502

2011-2-24

PMC3052196

Background ========== Gastrointestinal stromal tumors (GIST) are the most frequent mesenchymal tumors of the gastrointestinal tract and are thought to originate from the interstitial cells of Cajal \[[@B1]\]. The management of advanced and metastatic GIST has considerably improved with the use of imatinib mesylate (IM). Approximately 50 to 70% of unselected patients with advanced or metastatic GIST respond to IM and the median progression-free survival (PFS) is 20 to 24 months \[[@B1],[@B2]\] in a highly chemo-resistant disease \[[@B3]\]. IM, originally designed as a specific inhibitor of the Bcr-Abl kinase for the treatment of chronic myelogenous leukaemia, was also shown to be a potent inhibitor of the tyrosine kinase activities of KIT, PDGFR and CSF1R. *KIT*or *PDGFRA*mutations are considered an early event in the oncogenesis of GIST \[[@B4],[@B5]\] and are found in roughly 90% of cases. IM is considered a standard of care for patients with advanced disease and as adjuvant therapy for completely resected localised GIST \[[@B6]-[@B8]\]; however, its role in the neoadjuvant setting is currently under investigation. The outcome of patients with locally advanced and unresectable GIST is generally considered to be similar to that of patients with metastatic disease. Although first-line treatment with IM produces high rates of disease control in patients with advanced disease, most patients experience disease progression due to the emergence of molecularly resistant clones within 2-3 years after treatment initiation. This observation led several authors to investigate the value of surgical excision of residual disease following response to IM, before the development of secondary resistance. Several publications have reported on the feasibility of surgery following primary treatment with IM \[[@B9]-[@B14]\], but little is known about the exact benefit in terms of progression-free or overall survival. Furthermore, all these studies included patients with both locally advanced and metastatic disease. All were retrospective, except the recently reported RTOG-0132 study \[[@B11]\]. We report here the retrospective analysis of patients with locally advanced non metastatic GIST who received primary medical therapy with IM in the BFR14 prospective trial \[[@B15]\], with special attention to the patients who underwent secondary surgery of their primary tumor. Methods ======= BFR14 population ---------------- The BFR14 trial is a phase III trial randomizing interruption *versus*continuation of imatinib beyond one year of treatment for non progressive patients \[[@B15]\]. After the results of the randomisation at one year were known, the protocol was amended to allow randomisation after three years of treatment, and more recently after five years of treatment \[[@B16]\]. Inclusion criteria were: age at least 18 years, histological confirmation of locally advanced and/or metastatic GIST, immunohistochemical documentation of c-KIT (CD117) expression, and Eastern Cooperative Oncology Group performance status of 0 to 3. Patients had to be previously untreated with imatinib, with no history of previous malignancy, and were required to have normal renal, cardiac, and hepatic functions. No concurrent anticancer therapy was allowed. All patients gave written informed consent before inclusion. Imatinib was given orally at 400 mg per day, as a single daily dosing. Clinical and biological tolerance was assessed weekly during the first month of treatment, every 2 weeks the following month, then monthly for three months, and every three months thereafter. Initial assessment included a complete history, clinical examination, serum biochemistry, liver function test, whole blood count and computed tomography scan (CT scan) and/or magnetic resonance imaging (MRI) of the tumor. Imaging techniques were repeated after 6 and 12 weeks of treatment and every 3 month thereafter. Response was graded according to the RECIST criteria \[[@B17]\]. Definition of non metastatic locally advanced GIST -------------------------------------------------- To be eligible for the present substudy, patients of the BFR14 study were to have primary non metastatic GIST as assessed by the local multidisciplinary team at each participating site, and no prior surgery, thus excluding patients with recurrent GIST. Statistical analysis -------------------- Data were described using the median and range for continuous variables and using percentages with 95% confidence intervals for qualitative variables. Comparisons were performed using the chi-square test, Fisher\'s exact test, or Wilcoxon\'s rank-sum test as appropriate. Survival times were calculated from the date of entry in the BFR14 trial (i.e. initiation of IM treatment) and were displayed using the Kaplan-Meier method \[[@B18]\]. Progression-free survival (PFS) was defined as the time from the date of inclusion (start of IM) to the date of progression on IM 400 mg/day or death. Overall survival (OS) was defined as the time from the date of inclusion to death of any cause. Differences in survival distributions were tested using the Log-Rank test. Differences were considered statistically significant when p ≤ 0.05. Statistical analyses were performed using the SPSS 12.0.1 (SPSS Inc, Chicago, IL) software package. Results ======= As of April 2010, 434 patients have been included in the BFR14 trial. Sixty patients were registered as having no known metastasis initially. Of these 60 patients, only 25 had no previous history of surgery for GIST and were included in the present substudy. The 35 remaining cases included patients with locally advanced disease who received IM as additional therapy following R2 resection or patients with locally advanced recurrent disease. Characteristics of these 25 patients and response to IM are presented in Table [1](#T1){ref-type="table"}. Median age was 65.5 (range 39.8-80.5) years, 16 patients were males and the median performance status was 1 (range 0-3). Median tumor size at baseline was 15 cm. No complete response was seen following treatment with IM, 15 patients (60%) had a partial response (PR) after a median of 4.0 (range 1.4-12.8) months, 7 patients (28%) had stable disease (SD) as their best response, while 3 (12%) patients had progressive disease (PD) as their best response (Figure [1](#F1){ref-type="fig"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patients\' main characteristics ::: ----------------------------------------------------------------------------------------------------------- All Patients who\ Patients who\ underwent surgery were operated --------------------- ------------- ------------------- --------------- ---------- ------------- ---------- 25 9 16 Age  Median\ 65.5\ 60.0\ 69.7\ (range) (39.8-80.5) (39.8-80.5) (40.7-81.9) Gender  Male 16 64% 6 67% 10 63%  Female 9 36% 3 33% 6 38% Tumor location 0%  Stomach 4 16% 1 11% 3 19%  Small intestine 7 28% 4 44% 3 19%  Peritoneum 7 28% 1 11% 6 38%  Oesophagus 2 8% 0 0% 2 13%  Rectum 4 16% 3 33% 1 6%  Pelvis 1 4% 0 0% 1 6% Tumor size (mm)  Median (range) 150 (36-280) 150 (45-280) 149 (36-200)  ≤ 50 2 8% 1 11% 1 6%  50 \< - ≤ 100 7 28% 3 33% 4 25%  \> 100 16 64% 5 56% 11 69% WHO PS 0%  0 10 40% 4 44% 6 38%  1 10 40% 5 56% 5 31%  2 2 8% 0 0% 2 13%  3 1 4% 0 0% 1 6%  Not reported 2 8% 0 0% 2 13% Best response to IM  PR 15 60% 6 67% 9 56%  SD 7 28% 2 22% 5 31%  PD 3 12% 1 11% 2 13% ----------------------------------------------------------------------------------------------------------- ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Waterfall plot of patients\' best response**. Dark grey bars represent patients who did not undergo resection of their primary tumor. Light grey bars represent patient who had their primary tumor resected. :::  ::: Nine of the 25 patients (38%) underwent surgery after a median of 7.3 (range 3.4-12.0) months of IM treatment. Surgery was performed by the local surgeons who had initially refuted operation. In 4 patients tumor shrinkage with IM allowed resection at the price of a less morbid operation. In 2 patients surgical resection was performed as a salvage procedure for disease progressing on IM (primary progession in one case and progression following initial response in another case). The reason(s) that lead to the decision to operate each patient are listed in table [2](#T2){ref-type="table"}. Patients who underwent surgery tended to be younger (median 60.0 *vs*69.7, p = 0.174) and have better performance status (PS) (PS 0-1 9/9 *vs*11/16, p = 0.253) than the non-operated patients, although these differences were not statistically different (table [1](#T1){ref-type="table"}). The sex ratio was similar between the two groups. Non-operated patients were more likely to have GIST from the peritoneum or the oesophagus (7/16 *vs*1/9, p = 0.040). Baseline tumor size was comparable between patients who were operated and those who were not (p = 0.671, Mann and Whitney test). The response rate was slightly superior in the operated group (67% *vs*56% in the non-operated group), but the difference was not statistically significant (p = 0.691). The clinical benefit rate (PR+SD) was similar in the two groups (89% and 88%, p = 1.0). Among the 9 patients who underwent surgery, 6 had a partial response after a median of 4.2 months of treatment with IM, and surgery was microscopically complete (R0) for 5 of them and macroscopically incomplete in one of them (R2). Two patients underwent surgery with a volumetric response which was less than a partial response, i.e. disease stabilization according to RECIST; and surgery was complete for both of them (R0). One patient underwent surgery while his disease was progressing according to RECIST: surgical excision was macroscopically complete but with positive margins, and this patient remains in complete response (CR) 66.4 months after the procedure despite his refusal to continue IM post-operatively. The other eight operated patients accepted to continue IM after surgery as specified by the study protocol. Three of 9 patients progressed 13.3, 17.0 and 24.8 months after surgery; the other 6 patients are currently in CR at a median of 55.4 (range 50.0-66.5) months after surgery. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Reasons for operation following treatment with Imatinib ::: Patient N Age at surgery Tumor location Best % tumor shrikage Time to surgery reason for operation ----------- ---------------- ---------------- ----------------------- ----------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------- 4 43 Rectum -47 12,1 Significant response after 12 months on imatinib, enabling tumor resection 6 66 Mesentery -63 6,7 Large tumor lesion, decision to operate following tumor shrinkage on imatinib 7 81 Rectum -47 8,1 After initial response, patient had early signs of progression (increased blood flow on DCE-ultrasound) and was therefore operated before actual RECIST progression 8 40 Rectum -46 7,3 Surgery enabled following tumor shrinkage 12 43 Small bowel -74 6,5 Surgery planned prior to treatment with imatinib (true neoadjuvant) 13 61 Small bowel -74 11,7 Surgery enabled following tumor shrinkage 14 71 Small bowel 20 3,4 Rapid progression on imatinib 400 mg/d, dose increased to 600 mg/d which was poorly tolerated, salvage surgery seemed feasible. Resection was R1 15 76 Stomach 5 4,4 No response on imatinib with poor tolerance. Following surgery this patient was restarted on a lower dose of IM. 16 50 Small bowel -19 7,3 Stable disease after 6 months on imatinib, surgery was deemed feasible by surgeon. ::: With a median follow-up of 53.5 (range 4.4-77.0) months, the median PFS is 32.1 months for the 25 patients (Figure [2](#F2){ref-type="fig"}), while median OS is not reached (not shown). PFS was significantly longer (figure [3](#F3){ref-type="fig"}) for the 9 patients who underwent surgery after IM than for the 16 patients who continued medical treatment alone (median PFS not reached *vs*23.6 months respectively, p = 0.0322). Similarly, OS (figure [4](#F4){ref-type="fig"}) was longer for the resected group (median OS not reached *vs*29.7 months respectively, p = 0.0154). This difference in PFS and OS persisted when analysis was limited to patients with PR or SD on IM, after exclusion of progressive patients, although the difference in PFS did not reach statistical significance in this analysis (median 29.7 vs not reached, p = 0.0998 and 42.2 vs not reached, p = 0.0333 for PFS and OS respectively). When analysing only the 15 patients who had PR to IM, the median PFS was 29.7 months for non-resected patients *vs*not reached for resected patients (p = 0.2829). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Progression-free-survival**. :::  ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### **Progression-free survival according to surgical status**. :::  ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### **Overall survival according to surgical status**. :::  ::: Discussion ========== Imatinib mesylate is the new paradigm of treatment targeting the initial causal molecular event in solid tumors. However, it is still unclear whether patients with locally advanced or metastatic disease are cured, as shown by the results of the BFR14, which showed consistent disease progression following treatment interruption after 1, 3 and 5 years of treatment \[[@B15],[@B16]\]. The role of surgery during or following treatment with IM for locally advanced disease remains controversial. Several expert centers throughout the world have reported their results with surgical resection following treatment with IM for both locally advanced and metastatic GIST\[[@B9]-[@B14]\]. All these studies are unicentric and retrospective. Most of them have underlined the strong correlation between the disease status regarding response to IM before surgery and the outcome after surgery. Patient with multifocal progressive disease at the time of surgery have a short PFS following surgery (3-6 months), while patients with unifocal disease progression have a 6 months median PFS with some patients remaining free of progression after long-term follow-up \[[@B12],[@B13]\]. The recently reported RTOG0132/ACRIN6665 trial \[[@B11]\] is the first study to prospectively assess the role of preoperative IM (during 8 to 12 weeks) in patients with primary locally advanced (≥ 5 cm: 30 patients) and/or metastatic/recurrent (≥ 2 cm: 20 patients) GIST. In this study, toxicity was minimal and did not modify post-operative morbidity. However, because it was a single arm phase II trial, this study did not answer the question of the benefit of surgery in patients with locally advanced initially inoperable GIST. Our present study is the first multi-centric series to address the issue of benefit of surgery after neoadjuvant IM in this setting. We show that among 25 patients with non-metastatic locally advanced GIST, 9 patients (36%) were selected to undergo surgical resection following primary medical treatment with IM. These 9 patients had improved PFS and OS compared to non-operated patients, with survival rates close to those observed for localised intermediate or high risk GIST, whereas survival of non-operated patients was similar to that of patients with metastatic disease. Although these results suggest an improved outcome for operated patients, this study has some obvious limitations. One of these limitation is that patients were selected and not randomised to undergo surgery and were therefore more likely to benefit from the procedure based on medical judgement by the investigators at each site. Furthermore, our series is small and retrospective, precluding any definitive conclusion. As previously mentioned our observation is likely biased since selection of patients for surgery may be linked to other prognostic factors such as tumor location, patient\'s age, performance status as reflected by the differences (though not significant) seen in our series between the operated and non-operated groups. The response to IM may be another source of bias as more patients had a PR in the operated group than in the non-operated group. However, survival remained better in the operated group even when considering only patients with partial response or patients with clinical benefit (PR or SD). A possible source of difference of survival between the two groups may be the randomisation (see the \"patients and methods\" section). Six of 25 patients were randomised, two in the IM continuation arm and four in the interruption arm. All of the four patients randomised to interruption were in the non surgical group, therefore introducing a bias. However, PFS and OS were still significantly better in the surgery group when these four cases were removed from analysis (9.0 months *vs*median not reached p = 0.0037 and 26.3 months *vs*median not reached, p = 0.0128 respectively for PFS and OS). Another bias source of this multicentric study lies in the inclusion criterion of initial unresectability, which was left at the treating physician\'s discretion. Therefore, some patients may have had truly unresectable disease, while others may have had disease that was actually resectable at the price of a major procedure, in which case primary medical treatment appeared to be the best option. Resectability, before and after IM, was assessed by multidisciplinary teams, including surgeons expert in GIST management. Conclusions =========== Overall, this and other reports can not lead to any definitive conclusion regarding the benefit of surgery in patients with locally advanced GIST treated with IM. This benefit can only be demonstrated in randomised prospective trials, which are ongoing in the metastatic setting. However, since patients with locally advanced disease who become operable following IM appear to benefit from resection of the primary tumor, we think that surgery should be proposed, or at least discussed, in this subgroup of patients when the disease no longer responds to IM. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= AB contributed to provision of patients, participated in data acquisition, analysed the data, and helped draft the manuscript; PAC designed the study, contributed to patient provision, participated in data acquisition, analysed the data, and drafted the manuscript; FB contributed to patient provision and helped draft the manuscript; JF designed the study and contributed to patient provision; IRC and BB, AA contributed to patient provision and helped draft the manuscript, MR and DC contributed to provision of patients; DP contributed to the design of the BFR14 study and participated in data acquisition; JYB and AL contributed to the design of the BFR14 study, to patient provision and helped draft the manuscript. All authors have read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/11/72/prepub> Acknowledgements ================ The BFR14 trial was funded by a research grant from Novartis (Basel, Switzerland)

PubMed Central

2024-06-05T04:04:19.027148

2011-2-15

PMC3052197

Background ========== Glioma, the most common tumor of the central nervous system (CNS), frequently leads to death. Glioma is derived from brain glial tissue and comprises several diverse tumor forms and grades. Treatment of malignant gliomas is often palliative due to their infiltrating nature and high recurrence. Despite advances in surgery, chemotherapy and radiation gradually result in therapy-resistance. However, genetic events that lead to gliomas are mostly unknown. Recent researches highlight the importance of cancer-initiating cells in the malignancy of gliomas \[[@B1]-[@B3]\]. These cells have been referred to as glioma stem cells (GSC), as they share similarities to normal neural stem cells (NSCs) in the brain. There is increasing evidence that malignant gliomas arise from and contain these minority tumor cells with stem cell-like properties. This subpopulation of tumor cells with the potential for self-renewal and multi-lineage differentiation that recapitulates the phenotype of the original glioma \[[@B4]-[@B8]\], plays an important role in glioma initiation, growth, and recurrence. Eliminating GSCs from the bulk tumor mass seems to be a prosperous therapeutic strategy \[[@B9],[@B10]\]. Therefore, it is extremely important to understand the signal pathways that contribute to the formation and maintenance of GSCs. A number of signal pathways are involved in the formation and maintenance of stem cells, many of which are closely conserved across species. Notch signaling, an evolutionarily conserved pathway mediating direct cell-cell interaction and signaling, plays a pivotal role in the maintenance of NSCs \[[@B11]\]. The functions of the Notch pathway in cancer formation have been gradually established, and recent data have also implicated a role for Notch signaling in GSCs \[[@B12]\]. Notch is a family of hetero-dimeric transmembrane receptors composed of an extracellular domain responsible for ligand recognition, a transmembrane domain, and an intracellular domain involved in transcriptional regulation. When Notch receptor is triggered by the ligands on the neighboring cells, the intracellular domain of the Notch receptor (NICD) is released from the membrane, after successive proteolytic cleavages by the γ-secretase complex \[[@B13],[@B14]\]. NICD then translocates into the nucleus and associates with the transcription factor RBP-J, the DNA recombination signal binding protein-Jκ. The NICD-RBP-J complex further recruits other co-activators, and activates the expression of downstream genes associated with cell proliferation, differentiation and apoptosis \[[@B15]\]. It is believed that γ-secretase inhibitors (GSI) decrease the activity of Notch signaling and slow the growth of Notch-dependent tumors such as medulloblastoma \[[@B12]\]. Rapid proliferation, self-renewal ability and multipotential differentiation are the hallmarks of both normal NSCs and GSCs. Similarities in the growth characteristics and gene expression patterns of normal NSCs and brain tumor CSCs suggest that pathways important for NSCs are probable targets for eliminating brain tumor CSCs. The RBP-J-mediated canonical Notch pathway plays several significant roles in the maintenance and differentiation of NSCs \[[@B16]-[@B18]\]. During embryogenesis, Notch signaling is required to maintain all NSC populations, and to repress the differentiation of NSCs into intermediate neural progenitors (INPs) in vivo \[[@B19]-[@B21]\]. Along with later development, Notch signal commits NSCs to an astroglia fate, while repressing neuronal differentiation \[[@B22]\]. In adult, Notch signaling modulates cell cycle in order to ensure brain-derived NSCs retain their self-renewal property \[[@B23]\]. Increasing evidence has shown that there is a link between tumorigenesis and aberrantly activated Notch signaling \[[@B24],[@B25]\]. Notch1 and its ligands, Dll1 and Jagged1, were overexpressed in many glioma cell lines and primary human gliomas. When the expression of Notch1, Dll1 or Jagged1 was down-regulated by RNA interference, apoptosis and proliferation inhibition in multiple glioma cell lines were induced \[[@B26]\]. Depletion of Hey1, a member of Hes-related family downstream effectors of Notch signaling, by RNA interference also reduces proliferation of glioblastoma cells in tissue culture \[[@B27]\]. Moreover, the blockade of Notch signaling directly caused cell cycle exit, apoptosis, differentiation, and reduced the CD133-positive cells in medulloblastoma and glioblastoma cell lines while Notch activation enhances the expression of Nestin, promotes cell proliferation and the formation of NSC-like colonies and plays a contributing role in the brain tumor stem cells \[[@B28]-[@B30]\]. However, the exact roles of Notch signaling in the proliferation and differentiation of patient-derived GSCs have not been clearly elucidated. In this study, we explore the roles of Notch signaling in patient-derived GSCs with parallel analysis of normal NSCs by using GSI-mediated inhibition of Notch signaling in vitro. The results showed that when Notch signaling was inhibited, the proliferation and self-renewal ability of GSCs from human primary gliomas were attenuated. In addition, the blockade of Notch signaling in GSCs increased their differentiation into the downstream neural cell types, and promoted their conversion from stem cells into INP-like cells. Interestingly, although inhibition of Notch signaling definitely decreased the proliferating GSCs in long term culture, we found that the percentage of G2+M phase-GSCs were almost undisturbed at the initial stage of GSI treatment. To summarize, our results suggested that Notch signaling maintained GSCs by promoting their self-renewal and inhibiting their differentiation into INP-like cells, and supported that Notch signal inhibitors might be prosperous candidates of the treatments targeting CSCs for gliomas. Methods ======= Glioma samples -------------- Glioma tissues were obtained from 9 adult patients with pathologically diagnosed grade 2 to grade 4 gliomas, at the Department of Neurosurgery in Xijing Hospital, Fourth Military Medical University, under the guidance from the Medical Ethnic Committee of the Fourth Military Medical University. The summary of the patient population is outlined in Additional file [1](#S1){ref-type="supplementary-material"}: Table S1. Neurosphere culture ------------------- Neurosphere cultures were performed as described previously with some modifications \[[@B21]\]. Briefly, for the culture of NSCs, the brains from embryonic (E) day 12.5 C57BL/6 mice were dissected under a stereomicroscope. And for the culture of GSCs, tissues from patient specimen were acutely minced after sampling. The tissues were then washed, mechanically dissociated by repetitive pipette. Single cells were primarily plated in serum-free Dulbecco\'s modified Eagle\'s medium (DMEDM)/F12 medium containing 20 ng/ml basic fibroblast growth factor (bFGF, human recombinant, Sigma), 20 ng/ml epidermal growth factor (EGF, mouse submaxillary), the B-27 (1:50, GIBCO), penicillin (100 U/ml) and streptomycin (0.1 mg/ml). Cells were cultured at a density of 1 × 10^5^cell/ml in 24-well plates (0.5 ml/well), and were fed every 3 days by adding fresh medium supplemented with GSI or DMSO with indicated concentrations. Animal experiments were reviewed and approved by the Animal Experiment Administration Committee of the Fourth Military Medical University. Neurosphere assays ------------------ After 7 days from primary culture the numbers of primary spheres were counted under a microscope (Additional file [1](#S1){ref-type="supplementary-material"}: Figure S2) \[[@B21]\]. And for the expression of target genes, neurospheres were harvested on the 5th day of culture for RNA extraction, cDNA synthesis, and real-time reverse transcription-polymerase chain reaction (RT-PCR). Primary neurospheres were harvested and dissociated mechanically into single cell suspensions, and were replated at 1 × 10^5^cells/ml in 24-well plates. Cells were then cultured for another 7 days until secondary spheres formed \[[@B31]\], which were quantified by counting. On the 7th day of primary culture, neurospheres were plated onto poly-D-lysine (Sigma) coated glass cover slips in DMEM/F12 containing 10% fetal bovine serum (FBS) for another 7 days. On the third day of differentiation, neurospheres were photomicrographed and their neurites were counted and measured, then on the 7th day of differentiation culture, immunofluorescence staining was performed as described below. Immunofluorescence ------------------ Undifferentiated neurospheres were plated onto poly-D-lysine coated glass cover slips in serum-free medium for 4 h. Then cells were directly fixed in 4% paraformaldehyde at 4°C for 10 min, and incubated with primary antibodies overnight at 4°C, followed by species-specific secondary antibodies. Samples were visualized under fluorescence microscope (FV-1000, Olympus, Japan). Immunofluorescence for differentiated neurospheres was performed in a similar way. Cells were additionally counterstained with Hoechst. Primary antibodies used included rabbit anti-Nestin serum (1:200, Sigma), rabbit anti-glial fibrillary acidic protein (GFAP, 1:200, Sigma), mouse anti-mitogen-activated protein 2 (MAP2, 1:200, Sigma). FITC-conjugated goat anti-mouse IgG and Cy3-conjugated goat anti-rabbit IgG (1:400, Jackson ImmunoResearch) were used as the secondary antibodies. Quantitative RT-PCR ------------------- Total RNA of neurospheres was isolated using the Trizol reagent (Invitrogen). cDNA was synthesized and was used for real-time PCR with a kit (SYBR Premix EX Taq, Takara, Kyoto, Japan) and the ABI PRISM 7300 real-time PCR system, with human GAPDH and mouse β-actin as the reference controls. Primers used for real-time PCR were summarized in Additional file [1](#S1){ref-type="supplementary-material"}: Table S3. DNA content analysis -------------------- Spheres were dissociated mechanically into single cell suspensions in the culture medium. Cells were then washed and resuspended in PBS, and were fixed with ethanol at room temperature for 20 min. Cells were resuspended in PBS containing 50 μg/ml of propidium iodide and 0.1 mg/ml RNase A for 10 min, and were analyzed for ploidy using a flow cytometry (BD Biosciences). Data analysis was performed using the CellQuest software (BD Biosciences). Statistics ---------- Independent cultures from at least three samples were used for each experiment (Additional file [1](#S1){ref-type="supplementary-material"}: Table S2). For immunofluorescence, cells were counted by Image-ProPlus 6.0, and only cell bodies that were labeled with immunoreactivity were included. Proportions of immunoreactive cells in the total population of cultured cells revealed by Hoechst staining were calculated, and at least 5 microscopic fields per specimen were selected. For neurite analysis, neurites of 30 neurospheres from each culture in the presence of GSI or DMSO were measured. The total numbers of neurites per tumor spheres were counted via photomicrographs taken by a phase contrast microscopy, and the average of the length of neuritis per tumor spheres were measured by Image-ProPlus 6.0. Each experiment was repeated for at least three times. Data were expressed as mean ± s.e.m, and the difference between the two groups was analyzed with the Student\'s t-test, with *P*\< 0.05 as statistically significant. Results ======= Formation of neurosphere-like colonies from primary glioma specimens -------------------------------------------------------------------- Nine specimens of gliomas were used in the current studies, including 3 oligoastrocytomas, 3 oligodendrogliomas, 2 astrocytomas, and 1 glioblastoma, and the specimens were graded according to the WHO grading scheme (Additional file [1](#S1){ref-type="supplementary-material"}: Table S1). Tumor tissues were dissociated mechanically into a single cell suspension and were cultured in serum-free DMEM/F12 medium supplemented with EGF and bFGF. Seven primary gliomas of the nine gave rise to proliferating tumor spheres. Regardless of pathological subtype and grade, neurosphere-like clusters, or tumor spheres, first appeared within 72 h of primary culture and increased their numbers and diameters quickly during 7 days after the onset of the culture (Figure [1A](#F1){ref-type="fig"}). In order to estimate whether these tumor spheres showed NSC properties, we stained the tumor spheres from patients with anti-Nestin antibody. The result showed that these tumor spheres expressed Nestin, a marker of NSCs (Figure [1B](#F1){ref-type="fig"}). The multipotency of these human glioma cell-derived tumor spheres was confirmed by differentiation assay in vitro. We estimated the differentiation capacity of tumor spheres in differentiating conditions by examining the types of molecular markers expressed by neurons and glial cells. We observed that these cells could differentiate into GFAP-positive astrocyte- and MAP2-positive neuron-like cells (Figure [1C](#F1){ref-type="fig"}, [1D](#F1){ref-type="fig"}). In addition, a local recurrence tumor also could produce tumor spheres in growth medium (data not shown). Tumor spheres could be passed at least for five generations by mechanical dissociation and their stemness and multipotency could be maintained in serum-free medium supplemented with growth factors for at least one month. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Patient glioma-derived stem cells have the ability to form neurosphere-like colonies and gave rise to the downstream neural cell types of NSCs**. (A) Photomicrographs of typical primary tumor spheres from one glioma tissue at 72 h after plating. (B) Undifferentiated primary tumor spheres expressed high levels of Nestin (red), a marker of NSCs. (C, D) The tumor spheres-derived from human glioma were cultured in differentiation conditional medium for 7 days, and differentiated into neural cells expressing specific molecular markers of GFAP (C, red) and MAP2 (D, green). Scale bar, 100 μm in A, and 50 μm in BCD. :::  ::: Blockade of Notch signaling attenuates the proliferation and self-renewal ability and promotes differentiation of normal NSCs ----------------------------------------------------------------------------------------------------------------------------- Stem cell-like cells in brain tumors share many similarities with normal neural stem/progenitor cells and may require Notch signal for their survival and growth. In vitro, NSCs proliferate and form clonal spheres referred to as neurospheres. GSI reduced the proliferation of mouse embryonic brain-derived NSCs in a dose-dependent fashion (Additional file [1](#S1){ref-type="supplementary-material"}: Figure S1). The number of neurospheres was decreased in the presence of GSI, compared with the control treated with DMSO (Figure [2A](#F2){ref-type="fig"}). In order to confirm that GSI effectively blocked Notch signaling in NSCs in our culture system, we test the expression of Hes1 and Hes5, both of which are downstream molecules of the Notch signaling \[[@B12]\]. Total RNA was prepared from neurospheres on the fifth day of 25 μmol/L GSI treatment and was used for RT-PCR. The expression of Hes1 and Hes5 decreased remarkably in NSCs, suggesting that GSI at this concentration could inhibit Notch signaling effectively (Figure [2B](#F2){ref-type="fig"}, [2C](#F2){ref-type="fig"}). We quantitatively analyzed the number of primary neurospheres in the presence of GSI, and found that there was a significant decrease in the number of neurospheres upon GSI treatment at 25 μmol/L (Figure [2D](#F2){ref-type="fig"}). In order to determine the possible effect of GSI on the NSCs self-renewal ability, we harvested the spheres and dissociated them into a single cell suspension by soft pipeting. When replated in the presence of GSI, the number of secondary neurospheres significantly decreased after 7 days culture (Figure [2E](#F2){ref-type="fig"}). These results suggested that the proliferation of NSCs was slowed by inhibiting Notch signaling and the self-renewal ability, a key NSC behavior, was at least partially depleted. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Blockade of Notch signaling attenuates proliferation and self-renewal ability of normal mouse NSCs**. (A) Photomicrographs of neurospheres derived from E12.5 mouse brain at 72 h after primary culture, with GSI or DMSO supplemented. (B, C) Total RNA was prepared from neurospheres on the 5th day of GSI or DMSO treatment. And the expressions of Hes1 and Hes5 were measured by RT-PCR (B) and Real-time PCR (C), with β-actin as the reference control (n = 3, Hes5, *P*= 0.006, Hes1, *P*= 0.006). (D, E) Equal number of cells (1 × 10^5^/ml) were plated in the growth medium, and the number of primary (n = 3, *P*= 0.010) (D) and secondary (n = 3, *P*= 0.043) (E) neurospheres were counted 7 days after plating. \*, *P*\< 0.05, \*\*, *P*\< 0.01. :::  ::: Notch signaling has been shown to inhibit the differentiation of NSCs to INPs \[[@B21]\]. In our study, we tested the expression of molecular markers of INPs in primary neurospheres treated with GSI or DMSO. Quantitative RT-PCR showed that the mRNA levels of Glast, which is indicative of the frequency of NSCs, were decreased, while that of Mash1 and Tubulin α1, both of which are markers of INPs, were increased (Figure [3A](#F3){ref-type="fig"}, [3B](#F3){ref-type="fig"}). These results indicated an augmented differentiation from NSCs into INPs upon the blockade of Notch signaling by GSI. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### **Blockade of Notch signaling promotes the differentiation of normal mouse NSCs into INPs and downstream neural cell types**. (A, B) Total RNA was prepared from GSI or DMSO treated neurospheres derived from E12.5 mouse brain on the 5th day of culture. And the expressions of Glast, Mash1 and Tubulin α1 were measured by RT-PCR (A) and Real-time PCR (B), with β-actin as the reference control (n = 3, GLAST, *P*= 0.003, Mash1, *P*= 0.043, Tubulin α1, *P*= 0.046). (C, D) Immunofluorescence. Differentiated NSCs were stained with anti-GFAP, or anti-MAP2 antibodies after cultured on cover slips in differentiation conditional medium for 7 days. Stained samples were examined under a fluorescence microscope. (E, F) Quantification and comparison of neurons (MAP2^+^) or astrocytes (GFAP^+^) in GSI-treated and control NSCs. Cells were counterstained with Hoechst, to permit counting of cell nuclei in at least 5 microscopic fields per specimen (n = 3, E, *P*= 0.021, F, *P*= 0.031). \*, *P*\< 0.05, \*\*, *P*\< 0.01. Scale bar, 50 μm for C and D. :::  ::: To further study the effect of inhibiting Notch signaling on NSC differentiation, we used the neurosphere differentiation assay in vitro. When spheres were cultured adherently on poly-D-lysine coated glass cover slips without growth factors, they began to differentiate into cells bearing specific markers of neurons and astrocytes. We quantitatively compared the cell types produced by neurospheres in the GSI-containing medium with that of the control. All of the neurospheres gave rise to cells with the molecular markers of neurons or astrocytes (Figure [3C](#F3){ref-type="fig"}, [3D](#F3){ref-type="fig"}). However, the percentage of MAP2^+^cells increased significantly in the presence of GSI, from 29.0 ± 10.4% to 66.5 ± 8.4%, and the percentage of GFAP^+^cells in GSI-treated neurospheres was elevated from 8.7 ± 3.0% to 26.9 ± 6.6% (Figure [3E](#F3){ref-type="fig"}, [3F](#F3){ref-type="fig"}). These results suggested that inhibiting Notch signaling in NSCs leads to an increase in the number of differentiated cells. Decreased proliferation and self-renewal ability of GSCs upon GSI treatment --------------------------------------------------------------------------- Although Notch signaling has been shown to play critical roles in the maintenance of normal NSCs, whether this signaling might be involved in tumor stem cells is not fully clear. To determine whether Notch signaling activity was required during growth of GSCs, we investigate the effect of GSI on proliferation and self-renewal of GSCs. After Notch signaling was inhibited in GSCs by GSI treatment at 25 μmol/L, the expressions of Hes5 and Hes1, the specific and direct downstream targets of the Notch/RBP-J transcription complex were identified by RT-PCR and real-time PCR as described previously. After 5 days of GSI treatment, Hes5 and Hes1 expression markedly decreased (Figure [4A](#F4){ref-type="fig"}, [4B](#F4){ref-type="fig"}), and no obvious cell death was observed, indicating no effect on cell viability (data not shown). These results indicated that Notch signaling was efficiently blocked by GSI treatment in GSCs. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### **Attenuated proliferation and self-renewal ability of patient-derived GSCs on the blockade of Notch signaling**. (A, B) Total RNA was prepared from primary tumor spheres on the 5th day in the presence of GSI or DMSO. And the expressions of Hes5 and Hes1 were measured by RT-PCR (A) and Real-time PCR (B), with human GAPDH as the reference control (n = 5, Hes5, *P*= 0.046, Hes1, *P*= 0.002). (C, D) Equal number of cells (1 × 10^5^/ml) form brain tumor tissues were plated in the growth medium, the number of primary (n = 4, *P*= 0.008) (C) and secondary (n = 4, *P*= 0.041) (D) tumor spheres were counted 7 days after plating. \*, *P*\< 0.05, \*\*, *P*\< 0.01. :::  ::: Next, we quantitatively compared the proliferation and self-renewal ability of GSI-treated GSCs with that of the controls. The number of the primary tumor spheres in the presence of GSI decreased significantly, from 51.5 ± 2.8 to 34.8 ± 3.3 (Figure [4C](#F4){ref-type="fig"}). Self-renewal ability of the tumor spheres was assayed by dissociating and replating the primary tumor spheres. Our results showed that GSI-treated GSCs generated a decreased number of secondary tumor spheres (17.5 ± 2.3), than the number of controls (31.7 ± 5.6) (Figure [4D](#F4){ref-type="fig"}). These results showed that the proliferation and self-renewal ability of GSCs also could be attenuated by inhibiting Notch signaling. Blockade of Notch signaling promotes the differentiation of GSCs ---------------------------------------------------------------- The previous result indicated that inhibiting Notch signaling promotes the normal NSCs to differentiate into neurons and astrocytes, both of which are the downstream neural cell types of NSCs. Therefore, we investigated whether the GSI treatment promoted GSCs differentiation. Interestingly, after 3 days, approximately 18.7 ± 0.9 neurites grew out from each tumor spheres cultured in the medium with GSI, compared to only 6.7 ± 0.9 from that cultured with DMSO. Meanwhile, the average length of neurites increased from 206.0 ± 13.1 μm in tumor spheres culture with DMSO to 269.7 ± 28.4 μm in GSI-treated tumor spheres (Figure [5B](#F5){ref-type="fig"}, [5C](#F5){ref-type="fig"}). In order to further confirm whether these cells are the downstream neural cell types, immunofluorescence was performed on differentiated primary GSCs using the specific markers of neurons and astrocytes on the 7th day in differentiating conditional medium (Figure [5D](#F5){ref-type="fig"}, [5E](#F5){ref-type="fig"}). We quantitatively compared the cell types produced by neurospheres in the GSI-treated group with that of the control. The percentages of MAP2^+^cells and GFAP^+^cells increased significantly, as high as 51.6 ± 6.1% and 44.0 ± 1.7%, respectively (Figure [5F](#F5){ref-type="fig"}, [5G](#F5){ref-type="fig"}). These results suggest that inhibiting Notch signaling also promotes the differentiation of GSCs. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### **Augmented neurite outgrowth and enhanced differentiation of patient-derived tumor spheres on the blockade of Notch signaling**. (A) Photomicrographs of differentiated tumor spheres at 72 h after plated in differentiation conditional medium supplemented with GSI or DMSO. (B, C) Comparison of neurites number (n = 3, *P*\< 0.001) (B) and length (n = 3, *P*= 0.041) (C) between tumor spheres in the presence of GSI and DMSO. (D, E) Immunofluorescence. Differentiated tumor spheres were stained with anti-GFAP, or anti-MAP2 antibodies after cultured on cover slips in differentiation conditional medium for 7 days. Stained samples were examined under a fluorescence microscope. (F, G) Quantification and comparison of the percentages of neurons (MAP2^+^) (n = 3, *P*\< 0.001) (F) or astrocytes (GFAP^+^) (n = 3, *P*\< 0.001) (G) in the total cell number revealed by Hoechst counterstaining, between GSI-treated and control GSCs. Scale bar, 100 μm for A, and 50 μm for D and E. \*, *P*\< 0.05, \*\*, *P*\< 0.01. :::  ::: Blockade of Notch signaling promotes the conversion of GSCs to INP-like cells ----------------------------------------------------------------------------- The previous report indicated that blockade of Notch signaling in the CNS increased the frequency of INPs in vivo \[[@B21]\]. Precocious differentiation of NSCs into INPs might exhaust the NSC pool. Therefore, we investigated the effect of inhibiting Notch signaling on the frequency of GSCs and INP-like cells in glioma specimen. In an attempt to distinguish GSCs and INP-like tumor cells, we examined the expression of several markers that could distinguish NSCs from INPs by quantitive RT-PCR \[[@B20],[@B21]\]. Compared with the controls, the primary tumor spheres in the presence of GSI expressed lower Glast and CD133, which are indicative of the frequency of NSCs and GSCs. In contrast, Mash1 was highly expressed in GSI-treated tumor spheres (Figure [6A](#F6){ref-type="fig"}, [6B](#F6){ref-type="fig"}), although the expression level of another INP marker, Tubulin α1 was comparable between the GSI-treated tumor spheres and that of control. Altogether, these results suggested that blockade of Notch signaling may promote the conversion of GSCs to INP-like tumor cells. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### **GSI-treated primary tumor spheres show similar gene expression profile of INPs**. (A, B) cDNA was prepared from total RNA isolated from primary tumor spheres, treated with GSI or DMSO for 5 days respectively, and the expressions of GLAST (*P*= 0.002), CD133 (*P*= 0.015), Mash1 (*P*= 0.050) and Tubulin α1 (*P*= 0.116), were measured by RT-PCR (A) and Real-time PCR (n = 5) (B), with GAPDH as a reference control. \*, *P*\< 0.05, \*\*, *P*\< 0.01. :::  ::: GSCs show resistance to GSI treatment compared with NSCs -------------------------------------------------------- To gain further perspective on the dynamics of cellular proliferation accompanying differentiation, we treated NSCs and tumor spheres at a series of time points following GSI treatment with propidium iodide and examined cell cycle via FACS analysis. Compared with the controls, nearly 15.5 ± 0.5% of the NSCs treated with GSI for 24 h are in the G2+M phase, and then sharply decreased to less than 8.2 ± 1.7% at 72 h (Figure [7A](#F7){ref-type="fig"}). In contrast, the ratio of GSCs in the G2+M phase were slightly elevated at 48 h, and then declined insignificantly at 72 h (Figure [7B](#F7){ref-type="fig"}). The result showed that GSI treatment significantly reduced the ratio of the G2+M phase NSCs, but there is no obvious effect on the cell cycle of GSCs. Therefore, NSCs are more sensitive to GSI, while GSCs display a certain degree of resistance to GSI at the early stage of the treatment. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### **Different effects of GSI-treatment on the cell cycle of NSCs or GSCs**. (A, B) Comparisons of cell cycle between NSCs (A) and GSCs (B) in the presence of GSI or DMSO at 24 h, 48 h, 72 h using flow cytometry. Data represent as mean ± SD from three independent experiments. (n = 3, 24 h, *P*= 0.006, 48 h, *P*= 0.013). \*, *P*\< 0.05, \*\*, *P*\< 0.01. :::  ::: Discussion ========== Tumor stem cells such as GSCs have been considered as a novel target for the therapy of the malignant tumors, because these cells are supposed to play an important role in tumor initiation, growth, and recurrence \[[@B4]\]. Similarities in the growth characteristics and gene expression patterns of normal NSCs and GSCs suggest that pathways important for NSCs are probable targets for oncogenic brain tumor stem cells. In the present study, 1) we isolated GSCs from the human glioma tissues; 2) Like NSCs, these cells had the ability to form spheres in serum-free medium supplemented with growth factors and differentiated into downstream neural cell-like cells; 3) By GSI treatment, the number of GSCs-derived primary neurospheres and secondary neurospheres were markedly reduced compared with those treated with DMSO, indicating that in the long term culture (7-14 days), the proliferation and self-renewal ability of GSCs was ultimately reduced, upon the blocking of Notch signaling; 4) However, within 72 h culture, GSCs showed a certain degree of GSI-resistance, with undisturbed proliferation ability upon GSI treatment; 5) In addition, we showed that on blocking Notch signaling, GSCs are much biased to differentiate into INP-like cells, and ultimately neurons and glial cells in vitro. All these results suggest a promising preclinical application of Notch signaling antagonist (e.g., GSI)-based CSCs-targeting therapy in malignant glioma patients. The frequency of GSCs in tumor tissue ------------------------------------- Although CSCs have been identified as an important factor in tumor initiation and growth, their characteristics remain obscure concerning their heterogeneity. Here we found in our experiments that although seven of the nine human gliomas gave rise to proliferating tumor spheres, different numbers of spheres arisen from equal primary glioma cells among tumor samples. It should be noted that the specimens which did not give rise to proliferating neurospheres were patient \#1 (oligoastrocytoma, gradeII) and \#7 (anaplastic astrocytoma, grade III), with comparable tumor grades with the other specimens (Additional file [1](#S1){ref-type="supplementary-material"}: Table S1). Because samples are usually drawn from the periphery of the ablated tumor bulk, these two specimens might contain a certain amount of normal tissues. Overall, equal number of cells from high-grade and recurrent tumors, such as giant cell glioblastoma (WHO grade IV) and oligoastrocytoma (WHO grade III, recurrent tumor) often generate more primary tumor spheres. Due to limited number of samples, our accumulated results could not statistically lead to the conclusion that high-grade tumors contain more GSCs at present. However, the tendency described above indicated that the original frequency of GSCs might be different among samples according to tumor grades, or the GSCs from high-grade tumor tissues might show more typical properties of stem cells with higher proliferation and self-renewal ability. INP-like cells in GSCs population --------------------------------- In normal development of the brain, neurons and glia are generated from both NSCs and more limited INPs. And blockade of Notch signaling in NSCs have been shown to promote their conversion into INPs \[[@B20],[@B21]\]. GSCs can differentiate into neurons and astrocytes in culture medium with serum, as shown by our results and previous studies \[[@B4]\]. Like INPs, it is possible that an intermediate glioma progenitor cells (IGPs) also exist, linking the GSCs-IGPs-Neuron/glia hierarchy in tumor microenvironment \[[@B32]\]. Our results show that blocking Notch signaling in the primary tumor spheres leads to down-regulated mRNA level of CD133, a well accepted marker of GSCs at present, indicating a decrease of GSCs. Simultaneously, the mRNA level of Hes5 and Glast, two markers highly expressed in NSCs were also decreased, while that of Mash1, a marker up-regulated in INPs was increased in primary tumor spheres after being treated with GSI \[[@B20]\]. In addition, Tubulin α1, an INP marker, seems can not distinguish IGPs from GSCs. Since GSI-treated primary tumor spheres could still gave rise to secondary spheres, unless much fewer than that derived from control primary spheres, the CD133^low^/GLAST^low^/Hes5^low^/Mash1^high^IGPs might exist, with its number increased and proliferating ability decreased after the blockade of Notch signaling. Therefore, inhibiting Notch signaling might have therapeutic potential for human gliomas by exhausting GSCs and instruct them into less proliferative IGPs and differentiated neural cell types. Double positive cell types in the derivatives of GSCs ----------------------------------------------------- Although tumor-derived stem cells had many similarities to normal NSCs, it is important to note that differences might exist between them. Sphere differentiation assay on the specimen of 4\# patient demonstrated that GSCs could give rise not only to neurons and glia but also to a few cells that expressed both Map2 and GFAP, the molecular markers of astrocyte and neuron, respectively (Figure S3). Previous studies have reported similar abnormal cells in culture derived from pediatric and adult brain tumors \[[@B4],[@B33]\], indicating that such dual-fate cells might represent a significant fraction of GSCs derived progeny. These Map2^+^/GFAP^+^cells sometimes appeared larger than other cells derived from the same sphere (Figure S3). In addition, the GFAP positive glial cells derived from GSCs showed abnormal morphology, with slim cell bodies and neurites, compared with that derived from NSCs (Figure [3D](#F3){ref-type="fig"}, Figure [5E](#F5){ref-type="fig"}). Although morphological differences might exist between mouse and human glial cells, previous research on normal human tissue demonstrated that GFAP staining of human glial cells showed similar morphology with that of mouse glial cells \[[@B34]\]. Therefore, the morphological difference of GFAP positive glial cells might be attributed to whether they are NSC-derived or GSC-derived. Genetically, the generation of the double-positive cells and dysmorphic glial cells may accompany with gene mutation or abnormal activation of some signal pathways, leading to aberrant reprogramming procedure of GSCs, compared with normal differentiation of NSCs. GSI-resistance of GSCs at the early stage of GSI treatment ---------------------------------------------------------- In our study, we found that the numbers of both primary and secondary tumor spheres were decreased in the long run (7-day culture) after GSI treatment compared with the controls. However, cell cycle analysis results showed that although Notch blockade significantly reduced the ratio of the G2+M phase in NSCs, there is no obvious effect on the percentage of proliferating GSCs within 72 h after GSI treatment. These results indicate that, compared with NSCs, another distinctive feature of GSCs was that the former are more sensitive to GSI, while the latter displays a certain degree of resistance to GSI treatment at the early stage of the treatment. Due to the limited amount of primary glioma specimens, the cell cycle analyses were executed on primary tumor spheres from three independent tumor samples. Therefore, the resistance to GSI in GSCs at the early stage of the cell cycle might be a general characteristic in gliomas, or it only represents a few cases of glioma patients which might display resistance in the preclinical trial of GSI treatment. Previous research show that treatment with dipeptide GSI resulted in a marked reduction in medulloblastoma growth \[[@B35]\]. More recently, a clinical trial for a Notch inhibitor, MK0752 (developed by Merck, Whitehouse Station, NJ), has been launched for T-cell acute lymphatic leukemia and breast cancer patients (<http://www.clinicaltrials.gov/ct/show/NCT00100152>). Although GSI seems to be a promising reagent targeting GSCs by interfering Notch signaling, our results suggested that its effect might be limited to some glioma patients. Therefore, drug combination should be used at the early stage of therapy. However, since our results are based on in vitro culture system of patient-derived samples, more accurate conclusion could be drawn from animal models or preclinical trials in future study. The mechanisms of Notch signaling in regulating the proliferation and differentiation of GSCs --------------------------------------------------------------------------------------------- The mechanistic links between Notch signaling and the proliferation and differentiation of GSCs were presumably governed by more than one mechanism. In our study, the decreased proliferation and increased differentiation of GSCs upon GSI treatment are accompanied with down-regulation of Hes1 and Hes5, the canonical Notch downstream effectors. In addition, the expression level of Mash1, a proneural gene antagonized by the Hes genes was up-regulated in GSI-treated primary tumor spheres. Therefore, the canonical Notch-CBF1-Hes axis seems also play critical roles in the proliferation and differentiation of GSCs, as its function in NSCs \[[@B11]\]. On the other hand, Notch signaling has been shown to have both negative and positive influences on cell cycle progression \[[@B11],[@B36]\]. In the present study, we observed that the proliferation of GSCs decreased significantly in the long term culture, although GSI resistance of three glioma samples was present (see above). Mutations of p53, pTEN and H-Ras, have been identified in tumor tissues of giloma patients. And Notch signaling has been shown to crosstalk with p53 and pTEN signaling pathway, two major regulators of cell cycle \[[@B37],[@B38]\]. In addition, down-regulation of Notch signaling in H-Ras-transformed human breast cells led to a significant decrease in their proliferation \[[@B39]\]. Therefore, how Notch signaling promotes the cell cycle of GSCs is yet to be explored, on the scenery of the complex signal crosstalk and genetic circuitry. Conclusion ========== Our data indicate that like NSCs, Notch signaling maintains the patient-derived GSCs by promoting their self-renewal and inhibiting their differentiation, and support that Notch signal inhibitor might be a prosperous candidate of the drug treatment targeting CSCs for gliomas, however, with GSI-resistance at the early stage of treatment. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= YYH and MHZ carried out tissue culture, animal experiments and gene expression analyses, participated in study design and manuscript preparation. GC and LLi carried out specimens collection. LLiang, FG and YNW helped histological examination and immunohistochemistry staining. LAF and HH designed the study and prepared the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/11/82/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional file 1 **Hu et al Supplementary materials**The file contains Table S1-S3, Figure S1-S3 and their figure legends ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was supported by grants from the National Natural Science Foundation of China (30700693, 30973370, 31000485, 81071874) and the Ministry of Science and Technology of China (2009CB521706).

PubMed Central

2024-06-05T04:04:19.030390

2011-2-22

PMC3052198

Background ========== The number of people being diagnosed with cancer worldwide is rising sharply, mainly as a result of population growth, ageing and increases in the prevalence of several lifestyle factors related to cancer risk \[[@B1],[@B2]\]. Combined with improvements in survival, due in part to both earlier detection and better treatment, more people are living with a diagnosis of cancer than ever before, and this increasing trend is likely to continue \[[@B3]\]. One of the consequences of surviving cancer is the increased likelihood of being diagnosed with a second primary cancer. Among other things, these second cancers may be the result of lifestyle choices, genetics, environmental exposures (all of which may also be related to the first cancer) and late effects of treatment \[[@B3],[@B4]\]. For the patient and their treating clinician, quantifying and characterising the risks for second malignancies has important implications for screening (particularly when effective screening methods such as mammography are available), prevention strategies and counselling \[[@B3],[@B5],[@B6]\]. Identifying cancers which have an elevated likelihood of occurring together is also a useful starting point for investigating possible shared aetiologies and mechanisms of carcinogenesis \[[@B7]\]. This study reports for the first time the relative risks of a second cancer for people diagnosed with a first primary cancer in Queensland, Australia. Methods ======= A retrospective cohort design was used for this analysis. De-identified case records were obtained from the Queensland Cancer Registry (QCR), a population-based registry covering the entire state. All public and private hospitals, nursing homes and pathology services throughout Queensland are required by law to notify the QCR about any patients diagnosed with cancer, except for non-melanoma skin cancers \[[@B8]\]. The study cohort included all Queensland residents diagnosed with a first primary invasive cancer between 1982 and 2001 who survived for a minimum of 2 months. Since classifications for childhood cancers are different to adult cancers \[[@B9]\], we decided to restrict the cohort to people who were 15 years or older at the time of first diagnosis. A small number of records were excluded because information was missing for age (n = 20), or because the person was known to have had a first primary cancer diagnosed prior to 1982 (n = 76). The cohort was followed up until 31^st^December 2006 (allowing a potential minimum of five years and a maximum of 25 years after the initial diagnosis) to ascertain the occurrence of second primary invasive cancers. Histologically similar cases of cancer at the same body site were included, unless the medical record indicated that the tumour was recurrent or metastatic. Synchronous primary cancers (those diagnosed within 2 months of the first primary cancer) \[[@B10]\] were excluded because they were more likely to have been diagnosed as a result of detection bias \[[@B6]\]. Third or subsequent primary cancers were not considered in this analysis. Cancers of various body sites were analysed in a separate group if they averaged more than 100 eligible first primary cancers per year. These cancers are shown in Table [1](#T1){ref-type="table"}. Cancers of the head and neck (ICD-O-3 topography codes C00-C14 and C30-C32) were grouped together prior to analysis, as were cancers of the colon and rectum (C18-C20 and C218). All remaining types of cancer were analysed collectively in the category labelled \"Other\", including cases where site was ill-defined or unknown. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Characteristics of the study cohort ::: First Primary Cancers Second Primary Cancers ------------------------------------ ----------------------- ------------------------ -------- ------- **Total** 204,962 100.0 23,580 100.0 ***Sex***  Males 110,961 54.1 14,388 61.0  Females 94,001 45.9 9,192 39.0 ***Age at first diagnosis***  15-49 years 41,690 20.3 3,150 13.4  50-64 years 60,573 29.6 7,687 32.6  65 years and over 102,699 50.1 12,743 54.0 ***Period of first diagnosis***  1982-1986 36,216 17.7 4,716 20.0  1987-1991 43,904 21.4 5,841 24.8  1992-1996 57,381 28.0 6,972 29.6  1997-2001 67,461 32.9 6,051 25.7 ***Follow-up interval***  2 months to less than 1 year n.a. 2,511 10.6  1 year to less than 5 years n.a. 8,924 37.8  5 years to less than 10 years n.a. 7,286 30.9  10 years or longer n.a. 4,859 20.6 ***Type of first primary cancer***  Head and neck (C00-C14,C30-C32) 10,942 5.3 1,989 8.4  Oesophagus (C15) 2,075 1.0 105 0.4  Stomach (C16) 4,103 2.0 206 0.9  Colorectal (C18-C20,C218) 27,814 13.6 3,046 12.9  Pancreas (C25) 2,817 1.4 49 0.2  Lung (C33-C34) 17,347 8.5 729 3.1  Melanoma (C44,M872-M879) 29,289 14.3 5,092 21.6  Breast - female (C50) 26,725 13.0 2,962 12.6  Cervix (C53) 3,492 1.7 300 1.3  Uterus (C54) 3,527 1.7 409 1.7  Ovary (C56) 2,864 1.4 164 0.7  Prostate (C61) 23,122 11.3 3,006 12.7  Kidney (C64-C66,C68) 4,927 2.4 655 2.8  Bladder (C67) 8,719 4.3 1,680 7.1  Brain & CNS (C70-C72) 2,566 1.3 50 0.2  Thyroid (C73) 2,302 1.1 215 0.9  Non-Hodgkin lymphoma (M967-M972) 6,618 3.2 689 2.9  Lymphoid leukaemia (M982-M983) 2,645 1.3 422 1.8  Myeloid leukaemia (M984-M993) 2,195 1.1 125 0.5  Myeloma/PCT (M973) 2,174 1.1 134 0.6  Other 18,699 9.1 1,553 6.6 n.a. = not applicable; CNS = central nervous system; PCT = plasma cell tumours. ::: Person years at risk (PYAR) among people diagnosed with a first primary cancer was calculated as the time from 2 months after diagnosis until 31 December 2006, date of death or date of diagnosis of a second primary cancer, whichever came first. Data were stratified by type of first primary cancer, type of second primary cancer, sex, age at first diagnosis, period of first diagnosis and follow-up interval. Analysis by period of first diagnosis was restricted to five years of follow-up to allow more consistent comparisons across time. The expected number of second primary cancers in each stratum was calculated by multiplying the sum of PYAR by the cancer-specific incidence rate experienced by the general Queensland population, matched by sex, age group and time period. Standardised incidence ratios (SIRs) were then obtained by dividing the observed number of cases of second primary cancer by the expected number. The SIR is thus used to estimate the risk of a cancer patient developing a second primary malignancy relative to the incidence of cancer among the general population. Confidence intervals (CIs) for the SIRs were derived from the Poisson distribution \[[@B11]\] and calculated at the 95% level of certainty. All analyses were conducted using SAS v9.2 for Windows. Data required for this study was non-identifiable so no ethics committee approval was necessary. Results ======= The basic characteristics of the study cohort are summarised in Table [1](#T1){ref-type="table"}. Among the 204,962 eligible cancer patients, a total of 23,580 second invasive primary cancers were observed during 1,370,247 years of follow-up (median follow-up = 5.5 years per person, interquartile range = 1.3 to 10.2 years per person). In terms of absolute numbers, second primary cancers were more common among males and increased with older age, which is consistent with the distribution of first primary cancers. About one in ten (10.6%) of the second primary cancers were diagnosed within a year of the first diagnosis, while more than one in five (20.6%) were diagnosed at least 10 years afterwards. The highest proportions of second primary cancers occurred following an initial diagnosis of melanoma (21.6%), colorectal cancer (12.9%), prostate cancer (12.7%), or female breast cancer (12.6%). Relative risk of second primary cancers by sex ---------------------------------------------- Compared to the incidence of cancer in the general Queensland population, both males (SIR = 1.22; 95% CI = 1.20-1.24) and females (SIR = 1.36; 95% CI = 1.33-1.39) in the study cohort exhibited an increased risk of developing a second primary cancer (Table [2](#T2){ref-type="table"}). Significantly increased relative risks of invasive cancer were recorded among males following diagnosis of head and neck cancer, oesophageal cancer, lung cancer, melanoma, kidney cancer, bladder cancer, thyroid cancer, non-Hodgkin lymphoma, lymphoid leukaemia or myeloid leukaemia. In contrast, males initially diagnosed with either prostate or stomach cancer subsequently experienced a significantly lower risk of cancer compared to the general population. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Relative risk of second primary cancer by type of first primary cancer and sex, Queensland, 1982-2006 ::: -------------------------------------------------------------------------------------- Sex -------------------------- ---------- ------------------- ---------- ----------------- **First primary cancer** **Obs**. **SIR**\ **Obs**. **SIR**\ **(95% CI)** **(95% CI)**  Head and neck 1,620 **1.70**\ 369 **1.84**\ **(1.62-1.79)** **(1.66-2.04)**  Oesophagus 78 **1.43**\ 27 1.06\ **(1.13-1.79)** (0.70-1.54)  Stomach 150 ***0.83***\ 56 1.00\ ***(0.70-0.98)*** (0.75-1.30)  Colorectal 1,899 1.04\ 1,147 **1.08**\ (0.99-1.08) **(1.02-1.15)**  Pancreas 31 0.92\ 18 0.85\ (0.63-1.31) (0.50-1.34)  Lung 552 **1.18**\ 177 **1.41**\ **(1.08-1.28)** **(1.21-1.64)**  Melanoma 3,267 **1.77**\ 1,825 **1.70**\ **(1.71-1.83)** **(1.62-1.78)**  Breast - female n.a. n.a. 2,962 **1.31**\ **(1.27-1.36)**  Cervix n.a. n.a. 300 **1.36**\ **(1.21-1.52)**  Uterus n.a. n.a. 409 **1.16**\ **(1.05-1.28)**  Ovary n.a. n.a. 164 1.12\ (0.95-1.30)  Prostate 3,006 ***0.83***\ n.a. n.a. ***(0.80-0.86)***  Kidney 434 **1.39**\ 221 **1.69**\ **(1.27-1.53)** **(1.47-1.93)**  Bladder 1,370 **1.40**\ 310 **1.61**\ **(1.33-1.48)** **(1.43-1.79)**  Brain & CNS 31 1.05\ 19 0.94\ (0.71-1.49) (0.57-1.47)  Thyroid 70 **1.47**\ 145 **1.35**\ **(1.15-1.86)** **(1.14-1.59)**  Non-Hodgkin lymphoma 411 **1.37**\ 278 **1.36**\ **(1.24-1.51)** **(1.20-1.53)**  Lymphoid leukaemia 297 **1.66**\ 125 **1.40**\ **(1.48-1.86)** **(1.17-1.67)**  Myeloid leukaemia 87 **1.44**\ 38 **1.25**\ **(1.15-1.77)** **(0.88-1.71)**  Myeloma/PCT 93 1.04\ 41 0.89\ (0.84-1.27) (0.64-1.21)  Other 992 **1.23**\ 561 **1.37**\ **(1.16-1.31)** **(1.26-1.49)**  All cancers combined 14,388 **1.22**\ 9,192 **1.36**\ **(1.20-1.24)** **(1.33-1.39)** -------------------------------------------------------------------------------------- Obs. = Observed number of second primary cancers; SIR = standardised incidence ratio; CI = confidence interval; n.a. = not applicable; CNS = central nervous system; PCT = plasma cell tumours. SIRs shown in normal bold font indicate significantly increased risk; SIRs shown in bold italics indicate significantly decreased risk ::: Within the female cohort, the relative risk of a second cancer was higher for those diagnosed with head and neck cancer, colorectal cancer, lung cancer, melanoma, breast cancer, cervical cancer, uterine cancer, kidney cancer, bladder cancer, thyroid cancer, non-Hodgkin lymphoma, lymphoid leukaemia or myeloid leukaemia. There were no types of cancer for which female survivors had a significantly lower risk of developing a second invasive cancer in relation to the general population. Relative risk of second primary cancers by age group at diagnosis ----------------------------------------------------------------- The risk of a second malignancy was higher in comparison to the general population for each of the three age groups for all first primary cancers combined, but tended to decrease as age at first diagnosis increased (Table [3](#T3){ref-type="table"}) - 15-49 years (SIR = 1.84; 95% CI = 1.77-1.90), 50-64 years (SIR = 1.39; 95% CI = 1.36-1.42) and 65 years and older (SIR = 1.23; 95% CI = 1.20-1.25). Different patterns emerged within the various cancer-specific cohorts. An elevated relative risk across all three age groups was found following head and neck cancer, lung cancer, melanoma, female breast cancer, cervical cancer, kidney cancer, bladder cancer, non-Hodgkin lymphoma and lymphoid leukaemia. For oesophageal and colorectal cancer, significantly increased relative risks were only observed among those aged under 65 years at first diagnosis, while for pancreatic and brain cancers the risk was elevated in the younger age groups but was lower than expected for people aged 65 years and over. Consistently decreased relative risks were recorded within each age group for males with prostate cancer, although the SIR was not statistically significant for those aged 15-49 years. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Relative risk of second primary cancer by type of first primary cancer and age group at first diagnosis, Queensland, 1982-2006 ::: ----------------------------------------------------------------------------------------------------------------------------------------- Age group at first diagnosis -------------------------- ------------------------------ ----------------- ---------- ------------------- ---------- ------------------- **First primary cancer** **Obs**. **SIR**\ **Obs**. **SIR**\ **Obs**. **SIR**\ **(95% CI)** **(95% CI)** **(95% CI)**  Head and neck 223 **2.07**\ 884 **2.20**\ 882 **1.77**\ **(1.80-2.36)** **(2.06-2.35)** **(1.65-1.89)**  Oesophagus 6 **4.00**\ 39 **1.70**\ 60 1.16\ **(1.47-8.70)** **(1.21-2.32)** (0.88-1.49)  Stomach 17 1.59\ 60 1.04\ 129 0.91\ (0.93-2.55) (0.79-1.34) (0.76-1.08)  Colorectal 188 **1.58**\ 994 **1.15**\ 1,864 1.03\ **(1.36-1.82)** **(1.08-1.22)** (0.99-1.08)  Pancreas \-- \-- 26 **1.60**\ 23 ***0.61***\ **(1.05-2.35)** ***(0.39-0.92)***  Lung 52 **2.46**\ 270 **1.55**\ 407 **1.24**\ **(1.84-3.23)** **(1.37-1.74)** **(1.12-1.36)**  Melanoma 1,121 **2.05**\ 1,755 **1.80**\ 2,216 **1.74**\ **(1.94-2.18)** **(1.72-1.89)** **(1.67-1.81)**  Breast - female 689 **1.86**\ 1,047 **1.29**\ 1,226 **1.15**\ **(1.72-2.00)** **(1.21-1.37)** **(1.08-1.21)**  Cervix 129 **1.33**\ 94 **1.49**\ 77 **1.27**\ **(1.11-1.57)** **(1.21-1.83)** **(1.01-1.59)**  Uterus 47 **1.63**\ 164 1.07\ 198 **1.16**\ **(1.20-2.17)** (0.91-1.25) **(1.00-1.33)**  Ovary 33 1.25\ 64 1.12\ 67 1.06\ (0.86-1.76) (0.87-1.44) (0.82-1.34)  Prostate 7 0.81\ 554 ***0.87***\ 2,445 ***0.82***\ (0.32-1.66) ***(0.80-0.95)*** ***(0.79-0.85)***  Kidney 52 **1.63**\ 230 **1.59**\ 373 **1.57**\ **(1.22-2.14)** **(1.39-1.81)** **(1.42-1.74)**  Bladder 81 **2.02**\ 465 **1.63**\ 1,134 **1.70**\ **(1.60-2.51)** **(1.49-1.79)** **(1.60-1.80)**  Brain & CNS 27 **1.64**\ 15 0.94\ 8 ***0.48***\ **(1.08-2.39)** (0.52-1.55) ***(0.21-0.95)***  Thyroid 65 1.22\ 84 **1.39**\ 66 **1.37**\ (0.94-1.55) **(1.11-1.72)** **(1.06-1.74)**  Non-Hodgkin lymphoma 81 **1.60**\ 261 **1.57**\ 347 **1.23**\ **(1.27-1.99)** **(1.39-1.77)** **(1.10-1.36)**  Lymphoid leukaemia 27 **2.68**\ 131 **1.69**\ 264 **1.57**\ **(1.77-3.90)** **(1.42-2.01)** **(1.39-1.77)**  Myeloid leukaemia 22 **2.23**\ 33 1.39\ 70 **1.35**\ **(1.40-3.38)** (0.96-1.95) **(1.05-1.70)**  Myeloma/PCT 8 1.40\ 47 1.35\ 79 0.89\ (0.61-2.77) (0.99-1.79) (0.70-1.10)  Other 273 **1.69**\ 472 **1.43**\ 808 **1.20**\ **(1.49-1.90)** **(1.30-1.56)** **(1.12-1.29)**  All cancers combined 3,150 **1.84**\ 7,687 **1.39**\ 12,743 **1.23**\ **(1.77-1.90)** **(1.36-1.42)** **(1.20-1.25)** ----------------------------------------------------------------------------------------------------------------------------------------- Obs. = Observed number of second primary cancers; SIR = standardised incidence ratio; CI = confidence interval; CNS = central nervous system; PCT = plasma cell tumours. SIRs shown in normal bold font indicate significantly increased risk; SIRs shown in bold italics indicate significantly decreased risk. SIRs for pancreatic cancer for the age groups 15-49 years and 50-64 years have been combined due to less than 5 observed cases aged 15-49 years. ::: Relative risk of second primary cancers by time period of first diagnosis ------------------------------------------------------------------------- There was some evidence that the more recently a first primary cancer was diagnosed the higher the relative risk of a second primary cancer, with a gradual increase across the four time periods for all cancers combined (Table [4](#T4){ref-type="table"}): 1982-1986 (SIR = 1.14; 95% CI = 1.08-1.20), 1987-1991 (SIR = 1.22; 95% CI = 1.17-1.28), 1992-1996 (SIR = 1.36; 95% CI = 1.31-1.41) and 1997-2001 (SIR = 1.46; 95% CI = 1.41-1.50). In particular, the SIRs for people with colorectal cancer, lung cancer, breast cancer, thyroid cancer, non-Hodgkin lymphoma, lymphoid leukaemia and myeloid leukaemia were only significant for the later time periods. Among men with prostate cancer, the risk of developing a second primary cancer was lower than the incidence of cancer experienced by males in the general population irrespective of the year of first diagnosis, except for the latest period (1997-2001). Several cancer-specific cohorts had SIRs that were significantly higher than expected in each time period, including head and neck cancer, melanoma, kidney cancer and bladder cancer. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Relative risk of second primary cancer by type of first primary cancer and time period of first diagnosis, Queensland, 1982-2006 ::: -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Time period of first diagnosis -------------------------- -------------------------------- ------------------- ---------- ------------------- ---------- ------------------- ---------- ----------------- **First primary cancer** **Obs**. **SIR**\ **Obs**. **SIR**\ **Obs**. **SIR**\ **Obs**. **SIR**\ **(95% CI)** **(95% CI)** **(95% CI)** **(95% CI)**  Head and neck 165 **1.92**\ 204 **1.88**\ 271 **1.98**\ 255 **1.93**\ **(1.64-2.23)** **(1.63-2.16)** **(1.75-2.23)** **(1.70-2.18)**  Oesophagus 8 1.53\ 10 0.90\ 23 **1.60**\ 23 1.19\ (0.66-3.02) (0.43-1.66) **(1.01-2.39)** (0.76-1.79)  Stomach 18 0.75\ 27 1.02\ 30 0.91\ 36 1.09\ (0.44-1.18) (0.67-1.49) (0.62-1.31) (0.76-1.51)  Colorectal 198 0.89\ 273 0.95\ 411 1.07\ 559 **1.20**\ (0.77-1.02) (0.84-1.07) (0.97-1.18) **(1.11-1.31)**  Pancreas 7 1.20\ 5 0.61\ 10 0.98\ 12 0.83\ (0.48-2.47) (0.20-1.42) (0.47-1.81) (0.43-1.45)  Lung 71 1.06\ 93 1.14\ 147 **1.61**\ 148 **1.30**\ (0.83-1.34) (0.92-1.40) **(1.36-1.89)** **(1.10-1.53)**  Melanoma 311 **1.81**\ 471 **1.89**\ 639 **2.04**\ 835 **2.02**\ **(1.62-2.02)** **(1.72-2.07)** **(1.88-2.20)** **(1.89-2.16)**  Breast - female 152 0.97\ 199 0.94\ 359 **1.32**\ 504 **1.37**\ (0.82-1.14) (0.81-1.08) **(1.19-1.47)** **(1.25-1.50)**  Cervix 24 1.29\ 38 **1.88**\ 34 **1.64**\ 26 **1.55**\ (0.83-1.92) **(1.33-2.58)** **(1.14-2.29)** **(1.01-2.27)**  Uterus 30 1.14\ 35 1.18\ 35 0.87\ 63 1.30\ (0.77-1.63) (0.82-1.64) (0.60-1.21) (0.99-1.66)  Ovary 16 1.36\ 10 0.64\ 26 1.35\ 24 1.02\ (0.78-2.20) (0.31-1.18) (0.88-1.98) (0.66-1.52)  Prostate 163 ***0.64***\ 291 ***0.76***\ 603 ***0.84***\ 655 0.95\ ***(0.54-0.74)*** ***(0.68-0.85)*** ***(0.77-0.91)*** (0.88-1.02)  Kidney 44 **1.48**\ 73 **1.79**\ 92 **1.53**\ 160 **2.04**\ **(1.08-1.99)** **(1.40-2.25)** **(1.24-1.88)** **(1.74-2.38)**  Bladder 137 **1.46**\ 189 **1.78**\ 238 **1.84**\ 353 **2.22**\ **(1.22-1.72)** **(1.53-2.05)** **(1.61-2.09)** **(2.00-2.47)**  Brain & CNS \-- \-- 5 1.00\ \-- \-- 9 1.10\ (0.33-2.34) (0.50-2.09)  Thyroid 9 1.25\ 12 1.25\ 29 **1.64**\ 46 **1.67**\ (0.57-2.38) (0.64-2.18) **(1.10-2.36)** **(1.22-2.23)**  Non-Hodgkin lymphoma 42 1.09\ 62 1.15\ 88 **1.25**\ 144 **1.49**\ (0.79-1.48) (0.88-1.48) **(1.00-1.54)** **(1.25-1.75)**  Lymphoid leukaemia 24 1.00\ 33 1.25\ 63 **1.68**\ 88 **1.71**\ (0.64-1.48) (0.86-1.75) **(1.29-2.15)** **(1.37-2.10)**  Myeloid leukaemia 13 1.23\ 11 0.91\ 26 1.52\ 34 **2.05**\ (0.66-2.11) (0.45-1.62) (0.99-2.22) **(1.42-2.86)**  Myeloma/PCT 21 1.37\ 7 ***0.46***\ 19 0.83\ 39 1.05\ (0.85-2.10) ***(0.19-0.95)*** (0.50-1.29) (0.75-1.44)  Other 79 0.94\ 136 **1.22**\ 241 **1.43**\ 318 **1.43**\ (0.74-1.17) **(1.03-1.45)** **(1.26-1.62)** **(1.28-1.60)**  All cancers combined 1,534 **1.14**\ 2,184 **1.22**\ 3,386 **1.36**\ 4,331 **1.46**\ **(1.08-1.20)** **(1.17-1.28)** **(1.31-1.41)** **(1.41-1.50)** -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Obs. = Observed number of second primary cancers; SIR = standardised incidence ratio; CI = confidence interval; CNS = central nervous system; PCT = plasma cell tumours. Analysis by period of first diagnosis was restricted to five years of follow-up to allow more consistent comparisons across time. SIRs shown in normal bold font indicate significantly increased risk; SIRs shown in bold italics indicate significantly decreased risk. Estimates are not shown for Brain & CNS for the periods 1982-1986 and 1992-1996 due to observed cell counts of less than 5. ::: Relative risk of second primary cancers by follow-up interval ------------------------------------------------------------- The risk of a second primary cancer for all survivors combined remained consistently elevated compared to the general population during each follow-up interval (Table [5](#T5){ref-type="table"}): 2 months to less than 1 year after first diagnosis (SIR = 1.32; 95% CI = 1.27-1.37), 1 year to less than 5 years (SIR = 1.33; 95% CI = 1.31-1.36), 5 years to less than 10 years (SIR = 1.39; 95% CI = 1.35-1.42) and 10 years or longer (SIR = 1.28; 95% CI = 1.24-1.31). A significantly increased relative risk across all follow-up intervals was also observed following a diagnosis of head and neck cancer, melanoma, kidney cancer, bladder cancer or lymphoid leukaemia, while the risks were significantly higher from 1 year or longer after diagnosis for colorectal cancer, lung cancer, female breast cancer or non-Hodgkin lymphoma. The relative risk of developing another cancer was significantly higher among cervical cancer survivors up to 10 years after the initial diagnosis but not after that. Prostate cancer patients were found to have subsequent risks that remained lower than the matching population, particularly 10 or more years after initial diagnosis. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Relative risk of second primary cancer by type of first primary cancer and follow-up interval, Queensland, 1982-2006 ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------- Follow-up interval -------------------------- -------------------- ------------------- ---------- ------------------- ---------- ------------------- ---------- ------------------- **First primary cancer** **Obs**. **SIR**\ **Obs**. **SIR**\ **Obs**. **SIR**\ **Obs**. **SIR**\ **(95% CI)** **(95% CI)** **(95% CI)** **(95% CI)**  Head and neck 164 **1.67**\ 731 **2.00**\ 651 **2.10**\ 443 **1.89**\ **(1.42-1.95)** **(1.86-2.15)** **(1.94-2.26)** **(1.72-2.08)**  Oesophagus 22 1.22\ 42 1.31\ 32 **1.76**\ 9 1.11\ (0.77-1.85) (0.95-1.77) **(1.20-2.48)** (0.51-2.10)  Stomach 36 0.97\ 75 0.95\ 55 1.04\ 40 0.99\ (0.68-1.35) (0.74-1.18) (0.78-1.35) (0.71-1.35)  Colorectal 310 1.03\ 1,131 **1.07**\ 912 **1.11**\ 693 **1.14**\ (0.92-1.16) **(1.01-1.13)** **(1.04-1.18)** **(1.06-1.23)**  Pancreas 14 0.74\ 20 1.01\ 15 0.99\ \-- \-- (0.40-1.24) (0.62-1.57) (0.55-1.64)  Lung 147 1.06\ 312 **1.46**\ 183 **1.71**\ 87 **1.35**\ (0.89-1.24) **(1.30-1.63)** **(1.47-1.98)** **(1.08-1.67)**  Melanoma 458 **2.15**\ 1,798 **1.92**\ 1,553 **1.80**\ 1,283 **1.63**\ **(1.95-2.35)** **(1.84-2.02)** **(1.71-1.89)** **(1.55-1.73)**  Breast - female 194 1.02\ 1,020 **1.25**\ 1,019 **1.44**\ 729 **1.36**\ (0.88-1.18) **(1.17-1.33)** **(1.35-1.53)** **(1.26-1.46)**  Cervix 33 **2.13**\ 89 **1.46**\ 89 **1.43**\ 89 1.09\ **(1.46-2.99)** **(1.18-1.80)** **(1.15-1.76)** (0.87-1.34)  Uterus 29 1.03\ 134 1.15\ 147 **1.36**\ 99 1.00\ (0.69-1.48) (0.96-1.36) **(1.15-1.60)** (0.81-1.21)  Ovary 22 1.22\ 54 1.04\ 46 1.18\ 42 1.11\ (0.77-1.85) (0.78-1.35) (0.86-1.57) (0.80-1.50)  Prostate 344 ***0.81***\ 1,368 ***0.84***\ 988 ***0.88***\ 306 ***0.66***\ ***(0.73-0.90)*** ***(0.80-0.89)*** ***(0.82-0.93)*** ***(0.59-0.74)***  Kidney 115 **2.45**\ 254 **1.57**\ 177 **1.42**\ 109 **1.35**\ **(2.02-2.94)** **(1.38-1.77)** **(1.22-1.65)** **(1.11-1.63)**  Bladder 249 **2.40**\ 668 **1.73**\ 477 **1.61**\ 286 **1.39**\ **(2.11-2.72)** **(1.61-1.87)** **(1.47-1.76)** **(1.23-1.56)**  Brain & CNS 6 0.60\ 12 0.77\ 18 1.44\ 14 1.28\ (0.22-1.31) (0.40-1.34) (0.85-2.28) (0.70-2.15)  Thyroid 18 1.65\ 78 **1.53**\ 63 1.21\ 56 1.16\ (0.98-2.61) **(1.21-1.91)** (0.93-1.55) (0.87-1.50)  Non-Hodgkin lymphoma 74 1.23\ 262 **1.31**\ 209 **1.46**\ 144 **1.49**\ (0.97-1.55) **(1.16-1.48)** **(1.27-1.67)** **(1.26-1.76)**  Lymphoid leukaemia 52 **1.82**\ 156 **1.40**\ 143 **1.89**\ 71 **1.78**\ **(1.36-2.39)** **(1.19-1.64)** **(1.59-2.23)** **(1.37-2.22)**  Myeloid leukaemia 27 **1.59**\ 57 **1.44**\ 28 1.40\ 13 1.41\ **(1.05-2.31)** **(1.09-1.87)** (0.93-2.03) (0.75-2.41)  Myeloma/PCT 18 0.76\ 68 1.02\ 36 1.26\ 12 1.12\ (0.45-1.21) (0.79-1.29) (0.88-1.74) (0.58-1.95)  Other 179 **1.22**\ 595 **1.35**\ 447 **1.36**\ 332 **1.33**\ **(1.05-1.42)** **(1.25-1.47)** **(1.24-1.49)** **(1.19-1.48)**  All cancers combined 2,511 **1.32**\ 8,924 **1.33**\ 7,286 **1.39**\ 4,859 **1.28**\ **(1.27-1.37)** **(1.31-1.36)** **(1.35-1.42)** **(1.24-1.31)** ---------------------------------------------------------------------------------------------------------------------------------------------------------------- Obs. = Observed number of second primary cancers; SIR = standardised incidence ratio; CI = confidence interval; CNS = central nervous system; PCT = plasma cell tumours. SIRs shown in normal bold font indicate significantly increased risk; SIRs shown in bold italics indicate significantly decreased risk. SIRs for pancreatic cancer for 5 years to less than 10 years after first diagnosis and 10 years or longer after first diagnosis have been combined due to less than 5 observed cases 10 years or longer after first diagnosis. ::: Relative risk by type of first and second primary cancers and sex ----------------------------------------------------------------- The relative risks for specific second primary cancers varied substantially according to the type of first primary cancer (Figure [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}). Within the melanoma cohort (Figure [1c](#F1){ref-type="fig"} and [1d](#F1){ref-type="fig"}), both males and females were over six times more likely to be diagnosed with another primary melanoma compared to the general population. They also had significantly increased relative risks for several other cancers, including thyroid cancer and lymphoid leukaemia (both males and females), brain cancer, non-Hodgkin lymphoma, prostate cancer and colorectal cancer (males only) and kidney cancer and breast cancer (females only). However, lung cancers occurred less often than expected among males with a first primary melanoma. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Relative risk following all cancers combined, melanoma or colorectal cancer by type of second primary cancer and sex, Queensland, 1982-2006**. CNS: central nervous system; PCT: plasma cell tumour.Vertical black line indicates SIR point estimate; grey shading indicates SIR 95% confidence interval. X-axes are shown on a log scale. :::  ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Relative risk following prostate cancer, breast cancer, head and neck cancer or lung cancer by type of second primary cancer and sex, Queensland, 1982-2006**. CNS: central nervous system; PCT: plasma cell tumour. Vertical black line indicates SIR point estimate; grey shading indicates SIR 95% confidence interval. The SIR for a second primary prostate cancer following a first primary prostate cancer was 0.3 and is not shown in Figure 2a. No cases of lymphoid leukaemia following head and neck cancer were recorded for females (Figure 2d). X-axes are shown on a log scale. Note different scale endpoints for x-axis used in Figures 2d, 2e and 2f. :::  ::: An elevated relative risk of melanoma was also observed for both sexes in the colorectal cancer cohort (Figure [1e](#F1){ref-type="fig"} and [1f](#F1){ref-type="fig"}). In addition, male colorectal cancer survivors had higher risks for lymphoid leukaemia, oesophageal cancer and kidney cancer but a lower risk of developing a second primary colorectal cancer compared to the general population, while females with colorectal cancer experienced a subsequent increased risk of breast cancer. Men diagnosed with prostate cancer had significantly increased relative risks for thyroid cancer, melanoma, bladder cancer, kidney cancer, non-Hodgkin lymphoma and colorectal cancer but a significantly decreased risk of lung cancer (Figure [2a](#F2){ref-type="fig"}). Instances of primary prostate cancer occurring twice in the same person were rare. Female breast cancer survivors were more likely to be diagnosed with uterine cancer, myeloid leukaemia, stomach cancer, breast cancer, ovarian cancer, melanoma, kidney cancer or colorectal cancer than were the general population (Figure [2b](#F2){ref-type="fig"}). People initially diagnosed with head and neck cancer were found to have elevated relative risks for a second cancer of the head and neck, oesophageal cancer, lung cancer and non-Hodgkin lymphoma (Figure [2c](#F2){ref-type="fig"} and [2d](#F2){ref-type="fig"}). Males with head and neck cancer were also at increased risk of developing melanoma, colorectal cancer or bladder cancer. All lung cancer patients had a significantly increased risk of oesophageal and head and neck cancers compared to other residents of Queensland (Figure [2e](#F2){ref-type="fig"} and [2f](#F2){ref-type="fig"}), while males with lung cancer experienced high relative risks of being subsequently diagnosed with kidney, pancreas or bladder cancers and females with lung cancer had a significantly increased risk of developing a second primary lung cancer. Lymphoid leukaemia occurred less often among males following lung cancer than in the general population, although both the observed and expected number of cases were small. Discussion ========== Cancer patients in our study cohort were at significantly higher risk of a second diagnosis compared to the underlying incidence rates experienced in the entire population of Queensland. Although there was some variation in the size of the estimated relative risks, a consistent pattern of increased risk following all first primary cancers combined was seen for both males and females, and across all age groups at first diagnosis, time periods of first diagnosis and follow up intervals. Our data offers further evidence of significant associations between particular types of first and second primary cancers that have been previously documented elsewhere in Australia and around the world \[[@B6],[@B12]-[@B16]\]. Some examples include the mutual increased risks between head and neck, lung and oesophageal cancers and the relationship between melanoma and both prostate and female breast cancers. Much of the elevated risk of being diagnosed with a second malignancy can be attributed to risk behaviours (such as smoking, harmful levels of alcohol consumption and poor diet), inherited susceptibilities and/or the medical treatment that cancer survivors have received \[[@B3],[@B4],[@B6]\]. On occasions, treatment can also have the opposite effect of reducing the risk of subsequent diagnosis. As was noted in another recent study of second primary cancer \[[@B13]\], the overall lower SIRs observed among prostate cancer survivors are mainly due to extremely low rates of reoccurrence as the organ is often completely removed as part of treatment. Given that prostate cancer is the most common cancer diagnosed for males in Queensland, this also helps to explain why the total relative risk is lower for male survivors compared to female survivors. Another potential explanation for the difference in risk between cancer survivors and the general population is that some demographic factors may not be comparable between the two groups. Although the relative risks were based on calculations matched by sex, age group and time period, it is possible that other qualities, such as socioeconomic status, may vary between people who have been diagnosed with cancer and those who have not. This is likely to be the reason for at least part of the reduced risk of lung cancer (which has higher incidence among lower socioeconomic groups) for males following melanoma (which is more common in segments of the community with higher socioeconomic status) \[[@B6],[@B17]\]. A similar explanation could be used for the deficit of lung cancer after a diagnosis of prostate cancer. In contrast, the high reciprocal relative risks of melanoma with both female breast cancer and prostate cancer may correlate with the higher incidence for each of these malignancies among more affluent populations \[[@B18]-[@B20]\]. One of the interesting relationships that emerged from the analysis was the variation in relative risk of second primary cancer by age at first diagnosis following cancer of the brain and central nervous system. Younger survivors (15-49 years old) had an increased risk of being diagnosed with a second primary cancer while survivors aged 65 years or over had a decreased risk. This is most likely due to the various histopathological subtypes of brain tumours which are more common in the different age groups. For example, glioblastoma tends to be diagnosed at an older age and is associated with poor survival, allowing a limited time for treatment-related second primary cancers to appear \[[@B6]\]. The reasons behind an increased relative risk of second primary cancer following certain types of cancer among survivors who were diagnosed more recently are unknown. Tsukuma *et al.*\[[@B16]\] described a similar pattern in Japan, and suggested that apparent increases in risk may be due to improved follow-up and surveillance of cancer patients. Another possible cause is the change in treatment modalities over time \[[@B4],[@B21]\]. Second primary cancers that arise due to the effects of treatment for the initial cancer might be expected to occur many years after the first diagnosis. However, the increase in relative risk remained fairly consistent irrespective of time since diagnosis, in accordance with results published elsewhere in Australia \[[@B13]\] and the United States \[[@B6]\]. The main strengths of this study include the extensive population-based coverage achieved by the QCR for the reporting of cancers among Queensland residents, combined with a high level of histological verification (88% in 2006) \[[@B8]\], which is important for distinguishing between new primary cancers and metastases of an existing cancer. Since all data used in this study have been collected prospectively for administrative purposes, and coded independently of the hypotheses, the opportunity for recall or information bias has been removed. Increased medical surveillance of newly-diagnosed cancer patients may introduce a detection bias for second primary cancers. The likelihood of this happening has been reduced by only considering metachronous primary cancers, with a two-month window between first and second diagnosis. It is also possible that some second primary cancers were incorrectly classified as first primary cancers, especially for cancers diagnosed soon after the establishment of the QCR in the early 1980s; we are unable to quantify the impact of this on the observed results. While acknowledging that the study cohort and comparison population were not independent, with the population containing people already diagnosed with cancer, this proportion was less than 0.5% in any given year. Finally, as a result of the large number of comparisons made, the possibility that some of the SIR estimates have been spuriously identified as statistically significant needs to be considered, particularly those based on small numbers of primary or secondary cancers, and these results should therefore be interpreted with due caution. Conclusions =========== Our results demonstrate that cancer survivors in Queensland, Australia, like those in other countries, are confronted by a very real, ongoing risk of developing a second primary cancer that is significantly higher than the incidence of cancer experienced by the general population. Some first and second primary cancers share common aetiologies, making it imperative that cancer patients adopt a healthier lifestyle in order to lessen their chances of subsequent diagnoses \[[@B4],[@B22]\]. Recent studies have reported little difference in the health behaviours of cancer survivors compared to the wider community \[[@B23]-[@B25]\], suggesting that health promotion efforts may need to be specifically targeted at people diagnosed with cancer due to their subsequent increased risk. Further work is needed to determine exactly how beneficial changes to lifestyle may be in regard to the risk of developing a second primary cancer \[[@B26]\]. Even so, some second primary cancers will be unavoidable, and as the number of cancer survivors continues to grow, the importance of ongoing medical supervision and screening to detect second primary cancers at an earlier stage and thereby improve the effectiveness of treatment will remain critical. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= DY conducted the statistical analysis and drafted the manuscript. PB conceived the study and edited the draft manuscript. Both authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/11/83/prepub> Acknowledgements ================ The authors wish to thank the staff working in the Queensland Cancer Registry who provided the data extract used in this analysis.

PubMed Central

2024-06-05T04:04:19.033732

2011-2-23

PMC3052209

Background ========== Accumulating knowledge suggests that domestic violence occurring during pregnancy is a serious public health issue due to the risk for adverse maternal and fetal health outcomes \[[@B1]-[@B3]\]. Labour dystocia, another serious complication in obstetrics, has also been increasingly highlighted during the past decades \[[@B4]-[@B9]\]. Labour dystocia is defined as a slow or difficult labour or childbirth. According to Kjaergaard et al. \[[@B10]\] the term \'dystocia\' is frequently used in clinical practice, yet there is no consistency in the use of terminology for prolonged labour or labour dystocia \[[@B4],[@B6],[@B11],[@B12]\]. However, labour dystocia accounts for most interventions during labour \[[@B4],[@B6],[@B7]\]. Although both labour dystocia \[[@B4],[@B7]\] and domestic violence during pregnancy \[[@B1],[@B2]\] are associated with adverse maternal and fetal outcome, evidence in support of a possible association between experiences of violence and labour dystocia is sparse. One recent study from Iran has shown an association between experienced abuse by an intimate partner and labour dystocia, and such abuse included psychological threats as well as physical, or sexual abuse \[[@B13]\]. Although the demographic background of women exposed to domestic violence may vary widely, some women are more vulnerable and at increased risk \[[@B14]\]. Disadvantaged women, with low socio-economic status \[[@B15]-[@B17]\] and younger age, \[[@B18]\] as well as single women at younger age, \[[@B15]-[@B17]\] certain ethnic groups \[[@B15],[@B17],[@B19]\] and even women with a partner born outside Europe \[[@B17]\] are more likely to be exposed to domestic violence. Also unhealthy maternal behaviour such as smoking \[[@B20]-[@B23]\] and use of alcohol and drugs during pregnancy are more common among women who live in violent relationships \[[@B20],[@B21]\]. Pregnant women exposed to violence have a greater risk of delivering babies with low birth weight, \[[@B20],[@B22],[@B24]\] premature labour, \[[@B22],[@B25]\] abruption of placenta \[[@B25]\] and fetal trauma \[[@B22],[@B24],[@B25]\] or death \[[@B22],[@B24],[@B26]\] and are also at increased risk of caesarean section \[[@B25]\]. Some identified risk factors for dystocia are high maternal age, \[[@B10],[@B11]\] short maternal height, \[[@B27],[@B28]\] overweight, \[[@B10]\] obesity \[[@B29]\] and smoking \[[@B30]\]. Also, high fetal weight increases the risk for prolonged labour \[[@B31]\] and labour dystocia \[[@B32]\]. Further, up to 50% of unplanned caesarean sections among nulliparous women are related to labour dystocia \[[@B4],[@B6]\]. Already thirty years ago, Lederman et al. \[[@B33]\] showed that physical and psychosocial characteristics of the woman, such as maternal emotional stress related to pregnancy and motherhood, partner and family relationships, and fears of labour were significantly associated with less efficient uterine function, higher state of anxiety, higher epinephrine levels in plasma and longer length of labour. The higher levels of epinephrine may disrupt the normal progress in labour or the coordinated uterine contractions explained by an adrenoreceptor theory \[[@B34]\]. Subsequently, Alehagen et al. \[[@B35]\] confirmed significantly increased levels of all three stress hormones from pregnancy to labour and drastically increased levels of epinephrine and cortisol compared with nor-epinephrine, indicating that mental stress is more dominant than physical stress during labour. Maternal psychosocial stress, family functioning and fear of childbirth may have an association with specific complications such as prolonged labour or caesarean section \[[@B36]\]. History of sexual violence in adult life is associated with an increased risk of extreme fear during labour, \[[@B37]\] and fear of childbirth in the third trimester has been shown to increase the risk of prolonged labour and emergency caesarean section \[[@B38]\]. Thus, the current body of evidence in this area would support the hypothesis that experience of violence before and/or during pregnancy increases the risk of labour dystocia. The aim of this study was to investigate whether self-reported history of violence or experienced violence during pregnancy is associated with increased risk of labour dystocia in nulliparous women at term. Methods ======= The material used in this study originates from the Danish Dystocia Study (DDS), a population-based multi-centre cohort study, and 8099 nulliparous women were potentially eligible for inclusion in the study \[[@B8]-[@B10]\]. However, 6356 women were invited to the DDS study (external drop-out was 21.5%) and 5484 women accepted participation. For the current sub-study, a data set on 2652 nulliparous women who fulfilled the inclusion criteria (showed below) was available for analyses of exposure to violence before and during pregnancy. Among these, 985 (37.1%) met the protocol criteria for labour dystocia (Table [1](#T1){ref-type="table"}). These diagnostic criteria are in accordance with the American College of Obstetrics and Gynecology (ACOG) criteria for dystocia in labour\'s second stage \[[@B6]\] and also with the criteria for labour dystocia in first and second stage described by the Danish Society for Obstetrics and Gynecology \[[@B39],[@B40]\]. The diagnosis prompted augmentation (i.e. with oxytocin stimulation) \[[@B8]-[@B10]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Definition of stages and phases of labour and diagnostic criteria for dystocia for current sub-study \[[@B8]-[@B10]\]. ::: Stage of labour Definition of stages and phases Diagnostic criteria for dystocia -------------------- --------------------------------------------------------------------------- -------------------------------------------------------------------- ***First stage*** ***From onset of regular contractions leading to cervical dilatation*** Latent phase Cervix dilatation 0 - 3.9 cm Not given in this phase Active phase Cervix dilatation ≥ 4 cm \< 2 cm assessed over four hours ***Second stage*** ***From full dilatation of cervix until the baby is borne*** Descending phase From full dilatation of cervix to strong and irresistible urge to push No descending ≥ 2 hours or ≥ 3 hours if epidural was administrated Expulsive phase Strong and irresistible pushing during the major part of the contractions No progress ≥ 1 hour ::: Data were collected prospectively between May 2004 and July 2005. Participants were recruited from nine obstetric departments in Denmark with annual birth rates between 850-5400 per year. The departments were four large university hospitals, three county hospitals, and two local district departments. Recruitment of the women took place in the antenatal clinics at 33 gestational weeks, and baseline information was collected at 37 gestational weeks. *Inclusion criteria*were Danish speaking (i.e. reading/understanding) nulliparous women at 18 years of age or older, with a singleton pregnancy in cephalic presentation and no planned elective caesarean section or induction of labour. *Exclusion criteria*were nulliparous women with a delivery \< 37 or \> 42 weeks of gestation, induction, elective caesarean section and breech presentation (n = 1115 or 17.5% in DDS). All data were based on a self-administrated questionnaire and on information contained in obstetric records filled out by the midwives at admission and postpartum. Forty percent of the questionnaires were completed in an internet version. Fourteen (0.5%) of the 2652 women did not answer the questions about violence and were classified as having no exposure to violence. Eight items in the questionnaire dealt with violence and originated from the short form of the Conflict Tactics Scale (CTS2S) \[[@B41]\]. This instrument has been used in large population-based studies in Denmark, and translation from English to Danish and back translation to English were performed prior to the Danish Health and Morbidity survey 2000 \[[@B42]\]. The questions were adapted for a pregnant cohort in the DDS \[[@B8]-[@B10]\]. Three alternatives were provided as possible answers to the various exposure questions: \'yes during this pregnancy\', \'yes earlier\', and \'no never\'. Women were not required to provide information concerning the number of episodes of violence that had occurred (Additional file [1](#S1){ref-type="supplementary-material"}). \'*History of violence\'*was defined as experience of violence ever in lifetime before and/or during pregnancy, *\'Violence before pregnancy\'*as experienced violence ever in lifetime before pregnancy, \'*Violence during pregnancy\'*as experienced violence during pregnancy (with or without violence before pregnancy) and \'*Violence for the first time during pregnancy\'*as experienced violence during pregnancy without experienced violence before pregnancy. Further, for the purpose of analysis, violence was categorized as i) threat of violence, ii) physical violence, iii) sexual violence, and iv) serious violence. However, a more detailed description of the prevalence of violence will be published elsewhere by another research group. For the purpose of the current sub-study, the concept *domestic violence*was defined as exposure to psychological and/or physical abuse by \'Your husband/Co-habitant\' or \'A person you know very well in your family\', according to the first two alternatives in question 9 in the questionnaire (Additional file [1](#S1){ref-type="supplementary-material"}). Background and lifestyle factors were classified as follows. *Maternal age*was classified as 18-24, 25-29, 30-34 and \>34 years. *Country of origin*was classified according to whether the woman was born in Denmark, in another Nordic country, or in other country. *Cohabiting status*was divided into yes or no. *Educational status*was dichotomised as ≤ 10 years or \> 10 years and *employment status*as employed or unemployed (including voluntary unemployed or studying). *Smokingstatus*was classified as \"yes\" (if the woman was a daily smoker or was smoking at some point during pregnancy) or \"no\" (never smoked or alternatively, if she had ceased before pregnancy) and *use of alcohol*as \"yes\" (if the woman had been drinking alcohol during pregnancy at the time when the questionnaire was administered) or \"no\" (if the woman had been drinking solely alcohol-free drinks). *Body mass index*(BMI) was calculated from maternal weight and height before the pregnancy and classified as normal or low weight if BMI was ≤ 25, or overweight when \> 25. Infant *birth weight*was dichotomised as \< 3500 g or ≥ 3500 g and *deliverymode*as partus normalis (PN) or instrumental delivery, including caesarean section and vacuum extraction (VE). Ethics ------ Since no invasive procedures were applied in the study, no Ethics Committee System approval was required by Danish law. The policy of the Helsinki Declaration was followed throughout the data collection and analyses. Written consent was obtained and person-specific data were protected by codes. Permission to establish the database was obtained from the Danish Data Protection Agency (j. no. 2004-41-3995). Statistical methods ------------------- Chi-square analysis was used to investigate differences in background characteristics between women who were exposed to violence and women not exposed to violence. Odds ratios (OR) and 95% confidence intervals (95% CI) were calculated for the crude associations between various background- and lifestyle characteristics and labour dystocia, with dystocia as the dependent variable for logistic regression. Age was dichotomised as ≤ 24 or \>24 years and country of origin as Danish or non-Danish. Univariate logistic regression was used to analyse the crude odds ratios for dystocia in relation to various background- and lifestyle characteristics and self-reported history of violence. Further, multiple regression was used to analyse domestic violence (solely) and history of violence as independent variables (two different analysis) together with the other well-documented maternal factors (maternal age, BMI and smoking) associated with dystocia. Odds ratios were used as estimates of relative risk. Statistical significance was accepted at p \< 0.05. Statistical analyses were performed using the Statistical Package for Social Sciences (SPSS) version 16.0 for Windows. Results ======= Table [2](#T2){ref-type="table"} provides a descriptive overview of the maternal characteristics for the total cohort of 2652 women, with and without self-reported experience of \'history of violence\', \'violence before pregnancy\' and \'violence during pregnancy\'. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Descriptive overview of maternal characteristics in nulliparous women who have reported experienced violence **before and/or during pregnancy**compared to women not exposed to violence (n = 2652). ::: --------------------- -------------- ------------------------- --------------- ---------------------- ------------------ ---------------------- --------------- ------------------ -------------- ---------- **Characteristics** **Total** **History of violence** **P**\ **Violence**\ **P**\ **Violence**\ **P**\ **(2-sided)** **before pregnancy** **(2-sided)** **during pregnancy** **(2-sided)** \ **Not exposed**\ **Exposed**\ **Not exposed**\ **Exposed**\ **Not exposed**\ **Exposed**\ **n (%)** **n (%)** **n (%)** **n (%)** **n (%)** **n (%)** **n (%)** **Total** 2652 (100.0) 1712 (64.7) 940 (35.4) 1738 (65.5) 914 (34.5) 2586 (97.5) 66 (2.5) **Age, years** 18 - 24 440 (16.5) 233 (13.6) 207 (22.1) \< 0.001 236 (13.6) 204 (22.4) \< 0.001 420 (16.3) 20 (30.8) 0.02 25 - 29 1300 (49.0) 884 (51.6) 416 (44.4) 901 (51.8) 399 (43.8) 1274 (49.3) 26 (40.0) 30 - 34 728 (27.5) 476 (27.8) 252 (26.9) 481 (27.7) 247 (27.1) 712 (27.6) 16 (24.6) \> 34 180 (6.8) 119 (7.0) 61 (6.5) 120 (6.9) 60 (6.6) 177 (6.7) 3 (4.6) *Missing* 4 (0.2) **Country of**\ **origin** Denmark 2452 (92.5) 1577 (92.1) 875 (93.1) NS 1603 (92.2) 849 (92.9) NS 2390 (92.4) 62 (93.9) NS Nordic countries 54 (2.0) 38 (2.2) 16 (1.7) 38 (2.2) 16 (1.8) 53 (2.0) 1 (1.5) Other countries 146 (5.5) 97 (5.7) 49 (5.2) 97 (5.6) 49 (5.4) 143 (5.5) 3 (4.5) *Missing* 0 (0.0) **Cohabiting**\ **status** Yes 2517 (94.9) 1645 (99.7) 872 (98.8) 0.004 1668 (99.7) 849 (98.7) 0.003 2461 (99.5) 56 (94.9) \< 0.001 No 16 (0.6) 5 (0.3) 11 (1.2) 5 (0.3) 11 (1.3) 13 (0.5) 3 (5.1) *Missing* 119 (4.5) **Education**\ **status** \> 10 years 2128 (80.3) 1436 (84.7) 692 (74.7) \< 0.001 1457 (84.6) 671 (74.6) \< 0.001 2083 (81.5) 45 (68.2) \< 0.006 ≤10 years 494 (18.6) 260 (15.3) 234 (25.3) 265 (15.4) 229 (25.4) 473 (18.5) 21 (31.8) *Missing* 30 (1.1) **Employment**\ **status** Employed 1849 (69.7) 1237 (74.1) 612 (66.8) \< 0.001 1255 (74.0) 594 (66.7) \< 0.001 1805 (71.6) 44 (68.8) NS Unemployed 737 (27.8) 433 (25.9) 304 (33.2) 441 (26.0) 296 (33.3) 717 (28.4) 20 (31.2) *Missing* 66 (2.5) **Smoking** No 1995 (75.2) 1377 (80.9) 618 (65.9) \< 0.001 1396 (80.8) 599 (65.7) \< 0.001 1953 (75.9) 42 (63.6) 0.022 Yes 645 (24.3) 325 (19.1) 320 (34.1) 332 (19.2) 313 (34.3) 621 (24.1) 24 (36.4) *Missing* 12 (0.5) **Use of alcohol** No 1895 (71.5) 1240 (77.0) 655 (73.3) NS 1256 (76.7) 639 (73.6) NS 1851 (75.0) 44 (69.8) NS Yes 637 (24.0) 398 (24.3) 239 (26.7) 408(24.5) 229 (26.4) 618 (25.0) 19 (30.2) *Missing* 120 (4.5) **BMI** Normal or low\ 1954 (73.7) 1261 (77.5) 693 (77.9) NS 1282 (77.6) 672 (77.8) NS 1902 (77.6) 52 (80.0) NS (≤ 25) Overweight\ 563 (21.2) 366 (22.5) 197 (22.1) 371 (22.4) 192 (22.2) 550 (22.4) 13 (20.0) (\> 25) *Missing* 135 (5.1) --------------------- -------------- ------------------------- --------------- ---------------------- ------------------ ---------------------- --------------- ------------------ -------------- ---------- Statistical significance is accepted at p \< 0.05 ^†^Same women can occur in more than one group ::: Among the 940 (35.4%) women who reported experience of \'history of violence\', 914 (97.2%) reported experienced \'violence before pregnancy\'. Also, 66 (2.5%) women reported violence during current pregnancy (Table [2](#T2){ref-type="table"}). Of these women, 26 (39.5%) were exposed to \'violence for the first time during pregnancy\'. All women exposed to violence for the first time during their first pregnancy were Danish, three (11.5%) women in the age group 18 - 24 years, 17 (65.4%) at age 25- 29, five (19.2%) at age 30-34 and one (3.8%) \>34 years. Three (11.5%) women were not cohabiting, five (19.2%) had ≤ 10 years education, eight (30.8%) were unemployed, seven women were smokers (26.9%), ten (38.4%) were alcohol consumers at the 37^th^week of gestation, and five (19.2) had BMI \> 25. Of the 940 women who had a \'history of violence\', 697 (77%) answered a question concerning whom the perpetrator was. Thirty-seven percent had been exposed to domestic violence. Further, 22% to violence by someone they knew very well (not family member) and 15% by someone they knew superficially (family or other). The perpetrator was a stranger in 26% of the cases. Of the 66 women who had been exposed to violence during pregnancy, 53 (80%) answered the question about the perpetrator, and in 23 (43.0%) cases they were exposed to domestic violence. The median age of all nulliparous women was 28 years. In the violence-exposed group significantly more women were in the 18-24 age categories in all three violence exposure groups (p \< 0.001, p \< 0.001, p = 0.020). No differences in exposure to violence were found in relation to country of origin. In the total sample, 94.9% of the women (n = 2517) were cohabiting. Across all categories of exposure to violence, such exposure was proportionally more often reported by non-cohabiting women (p = 0.004, p = 0.003 respectively p \< 0.001) albeit only 16 (0.6%) of the women were not cohabiting. Slightly more than eighty percent (80.3%) of the women had more than 10 years of schooling. Exposure to \'history of violence\' and \'violence before pregnancy\' was more frequently reported by women who had a lower educational level (≤ 10 years) compared to women not exposed (p \< 0.001), as well as in the group \'violence during pregnancy\' (p \< 0.006). Over two-thirds (69.7%) of the women were employed. The exposed group differed from the non-exposed group before pregnancy in that more women were unemployed (p \< 0.001). However, there was no significant difference in employment status among the group of 66 (2.5%) women who were violence-exposed during pregnancy (Table [2](#T2){ref-type="table"}). More than twenty-four percent (24.3%) of these nulliparous women were smokers at term or at some point during pregnancy. Exposure to violence was proportionally more often reported by smokers than by non-smokers across all categories (p \< 0.001, p \< 0.001, p = 0.022). Twenty-four percent of the nulliparous reported that they consumed alcohol during pregnancy, in 37^th^week of pregnancy (Table [2](#T2){ref-type="table"}). The quantity ranged between 1 to 10 units of alcoholic beverages per week. However, there were no significant differences in alcohol consumption between violence-exposed or unexposed women. No differences in exposure to violence were found in relation to BMI. Crude odds ratios showed no association between experiences of \'history of violence\' and dystocia (n = 940) OR 0.91, 95% CI (0.77-1.08), \'violence before pregnancy\' and dystocia (n = 914) OR 0.90, 95% CI (0.77-1.07), \'violence during pregnancy\' and dystocia (n = 66) OR 0.90, 95% CI (0.54-1.50), or \'first time violence during pregnancy\' (n = 26) OR 1.24, 95% CI (0.56-2.71) and dystocia. Moreover, no significant associations were found between dystocia at term and any of the various categorizations of violence: i) \'threat of violence\' OR 0.97, 95%CI (0.79-1.18), ii) \'physical violence\' OR 0.93, 95%CI (0.78-1.11), iii) \'sexual violence\' OR 1.18, 95%CI (0.85-1.62) and iv) \'serious violence\' OR 1.00, 95%CI (0.81-1.23). A multiple regression done with \'domestic violence\' (solely) as an independent variable together with already known factors as maternal age, BMI and smoking associated with dystocia showed no significant association to dystocia at term, OR 1.23 95% CI (0.89 - 1.69). Women older than 24 years and women with pre pregnancy overweight had significantly increased risk for dystocia at term with OR 1.53 95% CI (1.16 -2.00) respectively OR 1.31 95% CI (1.07-1.62). Further, multiple regression with \'history of violence\' as an independent variable together with age, BMI and smoking showed no association to dystocia at term with OR 0.98 95% CI (0.81-1.18). Table [3](#T3){ref-type="table"} shows the relationship between background and lifestyle characteristics and the risk (crude odds ratios) for dystocia in women with and without exposure to \'history of violence\'. Women older than 24 years had significantly increased risk for dystocia at term, irrespective of exposure to violence (exposed: OR 1.64, 95% CI: 1.16-2.30; unexposed: OR 1.36, 95% CI: 1.02-1.83). Also, women who consumed alcohol during pregnancy and had experienced exposure to \'history of violence\' had an increased risk for dystocia at term (exposed: OR 1.45, 95% CI: 1.07-1.96). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Maternal background characteristics as risk factors for dystocia in nulliparous women with and without experience of history of violence, as shown by crude odds ratios (OR) and 95% confidence intervals. ::: ------------------------------------ ------------------------------ ------------------------------ -------------------------- ------------------ **Characteristics** **History of violence**\ **No history of violence**\ **(n = 940)** **(n = 1712)** **Total cases of dystocia**\ **Total cases of dystocia**\ **(n = 337)**\ **(n = 648)**\ **Dystocia/no dystocia** **OR 95% CI** **Dystocia/no dystocia** **OR 95% CI** Age \> 24 years 279/449 1.64 (1.16-2.30) 574/905 1.36 (1.02-1.83) Non-Danish 21/44 0.84 (0.49-1.44) 58/77 1.26 (0.88-1.80) Not cohabiting 4/7 1.02 (0.29-3.52) 1/4 0.41 (0.05-3.64) Low educational status (≤10 years) 85/149 1.00 (0.74-1.38) 84/176 0.76 (0.57-1.00) Unemployed 120/183 1.23 (0.93-1.63) 154/279 0.89 (0.71-1.12) Smoking 118/202 1.06 (0.80-1.41) 122/203 0.98 (0.77-1.26) Alcohol consumption 100/139 1.45 (1.07-1.96) 144/254 0.93 (0.74-1.18) Overweight \> 25 BMI 80/117 1.26 (0.91-1.75) 156/210 1.26 (0.99-1.60) ------------------------------------ ------------------------------ ------------------------------ -------------------------- ------------------ ::: Women giving birth to an infant with a birth weight of 3500 g or more (n = 1231) had significantly increased risk of dystocia irrespective of exposure to violence (exposed (n = 424): OR 2.0, 95% CI: 1.49-2.69; unexposed (n = 807): OR 1.39, 95% CI: 1.12-1.71). Women with dystocia had significantly increased risk for instrumental deliveries (n = 632) compared to normal deliveries, irrespective of exposure to violence (exposed (n = 221): OR 4.45, 95% CI: 3.23-6.11; unexposed (n = 410): OR 4.21, 95% CI: 3.33-5.33). Discussion ========== More than one third (35.4%) of the women in this study had been exposed to violence ever in their lifetime, i.e. before and/or during pregnancy. However, no association was found between experienced violence and labour dystocia in nulliparous women at term. Therefore, our findings suggest that women who have been exposed to violence ever in lifetime before and/or during pregnancy are not at a higher risk of prolonged delivery process at term. However, as this is the first study ever with the specific aim to examine the potential association between history of violence and labour dystocia, the current results should be regarded as only preliminary, and further research is needed in order to confirm these apparently negative findings. Nevertheless, recent findings by Khodakarami et al.\[[@B13]\] did show an association between experienced intimate partner violence and labour dystocia. However, Khodakarami et al.\[[@B13]\] did not define dystocia, and also, our definition of experienced domestic violence is somewhat broader, which makes it difficult to compare the results. Yet, in our study, the odds of having dystocia if exposed solely to domestic violence were increased by 23%, albeit not significantly. These two major challenges in obstetrics thus appear mostly to have different underlying risk factors, although smoking is common to both exposure to violence \[[@B20]-[@B23],[@B30]\] and prolonged labour \[[@B30]\], which can in turn lead to labour dystocia. The subjects investigated in our study are primarily Danish women (92.5%), i.e. they were born in Denmark and have Danish ethnicity. Due to ethical considerations, women younger than 18 years were excluded in this study in respect for Danish law regarding autonomy, because otherwise parental consent would have been necessary for participation in the study. The mean age of the nulliparous women was rather high, i.e. 28 years. In accordance with results from previous studies, \[[@B16]-[@B18]\] younger age (\< 24 years) is a risk group for exposure to violence. The results in our study showed that women older than 24 years with or without experience of violence had significantly increased risk for dystocia at term, although in the non-violence exposed group, the association may be regarded as marginally significant due to the lower limits of the confidence interval. Earlier studies have shown that increasing maternal age has a strong association with labour dystocia \[[@B10],[@B11]\]. Women exposed to violence were more often smokers, in accordance with what several international studies have shown, \[[@B21]-[@B23]\] even though smoking has been decreasing in Denmark during the last decade, especially in the age-group 25-44 years \[[@B42]\]. A nation-wide study in Denmark showed that in the year 2005, smoking prevalence at some point in pregnancy was 16% \[[@B43]\]. However, our study had the same definition of smoking as in the study of Egebjerg Jensen et al.\[[@B43]\], and the prevalence of smoking during pregnancy was higher, i.e., 24.3% in our study. It is alarming if the smoking prevalence is increasing during pregnancy. Another background variable that might be of importance for an association between exposure to violence and labour dystocia is alcohol. In the current study, women who had experience of violence and who also were alcohol consumers during late pregnancy had higher risk of dystocia at term compared to non-violence exposed women. The calculated odds ratio was significant (p = 0.017), albeit the strength of the association may perhaps best be regarded as modest in the current context, in that these are crude odds ratios, i.e. unadjusted for any other background characteristics. In accordance with earlier results, \[[@B20],[@B21]\] unhealthy maternal behaviour such as use of alcohol and drugs during pregnancy are more common among women who live in violent relationships. Yet, to our knowledge associations between consumption of alcohol during the third trimester in pregnancy and experience of violence as a risk factor for labour dystocia have not been described in the literature before. These findings are difficult to interpret and need further investigation. In the present study 2.5% (n = 66) of nulliparous women were exposed to violence during the pregnancy and 39.5% (n = 26) of them had never been exposed to violence previously. Thus, the violence was initiated during their first pregnancy. The size of this group was however limited and these results would need to be investigated further. Transition into a new social role can be experienced as a very stressful event for the father to-be \[[@B44]\] and may lead to increased pre-existing strains in the couple\'s relationship to such an extent that the partner uses psychological or physical violence towards the mother to-be. However, our definition of \'history of violence\' in this study includes all experienced violence during and before pregnancy, and thus, intimate partner violence is only one possible component. It should be noted that the current results regarding prevalence of exposure to violence may conceivably represent an underestimate of the true rates. Technical errors affected the internet data collection (40% of the material), such that women were unable to report whether they were exposed to violence during current pregnancy or not. More specifically, they were only provided with two alternatives of answers in the questionnaire, instead of three. Also, the true prevalence of physical and psychological abuse in pregnant women is difficult to estimate since women who are exposed to violence may be afraid to report such violence in fear of abuse escalation \[[@B24]\]. First time pregnancy may escalate existing stressors in the couple\'s relationship which can lead to psychological or physical abuse and this in turn may result in prolonged labour \[[@B33]-[@B36]\]. Nevertheless, in the current study, there was no association between exposure to \'first time violence during pregnancy and dystocia\'. However, there were only 26 women in this group. Despite the limited size of this group, the odds of having dystocia were increased by almost 25%, albeit not significantly. Thus, the question remains as to whether a significant association between dystocia and exposure to first time violence during pregnancy would be obtained in a larger sample. A potential weakness in the current study is the small number of individuals in some of the sub-group analyses. In current study overweight pre pregnancy showed significant increased risk of more than 30% to having dystocia at term irrespective if exposed solely to domestic violence or to history of violence. Kjaergaard et al.\[[@B10]\] has already presented overweight as a riskfactor for labour dystocia from the DDS \[[@B8]-[@B10]\]. Some potential obstetrical risk factors for dystocia were also analysed in relation to violence. Our findings showed that delivering a baby with a birth weight ≥ 3500 g was associated with dystocia at term without any association with exposure of violence. Yet, Kjærgaard et al.\[[@B8]\] have already shown on the DDS material that expecting a child with a birth weight \> 4000 g was associated with increased risk of dystocia. Indeed, high birth weight as a predisposing factor for prolonged labour and labour dystocia is well-described in the literature \[[@B31],[@B32]\]. Women exposed to violence more often give birth to low birth weight babies \[[@B20],[@B22],[@B24]\]. However, birth weight is probably not the sole explanation for labour dystocia, and women may have prolonged second stage without any correlation to birth weight \[[@B45]\]. It should also be noted that some studies have found no association between violence and low birth weight \[[@B14],[@B46]\]. Furthermore, unknown factors such as psychosocial stress may also have some importance in this context. However, Nystedt et al. \[[@B47]\] could not find a link between a low level of psychosocial resources in early pregnancy and increased risk for prolonged labour. The etiology of the diagnosis labour dystocia appears to be multifaceted and therefore complex. In addition, although instrumental delivery is a well-known independent consequence of dystocia, \[[@B4],[@B6]\] we did not find any association between instrumental delivery and experience of violence with labour dystocia. Women with labour dystocia had significantly increased risk for instrumental deliveries, irrespective of exposure to violence or not, a finding which is unremarkable. Previous studies have found that women reporting physical violence during pregnancy are more likely to be delivered by caesarean section than those who are not exposed to physical violence \[[@B25],[@B48]\]. However, it is important to keep in mind that in the current sample, only nulliparous women at term were included and thus all premature deliveries were excluded. Methodological discussion ------------------------- The results of this study might potentially be biased due to selection or misclassification. However, we do not find any reason to believe that systematic selection bias or misclassification occurred. The current cohort design based upon prospectively collected data enabled the comparison of risk of labour dystocia among women exposed and un-exposed to violence during the same time period. The population in this study consisted only of nulliparous women which made the cohort a homogeneous group in that respect. Also, the concept \'dystocia\' was very well defined, in accordance with ACOG criteria for dystocia in labour\'s second stage \[[@B6]\] and with the criteria for dystocia in the first and second stage described by the Danish Society for Obstetrics and Gynecology, \[[@B39],[@B40]\] which means that the composition of the group defined with labour dystocia is homogeneous. However, our results raise the question as to whether these criteria for labour dystocia are relevant for the diagnosis. Labour dystocia is still a poorly defined phenomenon which might be categorized with respect to clinical diagnosis \[[@B12]\]. It may well be that the current definition with a time span of four hours is too short, and therefore the prevalence of dystocia may be overestimated. The use of a lengthier time criteria might lead to a reduced number of cases diagnosed as dystocia, but would probably yield a more accurate estimate. The extent to which this in turn might lead to a stronger association between experienced violence and labour dystocia is unknown. The internal non-response rate of the questions about violence was only 0.5% that is, only 14 women in this cohort did not answer the violence questions at all. The limited number of women with missing information on violence exposure is unlikely to have affected the results in any major way, and we can only speculate as to whether these women were exposed to violence or not. However, as mentioned above, technical errors due to the use of the internet for data collection (40% of the answers at baseline) provided only two alternatives for answers regarding violence exposure, i.e. \'yes earlier\', or \'no never\', instead of three alternatives. Misclassification of responses could potentially have led to an underreporting of exposure to violence during pregnancy at term. MacMillan et al.\[[@B49]\] found that computer-based screening did not increase prevalence, and that written screening methods yielded fewest missing data. The questions measuring violence used for this sub-study have been previously validated and used in a Danish general population \[[@B42]\]. However, since the questions have not been adapted to a pregnant cohort before, this may have influenced the findings obtained. Further, it is possible that the rather broad time frame for experienced violence investigated in the current study is not relevant for a study of obstetric outcome. However, according to Eberhard-Gran et al., \[[@B37]\] history of sexual violence in adult life is associated with an increased risk of extreme fear during labour. In our hypothetical model excessive stress, fear and anxiety are related to dysfunctional labour. Screening for violence is not a routine in all countries. If it could be known for the midwife and the obstetrician prior to delivery that the woman had been exposed to excessive stress due to domestic violence before or during pregnancy, then health care practitioners could provide closer monitoring throughout pregnancy and during delivery. The caring process could be more carefully scrutinised to the unique woman\'s needs. However, the extent to which closer monitoring would decrease risk for labour dystocia is still an unanswered question. Conclusions =========== The hypothesis that nulliparous women who have been exposed to violence are more prone to labour dystocia during childbirth at term has not been confirmed. Due to the current scarcity of studies exploring a possible association between violence and labour dystocia, two major contributors to adverse maternal and fetal outcome, the extent to which a relationship might exist would need further investigation. In this regard, it would also be beneficial if the criteria for the definition dystocia could be further evaluated. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= All authors contributed to the planning of the study. Analyses were planned by all authors. HF performed the analysis and all authors interpreted the results. HF wrote the drafts of the manuscript, which the other authors commented on and discussed. All authors approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2393/11/14/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional file 1 **Appendix**. Questions concerning violence used in the current study. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ The authors would like to thank Hanne Kjaergaard (HK) who gave us permission to use the database from the Danish Dystocia Study, Mahnaz N Moghadassi, statistician, Department of Community Medicine, Division of Social and Preventive Medicine, Lund University, Sweden, for assistance and help with performing the analyses, and Malmo University, Faculty of Health and Society for economic support.

PubMed Central

2024-06-05T04:04:19.041589

2011-2-21

PMC3052210

Background ========== Reovirus is a nonenveloped dsRNA virus which is highly prevalent in the human population but produces few clinical symptoms. Great interest has surrounded the use of reovirus as an oncolytic agent due to its ability to infect and induce death in a range of human malignancies whilst sparing normal cells. Furthermore reovirus has completed a number of early clinical trials and is now being tested in the phase III setting \[[@B1]-[@B3]\]. Initial studies indicated that the tumour specific oncolytic activity was dependent upon the presence of an activated Ras signalling pathway \[[@B4]\], although recent data has indicated that susceptibility to reovirus infection may be influenced by additional complex mechanisms \[[@B5],[@B6]\]. Previous work in our laboratory has indicated that human melanoma cell lines, as well as freshly resected tumour, undergo reovirus-induced apoptotic death in a Ras/RalGEF/p38 dependent manner, and that this death is accompanied by the release of inflammatory chemokines and cytokines \[[@B7]\]. The release of pro-inflammatory mediators following viral infection of tumour cells has been observed with other oncolytic viruses such as Herpes Simplex Virus (HSV) \[[@B8]\] and Newcastle disease virus (NDV) \[[@B9]\]. As well as inducing direct oncolysis, several viruses, either naturally or via insertion of immune-activating genes, have been shown to stimulate anti-tumour immune responses, indicating their potential as immunotherapeutic as well as cytotoxic agents \[[@B10]\]. We have previously shown that reovirus can exert immunogenic effects against tumour cells by directly activating DC to stimulate innate NK/T cell cytotoxicity \[[@B11]\], and by reovirus-induced tumour cell death facilitating the priming of innate and adaptive anti-tumour responses in mouse and human model systems \[[@B12]-[@B14]\]. However, the immunogenicity of the pro-inflammatory milieu produced by reovirus-infected melanoma cells (independent of the effects of the virus itself which may be cleared rapidly in vivo), and the signalling pathways involved in initiating cytokine/chemokine production in tumour cells, have not been addressed. Chemokines can participate in the host response during infection and inflammation by directing immune effector cell migration. Four families of chemokines have been described based on the position of conserved cysteine residues \[[@B15]\]. Multiple chemokines can share one common receptor, and each chemokine can potentially bind to several different receptors, thereby allowing multiple biological outcomes depending upon the composition of the chemokine milieu and the cells within the environment \[[@B16],[@B17]\]. Furthermore, at sites of inflammation, chemokines can form heteromers, potentially inducing synergistic actions and enhancing leukocyte migration and activation \[[@B18]\]. Hence, the induction of multiple chemokines within an immunosuppressive tumour microenvironment has the potential to induce potent effects on immune effector cells to enhance therapy. For example, in a murine B16 melanoma model, ectopic expression and secretion of IP-10 by tumour cells increased the number of NK cells at the tumour site and prolonged NK cell dependent survival \[[@B19]\]. Data have also indicated a good correlation between CXCR3 expression on T cells and an improved clinical outcome in stage III melanoma patients \[[@B20]\]. The current study further investigates the chemokines and cytokines (including type I IFNs) induced by reoviral oncolysis and the signaling pathways responsible for the production of these pro-inflammatory mediators. We also determine the effects of reoTCM, specifically in the absence of active virus to exclude the direct consequences of viral immune activation, on chemotaxis, activation and effector functions of NK cells, DC and CTL. Materials and methods ===================== Cell Culture and Reovirus ------------------------- Skmel-28, Mel-624, Mel-888 and MeWo cells were grown in Dulbecco\'s Modified Eagle Medium (DMEM) (Invitrogen, Paisley, UK) supplemented with 10% (v/v) foetal calf serum (FCS) (BioSera, Crawley Down, UK) and 1% (v/v) L-glutamine (Sigma Aldrich). Peripheral blood mononuclear cells (PBMC) were obtained, with local ethics approval, from buffy coats of healthy blood donors by Ficoll-Hypaque density centrifugation. Human myeloid immature dendritic cells (DC) were generated from human PBMC by MACS CD14+ selection (Miltenyi Biotech) or monocyte adherence as previously described \[[@B21]\]. Monocytes were cultured in RPMI 1640 (Invitrogen Life Technologies) supplemented with 10% FCS, 1% L-glutamine (complete media), 800 U/ml GM-CSF (Schering-Plough) and 500 U/ml IL-4 (R&D systems) for 5 days. NK cells were isolated from human PBMC using MACS negative depletion kits following standard protocols as previously described (\> 90% purity) \[[@B11]\]. NK cells were routinely cultured in DMEM supplemented with 10% (v/v) human AB serum (Sera Laboratories International Limited), 5% (v/v) foetal calf serum and 1% (v/v) L-glutamine. Cytotoxic T lymphocytes (CTL) primed using reovirus infected mel-888 cells (moi 0.1) were generated in a 14 day priming assay as detailed below. For chemotaxis assays, NK, DC and CTL were resuspended in RPMI/0.5%Human AB/1%L-Glutamine. All cell lines were grown at 37°C in an atmosphere containing 5% CO~2~and were routinely tested for, and confirmed free of, *Mycoplasma*. Where indicated melanoma cells were treated with 2.5 mM 2-aminopurine (2-AP) (Sigma-Aldrich) or 50 μM caffeic acid phenethyl ester (CAPE) (Axxora, Nottingham, UK) for 1 hour prior to, and during, treatment with reovirus. Neither 2-AP nor CAPE at the doses used were directly toxic to melanoma cells (data not shown). Reovirus type 3 Dearing strain was provided by Oncolytics Biotech Inc. Virus titer was determined by a standard plaque assay using L929 cells. ELISA, Luminex and Flow Cytometry --------------------------------- IL-8 was detected using matched pair antibodies (BD Pharmingen). IFN-β was detected using an IFN-β kit (PBL Laboratories) as per the manufacturer\'s instructions. Eotaxin and IP-10 were detected using Luminex technology (Biosource) according to the manufacturer\'s instructions. To assess NK cell phenotypic activation, NK cells were labelled with CD69-APC/CD56-PE/CD3PerCP antibodies (BD Pharmingen, R and D Systems). CD69 expression was determined on gated CD56-PE^+ve^/CD3-PerCP^-ve^populations using a FACSCalibur instrument (BD Biosciences). Western Blotting ---------------- Melanoma cell lines were seeded in 10 cm dishes and left untreated or infected with 10pfu/cell reovirus for 4, 8, 12, 16, 20 and 24 hours prior to preparation of whole cell lysates. Nuclear fractions were prepared using the Active Motif Nuclear Extract Kit according to the manufacturers\' instructions. Whole cell lysates were prepared using RIPA buffer supplemented with a protease inhibitor cocktail (25 μl/ml, Sigma). 10 μg of total protein was loaded per lane for whole cell lysates. 20 μg of protein was loaded per lane for nuclear fractions. Proteins were separated on a 10% SDS polyacrylamide gel and transferred to a nitrocellulose membrane. Membranes were probed with antibodies against total PKR (BD Transduction Laboratories), I-κB (Santa Cruz) and p65 NF-κB (Upstate). Antibody binding was detected using the Odyssey system (LI-COR Biosciences UK, Cambridge, UK). Equal lane loading was confirmed using monoclonal antibody against β-actin (Sigma). RNA Interference (RNAi) Studies ------------------------------- siRNAs were purchased from Ambion and reconstituted to a concentration of 20 μM according to the manufacturers\' instructions. An siRNA oligonucleotide against PKR (PKRV) and an irrelevant (green fluorescent protein-specific) control siRNA were used. The sequences were, PKRV 5\' AAGGUGAAGGUAGAUCAAAGA-3\', Control 5\' AAGGACGACGGAAACUACAAG-3\'. Melanoma cells in 24 well plates and 10 cm dishes were transfected with 100 nM (this concentration was optimized by initial titration experiments) of siRNA. Control siRNA transfections using equivalent concentrations were included in each experiment. Transfections were carried out in serum-free media using OligofectAMINE (Invitrogen) for 6 hours. Media were replenished with serum and supplements and cells were cultured in the continued presence of siRNA overnight. After the overnight incubation pre-existing media was removed from the 24 well plates and replaced with 1 ml normal growth media, and cells were left untreated or infected with reovirus (0-10pfu/cell). Supernatants from these cells were collected at 24 and 48 hours post reovirus infection. For 10 cm dishes the pre-existing media was removed and whole cell lysates were prepared and separated by SDS-PAGE prior to probing with total PKR antibody. Equal lane loading was confirmed using a monoclonal antibody against β-actin. Priming of Mel-888 Cytotoxic T Lymphocytes (CTL) ------------------------------------------------ Mel-888 cells were seeded into tissue culture flasks and allowed to adhere. DC, resuspended in a 50:50 mix of DC media/reoTCM (from Mel-888), non-reoTCM, or DC media/DMEM complete, were added to the Mel-888 cells at a 1:3 ratio (DC:tumour cell) overnight. Suspension cells were aspirated, leaving the tumour monolayer intact, and cells were pelleted as previously described \[[@B13]\]. Tumour loaded DC were then resuspended in CTL media \[RPMI supplemented with 7.5% (v/v) human AB serum (Sigma), 1% (v/v) L-glutamine, 1% (v/v) sodium pyruvate (Life Technologies), 1% (v/v) non-essential amino acids (Life Technologies), 1% (v/v) HEPES (Life Technologies), 20 μmol/L 2 -mercaptoethanol (Sigma), and mixed with autologous PBMC at a ratio of 1:10 to 1:30. Cultures were supplemented with 5 ng/ml IL-7 (R&D systems) from day 1 and 30 U/ml IL-2 (R&D systems) on day 4 only. Cultures were re-stimulated using the same protocol at day 7. Cells were harvested at day 14 and chromium release, CD107 degranulation and intracellular IFN-γ assessed as detailed below. NK, DC and CTL chemotaxis ------------------------- 1.5 × 10^5^melanoma cells were plated in 6 well plates and left overnight. Pre-existing media was removed and replaced with DMEM/2%FCS/1% L-Glutamine and cells were infected with 1 or 10pfu/cell reovirus for 48 hours. Supernatants were collected and passed through a 0.2 μm Acrodisc syringe filter (Pall Life Sciences) and then through a Viresolve^NFR^filter (Millipore) to remove virus. Chemotaxis was assessed using a transwell system. 650 μl filtered TCM from each of the four melanoma cell lines was added to the lower chamber of triplicate wells in a 24-well plate and a 5 μm (NK cells and CTL) or 8 μm (DC) Thincert™ membrane (Greiner Bio-One) placed in the well. 5 × 10^5^NK, DC or CTL (the anti-Mel-888 CTL used in these chemotaxis experiments had been previously primed using reovirus-infected Mel-888-loaded DC as described \[[@B12]\]), resuspended in 100 μl of low serum media were placed in the upper chamber and plates were incubated at 37°C for 3 hours. Cells in the lower wells were harvested, washed in FACS buffer and labelled with 3 μl of CD56-PE (Serotec) and 3 μl CD3-FITC (BD Biosciences) (NK cells), CD11c-PE (BD Biosciences) (DC) or 3 μl CD3-FITC and CD8-PerCP (CTL) for 30 minutes at 4°C. Cells were washed, resuspended in 300 μl FACS buffer and transferred to a TruCount™ tube (BD Biosciences) containing a known number of fluorescent beads to provide the internal counting control. A cell to bead ratio was determined for each tube and a migration index was calculated by normalising this ratio to those of controls in which TCM harvested from uninfected tumour cells (\'non-reoTCM\') was used. CD107 degranulation and intracellular IFN-γ staining ---------------------------------------------------- CD107 is a marker of lysosomal granule exocytosis and was used as a marker of NK cell and CTL degranulation as previously described \[[@B14]\]. NK cells and CTL were incubated at a 1:1 ratio with tumour targets (K562, Mel-888 or SKOV-3), in the presence of CD107a-FITC and CD107b-FITC antibodies (BD Bioscience). Brefeldin A (10 μg/ml) was added after 1 hour and cells were incubated for a further 4 hours. Cells were then labelled with CD8 PerCP (for CTL) or CD56-PE, CD3-PerCP and Dead Cell Discriminator (Miltenyi Biotech) for NK cell cultures, and fixed in 1% PFA. Analysis was performed by gating on CD8 populations (CTL) or on CD56^+ve^populations, and excluding cells labelled with Dead Cell Discriminator and CD3 in the FL3 channel (NK cells). For determination of intracellular IFN-γ, cells were treated as above, permeabilised with 0.3% saponin and labelled with IFN-γ-FITC (BD Pharmingen) prior to flow cytometric analysis. ^51^Chromium cytotoxicity assay ------------------------------- Cytotoxicity of CTL was measured using a standard 4 hour ^51^Cr assay \[[@B21]\] against Mel-888 or irrelevant SKOV-3 targets. Percent lysis was calculated using the formula: % lysis = 100 × (cpm experiment - cpm spontaneous release)/cpm maximum release - cpm spontaneous release. Results ======= Reovirus-infected melanoma cells secrete eotaxin, IP-10 and IFN-β ----------------------------------------------------------------- We have previously shown that reovirus infection of human melanoma cell lines induces a potentially immunogenic form of cell death with the release of RANTES (CCL5), IL-8, MIP-1α (CCL3) and MIP-1β (CCL4) \[[@B7]\]. In addition, DC pulsed with reovirus-infected melanoma cells secrete the chemokines CCL2, 3, 4, 5, 7, 8, 11 (eotaxin), and CXCL10 (IP-10), and these culture supernatants induce NK cell chemotaxis \[[@B14]\]. We sought to extend these studies by first testing for secretion of this wider panel of chemokines from melanoma cell lines in response to direct reovirus infection. Having previously observed secretion of IFN-β (but not IFN-α) by DC pulsed with reovirus-infected Mel888 cells \[[@B14]\], we also tested for type 1 IFN production by reovirus-infected melanoma cells. Although the defective anti-viral IFN response of tumour cells to oncolytic viruses can account for tumour-specific oncolysis \[[@B22]\], this response may not be completely abrogated. IFN-β in particular may be functionally significant within the tumour microenvironment, since its secretion by reovirus-infected Mel888-loaded DC activates NK cells for enhanced cytotoxicity \[[@B14]\]. In addition to RANTES, MIP-1α, MIP-1β and IL-8 \[[@B7]\], we detected eotaxin and IP-10 (but not CCL2, 7 and 8 - data not shown) in all four cell lines tested. Levels of eotaxin were relatively low (Figure [1A](#F1){ref-type="fig"}), whilst IP-10 was secreted at higher levels by all cell lines (Figure [1B](#F1){ref-type="fig"}). Eotaxin is a member of the CC family of chemokines and can selectively recruit eosinophils \[[@B23]\], which have been associated with increased survival in a range of cancers \[[@B24]\]. IP-10 is a CXC chemokine, and a chemoattractant for monocytes, T lymphocytes and NK cells which has been shown to elicit immune-mediated anti-tumour effects *in vivo*\[[@B25]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Melanoma cell lines produce eotaxin, IP-10 and IFN-β in response to reovirus infection**. Melanoma cell lines were treated with 0 (open bars), 1 (pale grey bars) or 5 (dark bars) pfu/cell reovirus and supernatants were collected after 48 hours and assayed for eotaxin (A), IP-10 (B) and IFN-β (C). Results shown are representative of 3 independent experiments. :::  ::: Type 1 IFNs are secreted by normal cells to attenuate viral infections, and mediate multiple immunoregulatory functions that affect innate and adaptive responses \[[@B26]\], including phenotypic and functional maturation of DC \[[@B27]\] in the context of defence mechanisms against tumours \[[@B28]\]. Dysfunctional IFN pathways in cancer cells have been proposed as a mechanism by which replication and cell lysis for viruses such as vesicular stomatitis virus (VSV), vaccinia virus (VV), measles and NDV is restricted to tumour cells during oncolytic virotherapy \[[@B29]\]. Moreover, IFN-β has been genetically engineered into oncolytic viruses to improve the therapeutic index between normal and malignant cells \[[@B30]\], and to support priming of anti-tumour immunity \[[@B31]\]. Therefore, we tested whether type 1 IFNs were secreted by reovirus-infected melanoma cells, and found that IFN-β (Figure [1C](#F1){ref-type="fig"}), but not IFN-α (data not shown), was produced by all 4 cell lines. These data indicate that reovirus infection of melanoma cells induces inflammatory chemokines capable of recruiting immune effector cells, as well as IFN-β, which can support priming of anti-tumour immunity in the context of oncolytic virotherapy. Reovirus infection activates NF-κB in melanoma cells leading to chemokine/cytokine secretion -------------------------------------------------------------------------------------------- Next, we investigated the signalling pathways involved in chemokine/cytokine production following reovirus infection of melanoma cells. We focused on NF-κB, as reovirus infection induces NF-κB nuclear translocation to activate pro-apoptotic gene expression in cultured HeLa cells \[[@B32]\]. Furthermore, several of the chemokines/cytokines produced in our system, such as IL-8, RANTES and IFN-β, are known NF-κB dependent genes \[[@B33]\]. NF-κB resides in an inactive cytoplasmic form in conjunction with I-κB. Following I-κB degradation, NF-κB translocates to the nucleus to initiate transcription. Therefore, I-κB degradation and increased expression of the p65 NF-κB subunit in nuclear fractions can be used as indirect indicators of NF-κB activation. In all 4 melanoma cell lines I-κB degradation was observed within 16 hours of reovirus infection, which coincided with an increase in nuclear p65 NF-κB expression (Figure [2A, B](#F2){ref-type="fig"}). To confirm a role for NF-κB, we used IL-8 and IFN-β as representative chemokines/cytokines, and pre-treated the melanoma cell lines with the NF-κB small molecule inhibitor CAPE \[[@B34]\] prior to infection with reovirus. In all cell lines pre-incubation with 50 μM CAPE led to significant decreases in IL-8 levels at all doses of reovirus used (p \< 0.05) (Figure [2C](#F2){ref-type="fig"}). Similar inhibition of IFN-β secretion was observed following CAPE pre-treatment (data not shown). Taken together these data confirm that reovirus infection of melanoma cells induces NF-κB activation to initiate transcription of chemokines and cytokines. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Reovirus infection activates NF-κB in melanoma cells to induce cytokine secretion**. (A) Melanoma cell lines were seeded in 100 mm dishes and treated with 10pfu/cell reovirus. At 4, 8, 12, 16, 20 and 24 hours post-infection whole cell lysates were prepared and I-κB assessed by western blot. (B) Melanoma cell lines were seeded as in (A), and nuclear fractions were prepared and western blotted for NF-κB p65. Densitometry data is shown underneath each blot. (C) Melanoma lines were seeded in 24 well plates and pre-treated with 50 μM CAPE, or equivalent DMSO solvent concentrations, for 2 hours prior to addition of reovirus at the indicated doses. Supernatants were collected after 48 hours and IL-8 levels determined using ELISA. Data are representative of at least 3 independent experiments. \* indicates *P*\< 0.05, by Student\'s t-test. :::  ::: Chemokine and cytokine production by reovirus-infected melanoma cells is mediated by a PKR dependent pathway ------------------------------------------------------------------------------------------------------------ The double stranded RNA genome of reovirus is detectable by several cellular molecules which can activate multiple signalling pathways. Having established that reovirus infection induces NF-κB activation, we next sought to identify upstream mediators that might provide a link between dsRNA detection and NF-κB activation. A major candidate was the serine/threonine protein kinase PKR, which binds to, and is activated by, dsRNA. PKR can inhibit viral translation via phosphorylation of the translation initiation factor eIF-2α and ras-related defective PKR signaling has been implicated in the tumour specificity of reovirus replication and oncolysis \[[@B35]\]. PKR is also involved in the anti-viral type 1 IFN response, which is at least partially functional in our system, as demonstrated by secretion of IFN-β following reovirus infection (Figure [1C](#F1){ref-type="fig"}). Significantly, PKR is involved in the canonical NF-κB signalling transduction pathway \[[@B36]\], and can induce NF-κB activation via phosphorylation of I-κB \[[@B37]\]. We investigated the role of PKR in inflammatory chemokine/cytokine secretion by reovirus-infected melanoma cells, again using IL-8/IFN-β as representative readouts. Initial western blot analysis confirmed that all cell lines expressed baseline levels of total and phosphorylated PKR, which did not change on reovirus infection (data not shown). Cells were then pre-treated with the PKR inhibitor 2-AP prior to reovirus infection and significant reductions in IL-8 (p \< 0.05) (Figure [3A](#F3){ref-type="fig"}) and IFN-β (data not shown) were observed in 3 out of 4 cell lines. To further confirm these findings we used siRNA to specifically knockdown PKR expression \[[@B38]\]. Mel-624 cells were used, following initial optimization studies, as these were found to have the highest transfection efficiency of the 4 cell lines (data not shown). PKR siRNA decreased total PKR expression by approximately 50% compared with control siRNA treated cells (Figure [3B](#F3){ref-type="fig"}). This knockdown was found to correlate to an approximate 40% reduction in IL-8 secretion 24 hours post reovirus infection (p \< 0.05) (Figure [3C](#F3){ref-type="fig"}). These data confirm a role for PKR, in addition to NF-κB, in induction of inflammatory chemokines/cytokines upon reovirus infection and oncolysis. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### **PKR mediates cytokine production by reovirus treated melanoma cells**. (A) Melanoma cell lines were seeded in 24 well plates and pre-treated with 2.5 mM 2-AP or equivalent PBS controls, for 2 hours prior to addition of reovirus at the indicated doses. Supernatants were collected after 48 hours and IL-8 levels determined. Data are representative of at least 3 independent experiments. \* indicates *P*\< 0.05, by Student\'s t-test. (B) Mel-624 tumour cells were seeded in 100 mm dishes and transfected with 100 nM PKRV or irrelevant control siRNA. Cell lysates were prepared and western blotted for total PKR, with β-actin used to confirm equal track loading. (C) Mel-624 cells were seeded in 24 well plates and transfected with 100 nM PKRV or irrelevant control siRNA; reovirus was then added at the indicated doses. Supernatants were collected 24 and 48 hours later and IL-8 levels determined. Data are representative of two independent experiments. \* indicates *P*\< 0.05 by Student\'s t-test. :::  ::: Virus-filtered tumour conditioned media from reovirus-treated melanoma cells (reoTCM) induces a chemotactic response in NK cells, DC and CTL -------------------------------------------------------------------------------------------------------------------------------------------- Previously identified components of reoTCM, such as MIP-1α, MIP-1β, RANTES, \[[@B7]\], Figure [1B](#F1){ref-type="fig"}) are chemoattractants to a variety of immune cell types. We tested whether reoTCM could induce a chemotactic response in relevant immune effector cells (NK, DC, CTL). To address the potential immunogenic bystander effects of the chemokines and cytokines independent of direct consequences of the virus itself \[[@B11]\], reoTCM was passed through a Viresolve^NFR^filter, and successful removal of reovirus was confirmed by a negative plaque assay on L929 cells (data not shown). Isolated NK cells actively migrated toward reoTCM (Figure [4A](#F4){ref-type="fig"}), with an approximate 2-3 fold increase in migration in 3 out of 4 cell lines (p \< 0.05). Similar chemotactic responses were observed in DC with reoTCM derived from the same three cell lines (p \< 0.05) (Figure [4B](#F4){ref-type="fig"}). Anti-tumour effector CTL generated by priming in the presence of reo-infected mel-888 \[[@B12]\] gave comparable results (Figure [4C](#F4){ref-type="fig"}); increases in cell migration were observed for MeWo, SKMEL-28 (p \< 0.05) and Mel-888 (although this did not reach statistical significance). Mel-624 reoTCM failed to induce a significant chemotactic response in any cell type. Interestingly, this cell line exhibited lower levels of MIP-1α, MIP-1β and IP-10 secretion following reovirus infection (\[[@B7]\], Figure [1B](#F1){ref-type="fig"}). Although these data do not precisely define the single/multiple chemokines responsible for chemotaxis, they show that NK cells, DC and CTL are capable of actively migrating toward melanoma cells undergoing reoviral induced cell death. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### **Isolated NK cells, DC and CTL migrate toward virus filtered reoTCM**. 5 × 10^5^NK cells (A), DC (B) or CTL (C) were resuspended in RPMI + 0.5% human AB serum + 1% L-Glutamine and placed in a 5 μm (NK + CTL) or 8 μm (DC) Thincerts™ to separate cells from virus filtered reoTCM/non-reoTCM. Migration was assessed after 3 hours by labelling cells with CD11c-PE (DC), CD56-PE/CD3-FITC (NK cells) or CD3-FITC/CD8-PerCP (CTL) and then using Trucount™ tubes to provide an internal counting control. A cell:bead ratio was determined for each tube and a migration index calculated by normalising this ratio to those of non-reoTCM controls. Data represents means of triplicate wells +/- SEM and is representative of 4 independent donors. \* indicates *P*\< 0.05 by Student\'s t-test. :::  ::: ReoTCM from reovirus-treated melanoma cells induces phenotypic and functional activation of NK cells ---------------------------------------------------------------------------------------------------- Previous work in our laboratory has demonstrated that secretion of IFN-β by DC loaded with reovirus infected Mel-888 cells stimulated NK cell cytotoxicity toward melanoma targets \[[@B14]\]. Having demonstrated that reoTCM contains IFN-β and that virus-free reoTCM can be chemoattractant to immune effector cells (Figure [1C](#F1){ref-type="fig"}, [4A](#F4){ref-type="fig"}), we investigated the immunogenic potential of reoTCM with regard to NK cell activation and innate anti-tumour immune priming. Culture of isolated NK cells in Mel-888 reoTCM upregulated the expression of the NK activation marker CD69 (Figure [5A](#F5){ref-type="fig"}) and increased levels of NK degranulation and intracellular IFN-γ (Figure [5B](#F5){ref-type="fig"}) following co-culture with Mel-888 targets. ReoTCM-treated NK cell degranulation was also demonstrated against K562 targets (although not in the absence of targets - data not shown) demonstrating the non-specific nature of this innate response. Hence, the pro-inflammatory environment induced by reovirus infection of melanoma cells is capable, even in the absence of replicating virus, of inducing lytic activity and intracellular IFN-γ in activated NK cells, thereby potentially supporting innate anti-tumour effects within a treated tumour. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### **NK cells cultured in reoTCM are activated against melanoma cell targets**. (A) NK cells were cultured in the presence of virus-filtered reoTCM/non-reoTCM overnight and CD69 expression determined. (B) NK cells were cultured in the presence of filtered reoTCM or non-reoTCM for 48 hours and then co-cultured with Mel-888 tumour targets prior to CD107 and intracellular IFNγ assessment. Results shown are gated on CD56^+ve^/CD3^-ve^cell populations and are representative of 4 independent donors. :::  ::: Tumour cell-pulsed dendritic cells cultured in reoTCM prime human naïve anti-melanoma CTL ----------------------------------------------------------------------------------------- We have previously shown that reoTCM can induce phenotypic maturation of DC \[[@B11]\], and that DC loaded with infected Mel-888 cells can prime tumour specific CTL in the continued presence of reovirus \[[@B12]\]. To address the adaptive immunogenic potential of virus-free reoTCM, DC were cultured in reoTCM (or medium from uninfected non-reoTCM controls), loaded with Mel-888 cells and added to PBMC. The generation of anti-tumour CTL over 14 days was then assessed as previously described \[[@B13]\]. PBMC proliferation was measured using trypan blue exclusion to determine viable cell number and was greater in reoTCM priming cultures than non-reoTCM controls (Figure [6A](#F6){ref-type="fig"}). Moreover, this proliferation was associated with increased levels of IFN-γ in the priming culture supernatants, consistent with an evolving Th1 adaptive T cell response (Figure [6B](#F6){ref-type="fig"}). A chromium cytotoxicity assay was used to determine the lytic ability of CTL generated by reoTCM and non-reoTCM-treated tumour-loaded DC. Whilst some specific anti-Mel888 CTL activity was seen under non-reoTCM DC conditions, levels of killing were significantly higher when reoTCM-conditioned DC were used for CTL priming (approximately 70% lysis compared with 30%) (p \< 0.05) (Figure [6C](#F6){ref-type="fig"}). No killing of irrelevant SKOV-3 tumour targets was observed. In addition, CTL CD107 degranulation and intracellular IFN-γ, in the presence of Mel888 targets (but not SKOV-3, data not shown), was also higher after reoTCM priming compared with their non-reoTCM counterparts (Figure [6D](#F6){ref-type="fig"}). Although these cytotoxicity assays do not address the MHC class I restriction or antigen specificity of killing (which are potential confounding factors when using an allogeneic tumour cell line as an antigen source for human CTL priming), we have previously shown in this system that specific anti-Mel888 CTL include T cells which recognize the tumour-associated antigen MART-1 \[[@B12],[@B13]\], demonstrating that these responses include targeting of antigens relevant to anti-tumour therapy. Hence, in addition to activation of innate NK cell anti-melanoma activity, virus-free reoTCM is able to support effective priming of a specific adaptive CTL response. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### **ReoTCM effectively supports priming of specific CTL by tumour cell-loaded DC**. PBMC were incubated with autologous DC that had been cultured overnight with Mel-888 tumour cells in the presence of reoTCM/non-reoTCM. The PBMC were restimulated 7 days later and assayed at 14 days. (A) Lymphocyte proliferation was determined via trypan blue exclusion (2 representative donors are shown). (B) IFN-γ levels in CTL supernatants were determined by ELISA (2 representative donors are shown). (C) Cytotoxicity of lymphocytes primed in the presence of reoTCM versus non-reoTCM was determined by ^51^Cr release assay using Mel-888 tumour cells as specific targets and SKOV-3 as irrelevant controls. One donor is shown as representative of 2 independent experiments. \* indicates *P*\< 0.05 by Student\'s t-test. (D) CTL as in (C) were further assayed for CD107 degranulation and intracellular IFN-γ on co-culture with Mel-888 targets. The results of one donor are shown, representative of at least four independent experiments. :::  ::: Discussion ========== Reovirus is a tumour-specific oncolytic virus currently under clinical investigation \[[@B39],[@B40]\]. We, and others, have shown that reovirus is one of several therapeutic viruses whose activity can be mediated via activation of an anti-tumour immune response, as well as the direct oncolytic effect of viral replication in tumour cells \[[@B10]\]. Whether the immune response to viral therapy is problematic, due to rapid systemic inactivation of the agent, or actively therapeutic, via provision of a \'danger\' signal within an otherwise immunosuppressive tumour microenvironment, likely depends on multiple factors. These include route of virus delivery (intratumoural *versus*intravenous), the pre-existing immune status of the patient and the mechanisms by which the virus naturally, or via genetic modification, targets tumour cells. Consequently, various immunomodulatory strategies have been employed to improve oncolytic viral therapy, ranging from immunosuppression to improve viral persistence in the circulation \[[@B41]\], through to enhancement of immune activation via insertion of transgenes, such as GMCSF, into the viral genome \[[@B42]\]. We have previously shown that i) reovirus induces apoptotic death in human melanoma cells and that this death is associated with secretion of inflammatory chemokines/cytokines \[[@B7]\], ii) reovirus directly activates DC in the absence of tumour cells \[[@B11]\], and iii) reovirus-infected melanoma cells can activate innate and adaptive arms of the anti-tumour immune response \[[@B12]-[@B14]\]. However, these data were generated in the continued presence of active virus. Since reovirus itself is directly immunostimulatory \[[@B11]\], removal of the virus from reoTCM via filtration allowed us to specifically investigate the additional, bystander immunogenic effects of the inflammatory environment potentially generated in treated tumours. The immunogenic component of human reoviral therapy may have particular clinical relevance, since levels of reovirus replication in freshly resected melanoma cells may be low \[[@B7]\]. Furthermore, the switch from a suppressive to an inflammatory tumour milieu may persist even after the virus has been cleared \[[@B43]\]. We extended our previous analysis of the chemokines and cytokines produced by reovirus-infected human melanoma lines, and showed that eotaxin and IP-10 were also secreted (Figure [1](#F1){ref-type="fig"}). Interestingly, we also detected IFN-β (but not IFN-α) under these conditions, illustrating that an anti-viral type 1 IFN response is partially functional in these tumour cells. This is particularly important as IFN-β is involved in innate immune activation by DC loaded with reovirus-infected cells \[[@B14]\]. Moreover, IFN-β has been engineered into other oncolytic viruses to increase the therapeutic index between malignant and normal cells, and to enhance anti-tumour immune activation \[[@B30]\]. Although the mechanisms responsible for the inflammatory response of tumour cells to reovirus infection have not been addressed to date, previous studies have implicated a role for NF-κB since i) reovirus infection initiates translocation of the p50/p65 NF-κB subunits to the nucleus and activates pro-apoptotic gene expression \[[@B32],[@B44]\], ii) reovirus induces apoptosis in melanoma cells \[[@B7]\], and iii) NF-κB is involved in the production of chemokines and cytokines such as IL-8 and IFN-β \[[@B33]\]. This study confirms that reovirus infection of melanoma cells activates NF-κB, as assessed by I-κB degradation and accumulation of nuclear p65, and that blocking NF-κB with the small molecule inhibitor CAPE significantly decreases production of IL-8 and IFN-β (Figure [2](#F2){ref-type="fig"}). Importantly, this effect was seen across all 4 cell lines, suggesting that common signalling pathways are activated following reovirus infection of melanoma. To address viral sensing and signaling molecules that may lie upstream of NF-κB, we investigated the dependence of IL-8 and IFN-β production on PKR, as one of a number of candidate dsRNA sensors. PKR is involved in the tumour specificity of reoviral oncolysis (although the precise mechanism remains to be fully elucidated), and the anti-viral type 1 IFN response \[[@B35]\]. We found, via small molecule blockade and siRNA knockdown, that PKR is involved in the inflammatory response of melanoma cells following reovirus infection (Figure [3](#F3){ref-type="fig"}). Although we have been unable to detect any significant change in total or phosphorylated PKR following reovirus infection of melanoma cell lines (data not shown), these data suggest that dsRNA detection by PKR initiates activation of NF-κB-dependent chemokines and cytokines in tumour cells. These findings are in agreement with previous observations following direct infection of DC \[[@B11]\]. However, further work is required to fully characterize the signaling pathways responsible for the immunogenic nature of reovirus-induced tumour cell oncolysis. Filtered reoTCM induced a chemotactic response in NK cells, DC and anti-tumour CTL (previously primed using reovirus-infected tumour cells) (Figure [4](#F4){ref-type="fig"}), suggesting that the immunogenic milieu in treated tumours has the potential to recruit a range of immune cells capable of viral detection and innate/adaptive effector functions. With regard to improving access of primed CTL to tumours, this finding is consistent with previous murine data showing greater persistence of adoptively transferred antigen specific T cells within tumours undergoing VSV-mediated oncolysis \[[@B45]\]. To date we have not specifically identified which of the secreted chemokine(s) are responsible for NK cell, DC and CTL migration. It is possible that multiple chemokines may act in combination to engage a variety of receptors to induce a particular physiological response \[[@B16]\]. The ability of reoTCM to support activation of innate (Figure [5](#F5){ref-type="fig"}) and adaptive (Figure [6](#F6){ref-type="fig"}) immune responses against human melanoma cells shows that the immunogenic effects of reovirus induced cell death are not dependent on the continued presence of virus once an initiating danger signal has been delivered. Therefore, even if viral replication in patients is limited by neutralization, for example following repeated administration \[[@B46]\], the immunogenic response of tumour cells to reovirus infection may be sufficient to induce continuing anti-tumour effects. Overall, the present study shows that reovirus infection of human melanoma cells induces a range of chemokines and cytokines capable of inducing a chemotactic response in NK cells, DC and primed CTL. This inflammatory response is dependent upon NF-κB and PKR and is sufficient, in the absence of live virus, to support priming of innate and adaptive anti-tumour immunity. This data supports the potential of bystander activation of human anti-tumour immunity by reovirus killing of tumour cells, even if persistent viral replication is limited by the anti-viral immune response. Abbreviations ============= TCM: Tumour conditioned media; RANTES: regulated on activation normal T expressed and secreted; MIP: macrophage inflammatory protein; IFN: Interferon; PKR: Protein kinase R; NK: Natural Killer; DC: dendritic cells; CTL: cytotoxic T lymphocytes; NDV: Newcastle Disease Virus; HSV: Herpes Simplex Virus; DMEM: Dulbecco Modified Eagle Medium; FCS: Foetal Calf Serum; PBMC: peripleral blood mononuclear cells; CAPE: caffeic acid phenethyl ester; 2-AP: 2-aminopurine; FACS: fluorescent activated cell sorting; VSV: vesicular stomatitis virus; GMCSF: granulocyte macrophage colony-stimulating factor; siRNA: small interfering RNA; SDS-PAGE: sodium dodecyl sulphate polyacrylamide gel electrophoresis; Competing interests =================== Oncolytics Biotech Inc: KH/RV/HP/AM, commercial research grant. MC, employee. Authors\' contributions ======================= LS contributed to conception and design, acquisition of data, analysis and interpretation of data and wrote the manuscript. FE, RP, HP, KH, PS, RV and EI contributed to conception, analysis and interpretation of data (as did MC who also provided clinical grade reovirus (Reolysin). AM conceived the study, participated in its design and coordination and co-wrote the manuscript. All authors read and approved the final manuscript.

PubMed Central

2024-06-05T04:04:19.046293

2011-2-21

PMC3052211

Introduction ============ Continuous technological advances in molecular biology have paved the way for new discoveries in cancer research. In particular, high-throughput profiling of cancer tissue specimens and body fluids has been extensively used in order to unveil specific molecular fingerprint of cancer \[[@B1]-[@B4]\]. Such strategy holds great promise for diagnostics purposes, as it might distinguish between different patients\' prognostic subgroups (good/poor), which could provide the foundation for an individual therapeutic approach towards each patient (tailored therapy). However, in spite of these enormous efforts to elucidate cellular and molecular mechanisms underlying tumorigenesis, cancer still represents one of the deadliest scourges of the modern world. Poor outcomes of current therapies, in particular poor prognosis for patients in advanced stages of solid tumours, have opened the possibility that tumour cells include a population of cells responsible for the initiation of tumour development, growth and its ability to metastasize and reoccur. Because these cells share some similarities with stem cells, they are referred to as cancer stem cells (CSCs). CSC are undifferentiated cells characterised by three major features: (1) potential to differentiate into several or all types of cells that are produced by the original tumour; (2) self - renewal ability; and (3) capacity to maintain the \'stem cell pool\' and the most mature tumour elements for unlimited time periods \[[@B5]\]. CSC could originate from tissue-specific stem cells and bone marrow stem cells, and somatic cells that undergo trans-differentiation processes, or can result from the fusion or horizontal gene-transfer processes. The self-renewal and differentiation ability of CSC gives rise to all tumour cell types, and thereby produces tumour heterogeneity. This relatively new perspective, the so-called \"cancer stem cell\" concept, casts new light on the origins of cancer. The relationship and differences between normal and malignant stem cells remain unclear. In many instances, normal stem cells, tumour stem cells and metastatic stem cells share some common traits. Neoplastic stem cells were indeed shown to express similar antigen pattern and to display similar functional properties in comparison with normal stem cells. Moreover, it has been shown that for the maintenance and activation of both, normal stem cells and tumour stem cells, the Wnt/beta-catenin signalling, Notch and PTEN pathways are crucial \[[@B6]\]. Furthermore, growth of both, normal and neoplastic stem cells, is often mediated by the same cytokines \[[@B7]\]. Importantly, cancer/metastatic stem cells might be discerned from embryonic stem cells by their propensity to differentiate into the cell types within a particular organ (tumour). Therefore, it is tempting to believe that tumour arises from tissue stem cells, and that cellular components bearing stem-like properties govern tumour formation. If cancer arises from rare population of cells with stem-like characteristics, then it is plausible to presume that these stem cells differ from \"normal\" stem cells in high rate of mutations. It is widely accepted that stem cells undergo multiple mutations that are also required for carcinogenesis, most probably due to their long-lived nature \[[@B8]\]. Deregulation of self-renewal mechanisms (*e.g.*Wnt/beta-catenin, Notch and Hedgehog signalling pathways), which drive the stem cell expansion, might be the early key event precipitating the formation of CSCs in the particular tissue during the onset of carcinogenesis. This hypothesis is further corroborated by the fact that oncogenes may affect different stem cells and progenitor cells resulting in phenotypic differences in tumours, whereby it was shown that transgenes encoding components of the Wnt/beta-catenin signalling pathway preferentially induce mammary cancers from progenitor cells \[[@B9]\]. Activation of oncogenes and inactivation of some tumour-suppressor genes as the consequence of genomic instability might drive transformation of normal stem cells to CSCs. Several genes including AKT, TRAIL and CXCL12 are recognised as candidate genes for cancer stem cell progression and latent metastasis \[[@B10]\]. At last, cancer/metastatic stem cells might exhibit higher expression levels of some genes (*e.g.*CXCR4, SDF1, VEGF), anti-apoptotic proteins (Bcl-2 family inhibitors of apoptosis) and transporter proteins (BCRP and P-glycoprotein), and might remain in the G~0~phase accounting for their resistance to chemotherapy. Methods that unequivocally identify CSC *via*specific cell-surface protein markers might be diverse but have many pitfalls. *In vivo*assays using NOD/SCID mice are expensive and time consuming. Therefore, *in vitro*long-term growth assays including sphere-formation assays, serial colony-forming unit assays and label-retention assays are often used in order to screen for stem cell fractions or CSC-regulating compounds \[[@B6]\]. However, the major drawbacks of these methods include: (1) lack of tissue and tumour specificity; (2) inability to isolate CSCs according to the degree of tumour differentiation; (3) lack of species-specificity of CSC receptors and their ligands and homing receptors in the tissue environment; and (4) stem cell plasticity. At last, the general problem for all *in vitro*studies is the selection pressure upon the cultured cells, resulting in the selection of certain cell population permissible to survive and proliferate under specific conditions. Isolation of CSCs is also hampered by the lack of specific CSC antigens or typical antigen combinations not identified so far. These antigens might include regular stem cells antigens such as cytokine receptors, homing receptors (integrins, selectin-ligands, chemokine receptors, cytoadhesion molecules and ligands of matrix molecules such L1 or CD44) and various drug transporters \[[@B5],[@B6]\]. Metastases show a great variety of clinical presentations/manifestations, mostly in correlation with the primary tumour localization \[[@B11]\]. For example, breast cancer metastases can remain latent in the several years follow-up after surgical removal of the primary lesion, whereas metastases in patients with detected pancreatic cancer and small-cell lung carcinoma are often widespread at the time of cancer diagnosis. Furthermore, glioblastoma is locally progressive and invasive, but rarely affecting secondary sites outside CNS. The efficacy of common treatment regimens including radical excision of the primary tumour followed by radiotherapy and/or chemotherapy is limited and often fails to cure the patient. Recent findings imply that some treatments (*e.g.*irradiation at certain regimen, Taxol etc.) do not target the CSCs responsible for tumour development. Moreover, such treatment might have a completely opposite effect, *e.g.*to produce more cancer cells capable of metastasizing \[[@B12]-[@B14]\]. Indeed, it is widely accepted that CSC have a dormant nature and abundantly express drug transporters \[[@B15]\], which could provide a plausible explanation for resistance to standard chemotherapy known to target dividing cells. Consequently, 90% of deaths of the patients with solid tumors are attributed to local invasion and distant metastases \[[@B11]\]. Metastatic model - \"serial\" or \"parallel\"? ---------------------------------------------- The molecular mechanisms of metastasizing are still covered by the veil of mystery. The application of genomic profiling methods (DNA microarray) combined with either animal models of metastasis or laser capture microdissection (LCM) providing *in vivo*insight into molecular processes underlying metastatic progression have fostered the research of its complex molecular nature. So far, the so-called \"serial\" model of clonal progression has been generally accepted explaining that metastatic cells originate from the primary tumour and represent the end stage of tumorigenesis. Although the metastatic genotype contains additional genetic mutations, its spectrum of aberrations is thought to be similar to that found in the primary lesions \[[@B16]\], as augmented by the findings of kariotypic and genomic analyses of the breast, bladder, colon and kidney malignancies \[[@B17]-[@B24]\]. Surprisingly, some of these studies indicate the existence of metastases with lack of (without) genetic similarity to the primary tumour \[[@B18],[@B20],[@B25]\]. Bissig *et al.*\[[@B18]\] showed that 30% of renal cell metastases have almost completely different genotype in comparison with the primary lesion cells isolated from the same patient. Furthermore, a genomic study of metastatic breast cancer \[[@B20]\] presented a significant amount of breast metastases, which do not show strong clonal resemblance to the tumour of primary origin. Finally, the analysis of metastatic lesions at several sites in the same individuals showed a substantial evolutionary divergence between metastatic lesions and primary tumour, as well as between the metastases themselves. Schmidt-Kittler *et al.*\[[@B25]\] reported unexpected results providing novel insight into metastatic progression. In this study, patients were divided into two groups according to clinical presence or absence of metastatic dissemination (M0- no metastatic disease; M1-metastasis positive). The results of comprehensive genomic analyses of the primary tumour specimens and single cytokeratin-positive (CK+) epithelial cells from the bone marrow of the corresponding patients were as follows: (1) the CK+ cells from M0 patients showed about half as many genomic aberrations as those from M1 patients; and (2) most of the CK+ cells from M0 patients showed little similarity to the primary tumour. To sum up, cells from M0 patients showed whole chromosome copy number aberrations, while cells from M1 patients showed sub-chromosomal changes typical for aberrations that appear during telomere crisis. The authors suggest that eventual appearance of clinically detectable metastases in M0 patients could be a consequence of early disseminated cells, which evolve and pass through crisis independently of the primary tumour. Due to slow evolution of these disseminated cells, as well as their persistence during a long period, the results of this study offer an explanation for the appearance of clinically evident metastatic disease years after treatment of the primary tumour that was presumed to be successful and curative \[[@B16]\]. A comparison of genetic fingerprints among disseminated cells from M0 patients with those from metastatic cells that will arise during the follow-up after primary tumour treatment is of utmost importance for the \"parallel\" theory. Similarly, Hüsemann *et al.*\[[@B26]\] showed that cancer cells can spread systematically from earliest epithelial alterations in HER-2 and PyMT transgenic mice and from ductal carcinoma *in situ*(DCIS) in women, thus providing additional evidence that supports the novel outlook on the onset of metastatic spreading and its parallel progression (evolution) during tumorigenesis. This \"parallel\" concept is, nevertheless, based on several pioneering studies \[[@B27],[@B28]\], the first one being the experimental documentation in human leukemias \[[@B29]\] that shaped the concept of \'cancer stem cells\'. Since initiation of metastasis process is inherent to CSCs, the metastatic progression should be thus studied as an independent and \"parallel\" process in tumorigenesis governed by the so-called EMT (epithelial to mesenchymal transition) that occurs among tumour cells. The majority of tumours are epithelial but exert mesenchymal characteristics. During tumour invasion, tumour cells move from the primary tumour site and, similarly to the cells in epithelia during normal embryonic development, lose epithelial characteristics and cell-to-cell contacts, and acquire the mesenchymal gene expression \[[@B30]\]. Such cells are then capable of invading distant sites (Figure [1](#F1){ref-type="fig"}). The changes occurring in the cell-to-cell adhesions forces as well as in the cytoskeletal cortex association to plasma membranes are therefore central to the invasion, migration and intravasation of tumour cells. According to this, epithelial cells might be somehow \"induced\" to become metastatic by several factors such as chemokine CCL5 \[[@B31]\] and transcription factors FOXC2, Twist, Snug, Snail and ZEB1 \[[@B13],[@B32]\] that were shown to drive the EMT programme. The EMT process might account for some similarities found between embryonic stem cells and the stem cell-like traits in neoplastic cells. One must, however, always have in mind that not every EMT-inducing factor discovered will necessarily elicit a stem-like profile. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **The epithelial-mesenchymal transition (EMT) occurs at the primary tumour site where epithelial cells lose tight junctions and apico-basal polarity**. The remodelling of the cytoskeleton occurs as well. The invasion process through the extracellular matrix (ECM) is frequently led by so called tumour-associated fibroblasts. EMT can induce stem-cell-like properties in cells. The question still remains whether the existing cancer stem cells or rather those cells that escaped the primary tumour and acquired the stem cell like phenotype through the EMT process induce distant metastasis. :::  ::: As recently nicely reviewed by Hurt and Farrar \[[@B33]\], the question remains whether the existing CSCs or rather those cells that escaped the primary tumour and acquired the stem cell-like phenotype through the EMT process, induce distant metastases. Further research based on the use of powerful imaging methods and in line with the *in vivo*study performed by Condeelis *et al.*\[[@B34]\] showing the convincing images of individual cells delaminated from primary tumours, will be needed to resolve this issue. Taking all this together, new therapeutic modalities will be required to target independently evolved metastatic cells after early separation from the primary tumour, which will lay the groundwork for further improvement of therapeutic efficacy. Organ tropism and metastatic gene signature ------------------------------------------- The molecular basis of organ tropism, one of the main features of metastasis, is still obscure. It has been well documented that different types of cancer produce metastases at preferred secondary sites, depending on the tissue susceptibility to specific metastatic cells. Bone metastasis is often associated with breast, prostate or lung cancer, while it is rarely detectable in patients with diagnosed colorectal cancer \[[@B35]\]. This preferential development of macrometastases at distant secondary organs can be partially explained by the pattern of blood flow. Nevertheless, molecular interactions between metastatic cells (seeds) and stromal microenvironment (soil) have been proven to mediate efficiency of metastatic formation and its tissue specificity \[[@B35]\]. When it comes to the role of tumour microenvironment in tumour dissemination, early changes observed in tissue before evidence of carcinogenesis might be critical for tissue-specific metastasis. For example, the inflammatory response, matrix remodelling and increase in reactive oxygen species often precede tumour metastasis. Accordingly, it has been recently shown that alteration in the expression of metalloproteinase 9 precedes metastasis in lung \[[@B36]\]. The same studies also showed that arrival of bone marrow-derived hematopoietic progenitors expressing vascular endothelial growth factor receptor 1 in distant sites, which fosters inflammation and sustains tumour growth, represents some of the early changes present in the local microenvironment and necessary for metastases to occur \[[@B37]\]. Genetic and phenotypic characteristics of tumours have also been studied as factors linked to organ-specific metastases \[[@B38]-[@B40]\]. Kakiuchi *et al.*\[[@B39]\] reported differential gene expression profiles for lung, liver, kidney and bone metastases. Another research group \[[@B38]\] investigated metastatic potential of human breast cancer using MDA-MB-231 cell line, originally derived from the pleural effusion of a patient with metastatic dissemination. After injection into immunodeficient mice, metastases were detected in bone and adrenal medulla. Subsequently, human cells were re-isolated from the osteolytic bone metastases, and sublines with high metastatic potential were observed. Microarray profiling of these sublines revealed that the gene expression signatures directly correlated with metastases. The authors were able to distinguish cells with tendency to disseminate to bone or adrenal medulla, as the two groups showed differential gene expression patterns. A set of CXCR4, IL-11, CTGF and MMP1 genes, if co-expressed with the gene encoding osteopontin, correlated with bone-specific metastatic potential. Each of these genes, when expressed alone, failed to cause high metastatic potential. Thus, the set of genes found in the bone metastasis signature seems to be a causative factor in metastasizing to the bone. Furthermore, Minn *et al.*\[[@B40]\] identified a set of 54 genes with differential expression in lung-tropic breast cancer sublines in comparison with bone-tropic lines. The identified set of 54 genes in lung-tropic gene signature is probably not the only gene set responsible for metastasizing to the lungs, because a small amount of investigated primary tumours expressed this specific signature. Therefore, it is logical to presume that other lung-tropic metastatic signatures are yet to be discovered. Recent gene expression profiling studies revealed an association between Src pathway activity and late-onset bone metastasis in breast cancer, which is independent of hormone receptor status and breast cancer subtype \[[@B10]\]. Src activity was shown to be required for CXCL12 activation of the AKT cell survival pathway and for the resistance of metastatic breast cancer cells to the pro-apoptotic effects of TRAIL, both of which are predominantly expressed in the bone metastasis microenvironment. In view of these findings, targeting Src signalling pathway might provide a novel strategy to suppress the survival of disseminated cancer cells. Although it is widely accepted that some primary tumours are predestined to metastasize to specific organ, it is still unclear when and how exactly they acquire organ-tropic gene signatures \[[@B41]\]. Furthermore, the complex multistep nature of metastasis process suggests that vast arrays of genes are responsible for its regulation. Addressing the question of metastatic prediction, recent analyses of human cancer specimens using DNA microarray technology suggest that patients can be divided into prognostic subgroups, based on the \"good\" or \"poor\" gene expression signature of the primary tumour, which predicts the risk of metastasis appearance after tumour resection, particularly evaluated in breast cancer \[[@B4],[@B42],[@B43]\]. Traditionally, in clinical practice, the potential of its metastatic recurrence is in correlation with primary tumour size and histological grade. In a recent study, Ramaswamy *et al.*\[[@B44]\] compared gene signatures of primary adenocarcinomas and metastases from a similar set of adenocarcinomas. Their analysis revelaed a \"metastatic gene signature\" common to many different tumour types. When detected in some primary tumours, this signature indicated its pre-existed tendency to metastasize and, therefore, had a significant prognostic value. Nevertheless, evaluation of this metastatic signature once the primary tumour is diagnosed distinguishes tumours with local growing potential from those preconfigured to disseminate to distant sites \[[@B44],[@B45]\]. However, it remains unclear whether these metastatic signatures are causative factors in metastatic spreading or indirect indicators of metastatic potential \[[@B46]\]. Besides screening for genes linked to metastasis organ tropism, the stem cell theory might provide an additional explanation for this phenomenon. In fact, it has been well documented that CSCs express a G-protein-coupled seven-span transmembrane receptor CXCR4 on their surface similarly to normal stem cells for different organs/tissues \[[@B47]\]. Facts about the role of chemokines, small pro-inflammatory chemoattractant cytokines that bind to G-protein transmembrane receptors of target cells, in mediating cell trafficking might additionally illuminate how metastases are attracted to specific organs. For example, a stromal-derived chemokine SDF-1 exclusively binds to CXCR4 and is highly expressed in lymph nodes, lung, liver or bones. Therefore, metastasis of CXCR4+ tumour cells might be driven specifically to these organs through the SDF1-CXCR4 axis \[[@B47]\]. This assumption is augmented by the findings revealing that several CXCR4+ cancers such as breast, ovarian and prostate cancer metastasize to bones from bloodstream in a SDF-1 dependent fashion \[[@B47],[@B48]\]. However, possible therapeutic strategies based on modulation of the SDF1-CXCR4 axis or other chemokines involved in stem cell trafficking should be carefully considered, as this signalling pathway is normally involved in the trafficking of stem cells. Molecular mechanisms of metastasizing ------------------------------------- Metastasis is apparently central in terms of clinical management of cancer, as the preponderance of patients\' deaths are associated with disseminated disease rather than the primary tumour. Moreover, patients with small primary tumours and node negative status (T1N0) at surgery often (15% to 25%) develop distant metastases \[[@B49]\]. As summarized in this paper, recent literature data support the concept of metastasis as a second disease imposed on the primary tumour, where the outcome of metastasis is determined by the interplay between the specific subpopulation of metastatic cells and host homeostatic factors in specific organ microenvironment including vasculature. However, over the past decades, the so-called progression model has been widely accepted among clinicians and researchers. This model depicted metastasis as a result of several consecutive mutational events occurring either in subpopulations of the primary tumour or disseminated cells, and yielding a small fraction of cells that acquire full metastatic potential. Recently, several microarray studies have prompted reexamination of the progression model, as they enable identification of gene signatures that can distinguish metastatic from non-metastatic tumours, and postulate that metastatic propensity is established early in oncogenesis \[[@B44],[@B45]\]. Accordingly, metastatic genes can be classified into several groups: metastasis initiation genes, metastasis progression genes and metastasis-virulence genes \[[@B11]\]. Metastasis initiation genes provide an advantage in a primary tumour and enable tumour cells to enter the blood-flow. Most genes mediating tumour cell motility, invasion or angiogenesis belong to this group. This class includes genes involved in EMT and caspase 8, whereby loss of caspase 8 function protects tumour cells from programmed death due to release of integrin-regulated anchoring at the invasive front \[[@B30],[@B50],[@B51]\]. The second class of metastatic genes, specified as the metastasis progression genes fulfil some rate-limiting functions in primary tumour growth and other specific functions in metastatic colonization. These genes can be found within organ-tropic gene signatures, thus causing specific advantage restricted to particular distant organ. Nevertheless, while metastasis-virulence genes participate in metastatic colonization due to selective advantage in secondary sites, they do not affect the primary tumour development \[[@B11]\]. Therefore, these genes promote aggressiveness of metastatic tumour cells at the level of colonized distant organs. Recent study dealing with organ-specific metastases in breast cancer reported on the lung-tropic metastatic gene set comprising 54 genes, among which only 18 were expressed in primary tumours. These results provide an evidence for the existence of metastasis progression genes and metastasis virulence genes \[[@B40],[@B52]\]. Both groups of genes enable circulating tumour cells to colonize the lungs. The 18 genes expressed in the primary tumour induce primary tumorigenesis, leading to the larger tumour size at the time of diagnosis. Due to vascular-remodelling programme, three out of those 18 genes (EREG, COX2 and MMP1) facilitate tumour angiogenesis and intravasation in mammary tumours. In addition, they act as a mediator in tumour cells extravasation from the lung capillaries. Accordingly, the EREG, COX2 and MMP1 genes are classified as metastasis progression genes. Introduction of improved molecular research methods has opened new chapter in metastasis research by specifically disclosing some of the underlying molecular events in metastatic progression. Molecules which play an important role in processes involving cell-cell adhesion, migration, proteolysis, chemotaxis, angiogenesis and signal transduction have been investigated with aim to decipher their activity in evolution of metastasis \[[@B35]\]. Five candidate metastasis genes, namely CXCR4, IL-11, CTGF, MMP1 and osteopontin from the bone metastasis signature reported by Kang *et al*\[[@B38]\] encode secreted cytokines or cell surface receptors, which is consistent with traditionally accepted idea that metastasis formation is a consequence of pathological interactions between tumour cells and stromal microenvironment. The evidence for this widely accepted theory is provided by Hu *et al.*\[[@B53]\], who used a xenograft model of human DCIS and primary human breast tumours. The authors found that myoepithelial cells and fibroblasts mediate the transition from DCIS to invasive carcinoma. In the absence of normal myoepithelial cells, co-injection of fibroblasts promoted progression of *in situ*to invasive carcinoma. On the contrary, co-injection of normal myoepithelial cells effectively suppressed tumour weight, despite presence of progression-promoting fibroblasts. The obtained results were not the upshot of permanent genetic aberrations in the epithelial tumour cells. Furthermore, TGF-β and Hedgehog signalling were recognized to have a critical role in breast tumour progression, while decreasing TGF-β and Hedgehog pathway activity *via*TGFBR2/SMAD4 downregulation and Gli2 expression resulted in the loss of myoepithelial cells and accelerated invasion. Even if a wealth of evidence supports this model, some paradoxes still remain. First, in many tissues where tumours arise, normal mature cells have a short lifespan. Consequently, the chances to accumulate mutations required for tumour development are rather limited. Secondly, patients diagnosed with disseminated disease with unknown primary cancer metastatic disease have no clinically detectable primary tumour or only a small, well differentiated lesion that is found at autopsy \[[@B54]\]. Furthermore, although variant clones with high metastatic capacity can be identified in populations, it is frequently observed that these variants revert to a low-metastatic capacity after several generations \[[@B55],[@B56]\]. The so-called transient metastatic compartment model proposed by Weiss *et al*\[[@B57]\] suggests that all viable cells in a tumour might acquire metastatic capacity as explained by the progression model, but due to their position in the primary tumour and random epigenetic events, only a small fraction of these cells are competent to metastasize at a given moment in time. Thus, not all cells within a tumour preserve the capacity to disseminate to secondary sites as a result of accidental, or microenviromentally induced epigenetic events or inadequate access to vasculature. However, this model fails to explain the clonal nature of metastases. If every cell had a metastatic ability modulated only by momentary epigenetic events, then it would be less likely that significant proportions of secondary tumours would appear to be of clonal origin. Another intriguing theory, the so-called cell fusion theory, explains that the acquisition of metastatic phenotype occurs when a healthy migratory leukocyte fuses with a primary tumour cell. Such a \'hybrid\' has the innate blood cell ability to migrate through the body while still keeping the uncontrolled cell processes as occurring in the tumour cells \[[@B58],[@B59]\]. Interestingly, the fusion of genetic and cytoplasmic material between cells of different origins is an important physiological process during development. Generally, cell fusion and horizontal gene-transfer events could be important in the development of the CSCs \[[@B60]\]. Substantiating these theses, the i*n vitro*fusion of cells was proved to produce subclones with varying metastatic potentials \[[@B61],[@B62]\], and over 30 reports confirming cancer cell fusion in animal tumour models have been published so far \[[@B63],[@B64]\]. In addition, a number of factors such as cellular or viral fusion proteins or environmental factors may provoke cell fusion of tumour and normal cells. Such fused cells will probably die or become quiescent. Nevertheless, there is a small fraction that will maintain the ability to proliferate and generate malignant cells. It has been shown that endothelial cells in solid tumours might be aneuploid with multiple chromosomes and multiple centrosomes, implying the possibility of cell fusion between tumour cells and endothelial cells \[[@B65]\]. This theory has yet to be confirmed in humans as well. Technological advances in metastasis research: proteomics --------------------------------------------------------- Global transcriptome profiling of cancer has undoubtedly improved our knowledge of tumour biology, and has led to the discovery of novel therapeutic and imaging targets, as well as potential prognostic and predictive biomarkers for diverse cancer types. However, transcriptomics can predict neither the expression level nor the functional status determined by folding, post-translational modifications, cellular localization and molecular interactions of the key signalling molecules in complex protein networks integral to cancer pathogenesis. For example, how exactly and when the EMT process and other pro-survival signalling cascades in cancer cells are triggered, still remains to be elucidated. Cancer may be genetically based, but on the functional level, it is a proteomic disease \[[@B66]\], because tumour progression, invasion, and metastasis depend on the functional activity of many proteins, such as growth factors and proteases.- Consequently, molecular oncologists have been turning more to proteomics technologies (Table [1](#T1){ref-type="table"}) as to identify novel protein biomarkers specifically associated with metastatic transition, and to decipher signal transduction pathways that propel the cells down the road towards metastasis. Such technological approach should provide an early detection and prediction of metastatic processes, and reveal novel targets for drug development and therapeutic intervention. There is now an enormous wealth of literature data on employing proteomics in metastasis research covering practically all tumour types, and discovered biomarkers might be correlated with the metastatic process and/or cancer stem cell phenotype. Some recent examples will be briefly discussed below. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Overview of the most common proteomics technologies in the research of tumour invasion and metastasis ::: --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Proteomics method Abbreviation Basic principle Biological Application Advantages Limitations ------------------------------------------------------------------------------------------------------------ -------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------- Two-dimensional gel electrophoresis 2-DE Proteins are first resolved by their isoelectric points, and then by molecular weights Separation of proteins in complex biological samples High resolution Very sensitive\ Limited automation\ Direct detection of post-translational modifications Problematic gel-to-gel reproducibility\ Problematic recovery of hydrophobic and large molecular weight proteins\ Limited dynamic range of detection Two-dimensional difference gel electrophoresis 2D-DIGE Samples are labelled with two spectrally distinct fluorescent cyanine dyes, and run on the same 2-DE gel; the two gel images corresponding to each dye scan are then overlaid, and the intensities of paired spots are compared across the gel images Quantification of the differences in protein expression between different samples High sensitivity\ Expensive fluorophores, equipment and software Accurate quantitation\ Good reproducibility Matrix assisted laser desorption ionisation time-of-flight mass spectrometry MALDI-TOF MS Tryptic digests of sample proteins are co-crystallized with matrix, and spotted onto MALDI plate; ionization occurs by pulsed laser radiation primarily absorbed by the matrix, causing desorption and ionization of the analyte; the resulting peptide ions are directed into TOF mass analyzer, where peptide masses are measured by determining the time required for the ions to traverse the length of the flight tube and reach detector Protein identification\ Produces less raw data than other MS techniques\ Requires previous separation of protein mixture\ Amino-acid sequencing\ Data are relatively easy to interpret since most peptides carry only one charge and are present as a single peak in a spectrum Hampered identification of small acidic and integral membrane proteins Determination of the type and position of post-translational modifications Multidimensional protein identification technology MudPIT Mixture of tryptic peptides is resolved by the microcapillary column packed with reversed-phase resin followed by strong cation exchange resin; peptides are eluted directly from the column into the mass spectrometer to be rapidly analyzed Large-scale protein analysis of complex biological mixtures\ Detects proteins of wide range of pI, abundance and sub-cellular distribution\ Time-consuming Requires experienced personnel Does not detect protein activity nor interactions Limited throughput Generates the vast stream of raw data Identification of protein complexes\ Employed directly on crude samples\ Determination of post-translational modifications\ Easily automated High resolving power\ Quantitative analysis of protein expression High sensitivity Surface enhanced laser desorption and ionization time-of-flight mass spectrometry (ProteinChip Technology) SELDI-TOF MS Protein solutions are applied to the spots of ProteinChip Arrays that contain either chemically (anionic, cationic, hydrophobic, hydrophilic, or metal ion) or biochemically (immobilized antibody, receptor, DNA, enzyme, etc.) active surface retaining proteins according to their specific physicochemical properties; after adding matrix solution to bound proteins, the latter are ionized with nitrogen laser and their molecular masses measured by TOF mass analyzer. As a result, unique protein abundance profiles of species bound to the chip surface are obtained. Biomarker discovery\ Suitable for crude biological samples (body fluids, cells)\ Additional MS analysis needed for determining the identity of differentially expressed protein species Characterization of protein-protein and protein-DNA interactions and post-translational modifications (glycosylation and phosphorylation) High-throughput capability\ High sensitivity\ Detects proteins with molecular weights lower than 6-kDa\ High precision and reproducibility Isotope-coded affinity tags ICAT Two different protein samples are labelled at cysteines with the isotopically light and heavy ICAT reagents, combined and digested with trypsin; ICAT-labeled peptides are isolated by avidin affinity chromatography and analyzed by HPLC coupled to a tandem mass spectrometer; the ratio of ion intensities from co-eluting ICAT-labeled pairs permits the quantification, while a subsequent\ Sequence identification and quantification of proteins in complex mixtures\ Selects only cysteine-containing peptides and thus effectively reduces the complexity of the peptide mixtures Incomplete proteome coverage (10-20% of the whole cell proteome) MS/MS scan provides the protein identification Analysis of protein changes in specific subcellular fractions Laser-capture microdisscetion LCM A stained tissue slide is placed under a microscope, and a specific thermoplastic polymer film is placed over the tissue; the cells of interest are shot by an infrared laser pulse, which melts and fuses the film around the targeted cells; the cells embedded in the polymer are lifted away from the remaining tissue Isolation of pure cell populations from heterogeneous tissue sections prior to proteomic analyses focused on the investigation of novel biomarkers and drug targets High-throughput\ Requires competency in identifying the cells of interest\ Reduces sample heterogeneity\ Limited timeframe for microdissecting fresh frozen tissue. Increases the specificity of signals obtained in downstream protein analysis Reverse-phase protein microarrays RPMA Cell lysates are arrayed on nitrocellulose-coated glass slides binding denatured proteins; the slide is probed with a single antibody specific for an antigen of interest; upon signal development and imaging, the relative proportion of the analyte protein molecules can be compared between test samples on the array Functional mapping of known cell-signalling networks or pathways\ High-throughput\ The lack of availability of high-quality, specific antibodies\ Characterization of protein-protein, protein-DNA, and/or protein-RNA interactions Requires low sample volume\ Hampered analysis of low-abundance post-translational events Extremely sensitive analyte detection\ Good reproducibility, sensitivity, and robustness --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ::: 2-DE followed by MALDI-TOF MS analysis represents the workhorse of the vast majority of published proteomics studies on metastasis research, and this approach has proved effective in measuring protein expression patterns within cells, tissues and bodily fluids uncovering many novel metastasis-related proteins, such as chloride intracellular channel 1 (CLIC1). CLIC1 was specifically correlated with metastasis of gallbladder carcinoma, the most frequent form of bile duct cancer, as its expression was significantly up-regulated in the highly metastatic gallbladder cancer GBC-SD18H cell line when compared to the poorly metastatic GBC-SD18L cell line \[[@B67]\]. In addition, the overexpression of CLIC1 promoted cell motility and invasion of GBC-SD18L cells, while RNA interference of CLIC1 remarkably decreased cell motility and invasive potency of GBC-SD18H cell line. Similarly, proteomics profiling of tumor tissues from gastric cancer revealed CLIC1 to be significantly up-regulated in 67.9% of the patients \[[@B68]\]. This study revealed significant correlation between elevated expression of CLIC1 and lymph node metastasis, lymphatic invasion, perineural invasion, advanced pathological stage and poor survival in gastric cancer, which highlights the role of CLIC1 in tumor invasion and metastasis in gastric cancer. S100A11 protein is a calcium-binding protein implicated in a variety of biological functions such as proliferation and differentiation, whose relation with tumor progression and invasion was substantiated by recent proteomics studies. Tian *et al.*\[[@B69]\] carried out comparative 2-DE/MALDI-TOF MS analysis of non-metastatic and highly metastatic non-small cell lung cancer (NSCLC) cell lines and found S100A11 to be specifically up-regulated in the metastatic cell line. Immunohistochemical staining of 65 primary NSCLC tissues and 10 matched local positive lymph node specimens confirmed that the over-expression of S100A11 in NSCLC tissues was significantly associated with higher tumor-node-metastasis stage and positive lymph node status, implying regulatory role of this protein in promoting invasion and metastasis of NSCLC \[[@B69]\]. Using the same proteomics strategy, S100A11 was also reported to be up-regulated in metastatic hepatocellular carcinoma (HCC) tissues \[[@B70]\], and to rise in the expression level with the progression of colorectal cancer \[[@B71]\]. In view of these studies, S100A11 protein might be considered as useful candidate molecule for early diagnosis and intervention of NSCLC, HCC and colorectal metastases. The development of fluorescent dye labels allowing the comparison of two different samples on a single gel referred to as 2-D fluorescence difference gel electrophoresis (2-D DIGE) has improved reproducibility, more accurate quantitation and spot statistics between gels in comparison with conventional 2-DE. This method was successfully applied to screening potential biomarkers for early detection of prostate cancer and lung squamous carcinoma metastases. Pang *et al.*\[[@B72]\] analyzed protein samples from localized and lymph node metastatic prostate cancer (LNM PCa) as well as benign prostatic hyperplasia tissues, and found increased expression of e-FABP5, MCCC2, PPA2, Ezrin and SLP2 along with reduced expression of SM22 in LNM PCa tissues. Importantly, e-FABP5 levels were significantly increased in the sera of patients with LNM PCa. These findings were in line with the previous studies revealing an over-expression of e-FABP5 protein in PCa tissues \[[@B73]\]. Overexpression of e-FABP5 was shown to induce metastasis by up-regulating VEGF, which plays a crucial role in the metastatic cascade \[[@B74]\]. Therefore, increased e-FABP expression is a possible target to inhibit the malignant progression of prostate cancer cells. Using the same technique, Yao *et al.*\[[@B75]\] compared the protein profiles between laser capture-microdissected (LCM) lung squamous carcinoma (LSC) cells with and without lymph node metastasis (LNM), and found rise in the expression level of HSP27, Annexin A2, and CK19, whereas 14-3-3 σ had reduced expression in LNM LSC. Additional immunohistochemical analyses confirmed that these proteins are indeed correlated with several clinicopathological variables and prognosis of LSC. The increasing use of high-throughput platforms for the analysis of protein expression levels driven by technological improvements in mass spectrometry and array-based technologies has pushed the boundaries of clinical oncoproteomics. Serum protein pattern profiling by ProteinChip Technology (SELDI-TOF MS) has emerged as novel approach to discover protein signatures capable of discriminating patients with primary cancer from those with metastasis, as demonstrated by several recent studies on laryngeal squamous cell carcinoma \[[@B76]\], colorectal \[[@B77]\], ovarian \[[@B78]\], lung \[[@B79]\], prostate \[[@B80]\], breast \[[@B81]\] and gastric cancer \[[@B82]\]. Collectively, these studies clearly showed diagnostic and prognostic value of ProteinChip technology. In addition, Goncalves *et al.*\[[@B83]\] used this proteomics approach on a high-risk early breast cancer population receiving standard adjuvant chemotherapy, and managed to identify a post-operative serum proteomic profile that might predict metastatic relapse. Currently, there are no satisfactory screening and early diagnostic strategies for metastatic cancer. Due to its quantification capability and reproducibility, SELDI-TOF MS represents a serious candidate tool for rapid and accurate high-throughput screening of cancer patients. Major breakthrough in metastasis research represents the combination of LCM and protein microarray technologies, which has been applied to the analysis of human metastatic breast and ovarian cancer tissue samples in phase II clinical trials at the National Institutes of Health National Cancer Institute \[[@B84]\]. Similarly, Sheehan *et al.*\[[@B85]\] utilized reverse phase protein microarray (RPMA) technology to profile a matched cohort of primary and metastatic ovarian carcinomas using phosphorylation-specific antibodies. Strikingly, the metastatic signatures were clearly very different from the primary tumor taken at the same time at surgery, and these fingerprints appeared to be virtually patient-specific, which underscores the critical need for patient-tailored therapy designed to specifically target the disseminated cells. The same study revealed several phosphorylated proteins that were differently expressed between primary and metastatic tissues, including the phosphorylated forms of c-Kit, Ask, myristoylated alanine-rich C kinase substrate, IκBα, and Ras-GRF \[[@B85]\]. Importantly, metastasis correlated with activation of c-Kit, which was previously demonstrated to pertain to advanced stage and chemotherapy resistance in serous ovarian carcinomas. RPMA technology has also proved beneficial in investigating the cellular events that accompany metastatic progression, as exemplified by Paweletz *et al*\[[@B86]\] who used RPMAs to compare LCM-specimens of histologically normal prostate epithelium, prostate intraepithelial neoplasia, and invasive prostate cancer. Amplification of antibody-antigen complexes on the microarrays revealed a statistically significant increase in phosphorylation of Akt and a decrease in phosphorylation of Erk in premalignant and invasive prostate cancer. The authors draw a conclusion that these shifts in protein abundance indicated activation of pro-survival signaling pathways with cancer invasion. The same study also demonstrated that downstream components of the apoptotic cascade, namely cleaved and noncleaved caspase-7 and PARP were also shifted towards prosurvival function during cancer progression \[[@B86]\]. Based on obtained data, the authors proposed a hypothetical model of prostate cancer progression according to which activation of Akt suppresses apoptosis, probably *via*inactivation of its substrate GSK3-β, which might cause an imbalance between cell proliferation and death leading to the accumulation of cells within prostate gland \[[@B86]\]. Simultaneously, transient ERK activation and turning on pro-survival pathways might be associated with cellular migration responsible for invasion. Akt seems to play a central role in prostate cancer metastasis, as its activation could foster cell motility and survival during stromal invasion. Despite rapid development of proteomics technologies, cancer stem cell proteomics is still in its infancy. This could be ascribed to the extremely low availability of putative CSCs, as these cells represent only a small fraction of the overall cancer cell population. In addition, the methods for isolating a large enough sample of pure CSCs have not been developed yet \[[@B87]\]. However, recent proteomics studies on leukemic \[[@B88],[@B89]\] and pancreatic cancer stem cells \[[@B90]\] are bright examples on how to successfully overcome the issue of sample limitations in CSC research. Tibes *at al.*\[[@B88]\] demonstrated that RPMAs could be reliable, reproducible high-throughput approach for analyzing protein expression and phosphorylation status in primary acute myelogenous leukemia cells, cell lines and stem cells. Importantly, this group of authors found that leukemic stem cells had apparently different protein signature compared with normal stem cells, and observed different levels of protein expression when normal and leukemic CD34+/CD38+ and CD34+/CD38- cells were compared, or when leukemic and normal stem cells were compared. Novel perspective on human leukemogenesis was provided by Ota *et al.*\[[@B89]\], who utilized 2-DE/MS for protein profiling of isolated AC133^+^leukemic blasts from 13 individuals with acute leukemia or related disorders. They detected 10 differentially expressed proteins including NuMA, heat shock proteins, and redox regulators. An over-expression of HSP70 family proteins in leukemic blasts was proposed to be directly associated with leukemogenesis, or to the development of drug resistance. Importantly, the finding that the abundance of the nuclear mitotic apparatus protein (NuMA) in leukemic blasts was related to the number of chromosomal abnormalities raised the possibility that over-expression of NuMA perturbs cell cycle progression by inhibiting mitosis resulting in the chromosomal instability. The finding that the forced expression of NuMA resulted in G2/M arrest and apoptosis clearly shows that some other genetic events, besides aberrant expression of NuMA, are required for malignant transformation to leukemic cells. Characterization of the CSCs opens a new avenue for designing novel therapeutic strategies against cancer. However, crucial to this task will be identification of specific cell surface antigens (markers) (Table [2](#T2){ref-type="table"}) for CSCs detection and isolation from the heterogeneous tumor population. Hereby, membrane proteomics plays an important role, as demonstrated by He *et al.*\[[@B91]\], who employed the combination of lectin microarray and liquid chromatography-tandem mass spectrometry (LC-MS/MS) to discover novel cell surface glycoprotein markers of a glioblastoma-derived stem-like cell line. These authors identified six differentially expressed proteins between the stem-like glioblastoma neurosphere culture and traditional adherent glioblastoma cell line, whereby receptor-type tyrosine-protein phosphatase zeta, Tenascin-C, Chondroitin sulfate proteoglycan NG2, Podocalyxin-like protein 1 and CD90 were up-regulated, and CD44 was down-regulated \[[@B91]\]. Further elucidation of the biological roles of these proteins might prove important for an early diagnosis and improved treatment of glioblastoma. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Representative cell surface markers for human cancer stem cells ::: Type of cancer stem cells Cell surface markers --------------------------------------- -------------------------------------------------- Acute myelogenous leukemia; CD34+, CD38-, CD44, CD123+ Chronic myeloid leukemia CD34+, CD38-, CD123+ B-acute lymphogenous leukemia CD34+, CD38-, CD19+ Ph1-acute lymphogenous leukemia CD34+, CD38- Blast-crisis CML CD34+, CD38+, CD123+ Myeloproliferative disorder CD117+ Glioblastoma CD133+ Medulloblastoma CD133+ Pilocytic astrocytoma CD133+ Anaplastic ependymoma CD133+ Breast CD44+, CD24^-/low^, ESA+ Prostate CD133+/alpha 2 beta 1 integrin/CD44+ CD44+/CD24- Ovarian cancer CD44+, MyD88+ Colon cancer CD133+, CD44+, CD166+, E-CAMhig Pancreatic cancer CD133+, CD44+, CD24+ Hepatocellular cancer CD133+ Head and neck squamous cell carcinoma CD44+ Bone sarcomas Stro-1+, CD105+, CD44+ Melanoma CD20+, CD133+ Lung cancer CD133+ Liver CD133+, CD90+ Central nervous system CD133+ ::: Although yielding many useful information on cancer invasiveness and progression, conventional genomic and proteomic platforms are still limited in their capacity to identify changes in protein activity caused by post-translational mechanisms \[[@B92]\]. Most proteomic techniques provide information on protein abundance, which does not necessarily correlate with enzyme activity, since most enzymes are expressed as inactive zymogens or reside in complex with their endogenous inhibitors \[[@B93]\]. As an alternative, a new strategy termed activity-based protein profiling (ABPP) has emerged based on the use of enzyme family-specific activity-based chemical probes (ABPs) linked to specific reporter groups that by nature only target and subsequently tag the active form of these enzymes both *in vitro*and *in vivo*\[[@B94]\]. Jessani et al. \[[@B95]\] used this approach to profile serine hydrolase activities across a panel of human breast and melanoma cancer cell lines, and found that highly invasive cancer cells exerted secreted/membrane serine hydrolase activity profiles nearly orthogonal to those displayed by their less aggressive counterparts indicating that invasive cancers may share proteomic signatures that are more reflective of their cellular phenotype than tissue of origin. Importantly, they detected the up-regulation of two enzyme activities in invasive cancer lines, namely urokinase, a secreted serine protease with previously established role in tumor progression, and a membrane-associated serine hydrolase KIAA1363, the latter implicated as a new marker of tumor progression \[[@B95]\]. Similar study combining ABPP and metabolomics (rapid, high-throughput characterization of the small molecule metabolites including any metabolic intermediates, hormones and other components of signaling pathways found in an organism \[[@B96]\]) established a central role of KIAA1363 in an ether lipid signaling network bridging platelet-activating factor and lysophosphatidic acid \[[@B97]\]. As evident from these studies, integration of several different global profiling technologies may illuminate biochemical networks pertinent to cancer development and progression. In this respect, metabolomic profiling has already proved beneficial in characterizing the metabolic features of hepatocellular carcinoma \[[@B98]\], breast cancer \[[@B99]\], renal cell carcinoma \[[@B100]\] and prostate cancer \[[@B101]\] metastases. Metabolic alterations during cancer progression and metastasis revealed by these studies might provide new putative diagnostic and prognostic biomarkers as well as new therapeutic targets, such as e.g. sarcosine, an *N*-methyl derivative of glycine elevated most robustly in metastatic prostate cancer and detectable in the urine of men with organ-confined disease \[[@B101]\]. However, metabolic profiling studies involving CSCs to identify the key metabolites inherent to tumor progression are still scarce. The pioneering work in this field was performed at the American company Stemina Biomarker Discovery, whose scientists are focused on identifying metabolomic biomarkers that are potential indicators of drug efficacy against CSCs for the establishment of novel drug screening assays (<http://www.stemina.com/web/publications.php>). Specifically, they found small molecules unique to CSCs derived from glioblastoma multiforme (GBM), and identified unique metabolomic footprint of three different brain tumor stem cell lines (BTSC). Besides their potential as a screening tool for measuring GBM presence and progression, identified metabolite molecules might also serve as therapeutic targets in order to manufacture drugs specifically targeting GBM and BTSC cells. Conclusion ========== Many hypotheses have been postulated to explain the intricate nature of the metastatic process, but none of them completely accounted for the actual biological and clinical observations. Consequently, metastasis still remains an open issue with only few metastasis-inducing proteins experimentally validated so far. Global scale proteomics studies undoubtedly revealed specific metastatic markers often related to cell-signalling processes; however, they proved to be patient-specific rather than type- or tumor stage-specific, which necessitates a need for individual therapeutic approach towards each patient. Due to inconsistency between experimental results and clinical observations, a novel metastatic paradigm where serial and parallel metastatic processes are adequately integrated is needed to account for these differences. In addition, organ-tropic gene signatures were shown to bear a potential to improve patient risk stratification and therapeutic treatment. New diagnostic tools are therefore urgently required in clinical practice to detect patients who will benefit from the adjuvant chemotherapy after primary tumour resection. Consequently, this will lead to the improvements in patient treatment and reduction of adverse effects in patients who are traditionally unnecessarily treated with chemotherapy. While study of the metastases molecular mechanisms is far from trivial, the results discussed in this paper suggest that the benefits to our understanding of the cellular basis of metastasis more than justify the efforts employed. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= SKP coordinated the preparation of the manuscript, designed the manuscript structure, performed the literature search, wrote the majority of manuscript text and prepared accompanying Figure [1](#F1){ref-type="fig"} and Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}. MS has been involved in extensive literature search, language revision, preparation of Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"} and wrote the majority of the section \"Technological advances in metastasis research: proteomics\". HB has been involved in literature search and revision of clinical aspects in the manuscript and wrote majority of section \"Organ tropism and metastatic gene signature\". SS contributed to the main manuscript idea, concept preparation, literature search and revised the final manuscript. KP proposed the manuscript idea, and has been involved in literature search, manuscript writing (in particular sections dealing with concept of cancer stem cells and molecular mechanisms underlying metastases), review of clinical aspects and final manuscript review. All the authors read and approved the final manuscript. Acknowledgements ================ This work was supported by the Foundation of Croatian Academy of Sciences and Arts and the Croatian Ministry of Science, Education and Sports (grants number 335-0982464-2393, 335-0000000-3532, 101-0982464-2277).

PubMed Central

2024-06-05T04:04:19.049424

2011-2-22

PMC3052237

Background ========== The epidermal growth factor receptor (EGFR) is over-expressed in up to 90% of squamous cell carcinoma of the head and neck (SCCHN) and has been postulated to be a key molecular target in this malignancy \[[@B1]\]. EGFR signal transduction leads to cell proliferation, invasion, angiogenesis and metastasis \[[@B2]\]. EGFR overexpression and aberrant EGFR gene copy number (EGFR GCN) have been associated with poorer prognosis and disease-specific survival in SCCHN \[[@B1],[@B3],[@B4]\]. Therapies targeted against EGFR have demonstrated modest activity either alone or in combination with chemotherapy in both locally advanced \[[@B5]\] and recurrent and/or metastatic SCCHN \[[@B6]-[@B10]\]. No validated biomarkers exist to predict the response to EGFR inhibitors in SCCHN. The most common EGFR truncation mutation, EGFR variant III (EGFRvIII), harbors an in-frame deletion of exons 2 to 7 (801 bp), resulting in a truncated extracellular EGF-binding domain that is constitutively activated and ineffectively ubiquinated \[[@B11],[@B12]\]. EGFRvIII is found in many human cancers and is present in \~40% of glioblastomas and 5% of lung squamous cell carcinomas, where it confers tumorigenicity and dose-dependent resistance to gefitinib in pre-clinical models \[[@B13],[@B14]\]. The prevalence of EGFRvIII in SCCHN was first reported as 43% in one study of 33 SCCHN tumors \[[@B15]\]. EGFRvIII-transfected SCCHN cells had decreased apoptosis in response to cisplatin and decreased growth inhibition following treatment with the EGFR monoclonal antibody cetuximab compared with controls \[[@B15]\]. EGFRvIII is an interesting therapeutic target because unlike wild-type EGFR, EGFRvIII is not found in normal tissue. EGFRvIII is proposed to account for limitations in response to current EGFR inhibitors, however in patients with SCCHN tumors harboring EGFRvIII response to EGFR tyrosine kinase inhibition (TKI) is unknown. HPV infection is a risk factor for the development of SCCHN. HPV DNA is found in 20-30% of SCCHN and up to 40-66% of SCCHN of the oropharynx \[[@B16],[@B17]\]. HPV positive oropharyngeal tumors are clinically and molecularly distinct from HPV negative tumors \[[@B18],[@B19]\] and associated with a more favorable prognosis \[[@B20]\]. HPV positive status prospectively predicts survival and response to induction chemotherapy and chemoradiation in stage III or IV oropharynx cancers \[[@B21],[@B22]\] and better response to radiotherapy alone \[[@B23]\]. The combination of low HPV titers and high EGFR expression was associated with worse overall survival in oropharynx cancer \[[@B22]\]. Inactivation of pRb by HPV E7 protein results in overexpression of p16 protein, thus p16 immunostaining has served as a surrogate marker for HPV-associated SCCHN. Patients with tumors lacking both p16 expression and HPV (p16-/HPV-) had the worst disease-specific survival compared to tumors with p16+/HPV+, p16-/HPV+ or p16+/HPV- types \[[@B24]\]. Despite the importance of HPV in the pathogenesis and prognosis of SCCHN in response to chemotherapy and radiation, the role of HPV DNA and response to EGFR inhibitors in SCCHN is unclear. c-MET, a proto-oncogene tyrosine kinase receptor, is overexpressed in SCCHN, and its ligand, hepatocyte growth factor (HGF), stimulates cell proliferation, motility and invasion \[[@B25]\]. c-MET overexpression has been associated with disease progression in oral squamous cell carcinoma (OSCC) \[[@B26]\]. Elevated serum HGF is associated with resistance to chemoradiation and reduced survival \[[@B27]\]. c-MET amplification and mutations of MET confer an invasive phenotype associated with metastases in SCCHN \[[@B28]\]. Ligand-independent constitutive activation of c-MET via its heterodimerization with EGFR has been identified as a contributing mechanism of acquired resistance to cetuximab in SCCHN \[[@B29]\]. The role of c-MET in response to EGFR TKI in the clinical setting in SCCHN is unknown. In this study, we examine the prevalence of EGFRvIII, HPV, p16, c-MET and EGFR GCN in patients with R/M SCCHN and explore the potential prognostic and predictive roles of these biomarkers in patients treated with or without EGFR TKI. We hypothesized that EGFRvIII and c-MET would be associated with poorer prognosis or response to EGFR TKI, while HPV and p16 expression would predict improved clinical outcomes and response to treatment. Methods ======= Patients -------- We obtained approval from the University Health Network Research Ethics Board to evaluate the archival formalin-fixed paraffin embedded (FFPE) tumor specimens of patients with R/M SCCHN who were treated in four phase II trials for R/M SCCHN at Princess Margaret Hospital conducted from 2000-2005. Two of the four trials involved the EGFR TKI erlotinib (phase II trial of erlotinib \[[@B8]\], phase II trial of erlotinib and cisplatin \[[@B7]\]) and the remaining two trials used other non-EGFR targeted agents (phase II trial of the kinesin spindle protein inhibitor ispinesib \[[@B30]\], phase II trial of the multi-kinase antiangiogenic inhibitor sorafenib \[[@B31]\]). The medical records and case report forms were reviewed to obtain patient demographics, primary tumor site, treatment details and clinical outcome (response rate, time to progression and overall survival). Specimen Characteristics ------------------------ Archival FFPE tumor specimens were available in 35 of 48 patients (73%) treated with erlotinib and 18 of 37 (49%) patients treated with non-EGFR targeted agents. H&E stained sections were examined by a histopathologist (B.P-O.) to confirm the presence of \>80% tumor in the specimens evaluated. Assay Methods/Molecular Assays ------------------------------ ### EGFRvIII Mutation Detection #### RNA Isolation RNA was isolated in tumor area on the FFPE slides guided by H&E-stained serial sections. The tissues were deparaffinized by xylene and ethanol. Total RNA from paraffin-embedded tissues was extracted using RecoveryAll™ Total Nucleic Acid isolation Kit (Ambion Diagnostics, Streetsville, Ontario, Canada). #### Real-time RT-PCR Reverse transcription was done using TaqMan Reverse transcription reagent kit (Roche, Branchburg, New Jersey) according to the manufacturer\'s protocol. Reverse transcription reaction was done in a total volume of 25 mL including RNA template, 1.25 mL random hexamer, reverse transcription buffer, 5.0 mL dNTP, 5.5 mL MgCl, RNase inhibitor, and M-MLV reverse transcriptase. Real-time PCR was performed in duplicate in 25 mL reaction volumes using Platinum SYBR Green qPCR SuperMix-UDG (Invitrogen, Carlsbad, California) and a 7900HT instrument (Applied Biosystems, Foster City, California). The amplification conditions were: 50°C for 2 min., 95°C for 2 min., 40 cycles of 95°C for 15 sec. and 60°C for 1 min. #### Data Analysis of real-time PCR A mixture of at least eight normal FFPE tissue samples was used as a wildtype, normal control. The EGFRvlll cell line (u373flag*EGFRvlll*) was used as a positive control. The relative expression of *EGFR*exon 4 to *EGFR*exon 9 was determined using the delta delta Ct (ΔΔCt) method. All samples were run in duplicate, and the mean Ct number was used for data analysis. The difference in Ct values (Δthreshold cycle, ΔCt = Exon 9 Ct - Exon 4 Ct) was calculated for each RNA sample. The ΔCt from the normal tissue mixture was then subtracted from the ΔCt of the test sample to generate a ΔΔCt. A negative result occurs when the fold change (exon 9: exon 4 calculated as 2^-\ ΔΔCt^) is less than a value of 5. This value was arbitrarily chosen to ensure that no false positives were called. A positive result occurs when the fold change is the same as or greater than that of the positive control (7). When the fold change of tested samples falls between that of the normal control and the positive control (i.e. between 5 and 7) the results are considered inconclusive. ### HPV DNA Detection The Roche Linear Array HPV Genotyping kit (Roche Molecular Diagnostics, Pleasanton, California) was used to detect 37 low- and high-risk HPV types from FFPE tissues. In brief, FFPE sections were deparaffinized and DNA was extracted using a column based method (QIAamp, Qiagen, Valencia, California). HPV detection was performed using PCR amplification followed by hybridization of the amplified products to oligonucleotide probes and subsequent colorimetric determination. All experiments included an HPV positive control and an HPV negative control. HPV DNA by in situ hybridization (ISH) using the INFORM HPV III Family 16 probe (Ventana Medical Systems Inc., Tucson, Arizona) which detects genotypes 16, 18, 31, 33, 39, 35, 45, 51, 52, 56, 58 and 66, was performed according to the manufacturer\'s guidelines using the Ventana Benchmark automated slide staining system. All experiments included an HPV positive control and an HPV negative control. Slides were scored as positive if a punctate or diffuse pattern of signal were observed in the tumor nuclei. ### P16 and c-MET Detection Immunohistochemistry (IHC) for p16 and c-MET using the Ventana Benchmark XT auto-immunostainer (Tucson, Arizona) was performed on FFPE sections cut at 4 mm thick. Standardized staining protocols were provided by Ventana for the CINtec p16 Histology kit (MTM Laboratories Inc, Westborough Massachusetts) and c-MET antibody (SP44, rabbit monoclonal, Ventana Medical Systems Inc., Tucson Arizona). Controls were included in each assay, comprising of positive tissue controls and negative controls. All p16 and c-MET IHC slides were reviewed independently by two observers (B.P.O. and N.G.C.) without knowledge of EGFRvIII, HPV status or clinical outcome. p16 staining in SCCHN is generally observed to be dichotomous and scored as absent (weak or no staining) or present (strong and diffuse staining) \[[@B32]\]. c-MET IHC slides were assigned a semi-quantitative score based on the product of an intensity score (0 = no staining or equal to background, 1 = weak, or 2 = strong) and percent of area stained (0 = 0%, 1 = 1-30%, 2 = 31-60%, 3 = \>60%). Sections with an inter-observer variation were reassessed by a double-headed light microscope to achieve consensus. ### EGFR Gene Copy Number Archival tumor specimens were analyzed for EGFR GCN using fluorescent in situ hybridization (FISH) as previously described \[[@B33],[@B34]\]. One hundred non-overlapping interphase nuclei were scored for EGFR and CEP7 copy number and classified into six categories (University of Colorado Scoring system) by a reviewer blinded to clinical outcome (O.L.) \[[@B35]\]. Statistical Methods ------------------- Descriptive statistics were used to summarize the study cohort and to estimate the parameters of interest. Ninety-five percent confidence intervals were obtained for estimates of the presence of EGFRvIII, HPV, p16, c-MET and EGFR GCN. Exploratory analyses were performed to characterize the relationships between EGFRvIII, HPV, p16, c-MET and EGFR GCN with baseline patient characteristics and outcomes. Only patients with conclusive EGFRvIII results were included in the correlation analyses. The Kaplan-Meier method was used to estimate the overall survival and time to progression. All biomarkers were examined in univariate analysis of overall survival and time to progression using Cox proportional hazards model. Only those which were significant at 0.10 (two-sided) level in the univariate analysis were entered in the multivariate analysis and markers that remained significant at 0.05 (two-sided) level in the multivariate analysis were considered significant prognostic factors. Statistical analyses were performed using the SAS 9.1 software package (SAS Institute, Cary, North Carolina). Results ======= Patients -------- The clinical characteristics of the 53 patients in our study are described in Table [1](#T1){ref-type="table"}. For the entire cohort, the overall response rate (CR+PR) to study treatment was 4/53 (7.5%), median time to progression (TTP) was 1.8 months (95% CI 1.6-2.7) and median overall survival (OS) was 5.9 months (95% CI 4.5-8.7). Patients in the erlotinib group had a higher median OS of 7.9 months (95% CI 4.7-9.8) compared to patients in the non-erlotinib group with median OS of 4.2 months (95% CI 2.9-7.0) (p = 0.011). The erlotinib group had a higher TTP than the non-erlotinib group 2.7 months (95% CI 1.6-3.5) vs 1.5 months (95% CI 1.3-1.8) (p = 0.0009). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Clinical characteristics of the entire study cohort (n = 53) ::: Clinical Characteristic Number ------------------------- --------------------------- ------------ Median Age (Range) 56 (15-78) Gender Female:Male 12:41 ECOG Performance Status 0:1:2 15:34:4 Locoregional Recurrence Yes:No 45:8 Distant Metastases Yes:No 19:34 Primary Tumor Site Oropharynx 20 Larynx 14 Oral cavity 10 Hypopharynx 2 Neck mass unknown 4 Paranasal sinus 3 Histologic Grade Well differentiated 5 Moderately differentiated 33 Poorly differentiated 14 Infiltrating basaloid 1 Prior Therapy Chemotherapy 21 Radiation Therapy 51 Surgery 41 Race Asian 8 Black 42 Caucasian 9 Other 1 Smoker Current 33 Former 3 Never 15 Unknown 2 Erlotinib Yes: No 35:18 Best Response Partial response 4 Stable disease 20 Progressive disease 24 Inevaluable 5 ::: ### Expression of EGFRvIII mutation by real-time PCR As the previously reported immunohistochemistry-suitable antibody \[[@B15]\] against EGFRvIII is no longer available, EGFRvIII expression is analysed using the RT-PCR method. The presence of EGFRvIII mutation was detected in 22 patients (42%) (Table [2](#T2){ref-type="table"}), negative in 19 patients and inconclusive in 12 patients (Table [3](#T3){ref-type="table"}). The median EGFRvIII fold change was 6.8 (0.56 to 576.36) for all patients, 15.0 (4.1 to 576.36) for patients in the EGFRvIII positive group, 1.8 (0.6 to 4.3) for patients in the EGFRvIII negative group, and 6.5 (6.2 to 6.8) for patients in the inconclusive group. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### EGFRvIII mutation positive detected by RT-PCR (n = 22) ::: Case Treatment Primary Site Specimen Site EGFRvIII by RT-PCR EGFRvIII Fold Changes HPV DNA by Linear Array P16 IHC MET score by IHC EGFR FISH ------ ----------------------- --------------------------- ---------------------- -------------------- ----------------------- -------------------------- --------- ------------------ --------------- 1 Erlotinib hypopharynx Untreated primary \+ 11.33 \- \- High Low polysomy 2 Erlotinib oral cavity Untreated primary \+ 7.01 \- \- High Disomy 3 Erlotinib larynx Untreated primary \+ 61.77 \- \- Low Low polysomy 4 Erlotinib larynx Local recurrence \+ 26.64 \- \- Low High polysomy 5 Erlotinib neck mass unknown primary Untreated lymph node \+ 60.04 33 + \+ High Low trisomy 6 Erlotinib oral cavity Untreated lymph node \+ 218.49 \- \- High Amplification 7 Erlotinib oropharynx Unknown primary \+ 15.66 16 + \+ High Low trisomy 8 Erlotinib oral cavity Local recurrence \+ 8.94 \- \- High Low polysomy 9 Erlotinib + Cisplatin neck mass unknown primary Untreated lymph node \+ 127.84 16+ NE NE Low trisomy 10 Erlotinib + Cisplatin larynx Node recurrence \+ 576.36 16+ \+ High Disomy 11 Erlotinib + Cisplatin larynx Untreated primary \+ 8.26 \- \- Low Failed 12 Erlotinib + Cisplatin larynx Untreated primary \+ 17.38 \- \- High Low polysomy 13 Erlotinib + Cisplatin oropharynx Untreated primary \+ 14.28 16+ \+ High Disomy 14 Erlotinib + Cisplatin oral cavity Untreated primary \+ 69.68 \- NE NE Low polysomy 15 Erlotinib + Cisplatin oral cavity Untreated primary \+ 11.11 53+, 58+, 6+, 52+ \- High Low polysomy 16 Sorafenib oropharynx Untreated primary \+ 7.71 16 + \+ Low Low polysomy 17 Sorafenib oropharynx Untreated primary \+ 7.93 16 + \+ Low Low trisomy 18 Ispinesib neck mass unknown primary Untreated lymph node \+ 4.12 16+, 53+, 51 + \+ High High polysomy 19 Ispinesib oropharynx Untreated primary \+ 15.61 16+, 53+, 33+, 51+, 58 + \- High High polysomy 20 Ispinesib larynx Untreated primary \+ 218.26 \- \- High Disomy 21 Ispinesib oropharynx Local recurrence \+ 29.25 16+, 53+, 51+ \+ Low Disomy 22 Ispinesib hypopharynx Local recurrence \+ 9.31 16+, 53+, 58+, 52+ \- High Low polysomy Abbreviations: +, positive; -, negative; NE, not evaluable ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### EGFRvIII mutation negative detected by RT-PCR (n = 19) and inconclusive cases (n = 12) ::: Case Treatment Primary Site Specimen Site EGFRvIII by RT-PCR EGFRvIII Fold Changes HPV DNA by Linear Array P16 IHC MET score by IHC EGFR FISH ------ ----------------------- --------------------------- ---------------------- -------------------- ----------------------- ------------------------- --------- ------------------ ---------------------------- 23 Erlotinib larynx Local recurrence \- 1.76 \- \- High Low polysomy 24 Erlotinib larynx Untreated primary \- 0.95 \- \- NE NE 25 Erlotinib oropharynx Untreated primary Incon. NE 16 + \- Low Failed 26 Erlotinib larynx Untreated Incon. NE \- \- NE NE 27 Erlotinib oropharynx Local recurrence \- 2.56 16 + \- High Failed 28 Erlotinib neck mass unknown primary Local recurrence \- 1.81 \- \- High Low polysomy 29 Erlotinib paranasal sinus Untreated primary \- 0.56 \- \- High High polysomy 30 Erlotinib oropharynx Untreated primary \- 3.16 16 + \+ High Low trisomy 31 Erlotinib larynx Untreated primary \- 3.17 \- \- High Low polysomy 32 Erlotinib oropharynx Local recurrence NE NE \- \+ Low Low polysomy 33 Erlotinib larynx Untreated primary \- 3.16 \- \- High High trisomy 34 Erlotinib + Cisplatin oropharynx Untreated primary Incon. NE Incon. \+ Low NE 35 Erlotinib + Cisplatin larynx Untreated primary \- 4.29 \- \- Low Low polysomy 36 Erlotinib + Cisplatin oral cavity Untreated primary \- 0.9 \- \- Low High polysomy 37 Erlotinib + Cisplatin paransal sinus Local recurrence Incon. NE 16+, 53+ Incon. Low Low polysomy 38 Erlotinib + Cisplatin larynx Local recurrence Incon. NE 16+, 53+, 33+, 51+ \+ High Disomy 39 Erlotinib + Cisplatin paranasal sinus Local recurrence \- 2.59 6+ \- High NE 40 Erlotinib + Cisplatin oral cavity Untreated primary \- 2.01 \- \- High Disomy 41 Erlotinib + Cisplatin oropharynx Untreated primary Incon. NE \- \+ High Low trisomy 42 Erlotinib + Cisplatin oral cavity Untreated primary Incon. NE \- \- High High polysomy 43 Sorafenib oral cavity Untreated primary Incon. 6.8 \- \- Low Low polysomy 44 Sorafenib oropharynx Untreated primary \- 0.99 \- \+ High Amplification/High trisomy 45 Sorafenib oral cavity Local recurrence Incon. NE Incon. \- Low Low polysomy 46 Sorafenib oropharynx Untreated primary Incon. 6.22 16 + \+ Low Low trisomy 47 Sorafenib larynx Local recurrence \- 2.28 \- \+ High High polysomy 48 Sorafenib oropharynx Untreated primary \- 0.64 \- \- High Low polysomy 49 Sorafenib oropharynx Local recurrence \- 1.4 \- \- High High polysomy 50 Sorafenib oropharynx Local recurrence \- 1.3 \- \- High High polysomy 51 Sorafenib oropharynx Local recurrence \- 1.76 \- \+ Low Monosomy/Disomy 52 Ispinesib oropharynx Untreated lymph node Incon. NE \- \- High Low polysomy 53 Ispinesib oropharynx Node recurrence \- 4.05 16 + \- Low Amplification Abbreviations: +, positive; -, negative; Incon., inconclusive; NE, not evaluable ::: Patients with tumors harboring the EGFRvIII mutation had similar clinical characteristics to patients without the EGFRvIII mutation (Table [4](#T4){ref-type="table"}). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Presence of the EGFRvIII mutation is not significantly associated with any clinical characteristics ::: Clinical Characteristic EGFRvIII absent (n = 19) EGFRvIII present (n = 22) p-value ------------------------------------------- -------------------------- --------------------------- -------------------- Male - no., (%) 16 (84%) 18 (82%) 0.839 (Fisher\'s) Age - mean, (+/-SD) 53.5 (+/- 11.6) 55.1 (+/- 14.1) 0.685 (t-test) Oropharynx - no., (%) 8 (42%) 6 (27%) 0.318 (Chi-square) Larynx - no., (%) 6 (32%) 6 (27%) 0.763 (Chi-square) Oral Cavity - no., (%) 2 (11%) 5 (23%) 0.271 (Fisher\'s) Distant metastasis - no., (%) 4 (21%) 12 (55%) 0.053 (Fisher\'s) Locoregional recurrence - no., (%) 18 (95%) 16 (73%) 0.099 (Fisher\'s) Well-moderately differentiated - no., (%) 15 (79%) 14 (64%) 0.325 (Fisher\'s) Poorly differentiated - no., (%) 4 (21%) 8 (36%) Prior chemotherapy - no., (%) 9 (47%) 8 (36%) 0.476 (Chi-square) Prior radiotherapy - no., (%) 17 (89%) 22 (100%) 0.209 (Fisher\'s) Prior surgery - no., (%) 17 (89%) 17 (77%) 0.419 (Fisher\'s) Caucasian - no., (%) 15 (79%) 18 (82%) 0.562 (Fisher\'s) Erlotinib treatment - no., (%) 12 (63%) 15 (68%) 0.735 (Chi-square) No erlotinib treatment - no., (%) 7 (38%) 7 (32%) ::: #### EGFRvIII is associated with disease control In univariate analysis, the presence of EGFRvIII was associated with better disease control (Table [5](#T5){ref-type="table"}). Median EGFRvIII fold changes were higher for patients with disease control than patients with progressive disease (11.11 vs. 3.16, p = 0.04). No significant difference was observed between erlotinib-treated (p = 0.21) versus non-erlotinib (p = 0.10) treated patients due to the small sample size (Table [5](#T5){ref-type="table"}). The presence of EGFRvIII mutation was not associated with TTP (HR 0.94 (95% CI 0.33-2.71), p = 0.91) or OS (HR 0.91 (95% CI 0.32-2.60), p = 0.85) (Figure [1](#F1){ref-type="fig"}). ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### EGFRvIII mutation is associated with disease control ::: EGFRvIII Best Response -------------------------------- --------------- ---------- -------------- EGFRvIII absent by ΔΔCt 12 (67%) 6 (33%) P = 0.0099 EGFRvIII present by ΔΔCt 5 (25%) 15 (75%) (Chi-square) EGFRvIII mean fold change 10.62 63.76 P = 0.04 EGFRvIII median fold change 3.16 11.11 (Wilcoxon) Erlotinib treated patients  EGFRvIII absent by ΔΔCt 6 (55%) 5 (45%) P = 0.21  EGFRvIII present by ΔΔCt 3 (33%) 10 (77%) (Fisher\'s) Non-erlotinib treated patients  EGFRvIII absent by ΔΔCt 6 (86%) 1 (14%) P = 0.21  EGFRvIII present by ΔΔCt 2 (29%) 5 (71%) (Fisher\'s) ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Overall survival by EGFRvIII mutation status (HR = 0.91, 95% CI: 0.32-2.60, p = 0.85)**. :::  ::: ### HPV DNA HPV DNA testing by PCR was positive in 20 patients (38%), negative in 31 patients and inconclusive in 2 patients (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). The most prevalent HPV subtype found in our analysis was the high-risk HPV-16 (18/20 cases). The majority of HPV-16 positive tumors were from the oropharynx (12/18). HPV positive tumor status was not significantly associated with disease control (Table [6](#T6){ref-type="table"}), TTP (HR 1.19 (95% CI 0.46-3.11), p = 0.722) or OS (HR 0.88 (95% CI 0.34-2.29), p = 0.788). ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### HPV by PCR, p16 by IHC, c-MET by IHC are not associated with disease control ::: Best Response --------------------- --------------- ---------- -------------- HPV PCR negative 14 (50%) 14 (50%) P = 0.86 HPV PCR positive 9 (47%) 10 (53%) (Chi-square) P16 IHC negative 14 (47%) 16 (53%) P = 0.67 P16 IHC positive 8 (53%) 7 (47%) (Chi-square) c-MET IHC \< or = 2 9 (56%) 7 (44%) P = 0.39 c-MET IHC \>2 12 (43%) 16 (57%) (Chi-square) ::: ### P16 P16 immunoreactivity was detected in 17 patients (32%), absent in 33 patients and inconclusive in 3 patients (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). The inter-observer variability rate was 6% and discrepant cases were resolved by consensus review. P16 expression was not associated with disease control (Table [6](#T6){ref-type="table"}), TTP (HR 0.50 (95% CI 0.19-1.32), p = 0.16) or OS (HR 0.61 (95%CI 0.24-1.55), p = 0.30). The discordance between p16 IHC and HPV DNA by PCR was 25% (Table [7](#T7){ref-type="table"}). To investigate this further, we performed HPV DNA by ISH. The discordance between p16 IHC and HPV DNA by ISH was lower at 16% (Table [7](#T7){ref-type="table"}) and all 7 discordant cases were p16-positive/HPV-ISH-negative. Of these 7 discordant cases, 2 cases were HPV-16 positive by PCR, 4 cases were HPV negative by PCR and 1 case was inconclusive by HPV PCR. ::: {#T7 .table-wrap} Table 7 ::: {.caption} ###### Concordance of p16 IHC status with HPV by PCR and HPV by ISH ::: P16 IHC positive P16 IHC negative ---------------------- ------------------ ------------------ -------------- **HPV PCR positive** 10 5 P = 0.0016 **HPV PCR negative** 5 25 (Chi-Square) **HPV ISH positive** 6 0 P = 0.0002 **HPV ISH negative** 7 31 (Fisher\'s) ::: ### C-MET Forty-nine patients had sufficient tumor samples for evaluation of c-MET. Eighteen patients (63%) had low c-MET scores of 0, 1 or 2 and 31 patients had high c-MET (\>2) (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). Less than 10% inter-observer variability was observed and discrepant cases were resolved by consensus review. High c-MET was not associated with disease control (Table [6](#T6){ref-type="table"}), TTP (HR 1.47 (95%CI 0.56-3.85), p = 0.43) or OS (HR 1.72 (95%CI 0.65-4.56), p = 0.27). ### EGFR Gene Copy Number Forty-five patients had sufficient tumor samples for evaluation of EGFR GCN by FISH. High EGFR GCN (amplification and high polysomy) was detected in 13 patients and low EGFR GCN (disomy, low polysomy) was detected in 33 patients (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). High EGFR GCN was not predictive for TTP (HR 0.99, p = 0.822) or OS (HR 1.10, p = 0.644). High EGFR GCN was not associated with the presence of EGFRvIII (p = 0.14 Fisher\'s exact test). Discussion ========== To the best of our knowledge, this is the first study to evaluate the role of EGFRvIII in a cohort of patients with R/M SCCHN treated with or without EGFR TKI. This study confirms that EGFRvIII mutation is common in R/M SCCHN, and may play a role in prognosis. We identified EGFRvIII mutation in 42% of 53 R/M SCCHN tumors. This is in keeping with the first description of EGFRvIII expression by IHC and RT-PCR in 42% of 33 SCCHN tumors sampled \[[@B15]\]. In vitro studies suggest that EGFRvIII mutated SCCHN cell lines are resistant to the anti-EGFR monoclonal antibody cetuximab \[[@B15]\]. In this study, EGFRvIII was not associated with an inferior response to erlotinib therapy. Importantly, we observed a significant association between the presence of EGFRvIII (mean fold change and copy number by RT-PCR) with greater disease control, regardless of treatment with erlotinib, suggesting that perhaps EGFRvIII may have a prognostic role. The prognostic or predictive significance of the EGFRvIII mutation in response to systemic therapy in patients with SCCHN has not been previously described. The potential prognostic role of EGFRvIII appears to be independent of any clinicopathologic characteristics. This is consistent with another study where EGFRvIII detected by IHC in 234 of 681 locally advanced SCCHN tumors (34%) was associated with increased tumor size but not stage or other clinical factors \[[@B36]\]. In our study, EGFRvIII was not associated with overall survival or TTP. To our knowledge, EGFRvIII has not been linked to survival in SCCHN. EGFRvIII has been described more extensively in glioblastoma where it results in enhanced proliferation and reduced apoptosis effects that are mediated through increased levels of activated Ras \[[@B37]\] and activation of the PI3K pathway \[[@B38]\]. However, the role of EGFRvIII as a prognostic or predictive marker of response to EGFR inhibitors in glioblastoma remains controversial. EGFRvIII and PTEN co-expression was associated with response to EGFR TKI in 26 patients out of a cohort of 49 patients with recurrent glioma and a validation set of 33 patients \[[@B39]\]. EGFRvIII has been reported as a prognostic marker for poorer survival in some studies \[[@B40],[@B41]\], but not in others \[[@B42],[@B43]\]. Conflicting results have been attributed to small sample sizes with incomplete clinical data and varying methods to detect EGFRvIII. The presence of activating mutations conferring a better prognosis has been reported with EGFR mutations in non-small cell lung cancer (NSCLC) \[[@B44]\] and with PIK3CA mutations in breast cancer \[[@B45]\]. Somatic activating mutations (exon 19 deletion and 21 point mutation) in the EGFR tyrosine kinase domain confer sensitivity to EGFR inhibitors in NSCLC. Patients with these mutations also had improved survival and response to chemotherapy alone \[[@B46]\] or placebo \[[@B47]\]. This suggests that EGFR mutations in NSCLC are a good prognostic factor independent of EGFR TKI, hence it may be more difficult to demonstrate the value of EGFR mutations as predictors of benefit to EGFR TKI \[[@B44]\]. The prognostic value of EGFRvIII in SCCHN needs to be verified, and its role as a predictive marker of response to EGFR inhibitor should remain a relevant therapeutic question. In this study, the prevalence of HPV, p16 and c-MET expression (38%, 32% and 63% respectively) was in keeping with the literature. We did not observe HPV, p16 and c-MET expression to be predictive of disease control, TTP or OS. This may be due to limitations of a small sample size. Consistent with prior reports \[[@B21]\], HPV-16 was the most common HPV subtype in our study. c-MET is a poor prognostic marker in OSCC \[[@B48]\], however the small proportion (11%) of OSCC in our study precludes any meaningful association. Limitations of this study include its small sample size, potential bias towards patients with available tumor specimens (larger tumor size), potentially variable fixation and quality of the archival tissues and potential variation in marker status of primary tumor compared with recurrent or metastatic tumors (to our knowledge, this is theoretical and has not been described). Due to the absence of an untreated control group in this study (\'control\' patients received sorafenib or ispinesib), our results cannot conclusively confirm the prognostic versus predictive value of a biomarker. Although our methods did not use an antibody for EGFRvIII detection, we acknowledge that the use of RT-PCR in FFPE samples has demonstrated superior accuracy relative to IHC tests \[[@B49]\] and may allow greater applicability to settings where frozen tissue is unavailable. Conclusion ========== Predictors of response to EGFR inhibitors in SCCHN remain elusive. Biomarkers are desperately needed to guide patient selection in SCCHN. EGFRvIII remains an interesting tumor-specific target worthy of further exploration as a prognostic or predictive marker of response to EGFR inhibitor therapy in SCCHN. Larger prospective randomized studies are required to distinguish the prognostic and predictive significance of EGFRvIII, HPV, p16, c-MET and EGFR GCN in SCCHN treated with EGFR inhibitors. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= NC participated in the study design, data acquisition, immunohistochemical interpretation and drafted the manuscript. BP-O participated in the histological examination, immunohistochemical interpretation and manuscript preparation. KZ, NAP, JH, TZ, M-ST, SK-R carried out immunostaining, ISH, PCR analysis, and prepared and reviewed the manuscript. OL performed the FISH analysis and reviewed the manuscript. LW performed the statistical analysis. EC participated in patient management and reviewed the manuscript. LS conceived of the study, participated in its design and coordination, and prepared and revised the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ This work was supported by an unrestricted grant from Pfizer Canada, Inc.

PubMed Central

2024-06-05T04:04:19.055320

2011-2-27

PMC3052238

Background ========== Host-parasite coevolution is the result of multiple adaptations and counter-adaptations evolving in concert within the constraints of a particular system. Hosts use diverse defence mechanisms that coevolve with the offensive mechanisms of the parasite. From phages to ectoparasites, the success of infection depends on a series of steps and for each of them, the hosts may have specific defence mechanisms \[[@B1],[@B2]\]. The following steps may be distinguished, with greater or fewer steps potentially existing depending on the system and the level of resolution: The host encounter with the parasite is the first step. During this step, the host may exhibit particular behaviours in order to avoid the parasite \[[@B3]\] and there may be polymorphism for such behaviours within species \[[@B4]\]. Once the encounter has taken place, parasites with a dormant stage may need to be activated in order to terminate diapause and initiate the infection process - for example, by endospore germination \[see, for example, \[[@B5]\]\]. After the activation step, endoparasites need to enter the host tissues. For many parasites, including the one studied here, this occurs through the attachment of the parasites to the host tissues but hosts may evolve to prevent this attachment. For example, plants often have very specific mechanisms to prevent fungal pathogens from entering leaf tissue \[[@B6]\] and some species produce layers upon their epithelium - the first barrier against infection - to obstruct parasite penetration (for example, mucus in coral protection \[[@B7]\] or salivary mucins to preserve the oral cavity health \[[@B8]\]). After attachment and entering its host, the next step of infection is proliferation. To counteract parasite growth, the host adapts physiologically (for example, iron-withholding \[[@B9]\]) or actively defends itself with an immune response. In the final step of infection, the parasite releases transmission stages to infect other hosts. It has been argued that the fact that infection trials often intermingle the effects of different infection steps strongly influences our interpretation of host-parasite interactions \[[@B1],[@B10],[@B11]\]. For example, if only one of the steps is specific, the entire infection process will be specific. The same is true for environmental effects and host genotype-parasite genotype interactions. Furthermore, even if each of the steps is under simple genetic control (one or few loci) the combination of all of them might behave as a quantitative genetic trait and become more difficult to investigate. Resolving the infection process into its component steps simplifies the complexity of the infection process and helps us to better understand host-parasite interactions. Evolutionary models of host-parasite interactions are usually based on relatively simple assumptions about the underlying genetics and the impact of the environment. They commonly consider binary (Yes/No) infection outcomes (for example, matching-allele matrix \[[@B12]-[@B14]\]), even though available experimental data suggests more quantitative outcomes when looking at host and parasite interactions \[[@B15]-[@B17]\]. Explicit analysis of individual steps of infection can help bring in line theoretical models and data concerning the entire infection. As little is known about the degree of specificity of the individual steps, the specificity attributed to host-parasite interactions is usually the combined effect of all steps. Although it is reasonable to assume that different steps are under the control of different genes, and are influenced by the environment to different degrees, it is possible that a single component of the infection pathway may explain most of the observed variation in host-parasite interactions. This is particularly important because understanding variation in host susceptibility is central for controlling disease and understanding evolution. Here, we use the *Daphnia*-*Pasteuria*host-parasite system to investigate which step(s) best explain(s) the high degree of host genotype by parasite genotype interactions reported for this system \[[@B18]-[@B20]\]. We analyse the contribution of host and parasite genetics, host gender, host phylogeny and of the environment for the dynamics of host-parasite co-evolution. Reproduction in planctonic crustacean *Daphnia*is primarily clonal, which is most suitable for dissociations of genetic and environmental effects of its interactions with parasites. *Daphnia*are frequently found to suffer from bacterial, fungal and microsporidial infections \[[@B21],[@B22]\]; among them the Gram-positive bacterium *Pasteuria ramosa*\[[@B21]-[@B23]\]*. P. ramosa*produces endospores for transmission (Figure [1a](#F1){ref-type="fig"} and [1b](#F1){ref-type="fig"}\[[@B21]\]) that can remain dormant for decades \[[@B24]\]. Transmission is waterborne and endospores do not have flagella. The infection process is unknown but penetration of the host cuticula has been observed for the congeneric species *P. penetrans*, a parasite of root-knot nematodes \[[@B25]\]. Inside the host, *P. ramosa*proliferates in the haemocoel and musculature, castrates females and is transmitted horizontally after the release of endospores from the dead host \[[@B26],[@B27]\]. The interaction of *D. magna*clones and *P. ramosa*clones has been shown to be specific \[[@B20]\]. *Pasteuria*was shown to impose strong selection on its host \[[@B28]\] and there is evidence for coevolution \[[@B29]\]. Furthermore, strong effects of the environment and genotype-environment interactions were reported for the overall infection process \[[@B30],[@B31]\]. The goal of this study is to disentangle the different steps of the infection process and to analyse how they are shaped by host and parasite genetics, and the environment. We aim at finding the step that explains the greatest variance for the strong host-parasite interactions reported for the overall infection process. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Scanning electron microscopic (SEM) and transmission electron microscopic (TEM) images of the activation and the attachment step of the infection process of *Pasteuria ramosa*in *Daphnia magna***. (a) SEM image of a resting stage of *P. ramosa*. (b) TEM image of resting stage before activation. The exosporium (ex) encloses the two peripheral fibres (pf) and the endospore (en). (c) SEM image of activated spores trapped by *Daphnia*phyllopods. (d) TEM image of activated spores in *Daphnia*oesophagus. Top left: the spore is in the process of activating and shedding the exosporium. Bottom right: the activated spore with its sombrero-like structure in cross-section. Spore coat (sc) surrounding the cortex (cx). (e) TEM image of peripheral fibres (pf) and its microfibres on the upper side (upf) and on the lower side (lpf). The upper side is more furnished in microfibres and is more likely to play a role in the attachment. (f) TEM image of *Pasteuria*attached to the *Daphnia*oesophagus wall (ew). The nomenclature were defined according to the nomenclature of *Pasteuria penetrans*in \[[@B58]\]. :::  ::: We consider the following steps of the infection process and will investigate in details the second and the third, previously undescribed: (i) encounter; (ii) activation (once in contact with *Daphnia*, parasite endospores need a signal to germinate); (iii) attachment (the parasite must attach to the host and cross the host epithelium); (iv) proliferation (parasite proliferation and spore production); and (v) termination (killing the host to release spores). Environmental and host clone effects have been shown for the encounter and the proliferation steps \[[@B4],[@B30],[@B32]-[@B35]\]. However, neither of them can explain the strong host genotype by parasite genotype interactions described for the overall infection process in this system. Here, we localize where the activation and attachment steps take place and test for the genetic and environmental factors that influence those steps. Results ======= Spore activation ---------------- We developed a new method that traces fluorescently-labelled spores of *P. ramosa*in the transparent *Daphnia magna*hosts in order to investigate the activation of parasite spores and the attachment of the parasite to the host. Within minutes of exposing the *Daphnia*host to the *P. ramosa*spores, we observed a characteristic change in spore morphology. Spores acquire a sombrero-like structure (Figure [1c](#F1){ref-type="fig"} and [1d](#F1){ref-type="fig"}) which corresponds to the shedding of the exosporium and the extension of the peripheral fibres. This morphology was never observed in spores not exposed to hosts. We call this morphological change in spore shape \'activation\'. Activation was found to happen regardless of the host clone or *Pasteuria*clone used and was observed in both resistant and susceptible *D. magna*clones (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Results of infection trials, spore activation tests and attachment-tests. ::: ----------------------------------------------------------------------------------------------------------------------------------- Infection trial Spore activation Attachment-test\ (attached out of five) ------------ ----------- -- ----------------- ------------------ ------------------------ ----- ----- ----- ------- ------- ------- HO1 Hungary R R R Yes Yes Yes 0 0 0 HO2 Hungary **S** **S** **S** Yes Yes Yes **5** **5** **5** HO3 Hungary R R R Yes Yes Yes 0 0 0 M5 Belgium R R R Yes Yes Yes 0 0 0 M10 Belgium **S** **S** **S** Yes Yes Yes **5** **5** **5** Iinb1\* Germany\* R R R Yes Yes Yes 0 0 0 Mu12 Germany R R R Yes Yes Yes 0 0 0 DG-1-106 Germany **S** R R Yes Yes Yes **5** 0 0 AL144 Finland R **S** **S** Yes Yes Yes 0 **5** **5** Xinb3\* Finland\* **S** R R Yes Yes Yes **5** 0 0 XI\* Finland\* R R R Yes Yes Yes 0 0 0 Xfa6\* Finland\* **S** R R Yes Yes Yes **5** 0 0 Kela-39-09 Finland R **S** **S** Yes Yes Yes 0 **5** **5** Kela-18-10 Finland **S** R R Yes Yes Yes **5** 0 0 ----------------------------------------------------------------------------------------------------------------------------------- We tested all combinations of three *Pasteuria ramosa*clones (C19, C14, C1) with 14 *Daphnia magna*clones. Infection trials results are defined by exposing *Daphnia*to the parasites and determining the infection status after 20 days. Resistant means that none of the replicates were infected. Activation was determined by observing spores in the gut of the host with a sombrero-like shape. *R*means that the host clone is totally resistant to the concerned parasite clone. *S*means that the host clone is susceptible to the concerned parasite clone. *Yes*means that the spores were activated. Labcross: \'Iinb1\' is \'Mu11\' (Belgium) selfed once; \'Xinb3\' is \'X\' (Finland) selfed three times; \'Xfa6\' is \'AL144\' selfed three times and crossed with \'Xinb3\'; \'XI\' is a cross between \'Iinb1\' and \'Xinb3\'. ::: Spore attachment ---------------- We used different combinations of hosts and parasite clones previously characterized to be resistant or susceptible to given *Pasteuria*clones \[[@B20]\]. We observed the fate of fluorescent spores of three parasite clones exposed to 14 *D. magna*host clones with the aim of identifying differences which correlate with the compatibility of a given host-parasite combination (Table [1](#T1){ref-type="table"}). The parasites attach to the host oesophagus for all susceptible (compatible) host-parasite combinations, but they never do so for the resistant combinations (Table [1](#T1){ref-type="table"}, Figures [1f](#F1){ref-type="fig"} and [2a](#F2){ref-type="fig"}). Thus, the result of this attachment-test was 100% consistent with the results of infection trials (Table [1](#T1){ref-type="table"}). For susceptible combinations the host oesophagus was densely covered with spores forming a dense layer in the oesophagus, while there were no spores attached in resistant combinations. We never observed ambiguous cases (for example, only few spores attached). While spores in the mid and end gut moved with the flow of the food, those attached to the oesophagus were not disturbed by passing boluses of food, indicating that they strongly adhere. In resistant combinations spores were never seen attached to the oesophagus and all spores passed with the flow of the food through the gut (Figure [2b](#F2){ref-type="fig"}). Thus, spore attachment in the oesophagus was very specific to the *D. magna*and *P. ramosa*genotype and consistent with resistant/susceptibility status for each combination. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Fluorescently labelled parasite spores attach to the oesophagus of susceptible, but not resistant, *Daphnia*clones**. (a) Picture of a susceptible *Daphnia magna*exposed to fluorescently labelled spores. The entire animal is shown. Parasites are attached on the epithelium of the oesophagus (arrow). Other labelled spores can be seen with the rest of the food in the end gut (arrowhead). (b) Picture of a resistant *D. magna*exposed to fluorescently labelled spores. The entire animal is shown. The oesophagus is free of parasite (arrow). Labelled spores can be seen with the rest of the food in the end gut (arrowhead). Note the autofluorescence of the mandibule. Extended focus images obtained by the camera Leica DFC 300FX and the program Leica Application Suite (Version 3.4.0, package \'Montage\'). Intensity, contrast and sharpness were increased with the same strength. :::  ::: Influence of gender and culture conditions ------------------------------------------ Activation of spores was observed in all treatments and in all host clone-*Pasteuria*clone combinations (Table [2](#T2){ref-type="table"}). In contrast, the specificity revealed by the attachment-test was found to be independent of host gender, temperature and culture conditions (single versus crowded; high versus low food, Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Influence of the environment and host gender on spore attachment, as determined by the attachment-test. ::: Treatments ------------- ----------- ------------ ----------- --------- --------- ----------- ----------- ----------- ----------- ----------- --------- ----------- ----------- *Pasteuria* *Daphnia* Kela Kela Kela Kela Kela Kela Kela Kela Kela Kela Kela Kela clone clone 39-09 18-10 39-09 18-10 39-09 18-10 39-09 18-10 39-09 18-10 39-09 18-10 C1 Female **6/6** 0/5 **9/9** 0/9 **10/10** 0/10 **10/10** 0/10 **10/10** 0/10 **10/10** 0/10 Male **10/10** 0/10 **9/9** 0/9 **9/9** 0/10 **10/10** 0/10 **9/9** 0/10 **10/10** 0/10 C19 Female 0/10 **7/7** 0/8 **8/8** 0/9 **10/10** 0/10 **10/10** 0/9 **9/9** 0/10 **10/10** Male 0/5 **10/10** 0/10 **9/9** 0/10 **10/10** 0/10 **9/9** 0/9 **9/9** 0/10 **10/10** Infection trials (see Table 1) showed that *Daphnia magna*clone Kela-39-09 is susceptible to *Pasteuria ramosa*clone C1 but resistant to C19. Kela-18-10 is resistant to C1, but susceptible to C19. LF = low food condition; HF = high food condition; single = *Daphnia*raised single in a 100 mL jar; crowded = *Daphnia*randomly picked from crowded cultures (high density). The bold characters highlights results where *P. ramosa*were attached to the *D. magna*oesophagus ::: Spore activation and resistance of other *Daphnia*species --------------------------------------------------------- Spores were found to be activated after exposure to all *Daphnia*species tested (Table [3](#T3){ref-type="table"}). We found that spores of the *P. ramosa*clone C19 were able to attach to the oesophagus and infect *D. dolichocephala*(Table [3](#T3){ref-type="table"}) but did not attach to the oesophagus or infect *D. arenata*, *D. galeata*, *D. barbata*, *D. similis*or *D. lumholtzi*. We also tested other species for spore activation of *P. ramosa*. Upon exposure to *Simocephalus vetulus*(Daphniidae) spores were readily activated but did not attach to the oesophagus or infect any of the individuals tested. Upon exposure to mosquito larvae (*Culex*spp.), which are also filter-feeding but are not crustaceans, *P. ramosa*spores were neither activated nor attached to the host. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Relationship between one *Daphnia magna*-derived clone of *Pasteuria ramosa*(clone C19) and several *Daphnia*species belonging to three different subgenera (*D. magna*belongs to the subgenus *Ctenodaphnia*). ::: Clones of *Daphnia*species Sub-genus Origin Infection trial Spore activation Attachment-test (attached out of 5) ---------------------------- ---------------- -------------- ----------------- ------------------ ------------------------------------- *D. arenata* *Daphnia* USA R Yes 0 *D. galeata* *Hyalodaphnia* Germany R Yes 0 *D. barbata* *Ctenodaphnia* Zimbabwe R Yes 0 *D. similis* *Ctenodaphnia* Israel R Yes 0 *D. lumholtzi* *Ctenodaphnia* Zimbabwe R Yes 0 *D. dolichocephala* *Ctenodaphnia* South Africa S Yes 4 Legend as in Table 1 ::: Discussion ========== The aim of the current study was to analyse two steps in the life cycle of a bacterial parasite, characterize the specificity of the interaction, with regard to genetic and environmental factors, and relate these findings to what is known about host-parasite coevolution in this system. We focused on the activation of the parasite\'s resting stages and on the attachment of the activated spores to the host tissue where it enters the host. Our study revealed that *P. ramosa*spores captured by the filter feeding *Daphnia*are indiscriminately activated by every *Daphnia*clone and *Daphnia*species tested (Tables [1](#T1){ref-type="table"} and [3](#T3){ref-type="table"}). Furthermore, activation was not only found to be independent of the host genotype or species and host gender but also of the environmental conditions (namely, density, temperature and food conditions). The following step of the infection process, however, the attachment of the activated spore to the oesophagus wall of the host, depended strongly on the combination of the *D. magna*and parasite genotype, but not on the host\'s gender, nor the environmental conditions in which they were kept (Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"}). Previous studies with the *Daphnia-Pasteuria*system were not able to disentangle the activation, attachment and proliferation steps. Thus, variation in infection success as reported in earlier studies \[[@B19],[@B32],[@B35]-[@B40]\] may be explained by the combined effects of these steps. However, the binary polymorphism found in infection trials with high doses of single parasite clone \[[@B20]\] correlates perfectly with the results of our attachment-test (Table [1](#T1){ref-type="table"}). This suggests that only *Pasteuria*clones able to attach to the oesophagus are able to infect the host. Ben-Ami *et al*. \[[@B38]\] proposed that *D. magna*might be either completely resistant or susceptible to *P. ramosa*depending on the genotype-genotype interaction. They called this the \'binary infection hypothesis\'. Our data are consistent with this hypothesis and further pinpoint which specific step of the infection process is responsible for the high degree of specificity. For a given combination of host and parasite genotypes, the activated spores are either able to attach and then infect or they do not attach and do not infect. We did not see any evidence for a graded (quantitative) form of interaction. Spore attachment is a key step in *Daphnia-Pasteuria*coevolution ---------------------------------------------------------------- The *Daphnia-Pasteuria*system has become one of the prime examples of antagonistic coevolution. Host and parasites show strong genetic effects for resistance, virulence and infectivity; genotype × genotype interactions have been reported within and across populations and selection acts rapidly in natural populations \[[@B18],[@B19],[@B40],[@B41]\]. Our study suggests that the parasite-dependent \[[@B28]\] host population structure and the coevolution \[[@B29]\] described for this system are mainly driven by the properties of a unique step, the attachment step. First, this step revealed very strong host genotype by parasite genotype interactions (Table [1](#T1){ref-type="table"}). Second, the attachment step is independent of the environmental conditions. Third, a recent study of *D. magna - P. ramosa*coevolution using resurrected host and parasite isolates from lake sediments showed a signal of fluctuating selection only for infectivity but not for parasite virulence \[[@B29]\]. Virulence (the parasite\'s effect on infected hosts) was also observed to evolve but at a slower rate \[[@B29]\]. The authors proposed that the difference between the evolution of virulence and infectivity resulted from different genes contributing to these traits. Here we give a mechanistic explanation for this finding. Infectivity depends on the attachment and, most likely, on the ligands present on the host and on the parasite. On the other hand, expression of virulence may depend on the host\'s immune response during the within-host proliferation step. It is likely that these processes are determined by different sets of genes. The identification of the attachment step as the key step in the coevolutionary dynamics in this system will allow us to improve our understanding of the patterns of antagonistic coevolution. For example, evolutionary models studying the coevolution of the infectivity and the virulence steps \[[@B42]\] can fit our system in relation to the coevolution of the attachment and the proliferation steps. Those models typically characterize infection outcomes as binary (Yes/No), while empirical data suggest they are more quantitative \[[@B15]-[@B17]\]. We show that we can observe a binary outcome when individual steps of the infection process are considered. Furthermore, our method provides a fast and reliable way to test individuals and populations for their susceptibility to *Pasteuria*. Ongoing research in our group showed that up to 400 *Daphnia*individuals can be tested in a day (P Luijckx, in preparation). The assay we developed makes it possible to test for susceptibility without the potentially confounding effect of the within-host proliferation step in the infection trials. From the environment to the host body cavity -------------------------------------------- The resting endospores of *P. ramosa*can remain dormant for decades under harsh environmental conditions \[[@B24],[@B29]\]. Before attachment to the host, the spores need to be activated (Figure [1d](#F1){ref-type="fig"}). The filter-feeding *Daphnia*capture particles, including parasites, from the water and transport them on a mucus-layered pathway from the phyllopods to the mouth. During this process, the parasite\'s exosporium opens by an unknown trigger, releasing the activated spore form within less than 10 min (Figure [1](#F1){ref-type="fig"}). Despite the fact that spore activation is a necessary step for the infection, this step is entirely unspecific with regard to *Daphnia*species and clone, host gender and the environmental conditions (Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"}). The signal that triggers spore activation may be related to chemical substances in the mucus of the filtering apparatus, but other factors (for example, mechanical) cannot be excluded. Once the activated spore enters the oesophagus, it will attach to the oesophagus wall if the host and the parasite genotype are compatible. There it presumably penetrates the gut wall and enters the host\'s body cavity. A similar attachment process on the cuticula is also known from *P. penetrans*but, in this case, the parasite seems to be able to attach to any area of the nematode\'s body surface \[[@B25]\]. It has been proposed that the lower part of *P. nishizawae*attaches to the host because this part is densely covered by microfibres \[[@B43]\]. In contrast, in *P. ramosa*it is the upper part of the peripheral fibres (Figure [1e](#F1){ref-type="fig"}) that are most densely covered with a layer of microfibres. These fibres may be involved in the attachment (Figure [1f](#F1){ref-type="fig"}). An endospore adhesin epitope, situated on the exosporium of *P. ramosa*, has been identified and it has been suggested that it may be a ligand responsible for the recognition and the binding onto the host \[[@B44]\]. However, according to our results, it is unlikely that this epitope is involved in the attachment because the exosporium of *P. ramosa*is removed during the activation step. A later study, analysing surface proteins of *P. ramosa*spores by two-dimensional gel electrophoresis, proposed that a collagen-like protein may be responsible for the binding onto the host but might suffer the same problem of the previous study \[[@B45]\]. We propose that later studies on candidate proteins responsible for the specific attachment to the host in this system should investigate the spores once activated. The development of *Pasteuria*, from the moment they attach to the oesophagus until the vegetative stage can be detected in the hemolymph (about 8 days at 20°C \[[@B46]\]), is unknown. Also, the penetration mechanism is poorly described. Sayre and Wergin \[[@B25]\] show a transmission electron micrograph of *P. penetrans*with a structure they call a \'germ tube\' crossing the host epithelium. Our hypothesis is that the endospore makes a hole across the host epithelium and injects its cortex into the host. As one response of *Daphnia*to wounding is an increase of Phenoloxidase (PO) activity \[[@B47]\], one might expect the penetration process to trigger an immune response but this remains an open question. However, resolving the infection process will allow the study of the immune response during the proliferation step without the confounding effect of genetic variation in the attachment step. Environment effects and the proliferation step ---------------------------------------------- We found that environmental effects do not influence the activation and attachment step (Table [2](#T2){ref-type="table"}). Excluding these steps, we suggest that the proliferation step is the one responsible for the reported sensitivity of the overall infection process for environment effects \[[@B32],[@B34]\]. The activation and the attachment steps seem independent of the host\'s immune system (defined as a system that is potentially able to kill parasites), while the proliferation step is likely to be governed by the host\'s immune system. The immune system may lead to variation between and within those *Daphnia*clones that allow *Pasteuria*attachment (and, thus, enable the parasite to enter the host), thereby contributing to local and temporal adaptation, maternal effects and induced resistance \[[@B29],[@B34],[@B48]\]. We suggest that future studies on host immunity should use only *Pasteuria*clones that can attach to a given clone of *Daphnia*so that all the variation observed is likely to originate from variation during the proliferation step. These factors highlight the importance of controlling the host and parasite genotypes and breaking down the infection process in order to understand the respective role of each step in host-parasite interactions. Resolving the infection process leads to a better understanding of host-parasite interactions --------------------------------------------------------------------------------------------- Resolving the infection process in its sequential steps has been proposed in a number of theoretical models \[[@B10],[@B11]\] but experimental data are scarce. Our approach is transferable to other host-parasite systems and our results suggest that this can provide important new insights into host-parasite interactions and their evolution. Increasing the degree of the resolution of the infection processes highlights a universe of possibilities of the different levels at which host and parasites interact. The different steps might differ in how they are influenced by the environment. They might also differ in which sets of genes regulate them. As it is probably the case for our study system, different steps of the infection process might follow distinct evolutionary dynamics and be explained by different model (for example, balancing selection, directional selection) \[[@B10],[@B11]\]. However, because of the sequentiality of the steps, it is possible that the selection on one might depend on the selection on other steps. We propose that analysing infection as a succession of well characterized steps will help to reconcile the empirical data with predictions based on alternative coevolutionary models (for example, Red Queen and Selective Sweep models). Spores of all *P. ramosa*clones tested, and which were isolated from natural *D. magna*populations, were activated by all *D. magna*clones as well as by six other *Daphnia*species (Table [3](#T3){ref-type="table"}) and even a Cladoceran from a different genus, *Simocephalus vetulus*. Also, apart from the natural host, *D. magna*, *D. dolichocephala*, also became infected following attachment of the activated spores to the host oesophagus. This suggests that the triggers for spore activation and, to a lesser extent, for attachment are phylogenetically conserved. This may facilitate the host range evolution of the parasite. Indeed, despite its high specificity on the level of the host clone, *P. ramosa*infections have been reported in several species within the family Daphniidae \[[@B49]\]. It will be necessary to test more clones of different *Daphnia*species in order to determine their pattern of susceptibility and resistance to the parasite. Importantly, phylogenetically conserved steps of the infection process can be ruled out as major factors in coevolution, but are, perhaps, the most appropriate targets for vaccine and drug development. In fact, the genes involved in some infections steps have been worked out for some systems \[[@B50],[@B51]\] and can be of use in biomedicine for diseases control \[[@B52],[@B53]\]. Conclusion ========== Our study highlights the explanatory power of resolving the steps of the infection process in order to better understand host-parasite interactions and coevolution. Attachment appears to be the crucial step for the previously observed high specificity in the *Daphnia-Pasteuria*system and we speculate that it is the crucial step for coevolution as observed in this system \[[@B29]\]. Our results reveal that each step can involve different interactions between host, parasite and environment and that certain steps can be phylogenetically conserved. With this knowledge, it will be easier to apply simple models of host-parasite interactions to this system and identify the mechanistic basis of trade-offs, maternal effects, genotype × environment interactions and coevolution. The logic of this procedure can equally be applied to other host-parasite systems but also to study other types of biotic interactions. Methods ======= Host and parasite ----------------- We used 14 isofemale lines (hereafter referred as clones) of *D. magna*and one clone each of six other *Daphnia*species (Tables [1](#T1){ref-type="table"} and [3](#T3){ref-type="table"}). Unless otherwise stated, *Daphnia*clones were kept in standard medium (ADaM, \[[@B54]\] modified by using only 5% of the recommended Selenium dioxide concentration) at 20°C and fed with the chemostat cultured unicellular algae, *Scenedesmus obliquus*. The parasites used were single genotypes of *P. ramosa*, C1, C14 and C19, characterized as clones in Luijckx *et al*. \[[@B20]\] and originated from *D. magna*populations in Moscow (Russia), Tvärminne (Finland) and Gaarzerfeld (Germany), respectively. Spore suspensions of *Pasteuria*were obtained by homogenizing infected *D. magna*in ADaM and quantifying spore density. The status of resistant or susceptible *D. magna*were defined previously \[[@B20]\]. The infection status of two further Finnish *D. magna*clones (\'Kela-39-09\' and \'Kela-18-10\') exposed to *Pasteuria*clones were tested with the same protocol. All infections in these experiments were done with naïve individuals born to naïve mothers, kept under high food conditions. These conditions were applied because they are known to minimize the triggering of immune effect \[[@B34],[@B35],[@B55]\]. Fluorescence labelling of spores -------------------------------- Fluorescently labelled spores of *P. ramosa*were produced by homogenizing infected *Daphnia*in ADaM, followed by centrifugation at 10 000 g for 5 min at room temperature. The spore pellet was suspended in 0.5 mL of 0.1 M sodium bicarbonate (pH 9.1) containing 2.0 mg/mL of fluorescein-5(6)-isothiocyanate (F3651-100MG, Sigma-Aldrich, Miss, USA), a green fluorescent dye that stains proteins unspecifically \[[@B56]\]. Spores were incubated in the dark for 2 h at room temperature with occasional vortexing. The suspension was centrifuged at 10 000 g for 5 min and the supernatant removed. The spore pellet was suspended in distilled water and, again, subjected to centrifugation. This process was repeated until the supernatant was clear. Labelled spore suspensions can be stored at 4°C in the dark for several months. The shape and location of the green labelled spores were examined in the transparent *Daphnia*using a microscope with fluorescent light (Leica DM 2500, at magnification 200 × and 400 ×) and filter cubes Leica B/G/R (bandpass filter excitation 420/30 nm; 495/15 nm; 570/20 nm - band pass filter suppression 465/20 nm; 530/30 nm; 640/40 nm). We increased the colour contrast by adding a red fluorescent dye to the medium in which the *Daphnia*were observed. This was done by preparing a solution of concentrated red dye (0.05 mL of DMSO with 0.0015 g of Tetramethylrhodamin-5-isothiocyanate; T0820-5MG by Sigma-Aldrich), which was homogenized in PBS to make the diluted dye (1 μL of concentrated solution with 10 mL of phosphate buffered saline). We added 1 μL of this solution to the *Daphnia*medium 10 min before observing the *Daphnia*. We obtained extended focus images using a camera Leica DFC 300FX and the program Leica Application Suite (Version 3.4.0, package \'Montage\'). The separation of the different steps and their specificity ----------------------------------------------------------- Adult *Daphnia*were put individually in 1 mL of medium in 24-well-plates and exposed for at least 1 h to around 17,000 labelled *P. ramosa*spores. Susceptible hosts exposed to labelled spores become infected, suggesting that the dye does little or no harm to the spores (data not shown). Spore activation ---------------- Pilot trials revealed that the labelled spores remain in their typical spherical shape as long as they are not in contact with a host. Upon contact with the host phyllopods (swimming and respiratory appendages of branchiopod crustaceans), spores with a sombrero-like shape are observed (Figure [1c](#F1){ref-type="fig"} and [1d](#F1){ref-type="fig"}). We called this process \'spore activation\'. We tested all combinations of 14 *D. magna*clones and three *P. ramosa*clones for spore activation (Table [1](#T1){ref-type="table"}). The same was done for one clone each of six further *Daphnia*species but only in combination with one *P. ramosa*clone (Table [3](#T3){ref-type="table"}). We used five replicates for each host-parasite combination \[in total (14 × 3 × 5) + (6 × 1 × 5) = 190; details in Table [1](#T1){ref-type="table"}\]. Spore attachment ---------------- After exposure, *Daphnia*were placed on a microscopic slide and we examined the complete *Daphnia*body under a fluorescent microscope. The transparent body of *Daphnia*allowed us to determine in which body region activated fluorescent spores attach in the living animals. Once we determined the specific area, we tested resistant and susceptible *Daphnia magna*clones (five replicates of 14 clones, details in Table [1](#T1){ref-type="table"}) for differences in attachment. The same was done with clones of other *Daphnia*species (five replicates of one clone per species; details in Table [3](#T3){ref-type="table"}). In order to validate the assignment of individuals with apparently no spores attached to their oesophagus, we viewed the oesophagus of slightly squashed animals at 400 × magnification. For each experiment, the examiner was not informed whether the animals belonged to a susceptible or to a resistant clone. In order to confirm that the *Daphnia*ingested spores, the gut content was inspected for the presence of spores. All exposed animals had spores in the faeces. We call this procedure to test for spore attachment the \'attachment-test\'. Influence of gender and culture conditions ------------------------------------------ In order to see if the specificity pattern observed in the attachment-test was dependent on host sex or culture conditions, 10 host individuals of each sex were tested in each of six treatments. This was done with *D. magna*clones \'Kela-39-09\' and \'Kela-18-10\"\'because these two *Daphnia*clones have the reverse pattern of infectivity to the two *P. ramosa*clones used and they are easily induced to produce male and female offspring in the laboratory. *Daphnia*were raised either at one of four temperatures (10°C, 15°C, 20°C, 25°C, with high food), two food levels (at 20°C, fed daily with 2.5 or 5 million algae) or two density levels (at 20°C, high food level, single *Daphnia*or *Daphnia*from crowded stock cultures; see Table [2](#T2){ref-type="table"}). These conditions were chosen to represent various environments that are common in natural *Daphnia*populations. We did not employ a full factorial design, as our interest was not in establishing reaction norms but in testing for the influence of non-genetic conditions in general. *Daphnia*of both clones raised under these conditions were exposed to fluorescently labelled spores of *P. ramosa*C1 and C19. Given the very clear effects observed with the four combinations of hosts and parasites used and the range of conditions tested, we do not believe that other combinations would drastically change our results. However, we cannot exclude with certainty that some combinations might lead to a different result. Resistance or susceptibility of other *Daphnia*species ------------------------------------------------------ One clone of each of six other *Daphnia*species (*D. arenata*, *D. dolichocephala*, *D. galeata*, *D. barbata*, *D. similis*and *D. lumholtzi*) were assayed for their propensity of oesophageal spore attachment using *P. ramosa*clone C19. For this assay, groups of five conspecific individuals were exposed to 200,000 *P. ramosa*spores of clone C19 in 20 mL medium. Four replicates per species were used. After 5 days we filled the jars to 100 mL and then changed the medium on a weekly basis. Animals were fed daily with 5 to 10 million algal cells per jar depending on the size of the *Daphnia*species. The infection status was investigated under a microscope with phase contrast (magnification 400 ×), at host death or 29 days after exposure. Electron microscopy ------------------- In order to prepare *Daphnia*for transmission electron microscopy (TEM), infected individuals were fixed on ice in 4% glutaraldehyde buffer in Sorensen\'s phosphate buffer (0.1 M KH~2~PO~4~and 0.1 M Na~2~HPO~4~) and kept in the dark for several hours. The animals were then rinsed five times on ice using the same buffer for a total of 5 min. Post-fixation was carried out with 1% OsO~4~in Sorensen\'s phosphate buffer on ice. After post-fixation, the *Daphnia*were again washed in Sorensen\'s phosphate buffer on ice, dehydrated in a graded acetone series, and finally embedded in the epoxydic resin EPON. Transversal and sagittal sections were made through the oesophagus. Semi-thin sections (diamond knife, 0.7-1 μm) were cut in order to approach the right spot on the resin block using a RMC MT 6000-XL (RMC Inc, AZ, USA) ultramicrotome. In order to identify regions of interest for TEM, the tissue was stained using Richardson\'s dye \[[@B57]\] and examined under a light microscope. In order to see parasite structures using TEM, 5-8 ultrathin sections (diamond knife, 60 nm) were cut after every 10 semi-thin sections. The ultra-thin sections were mounted on Formvar-coated copper grids and stained with uranyl acetate and lead citrate to enhance the contrast. Ultrathin sections were analysed using a FEI Morgagni™ transmission electron microscope at 80 kV equipped with a digital camera. For scanning electron microscopy (SEM), *D. magna*were fixed in 3% glutaraldehyde buffer in 0.1 M phosphate buffer for 2 h at 20°C. Samples were washed twice in distilled water for 5 - 10 s, dehydrated in graded ethanol series and critical point dried overnight (16 h). The specimens were coated with gold (20 nm) and viewed using a Philips XL 30 ESEM under high volume conditions from 5 to 15 kv. Authors\' contributions ======================= DD conceived and designed the study, performed the experiment, performed data analysis and drafted the manuscript. PL participated in the design of the study. FB performed the SEM. CL performed the TEM. DE conceived of the study, participated in its design and participated in drafting the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We thank M. Ackermann and L. Du Pasquier for their thoughtful discussions; N. Boileau, J. Hottinger, M. Kredler and U. Stiefel for their laboratory assistance; and J. Andras, P. Beldade, S. Zweizig and the three anonymous reviewers for their comments on the manuscript.

PubMed Central

2024-06-05T04:04:19.060017

2011-2-22

PMC3052239

Background ========== *EVC*was first identified through a positional cloning approach as the gene defective in patients with the recessive disorder Ellis-van Creveld syndrome (EvC) \[[@B1]\]. The consistent features of this condition are short ribs, short limbs, postaxial polydactyly, dental abnormalities and nail dysplasia. Failure to find *EVC*mutations in a number of consanguineous EvC families mapping to the same region of chromosome 4 led to investigation of other genes in the EvC critical interval and to the identification of mutations in a second gene, *EVC2*, which is in close proximity to *EVC*\[[@B2]\]. The two genes are in divergent orientation with their translational start sites separated by only 2.86 kb in the human genome \[[@B3]\] and 1.75 kb in the mouse \[[@B4]\]. The phenotype associated with mutations in either of the two genes is indistinguishable indicating that *EVC*and *EVC2*act in a common pathway \[[@B5],[@B6]\]. Mice lacking Evc, like patients with Ellis-van Creveld syndrome, have short limbs, short ribs and dental abnormalities. On histological analysis of the growth plate, they have epiphyseal shortening and defective periosteal induction compatible with a defect in Indian Hedgehog (*Ihh*) signalling. Studying expression of *Ihh*and its downstream targets by *in situ*hybridization demonstrated normal *Ihh*expression but diminished mRNA levels of the Ihh downstream targets, *Patched1*(*Ptch1)*, *Gli1 and Pthrp*. *In vitro*studies treating mouse embryonic fibroblasts (MEFs) and chondrocytes with the Hedgehog (Hh) agonist purmorphamine confirmed that Hh signal transduction is defective in cells lacking Evc \[[@B7]\]. Thus Evc is essential for Ihh signalling in the cartilage growth plate. We have previously shown that Evc localises to the base of primary cilia \[[@B7]\]. Seminal studies demonstrated that primary cilia are required for Sonic hedgehog (Shh) signalling \[[@B8],[@B9]\] and subsequent studies have confirmed that this is also the case for Ihh signalling \[[@B10],[@B11]\]. Proteins are transported from the base to the tip of the cilium by anterograde intraflagellar transport (IFT) and back to the base by retrograde IFT, outward transport being mediated by a kinesin motor and retrograde transport by a dynein motor. Key components of Hh signalling such as Ptch1 and Smoothened (Smo) have been shown to enter and leave the cilium depending on the activation status of the pathway \[[@B12]\]. Hh signalling is ultimately meditated by the Gli transcription factor family of proteins which are observed in the cilia as well as in the nucleus \[[@B13]\]. In the absence of Hh signal, Gli3 is processed into a repressor form Gli3R, transcription of Gli3 targets being dependent on the balance between activator full-length Gli3 and Gli3R. Gli3 processing is reduced both when anterograde IFT is disrupted and when retrograde IFT is disrupted \[[@B8]-[@B10],[@B14]\]. In *Evc*^*-/-*^mice, in contrast to IFT mutants, Gli3 processing appears normal on protein extracts from E14.5 limbs though as with the IFT mutants, the expression of gene targets of Hh signalling such as *Ptch1*and *Gli1*are diminished \[[@B7]\]. In this study we demonstrate that, in addition to Evc, Evc2 is also required for Hh signal transduction. We have shown direct interaction between Evc and Evc2, have investigated Evc and Evc2 subcellular localisation and discuss the significance of these findings. Results ======= Evc2 is a positive regulator of Hh signal transduction ------------------------------------------------------ Evc is known to be a positive regulator of Hh signalling both in the cartilage growth plate and in cultured chondrocytes and MEFs. The observation that *EVC*and *EVC2*mutations are associated with the same phenotype indicates that Evc2 is likely to be a positive regulator of Hh signalling. We have tested this hypothesis using siRNA to knockdown Evc2 expression in LIGHT2, Hh reporter cells \[[@B15]\]. These cells are a mouse fibroblast line that stably express a Gli-dependent firefly luciferase and a TK Renilla luciferase control to allow quantitation of Hh pathway activation. LIGHT2 cells transfected with Evc2 siRNA had reduced Evc2 protein levels and had a diminished response to the Smo agonist purmorphamine compared to controls transfected with a non-targeting siRNA (Figure [1a](#F1){ref-type="fig"}). This result was confirmed in osteoblast-derived MC3T3 cells co-transfected with the Gli-dependent firefly luciferase and TK Renilla luciferase plasmids (Figure [1b](#F1){ref-type="fig"}). We also assessed expression of the Hh target gene, *Ptch1*compared to expression of the *Hprt*housekeeping gene in Evc2 null MEFs by RT-PCR. *Ptch1*expression in response to purmorphamine in Evc2 mutant MEFs was reduced compared to wild-type MEF controls (Figure [1c](#F1){ref-type="fig"}). Since purmorphamine activates Smo, these data confirm that, like Evc, Evc2 is a positive regulator of the Hh signalling pathway. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Evc2 is a positive regulator of Hh signalling**. **(a and b)**. Representative reporter assays in LIGHT2 Gli-reporter cells (a) and MC3T3 cells (b) transfected with siRNA and treated with control DMSO (white bars) or Hh agonist purmorphamine (black bars). Purmorphamine increases expression of the Gli reporter (FF Luc) with respect to the Renilla control (Ren Luc). Evc2 knockdown was confirmed by western blotting (lower images). Evc2 (approximately 140 kDa) was detected using R1656 and α tubulin (50 kDa) was used as a loading control. Cells with reduced Evc2 protein levels (approximately 80% less than normal) have a significant reduction in Gli reporter expression in response to purmorphamine compared to those transfected with the non-targeting siRNA pool. **(c)**. The purmorphamine induced expression of the Hh responsive gene, *Ptch1*, is significantly diminished in Evc2 null MEFs (-/-) compared to wild-type controls (+/+) as assessed by RT-PCR. :::  ::: Evc and Evc2 are homologues resulting from an ancient gene duplication ---------------------------------------------------------------------- We investigated what the protein sequence of *EVC*and *EVC2*could tell us about their relationship to each other and to their function. Evc (a 1005 amino acid protein) has a predicted signal anchor sequence and Evc2 (a 1220 amino acid protein) has a predicted signal peptide sequence and a second predicted transmembrane region (amino acids 209 - 231), the only feature of the proteins C-terminal to these transmembrane sequences being coiled-coil regions. Sequence analysis predicts that they have resulted from an ancient duplication of a pre-existing gene whose occurrence predates the radiation of most metazoan lineages \[see additional file [1](#S1){ref-type="supplementary-material"}\]. *EVC2*was identified as showing significant sequence similarity to *EVC*over approximately 550 amino acids when the latter was used to query the non-redundant protein sequence database using PSI-BLAST (second iteration, *E*= 6 × 10^-14^). Next, *EVC*and *EVC2*orthologues were each found as gene predictions in the genome assemblies of amphioxus (*Branchiostoma floridae*), sea urchin (*Strongylocentrotus purpuratus*), gastropod snail (*Lottia gigantea*) and sea anemone (*Nematostella vectensis*). For example, a BLAST search of the nonredundant sequence database with sea anemone *EVC2*as query revealed significant similarity (*E*= 6 × 10^-25^) with human *EVC2*over 840 amino acids. Since these invertebrate species and mammals last shared a common ancestor about 670 to 820 million years ago \[[@B16]\], this indicates that the gene duplication giving rise to *EVC*and *EVC2*occurred in an early metazoan species. In amphioxus, sea urchin and snail, but not sea anemone, genome assemblies these two genes, as in vertebrates, lie in close proximity in a 5\' to 5\' head-to-head arrangement. This arrangement, and the close proximity of their transcription start sites, suggests that *EVC*and *EVC2*share a bidirectional promoter \[[@B17]\]. This, in turn, suggests that *EVC*and *EVC2*need to be co-ordinately expressed in order to maintain an appropriate stoichiometry, or because they function in the same biological pathway. An *EVC*-like gene is also predicted in the placozoan, *Trichoplax adhaerens*, the most basal metazoan known \[[@B18]\] thereby confirming the presence of *EVC*-like genes in the earliest metazoans. *EVC*and *EVC2*orthologues are absent from *Drosophila*and *Caenorhabditis elegans*genomes, indicating that they have been deleted since they last shared an ancestor with mammals. Direct protein interaction between Evc and Evc2 ----------------------------------------------- To search for Evc-interacting proteins we screened a cDNA library derived from E11 mouse embryos by the yeast two-hybrid assay using *Evc*sequence encoding amino acids 49 - 1005 as bait. This screen did not identify any known Hh pathway components but did identify Evc2 as an Evc-interacting partner. In order to map this interaction, deletion constructs for both Evc and Evc2 were generated and used in a directed yeast-two-hybrid assay (Figure [2a](#F2){ref-type="fig"}). Significant binding was observed with the following *Evc*constructs: expressing amino acid 49 - 1005, 463 - 1005, 222 - 873 and 222 - 800 and Evc2. No growth was observed with an *Evc*construct expressing amino acids 49 - 531 and restricted growth with the construct coding for amino acids 222 - 647. Thus interaction was observed for the four constructs containing the third and fourth coiled-coil regions but restricted interaction with the construct containing only the first three coiled-coil regions and no interaction detected with the constructs containing only the first two coiled-coil regions. Whilst the fifth and sixth coiled-coil regions were contained within constructs for which interaction was observed they were not tested independently of the fourth coiled-coil region. Significant binding was observed with the *Evc2*construct expressing amino acids 250 - 671 and Evc (Figure [2a](#F2){ref-type="fig"}). This portion of the Evc2 protein contains the first three predicted coiled-coil regions; interaction was not assessed for these three coiled-coil regions independently. We attempted to express full-length Evc and Evc2 proteins in mammalian cells to test for their interaction by co-immunoprecipitation (Co-IP). Full-length proteins could not be expressed at high levels in mammalian cells and were insoluble in standard Co-IP buffers. Therefore we co-expressed the shortest constructs for which strong interaction was observed (Evc amino acids 463 - 1005 and Evc2 amino acids 250 - 671) in HEK 293 cells with Flag and Myc-tags, respectively, to test whether Evc and Evc2 interact in a mammalian expression system. HEK 293 cells are ciliated renal cells \[[@B19]\]. Using antibodies against the tags, each protein was bound independently to protein G Sepharose beads. The beads were then extensively washed prior to elution of the interacting proteins and we detected co-immunoprecipitaton of the corresponding interacting partner by Western blot. Evc and Evc2 co-immunoprecipated with each other but not with vector or antibody controls confirming their interaction (Figure [2b](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Protein-protein interaction of Evc and Evc2**. **(a)**. Directed yeast two-hybrid assay. For each construct, the portion of Evc and Evc2 coding sequence expressed is depicted. Predicted coiled-coil regions are indicated by black boxes. Negative controls (empty vectors pGBKT7 and pGADT7) and positive controls (p53 and Large-T antigen) were included. + indicates colony growth/interaction, - indicates no colony growth/lack of interaction. **(b)**. FLAG-tagged Evc is immunoprecipitated by Myc-tagged Evc2 but not by the Myc epitope alone. Myc-tagged Evc2 is immunoprecipitated by FLAG-tagged Evc but not by the Flag epitope alone. A non-specific band corresponding to the IgG heavy chain (HC) is indicated. :::  ::: Evc and Evc2 colocalize at the basal body and cilia --------------------------------------------------- We have previously observed Evc at the distal end of the basal body *in vivo*and *in vitro*\[[@B7]\]. To study Evc2 localization we raised a new polyclonal anti-Evc2 antibody, R1656, and we observed Evc2 co-localization with Evc at the basal body in fibroblast, osteoblast and chondrocyte cultures and in renal derived IMCD3 cells (Figure [3a](#F3){ref-type="fig"} and not shown). In addition, we observed diffuse pericentriolar staining for Evc2 (Figure [3a](#F3){ref-type="fig"}). Specificity of the Evc2 immunostaining was demonstrated by antigen blocking and immunostaining in null Evc2 MEFs (data not shown). On testing additional cell lines to ascertain Evc/Evc2 localization we observed Evc along the length of the ciliary axoneme in all cilia of osteoblast-derived MC3T3 cells (Figure [3b](#F3){ref-type="fig"}). In these cells co-staining detected Evc2 mainly at the base of cilia (Figure [3c](#F3){ref-type="fig"}). Co-transfection of *Evc*and *Evc2-GFP*constructs into IMCD3 cells detected both proteins along the length of the cilium (Figure [4a-b](#F4){ref-type="fig"}), supporting the observations in MC3T3 cells. Co-transfection into MC3T3 and NIH3T3 cells also resulted in both proteins localizing to the cilia (results not shown). However, we did not observe basal body or cilia localization when transfecting constructs expressing either tagged Evc or Evc2 individually (Figure [4a](#F4){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### **Subcellular localization of native Evc and Evc2**. **Each row consists of images from one representative cell. (a)**. Evc (S43B; red) and Evc2 (R1656; green) colocalize at the base of the primary cilium in an IMCD3 cell. γ- and acetylated tubulin antibodies (both blue) identify the centrioles and the primary cilium respectively. Evc2 has an additional pericentriolar distribution. Scale bar 5 μm. **(b)**. Evc is located along the length of the ciliary axoneme in MC3T3 cells. Evc (S43B; red) colocalizes with the primary cilia marker acetylated tubulin (green). Scale bar 10 μm. **(c)**. Evc2 is concentrated at the base of the primary cilium and only partially colocalizes with Evc in MC3T3 cells. Evc (S43B; red) and Evc2 (R1656; green). Scale bar 10 μm. :::  ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### **Evc and Evc2 localize to the ciliary membrane in co-transfected IMCD3 cells**. **Each row consists of images from one representative cell. (a)**. Single transfection and co-transfection of Evc and Evc2-GFP constructs (the protein expressed is given on the left of each image). Expression of either protein alone does not result in localization on cilium. Evc is identified by S43B immunostaining (Evc ab, red), Evc2 by the GFP tag and cilia by acetylated tubulin staining (tubulin; green or red). Co-expression of both proteins results in their expression on the cilium. (cilia from two independent cells are shown). Merged images also show DAPI staining (blue). **(b)**. Immunodetection of Evc and Evc2-GFP co-expressed in cells treated with (+) or without (-) Triton X100 to permeabilize the membranes and allow access to the antibodies. Evc ab (S43B), raised to Evc aa 459 - 999, does not detect Evc in non-permeabilized cells; Evc2 ab 1 (Y20), raised to a Evc2 peptide between aa 50 - 100, detects Evc2 GFP in non-permeabilized cells and Evc2 ab 2 (R1656), raised to Evc2 aa 780 - 1124, does not detect Evc2 in non-permeabilized cells. **(c)**. Schematic representation of Evc and Evc2 on the ciliary membrane as determined from B, the C-termini of both proteins are intracellular. The regions for which we confirmed interaction are indicated by boxes and the predicted coiled-coil regions are shown as small black squares. Regions that are homologous (EVC: 339 - 889, and EVC2: 469 - 1018) are indicated by shaded rectangles. The epitope regions detected by Evc and Evc2 antibodies are indicated. Scale bar throughout 5 μm. :::  ::: Evc and Evc2 topography in the cilia membrane --------------------------------------------- Evc2 has a predicted signal peptide sequence and a second predicted transmembrane region (amino acids 209 - 231). To address whether Evc2 spans the cilia membrane and demonstrate its orientation in the membrane we immunostained co-transfected, non-permeabilized IMCD3 cells with antibodies generated against peptides N-terminal and C-terminal of the second transmembrane domain (Figure [4b](#F4){ref-type="fig"}). Y-20 antibody was raised against a peptide near the N-terminus of Evc2 between the two predicted transmembrane domains and R1656 against a C-terminal region of Evc2 (amino acids 780 - 1124) (Figure [4c](#F4){ref-type="fig"}, Evc2 ab 1 and 2, respectively). Y-20 detected Evc2-GFP along the length of the cilium in both permeabilized and non-permeabilized cells demonstrating that this portion of Evc2 is extracellular (Figure [4b](#F4){ref-type="fig"}, Evc2 ab1). However, R1656 only detects cilial Evc2-GFP in permeabilized cells, indicating that the region that is C-terminal of the transmembrane domain and detected by this antibody is intracellular (Figure [4bc](#F4){ref-type="fig"} Evc2 ab 2). S43B antibody (raised against Evc amino acids 459 - 999) detects native Evc in cilia in permeabilized MC3T3 cells, but does not detect Evc in non-permeabilized MC3T3 cells (result not shown) or in non-permeabilzed co-transfected IMCD3 cells (Figure [4bc](#F4){ref-type="fig"}, Evc ab). This indicates that the region of Evc C-terminal of the transmembrane domain is also intracellular. Thus, the regions of Evc and Evc2 that we have shown by Co-IP to interact are intracellular whilst the N-terminus of Evc2 is extracellular (Figure [4c](#F4){ref-type="fig"}). Evc2 is required for localization of Evc at the base of primary cilia --------------------------------------------------------------------- Since both Evc and Evc2 are co-dependent for cilia localization of expressed proteins, we tested if Evc2 is required for cilia localization of Evc in Evc2 null cells. MEFs derived from Evc2 mutant mice do not produce any Evc2 transcript or protein (Figure [5a](#F5){ref-type="fig"} and [6a](#F6){ref-type="fig"}). Evc and Evc2 are both detected at the base of primary cilia in wild-type MEFs (Figure [5c](#F5){ref-type="fig"} and not shown, respectively). Despite the presence of Evc transcript and protein (Figure [5a-b](#F5){ref-type="fig"}) Evc was not detected at the base of cilia in Evc2 null MEFs. This confirms that Evc2 is essential for Evc cilia localization. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### **Evc2 is required for the localization of Evc at the base of primary cilia**. **(a)**. RT-PCR amplification products from Evc2 null (-/-) and wild-type (+/+) MEFs. As expected, no Evc2 transcript was detected in the Evc2 null MEFs. A significant amount of Evc transcript was amplified in the Evc2 null MEF sample. Hprt transcript was amplified as a control. **(b)**. Western blot analysis of Evc protein in Evc2 null MEFs. The amount of β-actin detected on the same blot was used as a loading control. Evc (approximately 130 kDa) is present in Evc2 null (-/-) MEFs despite having reduced levels (approximately 50%) **(c)**. Representative immunofluorescent staining of Evc in MEF cells. Despite the presence of protein, Evc (red) was not detected at the base of primary cilia in Evc2 null MEFs (-/-). Primary cilia were identified by the presence of acetylated tubulin (green) and nuclei stained with DAPI (blue). Scale bar 10 μm. :::  ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### **Longitudinal and transverse sections of chondrocyte cilia obtained by transmission electron microscopy**. The longitudinal section and red lines (left) indicate the approximate position of transverse sections. Transverse sections of cilia are from wild-type (+/+) and Evc mutant (-/-) chondrocytes cells. In both genotypes, disorganized doublets were found in the distal region (red arrow heads) and the champagne glass structures were found in the proximal region (green arrow heads). Nine triplet microtubules and spike structures (blue arrow heads) could be identified in transverse sections of basal body. Scale bar 100 nm. :::  ::: Primary cilia of cells lacking Evc appear structurally normal ------------------------------------------------------------- Although cilia in mice lacking Evc look normal on standard microscopy \[[@B7]\] the possibility remained that the Hh signalling defect could be secondary to a structural abnormality of cilia. We therefore examined cilia structure in chondrocytes using transmission electron microscopy (TEM). In the basal body region, cilia from mutant chondrocytes showed the normal configuration of 9 triplet microtubules with transition fibers radiating from the triplets at the distal end of the basal body (Figure [6](#F6){ref-type="fig"}, blue arrowheads). The champagne glass structures that connect the microtubule doublets to the ciliary membrane seen in the transition zone \[[@B20]\] are present in the mutant chondrocytes (Figure [6](#F6){ref-type="fig"}, green arrowheads). Sections through the mid-portion of the cilia show 9 microtubule doublets with normal orientation and structure. In the distal region of the cilia there is some collapse of the microtubule ring in chondrocyte cilia from both wild-type and *Evc*^*-/-*^mouse (Figure [6](#F6){ref-type="fig"}, red arrowheads), a feature that has been reported before in normal cilia \[[@B21],[@B22]\]. All observed cilia showed complete triplet microtubule structure in the basal body and doublet microtubules in the ciliary region. No structural differences were observed between the cilia from wild-type and *Evc*^*-/-*^chondrocytes. Evc2, but not Evc, is found in the nucleus ------------------------------------------ Evc and Evc2 are detected on cilia in several different cell types by immunostaining. To determine if Evc and Evc2 are located elsewhere in the cell we performed subcellular fractionation on MEFs derived from wild-type and null mice. The purity of cytoplasmic and nuclear protein samples was confirmed using antibodies to alpha tubulin and c-jun, respectively (Figure [7](#F7){ref-type="fig"}). Specificity of the antibodies to Evc and Evc2 was confirmed by the absence of bands on Western blots of MEFs derived from null mice. We detected full-length Evc2 (approximately 140 kDa protein) both in the cytoplasmic and in the nuclear fractions but only detected full-length Evc (approximately 130 kDa protein) in cytoplasmic fractions. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### **Evc2 is present in both the cytoplasm and the nucleus of MEF cells**. Western blot analysis of total, cytoplasmic and nuclear proteins from wild-type (+/+) and Evc2 (-/-) (**a**) and Evc (-/-) (**b**) MEFs. The purity of the cytoplasmic and nuclear protein samples was confirmed using antibodies to α-tubulin (approximately 50 kDa) exclusively cytoplasmic and c-Jun (approximately 43 kDa), exclusively nuclear. Full-length Evc2 (140 kDa) was detected in both the cytoplasmic and nuclear samples. Full-length Evc protein (130 kDa) was only detected in the cytoplasm. Antibodies to both proteins detected several additional bands that are assumed to be non-specific as they are present in the null MEFs. :::  ::: Discussion ========== The fact that mutations in *EVC*and *EVC2*cause the same human phenotype, the chondrodysplasia Ellis-van Creveld syndrome, suggested that these two proteins act in the same pathway. We have shown that Evc and Evc2 are both positive regulators of the Hh pathway and that they interact directly with each other. The Hh pathway defect described in Evc and Evc2 mutants is downstream of the action of purmorphamine. It is known that purmorphamine directly activates Smo and promotes its entry onto the cilium \[[@B23],[@B24]\]. Thus, both Evc and Evc2 are positive regulators that act on, or downstream of, Smo in Hh signal transduction. Analysis of *EVC*and *EVC2*gene sequences predicts that they have arisen from an ancient duplication. Conservation of their 5\' to 5\' head to head arrangement and proximity of transcriptional start sites suggests that *EVC*and *EVC2*need to be co-ordinately expressed in order to maintain an appropriate stoichiometry, or because they function in the same biological pathway. It is interesting to note that *Evc*-like genes are absent from organisms that do not use cilia for Hh signalling (flies) or that do not have Hh signalling (worms). We studied Evc and Evc2 subcellular localization and observed that Evc2 co-localizes with Evc at the base of primary cilia in fibroblast and chondrocyte cultures and in renal-derived cell lines. On testing additional cell lines we observed additional Evc along the length of the ciliary axoneme in the osteoblast-derived MC3T3 cell line. One interpretation of the difference between the immunofluorescence analysis in this cell line and other cell lines, primary cell cultures and on tissue cryosections is that the amount of native protein normally present in cilia is below the sensitivity of immunofluorescence in most cells. Indeed, MC3T3 cells do express higher levels of Evc than other cells studied (Western blot analysis not shown). Cilia localization of co-expressed protein from constructs is in keeping with this interpretation but another possibility is that localization of Evc and Evc2 to the cilia is dependent on additional factors, analogous to the way in which localization of Ptch1, the Hh receptor, and Smo localization changes on addition of ligand. Ptch1 localizes to the cilium in the absence of ligand, but on ligand binding leaves the cilium upon which Smo enters the cilium and in turn regulates processing and activation of the three Gli transcription factors. We did not observe a change in Evc or Evc2 localization in MC3T3 and IMCD3 cells after treatment with purmorphamine (results not shown) which suggests that cilia localization of Evc and Evc2 is not dependent on pathway activation status. The observation that overexpressed Evc and Evc2 proteins localize to the cilium after co-transfection, but not when expressed individually, suggests that basal body and cilia localization is dependent on the interaction between Evc and Evc2. This was confirmed using MEFs from Evc2 null mice where Evc is present but does not localize to the base of primary cilia. Co-dependent localization of Evc and Evc2 could explain why mutations in either gene result in an indistinguishable patient phenotype. Given that both Evc and Evc2 possess transmembrane domains, observing them on the cilium led us to question whether they span the cilia membrane and, if so, their orientation in the membrane. We addressed this by comparison of immunofluorescence on permeabilized and non-permeabilized cells with antibodies generated against peptides N-terminal and C-terminal of the transmembrane domain of Evc2. These confirmed that Evc2 spans the cilia membrane and that the N-terminal portion is extracellular. Evc has a predicted signal anchor sequence and comparison of immunofluorescence signal on permeabilized and non-permeabilized cells demonstrated that Evc is intracellular. Furthermore, the regions of Evc and Evc2 that we have demonstrated to interact are both intracellular (Figure [4c](#F4){ref-type="fig"}). A question that arises when a Hh signalling defect results from loss of a cilia protein is whether this defect is due to a cilia abnormality or due to specific modulation of Hh signal transduction. Ultrastructural analysis of cilia of Evc mutant cells does not revealed any abnormalities suggesting that the mutant phenotype does not result from defective cilia but rather that Evc is a specific modifier of Hh signal transduction. In addition to the cilia localization, we detected Evc2 in MEF nuclear extracts by Western blotting. We detected Evc in cytoplasmic but not in nuclear extracts suggesting that Evc does not enter the nucleus, although we cannot exclude the possibility that Evc is present at undetectable levels in the nucleus. Evc and Evc2 were not observed in the nucleus by immunostaining, but a diffuse nuclear distribution may be difficult to distinguish from background staining. Also, overexpressed proteins were not observed in the nucleus when expressed alone or when co-expressed leading us to conclude that additional factors are required to transport Evc2 into the nucleus, a process that may depend on activation of the Hh pathway. The presence of Evc2 both in the ciliary membrane and in the nucleus is intriguing. There is a precedent for proteolytic processing and translocation to the nucleus of the C-terminal region of a cilia membrane protein in polycystin 1 (PC1) \[[@B25]\]. After cleavage, the C-terminal peptide of PC1 moves to the nucleus where it associates with Stat6 and p100 to activate gene transcription. The nuclear localization and the N-terminal extracellular portion of Evc2 may indicate an analogous role for Evc2 in the regulation of Hh target gene transcription in response to extracellular signals. One of the many recent surprises regarding Hh signalling was the finding that the Gli transcription factors localize to cilia as well as the nucleus. Further experiments are required to determine whether Ellis-van Creveld proteins move in conjunction with Gli proteins or via an independent pathway. Conclusions =========== We have previously shown that Evc is a positive modulator of Ihh signalling at the growth plate acting at or downstream of Smo. Here we have shown that Evc2 is also required for Hh signalling and that there is a direct physical interaction between the two proteins. We have shown that they localize to the membrane of primary cilia in a co-dependent manner and have deduced their orientation within the cilia membrane, the C-terminus of each being intracellular and only Evc2 having an extracellular portion. We have found that Evc2 is present in the nucleus but cannot detect Evc in nuclear extracts. We conclude that Evc and Evc2 are interacting proteins that together modulate Hh signal transduction. Methods ======= Cell culture ------------ All cells were cultured in medium containing 10% FBS (Invitrogen Ltd; Carlsbad, CA, USA) at 37°C and 5% CO~2.~Medium was obtained from Invitrogen Ltd. LIGHT2 cells were cultured in DMEM with 4.5 g/l glucose; MC3T3 cells in alpha MEM and IMCD3 and HEK 293 cells in DMEM/F12 (1:1) and DMEM respectively with non-essential amino acids. Mouse embryonic fibroblasts (MEFs) were established from E14.5 mutant and littermate control embryos \[[@B26]\] and cultured in DMEM without Na pyruvate, with non-essential amino acids and penicillin/streptomycin. Evc mice and the generation of *Evc*^-/-^MEFs was described previously \[[@B7]\]. *Evc2*^-/-^MEFs were derived from Evc2 null mice that were generated by replacing exon1 of *Evc2*with a reporter gene encoding the green fluorescence protein fused in frame to the first ATG of *Evc2*(unpublished data). Experiments were performed using MEFs cultured for less than eight passages. SiRNA knock-down and Hh assays ------------------------------ siRNAs were ON-TARGETplus SMARTpool for Mouse *Evc2*and ON-TARGETplus non-targeting siRNA pool 1 as a control (Dharmacon). Cells were transfected in triplicate on 12 well plates when 50% confluent. LIGHT2 cells were transfected with 100 pmoles siRNA/well using Dharmafect1 reagent (Dharmacon). MC3T3 cells were transfected with 80 pmoles siRNA/well using X-tremeGENE siRNA reagent (Roche Applied Science, Penzberg, Germany). One day later MC3T3 cells were co-transfected with the 8xGli-BS-Luc \[[@B27]\] and TK-Renilla plasmid (Promega WI, USA) at a ratio of 4:1 using Fugene HD reagent (Roche Applied Science, Penzberg, Germany). Cells were treated for 48 hours with purmorphamine (2 μM, Calbiochem, San Diego, CA, USA) or an equivalent amount of DMSO carrier as a negative control. LIGHT2 and MC3T3 cells were harvested 72 and 48 hours after transfection, respectively, and assayed for luciferase reporter expression using the Dual Luciferase Reporter assay system (Promega WI, USA) and a Luminoskan Ascent luminometer (Thermo Scientific, Walthem, MA, USA). The data was normalized by calculating the ratio of Firefly to Renilla luciferase readings (FF Luc/Ren Luc). Each experiment was repeated at least twice in triplicate. *P*values were calculated by *t*-test (Two Sample Assuming Unequal Variances). Evc2 protein knockdown in cell lysates was assessed by Western blotting and densitometry. For *Ptch1*RT-PCR in MEFs, assays were carried out in triplicate on 2 *Evc2*^*-/-*^and 2 *Evc2*^*+/+*^wild-type MEF cultures. Cells were treated with purmorphamine or DMSO as above for 48 hours. RNA was prepared using Trizol reagent (Invitrogen Ltd; Carlsbad, CA, USA) and first strand cDNA was synthesized using Superscript III (Invitrogen Ltd; Carlsbad, CA, USA). Simultaneous PCR amplification of *Ptch1*(nt 1944 - 2303 \[GenBank: [NM\_008957](NM_008957)\]) and *Hprt*(nt 108 - 294 \[GenBank: [NM\_013556](NM_013556)) was performed for 22 cycles in standard PCR conditions. Ratios of *Ptch1*to *Hprt*band intensity were determined for each culture before and after treatment. Sequence analysis ----------------- PSI-BLAST \[[@B28]\] searches employed default parameters, and mouse sequences as queries (unless otherwise stated) against the non-redundant protein sequence database held at the National Center for Biotechnology Information (Bethesda, MD). BLASTp and TBLASTn searches of *Lottia gigantea*and *Branchiostoma floridae*gene models and genome assemblies used web-based searches at the Joint Genome Institute <http://www.jgi.doe.gov/>. Signal peptides and anchors were predicted using SignalP-HMM \[[@B29]\]. Coiled coil sequences were predicted using Coils \[[@B30]\] and a threshold of *p*\> 0.5. The phylogenetic tree of *EVC*and *EVC2*sequences was constructed with the Fitch-Margoliash algorithm using a Poisson genetic distance and global optimization with bootstrapping (PMID: 5334057). Yeast Two-Hybrid analysis ------------------------- Mouse *Evc*in pAS2-1 vector was transformed into yeast strain AH109 and used as a bait to screen approximately 1.25 × 10^6^clones from a mouse 11-day embryo cDNA library constructed in the pACT2 vector and pre-transformed into yeast strain Y187 (Clontech, Mountain View, CA, USA). Positive interactions were identified by growth of mated bait and library cells on media lacking leucine, tryptophan, histidine and adenine at 30°C for 4 - 8 days. Positive colonies were confirmed by X-alpha-galactosidase activity assay. For the directed yeast two-hybrid studies, AH109 and Y187 yeast strains were transformed with *Evc*and *Evc2*constructs and mated. Matings between yeast containing the pGBKT7-p53 and pGADT7-T-antigen vectors were used as a control for positive interaction. Constructs ---------- For yeast two-hybrid library screening, mouse *Evc*sequence encoding amino acids 49 - 1005 which does not include the transmembrane domain was cloned into pAS2-1 vector (Clontech, Mountain View, CA, USA). For directed yeast two-hybrid analysis, mouse *Evc*fragments were cloned into pGBKT7 vector (Clontech, Mountain View, CA, USA) and mouse *Evc2*fragments were cloned into pGADT7 vector (Clontech, Mountain View, CA, USA). For co-immunoprecipitation studies, the *Evc*fragment (amino acids 463 - 991) was cloned into pCMV-3xFLAG-10 vector (3 × Flag fusion; Sigma-Aldrich, St. Louis, MO, USA) and the *Evc2*fragment (amino acids 250 - 671) into pCMV-3 vector (Myc fusion; Stratagene Corp; La Jolla, CA. USA). For subcellular localization studies, the complete mouse Evc coding region was cloned into pcDNA3.1(-) (Invitrogen Ltd; Carlsbad, CA, USA). The complete mouse Evc2 coding region was cloned into pEGFP-N1 (Clontech, Mountain View, CA, USA). The stop codon was mutated to allow translational read-through into the EGFP gene. All constructs were sequenced to confirm correct gene sequence and reading frame. Co-immunoprecipitation (Co-IP) ------------------------------ HEK 293 (Human Embryonic Kidney) were transiently transfected with the Myc- and 3 × FLAG-tagged constructs using GeneJammer reagent (Stratagene Corp; La Jolla, CA. USA), following the manufacturer\'s instructions. Transfections were performed in T75 flasks at 80% confluence, using 60 μl GeneJammer and 10 μg each plasmid, and were allowed to grow for 48 hours. Cells were resuspended in lysis buffer (50 mM HEPES pH7.4, 100 mM NaCl, 100 mM EDTA, 20 mM beta-glycerophosphate, 0.5% NP-40, 1 mM PMSF, Complete Protease Inhibitor Cocktail (Roche Applied Science, Penzberg, Germany) for 30 min, and spun in a microcentrifuge for 15 min at 4°C. Lysates were incubated for 1 hour at 4°C with Protein G Sepharose 4 Fast Flow beads (GE Healthcare, Uppsala, Sweden) to pre-clear, and spun in a microcentrifuge for 10 min at 4°C. Lysates were incubated 24 hours at 4°C with Protein G Sepharose 4 Fast Flow beads (GE Healthcare, Uppsala, Sweden) and 1 μg anti-Myc (9E10; Santa Cruz Biotech Inc; CA, USA) or anti-FLAG antibody (M2; Sigma-Aldrich, St. Louis, MO, USA). The beads were then washed extensively with lysis buffer. The co-immunopreciptates were analyzed by SDS-PAGE and Western blotting with anti-Myc (Santa Cruz Biotech Inc; CA, USA) or anti-FLAG (Sigma-Aldrich, St. Louis, MO, USA). Evc2 antibody production ------------------------ Amino acids 780 - 1124 of the mouse Evc2 protein (GenBank [BAC06589](BAC06589)) were expressed with a 6 × His tag in *E. coli*, purified by Ni^2+^chelation chromatography (Novagen) and used to immunize rabbit. Total IgGs were prepared from final serum (Protein G HiTrap, GE Healthcare, Uppsala, Sweden). Total rabbit IgGs often bind non-specifically to centrosomes therefore it was important that mouse Evc2 specific IgGs were isolated. For this, the antigen region was expressed in *Escherichia.coli*with a GST tag and purified on Glutathione sepharose 4B (GE Healthcare, Uppsala, Sweden). Specific anti-Evc2 IgGs, henceforth referred to as R1656, were purified by affinity to the GST-tagged Evc2 protein. Immunofluorescent staining -------------------------- Cells were fixed in 4% (w/v) paraformaldehyde (PFA) in PBS at 4°C for 10 minutes and permeabilized in 0.1% Triton X100 in PBS for 10 minutes (S43B and permeabilization experiments) or in PBS for 10 minutes (non-permeablization experiments); ice-cold MeOH/Acetone (1:1) for 6 minutes (R1656) or ice-cold methanol for 3 minutes (Y20). Primary antibodies were: sheep polyclonal anti-Evc (S43B \[[@B7]\]); rabbit polyclonal anti-Evc2 (R1656); goat polyclonal anti-Evc2 (Y-20, Santa Cruz Biotech Inc; CA, USA); anti-acetylated tubulin (Sigma-Aldrich, St. Louis, MO, USA) and mouse monoclonal anti-gamma tubulin (Sigma-Aldrich, St. Louis, MO, USA). Secondary antibodies were: donkey anti-sheep AlexaFluor 594 (Molecular Probes, Invitrogen Ltd; Carlsbad, CA, USA); goat anti-rabbit FITC (Jackson ImmunoResearch Labs Inc; PA, USA); goat anti-rabbit Cy3 (Sigma-Aldrich, St. Louis, MO, USA); donkey anti-goat FITC (Jackson Immuno Research Labs Inc; PA, USA); rabbit anti-goat Cy3 (Sigma-Aldrich, St. Louis, MO, USA); donkey anti-mouse AMCA (Jackson ImmunoResearch Labs Inc; PA, USA); horse anti-mouse TexasRed (Vector Labs, UK) and goat anti-mouse FITC (Sigma-Aldrich, St. Louis, MO, USA). Samples were mounted in Vectashield with or without DAPI (Vector Labs, UK) and images captured on an Axioplan 2 fluorescence microscope (Zeiss). Antibody blocking experiments were performed by preincubating primary antibody with approximately 2 μg of GST-Evc2 on beads and GST control (for R1656). At least ten cilia were visualized in each experiment. MC3T3 cells and MEFs were serum starved overnight prior to immunofluorescent staining to induce approximately 60% ciliation of cells. Ciliation of IMCD3 cells approached 100% without serum starvation. Transmission electron microscopy (TEM) -------------------------------------- Chondrocytes were isolated from the proximal tibial epiphyses of E18.5 wild-type and *Evc*^*-/-*^littermates. First tissue was dissected and washed in PBS. Cells were released from the extracellular matrix by sequential digestion with hyaluronidase (5 minutes, 1 mg/ml PBS), trypsin (10 minutes, 2.5 mg/ml PBS) and collagenase (5 hour, 3 mg/ml DMEM containing 10% FBS) at 37°C with constant rotation. The chondrocytes were incubated in DMEM for a maximum of seven days. We confirmed by real-time PCR that these cells retained chondrocyte expression profiles during this time period. Cells were grown on culture inserts (Nunc, Thermo Scientific, Walthem, MA, USA) and serum starved overnight prior to fixation to induce cilia production. Cells were fixed in 2% PFA/PBS at 4°C for 1 hour, dehydrated and embedded in LR White resin (EMS). Ultra thin sections (approximately 80 nm) were prepared on a RMC MT-XL ultramicrotome and stained on Pioloform filmed copper grids with 2% aqueous Uranyl Acetate and Lead Citrate (Leica UK Ltd). The ultra structure of 2 Evc^-/-^and 3 wild type cilia was observed with a Philips CM 100 Compustage (FEI) Transmission Electron Microscope and digital images collected using an AMT CCD camera (Deben). Cytoplasmic/nuclear fractionation and Western blotting ------------------------------------------------------ *Evc2*^*-/-*^MEFs were characterized by RT-PCR amplification of Evc2 (nt 533 - 830 \[GenBank: [AB083066](AB083066)\]); Evc (nt 1445 - 3060 \[GenBank: [AJ250841](AJ250841)\]) and Hprt (see above), and by western blotting. The cellular fractionation protocol was adapted from published methods \[[@B31]\]. Briefly null and control MEFs from T75 flasks were suspended in ice-cold cell swelling buffer containing 10 mM HEPES pH7.9; 10 mM KCl; 0.1 mM EDTA; 0.1 mM EGTA; 1 mM DTT; 0.5 mM PMSF and Complete protease inhibitors (Roche Applied Science, Penzberg, Germany) for 15 minutes. A sample was taken to provide total protein. Cytoplasmic proteins were released by vortexing in 4% NP-40 (Sigma-Aldrich, St. Louis, MO, USA) and collected in the supernatant after centrifugation for 30 seconds at 13500 g. The pellet was washed in cell swelling buffer three times and resuspended in three times the pellet volume of 20 mM HEPES pH7.9; 0.4 M NaCl; 1 mM EDTA; 1 mM EGTA; 1 mM DTT; 1 mM PMSF and protease inhibitors to release nuclear proteins. Nuclear proteins were collected from the supernatant after centrifugation for 5 minutes at 13500 g. Western blotting was performed using the following primary antibodies; rabbit anti-Evc2 (R1656 described here); sheep anti-Evc (S43G,\[[@B7]\]), mouse anti-α tubulin (clone B-5-1-2, Sigma-Aldrich, St. Louis, MO, USA); mouse anti-β actin (clone AC-15, Sigma-Aldrich, St. Louis, MO, USA) and rabbit anti-c-Jun (60A8, Cell Signalling Technology, Beverly, MA, USA). Secondary antibodies were peroxidase-conjugated, donkey anti-sheep (Jackson ImmunoResearch Labs Inc; PA, USA); goat anti-mouse (Thermo Scientific, Walthem, MA, USA) and goat anti-rabbit (Jackson ImmunoResearch Labs Inc; PA, USA). Peroxidase was detected using the SuperSignal West Dura extended duration substrate (Thermo Scientific, Walthem, MA, USA). Authors\' contributions ======================= HJB drafted the manuscript and demonstrated that Evc2 is required for Hh signalling; that Evc2 localizes to the basal body; that Evc/Evc2 cilia localization is co-dependent and that Evc2 is found in the nucleus. ST performed the yeast-2-hybrid analysis and Co-IPs. Y-N L performed the TEM analysis. JC demonstrated that native Evc and Evc2 localize to MC3T3 cilia. KM performed luciferase assays. CP performed the bioinformatic analyses. VRP prepared the construct to create Evc2 null mice used to derive MEFs and assisted in the preparation of the manuscript. JG conceived the study, participated in the design of the experimental work and interpretation of findings and drafted the manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional file 1 **A phylogenetic tree of *EVC*and *EVC2*sequences**. The tree was constructed with the Fitch-Margoliash algorithm using a Poisson genetic distance and global optimization with bootstrapping. Bootstrap values at internal nodes not supported at 100% are provided. Species abbreviations: Bf, *Branchiostoma floridae*; Hs, *Homo sapiens*; Lg, *Lottia gigantea*; Mm, *Mus musculus*; Nv, *Nematostella vectensis*; T, *Trichoplax adhaerens*; Tr, *Tetraodon nigroviridis*; and, Xt, *Xenopus tropicalis*. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was supported by BBSRC, MRC, the European Commission Framework Programme 6 Specific Targeted Research Project: LSHM-CT-2007-037471 and the Spanish Ministry of Science and Innovation (SAF-62291 and SAF-17901). JC is a Wellcome Trust Research Training Fellow. We would like to thank Hiroshi Sasaki (RIKEN Center for Developmental Biology, Japan) for providing the 8xGliBS firefly luciferase plasmid, Newcastle University, Electron Microscopy Research Services for assistance with the TEM analysis and Anita Wittner for the preparation of Figure [2](#F2){ref-type="fig"}.

PubMed Central

2024-06-05T04:04:19.064018

2011-2-28

PMC3052240

Background ========== Genomes are shaped by a series of processes involving substitutions, insertions, deletions, transpositions, shuffling of exons or chromosomes, lateral gene transfer, gene fusion or fission, *de novo*origination, and gene and genome duplications. The fate of the modified genes depends closely on the mutation type. For example, exon shuffling and substitutions are likely to have different outcomes on gene function and the subsequent putative fixation or maintenance of the new gene. Since the work of Ohno (1970) \[[@B1]\] gene and genome duplication has been considered as a primary driving force in the adaptive evolution of genomes and genetic systems. Duplication may even be considered as a \"master mutation\", as it promotes the accumulation of subsequent mutations on duplicates, as described below (substitutions, indels, etc.). A duplication can be segmental (from a few nucleotides to several thousand kilobases) or may cover the whole genome (an event also called polyploidization). Segmental duplication (or small-scale duplication) and polyploidization correspond to distinct evolutionary processes with widely different impacts. Segmental duplication is a frequent event that occurs in all eukaryote lineages as part of a \"continuous\" process \[[@B2]\]. By contrast, polyploidization is a much more infrequent and spectacular mutation event that leads to either extinction or re-diploidization. The diploidization process, involving non-homologous recombination events together with deletions and pseudogenizations of genes, generates duplicated chromosomes that differ in large segments, but still exhibit paralogous regions. Three polyploidization events have occurred in the last 150 million years in the *Arabidopsis thaliana*lineage. By contrast, no detectable polyploidization event has occurred in the *Drosophila*lineage for the last 600 million years. Segmental duplication of functional genes generates two copies, one of which generally loses its function rapidly through pseudogenization. Empirical data suggests that the majority of duplicates become pseudogenized in vertebrates. In the remaining cases, both duplicates are fixed either because one of the duplicates shifts toward a new function (neofunctionalization) or because the two copies subfunctionalize (*e.g.*, the copies differentiate their expression patterns through the evolution of different *cis-*regulatory modules). The nature of the duplication event (segmental or whole-genome) will influence the likelihood of specific genes being lost or fixed. For instance, local duplication of a gene whose product is involved in a large interacting network will generate supernumerary duplicates that will disturb the network stoichiometry (gene dosage principle). In this case, the duplication event will be counter-selected. On the other hand, if the same gene is duplicated through a polyploidization event, all the members of the network will likewise be duplicated, retaining the network stoichiometry and reducing the likelihood of the duplicate being counter-selected. In this review, the mechanisms of duplication, the processes of fixation or maintenance of duplicates and their impact in evolutionary history are discussed. We go on to describe experimental works that shed light on the fates of duplicates through classical events (such as neofunctionalization and subfunctionalization). Conceptual approaches are discussed that take into account evolutionary biology at the scale of gene, genome and population. Mechanisms of duplication ========================= Segmental duplication --------------------- The best-described mechanism causing segmental duplications from a few base pairs to several thousand kilobases is unequal crossing-over \[[@B3],[@B4]\]. One of the most famous examples of large segmental duplication is the primate major histocompatibility complex (MHC) \[[@B5]\]. Other mechanisms that generate segmental duplications are transposition and retrotransposition events. As transposition is a \"cut-and-paste\" process, it is not likely that transposition results in a duplication events unless it is associated with lateral gene transfer (LGT). In specific cases, \"cut-and-paste\" transposition can result in duplication i.e. if it occurs in germ cells prior to meiosis and moves DNA from one chromosomal set to the other. LGT seems to be frequent only in bacteria and archebacteria \[[@B6]\], in which it can generate duplication. On the other hand, retrotransposition seems a frequent event in many eukaryote lineages. In metazoa, the maximum size of the duplicated segment corresponds to a transcript messenger (with the exclusion of *C. elegans*, in which co-transcription of multiple genes is possible). Whole genome duplication: polyploidization ------------------------------------------ This event is recurrent in eukaryotes (especially plants) and has also been described in bacteria \[[@B7]-[@B9]\]. In eukaryotes, documented polyploidization events comprise both autopolyploidy (polyploidization within a species) and allopolyploidy (hybridization between closely-related species). Polyploidization events that occurred long ago in the history of various lineages are difficult to detect because of subsequent remodeling of the genome through gene losses and recombinations. Ancient genome polyploidization (termed paleopolyploidization) has, for example, occurred in yeasts, angiosperms and teleost fishes \[[@B10]\]. Polyploidization can occur *via*various mechanisms, such as genomic doubling, gametic non-reduction and polyspermy. Genome doubling and gametic non-reduction involve failure of cell division during mitosis and meiosis, respectively. Unreduced eggs seem common in both animals and plants, whereas unreduced spermatozoa seem to be common only in plants. An obstacle facing newly-formed tetraploid individuals is the fact that crossing with diploid relatives generates triploids, generally considered to be an evolutionary dead-end as they tend to produce aneuploid gametes owing to problems of chromosomal pairing and segregation during meiosis. However, it has been demonstrated that triploids can generate euploid (haploid, diploid, triploid) gametes at a low rate \[[@B11]\]. These euploid gametes can then produce triploid or tetraploid offspring. Thus triploids seem to be important, as they may facilitate the transition from diploidy to tetraploidy, and successful establishment of polyploidy appears to be facilitated by perenniality (because overlapping generations allow mating between triploids and their parents). Finally, selfing greatly facilitates polyploidization because it allows triploids to be maintained for several generations until stable polyploidy is generated. Interestingly, studies in plants have shown that the rate of polyploidy formation varies with environmental conditions and parental origin \[[@B12],[@B13]\]. For instance, a sudden freeze during egg development causes more frequent production of unreduced gametes. High rates of hybridization (e.g., in hybrid zones) may facilitate polyploidization, which in turn facilitates the generation of isolated lineages, as polyploids tend to be reproductively isolated from diploid ancestors. Studies in plants have shown that the mean frequency of diploid gametes found in hybrids (28%) is about 50 times greater that in non-hybrids (around 0.5%). Interspecific hybrids often experience severe meiotic disorders because of homologous chromosome miss-pairing. Mechanisms involved in the fixation and maintenance of duplicates ================================================================= Conceptual distinction between fixation and maintenance of duplicates --------------------------------------------------------------------- Before going further, we need to make a conceptual distinction between fixation and maintenance of duplicates in order to decipher step-by-step the fates of duplicates in the genome. Numerous studies of gene duplication have focused on the mechanisms and functional consequences of duplicated genes at the molecular or organism scale. A biased interpretation of the role of duplication is carried out whether the process of duplication itself is not clearly unravelled. Three steps are responsible for leading to the generation of preserved gene duplicates: i) mutational events (duplication), ii) fixation of duplicates, and iii) maintenance or preservation. In this review, we define fixation rate as the probability that a duplicate, regardless of its functionality, spreads into a population (*i.e.*, becomes fixed), and maintenance rate as the probability that a duplicate is stabilized in a population (preservation). We can propose the following theoretical classification for gene duplicates: Cat 1. Spreading difficult, maintenance difficult Cat 2. Spreading difficult, maintenance easy Cat 3. Spreading easy, maintenance difficult Cat 4. Spreading easy, maintenance easy. In the literature, duplication rate is a combination of the duplicating mutation and fixation rates. Most studies have used an empirical value for duplicating mutations (depending on specific lineages) and so the fixation rate is considered equal to the duplication rate. However, the rate of fixation of duplications cannot be used *a priori*to estimate rate of mutational origin \[[@B14]\]. Conceptual distinction between mutational generation (duplication), fixation and maintenance has critical implications for genome-scale studies as highlighted by experimental works. For instance, Davis and Petrov \[[@B15]\] investigated which types of genes are likely to generate functional and persistent duplicates and proposed that slowly evolving genes have a tendency to generate duplicates. Indeed, duplicated genes in the genomes of *Saccharomyces cerevisiae*and *Caenorhabditis elegans*have much slower rates of amino acid substitution, insertion and deletion than single copy genes. However, authors concluded that it is still unclear whether fixation, maintenance, or both of these steps together cause the bias towards the preferential duplication and highlighted that the relative importance of these two steps depends largely on the frequency with which duplicate genes are fixed by positive selection \[[@B15]\]. Duplication rate is also hard to estimate because of the difficulty in distinguishing true newly born duplicates from old ones that appear young because of gene conversion. Gene conversion is a homogenizing process between two homologous DNA fragments occurring during recombination. The divergence between two DNA fragments is biased and decreases dramatically following gene conversion. New models have been studied; for instance, Pan and Zhang propose an interesting strategy using unequal crossover and retrotransposition, to estimate rate that involves separate quantification of the rates of two different mechanisms of gene duplication and subsequent combination of the two rates, weighted according to their respective contributions to the overall gene duplication rate \[[@B16]\]. Conceptually, rate of duplication has to be considered as the resultant of a three-step process: duplicating mutations, fixation of duplicates, and finally maintenance of duplicates (long-term survival). Fates of genes after segmental duplication ------------------------------------------ Segmental duplication occurs in one individual within a breeding population, and the fixation of duplicates is constrained by classical variables of population genetics. Models of population genetics predict that an entirely redundant duplicate copy cannot be maintained in the genome for long, as harmful mutations will accumulate. Conversely, functional divergence will favor long-term retention of duplicates. Two major processes of divergence are possible; *(i)*neofunctionalization, where one copy retains the ancestral function while the other acquires a novel function (\[[@B1],[@B17]\], and *(ii)*subfunctionalization, where the ancestral functions of the progenitor gene are partitioned between the duplicates, so that the union of activities and expression patterns of the duplicates are equivalent to those of the progenitor gene \[[@B18],[@B19]\]. Modeling of the process predicts that subfunctionalization will be complex in populations with large effective sizes \[[@B20],[@B21]\]. He and Zhang \[[@B22]\] broadened the concept of neofunctionalization by considering that a duplicate may retain all, none, or part of the ancestral functions. We note that different authors emphasize different meanings of \'gene function\' \[[@B23]\]. For example, Hughes \[[@B18]\] refers to subfunctionalization of protein biochemical function, whereas Force *et al*. \[[@B19]\] emphasize subfunctionalized patterns of gene expression. One model, the duplication-degeneration-complementation model (DDC) describing the fate of duplicates was proposed by Force *et al*. \[[@B19]\] and illustrated in the works of van Hoof \[[@B24]\]. It involves complementary degenerative mutations in *cis*-regulatory modules: a fixed degenerative mutation in a regulatory module of duplicate A is followed by *(i)*accumulation of additional fixed degenerative mutations in the same copy, leading to its pseudogenization, or *(ii)*mutations, in copy B, of a complementary regulatory module (this second mutation occurs in a module that remains intact in copy A so that the two copies become essential for complete gene expression, preventing pseudogenization of either one), or *(iii)*by acquisition of a new function in copy B, through mutation in a complementary regulatory module (copy A is retained because it exhibits the original function associated with the non-mutated regulatory module). Obviously, probabilities of fixation of duplicates will also depend on the size of the population, and the selective pressures associated with the mutations in the corresponding *cis*-regulatory modules. Fate of genes after polyploidization ------------------------------------ One striking result concerning polyploids is that despite unstable genomes and rapid re-patterning, the addition to the genome of a complete set of chromosomes is remarkably well-tolerated in eukaryotes (and many current species descend from polyploid ancestors). Why is whole genome duplication more likely to generate lineages that persist over evolutionary time? Although the fate of genes after whole genome duplication depends on mechanisms similar to those discussed above for local duplication (pseudogenization, neofunctionalization and subfunctionalization), fixation of polyploidy cannot be discussed solely in the same terms as fixation of segmental duplicates, because polyploids tend to be at least partially isolated from the ancestral (non-polyploid) population by reproductive incompatibility. Furthermore, in the case of plants, polyploids can often be maintained through selfing and vegetative reproduction. We might expect the offspring of polyploid individuals to be necessarily polyploid. However, experiments on synthetic polyploids show that gene inactivation or subfunctionalization occur as early as the first generation *via*genome imprinting and genomic changes \[[@B25]\]. This process has been evidenced for both allo- and autopolyploids \[[@B26]\]. Song *et al*. \[[@B25]\] observed extensive genomic rearrangements and fragment losses within five generations of plant hybrids in the genus *Brassica*. Other studies report genomic changes soon after formation of wheat and *Arabidopsis*allopolyploids but not in cotton or cordgrass *Spartina*(a natural polyploid) \[[@B27]\]. In most of the examples studied, rapid genomic re-patterning has been observed in allopolyploids but not in autopolyploids. There are several reasons to suppose that hybridization may be responsible for re-patterning. For instance, transposable elements that are repressed within each parent lineage, but that can be activated in hybrids, could facilitate gene translocation and unequal crossing-overs. Josefsson *et al*. \[[@B28]\] found that maternally-derived siRNAs of hybrids were not sufficient to repress the retrotransposons originating from the parental genomes in *Arabidopsis thaliana*X *Arabidopsis arenosa*hybrids. Furthermore, divergence between centromeric histones from parental species may lead to chromosome segregation distortion and non-disjunction in hybrids. In addition, non-homologous recombination and non-reciprocal exchanges are particularly likely among homologous chromosomes with structural differences. Nevertheless, genomic re-patterning in polyploidy is not driven exclusively by hybridization. In autotetraploids of both *Candida albicans*\[[@B29]\] and *S. cerevisiae*\[[@B30]\], genome size reduction through chromosome losses has been observed. In the second study, haploid and tetraploid lines reverted to diploidy in 1800 generations. These experiments show that entire sets of chromosomes can be lost, although the exact mechanisms involved remain unknown. Genomic re-patterning may also increase the genetic variability of newly formed polyploid populations. This variability can be beneficial for the generated polyploid lineage as it can counteract the reduction of variability due to drastic reduction of population size, which is generally associated with polyploidization events. Surviving polyploids therefore probably form a biased subset of those that have been generated; we witness only lineages that have evolved towards particularly fit and stable genomic configurations soon after polyploidization. Besides drastic structural changes in their genomes, polyploids also often exhibit tissue-specific changes in gene expression (for review see \[[@B27]\]). This is especially so for allopolyploids, which can experience *(i)*changes in methylation \[[@B31]\], *(ii)*disruption of heterochromatin leading to retrotransposon activation \[[@B28]\] and *(iii)*alteration in imprinting and biased expression of homologs \[[@B32]\]. For example, Adams *et al*. \[[@B26]\] found that cotton allopolyploids differed from parental individuals in tissue-specific expression patterns for 11 out of 18 genes analyzed. As is the case for genome rearrangement, most changes in gene expression seem to be due to hybridization rather than to polyploidy *per se*\[[@B31],[@B33]\] and are correlated with divergence between the parental species \[[@B28]\]. Much smaller effects on gene expression were found in autopolyploids than in allopolyploids. Proteomic analyses in autopolyploid cabbage have shown very few expression changes \[[@B34]\]. Song *et al*. \[[@B25]\] also observed that less extensive genomic rearrangements occurred in allopolyploids when formed from more closely related species. Mechanisms favoring or opposing fixation or maintenance of duplicates --------------------------------------------------------------------- ### Neofunctionalization Neofunctionalization is context-dependent and may require multiple mutations. This process is better known as the \'Dykhuisen-Hartl\' effect \[[@B35],[@B36]\]: as one copy of a duplicated gene can freely mutate (a single copy must remain under pre-existing selective constraints), these mutations can lead to either pseudogenization or neofunctionalization. Mutations that are neutral in a particular environment can be positively selected by new environments or by epistatic interactions with subsequent mutations. The coefficient of selection can therefore vary in time. Neofunctionalization can be classified into two types: \"*stricto sensu\"*neofunctionalization and micro-neofunctionalization. The first type involves a radical shift in biochemical function or expression pattern, giving rise to a new function at high levels of organization \[[@B37],[@B38]\]. Examples in the literature include the crystalline proteins, the antifreeze proteins and many proteins from the major histocompatibility complex (MHC) (for a review, see \[[@B23],[@B37]\]. The second type, namely micro-neofunctionalization, involves a shift in the specificity of a metabolic activity, or in affinity for a given ligand, etc. \[[@B39]\]. Genes involved in recognition of the environment (such as olfactory receptors or MHC genes) probably originated through micro-neofunctionalization. Micro-neofunctionalization may be responsible for the surprising observation that bacterial strains can be recovered at higher-than-expected frequencies when they are plated on specific media in which mutations are advantageous. Hendrickson *et al*. \[[@B40]\] showed that increased mutation frequency was the direct consequence of an increase in the target gene copy number. For example, bacteria carrying a defective, but leaky, *lacZ*^-^allele produced more *lacZ*^*+*^revertants than expected when cultures were plated on lactose minimal medium. Most of these mutants appeared not during growth in liquid medium (*i.e.*, before plating), but after a period of very slow growth on the lactose plates. The mechanisms responsible for this effect are as follows: first, the *lacZ*^-^allele is strongly expressed, allowing bacteria to survive on lactose by producing very large amounts of the defective enzyme; this increase in gene expression is selectively advantageous, as it amplifies the minimal activity of the *lacZ*^-^allele to a level that permits cell survival. Second, the presence of multiple copies of the *lacZ*^-^allele makes more likely the chance occurrence of a mutation restoring the wild type (LacZ^+^) activity, and thus optimal growth on lactose. Once a gene copy reverts to the wild type allele, it spreads throughout the bacterial population and overruns the other gene copies, which rapidly disappear. This model proposed by Francino \[[@B41]\] is called \'adaptive radiation\'. It postulates an initial period of positive selection for gene amplification, followed by positive selection on the paralog copies for the acquisition of an advantageous phenotype. As proposed in the *lacZ*system, gene amplification could initially provide the means to reach biologically relevant levels of protein functionality, before neofunctionalization occurs. The evolution or expansion of multigenic families involved in sensory perception (*e.g.*, olfactory receptor families) may be partly explained by this process. ### Subfunctionalization *via*specialization Several authors have shown that neither neofunctionalization nor subfunctionalization alone can adequately account for retention. Analysis of the genome-wide patterns of yeast protein interaction and human gene expression for duplicate genes has revealed rapid subfunctionalization accompanied by prolonged, substantial \"*stricto sensu*\" neofunctionalization in a large proportion of duplicate genes, suggesting a new model, termed \"sub-neofunctionalization^\"^. A possible biological explanation is that subfunctionalization could be followed by neofunctionalization with positive selection because of the pleiotropic constraint release \[[@B18],[@B22],[@B42]\]. However, we could also consider that the term \"specialization\" would be more appropriate in this particular case. ### Gene duplications and genetic robustness Genetic robustness can be defined as *(i)*the ability of a biological system to withstand mutations due to redundancy, here the ability of duplicates to balance loss of function in other copies and *(ii)*participation in a biological network, *e.g.*, alternative metabolic pathways and regulatory networks. Duplicate genes undergo relaxed selection shortly after their duplication, which enables them to tolerate more nucleotide changes than their single-copy counterparts. Similar scenarios take place for whole genome duplication (WGD), where gene duplicates can tolerate up to 10 times more amino acid changes than old duplicates in vertebrates \[[@B43],[@B44]\]. Robustness is therefore essential in evolutionary innovation and phenotypic diversity. ### Dominant negative mutations Dominant negative mutations are mutations whose gene product adversely affects the function of the normal wild-type gene product within the same cell. Dominant negative mutations are therefore often more damaging than null mutations. The probability of gene loss will therefore be correlated with the proportion of possible dominant negative mutations. Cooke *et al*. \[[@B45]\] has hypothesized that partial protein damage has a stronger phenotype than null mutants caused by loss of gene expression. It has been stated above that duplicated genes with full redundancy can be expected to reduce to a single copy over time through the stochastic accumulation of mutations that harm one of the genes. In some genes, point mutations damaging protein integrity could also cause a defective phenotype. Thus the genetic redundancy cannot easily decay away through the accumulation of point mutations. The corollary of this is seen in knock-out studies showing that many genes can be removed in a single step that abolishes the expression of the protein-encoding gene, although point mutations of these proteins often have phenotypes. Cooke argued that naturally-occurring deletions are rarer than point mutations and to be viable must avoid damaging neighboring genes. Hence only a few deletion events suffice to knock out a gene. This effect will increase according to the interactiveness of the protein. Therefore, highly connected protein duplicates are retained because of gene dosage sensitivity and dominant negative counter-effect. ### Dosage sensitivity: dual consequences If the product of a duplicated gene belongs to a large protein complex, the duplication event can be counter-selected because it may generate imbalance among members of the protein complex. The gene dosage balance hypothesis (GDBH) proposes that such stoichiometric imbalances in macromolecular complexes are a source of dominant negative phenotypes \[[@B46],[@B47]\]. General evidence supporting the GDBH has been found for example in yeast and *Arabidopsis thaliana*: focusing on essential genes, Papp et al. \[[@B48]\] have shown that dosage-sensitive genes are at least twice as likely to encode proteins involved in complexes as genes with low dosage sensitivity. Furthermore, a statistically significant higher proportion of genes whose overexpression is lethal encode proteins involved in complexes \[[@B48]\]. Three predictions, largely confirmed by experimental work, can be made: *(i)*artificial overexpression of one subunit should be harmful, *(ii)*the strength of transcriptional co-regulation of subunits can be expected to reflect dosage sensitivity, and *(iii)*duplication of a single gene whose product is involved in a protein complex is likely to be harmful. Besides protein interaction stoichiometry, gene balance is believed to be governed by regulatory effects. Relative numbers of regulatory genes modify the expression of the target genes. Whenever there is competition between different offspring, dosage-sensitive gene losses will be counter-selected. Dosage sensitivity can be qualified by a dosage compensation effect. A *trans*-acting dosage effect can negatively affect the expression not only of genes located elsewhere in the genome, but also of the genes present on the same chromosome, yielding a compensation result. By contrast, duplication by WGD increases the dosage of all genes and so should not affect the balance. Analysis of duplicate genes arising from paleopolyploidization events in angiosperm, vertebrate, teleost, yeast and *Paramecium*phyla \[[@B49]\] support this prediction: the transcription factors along with proteins involved in protein binding, protein modification, and protein degradation, were more strongly retained than other protein functional classes \[[@B46]-[@B48]\]. ### Putative role of hitchhiking in gene duplication Local duplication can involve large genomic regions encompassing several genes. The probabilities of neofunctionalization, subfunctionalization or pseudogenization of each gene in the duplicon are unchanged, but positive or negative selection on a gene within the duplicon will influence the fate of the linked genes \[[@B50]\]. ### Intrinsic genome evolution Lynch and Conery \[[@B2]\] suggested an inverse correlation between population size and genome size (the difference in genome sizes among species being due to intron size, the presence of different repetitive elements, and the presence of duplicates). They suggested that purifying selection was intense in large populations, essentially precluding fixation of significantly damaging mutations, whereas mutation with substantial damaging effects could be fixed by random drift in small populations. Hence duplications might be fixed despite their potentially damaging nature. For instance, a newly inserted intron requires a critical mass of nucleotides (*n*= 20-40 nucleotides in range) to be accurately recognized and removed. Understanding the origins of eukaryotic genome complexity in adaptive terms is rendered difficult by the fact that each length increase of a gene raises its vulnerability to mutational inactivation, thereby favoring its elimination from the population. Relative contributions of polyploidization *versus*local duplication to genome evolution ======================================================================================== If duplications are considered as a major source of genomic novelty, then the frequency of such events will be crucial to the evolution of species. Participation in the creation of new genetic materials from whole genome duplication and local duplication will depend on the phylum. Some phyla show greater propensity than others to be polyploidized. For instance, the angiosperms contain 30 to 80 percent of species in a neopolyploid state, unlike the *Drosophila*lineages, which do not seem to have polyploidized for at least 600 million years. Regional duplication processes will probably have a deeper impact in species that do not polyploidize than in those that do. Around 15% of the genes in the human genome are believed to arise from duplication events, whereas gene duplicates account for 8-20% of the *Drosophila melanogaster*, *Caenorhabditis elegans*, and *Saccharomyces cerevisiae*genomes. However, these estimates are highly dependent on the sensitivity required to determine when a duplicate is detectable and others works hypothesized that almost all human genes resulted from ancient duplication \[[@B51]\]. Examples in vertebrates ----------------------- Genomic comparative analyses have revealed unexpected dynamics concerning family size, and once again underline the importance of gene and genome duplication in the history of evolution. In vertebrates, two rounds of whole genome duplications are thought to have played an essential role in the establishment of gene repertoires \[[@B52]\]. These events occurred during chordate evolution after the split of the urochordate and cephalochordate lineages, before the radiation of extant gnathostomes (jawed vertebrates). The rate of local duplication is estimated at between 1 gene per 100 and 1 gene per 1000 per million years \[[@B44],[@B53]\]. Calculations were performed as described below. Lynch and Conery assumed that the number of silent substitutions per site increased approximately linearly with time. The relative age distribution of gene duplicates within a genome can therefore be inferred indirectly from the distribution of silent substitutions. For all the species tested the highest density of duplicates tended to be contained within the youngest age classes, with the density dropping off rapidly with increasing silent substitution. A smooth decay was seen with species that had probably not recently polyploidized, such as *Homo sapiens*. This observation is explained by the birth and death process \[[@B54]\]; the youngest age category represents newly arisen duplicates, and the subsequent decline in frequency results from mutational processes that eliminate complete open reading frames (deletions or frameshift mutations). The rates of birth and loss of such genes can be derived directly from the observed age distribution, assuming that these rates have remained essentially constant within the age class employed in the analysis. From this analysis Lynch and Conery found that the average probability of duplication of a eukaryotic gene was 1 percent per million years. Cotton and Page also showed that a constant birth and death rate model was appropriate for gene duplication data, allowing the estimation of the rate of gene duplication and loss in vertebrates over the last 200 Myr (0.115 percent duplication, 0.74 percent losses). In this case, they used estimated times from fossils and molecular clock data. We must bear in mind that the aim of such analyses is to evaluate the average evolutionary properties of the members of duplicate gene pairs, and that some of the gene pairs will survive longer than average owing to positive selection (*via*neofunctionalization and subfunctionalization) or be shorter-lived than average owing to negative selection against the duplicate (see above). Dynamics of gene family size ---------------------------- The comparison of whole genomes reveals changes in the size of specific gene families among organisms \[[@B55]\], and several authors have found it possible to infer ancestral state and deduce which lineages in gene families have contracted or expanded. This approach enables us to classify gene families into the conceptual categories listed in III.1. The authors used a model of stochastic birth and death for the gene family that could be applied to multispecies genome comparisons. This model takes into account the branch length of phylogenetic trees, together with duplication and deletion rates, and so provides expectations for divergence in gene family size among lineages. This affords an estimate for the rate of fixation and loss for a given family. The analysis of Hahn *et al*. \[[@B55]\] is based on birth and death processes, but it might be more usefully considered as a fixation index and maintenance index. They analyzed gene families contained within the whole human genome, chimpanzee, mouse, rat and dog, and found that more than half of the 9990 families present in the mammalian common ancestor had either expanded or contracted along at least one lineage. They also found 164 families to be evolving non-randomly at *P*\< 10^-5^. With this cut-off threshold, they expected no family to be significant by chance. The most common biological functions assigned to these gene families included immune defense gene, neuron developments and intercellular communication and transport. Interestingly, comparisons of both synonymous to non-synonymous nucleotide divergence and regulatory sequence divergence also showed gene categories with these biological functions. The authors concluded that natural selection could act at many levels during adaptive molecular evolution. The real situation could probably be more complex, as a duplicate that is not important for the function will tend to be less constrained (some proteins are more constrained than others; environmental proteins may have less constrained sites). Specific functional studies on these families would thus be informative. Experimental work ================= Non-evolutionary-biology-based analyses --------------------------------------- Numerous analyses based on non-evolutionary approaches have been published. Below we describe two interesting examples that demonstrate the need to integrate evolutionary history into experimental work. ### Interactome analyses He and Zhang \[[@B22]\] analyzed the high confidence interaction data compiled by Von Mering and those annotated in the MIPS database. A total of 331 gene pairs and 745 singleton genes underwent the following analysis: the authors looked for duplicate pairs, numbers of specific partners and numbers of shared partners. After gene duplication the two duplicates have the same interaction partners. In the subfunctionalization model, each duplicate gradually loses partners, but the number of total partners remains constant over time. The mean number of total partners for duplicate genes was about 8.6. This was more than for a singleton gene, where a value of 4.7 was found (the difference was statistically tested). The authors concluded that the model that best explained their results was the neofunctionalization model. However, it is still possible that genes staying in single copy and duplicating genes have different number of partners, and evolutionary-based analysis should be performed to conclude. ### Expression analyses He and Zhang also analyzed human gene expression including the expression levels of 7565 human genes in 25 independent and non-redundant tissues \[[@B56]\]. They transformed the quantitative expression levels into discrete expression patterns (expressed or unexpressed). They analyzed expression patterns of 515 singletons and 1230 pairs of duplicate genes and found that the number of expression sites per duplicate pair was significantly greater than that per singleton gene. This refutes the pure subfunctionalization model. Using the synonymous mutation distance between duplicates as a clock, they examined how the number of expression sites had increased over time since duplication. To reduce random fluctuations, they put duplicates into seven bins depending on their divergence times and found that the number of expression sites and the times of duplication were positively correlated. Evolutionary-biology-based analyses ----------------------------------- Ideally, all studies concerning the function of duplicates should be integrated into evolutionary-based approaches where the history of a gene and the corresponding function have to be sought. Phylogenetic methods, for example, have been developed for inferring ancestral expression profiles or ancestral functions of homologs. In this case, the whole family history has to be integrated using the information at each branch of the tree in order to deduce the ancestral pattern or ancestral function at each node. Thus inferences about nodal values permit the estimation of evolutionary changes along each branch segment of an evolutionary tree \[[@B57],[@B58]\]. Because information is scant, authors use \"trio information\", *i.e.*, between one ortholog and two co-orthologs. In this case, the authors assume that orthologs retain the ancestral expression or function. ### Gene-centred analyses #### Subfunctionalization Hittinger and Caroll \[[@B59]\] investigated the evolution of one pair of duplicates in *S. cerevisiae*: GAL1 and GAL3. GAL1 encodes the galactokinase enzyme and the GAL3 gene encodes a co-inducer of galactokinase, able to sequester a repressor of the gene transcription factor activating the galactose use pathway. GAL1 and GAL3 have a co-ortholog in *Kluyveromyces lactis*. Phylogenetic analysis showed that GAL1/GAL3 duplicated in the *S. cerevisiae*lineage after the *K. lactis*and *S. cerevisiae*split. Compared with the *K. lactis*co-ortholog, GAL3 has lost its enzymatic activity, whereas GAL1 has changed its regulatory requirements. The outcome is a more tightly controlled and more highly inducible GAL1. The authors tested whether GAL1 and GAL3 duplication and subfunctionalization could have been fixed *via*positive selection and tested fitness differences in genetically manipulated *K. lactis*. Increased expression of the gene module providing galactokinase activity enhanced fitness, whereas overexpression of the module equivalent to GAL3 reduced fitness. In *S. cerevisiae*, there is no such conflict, and therefore the subfunctionalization could have been positively selected. #### Subfunctionalization and genetic robustness Hickman and Rusche \[[@B60]\] studied the duplicated histone deacetylases Sir2p and Hst1p in *S. cerevisiae*and found that these paralogs with non-overlapping functions could confer genetic robustness against null mutations through a substitution mechanism. Hst1p is a NAD(+)-dependent histone deacetylase that acts with Sum1p to repress a subset of mid-sporulation genes. However, the mutant deleted for hst1 showed much weaker derepression of target loci than the mutants deleted for sum1. The authors showed that this weak derepression of target loci in hst1Delta strains occurs partly because Sir2p substitutes for Hst1p. Sir2p helps to repress the mid-sporulation genes only in the absence of Hst1p and is recruited to target promoters by a physical interaction with the Sum1 complex. Also, when Sir2p associates with the Sum1 complex, the complex continues to repress in a promoter-specific manner and does not spread. In addition, SIR2/HST1 gene from *Kluyveromyces lactis*, a closely related species that diverged prior to the duplication, can suppress an hst1D mutation in *S. cerevisiae*as well as interact with Sir4p. These results suggest that the evolutionary path of duplicate gene preservation may be an important indicator for the ability of duplicated genes to contribute to genetic robustness. #### Subfunctionalization deduced from protein architecture Cusack and Wolfe describe how a bifunctional gene, encoding two proteins by alternative splicing, arose when the chloroplast gene RPL32 integrated into an intron of the nuclear gene SODcp in an ancestor of mangrove and poplar trees \[[@B61]\]. Mangrove retained the alternatively spliced chimeric gene, but in the poplar lineage, it underwent duplication and subfunctionalization, through complementary structural degeneration, to re-form separate RPL32 and SODcp genes. The partitioning process is considered to be a subfunctionalization because structural changes in the poplar genes indicated that after duplication a complementary loss of subfunctions of the ancestral chimeric gene occurred in its two daughter genes. The losses of exon X (encoding the RPL32 subfunction) in the *Poplar2*and *Poplar3*lineage, and of exons 4, 7 and 8 (encoding the SOD subfunction) in *Poplar1*, were caused by degenerative mutations that were probably selectively neutral because in each case the subfunction lost by one gene copy was maintained by the other. #### Neofunctionalization with functional evidence Zhang \[[@B62]\] reports that the gene encoding pancreatic ribonuclease was duplicated independently in Asian and African leaf-eating monkeys. Statistical analyses of DNA sequences, functional assays of reconstructed ancestral proteins and site-directed mutagenesis showed that the new genes acquired enhanced digestive efficiencies through parallel amino acid replacements driven by positive selection. They also lost a non-digestive function independently, under a relaxed selective constraint. These results demonstrate that despite the overall stochasticity, even molecular evolution has a certain degree of repeatability and predictability under the pressures of natural selection. ### Large scale analyses Growing information resulting from DNA sequence data enable us to carry out large-scale comparative analyses. #### Indirect information about biochemical function Scannell and Wolfe \[[@B63]\] studied genes for which either a single copy ortholog or double copy co-orthologs were available in eight yeast species (four of which diverged post-WGD while the four others diverged from an ancestor pre-WGD). They showed that, on average, proteins encoded by duplicate pairs evolved at least three times faster immediately post-WGD than single copy genes, to which they behave identically in non-WGD lineages. Although the high rate of duplicated genes subsequently declined rapidly, it has not yet reverted to the typical rate for single copy genes. They also showed that although duplicate gene pairs often have highly asymmetric rates of evolution, even the slower members of pairs showed evidence of bursts of evolution after duplication. Asymmetry after duplication was also evidenced in teleosts \[[@B64]\]. Kellis *et al*. demonstrate that the yeast *Saccharomyces cerevisiae*arose from ancient WGD, by sequencing and analysing a close specie *Kluyveromyces waltii*. Their results provide the first comparison across an ancient WGD event and offer the opportunity to study the long-term fate of a genome after duplication. In the majority of cases (95%), accelerated evolution concerned only one of the two paralogues. These results strongly support the model in which one of the paralogues retained an ancestral function while the other, relieved of this selective constraint, was free to evolve more rapidly \[[@B65]\]. This asymmetry could reflect positive selection or relaxation leading to neofunctionalization or subfunctionalization. #### Expression and functional shift analyses Tirosh and Barkai \[[@B66]\] developed a method to compare expression profiles from different organisms and applied it to analyze the expression divergence of yeast duplicated genes. Expression profiles of *S. cerevisae*duplicate pairs were compared with those of their co-orthologs in *C. albicans*. Duplicate pairs were divided into two classes: symmetric *versus*asymmetric rates of expression divergence. The expression of many of these duplicate pairs is highly correlated, suggesting that they were retained by selection for high protein dosage or evolved through other functional aspects such as protein structure or interaction. The asymmetric class includes 43 duplicate gene pairs in which only one copy showed a significant expression similarity to the *C. albicans*ortholog. Some of these cases may involve neutral evolution of gene expression of no functional significance, or they may involve regulatory neofunctionalization. Wapinski *et al*. \[[@B49]\] developed a procedure to resolve the evolutionary history of all genes in a large group of species. Their procedures were applied to 17 fungal genomes to create a genome-wide catalog of gene trees to determine precise orthology and paralogy relationships across these species. Gene duplication and loss are highly constrained by the functional properties and interaction partners of genes. Annotations were performed with the well-annotated *S. cerevisiae*. In particular, stress-related genes exhibited many duplications and losses, whereas growth-related genes showed selection against such changes. Whole genome duplication circumvents these constraints and relaxes the dichotomy, resulting in an expanded functional scope of gene duplication. By characterizing the functional fate of duplicate genes, they showed that duplicated genes rarely diverged with respect to the biochemical function, but typically diverged with respect to regulatory control. Gene duplication may drive the modularization of functional network through specialization, thereby disentangling cellular systems. Earlier observations suggested that paralogous modules were formed in massive duplication events. Wapinski *et al*. found that paralogous modules were rare even post-WGD and suggested an alternative mechanism. Many paralogous pairs genetically interact with each other despite having no shared physical interactions, which may induce a partial division of labor (subfunctionalization) between two paralogous proteins that become physically or temporally separated. Such specialization could modularize a molecular network by separating links within a network when duplicating a node. Thus increasing gene copy number may simplify a system rather than making it more complex. Modularization could relax opposing constraints on a single component and thus set in motion further specialization and refinement \[[@B49]\]. This report compares functional behaviour at different levels between duplicated genes, and shows that gene duplication innovates through regulatory divergence. After duplication of several genes, these can either migrate in a coordinated manner, resulting in two paralogous classes, or be dispersed into different classes. The authors expected coordinated migrations after simultaneous duplications. To test this hypothesis, they counted the number of paralogous gene pairs connecting each pair of gene classes (transcriptional classes, biochemical classes, etc.) and found that coordinated migration was rare. Gene classes (functional regulatory or transcriptional) rarely shared more than one or two paralogous relations regardless of the overall proportion of retained paralogs. The few observed paralog classes are very small and were formed gradually (from independent duplications). Thus paralogs dispersed individually. #### Subcellular localization shift Marques *et al*. \[[@B67]\] analyzed the possibility of neofunctionalization or subfunctionalization in subcellular localization. In their studies, the authors used first a non-evolutionary-based approach, and then an evolutionary-based one. The first non evolutionary-based approach hypothesized that divergent subcellular localization between duplicates was a consequence of sublocalization (subfunctionalization) alone. The joint number of different compartments per protein pair (combining both duplicates) would be expected to be the same as that of the common ancestral protein. Conversely, the number of compartments per pair should be higher than that of the progenitor if neo-localization contributed to sub-cellular diversification. However, sublocalization data for ancestral protein or for an outgroup were lacking. Thus to assess the contribution of neo- and sublocalization to the functional diversification of duplicates, the authors used the \"He and Zhang\" strategy. They used the average number of subcellular compartments of yeast singleton gene products as a proxy for the subcellular representation of WGD duplicate progenitors of yeast duplicates. They observed that the joint number of distinct compartments per WGD-derived duplicate with distinct cellular localization was significantly higher than that observed for singleton proteins. By contrast, there was no difference between the distributions of the number of subcellular compartments for WGD duplicates with the same subcellular distribution. This suggests that the increase in the number of compartments observed for the WGD-derived pair with distinct cellular localization was due to neolocalization events among these duplicates. The authors underline that their conclusions require a caveat: the types of proteins represented in the WGD-derived pair with distinct cellular localization may generally and *a priori*be present in a larger number of compartments. To check this, they compared the number of distinct compartments per WGD-derived pair with distinct cellular localization and singletons for proteins within the same Gene Ontology (GO) classes. This analysis showed that for all the GO classes tested, the joint number of compartments per WGD-derived pair with distinct cellular localization was significantly higher than that observed for singletons. This suggests that the elevated number of compartments for D-pairs (WGD-derived duplicates with distinct cellular localization) could be the result of neolocalization and not of a wide cellular representation of ancestral progenitor proteins, prior to duplication. They also used an evolutionary-based approach on a few families where the functional information was available in sister species (*K. waltii*). They then constructed the phylogeny for 45 yeast families, mapped the subcellular localizations of these onto the phylogenies and finally used a parsimony-based analysis to deduce the ancestral and derived states (function). In 16 families, the subcellular localization has remained fully preserved among members. For the remaining 29 families, they analyzed changes in protein location, assuming that the scenario requiring the smallest number of subcellular changes, given the observed data (parsimony principle), reflected the true pattern of events. For 16 of the 29 families, they inferred the most likely scenario of subcellular diversification. Eight families showed instances of neolocalization. For example, members of the ubiquitin-conjugating enzyme family, involved in protein degradation, are generally located in the cytoplasm and the nucleus. Conclusion ========== In the present review, we have discussed genome evolution *via*duplication and the mechanisms involved in the fixation and maintenance of the duplicates. The forces driving the fates of duplicate genes rely not only on duplication type (*i.e.*, segmental duplication or whole genome duplication), but also on several phenomena (opposing or compensatory) linked to population size, gene function and gene balance. Commonly, local duplication may be eliminated by passive losses following the genetic population laws, whereas in the case of whole genome duplication, duplicates or chromosome losses are an active, complex biological process resulting from an equilibrium disturbance in the cell. To decipher the role and impact of duplication in genome evolution, future works could be usefully reinforced in the following main directions: a\) The distinction between fixation and maintenance of duplicates needs to be biologically conceptualized and mathematically modeled in future studies. The birth and death process is described as a common mechanism to explain the dynamic of gene families \[[@B54]\]. Here we propose considering the birth and death process as a fixation and maintenance index, fixation being the probability that a duplicate will spread in a population, and maintenance being the probability that a duplicate is preserved in the long term. This distinction enables us to take into account essential variables such as population size along with functionality of genes. b\) We are tempted to consider duplication as a primary driving force in the adaptive evolution of genomes, but this \"master\" mutation still has to be integrated into an evolutionary context to assess its importance in genome evolution. Single correlations between duplicate numbers between families or their respective similarities are too scarce to unravel genome evolution, and history of the duplicates has to be clearly integrated into phylogenetic comparative methods \[[@B68]\]. In future studies, phylogenetic comparative methods should be considered as paradigm. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= Both authors planned and wrote the paper. Reviewer\'s comments ==================== Reviewer\'s report 1 -------------------- ### Dr. Etienne Joly, IPBS, UMR CNRS 5098, Toulouse, France This reviewer provided no comments for publication. The authors are grateful to the reviewer for helpful suggestions. Reviewer\'s report 2 -------------------- ### Dr. Lutz Walter, German Primate Center, Goettingen, Germany This informative review discusses mechanisms and consequences of genome evolution through duplications. The authors considered both whole genome and segmental duplications. The evolutionary consequences are discussed and the authors consider favouring as well as opposing mechanisms to fix and maintain such duplications. I have only a minor point concerning the fixation and maintenance of duplicates. This is a central point of the manuscript and, therefore, should deserve more description than just a few sentences. Examples maybe helpful for the readers. Author\'s response ------------------ We completely agree with the comment of Dr. Walter about the conceptual distinction between fixation and maintenance of duplicates. In our opinion, this is an essential and critical point to be considered in all modern and future studies based on the fates of duplicates and their role on genomes evolution. In line with the reviewer\'s suggestion, we now describe more deeply the theoretical and conceptual distinction between each of the steps leading to the generation of preserved gene duplicates. In addition, we discuss one example in which such a distinction is crucial to avoid several potential sources of error and bias in the estimates of evolutionary rates of duplication. Reviewer\'s report 3 -------------------- ### Dr. W. Ford Doolittle, Dalhousie University, Halifax, Nova Scotia, Canada This reviewer provided no comments for publication. Acknowledgements ================ We would like to thank Michel C. Milinkovitch for helpful and constructive discussions. We are grateful to the three expert reviewers for their time and thoughtful comments.

PubMed Central

2024-06-05T04:04:19.068105

2011-2-18

PMC3052241

Background ========== Xuesaitong (XST) injection, consisting of total saponins from *Panax notoginseng*(*Sanqi*), was widely used for the treatment of cardiovascular and cerebrovascular diseases in China. As total saponins (including ginsenosides and notoginsenosides) in the XST injection are its active ingredients, quality control of total saponins in the XST injection is critical for its safety, efficacy and stability. Single or simultaneous determination of main components of the total saponin extracts from *P. notoginseng*using high performance liquid chromatography-ultraviolet detection (HPLC-UV) \[[@B1]-[@B5]\], high performance liquid chromatography-evaporative light scattering detection (HPLC-ELSD) \[[@B6]\], high performance liquid chromatography-mass spectroscopy (HPLC-MS) \[[@B7]-[@B13]\] have been reported but over half of the total saponins were not quantified in these studies due to the lack of saponin references or poor chromatographic resolution. A comprehensive and systematic quality control of saponin extracts is much needed. Fingerprint analysis is currently developed for quality control in Chinese medicine \[[@B14]-[@B26]\] and has been accepted by the WHO for the assessment of herbal medicines \[[@B27]\]. The State Food and Drug Administration (SFDA) of China requires all herbal medicine-derived injections and related materials to use chromatographic fingerprints \[[@B28]\] in standardization. This article reports a novel fingerprint analytical method for quality control of the XST injection, which may be applicable to other herbal products. Over the previous studies \[[@B1]-[@B13]\], the new method features the following advantages. (1) The representative fingerprints show good chromatographic separation for most of visible peaks in the chromatographic profiles at 203 nm; (2) All main saponins (27 visible peaks in chromatographic profiles) are identifiable using high performance liquid chromatography-photo diode array detection/electrospray ionization tandem mass spectrometry (HPLC-PDA/ESI-MS^n^) technique, ten saponin references or data from literature \[[@B8]-[@B14]\]. Methods ======= Materials and reagents ---------------------- Acetonitrile and methanol (HPLC grade) were purchased from Merck (Darmstadt, Germany). Acetic acid glacial (HPLC grade) was from Tedia (Fairfield, OH, USA). The water used was purified by Milli-Q system (Millipore, USA). Reference compounds, namely notoginsenoside R~1~, ginsenoside Rg~1~, Rg~2~, Rh~1~, Rb~1~, Rb~2~, Rd, Re, 20(S)-Rg~3~and 20(R)-Rg~3~were purchased from Jilin University (Shenyang, China). The structures of these compounds are shown in Figure [1](#F1){ref-type="fig"}. Mixed standard stock solution containing accurately weighed reference compounds was directly prepared in 80% aqueous methanol (v/v). Working standard solutions were prepared by diluting the stock solution with 80% aqueous methanol (v/v) to obtain a series of concentrations for the calibration curves. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Structures of the investigated saponins in *P. notoginseng***. glc, β-D-glucose; glc\', α-D-glucosexyl, β-D-xylose; rha, α-L-rhamnose; araf, α-L-arabinose (furanose). Notoginsenoside I \*, H is instead of OH (C~12~) in 20S-form. SC1 \*\*, 6-O-β-D-xylopyranosyl -20-β-D-xylopyranosyl-(1→6)-β-D-glucopyranosyl dammar-24-ene-3β, 6α, 12β, 20(S)tetraol. :::  ::: HPLC instrumentationadditional 1 and chromatographic conditions --------------------------------------------------------------- An Agilent 1100 HPLC system (Agilent Technologies, USA) consisted of a quaternary solvent delivery system, an on-line degasser, an auto-sampler, a column temperature controller and ultraviolet detector coupled with an analytical workstation and an Ultimate™ XB-C~18~column, 5 μm, 250 mm × 4.6 mm i.d. (Welch Materials, USA) were used in the HPLC-UV experiments. Flow rate was 1.0 ml/min and sample injection volume was 10 μl. Detection wavelength was set at 203 nm and the column temperature was at 30°C. Mobile phase contained deionized water-acetic acid (A; 100:0.01, v/v) and acetonitrile-acetic acid (B; 100:0.01, v/v). The gradient elution was as follows: 19-21.2% B at 0-30 min; 21.2-26% B at 30-35 min; 26-28% B at 35-40 min; 28-38% B at 40-50 min; 38-55% B at 50-60 min; 55% B at 60-65 min; 55-80% B at 65-70 min; 80-95% B at 70-75 min. Re-equilibrium was 10 min; the total run time was 85 min. HPLC-MS^n^instrumentation and chromatographic conditions -------------------------------------------------------- Analysis was performed on an Agilent 1100 series LC system equipped with a binary solvent delivery system, an auto-sampler, a column temperature controller, a photo diode array detector and a Finnigan LCQ Deca XP^plus^ion trap mass spectrometer (Thermo Finnigan, USA) via an ESI interface. The chromatographic conditions were the same for HPLC-UV as described in the previous section. The operating parameters for MS in the negative mode were as follows: collision gas, ultrahigh-purity helium (He); nebulizing gas, high purity nitrogen (N~2~); ion spray voltage, -4.5 kV; sheath gas (N~2~) at a flow rate of 60 arbitrary units; auxiliary gas (N~2~) at a flow rate of 20 arbitrary units; capillary temperature, 350°C; capillary voltage, -15 V; tube lens offset voltage, -30 V. Full scan data acquisition was performed from *m/z*80 to 1800 in MS scan mode. The MS^n^spectra were obtained with the collision energy for collision-induced dissociation adjusted to 30%-40% of maximum and the isolation width of precursor ions was 2.0Th. Sample preparation ------------------ Ten samples of the XST injection (Batch No. 20090307, 20090510, 20090310, 20081018, 9042213, 20090312, 20090421, 20090512, 20090504, 20090203), manufactured by three Chinese pharmaceutical companies, were obtained either from pharmacies or factories. For HPLC-PDA-MS^n^analysis, a certain volume of the injection, according to its nominal content of total saponins, was transferred to a 50 ml volumetric flask and was diluted with 80% aqueous methanol (v/v) to obtain total saponins at a concentration of about 1 mg/ml. For HPLC-UV analysis, the injection was diluted with 80% aqueous methanol (v/v) to obtain total saponins at a concentration of about 0.5 mg/ml. Prior to analysis, the sample solutions were filtered through a 0.45 μm nylon membrane (Whatman, Britain). Spiked injection was produced by mixing sample solutions with the reference solutions at the ratio of 1:1. Data analysis ------------- Data analysis was carried out with Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine (version 2004A, National Committee of Pharmacopoeia, China) recommended by the SFDA. Results and discussion ====================== Optimization of HPLC separation ------------------------------- We optimized the separation conditions including the column, mobile phase, detection wavelength, elution gradient and column temperature in this study. Four reversed-phase columns, Agilent Zorbax Eclipse SB-C~18~columns (250 mm × 4.6 mm, 5 μm; 150 mm × 4.6 mm, 3.5 μm; 100 mm × 2.1 mm, 1.8 μm) and Ultimate™ XB-C~18~column (250 mm × 4.6 mm, 5 μm) were tested. The results showed that all four columns obtained good peak resolutions in 75 min, 75 min, 45 min and 75 min respectively; however, only two columns with the length of 250 mm (Zorbax Eclipse SB-C~18~and Ultimate™ XB-C~18~) produced more peaks in chromatograms. Ultimate™ XB-C~18~column (250 mm × 4.6 mm, 5 μm) was selected in the fingerprint analysis due to its lower cost than Zorbax Eclipse SB-C~18~column. The effects of mobile phase composition on chromatographic separation were also studied. The cetonitrile/water system produced more sharp peaks than the methanol/water system; the addition of 0.01% acetic acid in the acetonitrile/water system further improved the peak shape. Moreover, as the retention time of some components such as ginsenoside 20(S)-Rg~3~and 20(R)-Rg~3~was long, gradient elution was used in HPLC analysis. Satisfactory separation was achieved in 75 min. There was no strong absorption for most of saponins in the region of ultraviolet and visible spectra due to their structural characteristics, *eg*lack of conjugation groups in the molecular structures. As the end adsorption wavelength 203 nm is suitable for the assay of ginsenosides and notoginsenosides \[[@B1]-[@B5]\], it was selected as the detection wavelength in the experiment. Furthermore, the effects of column temperature on chromatographic separation were also examined. Four column temperatures, namely 20, 25, 30 and 35°C were tested. We found that at 30°C most peaks in chromatography had good resolutions; therefore, 30°C was chosen as the column temperature for the fingerprint analysis. HPLC-UV fingerprinting of the XST injection ------------------------------------------- To standardize the fingerprints, we analyzed ten samples using the optimized HPLC-UV method. Peaks found in all ten samples with good resolution were assigned as \'characteristic peaks\' and there were 27 characteristic peaks in the fingerprint chromatograms (Figure [2A](#F2){ref-type="fig"}). The software of Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine was used to evaluate these chromatograms. To exclude the effects of the solvent and baseline fluctuation, we selected the chromatographic data of these ten samples and treated them within the time frame of 28 min to 75 min. The similarities of chromatograms for the ten samples to the reference fingerprints were established using the means of all chromatograms (Additional file [1](#S1){ref-type="supplementary-material"}). The results showed that the ten samples possessed similarities to the reference fingerprints (Additional file [2](#S2){ref-type="supplementary-material"}). While the HPLC-UV fingerprints from different batches and companies varied, the 27 characteristic peaks were common in all samples. Therefore, the detection of these common peaks in HPLC fingerprints is useful in assessing the quality of the XST injection. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### **Chromatograms of (A) the representative fingerprint, (B) mixture standard compounds including (1) notoginsenoside R~1~, (2) ginsenoside Rg~1~, (3) ginsenoside Re, (9) ginsenoside Rb~1~, (11) ginsenoside Rg~2~, (12) ginsenoside Rh~1~, (13) ginsenoside Rb~2~, (15), ginsenoside Rd, (23) ginsenoside 20 (S)-Rg~3~and (24) ginsenoside 20 (R)-Rg~3~**. :::  ::: Identification of characteristic peaks -------------------------------------- HPLC-PDA/ESI-MS^n^was used for the components analysis and all 27 characteristic peaks were identified. In the ESI-MS experiment, the molecular weight of each peak was also obtained. By comparing with the ESI-MS^n^data and HPLC retention time of standard sanponins (Figure [2B](#F2){ref-type="fig"} and Additional file [3](#S3){ref-type="supplementary-material"}), we identified 10 peaks as notogisenoside R~1~, ginsenoside Rg~1~, Re, Rb~1~, Rg~2~, Rh~1~, Rb~2~, Rd and 20(S)-Rg~3~, 20(R)-Rg~3~. A total of 17 peaks were identified tentatively with the aid of the ESI-MS^n^data and HPLC retention time of some saponins from previous reports \[[@B1]-[@B13]\]. All the identification results are shown in Table [1](#T1){ref-type="table"}. In addition, The UV spectra of all peaks in the XST injection were obtained from the PDA chromatogram (Additional file [3](#S3){ref-type="supplementary-material"}). The results showed that among all the peaks in the chromatogram of the XST injection no strong UV absorption within the wavelength range from 210 nm to 400 nm was obtained, suggesting that the XST injection consisted of saponins with few other natural components possessing strong UV absorption, such as flavonoids, lignins, anthraquinones and alkaloids. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The identification results of saponins in the XST injection by LC/MS^n^ ::: **Peak No**. Identification Retention time (min) **MS\[M-H\]**^-^ MS data (*m/z*) -------------- ----------------------------- ---------------------- ------------------ ------------------------------------------------------------------------------------------------------ 1 Notoginsenoside R~1~ 34.89 932 799 \[M-H-Xyl\]^-^; 637 \[M-H-Xyl-Glc\]^-^; 475 Agl 2 Ginsenoside Rg~1~ 39.32 800 637 \[M-H-Glc\]^-^; 619 \[M-H-H~2~O-Glc\]^-^; 475 Agl 3 Ginsenoside Re 39.72 945 783 \[M-H-Glc\]^-^; 637 \[M-H-Glc-Rha\]^-^; 475 Agl 4 Notoginsenoside R~4~ 51.24 1240 1107 \[M-H-Xyl\]^-^; 1077 \[M-H-Glc\]\]^-^; 945 \[M-H-Xyl-Glc; 783 \[M-H-Xyl-2Glc\]^-^ 5 Ginsenoside Rf 51.89 800 637 \[M-H-Glc\]\]^-^; 475 Agl 6 Notoginsenoside Fa 52.17 1240 1107 \[M-H-Xyl\]^-^; 1077 \[M-H-Glc\]\]^-^; 945 \[M-H-Xyl-Glc; 783 \[M-H-Xyl-2Glc\]^-^ 7 Notoginsenoside I 52.39 1092 929\[M-H-Glc\]^-^; 767 \[M-H-2Glc\]^-^; 605\[M-H-3Glc\]^-^ 8 SC1 52.56 901 769 \[M-H-Xyl\]^-^; 637 \[M-H-2Xyl\]^-^; 475 Agl 9 Ginsenoside Rb1 53.48 1107 945 \[M-H-Glc\]^-^; 783 \[M-H-2Glc\]^-^; 621 \[M-H-3Glc\]^-^; 459 Agl 10 Notoginsenoside Fc 54.32 1209 1077 \[M-H-Xyl\]^-^; 945 \[M-H-2Xyl\]^-^; 783 \[M-H-2Xyl-Glc\]^-^; 621 \[M-H-2Xyl-2Glc\]^-^; 459 Agl 11 Ginsenoside Rg~2~ 54.75 783 637 \[M-H-Rha\]^-^; 621 \[M-H-Glc\]^-^; 475 Agl 12 Ginsenoside Rh~1~ 55.04 637 475 \[M-H-Glc\]^-^ 13 Ginsenoside Rb~2~ 55.30 1077 945\[M-H-Arap\]^-^; 915\[M-H-Glc\]^-^; 783\[M-HArap-Glc\]^-^; 621\[M-H-Arap-2Glc\]^-^; 459 Agl 14 Ginsenoside F~1~ 55.84 637 475 \[M-H-Glc\]^-^ 15 Ginsenoside Rd 57.16 945 783 \[M-H-Glc\]^-^; 621\[M-H-2Glc\]^-^; 459Agl 16 Notoginsenoside K 58.32 945 783 \[M-H-Glc\]^-^; 621\[M-H-2Glc\]^-^; 459Agl 17 Notoginsenoside T~5~/Unkown 61.70 752 619\[M-H-Xyl\]^-^; 457 Agl 18 Unkown 62.09 765 603\[M-H-Glc\]^-^ 19 Notoginsenoside T~5~/Unkown 62.42 752 619\[M-H-Xyl\]^-^; 457 Agl 20 Unkown 62.81 765 603\[M-H-Glc\]^-^ 21 Ginsenoside Rk~3~ 63.42 619 551 \[M-H-C~5~H~10~\]^-^ 22 Ginsenoside Rh~4~ 64.18 619 551 \[M-H-C~5~H~10~\]^-^ 23 20(S)-ginsenoside Rg~3~ 65.14 783 621 \[M-H-Glc\]^-^; 459 Agl 24 20(R)-ginsenoside Rg~3~ 65.86 783 621 \[M-H-Glc\]^-^; 459 Agl 25 Ginsenoside F~2~ 66.05 783 621 \[M-H-Glc\]^-^; 459 Agl 26 Ginsenoside Rk~1~ 72.47 765 603\[M-H-Glc\]^-^ 27 Ginsenoside Rg~5~ 72.89 765 603\[M-H-Glc\]^-^ ::: Determination of the main saponins in the XST injection ------------------------------------------------------- As shown in Figure [2A](#F2){ref-type="fig"}, 27 saponins were well separated, of which 25 were potentially identified (Table [1](#T1){ref-type="table"}). The ratio of total saponin peak area to all peaks (except for solvent peaks and baseline fluctuation in 0-28 min) in the chromatogram of each sample was beyond 95%. Thus, a method for quantification of the 27 saponins should provide a global and systematical evaluation for the quality control of the XST injection. However, it was difficult to obtain the reference compounds for all 27 saponins; we were only able to obtain ten, including notoginsenoside R~1~, ginsenoside Rg~1~, Re, Rb~1~, Rg~2~, Rh~1~, Rb~2~, Rd, 20(S)-Rg~3~and 20(R)-Rg~3~. Some reports \[[@B1]-[@B3]\] found that the slopes of regression equations for most of the determined saponins, such as notoginsenoside R~2~, R~4~, Fa, ginsenoside Rg~1~, Re, Rf, Rb~1~, Rg~2~, Rh~1~and Rd were approximately negatively correlated to their molecular weights by HPLC-UV at 203 nm (Additional file [4](#S4){ref-type="supplementary-material"}) and that the regression equations of some saponins with similar molecular weights were also close to each other under the same chromatographic condition (Additional file [5](#S5){ref-type="supplementary-material"}, [6](#S6){ref-type="supplementary-material"}, [7](#S7){ref-type="supplementary-material"}, [8](#S8){ref-type="supplementary-material"} and [9](#S9){ref-type="supplementary-material"}). Ten saponins, namely R~1~, ginsenoside Rg~1~, Re, Rb~1~, Rg~2~, Rh~1~, Rb~2~, Rd, 20(S)-Rg~3~and 20(R)-Rg~3~were quantitatively determined and the rest 17 saponins without standard references were semi-quantified using substitutive standard substances. The calibration curves for the quantification of each saponin were selected and listed in Table [2](#T2){ref-type="table"}. The developed analytical method was successfully applied to analysis of ten batches of the XST injection. All of the 27 characteristic peaks were determined simultaneously and the results are in Table [3](#T3){ref-type="table"}. In the XST injection, the content of ginsenoside Rb~1~was the most (26.17%-29.60%), followed by ginsenoside Rg~1~(20.50%-25.43%), Rd (6.82%-8.10%), notoginsenoside R~1~(5.29%-6.89%) and ginsenoside Re (2.91%-4.92%). The total content of the five saponins made up 61.69%-71.39% of the total saponins in the XST injection (total saponins nominal: 50 mg/ml). The ten saponins with available standard substances were quantitatively determined and made up 68.46%-75.85% of the total saponins nominal. Thus, combined with the semi-quantification data, 81.81%-95.71% components in the XST injection could be examined. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Calibration curves, detection limits and quantification limits of the saponins by HPLC-UV ::: **Peak No**. Saponins **M.W**. **Calibration curve**^**a**^ Linear range (μg/ml) ***R***^***2***^ LOD (μg/ml) -------------- ----------------------------- ---------- ------------------------------ ---------------------- ------------------ ------------- 21 Ginsenoside Rk~3~ 619 y = 6.7519x - 7.6085 22 Ginsenoside Rh~4~ 619 y = 6.7519x - 7.6085 12 Ginsenoside Rh~1~ 637 y = 6.7519x - 7.6085 4.28-68.5 0.9993 2.14 14 Ginsenoside F~1~ 637 y = 6.7519x - 7.6085 17 Notoginsenoside T~5~/Unkown 752 y = 5.4845x - 4.8387 19 Notoginsenoside T~5~/Unkown 752 y = 5.4845x - 4.8387 18 Unkown 765 y = 5.4845x - 4.8387 20 Unkown 765 y = 5.4845x - 4.8387 26 Ginsenoside Rk~1~ 765 y = 5.4845x - 4.8387 27 Ginsenoside Rg~5~ 765 y = 5.4845x - 4.8387 11 Ginsenoside Rg~2~ 783 y = 5.6715x - 5.6679 3.34-53.5 0.9993 1.67 23 20(S)-Rg~3~ 783 y = 5.4845x - 4.8387 2.95-47.3 0.9990 1.48 24 20(R)-Rg~3~ 783 y = 5.0923x - 2.8995 2.63-42.0 0.9994 1.75 25 Ginsenoside F~2~ 783 y = 5.4845x - 4.8387 2 Ginsenoside Rg~1~ 800 y = 5.1367x - 76.471 16.64-1065 0.9990 10.29 5 Ginsenoside Rf 800 y = 5.1367x - 76.471 8 SC1 901 y = 4.3254x - 5.0843 1 Notoginsenoside R~1~ 932 y = 4.3254x - 5.0843 10.26-492.5 0.9997 7.42 3 Ginsenoside Re 945 y = 4.4123x - 29.465 43.28-692.5 0.9993 4.73 15 Ginsenoside Rd 945 y = 4.1199x - 5.5681 16.64-532.5 0.9993 4.43 16 Notoginsenoside K 945 y = 4.1199x - 5.5681 13 Ginsenoside Rb~2~ 1077 y = 3.8757x + 2.4182 4.84-77.5 0.9995 1.95 7 Notoginsenoside I 1092 y = 3.8757x + 2.4182 9 Ginsenoside Rb~1~ 1107 y = 3.5815x - 29.548 15.98-1022.5 0.9992 7.91 10 Notoginsenoside Fc 1209 y = 3.5815x - 29.548 4 Notoginsenoside R~4~ 1240 y = 3.5815x - 29.548 6 Notoginsenoside Fa 1240 y = 3.5815x - 29.548 ^a^y: peak area of analyte; x: concentration of analyte (μg/ml) ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Contents (%) of the 27 saponins in the XST injection (total saponins nominal: 50 mg/ml) ^a^ ::: **Peak No**. Saponins S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 -------------- --------------------------------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- 1 Notoginsenoside R~1~(%) 6.64 5.29 6.89 6.47 6.27 5.86 5.33 6.41 6.07 6.35 2 Ginsenoside Rg~1~(%) 25.43 20.50 24.53 23.99 23.76 20.29 21.15 22.23 22.31 23.33 3 Ginsenoside Re (%) 3.43 2.91 4.92 3.61 3.55 3.56 3.35 3.04 3.03 3.69 4 Notoginsenoside R~4~(%) 1.52 1.19 1.24 1.33 1.28 1.33 1.31 1.11 1.15 1.38 5 Ginsenoside Rf (%) 1.24 0.95 0.98 1.15 1.15 0.97 1.03 1.03 1.03 1.00 6 Notoginsenoside Fa (%) 1.45 1.21 1.90 1.35 1.44 1.43 1.35 1.29 1.29 1.34 7 Notoginsenoside I (%) 0.89 0.62 0.17 0.80 0.80 0.76 0.81 0.73 0.66 0.83 8 SC1 (%) 0.65 0.51 2.28 0.56 0.62 0.46 0.54 0.52 0.49 0.54 9 Ginsenoside Rb~1~(%) 28.39 26.17 26.34 28.30 28.78 29.58 29.60 28.00 28.14 27.78 10 Notoginsenoside Fc (%) 1.30 0.94 0.99 1.13 1.12 1.06 0.98 1.05 1.05 1.15 11 Ginsenoside Rg~2~(%) 1.02 1.31 1.08 1.18 0.98 0.78 1.44 1.38 1.38 1.17 12 Ginsenoside Rh~1~(%) 1.77 3.06 2.25 2.22 1.65 1.06 2.90 3.19 3.22 2.17 13 Ginsenoside Rb~2~(%) 1.09 0.69 2.18 1.07 1.06 1.00 0.90 0.81 1.11 1.04 14 Ginsenoside F~1~(%) 0.76 1.77 0.29 1.14 0.85 0.50 1.59 1.90 1.88 1.24 15 Ginsenoside Rd (%) 7.50 6.82 7.25 7.22 7.24 7.27 8.10 7.41 7.48 7.18 16 Notoginsenoside K (%) 1.01 0.72 1.05 1.18 1.24 1.33 1.36 0.96 1.04 1.43 17 Notoginsenoside T~5~/Unkown (%) 0.39 0.69 0.58 0.69 0.47 0.39 0.79 0.87 0.86 0.83 18 Unkown (%) 0.30 0.37 1.11 0.45 0.36 0.23 0.56 0.50 0.50 0.46 19 Notoginsenoside T~5~/Unkown (%) 0.72 1.31 0.41 1.19 0.82 0.63 1.51 1.51 1.54 1.20 20 Unkown (%) 0.39 0.55 0.31 0.55 0.37 0.39 0.70 0.66 0.67 0.55 21 Ginsenoside Rk~3~(%) 0.90 2.30 1.59 1.78 1.10 0.80 2.35 2.52 2.57 1.77 22 Ginsenoside Rh~4~(%) 1.27 3.66 2.47 2.69 1.49 0.91 3.70 3.87 3.88 2.65 23 20S-Rg~3~(%) 0.37 1.01 0.75 0.81 0.44 0.43 1.21 1.09 1.14 0.83 24 20R-Rg~3~(%) 0.21 0.70 0.52 0.51 0.25 0.22 0.78 0.76 0.82 0.56 25 Ginsenoside F~2~(%) 0.36 0.38 0.23 0.28 0.14 0.10 0.78 0.42 0.43 0.25 26 Ginsenoside Rk~1~(%) 0.41 1.13 1.22 0.81 0.66 0.47 1.62 1.02 1.28 0.80 27 Ginsenoside Rg~5~(%) 0.32 1.30 1.17 1.05 0.65 0.46 1.95 1.31 1.50 1.03 **Total**(%) ^b^ 89.41 86.78 93.54 92.47 87.90 81.81 95.71 94.27 95.02 91.50 ^a^Mean values of samples (*n*= 3) ^b^Total content of the 27 saponins in samples ::: Conclusion ========== The fingerprint profiles of ten batches of samples showed 27 characteristic peaks. Ten of these 27 saponins in the XST injections were quantitatively determined with their standard references; the rest 17 saponins were semi-quantified with the substitutive standard references. Abbreviations ============= XST: Xuesaitong; HPLC-UV: high performance liquid chromatography-ultraviolet detection; HPLC-PDA/ESI-MS^n^: HPLC with photo diode array detection/electrospray ionization tandem mass spectrometry; HPLC-ELSD: high performance liquid chromatography-evaporative light scattering detection; HPLC-MS: high performance liquid chromatography-mass spectroscopy; SFDA: State Food and Drug Administration (China) Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= XHF designed the study. HY performed the fingerprint and quantitative analysis and wrote the manuscript. PYS and QS assisted HY to identify the characteristic peaks using HPLC-PDA/ESI-MS^n^. All authors read and approved the final version of the manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional file 1 **The chromatogram of similarity analysis of the fingerprints of 10 samples**. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 2 **The similarities of chromatograms of 10 samples (n = 3)**. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 3 **PDA Chromatograms**. standard compounds (A) and a XST injection (C), and total ion current chromatograms of standard compounds (B) and a XST injection (D). 1-27 were the characteristic peaks, listed in Table [2](#T2){ref-type="table"} ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 4 **Plots of slopes of calibration curves vs. molecular weights (MW) of saponins**. From literatures (A) \[Journal of Pharmaceutical and Biomedical Analysis 41 (2006) 274-279\], (B) \[Journal of Pharmaceutical and Biomedical Analysis 48 (2008) 1361-1367\], (C) \[Journal of Pharmaceutical and Biomedical Analysis 38 (2005) 45-51\], (D) \[Journal of Chromatography A 1011 (2003) 77-87\], (E) \[Journal of Shenyang Pharmaceutical University Vol. 20, No.1 (2003) 27-31\], and (F) \[Chinese Pharmaceutical Journal Vol. 38, No.9 (2003) 698-699\] ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 5 **The method validation for simultaneous determination of the twenty-seven saponins in XST injection**. The quantitative and semi-quantitative methods were validated and the semi-quantitative principle were discussed in detail. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 6 **Precisions and repeatability**. The results of precision and repeatability for simultaneous determination of the twenty-seven saponins ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 7 **Recovery**. The results of recovery for simultaneous determination of the twenty-seven saponins ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 8 **Plots of slopes of calibration curves vs molecular weights (MW) with different chromatography columns**. (A) Ultimate™ XB-C18 (250 mm × 4.6 mm, 5 μm), (B) Zorbax Eclipse SB-C18 (250 mm × 4.6 mm, 5 μm) and (C) Zorbax Eclipse SB-C18 (100 mm × 2.1 mm, 1.8 μm) ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 9 **Regression equation using different columns**. Columns: Zorbax Eclipse SB-C18 (250 mm × 4.6 mm, 5 μm) and Zorbax Eclipse SB-C18 (100 mm × 2.1 mm, 1.8 μm) ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was supported by the National S&T Major Project (No. 2009ZX09502-005 & 2009ZX09311-002) and Zhejiang Provincial Natural Science Foundation, China (R2080693).

PubMed Central

2024-06-05T04:04:19.073314

2011-2-24

PMC3052242

There was an error in Table 3. P-values at column 5, line 1 (0.05), line 7 (0.05) and line 9 (0.04) should be in bold. **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.077779

2011-2-28

PMC3052243

There was an error in Table 2. Please see the corrected Table 2 here: ::: {#pone-9ecf10d6-782c-4af6-a7cb-15d4b0e2305d-g001 .fig}  ::: **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.078150

2011-2-28

PMC3052255

Introduction {#s1} ============ Angiosperms manifest a wide array of floral displays, from species that produce only a single large flower per plant, to those that produce hundreds of tiny flowers on many inflorescences per individual. The attractiveness of these different floral displays is closely coupled to mating system, the type and breadth of pollinators and individual fecundity [@pone.0016111-Goodwillie1], [@pone.0016111-Primack1]. Studies across plant species show that species tend to produce fewer flowers per individual as the size of the flower increases [@pone.0016111-Charlesworth1], [@pone.0016111-Harder1], [@pone.0016111-Sakai1], [@pone.0016111-Sakai2], [@pone.0016111-Sargent1],[@pone.0016111-Worley1],[@pone.0016111-Worley2]. The causal agents of this trade-off have been investigated empirically in a number of plant species, indicating that hierarchical resource allocation [@pone.0016111-Obeso1], [@pone.0016111-Richards1], pollen discounting via geitonogamy in obligate out-crossing species [@pone.0016111-Goodwillie1], [@pone.0016111-Harder1], [@pone.0016111-Richards1] and pollen limitation [@pone.0016111-Harder2] are all important drivers in this trade-off. Few studies have, however, determined whether a flower size/number trade-off occurs across, as well as within, species (but see 7) and examined its consequences for plant community structure. The potential implications could include partitioning of pollinators among species as a function of flower size, differential pollen dispersal as a function of pollinator size and mobility relationships [@pone.0016111-Greenleaf1], and differences in fecundity among species. Surprisingly, interactions between plants and their pollinators have received limited attention as a driver of species diversity in tropical tree communities [@pone.0016111-Stone1], [@pone.0016111-Ghazoul1], [@pone.0016111-Sargent2]. Partitioning of pollinator services could limit competitive exclusion, and if these interactions can be shown to contribute to an equalisation of fitness, contribute to species coexistence [@pone.0016111-Chesson1]. We combine experimental ecology, field observation and molecular ecological approaches to investigate the ecological implications of a flower size/number trade-off in a clade of tropical forest canopy and emergent trees. The processes that determine the differential fecundity of individuals are poorly understood, and therefore the importance to demography of pre-dispersal stages of plant reproduction may have been underestimated. Studies that link floral traits to fecundity are also likely to reveal important implications of flower size vs number trade-offs for species\' differential responses to habitat degradation [@pone.0016111-Harder2], [@pone.0016111-Vamosi1]. For example, increased isolation of individual trees is of concern for the management of tropical tree species that have been heavily fragmented in recent years [@pone.0016111-Lowe1]. We used a group of 12 coexisting dipterocarp species ranging in flower size (calyx diameter from 1.2 mm to 10.2 mm) to explore the function of flower size and number and pollination systems in fruit production. Specifically, we combine measures of flower size, flower number and pollination success (as inferred from pollen tube growth data) with measures of paternity and genetic relatedness within mapped populations of two species to test the following explicit hypotheses. 1. Across species, flower size scales negatively with mean flower production per individual, which would be indicative of a flower size/number trade-off. 2. Across species, mean pollination success (as defined by the proportion of flower styles containing pollen tubes) increases with flower size. 3. Species with small flowers have small-bodied pollinators. 4. Species with smaller bodied pollinators have more limited pollen dispersal distances. 5. The proportion of flowers that give rise to fruit scales positively with flower size. 6. Spatial aggregation and limited pollen dispersal increase inbreeding in a small flowered species. These hypotheses encapsulate a set of trade-offs between flower size, pollinator body size and pollen dispersal that might have consequences for fruit production and fecundity, which are important components of plant fitness (see [Figure 1](#pone-0016111-g001){ref-type="fig"}). ::: {#pone-0016111-g001 .fig} 10.1371/journal.pone.0016111.g001 Figure 1 ::: {.caption} ###### Schematic illustrating how interactions between flower size and number, pollination biology, pollen dispersal and fruit production might contribute to equalisation of fitness components among species of Dipterocarpaceae. :::  ::: Tree species in the Dipterocarpaceae dominate lowland rain forests of Southeast Asia, accounting for up to 80 percent of the density and basal area of canopy trees [@pone.0016111-Ashton1]. Dipterocarps display variable flower size and morphology and are pollinated by insects including thrips, beetles, moths and giant honeybees [@pone.0016111-Ghazoul2], [@pone.0016111-Konuma1], [@pone.0016111-Sakai3], [@pone.0016111-Fukue1], [@pone.0016111-Kenta1], [@pone.0016111-Lee1], [@pone.0016111-Naito1]. Reproductively mature trees of many dipterocarps exhibit irregular mass flowering [@pone.0016111-Ashton1]. Dipterocarps produce hermaphroditic flowers and are considered predominately out-crossing, which is maintained through pre- and post- zygotic incompatibility mechanisms [@pone.0016111-Bawa1]. Apomixis has been reported in some species [@pone.0016111-Fukue1], [@pone.0016111-Ng1], but we found no evidence of this in the progeny arrays of our study species. Fruit dispersal occurs by gravity and wind assisted gyration for the majority of fruit falling beneath the mother tree or a within 20--30 m [@pone.0016111-Osada1], [@pone.0016111-Seidler1]. Limited seed dispersal leads to the aggregation of related individuals, and significant fine-scale spatial genetic structure has been observed in the majority of dipterocarp species studied [@pone.0016111-Kettle1]. Results {#s2} ======= Flower and fruit production as a function of flower size {#s2a} -------------------------------------------------------- Analyses of absolute (unscaled) log-transformed values of maximum potential flower production yielded a single most likely model ([Table S1a](#pone.0016111.s002){ref-type="supplementary-material"} Supplementary Information), in which flower production declined as a function of flower size for the comparison of 11 species (F~1,7~ = 35.88, *P* = 0.0005, [Fig. 2a](#pone-0016111-g002){ref-type="fig"}). In this model canopy species had lower values of unscaled maximum flower production than emergent species (F~1,7~ = 57.05, *P* = 0.0001), and the interaction between life form and flower size was also significant (F~1,7~ = 21.81, *P* = 0.0023). Analyses of maximum flower production scaled to tree size yielded three most likely models ([Table S1a](#pone.0016111.s002){ref-type="supplementary-material"} Supplementary Information). In all three cases scaled maximum flower production declined significantly with increasing flower size (e.g. F~1,7~ = 70.39, *P*\<0.0001 for the model with the lowest AIC value, [Fig. 2b](#pone-0016111-g002){ref-type="fig"})), but there was no difference between canopy and emergent species (F~1,7~ = 1.71, *P* = 0.23) and no interaction of life form and flower size (F~1,7~ = 2.13, *P* = 0.18). ::: {#pone-0016111-g002 .fig} 10.1371/journal.pone.0016111.g002 Figure 2 ::: {.caption} ###### Effects of calyx tube diameter (flower size, mm) on estimated flower and fruit production in 11 dipterocarp tree species. Effects of calyx tube diameter on estimated (a) maximum per capita flower production, (b) maximum per capita flower production scaled to basal area and mean species tree height, (c) maximum per capita fruit production, (d) maximum per capita fruit production scaled to basal area and mean species tree height, (e) mean percentage of flower styles with pollen tubes and (f) mean number of pollen tubes per pollinated flower for four canopy (open symbols) and seven emergent (closed symbols) dipterocarp species growing in Sepilok Forest Reserve, Sabah. Maximum flower and fruit production were estimated from the mean of the upper quartile of values per species (n = 2 to 14 individuals per species). :::  ::: In contrast to flower production (where small flowered species produce more flowers than larger flowered species), there was not a significant effect of flower size in any of the three most likely models analysing absolute values of maximum fruit production ([Fig. 2c](#pone-0016111-g002){ref-type="fig"}) or in the single most likely model of maximum fruit production scaled to tree size ([Fig. 2d](#pone-0016111-g002){ref-type="fig"}). Emergent species had a marginally greater value of maximum absolute fruit production than canopy species ([Fig. 2c](#pone-0016111-g002){ref-type="fig"}), and this term was significant in one of the three most likely models for this variable, but there was no difference between canopy and emergent species in maximum fruit production scaled to tree size ([Fig. 2d](#pone-0016111-g002){ref-type="fig"}). The nearest neighbour distance term was non-significant in all analyses of fruit production among species. The relationships between flower size and fruiting variables and pollination success (median number of pollen tubes per pollinated flower) were all supported by phylogenetically independent contrasts. Of the 12 species used in this study, only eight species exist in well resolved phylogenies [@pone.0016111-Gamage1], [@pone.0016111-Kamiya1]. With the exception of *S. xanthophylla* all the *Shorea* species in our sample fall within the Red Meranti clade [@pone.0016111-Kamiya1], with no apparent phylogenetic signal of calyx size. Pollination success {#s2b} ------------------- Analyses of the proportion of flowers pollinated identified five most likely models ([Table S1c](#pone.0016111.s002){ref-type="supplementary-material"}, Supplementary Information), but in all cases a positive effect of flower size was the only significant term (F~1,7~ = 55.97, *P*\<0.001, [Fig. 2e](#pone-0016111-g002){ref-type="fig"}). Similarly, flower size was the only significant term in all the five most likely models analysing interspecific differences in the mean number of pollen tubes per pollinated flower (F~1,9~ = 45.72, *P*\<0.0001, [Fig. 2f](#pone-0016111-g002){ref-type="fig"}). Differences in life form and in median nearest neighbour distance were not significant in these models of pollination success among species. Pollinator body size {#s2c} -------------------- The experimental exclusion of pollinators based on body size resulted in a significant decline in pollination success with increasing flower size (*t* = −6.18, *P*\<0.0001, [Fig. 2](#pone-0016111-g002){ref-type="fig"}). In small-flowered species (1.0--2.0 mm), such as *Shorea xanthophylla*, restricting access to pollinators had very little effect on the proportion of fruit produced relative to that expected among open pollinated flowers. In contrast, in the largest flowered species *Dipterocarpus grandiflorus* fruit set within treatments that limited pollinator access to small bodied insects was substantially reduced under all mesh sizes ([Fig. 3](#pone-0016111-g003){ref-type="fig"} see Dg). The response to mesh bag treatments was significant (*t* = 2.76, *P* = 0.011), but the interaction of flower size and mesh bag treatment was not significant. ::: {#pone-0016111-g003 .fig} 10.1371/journal.pone.0016111.g003 Figure 3 ::: {.caption} ###### Effect of pollinator exclusion using bags with mesh apertures of 0.2 mm (solid line, small closed symbols), 2 mm (dashed line, open symbols) and 5 mm (dotted line, large closed symbols) on pollination success one month after anthesis. Pollination success is defined as the mean percentage of flowers that gave rise to immature fruit for 8 to 10 bags per treatment on 2 to 6 replicate trees per species, expressed as a percentage of the mean value for an equivalent sample of open-pollinated flowers on the same tree. :::  ::: Pollen dispersal distance and genetic relatedness between assigned parents {#s2d} -------------------------------------------------------------------------- Paternity analysis revealed that short-distance mating events were far more common among the smaller flowered species *S. xanthophylla* (47% of matings within 50 m and 71% within 100 m of the mother tree) than the larger flowered species *P. tomentella* (13% within 50 m and 32% within 100 m) (χ^2^ = 17.96, *P*\<0.001). The median (Wilcox rank sum  = 7325, *P*\<0.0001) and mean (*t* = −4.609, *P*\<0.0001) pollen dispersal distances for all assigned progeny showed a similar pattern: 63 m and 119 m in *S. xanthophylla* and 164 m and 177 m in *P. tomentella* respectively. Because the density of potential pollen donors within 100 m of mother trees was generally greater for *S. xanthophylla* than *P. tomentella*, we also tested for differences in pollen dispersal between the two species using a subset of mother trees that had equivalent distances to nearest flowering neighbours. In the subset of *S. xanthophylla* mothers, 32 percent of matings were assigned at less than 50 m and 34 percent at less than 100 m. The results of the TWOGENER analysis of pollen dispersal are summarised in [Table 1](#pone-0016111-t001){ref-type="table"}. The differentiation in the pollen gene pool is significant in both species (*P*\<0.01) and almost twice as great in *S. xanthophylla* than in *P. tomentella* (0.0624 and 0.0364 respectively). In this analysis the mean pollen dispersal distance for *S. xanthophylla* was estimated at 56 m (S.E. 2.88) and for *P. tomentella* 416 m (S.E. 29.93). The effective number of pollen donors per mother tree (N*~ep~*) was 8.01 in *S. xanthophylla* and 13.73 in *P. tomentella* ([Table 1](#pone-0016111-t001){ref-type="table"}). The effective pollen neighbourhood area (*A~ep~*) ranged from 1.46 ha (*A~ep~^min^*) −14.62 ha (*A~ep~^max^*) in *S. xanthophylla* and from 8.43 ha (*A~ep~^min^*) −84.26 ha (*A~ep~^max^*) in *P. tomentella* ([Table 1](#pone-0016111-t001){ref-type="table"}). ::: {#pone-0016111-t001 .table-wrap} 10.1371/journal.pone.0016111.t001 Table 1 ::: {.caption} ###### Mean ± SEM calyx tube width (mm) for the 12 study species of Dipterocarpaceae sampled in Sepilok forest reserve (SFR). ::: {#pone-0016111-t001-1} Number of sample trees (n) ----------------------------------- ---------- ----------------------------- ---------- ------- ---------------------------- ----- --- *Shorea xanthophylla* (Sx) Canopy 40 m (25 m) - 35 m\* 1.2±0.1 19.7 29 26 3 *Shorea multiflora* (Sm) Canopy 40 m (No estimate) - 35 m\* 1.3±0.2 37.9 35 31 6 *Hopea beccariana* (Hb) Canopy 45 m (35 m\*) 1.4±0.1 200 14 11 6 *Shorea leprosula* (Sl) Emergent 60 m (60 m) - 65 m\* 2.2±0.2 85.1 55 33 6 *Shorea parvifolia* (Sp) Emergent 65 m\* (60 m) 2.3±0.1 69.7 6 6 0 *Shorea johorensis* (Sj) Emergent 50 m (65 m\*) 2.5±0.1 66.7 32 29 6 *Shorea macropteraa* (Smac) Canopy 40\*- 50 m (60 m) 3.0±0.2 118.2 6 6 0 *Shorea smithiana* (Ss) Emergent 60 m\* (55 m) 3.4±0.2 91.8 22 10 3 *Shorea beccariana* (Sb) Emergent 60 m (55 m\*) 3.6±0.3 39.5 53 38 0 *Parashorea tomentella* (Pt) Emergent 65 m (65 m\*) 4.2±0.1 50.5 29 25 6 *Shorea macrophylla* (Smph) Emergent 45 m (45 m) 4.9±0.2 N/A N/A N/A 2 *Dipterocarpus grandiflorus* (Dg) Emergent 45 m \* (45 m) 10.2±0.7 24.8 30 18 3 Mean tree heights are from Ashton (2004), except values in brackets which are from [@pone.0016111-Wood1] and those marked \*, which were measured directly at SFR or obtained from a published source equivalent to our own measure. Median two nearest neighbouring (NN) distance to flowering conspecific trees; number of sample trees (n) for a) quantification of flower and fruit production; b) pollination success, based on pollen tube analysis and c) trees for experimental pollinator exclusion. † Following Ashton [@pone.0016111-Ashton2]. ::: Genetic relatedness between mothers and assigned fathers was an order of magnitude higher in progeny of the small-flowered species *S. xanthophylla* than in the larger flowered species *P. tomentella* (mean pairwise kinship coefficient  = 0.033 (S.E. 0.008) and 0.002 (S.E. 0.01) respectively; *t* = 2.195, df = 229, *P* = 0.028). The mating system parameters based upon the progeny arrays indicate that both species are extensively outcrossing (*t~m~* = 0.996 and *t~m~* = 0.907 in *S. xanthophylla* and *P. tomentella* respectively). However, the parental inbreeding coefficient and biparental inbreeding were both greater in *S. xanthophylla* (see supporting information [Table S2](#pone.0016111.s003){ref-type="supplementary-material"}). Discussion {#s3} ========== Flower size/number trade-off and fecundity {#s3a} ------------------------------------------ We confirmed that flower size scaled negatively with flower number across 11 dipterocarp species. Although canopy species produced fewer flowers for their flower size than predicted by the scaling relationship for emergent species, all species fit the same negative relationship when flower production was scaled to tree size. These relationships suggest that the trade-off between flower size and potential flower production was consistent across all species, but subject to allometric constraints. These observations are robust to phylogenetic comparison These results are consistent with other studies that suggest that a flower size vs number trade-off is common in angiosperms [@pone.0016111-Sargent1], . Maximum fruit production per tree, in contrast to flower production, did not scale negatively with flower size. This indicates that the potential fecundity offered by a large number of small flowers was offset by the reduced likelihood that each of these flowers would be pollinated and develop into a mature fruit. Our results suggest that processes operating between flower production and fruit dispersal contribute to equalising fecundity among dipterocarp species. Below we interpret our findings and discuss the range of alternative hypotheses that could lead to the observed patterns of fecundity. Fruit production in larger fruited species could be resource limited, leading to larger fruits being produced in fewer numbers [@pone.0016111-Richards1]. However, we exclude this hypothesis for two reasons. First, the scaling relationships were determined for trees representing the upper quartile for flower production, which are least likely to be resource-limited among the sample. Secondly, fruit size scales positively with flower size in dipterocarps (C.R. Maycock *et al.*, unpubl. data), therefore small-flowered species would have lower overall per capita fruit mass, even though fruit number remains the same, a scenario that is at odds with the expectations of resource limited fruit production. Experimental exclusion of all potential pollinators except thrips reduced pollination success in large-flowered but not small-flowered species. This result suggests that the role of thrips in pollination is important for small-flowered species of dipterocarp such as *Shorea xanthophylla* but less so as flower size increases. Thrips are abundant in the flowers of all the dipterocarps examined in this study and are widespread in the flowers of many non-dipterocarps in the same forest (C.R. Maycock, *et al*. unpubl. data). Hence their abundance is unlikely to limit pollination success in the small-flowered species, and therefore we reject pollinator limitation as a determinant of the flower size-dependent reduction in pollination success. There are several concurring lines of evidence to support the hypothesis that not only do large-flowered species receive a higher proportion of pollen (as empirically demonstrated here for 11 species) but also that pollen received is of greater genetic quality (*i.e.*, more compatible). First, our examination of pollen tube growth among species showed that both the proportion of pollinated flowers and mean number of pollen tubes per pollinated flower was greater in larger-flowered species. This implies that either smaller-flowered species receive fewer per capita visits by pollinators than larger-flowered species, and/or that a higher proportion of pollination events result in the transfer of non-compatible self-pollen that is blocked at the stigmatic surface or at the early stages of pollen tube growth. Dipterocarp species are considered to be self-incompatible [@pone.0016111-Bawa1], and although there is evidence that selfing can lead to fruit production, these fruit are preferentially aborted at an early stage of maturation [@pone.0016111-Dayanandan1], [@pone.0016111-Momose1]. This behaviour may have evolved as a predator avoidance strategy [@pone.0016111-Ghazoul3]. Second, the pollinator exclusion experiment also revealed that fruit set increased as access to larger pollinators was permitted by larger mesh sizes in larger flowered species. The exception to this pattern was the largest flowered species, *Dipterocarpus grandiflorus*, which had significantly reduced fruit set even with the largest mesh (5 mm). This outcome suggests that the importance of larger pollinators increases as flower size increases and, for the largest flowered species, *D. grandiflorus*, pollinators that exceed 5 mm in size, such as giant honey bees *Apis dorsata*, large scarabid beetles and Sphingidae and Noctuidae moths, play an important role. We conclude that the role of small-bodied insects in dipterocarp pollination, in particular thrips, declines with increasing flower size. The relevance of this finding in the context of putative trade-offs is that the larger bodied pollinators may transfer pollen over greater distances and may therefore be more effective agents for outcrossing [@pone.0016111-Greenleaf1], [@pone.0016111-Haskell1]. ### Differential pollen dispersal {#s3a1} Our study provides evidence that larger bodied pollinators disperse pollen over greater distances than smaller bodied pollinators in dipterocarps. Our paternity analysis and indirect estimates of pollen dispersal demonstrated that short-distance pollen dispersal events were more frequent, and that the mean pollen dispersal distance were shorter, for the small-flowered species *Shorea xanthophylla* than for *Parashorea tomentella*, independent of density of flowering conspecifics. Progeny collected from a subset of *S. xanthophylla* mother trees with an equivalent density of flowering conspecifics to that of *P. tomentella* were assigned more frequently to nearer pollen donors (\<100 m) than for *P. tomentella*. Other studies have estimated mean pollen dispersal distances in dipterocarp species to lie in the range 175--207 m [@pone.0016111-Konuma1], [@pone.0016111-Kenta1], [@pone.0016111-Lee1]. For the five dipterocarp species where pollen dispersal distances have been quantified, representing flower sizes ranging over two orders of magnitude, mean pollen dispersal distance increases with flower size, with the lowest mean dispersal distances in *S. xanthophylla* and greatest dispersal distances in *Dipterocarpus tempehes* ([Table 2](#pone-0016111-t002){ref-type="table"}). Such a pattern remains inconclusive, as it is based on only five species, and other factors may influence pollen-mediated gene flow, including differences in the density of flowering conspecifics. Nevertheless, the results are suggestive that increasing flower size corresponds to higher pollen flow distances, which we interpret as being mediated by the larger-bodied pollinators attracted by the larger flowers. Hence, all else being equal, these larger-flowered species experience higher outcrossing rates and are less susceptible to inbreeding. This corresponds to a higher quality of pollination service for large flowered species, which compensates for the relatively few flowers produced by these species. ::: {#pone-0016111-t002 .table-wrap} 10.1371/journal.pone.0016111.t002 Table 2 ::: {.caption} ###### Summary table of pollen dispersal from TWOGENER analysis in two species of dipterocarp with contrasting flower size. ::: {#pone-0016111-t002-2} Species d/trees ha Φft δ (m) *N~ep~* *A~ep~^min^* (ha) *A~ep~^max^* (ha) ------------------- ------------ ------------------------------------------- ------------------------- --------- ------------------- ------------------- *S. xanthophylla* 5.48 0.0624[\*\*](#nt104){ref-type="table-fn"} 56.27 (43.08--69.48) 8.01 1.46 14.62 *P. tomentella* 1.63 0.0364[\*\*](#nt104){ref-type="table-fn"} 416.25 (292.84--539.67) 13.73 8.43 84.26 Density of adult trees (d); differentiation in pollen gene pool (Φ*~ft~*); mean pollen dispersal distance(*δ*); effective number of pollen donors (*N~ep~*); minimum effective pollen neighbourhood area (*A~ep~^min^*) and maximum effective pollen neighbourhood area (*A~ep~^max^*) following Smouse *et al* 2001. \*\* Significant at *P*\<0.01, values in parentheses are 95% confidence intervals. ::: Inbreeding and thus reduced fecundity through inbreeding depression are predicted to be greater in species where related individuals are highly aggregated. Seed dispersal limitation and habitat specialisation are common in dipterocarps [@pone.0016111-Osada1], [@pone.0016111-Seidler1], and both processes can lead to aggregation of related individuals [@pone.0016111-Hardy1], [@pone.0016111-Jones1], [@pone.0016111-Loveless1]. This is measurable as fine-scale spatial genetic structure (SGS) within populations. A strong SGS signal would not be expected if gene flow by pollen or seed dispersal is extensive. Dipterocarp seed dispersal is known to occur over short distances [@pone.0016111-Bagchi1], [@pone.0016111-Blundell1], [@pone.0016111-Blundell2], [@pone.0016111-Curran1], [@pone.0016111-Wells1], which suggests that long distance pollen flow is likely to be the primary agent undermining SGS. The evidence for fine-scale SGS observed in comparable studies of dipterocarp species [@pone.0016111-Richards1], [@pone.0016111-Lee1], [@pone.0016111-Ng1], [@pone.0016111-Kettle1], [@pone.0016111-Ng2], [@pone.0016111-Takeuchi1] suggests that SGS is common in the Dipterocarpaceae, but weaker in larger flowered species. This interpretation is consistent with our pollinator exclusion experiments, and other studies [@pone.0016111-Kenta1], [@pone.0016111-Harrison1], [@pone.0016111-Ghazoul4] that identify the importance of large insects such as *Apis dorsata* and sphingid moths for pollination of large-flowered dipterocarps. Significant fine-scale SGS combined with limited pollen dispersal may impose a fitness cost due to elevated inbreeding depression or incompatible matings in self-incompatible species ([Fig. 1](#pone-0016111-g001){ref-type="fig"}), and this is likely to be more substantial among smaller flowered species for the reasons given above. Such processes would lead to a genetic cost under limited pollen dispersal and is supported by evidence of a higher frequency of matings between more closely-related individuals in the smaller-flowered species *Shorea xanthophylla* than in *Parashorea tomentella*. High neighbourhood genetic relatedness has been shown to interact with density to increase the proportion of aborted fruit in the insect pollinated tropical tree *Jacaranda copaia* [@pone.0016111-Jones1], which supports our hypothesis that genetic processes contribute to the trade-off between flower size and pollination success ([Fig. 1](#pone-0016111-g001){ref-type="fig"}). ### Differential Fecundity {#s3a2} As a component of plant fitness, fecundity is represented in many models of population persistence and species coexistence [@pone.0016111-Chambers1], [@pone.0016111-Hubbell1], [@pone.0016111-Kohyama1], [@pone.0016111-Verhulst1]. However, empirical investigations of coexistence in tropical trees have focused more on plant responses to abiotic resource availability [@pone.0016111-Fine1], [@pone.0016111-Kraft1], [@pone.0016111-Russo1]. These studies do not support an unequivocal role for habitat partitioning in the maintenance of the diverse tropical tree communities observed in Borneo. While our study is limited to only a few species, it does highlight the potential importance of partitioning of the pollinator niche as an important biotic resource, and also provides the basis for postulating a series of trade-offs (depicted in [Figure 1](#pone-0016111-g001){ref-type="fig"}) linking fruit production to flower size, flower number and pollination systems within the Dipterocarpaceae. These mechanisms may well contribute to equalisation of fecundity among species, and warrant further investigation as a novel mechanism for promoting coexistence. Conservation implications {#s3b} ------------------------- Our findings have implications for the conservation of genetic and species level diversity. First, our results highlight the potential for differential vulnerability of species to inbreeding, genetic drift and pollen limitation as a result of habitat fragmentation [@pone.0016111-Ghazoul1], [@pone.0016111-Aguilar1], [@pone.0016111-Honnay1]. Larger-flowered dipterocarps may be less vulnerable to fragmentation as larger-bodied pollinators have a greater capacity to disperse between forest fragments [@pone.0016111-Ghazoul1], [@pone.0016111-Dick1], [@pone.0016111-SteffanDewenter1] and thus maintain larger effective populations of forest trees. Second, the genetic processes invoked here also have clear implications for tropical forest restoration, and specifically the collection of seeds for generating planting stock. The genetic diversity of seed collections from small flowered species may be more vulnerable to cryptic processes, such as an elevated proportion of inbred progeny, which may compromise the long-term viability of forest restoration [@pone.0016111-Kettle2]. Hence we emphasise the importance of maintaining the integrity of pollinator communities and dispersal corridors between forest fragments for tropical forest conservation. Conclusions {#s3c} ----------- Our study suggests that a trade-off exists between flower size and number that, coupled with a positive scaling between flower size and pollination success, reduces the variance in per tree fruit production between species. We have shown that species with small flowers produce them in greater numbers, but these flowers are pollinated by small-bodied insects that are less likely to disperse pollen over long distances. This gives rise to limited gene flow by pollination and, coupled with adult populations with strong spatial genetic structuring, increases the likelihood of mating between related individuals in small-flowered species. As mean flower size increases across species, flower number per tree declines, but the likelihood of receiving pollen from unrelated (more distant) conspecifics increases because larger-bodied pollinators such as sphingid moths and *Apis dorsata* become increasingly important as pollinators. Thus a trade-off embedded in the relationship between flower size and the pollination ecology of co-occurring tropical trees contributes to reducing variance in per capita fecundity among species and life-forms. Although we cannot infer that fecundity relates directly to life-time mean fitness, seed production is an important component of plant reproductive success. Inbreeding has been shown to have direct implications on differential seed mass in dipterocarps [@pone.0016111-Naito2] and germination rates and survival in other tropical tree species [@pone.0016111-Aldrich1], [@pone.0016111-Hufford1], [@pone.0016111-Hufford2]. It is therefore plausible that these processes reduce fitness inequalities among species and thereby delay competitive exclusion. This perspective suggests that partitioning of pollinators may contribute to maintaining the diversity of species in this ecologically and economically important family of tropical trees, and that the pollination component of the regeneration niche may play a more substantial role in maintaining species diversity than has been recognised previously. Materials and Methods {#s4} ===================== Study site and species {#s4a} ---------------------- The study site was the Sepilok Forest Reserve (SFR: 5°47′--5°52′ N, 117°55′--118°03′ E), which is about 24 km west of the state capital of Sandakan on the east coast of Sabah, Malaysia (SI [Fig. S1](#pone.0016111.s001){ref-type="supplementary-material"}). SFR is a "Class VI virgin jungle reserve" gazetted in 1930, managed by the Forest Research Centre Sabah for forest protection and research. SFR supports lowland dipterocarp and heath forest (described by Fox [@pone.0016111-Fox1]). Mean annual temperature falls in the range of 26.7--27.7°C and mean annual rainfall is 2929±134 mm (Malaysian Meteorological Service, unpubl. data). Flower size, flower number and fruit production {#s4b} ----------------------------------------------- We examined the trade-offs among flower size and total flower and fruit production across 12 dipterocarp species during two minor flowering events between July 2001 and November 2002 supplemented by information obtained during flowering events in 2006--2008 ([Table 3](#pone-0016111-t003){ref-type="table"}). Mean flower size of these species, measured as calyx tube diameter on an average sample of 700 flowers per species from 6 to 55 individuals, varied in the range 1.2 mm--10.2 mm ([Table 3](#pone-0016111-t003){ref-type="table"}). Flowering trees were located using binoculars along an extensive trail system, in an area of approximately 640 ha. A mean of 22 trees were sampled for each species ([Table 3](#pone-0016111-t003){ref-type="table"}). Diameter at breast height (dbh) and crown width were measured for each tree, and their positions were recorded using a GPS (Magellen 315, Thales Navigation, Santa Clara, USA). Similarly, the nearest two flowering neighbours for each tree sampled were located and their positions recorded as above. Phylogenetically independent contrasts (PICs) were used to examine trade-offs between flower size and total flower production, pollination success, the mean number of pollen tubes per pollinated flower or maximum total fruit production to take account of phylogenetic inertia within closely related species [@pone.0016111-Harvey1]. We applied the relationships determined from a recent dipterocarp phylogeny using *PgiC* gene sequence [@pone.0016111-Kamiya1] which includes seven of the 12 study species. The PIC analysis was conducted using the computer program Compare 4.6b on the transformed variables [@pone.0016111-Martin1]. Following PIC analyses, the Y-variable contrasts were regressed on the X-variable contrasts as outlined in [@pone.0016111-Purvis1]. ::: {#pone-0016111-t003 .table-wrap} 10.1371/journal.pone.0016111.t003 Table 3 ::: {.caption} ###### Summary statistics of pollen dispersal from paternity analysis in dipterocarp species with a range of flower size, including the number of genotyped offspring (n); progeny type used in the paternity analysis: mature embryos directly collected from mother (ME); forest floor seedlings (FS), or Saplings (Sp); seedlings germinated from seeds collected below putative mothers (GS); proportion of total embryos assigned at 80% confidence level (% *asgn* 80%); mean pollen dispersal distance across all assigned embryos (*MPDd*), and sample area over which paternity analysis was conducted (PN area); N/A  =  not available. ::: {#pone-0016111-t003-3} Species Calyx tubeWidth (mm) n Progeny type \% *asgn* 80% *MPDd* (m) S.E PN area (ha) source ------------------- ---------------------- ----- -------------- --------------- ------------ ------- -------------- ------------------------- *S. xanthophylla* 1.2 456 ME 68 119 8.4 100 This study *S. lumutensis* 2.2 182 GS 63 175 N/A 8 [@pone.0016111-Lee1] *P. tomentella* 4.2 408 ME 26 176 14.67 100 This study *N. heimii* 10.0 248 FS & Sp 37 191 N/A 42 [@pone.0016111-Konuma1] *D. tempehes* 10.0 335 FS 88 207 N/A 70 [@pone.0016111-Kenta1] ::: Four 1 m^2^ mesh traps raised 1 m above the ground were installed under each tree to estimate flower production. The traps were placed approximately 5 m from the trunk, and at 90° intervals from a random azimuth. All traps were installed prior to the commencement of flower fall and their contents were collected approximately every three days until the end of fruit-fall. The samples were air dried and sorted into fractions representing buds, flowers and fruit, and the fractions were counted and weighed. Flower counts were based on calyxes, which we presume to be subjected to less lateral dispersal by wind than corollas. Total flower production was estimated by multiplying the total number of buds, flowers and fruit captured by the traps (expressed per m^2^) by the projected area of the tree crown. The flowering events we sampled for this paper were not general flowering events sensu [@pone.0016111-Sakai4], and a high proportion of flowering individuals produced only a small number of flowers on relatively few branches. Such flowering events are thought to contribute very little to population-level fecundity [@pone.0016111-Maycock1]. Therefore we estimated mean maximum flower and fruit production from the upper quartile of values per species (n = 2--14 individuals per species). Because basal area alone is insufficient to account for differences in architecture and allometry among species and to account for effects of canopy exposure on flower production, we present values of maximum flower production scaled to tree size:where MHE/MSE is a scaling factor for height calculated by dividing the maximum height of emergents at the study site (MHE: 65 m) by the maximum height of the study species (MHS). Basal area was determined by direct measurement for each tree sampled. Maximum height of the emergent trees in the SFR and maximum height of each species were estimated to the nearest 5 m using a measuring line dropped from the canopy. Total fruit production was estimated from the mean number of fruit captured per trap multiplied by the projected area of the tree crown. Pollination success {#s4c} ------------------- We examined pollen tube growth as a measure of pollination success in 11 of the same 12 dipterocarp species using flower samples collected during 2001--2002 (*H. beccariana*, *S. beccariana*, *S. johorensis*, *S. leprosula*, *S. macroptera*, *S. multiflora*, *S. parvifolia*, *S. smithiana* and *S. xanthophylla*,) or 2006--2008 (*D. grandiflorus* and *P tomentella*) ([Table 3](#pone-0016111-t003){ref-type="table"}). Flower samples were obtained using mixed rope-climbing techniques to gain access to the canopy. Inflorescences were collected directly from each tree at five randomly selected locations within the crown. Flowers that had recently shed their corollas were removed from the inflorescences. We collected a minimum of 500 flowers from each tree for the *Shorea* spp and *H. beccariana*, while the low flower densities of *D. grandiflorus* and *P. tomentella* limited the sample to approximately 50 and 100 flowers per tree, respectively. All flowers were stored in formalin-acetic-alcohol (FAA). Mean calyx diameters for each tree were determined from direct measurement on 25 randomly selected flowers. The styles of the flowers were stained with 1 percent aniline blue solution and the number of pollen tubes in each were determined using fluorescence microscopy as outlined in Ghazoul *et al.*, [@pone.0016111-Ghazoul2]. We examined pollen tubes within 25200 flowers sampled randomly from the flower collections taken from a total of 638 individuals of the 11 species. Pollinator body size {#s4d} -------------------- Mesh bags were used to exclude visitors from inflorescences prior to anthesis on 2--6 trees from each of nine of the 12 study species that represented an order-of-magnitude range in flower size and flowered between May 2006 and May 2008 ([Table 1](#pone-0016111-t001){ref-type="table"}). Ten replicate inflorescences per tree were used for each of the following treatments: (a) open pollination (no mesh); (b) exclusion of large pollinators (5 mm mesh); (c) exclusion of medium-sized and large pollinators (2 mm mesh); and (d) exclusion of all pollinators except thrips (0.2 mm mesh). Control large mesh bags containing fly trap paper confirmed that large mesh bags did not inhibit visitation by smaller pollinators. The bags were removed at the end of anthesis. Fruit initiation and development were monitored every five days throughout the first month, and then every 10 days until fruit maturity. The effect of mesh bag treatments is expressed as percent pollination success relative to the open pollination treatment one month after anthesis. Pollen dispersal distances {#s4e} -------------------------- Pollen dispersal distance was estimated using paternity analysis based on samples of mature fruit collected from known mother trees of *Shorea xanthophylla* (mean calyx diameter 1.2 mm) and *Parashorea tomentella* (mean calyx diameter 4.6 mm). These species were selected because this contrast in flower size was predicted to generate differences in pollen dispersal distance based on our sampling of their flower visitor communities (C.R. Maycock *et al.*, unpubl. data), and because their densities of flowering individuals maximised sampling effort for the available resources. We applied identical sampling strategies in both species, combining the genotypes of mother and progeny with genotypes from all potential pollen donors in mapped populations within 500 m of each mother trees. Because flowering tree density can have a high variance, both within and among species, our sampling design enables us to explicitly control for this. ### DNA sampling, extraction and genotyping {#s4e1} All sampling was carried out between April 2006 and September 2008. 20--35 mature fruits were sampled from each of 20 and 17 mother trees of *Shorea xanthophylla* and *Parashorea tomentella* respectively. Fresh fruits were dissected and embryos removed and dried in Sigma™ silica gel prior to DNA extraction. In addition to sampling inner bark from each mother tree, inner bark was sampled from all flowering con-specifics within a 500 m radius of each mother and the location of all adult trees recorded with a GPS. For adult trees, fresh inner bark (cambium) was collected using a 2 cm diameter leather punch and a hammer following Colpaert *et al.* [@pone.0016111-Colpaert1]. The fresh samples were placed in labelled tea bags and dried in Sigma™ silica gel within 12 hours of sampling, and stored under the same conditions prior to DNA extraction. Total genomic DNA was extracted from approx. 0.03 g of silica dried material (inner bark, embryo or leaf, depending on sample type) and ground to a fine powder using a Qiagen Mixer-mill™ with Retsch © stainless steel 'cone balls' and the Qiagen DNAeasy™ 96-well-plate plant extraction system. For the paternity analysis the genotype of each mother tree, candidate father and embryo was determined at nine polymorphic, genomic, nuclear microsatellite loci. Primers are summarised in Supplementary Information [Table S3](#pone.0016111.s004){ref-type="supplementary-material"}. The forward primer of each locus was 5- prime end labelled with one of four florescent-labelled dyes (FAM, NED, PET or VIC) to enable multiplexing of multiple loci during fragment analysis. All PCR products were quantified using an ABI 3730xl DNA Analyser (Applied Biosystems) and sample genotypes scored using Genemapper v.4.0™ software (Applied Biosystems). To establish whether data from different microsatellite loci were independent, genotypic linkage disequilibrium was tested between all pairs of loci within each species adult sample using GENEPOP 4.0 [@pone.0016111-Rousset1] with the following settings; dememorization: 1000, batch: 100, iteration per batch: 5000. No evidence for significant linkage disequilibrium was detected. ### Direct estimates of pollen dispersal {#s4e2} Mother trees were sampled based upon the abundance of fruits and accessibility. Using multilocus genotypes (9 loci, see [Table S3](#pone.0016111.s004){ref-type="supplementary-material"} Supporting information) of progeny and known mother trees we applied a maximum likelihood exclusion analysis in CERVUS 3.0, to exclude candidate fathers [@pone.0016111-Kalinowski1], [@pone.0016111-Marshall1]. Candidate fathers were defined as all flowering conspecifics within a 500 m radius of each mother tree. Pollen dispersal distance was deduced for each embryo based on the position of the mother tree and the assigned pollen parent (father) within the 500 m radius. Simulations of paternity were run using the allele frequencies of all adult reproductive trees and the following settings: 10000 cycles; minimum number of loci typed 5, mothers as known parents, test for selfing (*i.e.*, mothers also included as candidate fathers); the flowering conspecifics within 500 m of each mother were set as the candidate fathers for progeny from that mother only; 1 percent for proportion of loci mistyped, and 87 percent for proportion of loci genotyped. The proportion of candidate fathers sampled was set at 80 percent, which was considered a conservative estimate based upon the observed frequency of flowering conspecifics across the entire SFR. Assignment was based upon the 80% confidence level of the critical LOD score. Based upon paternity assignment, 68 percent of progeny (310 of 456) from 20 mothers trees of *S. xanthophylla* and 26 percent of progeny (106 of 408) from 17 mother trees of *P. tomentella* could be assigned with an 80% confidence within 500 m of the mother tree. For all progeny that could not be assigned at this level of confidence the pollen donors were assumed to be beyond a 500 m radius of each mother tree. This assumption would be supported by evidence that the gene pool of the genotyped progeny differed significantly from that of the pollen donor pool within the 500 m radius. To test this we examined the frequency of rare alleles (abundance \<0.01) in the embryo cohorts of each species. In *P. tomentella* 92 percent of the rare alleles (\<0.01) are observed only in the embryos, not in the mothers or neighbouring adults, thus must have come from outside the 500 m radius we sampled. In contrast in *S. xanthophylla* only 49 percent of rare alleles were unique to the embryo cohort. ### Indirect estimate of pollen dispersal {#s4e3} Pollen dispersal was also evaluated using the TWOGENER [@pone.0016111-Austerlitz1], [@pone.0016111-Smouse1] approach with POLDISP [@pone.0016111-RobledoArnuncio1]. TWOGENER estimates the heterogeneity of the pollen gene pool sampled within a set of maternal trees (Φ *~FT~*) using AMOVA [@pone.0016111-Excoffier1]. This has the added advantage that it provides an estimate of contemporary gene flow over the scale of the landscape which is less sensitive to flowering tree density and complementary to the more localised estimate of the paternity analysis. The mean effective pollen dispersal distance (δ) was calculated assuming the following dispersal models; normal, bivariate normal and anisotropic normal, following [@pone.0016111-Smouse1], [@pone.0016111-Cloutier1], [@pone.0016111-deLacerda1] in *S. xanthophylla* and *P. tomentella*. The density of adult trees was calculated using our detailed spatial distribution data of our focal species within SFR ([Table 1](#pone-0016111-t001){ref-type="table"}). The effective number of pollen donors *N~ep~* and effective pollination neighbourhood area (*Â~ep~*) were calculated as follows: *N~ep~* = *1/2Φ~ft~* and *Â~e~ = N~ep~*/*d~e~* where *d~e~* is the effective density of trees. We used a range of values of *d~e~* including those derived from the TWOGENER analysis. Because the effective density of flowering adults may vary markedly from year to year we calculated *N~ep~* assuming a range of densities based upon the census density of reproductive adults (*d*) and *d*/2, *d*/4 and *d*/10 following the approach of [@pone.0016111-Born1] after [@pone.0016111-Smouse1]. Inbreeding of progeny {#s4f} --------------------- The relatedness between mothers and putative fathers, derived from the paternity analysis, was estimated for each of the assigned progeny using SPAGEDI 1.2 g [@pone.0016111-Hardy2]. Here we calculated the individual and mean kinship coefficient *F* [@pone.0016111-Loiselle1] for each parent pair assigned by CERVUS [@pone.0016111-Kalinowski1] within *S. xanthophylla* and *P. tomentella* (n = 229 and 106 respectively). Mean kinship coefficients *F* have been used to express the relatedness between mating individuals in ants [@pone.0016111-Hardy3] and are useful as they are directly comparable to measures of spatial genetic structure among species. We supplement the measures of relatedness between mating pairs with estimates of the number of pollen donors per seed crop (Nep) and multilocus outcrossing estimates using the program MLTR v 2.4 [@pone.0016111-Ritland1]. Statistical analysis {#s4g} -------------------- Linear models were used to investigate trade-offs among species\' between log-transformed flower size and: (1) log-transformed maximum flower production; (2) arcsine-transformed percentage of flowers pollinated; (3) log-transformed mean number of pollen tubes per pollinated flower; and (4) log-transformed maximum fruit production. We included life-form (canopy or emergent following Ashton [@pone.0016111-Ashton2]), as an additional independent variable in all models. For analyses of pollination success and fruit production (2--4) we also included median values of the mean distance to the two nearest flowering trees (for pollination success) or fruiting trees (for fruit production) as an independent variable to account for possible effects of differences between species in the density of reproductive trees. We compared models containing all possible combinations of one or more explanatory variables and their interactions and used AIC values to select the most likely models. Models with AIC values that differed by less than two were judged equally valid [@pone.0016111-Burnham1] but F statistics and probability values reported in this paper are taken from the model with the lowest AIC value. The effects of pollinator exclusion on pollination success were analysed using a generalised linear model with log-transformed calyx diameter and mesh bag treatment as fixed explanatory variables. The non-significant interaction term was removed from the model. Pollination success was expressed as the proportion of fruits within each treatment surviving to one month after the end of anthesis, relative to open pollination. To test the null hypothesis of no difference in the frequency of short distance mating we conducted a chi-squared test on counts of pollen dispersal within distance classes \<50 m, 50--99.9 m, 100--199.9 m and 200--500 m in *S. xanthophylla* and *P. tomentella.* Average pollen dispersal distances over all assigned embryos were compared between *S. xanthophylla* and *P. tomentella* using a Student *t*-test (arithmetic mean) and Wilcox test (median). To determine the significance of inbreeding of progeny as a function of flower size we used a Student *t*-test to test the null hypothesis of no difference in mean kinship coefficient. All statistics were conducted using R (R Development Core Team 2009). Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Location of Sepilok Forest Reserve, boundary demarked by orange line together with distribution of flowering trees of *Parashorea tomentella* and *Shorea xanthophylla* used for paternity analysis in. The two grey ellipses indicate the subset of mother trees from *S. xanthophylla* which had comparable local density of flowering conspecifics to *P. tomentella*. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### (**a**) AIC and Δ~AIC~ values from four candidate models of log-transformed absolute (unscaled) and scaled flower production for 11 dipterocarp species at Sepilok Forest Reserve, Sabah. The most likely models are shown in bold. (**b**) AIC and Δ~AIC~ values from 18 candidate models of arcsine square root -- transformed proportion of flowers pollinated and log-transformed mean number of pollen tubes per pollinated flower for 11 dipterocarp species at Sepilok Forest Reserve, Sabah. The most likely models are shown in bold. logFS, log-transformed flower size; LF, life-form; NND, median values of the mean distance to the two nearest flowering trees. (**c**) AIC and Δ~AIC~ values from 18 candidate models of log-transformed absolute (unscaled) and scaled fruit production for 11 dipterocarp species at Sepilok Forest Reserve, Sabah. The most likely models are shown in bold. logFS, log-transformed flower size; LF, life-form; NND, median values of the mean distance to the two nearest flowering trees. (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Mating system statistics for progeny of *Shorea xanthophylla* and *Parashorea tomentella* based upon 9 microsatellites loci. Number of progeny genotypes (*N*); multilocus outcrossing rate (*t~m~*); single locus outcrossing rate (*t~s~*); biparental inbreeding as defined by the difference between multilocus and single locus outcrossing rates (*t~m~ - t~s~*); Parental inbreeding coefficient. Values in parentheses are standard error (SE) based upon 100 bootstraps. (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### Summary of the 11 microsatellite primers used for paternity analysis and quantification of relatedness between assigned parents in two dipterocarp species *Shorea xanthophylla* and *Parashorea tomentella*. Number of alleles (Na); observed heterozygosity (*H~obs~*); expected heterozygosity (*H* ~e~); paternity non-exclusion probability at each locus (*N-P~E~*) and total exclusion probability over all loci (*P~E~*) given known mother. *^a^* Redesigned primers based on published primers. *^b^* Newly developed microsatellite primers. ^c^ Published primers. (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank the governments of Malaysia and the state of Sabah for permission to conduct research; the Sabah Forest Department for logistic support; Jeisin bin Jumian and Mike Charkow for field assistance; Will Goodall-Copestake and Viola Huschauer for laboratory assistance and Soon Leong Lee for advice on primers. Fragment analysis was conducted using the sequencing facility of the Genetic Diversity Centre of ETH Zurich. The manuscript benefited from the comments of the academic editor Jerome Chave and anonymous reviewers. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Financial support was provided by the British Ecological Society (<http://www.britishecologicalsociety.org/>) "Overseas Research Programme" and the Natural Environment Research Council (<http://www.nerc.ac.uk/>) grant numbers NER/A/S/2001/00835 and NE/D003822/1. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: CK CM JG PH DB. Performed the experiments: CK CM EK RS. Analyzed the data: CK CM DB. Contributed reagents/materials/analysis tools: CK CM EK RS. Wrote the paper: CK CM JG PH DB.

PubMed Central

2024-06-05T04:04:19.078466

2011-2-1

PMC3052256

Introduction {#s1} ============ The environmental conditions in which plants live vary constantly, during the day, with the seasons and as a consequence of climate. Plants are sessile organisms that are constrained to complete their live-cycle in one place and therefore must protect themselves against adverse conditions: their growth depends on their responses to the environment. Plants perceive variations in the environment and react, in accordance to the nature and the intensity of the stress (reviewed in [@pone.0016645-Lefebvre1]). In response to abiotic stress, many genes are either down-regulated, mostly genes involved in growth, or activated at the transcriptional level [@pone.0016645-Desikan1], [@pone.0016645-Seki1], [@pone.0016645-Seki2], [@pone.0016645-Kreps1], [@pone.0016645-Kilian1]. Up-regulated genes can propagate the stress signal (transcription factors, hormone synthesis...), be involved in homeostasis (osmoticum synthesis, ions and water transport, molecule protection by chaperones...) or damage limitation (detoxification, antioxidants...) (reviewed in [@pone.0016645-Wang1], [@pone.0016645-Bray1]). As a result, there is a constant readjustment of the balance between growth arrest in order to protect and save the organism and growth maintenance at the risk of exhausting water or nutrient supplies (e.g. [@pone.0016645-Hummel1]). The regulation of stomata aperture according to water availability is a good illustration of this equilibrium. Stomata are the pores through which plants lose water through evapo-transpiration and acquire CO~2~ for carbon fixation by photosynthesis: closing stomata economizes water, but reduces growth as CO~2~ influx is limited. Water depletion induces stomata closure, through an increase of the stress hormone ABA [@pone.0016645-Nilson1], [@pone.0016645-Wasilewska1]. The *ESK1* gene was initially identified by map-based cloning of the *Arabidopsis thaliana esk1-1* mutation obtained using an *in vitro* screen for constitutive freezing tolerance [@pone.0016645-Xin1]. Subsequently, it was shown that when *esk1* plants were grown in soil, they were more tolerant to freezing than wild type after a period of acclimation [@pone.0016645-ReyesDiaz1], [@pone.0016645-BouchabkeCoussa1]. *esk1* mutant phenotypes are those of a stressed plant: reduced stature and relative water content (RWC) compared to wild type [@pone.0016645-Xin1], [@pone.0016645-BouchabkeCoussa1], [@pone.0016645-Lugan1], it accumulates high levels of stress metabolites that are thought to maintain the internal osmotic status, such as the osmoticum proline [@pone.0016645-Xin1], [@pone.0016645-Lugan1] and its transcriptome resembles that of drought-stressed wild type plants [@pone.0016645-BouchabkeCoussa1]. Furthermore, the evapo-transpiration of *esk1* mutants is severely reduced and their water use efficiency (WUE) i.e. the amount of carbon incorporated in biomass per water used, is higher than wild type. The stressed phenotypes of *esk1* led us to examine whether *ESK1* was implicated in the action of the stress hormone ABA. We have examined ABA production in *esk1* mutants and the phenotypes of double mutants affected in ABA responses and *esk1*. Then, we addressed the origin of the very low evapo-transpiration of *esk1*. We observed that cross sections of *esk1* vascular tissue had an "*[ir]{.underline}regular [x]{.underline}ylem*" (*irx*) phenotype. Analysis of the xylem cell wall in *esk1* mutants highlighted significant modifications in chemical composition compared to the wild type. Furthermore, we assessed *esk1* root-hydraulic conductivity. Taken together, our results indicate that ESK1 is necessary for the formation of the functional xylem required for water transport to aerial plant tissues. Results {#s2} ======= ABA levels are altered in *esk1* plants {#s2a} --------------------------------------- *esk1* mutants were previously described as freezing [@pone.0016645-Xin1], cold and salt tolerant [@pone.0016645-BouchabkeCoussa1], [@pone.0016645-Lugan1]. A common factor in these stress responses is the accumulation of the phytohormone ABA [@pone.0016645-Thomashow1]. In order to determine whether ABA was implicated in inducing the stressed *esk1* state, ABA contents were measured in rosettes of wild type and two *esk1* mutants (*esk1-4* and *esk1-5*) submitted or not to mild drought ([Figure 1](#pone-0016645-g001){ref-type="fig"}) and compared to the ABA-deficient mutant *aba3-1*, affected in the sulfuration of the molybdenum cofactor required for the last steps of ABA biosynthesis [@pone.0016645-LeonKloosterziel1], [@pone.0016645-Schwartz1]. ABA measurements from the two *esk1* mutants were similar as expected for null mutant, with or without drought treatment ([Table S1](#pone.0016645.s003){ref-type="supplementary-material"}, LSD tests). ::: {#pone-0016645-g001 .fig} 10.1371/journal.pone.0016645.g001 Figure 1 ::: {.caption} ###### Basal ABA levels are high in *esk1* mutants. Box-plot of \[ABA\] determined for wild type Col-0 (wt), *aba3-1*, *esk1-4* and *esk1-5* under control and mild drought conditions. The \[ABA\] is reported as pg/mg of dry weight. Each box-plot was produced from results for 11 or 12 measurements. Red \* indicates significant difference with wild type for the same treatment and black \* indicates significant genotype x treatment interaction (*p*-value\<0.05) for each pair of wild type and the other three genotypes. :::  ::: In control conditions, ABA contents were significantly higher in *esk1* mutants than either wild type or *aba3-1*, the latter being due to its defect in ABA biosynthesis [@pone.0016645-LeonKloosterziel1]. Increased in ABA production is associated with stress and is in accord with *esk1* mutants being stressed *per se.* As previously described, ABA contents were significantly higher in drought conditions for wild type and *aba3-1*, however, ABA levels in *esk1* mutants were the same as those in non-stressed conditions ([Table S1](#pone.0016645.s003){ref-type="supplementary-material"}, 1D ANOVA). The slight increase in ABA levels in *aba3-1* plants in these conditions is probably because this is not a null allele [@pone.0016645-Xiong1]. We found significant genotype x treatment interactions for the wild type and the other three genotypes ([Figure 1](#pone-0016645-g001){ref-type="fig"}, black stars and [Table S1](#pone.0016645.s003){ref-type="supplementary-material"}, 2D ANOVA), but no interactions were significant when analyzing *esk1* and *aba3-1* ([Table S1](#pone.0016645.s003){ref-type="supplementary-material"}, 2D ANOVA). These results demonstrate that *esk1* mutants respond differently to drought stress than wild type. ESK1, ABA3 and likely OST1 act in independent pathways {#s2b} ------------------------------------------------------ To determine whether the higher ABA levels observed in *esk1* mutants were a direct or indirect consequence of ESK1 function, double mutants were generated with either *aba3-1* or the ABA signaling mutant *ost1/snrk2e.* OST1 is a SnRK2 protein kinase that acts in an early step of the ABA signaling pathway in guard cells [@pone.0016645-Mustilli1]. Water loss in double mutants was examined either indirectly as leaf temperature differences visualized using infrared thermal imaging of detached leaves, or by rapid dehydration of rosettes ([Figures 2](#pone-0016645-g002){ref-type="fig"} and [3](#pone-0016645-g003){ref-type="fig"}). ::: {#pone-0016645-g002 .fig} 10.1371/journal.pone.0016645.g002 Figure 2 ::: {.caption} ###### *esk1-5 ost1* double mutant phenotypes. A. Two representative plants of each genotype grown in standard non-stressed conditions: wild type (wt), *ost1*, *esk1-5* and *esk1-5 ost1* double mutant. B. False color infrared image of detached leaves from 4 week-old plants, 2 leaves from 2 plants were observed for each genotype. C. Rapid dehydration of detached rosettes from 4 week-old plants. Water loss was determined as % of initial fresh weight. Error bars represent standard error values (n = 4). :::  ::: ::: {#pone-0016645-g003 .fig} 10.1371/journal.pone.0016645.g003 Figure 3 ::: {.caption} ###### *esk1 aba3-1* double mutants phenotypes. A. Two representative plants of each genotype grown in standard non-stressed conditions: wild type (wt), *aba3-1*, *esk1-4*, *esk1-5*, *esk1-4 aba3-1* and *esk1-5 aba3-1* double mutants. B. False color infrared image of detached leaves from 4 week-old plants, 2 leaves from 2 plants were observed for each genotype. C. Rapid dehydration of detached rosettes from 4 week-old plants. Water loss was determined as % of initial fresh weight. Error bars represent standard error values (n = 4). :::  ::: The *aba3-1* and *ost1/snrk2e* single mutants displayed lower leaf temperatures than wild type due to their increased water loss upon dehydration, as previously reported [@pone.0016645-LeonKloosterziel1], [@pone.0016645-Mustilli1], [@pone.0016645-Merlot1] ([Figures 2B](#pone-0016645-g002){ref-type="fig"} and [3B](#pone-0016645-g003){ref-type="fig"}). For the *aba3-1* mutant, this was due to its impaired ability to synthesize ABA, whereas the *ost1* mutant is affected in the ABA signaling pathway that induces stomata closure, downstream of ABA production. In contrast, *esk1* leaves were hotter and water loss was reduced compared to wild type, in accordance with the reduced evapo-transpiration reported by Bouchabke-Coussa and co-workers [@pone.0016645-BouchabkeCoussa1] ([Figures 2B, C](#pone-0016645-g002){ref-type="fig"} and [3B, C](#pone-0016645-g003){ref-type="fig"}). Both *esk1-5 ost1* and *esk1-5 aba3-1* had leaf temperatures that were intermediate of those for the parental genotypes, as expected if the mutations affected independent pathways ([Figures 2B](#pone-0016645-g002){ref-type="fig"} and [3B](#pone-0016645-g003){ref-type="fig"}). Nonetheless, intermediate leaf temperatures could also be obtained if the two genes act in partially overlapping pathways that modulate stomata opening. Similar results were obtained in water loss measurements on rapid dehydration of detached rosettes ([Figures 2C](#pone-0016645-g002){ref-type="fig"} and [3C](#pone-0016645-g003){ref-type="fig"}). The *esk1* mutant plants lost only 30% of their weight as water, whereas *ost1* and *aba3-1* lost over 70%. Water loss from the *esk1-5 ost1*, *esk1-5 aba3-1* and *esk1-4 aba3-1* double mutants was always between these extremes. As shown in [Figure 2A](#pone-0016645-g002){ref-type="fig"}, *esk1-5 ost1* and *esk1-5* mutant plants were the same size. As the *ost1* mutation does not affect plant stature, despite having open stomata, this indicates that the factor influencing plant size is not the degree of stomata opening. In contrast, both *esk1* and *aba3-1* mutants display dwarf phenotypes and the double mutant plants were even smaller ([Figure 3A](#pone-0016645-g003){ref-type="fig"}), indicating that the growth limitations caused by these mutations are additive. Thus, *ABA3* and *ESK1* are independent in the determination of plant size and act in different pathways. The small size of *esk1* is not, therefore, directly due to the high levels of ABA accumulation. Taken together the double mutant phenotypes demonstrate that ESK1 does not intervene in growth regulation and water loss through ABA and the high ABA levels in mutants is an indirect consequence of *esk1* mutation. Hydraulic conductivity is reduced from the *esk1* root apparatus {#s2c} ---------------------------------------------------------------- The major driving force of plant water uptake is transpiration, which is controlled by stomata aperture. Entry of water into the plant is also regulated by the hydraulic conductance of the root apparatus and can potentially limit water availability in the shoots [@pone.0016645-Maurel1]. Given the reduced water loss, high leaf temperature and low RWC of *esk1* mutants [@pone.0016645-BouchabkeCoussa1] ([Figures 2](#pone-0016645-g002){ref-type="fig"} and [3](#pone-0016645-g003){ref-type="fig"}), we hypothesized that the overall water flux supplied by the xylem sap is reduced in *esk1* plants. In order to test this hypothesis, *esk1* root hydraulic conductance was measured and compared to wild type ([Figure 4](#pone-0016645-g004){ref-type="fig"}). ::: {#pone-0016645-g004 .fig} 10.1371/journal.pone.0016645.g004 Figure 4 ::: {.caption} ###### Hydraulic parameters of the root apparatus are modified in *esk1.* A. Root hydraulic conductance *L*o is expressed as ml of sap exudate per hour and per unit of exerted pressure. B. Root/rosette dry weight ratio. C. Root hydraulic conductivity is expressed as *L*o per unit of root mass. Wild type (wt). Black \* indicates significant difference with wt (*p*-value\<0.05). :::  ::: The root hydraulic conductance, or *L*o, represents the overall capacity of the root apparatus to transport water from the soil to the shoot. The root hydraulic conductivity, or *L*p~r~, is the conductance of a root surface unit: *L*o =  *L*p~r~\*S, with S related to the surface of exchange between the root and its substrate. *L*o and *L*p~r~ have been measured for wild type and two *esk1* mutants, as well as the root/rosette dry weight ratio ([Figure 4](#pone-0016645-g004){ref-type="fig"}). Despite the small size of the *esk1* lines, the dry matter allocation between the root apparatus and the rosette was similar to that of wild type. The *L*o of *esk1* lines was exceptionally low: 6% and 5% of wild type for *esk1-1* and *esk1-5* respectively. Part of this low root conductance can be explained by the small biomass of the root apparatus. Nonetheless, the *L*p~r~ values for the *esk1* lines were also extremely low, indicating that the mutant root system produced a higher resistance to water transport from the soil to the hypocotyl. Thus, water transport is impaired in *esk1* mutants in comparison to wild type and probably induces the low evapo-transpiration and the reduced RWC observed in the mutants. The *ESK1* promoter specifically drives expression in vascular tissue {#s2d} --------------------------------------------------------------------- To determine the site of *ESK1* gene function, expression analyses were carried out using a 1 kb *ESK1* promoter region fused to the *GUS* reporter gene. Seven independent insertion lines were analyzed and similar results for seedlings and mature plants were obtained: in all cases, expression was restricted to the vascular tissue and was initially observed a few days after germination, just after the appearance of the first differentiated vascular tissues, in the radicle and cotyledons ([Figure 5A](#pone-0016645-g005){ref-type="fig"}1, A3). In roots, GUS staining was absent from the meristematic and elongation zone and was only apparent in the differentiation zone ([Figure 5A](#pone-0016645-g005){ref-type="fig"}2). At ten days after germination, expression profile is still restrained to vascular tissues of all organs ([Figure 5A](#pone-0016645-g005){ref-type="fig"}3--A5). ::: {#pone-0016645-g005 .fig} 10.1371/journal.pone.0016645.g005 Figure 5 ::: {.caption} ###### Histochemical analysis of *GUS* reporter gene expression from the *ESK1* promoter. A. Expression in whole plantlets 4 days after germination (A1, A2) and 10 days after germination (A3, A4, A5) with magnification of root (A4) and leaf (A5). B. Transverse section of hypocotyls, before (B1, B2) and after floral transition (B3, B4). C. Transverse section of stem (C1, C2). Double-headed arrow indicates the position of cross-section in B2. Red arrowheads indicate interfascicular fibers (B3) or sclerenchyma (C1, C2) and black arrowheads indicate newly-formed xylem cell (B1, B2, B4, C1, C2). X, xylem; P, phloem. :::  ::: More detailed characterization of *ESK1-*promoter driven expression within the vascular tissue was performed on three independent homozygous transgenic lines, using sections of hypocotyls and stems. As shown in [Figure 5](#pone-0016645-g005){ref-type="fig"} (B1, B2, C1), GUS staining was restricted to the cells undergoing xylem vessel differentiation, in both hypocotyls and stems. Furthermore, *ESK1* also appeared to be expressed in the interfascicular parenchyma in the stem ([Figure 5C](#pone-0016645-g005){ref-type="fig"}1, C2). After floral induction, GUS staining was also observed in the fibers of the hypocotyls as well as xylem vessels ([Figure 5B](#pone-0016645-g005){ref-type="fig"}3, B4). These results reveal that the *ESK1* gene is specifically expressed in tissues undergoing secondary cell wall deposition. *esk1* mutants show altered vascular apparatus morphology {#s2e} --------------------------------------------------------- As *ESK1* expression was specific to vascular tissue and *esk1* mutants were defective in water transport, we conducted a detailed analysis of vascular tissue structure in different *esk1* mutants. Similar results were obtained with the three mutants studied and results obtained with *esk1-1* are presented. Transverse sections were performed on hypocotyls, roots and stems of 6 week-old plants grown on soil without water deficit ([Figure 6](#pone-0016645-g006){ref-type="fig"}). Xylem in vascular bundles of *esk1* roots and hypocotyls appeared collapsed ([Figure 6A](#pone-0016645-g006){ref-type="fig"}1, B). In hypocotyls, fibers also seem to be affected: cells appeared angular and the ring of fibers that surrounded xylem vessels seemed less regular than in wild type ([Figure 6A](#pone-0016645-g006){ref-type="fig"}2). This phenotype was reminiscent of the *irx* mutants that have collapsed xylem [@pone.0016645-Turner1]. When stems were only a few centimeters high, their xylem vessels seemed unaffected ([Figure S1](#pone.0016645.s001){ref-type="supplementary-material"}). However, at later stages, when they developed lateral branches, irregularly shaped cells and collapsed xylem elements were observed in stems at both the base ([Figure 6C](#pone-0016645-g006){ref-type="fig"}1) and mid-height level ([Figure 6C](#pone-0016645-g006){ref-type="fig"}2). ::: {#pone-0016645-g006 .fig} 10.1371/journal.pone.0016645.g006 Figure 6 ::: {.caption} ###### Comparison of vascular tissue of *esk1-1* and wild type, in hypocotyls, roots and stems. Vascular tissue is shown for *esk1-1* on the left hand side and for wild type (wt) on the right hand side. A. Transverse section of hypocotyl observed under UV light. A1, fibers (fib); A2, xylem cells (X). B. Transverse section of root observed under UV light. C. Transverse section of the stem base (C1) and at mid-height (C2) observed under UV light. Arrowheads indicate collapsed xylem. :::  ::: In view of this phenotype, we hypothesized that the very low *esk1* root hydraulic conductivity was due to restricted water transport through collapsed vessels. This is defined by Poiseuille\'s law, in which the major factor affecting the hydraulic conductance of a tube is its diameter. The theoretical conductivity of root tracheary elements (TE) was calculated for *esk1* mutant and wild type, based on maximal diameter measurements obtained from transverse sections of root. As shown in [Table 1](#pone-0016645-t001){ref-type="table"}, the flux calculated for *esk1* mutants is approximately five-fold lower than that of wild type. This establishes a direct consequence of the *irx* phenotype observed for *esk1* mutants on hydraulic conductivity through the root apparatus. ::: {#pone-0016645-t001 .table-wrap} 10.1371/journal.pone.0016645.t001 Table 1 ::: {.caption} ###### Low *esk1* root hydraulic conductivity is mainly explained by a reduction in tracheary element diameter. ::: {#pone-0016645-t001-1} wt (n = 19) *esk1-1* (n = 18) *esk1-5* (n = 19) ------------------ ------------- ------------------- ------------------- Mean radius 8.14 5.08 5.93 Standard error 0.14 0.27 0.16 Max radius 10.16 7.00 6.81 Max conductivity 4.18 0.94 0.84 \% of wt 100 22.6 20.2 The 4 to 5 larger xylem vessels of 6 week-old plants of *esk1* mutant alleles and wild type (wt) sections were analyzed for the maximal circular radius. The maximal conductivity (g) of the vessels was calculated according to Poiseuille\'s law. R = (8.η)/(π.r^4^) where g = 1/R (m^2^/s/MPa), η is the water viscosity coefficient (1E^−9^ MPa/s) and r is the maximal radius (µm). Max conductivity: maximal hydraulic conductivity calculated. Mean radius: mean of all radius measured (µm). Max radius: maximal radius measured (µm). ::: The chemical composition of *esk1* xylem is different from wild type {#s2f} -------------------------------------------------------------------- The chemical composition of the vascular tissue of hypocotyls at two developmental stages and the stem base was examined using FTIR spectral analysis performed on the xylem of transverse sections. Two representative *esk1* null mutants were analyzed and a Student\'s *t*-test was performed, corresponding to the differences between wild type and *esk1* spectra. For graphical representation, *t*-values were plotted against wave numbers to determine the significance of the differences at each wavelength ([Figure 7B](#pone-0016645-g007){ref-type="fig"} and [8B](#pone-0016645-g008){ref-type="fig"}). ::: {#pone-0016645-g007 .fig} 10.1371/journal.pone.0016645.g007 Figure 7 ::: {.caption} ###### The structure and composition of *esk1* xylem is modified in hypocotyls. A. Transverse section of hypocotyls from 3 week-old (3 wo) and 6 week-old (6 wo) *esk1-1* and *esk1-5* plants, with lignin stained in green with Carmine-green (3 wo), or lignin fluorescing under UV light (6 wo). B. Comparison of FTIR spectra obtained from xylem in hypocotyl sections of 3 week-old (3 wo) or 6 week-old (6 wo) wild type and *esk1* plants. A Student\'s *t*-test was performed on absorbance values of wild type versus mutant and plotted against wave numbers. The grey zone, between -2 and +2, corresponds to non-significant differences (*p*-value\<0.05) between the two genotypes tested. Significant positive *t*-values indicated a higher absorbance value in wild type than in *esk1* mutants. :::  ::: In accordance with the severe vascular tissue defects observed in both hypocotyls ([Figure 7A](#pone-0016645-g007){ref-type="fig"}) and stems ([Figure 8A](#pone-0016645-g008){ref-type="fig"}), the infrared absorbance spectra from *esk1* mutants were different from those of wild type. Furthermore, *esk1-1* FTIR profiles obtained from stems ([Figure 8B](#pone-0016645-g008){ref-type="fig"}) and from hypocotyls before and after fiber appearance ([Figures 7B](#pone-0016645-g007){ref-type="fig"}) presented the same differences to wild type indicating that the composition of the xylem cell walls was modified in a similar manner in these different tissues. ::: {#pone-0016645-g008 .fig} 10.1371/journal.pone.0016645.g008 Figure 8 ::: {.caption} ###### The structure and composition of *esk1* xylem is modified in stems. A. Transverse section of stems from *esk1-1* and *esk1-5* plants, with lignin fluorescing under UV light. B. Comparison of FTIR spectra obtained from xylem in basal stem sections of wild type and *esk1* plants. A Student\'s *t*-test was performed on absorbance values of wild type versus mutant and plotted against wave numbers. The grey zone, between −2 and +2, corresponds to non-significant differences (*p*-value\<0.05) between the two genotypes tested. Significant positive *t*-values indicated a higher absorbance value in wild type than in *esk1* mutants. :::  ::: Significant differences were observed in the absorption at wavelengths from 1800 cm^−1^ to 1700 cm^−1^, which correspond to ester linkages [@pone.0016645-Mouille1] or esterified pectins [@pone.0016645-Kacurakova1]: *esk1* absorbance was reduced compare to wild type. In addition, a highly significant peak was observed at wavenumbers around 1240 cm^−1^, assigned to C-O vibrations in pectic polysaccharides [@pone.0016645-Kacurakova1]. Other major differences were in a series of wavelengths between 1000 cm^−1^ and 830 cm^−1^, associated with crystalline polysaccharide components such as cellulose, and suggest a reduction in the amount of crystalline cellulose [@pone.0016645-Mouille1]. For example, 898 cm^−1^ corresponds to the β-linked glucan polymers [@pone.0016645-Kacurakova1]. A direct measurement of crystalline cellulose carried out on 6 week-old hypocotyls and stems confirmed that the *esk1* mutant contains lower levels of cellulose than the wild type ([Figure 9](#pone-0016645-g009){ref-type="fig"} and data not shown). In addition, significant differences were noted between wild type and *esk1* mutants at the wavelengths around 1370 cm^−1^ that are associated with CH~2~ stretches of cellulose and 1157 cm^−1^, attributed to cellulose C-O-C linkages [@pone.0016645-Kacurakova1]. Finally, significant differences in absorbance at other non-assigned wavelengths were also observed between wild type and mutants. ::: {#pone-0016645-g009 .fig} 10.1371/journal.pone.0016645.g009 Figure 9 ::: {.caption} ###### Crystalline cellulose content is reduced in *esk1-5* hypocotyl. Cellulose contents expressed as % of hypocotyl dry weight. Data from 4 measurements are plotted. The decrease in *esk1-5* cellulose content compared to wild type (wt) was significant using Mann and Whithney non parametric test (*p*-value\<0.05). :::  ::: Interestingly, in the transcriptome data previously obtained for *esk1-5* [@pone.0016645-BouchabkeCoussa1], no "core genes" participating directly in cellulose, hemicellulose, pectin or lignin biosynthesis or modification were down-regulated in *esk1-5* background. However, At1g23205, annotated as a pectin methyl esterase inhibitor (PMEI) was expressed 6.6 times less in *esk1-5* than in wild type, in accordance with the reduction in esterified pectins indicated by FTIR analysis. Discussion {#s3} ========== *esk1* mutants were initially described as freezing tolerant without cold acclimation [@pone.0016645-Xin1] and ESK1 was proposed to be a negative regulator of freezing tolerance [@pone.0016645-Xin2]. Later studies also revealed altered responses to other stresses such as drought and salt, thereby revoking the specificity of its action. Instead, ESK1 was proposed to play a role in water uptake or circulation [@pone.0016645-BouchabkeCoussa1], [@pone.0016645-Lugan1]. Nevertheless the precise function of ESK1 in these processes remained to be determined. ESK1 plays a role in xylem layout {#s3a} --------------------------------- Xylem conducts water and nutrients from the soil to sink organs. Its differentiation occurs in two steps which seem to be coupled [@pone.0016645-Chaffey1]. During the first step, only xylem vessels are generated, and in the second, corresponding to secondary growth, xylem and fibers are formed. TE then undergo programmed cell death (PCD), leading to functional cell corpses. These empty cells are connected end-to-end by perforated cell plates that allow fluid transport and can support the turgor pressure generated by transpiration without collapsing [@pone.0016645-Fukuda1]. Thus, differentiation of TE involves several steps: cell expansion, followed by the deposition of a thick secondary cell wall and cell death [@pone.0016645-Plomion1]. In seedlings grown *in vitro*, *GUS* expression from the *ESK1* promoter was observed in vascular tissue a few days after germination and may coincide with the occurrence of secondary cell wall deposition, even though the exact time point of the onset of secondary vascular development has not been determined in Arabidopsis [@pone.0016645-Elo1]. A close inspection of *ESK1*-promoter driven staining pattern in the xylem of hypocotyl before floral induction revealed that (i) only the last cell of the xylem cell file was stained and (ii) stained cells presented morphological differences, from dense stained cells, to cell corpses possibly corresponding to different stages of xylem vessel differentiation, rather than PCD. Moreover, at later stages of development, staining was also apparent in interfascicular fibers suggesting that *ESK1* expression occurs in tissues undergoing or that will undergo secondary cell wall deposition and not specifically in TE cell lines ([Figure 5](#pone-0016645-g005){ref-type="fig"}). Cytological analysis revealed that *esk1* mutants presented a strong *irx* phenotype in roots, hypocotyls and stems ([Figure 6](#pone-0016645-g006){ref-type="fig"}). The localization of *ESK1*-promoter driven expression in TE and fibers is in agreement with the collapsed xylem vessels and disorganized fibers observed for *esk1* mutants. Several *irx* mutants have been shown to be affected in their secondary cell wall synthesis and some of them display reduced resistance to compressive force [@pone.0016645-Turner1]. For instance, *irx1/cesa8*, *irx3/cesa7* and *irx5/cesa4* mutants are affected in subunits of the cellulose synthase complex [@pone.0016645-Bosca1], [@pone.0016645-Chen1], [@pone.0016645-Taylor1], [@pone.0016645-Taylor2]; *irx8* is deficient in hemicellulose [@pone.0016645-Persson1]; *cad-c cad-d* double mutant is affected in cinnamyl alcohol dehydrogenase and has 40% less lignin than wild type plants as well as collapsed xylem elements [@pone.0016645-Sibout1]. These *irx* mutants, apart from *cad-c cad-d*, also have a reduced stature. Moreover *irx3/cesa7* and *cad-c cad-d* have reduced rigidity, indicating the importance of the composition of the plant cell wall in determining the physical properties of the plants. In addition, the expression patterns observed for *ESK1* were similar to those for reporter genes under the control of transcription factor promoters that are involved in secondary cell wall synthesis [@pone.0016645-Zhong1]. Therefore, ESK1 may participate in the process leading to secondary cell wall deposition. The *esk1* rosette suffers from a defect in water supply from the root apparatus {#s3b} -------------------------------------------------------------------------------- As xylem vessels are conducting tubes that transport water, defects in their shape may have consequences on water distribution. Water supply is determined by the transport capacity of the root system, or root hydraulic conductance, with the degree of stomata opening controlling the rate of the transpiration driven water uptake into plants. The *esk1-5* mutant had high levels of ABA, causing stomata closure and consequently low levels of evapo-transpiration ([Figures 1](#pone-0016645-g001){ref-type="fig"}, [2](#pone-0016645-g002){ref-type="fig"} and [3](#pone-0016645-g003){ref-type="fig"}). In the *esk1-5 ost1* and *esk1-5 aba3-1* double mutants, water loss and leaf temperature, were intermediate between those of the single mutants ([Figure 2B, C](#pone-0016645-g002){ref-type="fig"} and [3B, C](#pone-0016645-g003){ref-type="fig"}). This higher evapo-transpiration compared to *esk1* single mutants was likely due to increased opening of stomata, which are, therefore, functional in *esk1.* Thus, the reduced levels of evapo-transpiration observed in *esk1* mutants were not due to permanently closed stomata. We hypothesized that this could be because water uptake or transport to the vegetative tissues is reduced by the *esk1* mutation. When hydraulic conductivity was assessed, this was found to be lower for *esk1* mutant roots ([Figure 4](#pone-0016645-g004){ref-type="fig"} and [Table 1](#pone-0016645-t001){ref-type="table"}). Water uptake by the roots occurs through radial transfer from the soil to the xylem vessels through the apoplast, the symplast or through the cell-to-cell pathway where it crosses cell membranes through water channels named aquaporins [@pone.0016645-Javot1], [@pone.0016645-Steudle1]. Apoplastic, symplastic or aquaporin water transport modification could partly explain the lower *L*pr measured for *esk1* roots compared to wild type ([Figure 4](#pone-0016645-g004){ref-type="fig"} and [Table 1](#pone-0016645-t001){ref-type="table"}). Nonetheless, transcriptome analyses did not find significant alterations in aquaporin transcript abundance in *esk1-5* plants [@pone.0016645-BouchabkeCoussa1]. A reduction in root system biomass could also cause a moderate decrease in water flow ([Figure 4A, C](#pone-0016645-g004){ref-type="fig"}), as would modifications to the root structure that generate resistance to water transport from the soil to aerial parts. Once water has reached the xylem, its flux largely depends on the diameter of the vessels. The fact that the shape of the xylem vessels in *esk1* roots was strongly altered could cause a higher resistance of *esk1* vessels to water flow, independent of the water supply status, in agreement with the particularly low hydraulic conductivity measured for the *esk1* lines ([Figure 4C](#pone-0016645-g004){ref-type="fig"}). The maximal conductivity calculated using Poiseuille\'s law predicts a decrease of 80% in *esk1* compared to wild type due to decrease in the diameter of roots vessels ([Table 1](#pone-0016645-t001){ref-type="table"}). This reduction is of the same order of magnitude as the decrease observed in root hydraulic conductivity measured. This suggests that the diameter of the xylem vessels is the main factor limiting water flux in xylem in *esk1* mutants. By contrast, in maize the radial resistance is usually rate-limiting and not the longitudinal hydraulic resistance within the xylem [@pone.0016645-Steudle1]. These results suggest that the origin of the disrupted water uptake and water circulation through *esk1* xylem would be the alterations observed in the vessels\' structure. Therefore, ESK1 appears to be necessary for the correct formation of functional xylem vessels. The *esk1* stressed phenotype is likely due to its TE default {#s3c} ------------------------------------------------------------- Our results tend to show that, as a result of collapsed TE, the overall volume of water circulating in *esk1* xylem is greatly reduced. In previous studies, the *esk1-5* mutant was described as constitutively stressed with regard to its physiology: smaller stature of the plants, reduced RWC and transpiration rate, better WUE, similar transcriptomic profile and metabolomic data under control and drought conditions [@pone.0016645-BouchabkeCoussa1], [@pone.0016645-Lugan1]. The increased osmolyte concentrations observed in *esk1* lines compared to wild type [@pone.0016645-BouchabkeCoussa1], [@pone.0016645-Lugan1] would explain the freezing-tolerant phenotype described by Xin and co-workers [@pone.0016645-Xin2]. In the current study, we showed that ABA levels in *esk1* mutants were higher than those of wild type in non-stressed conditions, which would be incoherent with a role for ESK1in stomata closure through ABA signaling downstream [@pone.0016645-Xie1]. Analysis of the double mutant *esk1-5 aba3-1* showed that this was not due to a direct role of ESK1 in ABA biosynthesis. Indeed, the double mutant phenotypes demonstrated that ESK1 did not intervene in growth regulation and water loss through ABA and that high ABA levels were an indirect consequence of *esk1* mutation. Moreover, double mutant studies and root *L*p~r~ measurement confirmed that *ESK1* could contribute to water circulation in the plant, as suggested earlier [@pone.0016645-BouchabkeCoussa1], [@pone.0016645-Lugan1]. This conclusion is consolidated by the fact that the size of *esk1-5 ost1* double mutant is identical to *esk1-5* ([Figure 2A](#pone-0016645-g002){ref-type="fig"}) and is, therefore, independent of the level of evapo-transpiration by the plant and only dependant on the amount of water in the plant. In addition, *esk1-5 aba3-1* double mutants were smaller than both dwarf single mutants due to the combined effects of ABA deficiency and reduced water transport. Like *esk1*, *irx1*/*cesa8* has a reduced stature and high ABA levels, together with collapsed xylem and reduced water loss on detached leaves [@pone.0016645-Chen1]. Chen and coworkers also observed an enhanced drought and osmotic stress tolerance in *irx1*/*cesa8* mutant compared to wild type [@pone.0016645-Chen1]. Moreover, *irx14* mutant, affected in a glycosyl transferase involved in cell wall glucuroxylan biosynthesis, has recently been shown to exhibit drought tolerance [@pone.0016645-Keppler1]. It will be interesting to see whether this is a general phenomenon among *irx* mutants. By integrating all the aspects of the *esk1* phenotype, a model can be obtained were the low hydraulic conductance of the root apparatus, due to xylem vessels malformation, induces an hydraulic signal to the shoot, resulting in ABA synthesis [@pone.0016645-Christmann1]. The consequences of high ABA levels are well documented, and include stomata closure, low evapo-transpiration and synthesis of metabolites like soluble sugars and proline. Xylem malformation in *esk1* is accompanied by alterations in secondary cell wall composition {#s3d} --------------------------------------------------------------------------------------------- *irx* phenotypes have been reported for mutants affected in their cell wall composition. FTIR analyses were performed on *esk1* sections targeting xylem, which are mainly composed of secondary cell wall. We observed differences between wild type and *esk1* that were attributed to reduced pectin esterification and crystalline cellulose in the mutant. The later was confirmed by crystalline cellulose measurements ([Figures 7](#pone-0016645-g007){ref-type="fig"}, [8](#pone-0016645-g008){ref-type="fig"} and [9](#pone-0016645-g009){ref-type="fig"} and data not shown). Although pectins are mainly present in primary cell walls [@pone.0016645-SanchezRodriguez1], [@pone.0016645-Wolf1], pectin methylesterification appears to be a prerequisite for lignin modifications during secondary cell wall deposition in xylem cells [@pone.0016645-Pelloux1]. A feedback loop which would interconnect the control of cellulose and pectin biosynthesis has already been suggested to participate in the homeostasis of wall rigidity [@pone.0016645-Burton1]. Compensation mechanisms have also been observed between primary and secondary cell walls, when studying the *irx3* mutant affected in a secondary cell wall specific cellulose synthase gene [@pone.0016645-Bosca1]. It is also worth noting that comparative FTIR spectroscopy performed on the *irx1* mutant clearly demonstrated that the cell wall modifications encountered in *esk1* mutants are not those of a cellulose synthase deficient mutant ([Figure S2](#pone.0016645.s002){ref-type="supplementary-material"}). All the modifications observed in *esk1* mutants would therefore be consistent with a role for ESK1 in the deposition or maintenance of a functional secondary cell wall, specifically in xylem and interfascicular fibers, its defect inducing a strong *irx* phenotype. Indeed, the altered secondary cell wall composition may result in a modification of the mechanical properties of the TE, so that resulting cell walls cannot provide enough strength to endure negative pressure, therefore inducing an *irx* phenotype [@pone.0016645-Turner1]. Recently, sequence analysis predicted ESK1 to be part of a novel group of proteins named the PC(Pmr5-Cas-1p)-esterase family. Members of this family share a putative N-terminal acylesterase domain and ESK1 was proposed to modify cell wall glycans through carbohydrate acylation [@pone.0016645-Anantharaman1]. ESK1 also contains a C-terminal Domain of Unknown Function 231 (DUF231) [@pone.0016645-Xin2]. Bischoff and co-workers found that members of the DUF231 gene family like *POWDERY MILDEW RESISTANCE 5* (*PMR5*), *TRICHOME BIREFRINGENCE* (*TBR*) *TBR-LIKE 3* (*TBL3*) and *ESK1*, harbor another plant-specific domain (TBL domain) that contains a conserved GDSL motif common to certain esterases and lipases [@pone.0016645-Bischoff1]. Interestingly, FTIR analysis and PME activity measured for *tbr* and *tbl3* etiolated seedlings also indicated reduced levels of esterified pectins in their cell walls compared to the wild type [@pone.0016645-Bischoff1]. Moreover, secondary cell wall cellulose contents were reduced in *tbr* and *tbl3.* Bischoff and co-workers concluded that the TBL/DUF231 genes might be pectin-binding proteins that are involved in the overall cell wall chemical balance [@pone.0016645-Bischoff2]. Bischoff and Scheible also propose that ESK1 should be renamed TBL29 (personal communication). Nevertheless, we cannot exclude a more general action of ESK1 in xylem formation, as such broad impact on secondary cell wall chemical composition has not previously been reported for any cell wall mutant. The precise function of proteins harboring a DUF231 remains to be determined. Conclusion {#s3e} ---------- The data presented here demonstrate that *ESK1* plays a major role in the formation of functional xylem vessels, which has dramatic consequences on water transport. *ESK1* is expressed in cells that develop into xylem and TE are collapsed in *esk1* mutants, a characteristic feature of *irx* phenotype. We establish for the first time in this study a clear link between this TE phenotype and the low water conductance, which explains their small stature, low evapo-transpiration, the stressed plant state and the stress tolerance previously reported [@pone.0016645-Xin1], [@pone.0016645-BouchabkeCoussa1], [@pone.0016645-Lugan1]. Our results favor the hypothesis that ESK1 is a protein involved in cell wall deposition, maturation or regulation, as suggested for other members of the DUF231 family [@pone.0016645-Anantharaman1], [@pone.0016645-Bischoff1], [@pone.0016645-Bischoff2], [@pone.0016645-Vogel1]. It is also possible that the defect in ESK1 could influence earlier steps with consequences on the correct assembly of cell walls in xylem and interfascicular fibers. Identification of the molecular function of ESK1, its involvement in stress responses and its potential role in cell wall formation will require further investigation. Materials and Methods {#s4} ===================== Plant material and growth conditions {#s4a} ------------------------------------ *esk1-1* was isolated by Xin *et al.* [@pone.0016645-Xin2] and *esk1-4* and *esk1-5* have been described by Bouchabke-Coussa *et al.* [@pone.0016645-BouchabkeCoussa1]. *ost1*/*snrk2e* [@pone.0016645-Yoshida1] was kindly provided by Dr J. Leung, *aba3-1* was identified by Leon-Kloosterziel *et al.* [@pone.0016645-LeonKloosterziel1] and *irx1-1* was described by Turner *et al*. [@pone.0016645-Turner1]. *esk1-1*, *esk1-4*, *esk1-5*, *ost1*/*srk2e* and *aba3-1* are in the Col-0 genetic background, *irx1-1* is in the L*er* background. Abiotic stress {#s4b} -------------- For ABA measurements, control culture conditions and mild drought stress were applied to individual plants, grown on Fertiss® propagation plugs. Seeds were stratified 3 or 4 days at 4°C in 0.1% (w/v) agar and then two seeds were sown onto each propagation plug. After the emergence of cotyledons, seedlings were removed to leave one plant on each plug. Plants were grown in controlled conditions in a growth chamber (22°C, light intensity 180 µmol m^−2^ s^−1^, 16 h photoperiod, 70% RH). Stress was applied at bolting: twenty-five days after sowing in our conditions. Each propagation plug was adjusted daily to the target substrate saturation with nutritive solution: 60% (w/v) for the control and 30% (w/v) for mild drought stress as previously described Bouchabke-Coussa *et al.* [@pone.0016645-BouchabkeCoussa1]. Plants were harvested after 7 days of stress implementation. Infrared thermography and water-loss assays {#s4c} ------------------------------------------- Rapid dehydration assays were carried out using 3-week-old plants grown on propagation plugs in a growth-chamber (22°C, light intensity 180 µmol m^−2^ s^−1^, 16 h photoperiod, 70% RH). Four rosettes per genotype were cut from their root system and water loss was measured as described previously [@pone.0016645-Lefebvre2]. For leaf temperature measurements, plants were grown for 4 weeks in soil in a glasshouse (22°C, minimum 13 h photoperiod, maximum light intensity of 500 µmol m^−2^ s^−1^). In a growth chamber (25°C, 150 µmol m^−2^ s^−1^, 50% RH) leaves from well-watered plants were detached, placed abaxial-side uppermost and images acquired using an A320 infrared camera equipped with a 45° lens (FLIR Systems; <http://www.flir.com>). ABA content determination {#s4d} ------------------------- Four independent experiments were carried out using plants at inflorescence emergence, either subjected to mild water deficit [@pone.0016645-BouchabkeCoussa1] or controls. Triplicate measurements were carried out for each condition and each measurement was performed using tissue from a pool of three rosettes. Tissue was freeze-dried prior to extraction of ABA. Samples were weighed and 33 ng of a standard ABA-d~4~ (Euriso-Top SA, France) were added as an internal standard. Freeze-dried material was ground in 3 ml acetone/water/acetic acid (80/19/1; v/v/v), then centrifuged (4630 *g*, 4°C, 3 min). After re-extraction of the pellets with 1 ml of extraction solvent and sonication for 20 min (25 Hz), the supernatants were combined and concentrated under nitrogen. The dry extract was dissolved in 150 µl acetonitrile/water (50/50; v/v), filtered and analysed by HPLC--electrospray-tandom mass spectrometry (HPLC-ES-MSMS). Sample components were separated on a reverse-phase column (Uptisphere C18 5 µm, 150×2 mm i.d, Interchrom) using a Waters 2695 separation module (Alliance) (Waters, Milford, MA, USA) equipped with a Waters 2487 dual UV detector, with a flow-rate of 0.15 ml/min and a binary gradient: acetonitrile 0.5% acetic acid (v/v) (A) and acetic acid 0.5% (v/v) (B). Typically, the solvent gradient was programmed as following: 0--5 min 20% A, 5--15 min 65% A, 15--20 min 100% A, before returning to the initial composition at 30 min. Separated molecules were ionized in the ESI source and analyzed with a Waters Quattro LC triple quadrupole mass spectrometer (Waters, Milford, MA, USA) operating in a Multiple Reaction Monitoring (MRM) scanning mode. Instrument parameters were set as follows: capillary 2.75 kV (negative mode), extractor 2 V, source block and desolvation gas temperatures 120 and 350°C, respectively. Nitrogen was used to assist the nebulisation and the desolvation (250 and 450 L/h respectively), argon was used as collision gas at 3.5 10^−3^ mbar. The parameters used for MRM quantification of ABA-d~4~ and ABA in positive mode were: cone potential 16 V, collision energy 10 eV for two transitions used m/z 267\>251 and 263\>247 respectively. The limit of detection (LOD) and limit of quantification (LOQ) were calculated from calibration curve and sample using the Quantify module of MassLynx version 4.1 software. Typically, for a 5 µl injection of sample prepared with 33 ng internal standard and reconstituted in 150 µl of 50/50 acetonitrile/H~2~O (v/v), the respective LOD and LOQ are 1 and 3 pg/mg dry mass. Data from the four independent experiments were combined for the statistical analysis. Multiple-samples comparisons were performed between genotypes for each treatment independently. Analysis of the effect of two variables, i.e. genotype and water treatment, and their interactions on the ABA concentration was performed by 2-D ANOVA for each pairs of genotypes. ANOVAs and multiple-samples comparisons were performed with the Statgraphics® software. Multiple-samples comparisons were carried out using the Fisher test (Least Significant Difference) at the 5% threshold. Box-plots graphs were performed with tools created by the Anastats association, available free at <http://www.viesanimales.org/stats/Download.htm>. Reporter constructs, GUS assays and cytological observations {#s4e} ------------------------------------------------------------ An approximately 1 kb fragment of the *ESK1* promoter was amplified from wild type Col-0 genomic DNA using 5′ attB1-TGGTTGGTGCCGTATACATA 3′ and 5′ attB2-CCAAGGTTGCATCTGTTTGT 3′ primers, where attB1 and attB2 contain sequences complementary to the Gateway™ recombination sequences. The PCR product was cloned in the pDONR207™ gateway entry vector (Invitrogen™) via BP recombination reaction. Subsequently an LR recombination reaction was carried out to introduce the fragment into the pBI101-R1R2 binary destination vector (F. Divol, J.-C. Palauqui and B. Dubreucq, Institute Jean-Pierre Bourgin, INRA, Versailles, France, unpublished data) as a fusion with the *GUS* reporter gene. Electro-competent C58C1 *Agrobacterium tumefaciens* were transformed with this construct, which was then used for agroinfiltration of wild type Arabidopsis flower buds. Seven independent transgenic lines, each with a single homozygous T-DNA insertion, were retained for subsequent analyses. Histochemical detection of *GUS* expression was performed as described by Jefferson *et al.* [@pone.0016645-Jefferson1]. Potassium ferricyanide/potassium ferrocyanide was used and concentration adjusted (from 1 to 3 mM) depending on the insertion line, the organ and the developmental stage considered. Samples were observed by light microscopy (LEICA DMR B DIC). For observation of hypocotyls or stems or GUS stained, *esk1* and wild type samples were embedded in 8% agarose and 70 µm sections cut using a vibratome. *esk1* and wild type sections were observed under UV light, where lignified tissues appears fluorescent, or under white light when stained with carmine-green which gives green coloration to lignin and pink coloration to cellulose [@pone.0016645-Cholet1]. Hydraulic conductance {#s4f} --------------------- Root water transport measurements were performed essentially as described in Javot *et al.* [@pone.0016645-Javot2]. Freshly excised root organs were inserted into a pressure chamber with the hypocotyls threaded through the lid of the chamber. Upon pressurization, the flow rate of the exuded sap was measured and plotted as a function of the pressure imposed. The gradient of the graph obtained provided the hydraulic conductance measure, or *L*o, which was then adjusted with respect to root dry weight in order to obtain the root hydraulic conductivity, or *L*p~r~. Data were obtained from 13 measurements for wild type, 9 measurements for *esk1-1* and 5 measurements for *esk1-5*. Box-plots graphs were performed with tools created by the Anastats association, available free at <http://www.viesanimales.org/stats/Download.htm>. For calculation of root maximal conductivity, ImageJ software was used to measure the maximal radius of the 4 to 5 larger xylem vessels. 4 images of root sections observed under UV light, from 3 different plants were used. The maximal conductivity (g) of the vessels was calculated according to Poiseuille\'s law. R = (8.η)/(π.r^4^) where g = 1/R (m^2^/s/MPa), η is the water viscosity coefficient (1E^−9^ MPa/s) and r is the maximal radius (µm). FTIR spectroscopy {#s4g} ----------------- Analyses were carried out on xylem tissue from hypocotyls or the basal region of branched stems using 50 µm thick vibratome sections of agarose-embedded tissue. For each genotype, 5 to 6 sections from 3 different plants were analysed. FTIR spectra were collected from a 50 µm x 50 µm window targeting xylem vessels; normalization of the data and statistical analysis (Student\'s *t*-test) were performed as described in Mouille *et al.* [@pone.0016645-Mouille1]. Crystalline cellulose extraction and dosage {#s4h} ------------------------------------------- Plant material was harvested and stored in absolute ethanol. Hypocotyl or stem pieces were ground and incubated twice in 70% ethanol for 60 min at 70°C. Pellets were then washed in acetone for 2 min at room temperature then vacuum-dried. After weighing samples, crystalline cellulose content was determined as described in Scott and Melvin [@pone.0016645-Scott1] and Updegraff [@pone.0016645-Updegraff1]. Two independent experiments were carried out with two replicates in each. A Mann and Whitney non parametric test was used to compare cellulose content between the two genotypes. Mann and Whitney non parametric test and box-plots graphs were performed with tools created by the Anastats association, available free at <http://www.viesanimales.org/stats/Download.htm>. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Vascular tissue structure in young stems of *esk1* and wild type.** Transverse section of young stems (2 to 3 centimeters high) from wild type (A, wt) and one representative *esk1* mutant plant (B, *esk1-1*), with lignin stained in green with Carmine-green. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **The structure and composition of *esk1-1* and *irx1-1* xylem are different.** Comparison of FTIR spectra obtained from xylem in basal stem sections of *irx1-1* and *esk1-1* plants and their respective wild types, L*er* and Col-0, respectively. A Student\'s *t*-test was performed on absorbance values of wild type versus mutant and plotted against wave numbers. The grey zone, between −2 and +2, corresponds to non-significant differences (*p*-value\<0.05) between the two genotypes tested. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Statistical comparisons of ABA measurements from wild type (wt), *esk1* and *aba3-1* mutants.** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Akira Suzuki, Georges Pelletier, Françoise Budar and Volker Bishoff for critical reading of the manuscript, and Liliane Laroche, Lilian Dahuron and Bernadette Trouve for plant culture. We are grateful to Richard Sibout, Jérôme Pelloux and Deborah Goffner for productive scientific exchanges. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the ANR-07-GPLA-003 grant (Agence Nationale pour la Recherche - Génoplante). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: VL HMN JT YB MDT. Performed the experiments: VL MNF AD HMN AMG JT YB GM. Analyzed the data: VL MNF AD HMN AMG JT YB GM MDT. Contributed reagents/materials/analysis tools: VL MNF AD AMG. Wrote the paper: VL HMN AMG YB GM MDT.

PubMed Central

2024-06-05T04:04:19.083327

2011-2-1

PMC3052266

There was an error in the third subheading of the Results section (\"African-Americans and Caucasians\"). The correct sentence is: Single locus tests of association in the African-Americans identified seven significant allelic associations at IL12B polymorphisms rs3212227 (p = 0.002), rs2421047 (p = 0.008), rs2288831 (p = 0.008), rs10631390 (p = 0.001), rs3212220 (p = 0.003), rs6894567 (p = 0.007), and rs17860508 (p = 0.002) and three significant genotypic associations at rs3212227 (p = 0.05), rs10631390 (p = 0.05) and rs6894567 (p= 0.03) (Table 2b).&rdquo; **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.087763

2011-2-28

PMC3052267

There was an error in the fourth paragraph of the Discussion section. \"Yoruba\" is not a Discussion subheading but the beginning of the sentence: \"Yoruba IL12B haplotype frequencies (parental haplotypes only) \...\" **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.088181

2011-2-28

PMC3052270

::: {#Fig_945 .fig}  ::: Protocol ======== Cell culture ------------ 1. Grow Chinese Hamster Ovary cells (CHO-K1) using standard cell culture procedures in F-12 Ham medium with GlutaMAX I containing 10% fetal calf serum, 50 U/ml penicillin, and 50 µg/ml streptomycin sulfate (Invitrogen). 2. Exchange the culture medium to serum-free media. Label the cell surface proteins were at different time points with CyDye DIGE Fluor Cy3 or Cy5 minimal dyes (see section *Cell Surface Labeling*, below). 3. Pool equal numbers of cells from each time point and label with CyDye DIGE Fluor Cy2. Use these as an internal standard for each 2-D gel. 4. The majority of the experiments can be performed with CHO-K1 cells, but mouse embryo fibroblasts (3T3 L1) and mouse ascites lymphoma lymphoblasts (EL4) can also used (data not shown). 5. Grow the two latter cell types in DMEM medium with GlutaMAX II, but, otherwise, use identical conditions used for the CHO-K1 cells. Cell surface labeling --------------------- 1. Carefully detach adherent cells non-enzymatically, counting and dividing into aliquots of 5--10 x106 cells. For cells growing in suspension, omit the detaching step. 2. Centrifuge the cell suspensions at about 800 x g for 5 minutes. Remove the supernatants containing the medium. 3. Wash the pellets by resuspendion in 1 ml ice cold Hank's Balanced Salt Solution (HBSS) pH 8.5. Centrifuged at 800 x g at 4°C for 2 minutes. 4. Remove the supernatant and resuspend the cell pellet in 200 µl ice cold labeling buffer (HBSS pH 8.5, 1 M urea). 5. Label the intact cells with 600 pmol CyDye DIGE Fluor minimal dyes for 20 minutes on ice in the dark. 6. Quench the reaction by adding 20 µl 10 mM lysine. Incubate for 10 minutes. 7. Wash the surface-labeled cells twice by resuspension in 500 µl HBSS pH 7.4, followed by centrifugation at 800 x g at 4°C for 2 minutes. Cell lysis and fractionation ---------------------------- 1. Lyse the surface-labeled cells in 150 μl cold lysis buffer (7 M urea, 2 M thiourea, 4% CHAPS, 30 mM Tris, 5 mM magnesium acetate pH 8.5). Leave on ice for at least 1 h with occasional vortexing. 2. Centrifuge the lysates at 10 000 x g at 4°C for 5 minutes. Transfer the supernatant to a new tube. This sample is the non-fractionated sample containing all cellular proteins. 3. In parallel, wash cell pellets, as above, then fractionate (using a membrane fractionation kit, Pierce) into membrane and cytosolic fractions prior to 2-D gel electrophoresis. The membrane fraction contains internal and cell surface membrane proteins. 4. For comparison, follow the standard Ettan DIGE procedure ^3^, and lyse, label, and, finally, fractionate the cells. Determine the protein concentration in the samples using the 2- D Quant Kit (GE Healthcare). 2-D electrophoresis ------------------- 1. Rehydrate Immobiline DryStrip gels, pH 3--11NL (24 cm) using Immobiline DryStrip Reswelling tray, 24 cm in 450 µl DeStreak Rehydration solution (0.5% IPG Buffer) overnight. 2. Apply the CyDye-labeled samples (corresponding to 50 µg total protein) to Immobiline DryStrip gels by anodic cup loading in the manifold and perform isoelectric focusing (IEF) using Ettan IPGphor II IEF System according to instructions. 3. After IEF, equilibrate the strips in two steps and place on top of large (26 x 20 cm) 12.5% polyacrylamide gels (SDS-PAGE). Overlay with 0.5% agarose (in running buffer containing bromophenol blue). Run 2-D electrophoresis using Ettan DALTtwelve Large Vertical System at 5 W/gel for 30 min, and then at 15 W/gel until the dye front reaches the bottom of the gel. Imaging and data analysis ------------------------- 1. After completing 2-D electrophoresis, scan the gels for Cy2, Cy3 or Cy5 using a Typhoon 9410 Variable Mode Imager. 2. Compare spot maps from membrane fractions, cytosolic fractions, and non-fractionated samples using DeCyder 2-D Differential Analysis Software ^4^. Post-staining ------------- 1. After imaging, silver stain the gels according to standard procedure ^5^. Protein identification ---------------------- 1. Grow, harvest, wash and lyse preparative amounts (approximately 1 mg total protein from 10 x106 CHO cells) of cells, as described above for a non-fractionated sample. Use a Cy5 cell surface labeled sample (see above) as a spike, and apply together with the unlabelled preparative amounts of protein. 2. Carry out 2-D electrophoresis as described above, but this time, aplly 600 µg unlabeled cell lysate together with 50 µg cell surface labeled spike by anodic cup application. Use reference markers to allow correct spot picking, and place between the glass plates before gel casting according to recommendations ^3^. 3. Scan the gel in Cy5 channel to obtain the Cy5 cell surface spot map, followed by total protein staining using Deep purple total protein stain ^3^. 4. Match together the two spot maps using the DeCyder 2-D software and create a pick list for all the spots corresponding to the Cy 5 labeled cell surface proteins. Pick cell surface proteins using the Ettan spot handling station, extract from the gel plugs, and trypsinate using the Ettan digester. Identify using MALDI-TOF mass spectrometry. **Table 1.** An Ettan DIGE experiment was performed using samples from serum depleted cells labeled according to the cell-surface protein labeling protocol in figure 1. SampleTime of serum depletion**Labelled with CyDye**Gel  number1-Cy 3, Cy 21230 minutesCy 5, Cy 2132 hoursCy 3, Cy 2244 hoursCy 5, Cy 22516 hoursCy 5, Cy 23   Discussion ========== Protein concentration --------------------- An overview of the two labeling workflows is shown in Figure 1. Since the cells are still intact when labeled according to the cell-surface protein labeling protocol, only the cell surface proteins are exposed to the dye. In the standard Ettan DIGE protocol, the cells are lysed before labeling and proteins inside as well as outside the cell are labeled (Fig 1). The relative amount of dye to protein in the cell-surface protein labeling protocol is not known, since cell-surface proteins cannot be specifically quantitated. However, it is known that cell surface proteins constitute a very low proportion of the cellular proteins ^2^. Approximately 5--10 x10^6^ cells to 600 pmol of dye were used. It may be possible to use fewer cells since only 12.5--25% of the nonfractionated sample in this study was used for 2-D electrophoresis. Fig 1. Overview of labeling workflow protocols ---------------------------------------------- Protein concentrations in the different fractions were determined using the 2-D Quant Kit. The total protein amount derived from 10 x10^6^ CHO-K1 cells was 920 μg in the non-fractionated sample, 225 μg in the membrane/hydrophobic fraction and 770 μg in the cytosolic/hydrophilic fraction. These amounts will most likely vary depending on cell type and cell size. The proportion of proteins that are labeled in the cell-surface protocol may be higher than the standard Ettan DIGE minimal labeling, which is 2-3% of total protein. It seems that only one dye molecule is attached per protein molecule, since the spot shape is round and vertical streaking of the low molecular proteins on the gels is absent (Fig 2 and 3). Two and more dye molecules per protein would cause vertical streaking due to increased molecular weight of the labeled protein. This phenomenon would mostly be seen with the low molecular weight proteins, since a change in molecular weight of these proteins would result in a larger shift on the gel compared to high molecular weight proteins. Cell-surface protein specific labeling Two identical samples of cells were surface labeled with CyDye DIGE Fluor Cy3. One was lysed and used directly for 2-D electrophoresis. The other sample was lysed and fractionated into membrane and cytosolic fractions. The entire fluorescent label appeared in the membrane fraction; the cytosolic fraction was devoid of any labeled proteins (Fig 2). The same gel with the cytosolic protein sample was silver stained and the result showed that there were proteins in the gel, but they were not labeled using the cell-surface protein labeling protocol. These results suggest that this new labeling protocol is specific for cell surface proteins. The CyDye DIGE Fluor minimal dye does not appear to enter the cell or pass through the cell membrane. The cells are kept on ice prior to the labeling and this may reduce any transport across the membrane. The time for CyDye DIGE Fluor minimal dye exposure is also relatively short (20 minutes), which is sufficient for protein labeling but not for entry into the cell. Another possible explanation for the lack of labeling inside the cell is that even if the dye passes across the membrane, the pH inside the cell is too low (\< pH 7.4) for an efficient labeling reaction to occur (optimal pH 8.5). The labeling reaction is quenched followed by washing of the cells, which further prevents any protein labeling after the cells have been lysed. This method has been successfully applied to human cell lines in vitro as well as to a complex biological system in vivo ^7^. Fig 2. Specificity of cell-surface protein labeling. ---------------------------------------------------- The cellsurface proteins of CHO-K1 cells were labeled with Cy3 and fractionated. The different fractions were separated by 2-D electrophoresis and scanned for Cy3 fluorescence (A). The same gels were then silver stained (B). Fractionation ------------- There are only minor differences in the spot pattern for the membrane-fractionated sample compared with the nonfractionated sample (Fig 2). The two spot maps were compared using DeCyder 2-D Differential Analysis Software and all the spots detected in the membrane fraction were also present in the non-fractionated sample. Fractionation, therefore, is not necessary to improve detection of cell surface proteins but can be used to verify lack of labeling of proteins inside the cells. Comparison between protocols ---------------------------- To be able to evaluate the advantages with the new cell surface protein labeling protocol, a comparison with the standard Ettan DIGE protocol was performed. Two identical samples from CHO-K1 cells grown in the same flask were labeled in parallel with the two different protocols, respectively (Fig 1). A cell-surface Cy5 labeled sample was run on the same gel as a Cy3 labeled cell lysate. The green spots (Cy3) on the gel (Fig 3A) represent the proteins labeled using the standard Ettan DIGE procedure followed by membrane fractionation. These spots are presumably membrane proteins including cell surface proteins as well as proteins from membranes inside the cell (ER, Golgi, mitochondrion, and nucleus). Standard Ettan DIGE labeling procedure followed by a membrane fractionation step was chosen for comparison, since it should give the highest probability for detecting the low abundant cell surface proteins. The red spots (Cy 5) on the gel (Fig 3A) are cell surface specific proteins labeled using the new cell surface protein labeling protocol that are not visible with the standard labeling procedure (green spots). Furthermore, the yellow spots represent overlapping proteins that occur in both samples using either procedure (Fig 3A). Another useful application would be to combine both labeling protocols to distinguish proteins on the cell surface from those on intracellular membranes. Moreover, information about relative changes in cell surface/membrane protein levels after various stimuli could be followed by using both labeling techniques simultaneously. Fig 3. ------ \(A) 2-D gel images of a CHO-K1 Cy5 cell-surface labeled sample (red spots, see protocol 1, Fig 1) and a membrane fractionated Cy3 sample (green spots, see protocol 2, Fig 1) labeled according to standard Ettan DIGE protocol run in the same 2-D gel. (B) DeCyder 2-D Differential Analysis Software views from the 2-D gel showing a cell-surface labeled protein not visible using the standard Ettan DIGE protocol. Identification of cell surface proteins --------------------------------------- Since the spot pattern is very different between a cell surface labeled sample and a total protein labeled sample, it was necessary to include a cell surface labeled spike to enable matching and identification of the cell surface spots in the preparative spot map. All cell surface proteins were picked. Cell surface proteins can be difficult to identify due to their low abundance. The actual protein amount in some spots may be insufficient for identification, since the cell surface proteins are visually enriched and not physically enriched using this protocol. To facilitate successful identification of low abundant cell surface proteins, the preparative amounts of total protein can be enriched for membrane proteins before application on 2-D electrophoresis. In these CHO cells, intracellular membrane proteins and cell surface proteins constitute approximately 20% of the total protein in the cells. For this cell type, the protein amount in the spots could potentially be increased by a factor 5 by enrichment of membrane proteins. Also, the use of narrow gradient pH intervals of the IPG strips will allow application of larger amounts of protein. In this study we used only 600 μg total protein, with no enrichment for membrane proteins, and a broad range IPG strip. We were still able to identify a large number of cell surface proteins, of which 82% were previously known as membrane associated proteins. Multiplexing ------------ To test the cell surface protein labeling protocol in an Ettan DIGE experiment using all three dyes ^6^, a series of samples from serum depleted CHO cells were collected and cell surface proteins labeled at different time points (Table 1). Samples were separated by 2-D electrophoresis. The preparation of an internal standard for a cell surface DIGE experiment is straightforward. In this case, Cy 2 cell surface labeled samples (from all time points) were pooled and used as an internal standard applied to each 2-D gel. All three CyDye DIGE Fluor minimal dyes labeled cell surface proteins similarly (data not shown). Changes in expression during serum starvation for many of cell surface proteins were detected using DeCyder 2-D software (Fig 4). Fig 4. ------ Change in expression of two cell-surface proteins during starvation of CHO-K1 cells. Spot maps were analyzed using DeCyder 2-D Differential Analysis software. Conclusions ----------- The new Ettan DIGE protocol for cell surface protein labeling is rapid, simple to use and highly specific for labeling cell surface proteins. Many novel cell surface proteins are only detectable when using the cell-surface protein labeling protocol. Over 80 new cell-surface proteins for CHO cells were detected using DeCyder 2-D Differential Analysis Software. Over 80% of the identified cell surface labeled spots were membrane associated proteins. Multiplexing is achieved using the three CyDye DIGE Fluor minimal dyes, and in combination with DeCyder 2-D software, this new protocol is a powerful tool for studying cell surface proteins with all the advantages obtained with the 2-D DIGE technology. We thank Professor Dontscho Kerjaschki, Corina Mayrhofer and Sigurd Krieger at the Institute of Clinical Pathology, University of Vienna, Austria for their collaboration. [^1]: Correspondence to: Asa Hagner-McWhirter at <asa.hagner-mcwhirter@ge.com>

PubMed Central

2024-06-05T04:04:19.088576

2008-11-26

PMC3052297

Introduction {#s1} ============ It is evident that the central nervous system has an impressive capability of forming and maintaining multiple long-term motor memories. We can acquire different motor skills such as bicycling, ice skating or driving a car and once acquired, these skills are often retained for a very long time [@pone.0017451-Karni1]. However, despite the versatility of our motor repertoire learning may be hindered and interference can occur if we engage in subsequent learning of different motor skills [@pone.0017451-BrashersKrug1], [@pone.0017451-Walker1]. Practice of a new motor task, A, leads to improved performance which can last many hours and days after practice. Immediately after practice, the "motor memory" of A is fragile and retrograde interference may occur if another task, B, is learned shortly afterwards [@pone.0017451-BrashersKrug1], [@pone.0017451-Shadmehr1]. In some instances, the memory of task A becomes more resistant to interference over time, such that practice of task B on the following day no longer disrupts the memory of A [@pone.0017451-Muellbacher1], [@pone.0017451-Krakauer1]. Consolidation is defined as the process, by which motor memories become increasingly stable with continued passage of time, and one mechanism of interference is disruption of consolidation [@pone.0017451-Muellbacher1]. In other cases, where it has been postulated that task B is viewed as a variant of task A, such as force-field adaptation or learning of one form of visuomotor rotation followed by another rotation through a different angle, then interference occurs between A and B even on day 2 despite the fact that A had been well learned on day 1 [@pone.0017451-Krakauer1], [@pone.0017451-Caithness1], [@pone.0017451-Goedert1], [@pone.0017451-Krakauer2]. This persistent interference has practical value. When we have two successive lessons practising the same skill (e.g. a golf swing), it makes some sense that the system should view them as a single continuous process of skill acquisition rather than two separate memories. It may be that in this case, performance of task B on day 2 re-engages the motor memory of task A, which then becomes susceptible to interference. Contextual cues (the feel of the golf club) have been speculated as one mechanism that might allow the brain to distinguish between different internal models and thereby learning the same versus different tasks [@pone.0017451-Krakauer2], [@pone.0017451-Cothros1]. In many studies on motor interference there has been an assumption that early interference occurs because the motor output required for the second task interacts with the motor representation of the previously learned task i.e. retrograde interference. This is consistent with the fact that interference after learning task A only occurs if task B is novel and being learned; it does not occur if we perform our normal daily activities. Presumably we cannot interfere with a memory unless we actively engage the system of memory formation itself. The mechanisms of interference may relate to disruption of motor memory consolidation (retrograde interference) or to persisting representations of previously learned motor skills (anterograde interference) [@pone.0017451-Cothros2]. A recent view is that switching between multiple motor skills may be problematic due to contextual retrieval effects [@pone.0017451-Krakauer1], [@pone.0017451-Cothros1]. However, little is known about what details of the new task are actually necessary for interference to occur. Here we ask what factors determine whether or not tasks will interfere with each other. In particular: (1) do the motor memories of A and B have to involve the same movement direction and muscles or just the same joints? (2) is the interference purely motor, or does the sensory input that is being used to improve performance on task B also play a role? The present experiments concern the case in which learning two different skills at the same joint shows early but not late interference. We propose that during repeated practice of a task, there are relatively rapid changes in the effectiveness of synapses in neural circuits that control the movement. Stabilisation of these changes to produce long-term modifications in transmission requires protein synthesis that may take several hours [@pone.0017451-BrashersKrug1], [@pone.0017451-Luft1]. During this period, if the cellular network is reengaged in another session of learning then consolidation of the first task will be disrupted. If this is true we predict that interference should be movement specific and occur only if the neurones that are engaged in the learning of that task are also engaged in learning the second task. We developed an accuracy-tracking task (AT) ([fig. 1A](#pone-0017451-g001){ref-type="fig"}) at the ankle that could be performed using the plantar flexor muscles triceps surae (soleus (SOL) and gastrocnemius), or the dorsiflexor muscle, tibialis anterior (TA). We then tested what signals might be important in interference by examining separately the contribution of activity in motor output pathways to, and sensory inputs from the plantar flexor muscles. As predicted, movement-specific interference was observed with a previously acquired ballistic force task (FT) involving the plantar flexor muscles ([fig. 1A](#pone-0017451-g001){ref-type="fig"}): Interference was observed when the accuracy-tracking task was performed with the same agonist muscles and the same movement direction as the ballistic task. In contrast, interference was not observed when the identical competing task was performed as the opposite movement direction involving the antagonist muscles. Furthermore, interference required learning to occur and sensory input from the trained muscles or movement was surprisingly effective in causing interference, whereas motor output in the absence of sensory input (subthreshold repetitive transcranial magnetic stimulation (rTMS) to the primary motor cortex (M1)) was ineffective. The present study is, to our knowledge, the first to demonstrate that peripheral nerve stimulation may cause interference. We argue that this is because sensory feedback constitutes an important error signal in motor learning. Interference was not observed when the competing task was practiced 4 hours after initial motor practice suggesting that one possible mechanism in the observed interference effects may be disruption of consolidation. ::: {#pone-0017451-g001 .fig} 10.1371/journal.pone.0017451.g001 Figure 1 ::: {.caption} ###### Methodological overview and experimental conditions. (**a**) The primary learning task was a ballistic force task (FT) performed as isometric ankle plantar flexion. In order to study interference effects practice of the primary task was interleaved with periods involving different activities e.g. practicing an accuracy-tracking task (AT). (**b**) Depicts the motor performance of a subject from Group 2 at the beginning and at the end of ballistic force task and accuracy task practice respectively. For the force task, the black traces represent plantar flexion torque. For the accuracy task black represents the target while red represents the exerted torque. (**c**) Subjects were divided in 12 groups. Practice consisted of FT learning periods of 8 minutes separated by a period of 20 min. or ∼4 hours. During the breaks subjects either practiced a secondary accuracy task (AT) with the FT agonist or antagonist muscles, performed a non-learning task, waited, received 1 Hz repetitive transcranial magnetic stimulation (rTMS) of the primary motor cortex or 1 Hz repetitive electrical nerve stimulation (rENS) of the agonist tibial nerve (TN) or the antagonist common peroneal nerve (CPN). :::  ::: Results {#s2} ======= A protocol allowing a direct test of both motor learning and interference effects was established. Participants were divided into 12 groups ([fig. 1C](#pone-0017451-g001){ref-type="fig"}) who all practised making ballistic pulses of maximal voluntary plantar flexion torque at the ankle in 2 or 3 sets of 35 trials (FT1, FT2 and FT3) separated by breaks of 20 minutes or 4 hours ([fig. 1A](#pone-0017451-g001){ref-type="fig"}). In the first set of trials, all groups improved their ability to generate maximal ballistic plantarflexor torque ([fig. 1B](#pone-0017451-g001){ref-type="fig"}) (mean ± s.e.m. increase from first to last trial) 32±3% (F~1,194~ = 122.7, p\<0.001). There were no differences between groups (F~11,194~ = 0.129, p = 0.998). During the breaks different groups were exposed to different interventions ([fig. 1C](#pone-0017451-g001){ref-type="fig"}). Subjects were also able to improve motor performance in the subsequent practice sessions, but the retention of the behavioural improvement markedly depended on the interventions. The subjects in Group 1,2 and 6--12 participated in 2 experimental sessions with a minimum of 2 weeks in-between (see [Methods](#s4){ref-type="sec"} section). When analyzing the baseline performance in theses experiments we observed that in session 2, baseline was 18±7% higher than in session 1 (t = 9.68 p\<0.01). This demonstrates long-term retention of the ballistic force task learning. Experiment I: Interference with retention of motor learning is specific for direction of movement {#s2a} ------------------------------------------------------------------------------------------------- Experiment I investigated the specificity of between-task interference. In the breaks between ballistic training (FT) sets, two groups of subjects practiced an accuracy task (AT) that involved tracking a moving target on a computer screen by generating low force. The task was identical for the two groups, but in Group 1, tracking was achieved by activating the plantar flexion muscles, whereas in Group 2, the tracking force was achieved by the opposite movement direction activating the dorsiflexor muscles ([fig. 1A and 1C](#pone-0017451-g001){ref-type="fig"}). The results obtained in Experiment I are illustrated in [Figure 2](#pone-0017451-g002){ref-type="fig"}. In order to test for differences in FT performance within and between intervention groups and training sets, data for individual subjects were entered into a two-way analysis of variance (ANOVA) with GROUP (Group1, 2, 6 and 9) and SET (change in motor performance during FT1, FT2, FT3 and between FT1-FT2 and FT2-FT3) as factors. The ANOVA yielded a significant main effect for SET F~(4,\ 169)~ = 38.95, p\<0.001. The main effect of group was non-significant, F~(3,\ 169)~ = 1.245, p = 0.28. However, the GROUP×SET interaction was significant, F~(12,\ 150)~ = 2.743, p = 0.006, signifying that the changes in ballistic motor performance within and between sets was different between intervention groups. ::: {#pone-0017451-g002 .fig} 10.1371/journal.pone.0017451.g002 Figure 2 ::: {.caption} ###### Between-task interference is movement and effector specific. Learning of the ballistic force task (FT), the accuracy task (AT) and between-task interference effects. (**a**) Learning curves for the two tasks. The ballistic force task was performed as plantar flexion whereas the accuracy task was performed as either plantar flexion (Group 1 -- red) or dorsiflexion (Group 2 - blue). Motor performance was normalized to baseline (initial) ballistic force and deviation (error) from optimal tracking target respectively. During practice subjects increased ballistic force in FT and decreased deviation in AT. Curves represent group average motor performance, error bars represent s.e.m (**b**) Increase in FT motor performance during FT practice and decrease in FT performance during AT practice. Bars represent group average ± s.e.m. An asterisk denotes significant difference (p\<0.05) in Bonferroni corrected tests. :::  ::: Post hoc pairwise comparisons revealed that there were no between-group differences in FT acquisition within FT1 but between FT1 and FT2 ballistic motor performance decreased 29±9% in Group 1, meaning that in the second set of ballistic trials their first trial was 98% of baseline performance. This decrease in FT performance after AT practice was significantly different from Group 2 (t = 2.417, p = 0.048) in which ballistic performance only decreased 13,8±4%. Although both Group 1 and 2 had to track an identical object in the accuracy task, only Group 1 in which both FT and AT was performed with the plantar flexor muscles showed catastrophic interference with the force task ([Fig. 2](#pone-0017451-g002){ref-type="fig"}). During FT2, both Group 1 and 2 again improved motor performance by 34±11% and 29±10% respectively with no difference between groups (t = 0.107, p = 0.9), but the same pattern of interference was however observed following AT2. Group1, which practiced AT using plantarflexion, produced 31±10% less force in the first trial of FT3 than on the last trial of FT2 set 2. This effect was significantly different from Group 2 (t = 2.843, p = 0.014), in which FT only decreased 12.7±6%. During FT3 both groups again improved motor performance 34±9% and 26±9% of baseline with no difference between groups (t = 1.252, p = 0.7). These results indicate that practice of the accuracy task lead to interference selectively in Group 1 when the two competing tasks engaged the same movement direction and the same agonist muscles. However, since both Group 1 and 2 were exposed to potentially interfering interventions inbetween FT sets, comparison of differences between these groups only allows interpretations on relative interference. Consequently, the ANOVA also included data from Group 9, which served as a control group in which subjects were only exposed to a sham intervention. Pairwise comparison of the effects observed in Group 1 and 2 to Group 9 revealed significant differences. Between FT1 and FTII the decrease in Group 9 motor performance was significantly different from Group 1 (t = 3.287, p = 0.003) but not Group 2 (t = 0.94, p = 0.93). Between FT2 and FT3 motor performance also decreased significantly more in Group 1 than Group 9 (t = 2.558, p = 0.047) whereas there was no difference between Group 2 and 9 (t = 0.4, p = 0.82). These results confirm that interference was selective for Group 1 in which the competing tasks involved the same movement direction and agonist muscles. Analysis of the parameter estimates obtained for the individual FT learning curves (y = y0×ax^b^) demonstrated significant effects of both SET (F~2,\ 107)~ = 43.25, p\<0.001) and GROUP (F~(3,107)~ = 4.26, p\<0.01) on *y0* and also a significant GROUP×SET interaction (F~(6,96~ = 3.42 p\<0.01). In Group 1 there was no difference in *y0* between sets indicating that there was no retention and interference was complete. In contrast, *y0* increased significantly from FT1 to FT2 (t = 2.198, p = 0.055) and FT 2 to FT3 (t = 4.12, p = 0.014) in Group 2 signifying significant FT retention after AT practice. This was also the case for Group 9 from FT1 to FT2 (t = 2.817, p = 0.01) and FT2 to FT3 (t = 2.43, p = 0.023). Within FT2 there was a significant difference in *y0* between Group 1 and 2 (t = 2.64, p = 0.017) and between Group 1 and 9 (t = 2.76, p = 0.016). This was also the case for FT3 in which *y0* was higher for both Group 2 (t = 3.033, p\<0.019) and Group 9 (t = 4.27, p\<0.01) compared to Group 1. Although the parameter estimate *a* decreased slightly between sets in Group 2 and 9 whereas it remained constant in Group 1 no significant differences were observed. The results demonstrate a remarkable specificity of interference in motor learning. In Group 1, complete interference and no FT retention was observed after competing AT training involving the same agonist muscles and movement direction. In contrast, practice of the competing task involving the antagonist muscles and opposite movement direction did not cause interference. There were also significant differences between groups in retention of the accuracy task. The two-way ANOVA performed on changes in AT performance within and between sets yielded a significant main effect for SET F~(3,\ 71)~ = 56.52, p\<0.001, and the GROUP×SET interaction was also significant, F~(3,\ 64)~ = 10.06, p\<0.001. Following FT1, the initial error in AT1 increased slightly in Group 1 compared to baseline. This effect was however not significantly different between groups (t = 2.433, p = 0.19). Within AT1, both groups reduced tracking error with no difference between groups (t = 2.955, p = 0.122), but between AT1 and AT2 tracking error increased significantly more in Group 1 than Group 2 (t = 3.343, p = 0.039). In AT2 both groups again improved accuracy. Thus, for both tasks, interference was strong and specific in the group which used plantar flexion in the tracking task as well as the ballistic task, whereas there was no interference if the tracking task was performed as dorsiflexion involving the antagonist muscles and movement direction. Experiment II: Ballistic motor learning consolidates over time and with increased initial training; Passage of time hinders between-task interference {#s2b} ----------------------------------------------------------------------------------------------------------------------------------------------------- Previous studies have suggested that interference between motor tasks may relate to disruption of early consolidation following motor learning [@pone.0017451-BrashersKrug1], and this has also been found for a ballistic task as in the present study [@pone.0017451-Muellbacher1], [@pone.0017451-Baraduc1]. It is however also possible that interference occurs through other (largely unknown) mechanisms and recent studies of sensorimotor adaptation have failed to demonstrate that time for consolidation stabilizes motor memories against interference [@pone.0017451-Caithness1], [@pone.0017451-Goedert1]. In Experiment II we allowed 4 hours to pass in between FT training sets. Group 3 had 3 hours and 40 minutes break after initial FT training. Following this break subjects engaged in AT training involving the same agonist muscles and movement direction (i.e. plantar flexion) corresponding to Group 1. Immediately after this the subjects engaged in FT2. Subjects in Group 4 did not practice the competing AT task. Instead these subjects had a break of 4 hours between FT sets. In Group 5 subjects also had 4 hours break between FT sets but the amount of training in FT1 was increased from 35 trials to 45 trials in order to investigate whether the learning consolidated with an increased amount of initial training ([Fig. 3](#pone-0017451-g003){ref-type="fig"}). ::: {#pone-0017451-g003 .fig} 10.1371/journal.pone.0017451.g003 Figure 3 ::: {.caption} ###### Ballistic motor learning consolidates over time and with increased initial training; Passage of time hinders between-task interference. (**a**) Learning curves for the FT task. Group 3 (red) practiced the competing accuracy task after a 3 h40 min break. Group 4 (blue) had 4 h break. Group 5 (green) had extra initial practice and 4 h break. Performance was normalized to baseline. Curves represent group average FT motor performance. Error bars represent s.e.m. (**b**) Increase in FT motor performance during practice and decrease in FT performance during breaks. Bars represent group average ± s.e.m. :::  ::: For FT performance, a 2-way ANOVA was used to test for changes within and between groups during FTI, FTII and between FT sets. This test revealed a significant main effect of set (F~2,71~ = 103.31, p\<0.001), no significant main effect for group (F~2,71~ = 0.0125, p = 0.99) and a tendency to a GROUP×SET interaction (F~4,63~ = 2.13, p = 0.08). During initial ballistic practice (FT1), Group 3--5 improved performance with no significant differences between groups. Group 3 improved FT performance by 37±7%, Group 4 improved performance by 36±6% and Group 5 improved performance by 41±7%. Analysis of the individual learning curves also revealed no differences between groups. 3 hours and 40 minutes after initial FT practice Group 3 engaged in AT practice leading to significantly reduced tracking errors (t = 20.81, p\<0.001). [Figure 3b](#pone-0017451-g003){ref-type="fig"} shows that after a break of 4 hours between FT training sets there was retention of the learning in Group 4. Contrary to what was observed in Experiment I, where learning of a competing task immediately after initial learning caused interference, no interference was observed for Group 3, in which subjects engaged in AT learning 3 hours and 40 minutes after initial FT practice. There was no significant difference between Group 3 and 4 in the change in FT motor performance between FT1 and FT2 (t = 0.11, p = 0.98. Analysis of the learning curve parameter estimates demonstrated a significant main effect of SET (F~1,\ 47)~ = 37.44, p\<0.001) on *y0* and a GROUP×SET interaction (F~(2,42~ = 1.136, p = 0.331). In both Group 3 (t = 3.049, p = 0.02), 4 (t = 3.265, p = 0.018) and 5 (t = 3.993, p\<0.01) *y0* increased significantly from FT1 to FT2. The parameter estimate *a* decreased between FT1 and FT2 (main effect F~1,47~ = 4,37, p\<0.01) with no significant differences between groups. These findings indicate that a break of almost 4 hours between learning sessions prevents the learning of a competing AT task from causing interference. In Group 5 we found a tendency that extended practice in the initial training set was followed by a smaller drop in performance during the break compared to Group 3 (t = 1.97, p = 0.068 and group 4 (t = 1.86, p = 0.08). When additionally comparing FT performance at the beginning and end of FT1 and FT2 for Group 3--5 in an additional two-way ANOVA we found a significant SET×GROUP interaction (F~6,84~ = 3.298, p = 0.03) revealing that at the beginning of FT2 Group 5 had a significantly higher FT motor performance than Group 3 (t = 2.651, p = 0.017) and 4 (t = 2.453, p = 0.025). This finding seems consistent with the notion that saturation learning improves retention [@pone.0017451-Krakauer2]. During the second ballistic force task training set all 3 groups improved performance with no significant differences between groups. Experiment III: Interference with retention of motor learning is not seen with a simple nonlearning task {#s2c} -------------------------------------------------------------------------------------------------------- In the interval between ballistic training sets, Group 6 ([Fig. 4](#pone-0017451-g004){ref-type="fig"}) performed plantar flexions at 1 Hz by approximately the same amount as in the ankle-tracking task, but without being required to be as precise as possible and without any visual feedback on motor performance ([Figure 4A](#pone-0017451-g004){ref-type="fig"}). Changes in FT performance within and between sets were entered into the two-way ANOVA previously described for Experiment I in order to allow comparison between groups. ::: {#pone-0017451-g004 .fig} 10.1371/journal.pone.0017451.g004 Figure 4 ::: {.caption} ###### Performing a non-learning task does not lead to interference. **a**) Learning curves for the FT task. Performance was normalized to baseline. Curves represent group average FT motor performance Error bars represent s.e.m During breaks subjects performed a non-learning task involving voluntary 1 Hz agonist contractions. (**b**) Increase in FT motor performance during practice and decrease in FT performance during breaks. Bars represent group average ± s.e.m. :::  ::: The decrease in motor performance observed between FT1 and FT2 ([Fig. 4B](#pone-0017451-g004){ref-type="fig"}) was significantly smaller in Group 6 compared to Group 1 (t = 3.291, p = 0.003) with no differences between Group 6 and Group 2 (t = 0.921, p = 0.83) or between Group 6 and Group 9 (t = 0.42, p = 0.89). The same tendency was observed between FT2 and FT3. Again, Group 6 FT performance decreased significantly less than Group 1 (t = 2.458, p = 0.047) with no difference to Group 2 and 9 indicating no interference. Analysis of the learning curves revealed an increase in baseline (*y0*) from the FT1 to FT2 (t = 3.071, p = 0.025) and from FT2 to FT3 (t = 2.751, p = 0.047). The slope parameter *a* decreased insignificantly between training sets. Conclusively, performance of this non-learning task in the breaks between FT set failed to produce interference with retention of the ballistic task ([Fig. 4](#pone-0017451-g004){ref-type="fig"}). We conclude that the interference observed in Experiment I is not caused by extensive use of the agonist muscle i.e. fatigue. More likely, engagement in a task, which produces motor learning, is essential for interference to occur. Experiment IV: Suprathreshold rTMS causes interference -- subthreshold rTMS does not {#s2d} ------------------------------------------------------------------------------------ We were surprised that voluntary contractions without any requirement to acquire skill did not cause interference. Muellbacher et al. [@pone.0017451-Muellbacher1] and Baraduc et al. [@pone.0017451-Baraduc1] previously examined a similar ballistic learning task in the hand and found that rTMS of the primary motor cortex (M1) at a similar rate (i.e. 1 Hz) and producing a similar amount of contraction as in our volitional task abolished retention. In Experiment IV we repeated their experiment by applying rTMS in the breaks between FT sets. In 2 groups of subjects 1 Hz rTMS was applied to M1 over the hotspot for ankle muscle activation at 115% (Group 7) and 90% (Group 8) of resting motor threshold (rMT). In Group 9 sham rTMS was applied. [Figure 5](#pone-0017451-g005){ref-type="fig"} shows that rTMS at 115% rMT interfered with retention of the ballistic learning, whereas subthreshold rTMS and sham rTMS did not. Differences in individual FT motor performance within and between sets were entered into a two-way ANOVA for Group 7--9. This analysis revealed a main effect of SET (F~2,71~ = 135.87, p\<0.001), a tendency to a main effect of GROUP (F~2,71~ = 2.575, p = 0.078) and a significant GROUP×SET interaction (F~4,63~ = 10.77, p\<0.001). There were no differences between groups in FT1. During the first period of rTMS at 115% rMT, ballistic force decreased 34±5% to 98% of baseline in Group 7 indicating complete interference. This decrease in FT performance was significantly less in Group 8 (90% rMT TMS) (t = 4.963, p = 0.002) and Group 9 (sham rTMS) (t = 6.025, p\<0.001) in which FT performance decreased 8±2% and 3±3%. No difference was observed between Group 8 and 9 (t = 1.061, p = 0.868) indicating that rTMS at 90% rMT did not cause interference. ::: {#pone-0017451-g005 .fig} 10.1371/journal.pone.0017451.g005 Figure 5 ::: {.caption} ###### Suprathreshold but not subthreshold rTMS leads to interference. Effects of FT motor practice and rTMS during breaks on motor learning and corticospinal excitability. (**a**) Learning curves for the FT task. Performance was normalized to baseline. Curves represent group average FT motor performance Error bars represent s.e.m. During the breaks Group 7 (red) received 115% rMT rTMS of M1, Group 8 (green) received 90% rMT rTMS of M1 and Group 9 (blue) received 1 Hz sham rTMS. (**b**) Increase in FT motor performance during practice and decrease in FT performance during breaks. Bars represent group average ± s.e.m (**c**) Motor evoked potential (MEP) recruitment curves for agonist (SOL) and antagonist (TA) before () and after () FT training, and after training + rTMS (red, green, blue). Abscissa represents stimulation intensity, the ordinate represents MEP amplitude normalized to MEP~max~ before training. An asterisk denotes significant difference (p\<0.05) in Bonferroni corrected tests. :::  ::: After the second period of rTMS at 115% rMT, FT motor performance again decreased by 33±5% to 103% of baseline in Group 7. This was significantly different from Group 8 (t = 4.227, p\<0.001) and Group 9 (t = 4.288, p\<0.001) in which FT performance decreased 12±2% and 10±3% respectively with no significant difference between these groups (t = 0.256, p = 0.99). Although rTMS at 115% rMT caused interference, the ability to improve FT performance with practice was not impaired. Both during FT2 and FT3 motor performance increased 35±5% and 35±6% with no difference to FT1 (t = 1.076, p = 0.98 and t = 0.263, p = 0.99). In Group 8 and 9 FT performance increased significantly less during FT2 compared to Group 7 (23±5%, t = 3.983, p\<0.001 and 23±4% t = 3.698, p\<0.001). The same tendency was found for FT3 during which FT performance increased 27±6% (t = 1.686, p = 0.195) and 23±6% (t = 2.12, p = 0.104) respectively. Analysis of the individual learning curves revealed no differences between groups in the first ballistic training set. For *y0*, there was a significant main effect of both GROUP (F~2,71~ = 4.04, p\<0.01) and SET (F~2,,71~ = 40.39, P\<0.001) and a GROUP×SET interaction (F~4,,63~ = 3.51, p\<0.01). With subsequent practice *y0* increased from FT1 to FT2 (t = 2.67, p = 0.016 and t = 3.14, p = 0.001) and from FT1 to FT3 (t = 2.75, p = 0.005 and t = 3.18, p\<0.001) in the 90% rMT and sham rTMS groups. There were no differences in *y0* between sets in the 115% rMT group indicating no retention, but significant differences between the 115% rMT rTMS group and the other two groups in FT2 (t = 2.80, p = 0.004 and t = 3.12, p\<0.001)) and FT3 (t = 2.84 p\<0.001 and t = 3.03, p\<0.001). The parameter estimate *a* displayed a significant main effect of SET (F~2,71~ = 3,37, p\<0.01). For Group 8 and 9 the estimate of *a* tended to decrease in FT2 and FT3, but there were no significant differences between groups. In conclusion 1 Hz rTMS at 115% rMT caused complete interference and abolished retention of the ballistic learning whereas rTMS at 90% rMT did not cause interference. We also examined corticospinal excitability before and after ballistic training by plotting the input-output relationships of motor evoked potential amplitudes (MEPs) in SOL and TA muscles. For SOL there was a main effect of STIMULATION INTENSITY (F~6,279~ = 38.75, p\<0.001), GROUP (F~4,279~ = 6.11, p\<0.01) and a significant STIMULATION INTENSITY×GROUP interaction (F~24,245~ = 2.60, p = 0.02). For TA there was only a main effect of STIMULATION INTENSITY (F~6,279~ = 23.42, p\<0.001). Corticospinal excitability increased in SOL after ballistic training: MEPs were significantly facilitated at stimulation intensities of 1.3 (t = 2.84, p = 0.015) and 1.4 rMT (t = 2.47, p = 0.032). However, this increase was abolished by rTMS at 115% rMT as reported previously [@pone.0017451-Muellbacher1], [@pone.0017451-Muellbacher2], [@pone.0017451-Chen1], [@pone.0017451-Lang1] - for review see [@pone.0017451-Fitzgerald1]: in fact SOL MEP amplitudes obtained at 1.3 rMT (t = 3.23, p\<0.001) and 1.4 rMT (t = 3.15, p = 0.001) were significantly smaller than after FT practice alone. Although less pronounced, subthreshold rTMS at 90% MT was also accompanied by a decrease in MEP amplitudes compared to post training values (1.3 rMT t = 2.43, p = 0.042, 1.4 rMT t = 2.42 p = 0.048). Following sham rTMS there were no significant differences from post training values. Both the training induced facilitation and the depression of corticospinal excitability observed following rTMS was effector-specific since it was observed only for the agonist SOL but not for the antagonist TA ([Fig. 5C](#pone-0017451-g005){ref-type="fig"}). Experiment V: Repetitive electrical stimulation of the nerve to the trained muscle, but not its antagonist causes interference {#s2e} ------------------------------------------------------------------------------------------------------------------------------ Why did suprathreshold rTMS lead to interference whereas subthreshold rTMS and voluntary movement did not? In Experiment V we examined the effect of producing ankle movement and afferent feedback by repetitive electrical stimulation (rENS) of the peripheral nerve to either the agonist plantarflexor muscles (SOL, tibial nerve (TN)) or the dorsiflexor muscle (TA; common peroneal nerve (CPN)). During the breaks between FT sets Group 10 received 1 Hz rENS of TN at 115% rMT, Group 11 received 1 Hz rENS of CPN at 115% rMT and Group 12 received 1 Hz rENS of TN at 90% rMT. [Figure 6](#pone-0017451-g006){ref-type="fig"} shows that rENS applied to the tibial nerve at 115% rMT in Group 10 interfered with retention of the ballistic learning, whereas subthreshold stimulation in Group 12 and suprathreshold stimulation of the antagonist nerve in Group 11 did not. A two-way ANOVA on differences in individual FT motor performance within and between sets revealed a significant main effect of SET (F~2,71~ = 86.84, p\<0.001) and a significant GROUP×SET interaction (F~4,63~ = 6.548, p\<0.001). ::: {#pone-0017451-g006 .fig} 10.1371/journal.pone.0017451.g006 Figure 6 ::: {.caption} ###### Repetitive electrical stimulation of the nerve to the trained muscles, but not its antagonist causes interference. (**a**) Learning curves for the FT task and effects of rENS during breaks. Performance was normalized to baseline. Curves represent group average FT motor performance Error bars represent s.e.m. During the breaks Group 10 (red) received 1 Hz 115% MT rENS of the agonist peripheral nerve (TN), Group 11 (green) received 1 Hz 115% MT rENS of the antagonist peripheral nerve (CPN) and Group 12 (blue) received 1 Hz 90% rENS of the gonist nerve. (**b**) Increase in FT motor performance during practice and decrease in FT performance during breaks. Bars represent group average ± s.e.m. An asterisk denotes significant difference (p\<0.05) in Bonferroni corrected tests. :::  ::: All groups improved motor performance with no between-group differences during FT1 ([Fig. 6A and 6B](#pone-0017451-g006){ref-type="fig"}). During the first break, FT performance decreased 32±9% in Group 10 to 103% of baseline. This decrease was significantly larger than what was observed for Group 11 (t = 4.586, p = 0.009) and Group 12 (t = 4.667, p\<0.007) in which FT performance decreased by 5.5±4% and 5±5% respectively. Between FT2 and FT3 similar observations were made. In Group 10, FT performance decreased 21±6% which was significantly more than Group 11 (7±3%) (t = 2.63, p = 0.028) and Group 12 (12±6%)(t = 2.45, p = 0.042). There were no significant differences between Group 11 and 12. Analysis of the individual learning curves revealed a significant main effect of both GROUP (F~2,71~ = 3.72, p\<0.01) and SET (F~2,,71~ = 29.94, P\<0.001) and a GROUP×SET interaction (F~4,,63~ = 3.22, p\<0.01). There were no differences between groups in the first ballistic training set. With subsequent practice however, *y0* increased from FT1 to FT2 (t = 2.46, p = 0.017 and t = 2.39, p = 0.013) and from FT1 to FT3 (t = 3.14, p = 0.003 and t = 3.24, p\<0.001) for Group 11 and 12 who received suprathreshold CPN stimulation and subthreshold TN stimulation. In contrast, there were no differences in *y0* between training sets for Group 10 who received rENS of TN at 115% rMT, but significant differences from Group 10 to Group 11 (t = 2.70, p = 0.008) and 12 (t = 2.64, p = 0.01) for FT2. Although less pronounced, this was also the case for FT3 *y0* (t = 2.23, p = 0.024) and (t = 2.17, p = 0.03). The parameter estimate *a* displayed a significant main effect of SET (F~2,71~ = 4.05, p\<0.01). For Group 11 and 12 the estimate of *a* tended to decrease in FT2 and FT3, but there were no significant differences between groups. In conclusion, retention of ballistic learning was subject to interference selectively by repetitive suprathreshold stimulation of the TN to the FT agonist muscles, whereas this interference was not observed if stimulation was below movement threshold. Suprathreshold stimulation of the CPN to the dorsiflexor muscle TA also did not cause interference. Discussion {#s3} ========== Practice of a new motor task is usually associated with an improvement in performance. This is generally thought to be mediated by experience-driven changes within the neural circuits involved in the trained task. Indeed, if we stop practicing and return the next day to the same task, we find that our performance has been maintained and may even be better than it was at the first day. This retention is a measure of our ability to form, store and retrieve a motor memory of the task [@pone.0017451-Walker1], [@pone.0017451-Robertson1]. However, if a second motor task is practiced after initial motor learning interference can occur and consequently motor performance on subsequent occasions may be no better than baseline on day one [@pone.0017451-BrashersKrug1], [@pone.0017451-Krakauer1]. How, why and when interference occurs does however remain controversial. In the present experiments, subjects learned to increase their skill in performing a primary ballistic force task. The "motor memory" of this skill was initially labile and between-task interference occurred in Experiment I if the subjects learned an accuracy-tracking task involving the same movement direction and agonist muscles shortly afterwards. However, in Experiment II if the competing task was introduced several hours after learning the ballistic task, then no interference occurred. Consolidation may be defined as a set of processes whereby a motor memory is stabilized with continued passage of time and becomes less susceptible to disruption from a competing memory [@pone.0017451-McGaogh1], [@pone.0017451-Krakauer3]. This implies that the ballistic motor learning consolidated. We speculated that this process of consolidation involves long-term stabilization of synaptic changes that are induced in specific circuits during the period of initial learning. For interference to occur a second task must involve activation of, and learning in, the same neural circuits. The results of Experiment I, II and III were consistent with this. In Experiment I, the results showed that subsequent practice of a different task, emphasizing accuracy rather than force, caused interference to the extent that ballistic task performance returned to baseline. Importantly, the observed interference was very specific. Interference was only observed when the two tasks involved the same direction of movement thereby engaging the same agonist muscles. When the two tasks were learned using an opposite direction of movement (involving the antagonist muscles) no interference was observed. This demonstrates that interference did not relate to the competing task per se since both groups practiced identical tasks. More importantly, it also demonstrates that interference is specific to the neural circuits encoding a specific movement (direction) and involving a specific set of muscles. Since both Group 1 and 2 were exposed to a potentially interfering intervention comparison between these groups only allowed conclusions on relative interference. However, comparison of these groups to a control group revealed selective interference in Group 1 and no interference in Group 2. It could be speculated that part of the FT performance gain during practice could be explained by a warm-up effect, which would not relate to learning as such. However, this is not very likely to explain all of the performance gain since significant retention of performance could be observed both 4 hours and 2 weeks later. In addition, warm-up effects would not explain the observed differences between intervention groups. It is noteworthy that not only did the accuracy task interfere with ballistic learning for Group 1, the ballistic task also interfered with retention of learning in the accuracy task. This effect was also observed specifically for Group 1 in which the two tasks were practiced with the same muscles and direction of movement. Although interference may seem more pronounced in the ballistic task, it is not possible to quantify asymmetry of interference between these tasks. In addition, the ballistic task was practiced for 8 minutes while the accuracy task was practiced for 20 minutes and may also influence the susceptibility to interference. Although the interference effects eliminated improvements due to previous practice it did not affect the ability of the subjects to increase motor performance during the following practice sessions. Rather the interventions may have interfered with the early labile motor memory processes thereby preventing consolidation as indicated by the findings of Experiment II. The question then becomes: what is the mechanism(s) by which one task interferes with a competing or conflicting task? Naturally multiple mechanisms may contribute differentially, but based on the findings of Experiment I and III interference requires competing learning processes in appropriate motor circuits. In Experiment III the subjects performed voluntary plantar flexions during breaks. These voluntary contractions engaged the same muscles and movement direction as the accuracy task but did not cause interference. So why did the accuracy task lead to interference while the voluntary task did not? During the voluntary contractions subjects did not engage in learning. No feedback was provided, nor were any task constraints reinforced meaning that there was no error signal. The results of Experiment III showed that the observed interference is not related to muscle activation per se, but that interference requires the subject to engage in acquisition of a skill, in order to promote learning in appropriate motor circuits. The lack of interference in the simple voluntary nonlearning task may be due to the fact that this was a task with which subjects were already familiar, perhaps due to the similarity of this task to the ballistic force task. It has previously been demonstrated that after initial motor learning, competing motor learning [@pone.0017451-BrashersKrug1], [@pone.0017451-Shadmehr1], [@pone.0017451-Krakauer1], pharmacological interventions [@pone.0017451-Luft1] and rTMS protocols [@pone.0017451-Muellbacher1] can cause interference. It is important to note that different interfering agents may utilize different mechanisms. One possible mechanism of (retrograde) interference in motor learning may be disruption of motor memory consolidation processes. If interference occurs through disruption of consolidation this would require that there is a limited time window during which the learning of skill B impairs future performance of A in the classical ABA paradigm. There is general agreement that practice of skill B can interfere with future performance of skill A, but there is however a large controversy as to whether this interference actually exhibits a temporal gradient. This may depend on the type of learning. In Experiment II we demonstrated that learning of the ballistic task did consolidate with passage of time and with increased initial training. Four hours after initial ballistic practice the accuracy task no longer caused interference. This is consistent with the findings of Muellbacher et al. [@pone.0017451-Muellbacher1] for ballistic motor learning. Although a critical role of a time window for consolidation has been observed previously for ballistic learning as in the present study [@pone.0017451-Muellbacher1], it has also been a topic of great controversy. In other forms of learning, recent studies have failed to confirm the consolidation window hypothesis [@pone.0017451-Caithness1], [@pone.0017451-Goedert1]. A large part of the studies on interference have focused on visuomotor adaptation, force-field learning and rotation adaptation learning. For these types of learning, interference from competing tasks is persistent and resistance to interference is not observed with passage of time as it is observed as it is in the current study. Miall et al. [@pone.0017451-Miall1] argued that this persistent interference could be mediated by anterograde mechanisms. In recent studies it has been demonstrated that interference in visuomotor rotation adaptation may be both retrograde, anterograde and due to contextual blocking of retrieval [@pone.0017451-Krakauer2]. In the current study, it does not appear that the interference observed in Group 1 was anterograde interference caused by after-effects from accuracy task learning. In that case we should still have observed interference in Group 3 since the time interval from B to A in the ABA paradigm was identical in the two groups. Concerning visuomotor rotation this type of learning does however consolidate over time and with increased initial practice [@pone.0017451-Krakauer1]. Different tasks naturally engage different networks and different types of learning may also consolidate differently and have different susceptibilities to interference. This is underlined by a recent study by Baraduc et al. [@pone.0017451-Baraduc1], which demonstrated that rTMS of the primary motor cortex disrupted retention of ballistic motor learning but not force-field adaptation learning. This likely relates to differences in underlying networks and the role of the primary motor cortex in the specific type of learning [@pone.0017451-Baraduc1]. In the present study, practice of the FT also produced specific increases in corticospinal excitability as evidenced by increased MEP amplitudes for the soleus muscle, but not for the antagonist TA muscle. This would appear significant since soleus was the agonist in the ballistic training task. This finding of increased corticospinal excitability is in agreement with previous studies on motor learning. Numerous studies have previously indicated a role of the primary motor cortex, M1 in skill acquisition [@pone.0017451-PascualLeone1], [@pone.0017451-PascualLeone2], [@pone.0017451-Karni2] - for review see [@pone.0017451-Sanes1] - and recently a few studies have also indicated a role of M1 in early motor memory consolidation [@pone.0017451-Muellbacher1], [@pone.0017451-Luft1], at least in certain types of tasks (see [@pone.0017451-Baraduc1] for details). Experiments IV and V tested the importance of motor output and sensory input in causing interference by disrupting the ballistic motor memory. Several studies have documented that low frequency rTMS can reduce cortical excitability transiently [@pone.0017451-Muellbacher1], [@pone.0017451-Muellbacher2], [@pone.0017451-Chen1], [@pone.0017451-Lang1], [@pone.0017451-Chen2] for review see [@pone.0017451-Fitzgerald1]. Like Muellbacher et al. [@pone.0017451-Muellbacher1] we found that application of suprathreshold 1 Hz rTMS of the contralateral motor cortex caused interference. Consistent with the findings of the present study Muellbacher et al. [@pone.0017451-Muellbacher1] found that 4 hours after initial learning of the ballistic task, rTMS did not cause interference. In control experiment rTMS of the occipital and dorsolateral prefrontal cortex did not cause interference and this led to the interpretation that M1 is involved in the early establishment of memory of the ballistic motor task following training. However, we also found that subthreshold TMS, which is also known to activate M1 and corticospinal outputs [@pone.0017451-Fitzgerald1], [@pone.0017451-DiLazzaro1], [@pone.0017451-DiLazzaro2], failed to interfere with motor memory. This difference between rTMS effects could simply be because subthreshold rTMS failed to activate sufficient neurones to produce interference, or it could indicate an important role for re-afferent feedback from the movements evoked by suprathreshold stimulation. Consistent with the TMS results presented here (see [fig. 5](#pone-0017451-g005){ref-type="fig"}) Lang et al. [@pone.0017451-Lang1] demonstrated that corticospinal excitability was suppressed more after suprathreshold as compared to subthreshold rTMS, but in addition, the evoked motor potentials were also suppressed following suprathreshold 1 Hz electrical stimulation of the peripheral nerve (rENS). The interpretation of these findings was that the intensity of stimulation has an impact on the after effects of rTMS and that reafferent feedback may contribute to the stronger suppression of corticospinal excitability observed following suprathreshold 1 Hz rTMS compared to subthreshold 1 Hz rTMS. In fact, Experiment V suggested that the (re)afferent feedback could be highly important since ankle plantar flexion produced by peripheral nerve stimulation of the agonist nerve (rENS), which generates strong sensory input, also interfered with motor memory consolidation. Again the interference effect observed following rENS was very selective and did not occur when the antagonist nerve was stimulated or when subthreshold intensities were used. Why then did a non-targeted voluntary movement of the ankle, which produces sensory input similar to that evoked by peripheral nerve stimulation at 1 Hz, fail to interfere with the formation of motor memory? The difference may be that sensory feedback produced by volitional movement is predictable whereas that produced by peripheral nerve stimulation alone (as well as input produced by suprathreshold TMS) is unexpected. In Experiment III the sensory feedback was generated naturally by voluntary movement and so would not conflict with the expectations. Peripheral nerve stimulation leads to artificially generated sensory signals. In the context of learning, the CNS may interpret this as an error signal indicating a discrepancy between expected movement and the actual movement signalled by sensory feedback. Recent research has suggested that the cerebellum has a crucial role in detecting such discrepancies, assists the cerebral cortex in transforming sensory signals from spinal modules to motor-oriented commands and that it updates motor programmes so that future movements are performed more optimally [@pone.0017451-Krakauer3], [@pone.0017451-Luft2]. In line with this notion, Chen et al. [@pone.0017451-Chen3] recently demonstrated that disruption of the human cerebellar thalamus which relays cerebellar signals to motor cortex significantly impaired the ability of the brain to form internal models of action. Several studies have suggested that motor skill acquisition progresses in multiple dissociable stages, which have different sensitivity to feedback error signals. Smith et al. [@pone.0017451-Smith1] proposed that at least two distinct processes with different learning rates and different capacities for retention are involved in motor learning. This may also relate to susceptibility to interference. One process proceeds rapidly but has poor retention. This phase is hypothesized to depend strongly on feedback error signals and may be located in the cerebellum. The other process evolves slowly and responds weakly to error but retains information well. Hadipour-Niktarash et al. [@pone.0017451-HadipourNiktarash1] recently suggested that M1 contributes to the slow processes that maintain motor memory. Furthermore recent experimental studies highlight the key role of the cerebellum in modulating excitability of M1 after sustained peripheral stimulation as in the present study [@pone.0017451-Luft2]. We consequently propose that unexpected sensory input can be interpreted as an error signal to update the synaptic efficiency in the neuronal circuitries in the sensory-motor system subserving motor performance during and following practice. This disrupts any ongoing plastic changes from previous learning unless these have been consolidated by changes in protein synthesis. Indeed, the remarkable muscle specificity that we have observed at least at the ankle joint suggests that the some of the synaptic changes could occur at lower levels of the motor output such as M1 or the spinal cord. The present findings add knowledge to the literature on interference and consolidation in motor learning in several respects. The results demonstrate that the observed interference effects are remarkably specific, consistent with the idea that interference occurs in neural circuits that are involved in a specific movement and activation of individual muscle synergies. Two important results reinforce each other: Experiment I and III demonstrated that between-task interference requires identical or overlapping circuits to be engaged in competing motor learning processes; Interference does not occur when learning is not involved. In such cases, there is no error signal and therefore no competing motor memory consolidation process. Secondly, the present study is, to our knowledge, the first to demonstrate that peripheral nerve stimulation may cause interference possibly through disruption of early motor memory formation. This emphasizes the importance of sensory feedback error signals in interference and motor learning. Methods {#s4} ======= Participants {#s4a} ------------ Sixty-one adults aged 20--42 years, (25±4, mean + s.d.), 39 males and 22 females trained a ballistic force task (FT) involving rapid plantar flexion with the left ankle joint. Participants were untrained to moderately trained and had no known medical or neurological conditions that could impair motor learning or performance. Participants who prior to participation had a history of training ballistic plantar flexion movements were excluded from the study. Participants were randomized into twelve different training groups described in details below (See [fig. 1C](#pone-0017451-g001){ref-type="fig"}). Participants in Groups 3, 4 and 5 only participated in one experimental session. Participants in all other groups were included in two different training groups, so that N = 8 for all groups except Group 1 and 2 in which N = 9. A minimum of 2 weeks between each participant\'s experimental sessions was given to minimize the influence of the first test on the second test. Baseline performance in session 2 was however significantly better than in session 1. This demonstrates long-term retention of the ballistic FT learning. The motor performance measures reported in the results section were always normalized to the baseline performance in the individual test and although the reuse of subjects could potentially affect the amount of interference observed in individual subjects, the reuse of subjects does not affect the conclusions of the study, since subject allocation was randomized and marked differences were observed between different intervention groups regardless of prior experience with the task. Written informed consent was obtained from all participants before the experiment. The experiments (KF 01-131/03) were approved by the local ethics committee of the Capital Region of Denmark (De Videnskabsetiske Komiteer for Region Hovedstaden) and followed the regulations expressed in the Helsinki declaration (1964). Force Task {#s4b} ---------- All twelve groups performed and practiced a ballistic force task (FT) consisting of 2 or 3 sessions of 35 ballistic isometric plantar flexions of the ankle joint. Group 5 performed 45 ballistic plantar flexions in the initial training set. Before the practice started participants were instructed to perform a 5 min. warm-up session on a bicycle ergometer. During motor practice the participants were seated in a custom build chair with their left foot attached to a force pedal. Before training subjects were instructed how to perform the task and allowed two test contractions in order to become accustomed to the task. Participants were instructed to produce as much torque as possible by pressing the force pedal (isometric contractions) using plantar flexor muscles within 250 ms, then relax and return to baseline within a total time window of 500 ms. Isometric conditions where chosen in order to minimize any contribution of antagonist muscle activation such as that seen as part of a triphasic activation pattern during rapid concentric movements in the upper limb [@pone.0017451-Marsden1]. The participants performed one ballistic isometric plantar flexion every 10 s. The participants were given visual feedback on a monitor placed in front of them. The monitor displayed a window with a trace of the force applied during each contraction, the specified time window for contraction and a continuously updated trend plot of the FT performance in all trials obtained during the whole set. The participants were instructed in each trial to perform a maximal ballistic contraction and to increase maximal force across trials. Trials in which a countermovement (defined as a downward deflection of the baseline) or "false start" (defined as an upward deflection of the baseline preceding the actual contraction) occurred were not included in the analysis. If such trials were noted during the experiments an extra trial was performed. During all training sessions participants were verbally encouraged to improve their performance as much as possible. The three FT training sets (FT1, FT2 and FT3) sessions were separated by periods of 20 minutes. During these periods the training groups were subjected to different interventions. Accuracy task {#s4c} ------------- During the two 20 min breaks between FT training Group 1 and 2 trained a second motor task. Contrary to the force task this second task emphasized maximal accuracy. The accuracy task (AT) training was performed in order to evaluate the effect of learning a second motor task on the recent motor learning from the first task. Secondly, it was the purpose to investigate whether any observed disruptive effects on motor memory consolidation were general or muscle/movement specific. The AT involved visuomotor tracking of a computer generated sinusoid curve with one cycle pr 10 sec on a monitor using only the plantar flexion (Group 1) or dorsiflexion (Group 2) force signal. The curve was displayed in windows of 8 seconds each and each subject performed a total of 120 windows in each 20 min period. Before beginning of the training subjects were instructed how to perform the AT and allowed two test trials in order to become accustomed to the task. The subjects were instructed to keep the force signal as close to the target curve as possible at all times and were verbally encouraged to improve their performance as much as possible. The maximal force produced during the AT task was very low (5% of MVC). Passage of time between initial learning and learning of a competing task and effect of extra practice {#s4d} ------------------------------------------------------------------------------------------------------ In Experiment II (Group 3 to 5) we investigated whether the FT learning consolidates over time and with increased initial learning. To elucidate whether the susceptibility to interference by learning a competing task was affected by passage of time Group 3 had a break of 3 hours and 40 minutes after training the FT. Following this break subjects engaged in the AT training involving the same movement direction and agonist muscles (i.e. plantar flexion). Immediately after this the subjects engaged in a second FT training set. Group 4 did not train the AT. Instead, these subjects had a break of 4 hours in-between FT training sets. In Group 5, subjects also had 4 hours break in-between FT sets, but the amount of training in the initial FT set was increased from 35 trials to 45 trials in order to investigate whether FT learning consolidated with increased initial training i.e. if extra learning affected retention. Voluntary contractions of agonist muscles during breaks {#s4e} ------------------------------------------------------- In order to investigate whether any effects observed in Group 1 or 2 could be related to fatigue or simple use of the agonist rather than learning a different task, Group 6 performed small isometric contractions (\<5% of MVC) of the soleus muscle at a frequency of 1 Hz during the 20 min breaks in between force task training with no learning requirement and no feedback on performance. Repetitive transcranial magnetic stimulation and peripheral nerve stimulation during breaks {#s4f} ------------------------------------------------------------------------------------------- To further elucidate the mechanisms and susceptibility to interference and the role of motor output and sensory feedback different interventions were applied during the breaks in between FT sets. Group 7 and 8 received 1 Hz rTMS of M1 for 20 minutes at intensities of 115% and 90% rMT for the SOL muscle. Group 9 received sham TMS. Group 10 and 11 received rENS of the tibial nerve at an intensity of 115% and 90% SOL rMT. Group 12 received 1 Hz rENS of the common peroneal nerve at an intensity of 115% TA rMT. These different stimuli were applied to evaluate the effect on newly formed motor memory. Each participant received a total of 1200 stimulations at a rate of 1 Hz during each pause. The stimulus parameters for the peripheral nerve stimulation were chosen in order to evoke motor responses and (re)afferent activation corresponding to the rTMS evoked responses. Recording and stimulation procedures {#s4g} ------------------------------------ Electromyographic (EMG) recordings from the TA and SOL muscles were obtained with bipolar surface EMG electrodes (0.5 cm diameter of electrodes; 2 cm distance between electrodes; Blue Sensor, Ambu Inc.,USA) over the belly of the muscles. Torque was measured with a custom build force pedal with a build in strain gauge. The EMG signals were amplified (×2000), using custom build EMG-amplifiers, filtered (band-pass, 25 Hz to 1 kHz) sampled at 2 kHz, and stored on a PC for off-line analysis (CED 1401+ with Signal 3.09 software, Cambridge Electronic Design Ltd., UK). Magnetic stimuli were delivered to the contralateral (right) hemisphere primary motor cortex (M1) by a Magstim Rapid^2^ stimulator (Magstim Company Ltd., Whitland, UK) via a custom-made 90 mm figure-of-eight coil (batwing design, Magstim Company Ltd., Whitland, UK). The optimal position of the coil for eliciting motor evoked potentials (MEPs) in the SOL muscle was established through a mini-mapping procedure of M1 and the coil was placed on the scalp over the hot-spot of the SOL representation with the handle of the coil pointing horisontally backward, so that the current in the brain flew in posterior-anterior direction. Motor threshold (MT) was defined, as the minimum intensity required to produce MEP amplitudes larger than 50 µV in 3 out of 5 trials. During all experiments involving TMS a Brainsight^tm^ image guided TMS navigational system (Brainsight-Frameless 1.7.7 Rouge Research, Canada) was used to monitor the position of the coil. rTMS experiments were performed in accordance with current safety recommendations [@pone.0017451-Wassermann1]. In the sham TMS condition (Group 9) a magnetic coil was placed on the scalp of the subject and a second coil was placed above this coil. Stimulations were delivered only through the second coil, thus not activating the motor cortex. TMS recruitment curves were obtained through application of stimulation intensities from 0.8--1.4 MT in steps of 0.1. Stimulations were delivered in a random sequence with 4 s inter stimulus interval. 5 stimuli were obtained at each intensity and the MEP was measured as the average peak-to-peak amplitude of five trials. In Group 10 and 11 soleus Hoffmann reflexes (H-reflexes) were elicited by electrical stimulation of the posterior tibial nerve (PTN) with a 1 ms square-wave pulse (model DS7A Digitimer, US) in the range 10--50 mA using a custom build ball-shaped mono-polar electrode placed in the popliteal fossa. The anode was placed proximal to the patella. H-reflex threshold was defined as the minimum intensity required to evoke a H-reflex visible in the online soleus EMG. In Group 12 activation of TA was elicited through electrical stimulation of the common peroneal nerve (CPN) with a 1 ms square-wave pulse (model DS7A Digitimer, US) in the range 10--50 mA using bipolar electrodes (Blue Sensor, Ambu Inc. USA) The anode was placed distal and lateral to the insertion of the patellae ligament and the cathode just below the neck of the fibula above CPN. In TA it was not possible to evoke a clear H-reflex in any participants. Consequently stimulations were applied at 115% M-wave threshold defined as the minimum intensity required to evoke an M-wave in the online TA EMG (for methods see e.g. [@pone.0017451-Schiepatti1]). Data analysis {#s4h} ------------- Motor performance in the force task (FT) was calculated as the peak ankle torque within the time window. The torque produced in each trial throughout training was normalized to baseline performance i.e. the torque produced in the first trial during training. Motor performance in the AT task was calculated as the mean difference between the position of the target sinusoidal curve and the force signal in each trial. This average deviation from optimal performance (error) was also normalized to baseline performance. This normalization of motor performance to baseline was performed in order to allow comparison. For quantification of improvement of performance during training sets and loss of performance between sets an average of the first 3 trials and the last 3 trials in each set was calculated respectively. Statistical analysis was performed on the data using Sigmaplot 11 (Systat Software Inc. USA). Before statistical comparison, all data sets were tested for normal distribution by a Kolmogorov-Smirnov test. Motor performance in the force task and the accuracy task in the different groups were compared by two-way analysis of variance (ANOVA) with time (set) and groups as factors. Initially, the increase in motor performance during FTI was tested against baseline for all groups. Following this, separate tests were completed for each of Experiment I to V. In these tests a two-way ANOVA was used to test for differences in force task performance within and between groups during FTI, FTII, FTIII and during the break between sets (FTI-FTII and FTII-FTIII). Post hoc pairwise comparisons were performed with Bonferroni tests. In Experiment I the statistical FT analysis included data from Group 1,2 and 9 in order to enable comparison of interference effects to a control (sham) group. It should also be noted that Experiment II consisted of a different number of sets and hence also a different number of statistical comparisons. A corresponding two way ANOVA was set up for the AT performance in Experiment I, while improvement in AT performance in Experiment II was tested as a paired t-test. To further elucidate how the different interventions caused interference with motor learning all FT learning curves were fitted to a three-parameter power function y = y0+ax^b^. Again, a two-way ANOVA was performed for each experiment on the parameter estimates with group and set as factors. MEP amplitudes were normalized to MEP~maxpre~ to allow comparison. MEP amplitudes were compared by a two-way ANOVA with treatment and time as factors. In all statistical tests multiple comparison analyses, post hoc Bonferroni tests were performed for all pairwise comparisons. All values are reported as mean ± s.e.m. unless stated otherwise. In all tests, statistical significance was assumed if p\<0.05. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The study was supported by Ludvig and Sara Elsass Foundation ([www.elsassfonden.dk](http://www.elsassfonden.dk)). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JLJ JBN. Performed the experiments: JLJ THP. Analyzed the data: JLJ THP. Wrote the paper: JLJ JCR JBN.

PubMed Central

2024-06-05T04:04:19.090131

2011-3-9

PMC3052298

Introduction {#s1} ============ Muscle satellite cells (SCs) are a population of cells that reside between the sarcolemma and basal lamina of myofibres [@pone.0017392-Mauro1] and have been shown to play an integral role in skeletal muscle repair [@pone.0017392-Parker1], hypertrophy [@pone.0017392-Rosenblatt1], [@pone.0017392-Schultz1], and hyperplasia [@pone.0017392-Tamaki1], [@pone.0017392-Walsh1], in humans and animals. Although SCs are key regulators of muscle growth during development and muscle adaptation following exercise [@pone.0017392-Carlson1]--[@pone.0017392-Dreyer1], the cellular regulation of human muscle SC function remains largely unexplored. Undoubtedly, the orchestration of events that govern SC function following damage involves a complex milieu of factors originating from the SC in addition to niche factors extrinsic to the SC [@pone.0017392-Christov1]. Identified regulators of human SCs include insulin like growth factor-1 [@pone.0017392-McKay1], hepatocyte growth factor [@pone.0017392-OReilly1], transforming growth factor beta [@pone.0017392-Carlson2] and Notch/Delta [@pone.0017392-Carlson1]. Recently, interleukin-6 (IL-6) has been implicated as playing a role in the regulation of human SCs in response to damaging eccentric contractions [@pone.0017392-McKay2]. Traditionally, IL-6 was considered an inflammatory cytokine [@pone.0017392-Spangenburg1], however, recent work has shown that IL-6 is produced by muscle [@pone.0017392-Hiscock1], released into circulation [@pone.0017392-Steensberg1] and can act on the muscle cells themselves. As such, IL-6 is now also referred to as a "myokine" [@pone.0017392-Keller1], [@pone.0017392-Pedersen1]. Importantly, IL-6 knockout mice demonstrated a blunted hypertrophic response and less SC-mediated myonuclear accretion compared to wild-type mice following compensatory hypertrophy [@pone.0017392-Serrano1]. Furthermore, SCs from IL-6^−/−^ mice demonstrated an impaired proliferative capacity, both *in vivo* and *in vitro*, which was shown to be related to a lack of IL-6-mediated signal transducer and activator of transcription-3 (STAT3) signalling [@pone.0017392-Serrano1]. We have recently reported an increase in IL-6 protein localized in SCs 24 hours following a damaging bout of muscle-lengthening contractions (MLC) in humans coinciding with an increase in *cyclin D1* expression and SC number [@pone.0017392-McKay2]. These data indicate that IL-6, acting via the janus kinase 2 (JAK2)/STAT3 signalling pathway, may be involved in SC proliferation/activation. STAT3 is a downstream target of IL-6 [@pone.0017392-Levy1], [@pone.0017392-Trenerry1], and in response to IL-6 binding, STAT3 is phosphorylated via JAK2. This leads to the subsequent homodimerization and translocation of p-STAT3 to the nucleus [@pone.0017392-Rawlings1]. Once in the nucleus, p-STAT3 binds to the γ-interferon activation sequence (GAS) element where it then promotes the transcription of downstream genes [@pone.0017392-Ivanova1]. These genes have been shown to be responsible for a number of cellular functions including proliferation, migration, as well as anti-apoptotic functions [@pone.0017392-Serrano1]. *cMyc* is a downstream target gene in the STAT3 signalling cascade. It has been shown to regulate cell-cycle kinetics through the up-regulation of a number of Cyclin proteins which are involved in the cell growth phase G1 [@pone.0017392-Fukada1]--[@pone.0017392-Masuda1]. Furthermore, STAT3 also regulates a number of its upstream signalling cascade members including IL-6, GP130, IL-6Rα and suppressor of cytokine signalling 3 (SOCS3). The STAT3 pathway is regulated in a negative feedback loop through interactions with JAK2 [@pone.0017392-Naka1]. SOCS3 can bind phosphotyrosines on JAK2 and physically block STAT3 from binding to JAK2. Additionally, SOCS3 can recruit ubiquitin-transferases leading to the ubiquitination and degradation of JAK2 [@pone.0017392-Rawlings1]. Based on previous work by McKay and colleagues (2009) showing that p-STAT3 co-localized with SCs we aimed to quantify SC localized p-STAT3 signalling over a time course. We hypothesized that the mRNA species of the IL-6/STAT3 signalling cascade would be up-regulated early following the MLC protocol along with a similar increase, as reported previously by McKay and colleagues (2009), in the IL-6^+^/Pax7^+^ cell population. Furthermore, using a time course directed at capturing STAT3 phosphorylation, we hypothesized that we would observe an increase in p-STAT3 specifically in the SC population coinciding with an increase in SC number. In addition, we hypothesized that cMyc, a downstream product of STAT3 signalling, would be detected with mRNA up-regulated in whole muscle and protein co-localized to SCs following an increase in STAT3 signalling in the SC population. Results {#s2} ======= To confirm that the MLC protocol caused muscle damage we examined serum creatine kinase (CK) levels over the time course. Serum CK peaked 24 hours post MLC, increasing over 300% from PRE levels (p\<0.05 vs. PRE) ([figure 1a](#pone-0017392-g001){ref-type="fig"}). Serum IL-6 also peaked 24 hours post-MLC (p\>0.05 vs. PRE), as well, increasing over 200% from PRE ([figure 1b](#pone-0017392-g001){ref-type="fig"}). These measures were correlated indicating a relationship between muscle damage and serum IL-6 (R^2^ = 0.3055; p\<0.001) ([figure S1a](#pone.0017392.s001){ref-type="supplementary-material"}) ::: {#pone-0017392-g001 .fig} 10.1371/journal.pone.0017392.g001 Figure 1 ::: {.caption} ###### Serum measures and Pax7 positivity. (**1a**) Average serum CK response in U/L. (**1b**) Average serum IL-6 response in pg/mL; note the similar serum responses between IL-6 and CK. (**1c**) Myf5 mRNA expression relative to GAPDH, expressed as fold change from PRE. (**1d**) Pax7^+^ cells as a percentage of total myonuclei over the time-course. (**1e**) Representative image at 40× magnification of a Pax7/Laminin stain with Pax7 in red, Laminin in green and DAPI in blue. Values are reported as mean ± S.E.M. \*p\<0.05 vs. PRE; † p\<0.05 vs. T1, ‡ vs. T3. :::  ::: In response to acute muscle damage we observed a 26.6% increase in Pax7^+^ cells 24 hours following the MLC protocol. Pax7^+^ cells per 100 myofibers increased from 15.5 at PRE to 19.6 (p\<0.05) 24 hours post exercise. When expressed as a percentage of total myonuclei we observed a 60.3% increase in satellite cell number (∼3% at PRE to ∼4.5% 24 hours post) (p\>0.05 ([figure 1d](#pone-0017392-g001){ref-type="fig"})). Satellite cells were quantified using a Pax7/Laminin co-stain ([figure 1e](#pone-0017392-g001){ref-type="fig"}) to ensure that all Pax7 cells were in the SC niche. Furthermore, myogenic regulatory factor 5 (*Myf5*), known for its role in SC proliferation was significantly up-regulated 1.8 fold 24 hours following the MLC (p\<0.05) ([figure 1c](#pone-0017392-g001){ref-type="fig"}), while no significant change was observed for myogenic regulatory factor 4 (*MRF4)* (data not shown), known for its role in differentiation. The number of SCs expressing IL-6 was quantified ([figure 2a--d](#pone-0017392-g002){ref-type="fig"}). While there were relatively few co-positive SCs PRE, there was a significant increase in IL-6^+^/Pax7^+^ cells at 3 and 24 hours post MLC vs. PRE (p\<0.05) with the peak observed at 3 hours post where 75% of Pax7^+^ cells were co-positive for IL-6 ([figure 2e](#pone-0017392-g002){ref-type="fig"}). ::: {#pone-0017392-g002 .fig} 10.1371/journal.pone.0017392.g002 Figure 2 ::: {.caption} ###### IL-6/Pax7 co-localization. (**2a**) Representative merged image of a muscle cross-section at 40× magnification stained for IL-6 in green, Pax7 in red, and nuclei in blue being DAPI^+^. The scale bar is 50 µm. (**2b--e**) **100×** magnification image of the inset box in 2a. Scale bar is 10 µm (**2b**) DAPI^+^ nuclei. (**2c**) Pax7^+^ nuclei. Note only one of the nuclei is Pax7 positive showing the specificity of the Pax7 stain (**2d**) IL-6^+^ nuclei. Note that both nuclei are IL-6 positive while there is less intense IL-6 staining in the fiber itself. (**2e**) Merged image showing the co-localization of DAPI, IL-6 and Pax7. (**2f**) IL-6^+^ cells as a percentage of Pax7^+^ cells. Values are reported as mean ± S.E.M. \*p\<0.05 vs. PRE. :::  ::: Previous attempts to quantify the STAT3 response did not show a significant change at the whole muscle level, thus the nuclear and cytoplasmic fraction of whole muscle was examined for a more specific analysis. Purity was assessed using lactate dehydrogenase (LDH) as a nuclear marker and p84 as a cytoplasmic marker ([figure S1b](#pone.0017392.s001){ref-type="supplementary-material"}). Cytoplasmic p-STAT3 was significantly up-regulated 24 hours vs. PRE (p\<0.05) when measured against t-STAT3 which remained unchanged over the time course ([figure 3a](#pone-0017392-g003){ref-type="fig"}). When the nuclear fractions were analysed; however, no detectable p-STAT3 and only traces of t-STAT3 were observed ([figure 3b](#pone-0017392-g003){ref-type="fig"}). JAK2 was also analyzed in the cytoplasmic fraction ([figure 3c](#pone-0017392-g003){ref-type="fig"}). The ratio of p-JAK2 to t-JAK2 was not significantly different at any time point ([figure 3d](#pone-0017392-g003){ref-type="fig"}). ::: {#pone-0017392-g003 .fig} 10.1371/journal.pone.0017392.g003 Figure 3 ::: {.caption} ###### Nuclear and cytoplasmic expression of STAT3 and JAk2. (**3a**) Ratio of phosphorylated to total STAT3 protein in the cytoplasmic fraction. (**3b**) Representative images of p-STAT3 and t-STAT3 in both the cytoplasmic and the nuclear fraction. (**3c**) Representative images of p-JAK2 and t-JAK2 in the cytoplasmic fraction. (**3d**) Ratio of phosphorylated to total JAK2 in the cytoplasmic fraction. Values are reported as mean ± S.E.M. \*p\<0.05 vs. PRE. :::  ::: Although changes in p-STAT3 were not detectable in the enriched nuclear fractions via western blot, immunofluorescent analysis of p-STAT3 ([figure 4a--b](#pone-0017392-g004){ref-type="fig"}) illustrated an increase in p-STAT3^+^/Pax7^+^ cells as a percentage of total Pax7^+^ cells ranging from 20% PRE to 40% at 1 hour, 50% at 3 hours, and 60% at 24 hours (all significantly different vs. PRE p\<0.05) ([figure 4c](#pone-0017392-g004){ref-type="fig"}). Diffuse p-STAT3 staining was observed throughout the muscle fibres and appeared to intensify over the time course ([figure 4a--b](#pone-0017392-g004){ref-type="fig"}), which agreed with the increased cytoplasmic p-STAT3 at 24 hours post observed using western blot analysis. Importantly, the diffuse p-STAT3 observed in the fibre almost never co-localized with non-satellite cell myonuclei. To further verify that p-STAT3 signalling was indeed occurring, downstream target genes of STAT3 were analyzed. *IL-6* mRNA peaked at 3 hours, up approximately 150 fold from PRE (p\<0.05) and remained elevated, up ∼80 fold from PRE (p\<0.05) at 24 hours ([figure 5a](#pone-0017392-g005){ref-type="fig"}). Both *IL-6Rα* ([figure 5b](#pone-0017392-g005){ref-type="fig"}) and *GP130* ([figure 5c](#pone-0017392-g005){ref-type="fig"}) mRNA showed significant increases peaking at 24 hours up 7 fold from PRE (p\<0.05) and 4.5 fold from PRE (p\<0.05) respectively. *SOCS3* also increased, peaking at 3 hours, up 13 fold (p\<0.05) and remaining elevated at 24 hours up 8 fold (p\<0.05) ([figure 5d](#pone-0017392-g005){ref-type="fig"}) from PRE. Furthermore, the expression of *SOCS3* was positively correlated (R^2^ = 0.5984, p\<0.001) with the expression of *IL-6* across time ([figure S1c](#pone.0017392.s001){ref-type="supplementary-material"}). The cell-cycle related gene *cMyc* was robustly up-regulated from PRE over the entire time course peaking at 4 hours up 15 fold (p\<0.05) ([figure 6a](#pone-0017392-g006){ref-type="fig"}). *cMyc* mRNA was positively correlated with *IL-6* mRNA (R^2^ = 0.2876, p\<0.001) ([figure S1d](#pone.0017392.s001){ref-type="supplementary-material"}) and *SOCS3* mRNA (R^2^ = 0.5406, p\<0.001) ([figure S1e](#pone.0017392.s001){ref-type="supplementary-material"}) illustrating a positive relationship in the temporal expression of these STAT3 target genes. ::: {#pone-0017392-g004 .fig} 10.1371/journal.pone.0017392.g004 Figure 4 ::: {.caption} ###### p-STAT3^+^/Pax7^+^ cells. (**4a**) Representative merged image of PRE at 40× magnification with inset box showing (**a-i**) DAPI^+^ nuclei, (**a-ii**) Pax7^+^ nuclei, (**a-iii**) no p-STAT3 stained nuclei and (**c-iiii**) a merged image. (**4b**) Representatvie merged image of T3 at 40× magnification with inset box showing (**b-i**) DAPI^+^ nuclei, (**b-ii**) Pax7^+^ nuclei, (**b-iii**) p-STAT3^+^ nuclei and (**b-iiii**) a merged image showing co-localiztion. Note that punctate p-STAT3 is not present at PRE but co-localizes with Pax7 at T3 and that there is an increase in diffuse fiber staining that occurs from PRE to T3. (**4c**) Percentage of p-STAT3^+^ SC as quantified over the time course peaking at T24. Values are reported as mean ± S.E.M. \*p\<0.05 vs. PRE. :::  ::: ::: {#pone-0017392-g005 .fig} 10.1371/journal.pone.0017392.g005 Figure 5 ::: {.caption} ###### STAT3 downstream genes. (**5a**) IL-6, (**5b**) IL-6Rα, (**5c**) GP130 and (**5d**) SOCS3 mRNA expression relative to GAPDH, expressed as fold change from PRE. Values are reported as mean ± S.E.M. \*p\<0.05 vs. PRE; † p\<0.05 vs. T1, ‡ vs. T3. :::  ::: ::: {#pone-0017392-g006 .fig} 10.1371/journal.pone.0017392.g006 Figure 6 ::: {.caption} ###### cMyc^+^/Pax7^+^ quantification at Pre and T24. (**6a**) cMyc mRNA relative to GAPDH, expression represented as fold changes from PRE. (**6b**) cMyc^+^ SC as a percentage of Pax7^+^ cells between Pre and T24, n = 7. (**6c & 6d**) Serial sections of the same muscle cross-section with higher magnification (100×) boxes showing the same nuclei in both images. (**6c**) Representative image at 20× magnification of a merged Pax7/Laminin costain where (**c-i**) Pax7 is in green, (**c-ii**) DAPI^+^ nuclei are in blue and (**c-iii**) laminin in red. Note the two Pax7^+^ nuclei which have been highlighted to show their individual Pax7 positivity, DAPI staining and location within the laminin boarder. These nuclei are highlighted again in (**6d**) where the same cells have been stained colourometrically for cMyc. (d-i) A magnified image of the inset box showing the brown cMyc staining co-localized to the nuclear marker hematoxalin The brown cMyc positive cells are clearly within the visible fiber boarder. \*p\<0.05 vs. PRE. :::  ::: To further verify the mRNA data illustrating an increased expression of *cMyc*, immunohistochemical analysis of cMyc was coupled with Pax7 in serial sections ([figure 6c--d](#pone-0017392-g006){ref-type="fig"}) and the number of cMyc^+^/Pax7^+^ cells was quantified PRE and 24 hours following MLC. The percentage of cMyc^+^/Pax7^+^ cells was significantly increased at 24 hours (∼40%) versus PRE (∼9%) ([figure 6b](#pone-0017392-g006){ref-type="fig"}). This provides additional evidence that the STAT3 signalling cascade was active as a consequence of the bout of MLC in SCs leading to proliferation of the SC population. Discussion {#s3} ========== Cell proliferation is a complex process regulated by a number of factors, such as IL-6, IGF-1, HGF and TNFα [@pone.0017392-McKay1], [@pone.0017392-Machida1], [@pone.0017392-Serrano1], [@pone.0017392-OReilly1], [@pone.0017392-Alvarez1], among others. It is a process vital for continued functional capacity and overall tissue survival of skeletal muscle. Skeletal muscle fibres are post-mitotic thus the satellite cell (SC) is the exclusive source of new nuclei for the maintenance of healthy skeletal muscle. Here we illustrate the potential of STAT3 signalling in promoting SC proliferation following acute muscle damage in humans. We have demonstrated that following muscle damage, phosphorylated STAT3 (p-STAT3) in SCs increases early (within one hour), inducing downstream target genes (i.e. *GP130* and *SOCS3*), which further regulate the increase in STAT3 production and response (as induced via IL-6), leading to increased *cMyc* expression, which drives cell proliferation. As SCs account for such a small percentage of total myonuclei (∼2--7%), even when the nuclear fraction of whole muscle homogenate was analyzed, no p-STAT3 was observed. This lack of observable p-STAT3 in the nuclear fraction combined with the increase in p-STAT3^+^/Pax7^+^ cells, measured via immunohistochemistry strongly support the premise that p-STAT3 signalling occurred almost exclusively in SCs in human muscle. Further evidence is provided by the expansion of the cMyc^+^/Pax7^+^ cell population from PRE to 24 hours expressed as a percentage of total Pax7^+^ cells. Collectively, the time-course utilized in this study was successful in capturing a rapid up-regulation in p-STAT3 signalling in SCs shown to be associated with an increase in SC number following exercise induced muscle injury. In agreement with previous studies [@pone.0017392-McKay2], IL-6 mRNA and protein in both the blood and co-localized to the SC increased in response to MLC. The ∼1.5 fold increase in serum IL-6 supports a number of studies reporting that IL-6 is released from the muscle into circulation in response to MLC [@pone.0017392-McKay2], [@pone.0017392-Pedersen1], [@pone.0017392-Fischer1]. *IL-6* mRNA was significantly up-regulated 3 hours and 24 hours post MLC, up ∼150 fold and ∼100 fold respectively. While the fold changes were of a greater magnitude in this study than in McKay and colleagues (2009), most likely due to the dietary controls and the training status of the subjects, the directional temporal pattern of expression was similar between the two studies. Finally, IL-6^+^/Pax7^+^ cells increased from PRE, where there were relatively few co-positive cells, to a peak at 3 hours where approximately 75% of all SCs were IL-6^+^. IL-6^+^/Pax7^+^ cells remained significantly up-regulated from PRE at 24 hours, with approximately 60% of Pax7^+^ cells expressing IL-6. It is known that SCs express the IL-6Rα [@pone.0017392-Keller2], [@pone.0017392-McKay2] allowing for IL-6 signalling to occur. Thus, an increase in SC mediated IL-6 synthesis suggests that IL-6 signalling occurs in an autocrine/paracrine fashion. As a consequence, we suggest that increased IL-6 signalling leads to the induction of p-STAT3 signalling in the SC as evidenced by the increases in the p-STAT3^+^/Pax7^+^ population and the up-regulation of p-STAT3 regulated genes. Previously p-STAT3 was found to be localized in the SC compartment; however, quantification of p-STAT3 over a time-course was not conducted in that study [@pone.0017392-McKay2]. To verify that p-STAT3 regulated the SC response, we quantified p-STAT3 immunofluorescence. The proportion of p-STAT3^+^ SCs was significantly elevated at T1 with ∼40% of SCs co-positive and peaking at T24 where ∼60% of SCs were co-positive, similar to that of IL-6^+^ SCs. Importantly, this early increase in p-STAT3^+^ SCs verifies the rapid signalling of this system and confirms that the timing of these events were indeed occurring early as proposed by McKay et al. [@pone.0017392-McKay2]. This provides temporal evidence that following a bout of damaging exercise, STAT3 is phosphorylated leading to downstream signalling events resulting in proliferation of human SCs. The significant increase in Pax7^+^ cells, the up-regulation of *Myf5* mRNA (a major regulator of SC proliferation) coupled with no significant change in *MRF4* mRNA (a major regulator of SC differentiation) and an increase in the percentage of cMyc^+^ SCs 24 hours after the MLC provide further evidence that human SC proliferation involved p-STAT3 signalling. In addition, because almost no other nuclei other than SCs were positive for p-STAT3 and the nuclear fraction western blots detected virtually no p-STAT3 as they were not sensitive enough to detect such a small percentage of cells expressing p-STAT3protein, it would appear that the IL-6 induced STAT3 signalling cascade occurred almost exclusively in SCs. This supports the observational report from McKay and colleagues who found that p-STAT3 was co-localized to SCs 4 hours after subjects had preformed 300 MLC [@pone.0017392-McKay2] and work done by Kami & Senba who co-localized p-STAT3 to c-Met^+^ cells in rats following freeze crush injury [@pone.0017392-Kami1]. Trenerry and colleagues also showed a transient response in myonuclei expressing p-STAT3 2 hours following leg extension exercise [@pone.0017392-Trenerry1], however no SC specific analysis was conducted in that study. It is also worth noting that there was a progressive increase in p-STAT3 staining in the muscle fibers over the time-course. This agrees with the stepwise increase in p-STAT3 observed in the cytoplasmic fraction. This may be explained by additional cellular functions of p-STAT3. p-STAT3 has been implicated in microtubule remodelling, matrix metalloproteinase production [@pone.0017392-Korzus1], [@pone.0017392-Gao1], and focal adhesion protein production [@pone.0017392-Reich1], [@pone.0017392-Gao1], leading to increased migratory potential of different cell types such as inflammatory and SCs. In addition, p-STAT has also been shown to play a role in regulating the general hypertrophic response through its interactions with the leptin receptor in skeletal muscle [@pone.0017392-Olmedillas1], and angiotensin II in cardiac muscle [@pone.0017392-Yue1]. Possibly, in response to muscle fiber damage, STAT3 is phosphorylated in damaged myofibers acting as a localized chemotactic stimuli for the SC and inflammatory cell populations to migrate toward. This allows for expeditious repair via phagocytosis of necrotic or damaged tissues [@pone.0017392-Kimura1], breakdown of connective tissues through matrix metalloproteinase activity [@pone.0017392-Senft1] and efficient migration of SCs to the areas of damage [@pone.0017392-Walker1]. There was no significant increase in p-JAK2 protein at any of the time points. As an upstream regulator of STAT3 phosphorylation we would expect a temporal profile for JAK2 phosphorylation similar to that of STAT3. Since cytoplasmic p-STAT3 peaked at 24 hours following MLC it is possible that we missed the increase in p-JAK2 as it may have occurred sometime between 3 and 24 hours. In addition to the increase in serum IL-6 and *IL-6* mRNA, we also observed an up-regulation of several STAT3-related downstream genes. *SOCS3* expression, downstream of STAT3 signalling serves as a negative regulator of JAK/STAT signalling. *SOCS3* functions via a classic feedback inhibition loop. Production of SOCS3 is up-regulated in response to JAK/STAT signalling whereby it then acts to suppress JAK/STAT signalling by inhibiting the activity of JAK2 kinase [@pone.0017392-Naka1], [@pone.0017392-Rawlings1]. Importantly, the temporal expression of SOCS3 was closely related to the temporal expression of IL-6 with a peak at 3 hours and remaining elevated at 24 hours post MLC. *cMyc*, a gene known to promote proliferation, and that is expressed downstream of STAT3 [@pone.0017392-Kiuchi1], was also associated with the same pattern of temporal expression as IL-6 and SOCS3 following MLC, as demonstrated through significant correlations with each of these genes ([figures S1d](#pone.0017392.s001){ref-type="supplementary-material"}--e). Two other necessary members of this signalling cascade are the receptor sub-complexes *IL-6Rα* and *GP130*. They both revealed increases in their mRNA expression over the time course with a peak at 24 hours. This is important as both of these mRNA species are up-regulated in response to STAT3 signalling and serve to autoregulate the pathway. Providing yet another level of support for STAT3 mediated SC proliferation was the increase in the number of cMyc^+^/Pax7^+^ cells observed at T24 versus PRE. Previous work by McKay and colleagues [@pone.0017392-McKay2] did not localize the downstream signalling events of the STAT3 pathway to the SC, looking only at different mRNA species in whole muscle. Importantly, this work definitively shows that the change observed in *cMyc* mRNA translates into significant protein expression specifically in the SCs during a time when these cells were proliferating. This implicates STAT3 as a primary inducer of SC proliferation in human muscle following MLC. The cMyc^+^/Pax7^+^ population was only analysed PRE and 24 hours post MLC as this was the time when the highest levels of proliferation were occurring as evidenced by the significant increase in Pax7^+^ cells. cMyc is a critical regulator of the cell cycle, responsible for the transition from G1 to S phase [@pone.0017392-Dang1]. It has also been shown to be up-regulated in response to IL-6 signalling [@pone.0017392-Nabata1] and in a STAT3 induced GP130 mediated manner [@pone.0017392-Kiuchi1]. One of its main downstream functions is to regulate the production of CyclinD1, a known regulator of the cell cycle, which we have previously reported to increase in response MLC. The increase in SCs expressing cMyc serve as direct evidence that p-STAT3 signalling is occurring in SCs leading to their proliferation. The STAT3 signalling cascade as induced by IL-6 has been shown here to be an important regulator for the proliferation of SCs early following a muscle damaging protocol. Collectively, we have shown that SCs up-regulate IL-6 following damage which likely acts in an autocrine/paracrine fashion to promote proliferation through STAT3 signalling. This IL-6 induced STAT3 signalling is evidenced by the up-regulation of the downstream genes regulated by p-STAT3. Furthermore, with p-STAT3 being observed almost exclusively in SCs, not in other myonuclei and only to a large degree at least 1 hour post muscle damage, it appears that it acts in a very specific manner through IL-6 leading to the proliferation of SCs and subsequent repair of muscle damage. Thus, as evidenced by the induction of cMyc protein it appears that STAT3 is a key signalling molecule in human SCs in response to physiological levels of muscle damage and contributes to the robust proliferation of SCs in the acute period following damage. Materials and Methods {#s4} ===================== Ethics Statement {#s4a} ---------------- All subjects were informed of the procedures and the potential risks associated with the study and gave written informed consent. This study was approved by the Hamilton Health Sciences Research Ethics Board (08-413) and conformed to the Declaration of Helsinki regarding the use of human subjects as research participants. Subjects {#s4b} -------- Twelve healthy males (age 21.2±1.6 yrs, height 178.2±5.5 cm and weight 82.6±11.5 kg) were recruited from the McMaster University community. Subjects were sedentary having done no lower body resistance exercise for at least the past 6 months and were non-smokers. Subjects were told to refrain from doing any moderate or strenuous exercise for two days prior to and during the study. Muscle Damage Protocol {#s4c} ---------------------- The muscle damage protocol was preformed as described previously [@pone.0017392-McKay3]. Briefly, subjects completed 300 unilateral isokinetic eccentric contractions of the *quadraceps femoris* using a Biodex dynamometer (Biodex-System 3, Biodex Medical Systems, Inc., Shirley USA) at 180°•s^−1^ over a 55° range of motion [@pone.0017392-McKay2], [@pone.0017392-McKay1], [@pone.0017392-OReilly1]. The experimental leg was chosen as the dominant leg. During each set investigators provided verbal encouragement so as to elicit maximal effort from the subjects. The entire duration of the protocol was approximately 30 min of which the muscle was under tension for about 10 min. The protocol employed in the present study has previously been shown to cause physiological muscle damage as evidenced by increased serum creatine kinase (CK) [@pone.0017392-McKay3]--[@pone.0017392-Beaton1], z-band streaming [@pone.0017392-Gibala1], [@pone.0017392-Gibala2], desmin disruption, a significant infiltration of the inflammatory related macrophages and neutrophils, as well as a significant myogenic response leading to an increase in satellite cell number based on both Pax7 and NCAM staining [@pone.0017392-McKay2], [@pone.0017392-McKay1], [@pone.0017392-OReilly1]. Muscle Biopsies {#s4d} --------------- Subjects reported to the lab at either 7, 7:30, or 8am having completed an overnight fast. Upon arrival to the lab subjects rested, had their PRE blood draw taken from the *antecubital vein* and then underwent the muscle damaging protocol. Once the subjects had completed the protocol they had one hour to rest. One hour post muscle damage the subjects had a second blood draw taken as well as two muscle biopsies; one from the exercise leg (T1) and one from the control leg (PRE). Three hours after the protocol the subjects had a third blood draw taken along with another biopsy from their exercise leg (T3). Subjects were then given a meal replacement beverage (Ensure, Abbott Laboratories, Abbott Park, Illinois, U.S.A.) The subjects were instructed to refrain from taking any anti-inflammatory drugs or doing any vigorous exercise. The subjects arrived the next morning at 8, 8:30, or 9am in a fasted state for the final (T24) blood draw and biopsy from the exercised leg. Muscle biopsies were obtained using the percutaneous needle [@pone.0017392-Bergstrom1], [@pone.0017392-Hennessey1] method using manual suction from the vastus lateralis under local anaesthetic (1% lidocaine) as described previously [@pone.0017392-Bourgeois1]. In total, four muscle biopsies were obtained from each subject, three in the exercise leg and one in the control leg (non-exercise leg). The muscle tissue was dissected free of adipose and connective tissue. The tissue was immediately divided into four sections for RNA and protein (western blotting and immunohistochemical) analysis, three of which were flash-frozen in liquid nitrogen and stored at −80°C for later analysis and the final piece embedded in Optimal Cutting Temperature (OCT) compound embedding medium and frozen in liquid nitrogen cooled isopentane. Blood Measures {#s4e} -------------- Blood samples were obtained from the antecubital vein immediately prior to the intervention and then again concurrently with the muscle biopsies one, three, and twenty-four hours after the muscle damage protocol. Approximately 8 mL of blood was taken from each subject at each time point. 4 mL was drawn into a heparinised tube while the other 4 mL was drawn into a non-heparinized tube to obtain plasma and serum samples respectively. Samples were allowed to sit on ice (plasma) or at room temperature (serum) for 15 minutes, centrifuged at 4000 RPM for 15 minutes, aliquoted into 600 µL aliquots and frozen down at −80°C for later analysis. Serum analysis was conducted for both creatine kinase activity and IL-6 protein. Creatine kinase activity was analysed using a commercially available kit (Pointe Scientific, Inc., Canton USA) with modifications to the protocol to allow for running the samples at 25°C. This included adding additional sample using a ratio obtained from the International Federation of Clinical Chemistry (IFCC) for conversion between 37°C and 25°C. Additionally, sample absorbance was measured every twenty seconds for twenty minutes to obtain the slope of the change in absorbance/minute. This value was used to calculate the concentration of creatine kinase in international units (U/L) which is defined as the amount of enzyme that catalyzes the transformation of one micromole of substrate per minute. Serum IL-6 was analysed using a commercially available high sensitivity Quantikine Enzyme-Linked ImmunoSorbent Assay (ELISA) kit according to the manufacturer\'s instructions (R&D systems Inc., Minneapolis USA). Samples were run in duplicate with all subjects on the same plate and an intra-assay CV of 6.9%. RNA Isolation {#s4f} ------------- RNA was isolated was conducted as previously described [@pone.0017392-McKay2]. RNA isolation was conducted using Trizol Reagent (Invitrogen Corporation, Carlsbad USA) and RNA purification was done using the Qiagen RNeasy mini kit (Qiagen Sciences, Mississauga Canada). The RNA was then quantified using a spectrophotometer (NanoDrop 1000, Thermo Fisher Scientific Inc., Wilmington USA). Reverse Transcription (RT) {#s4g} -------------------------- RNA was transcribed to cDNA using Applied Biosciences High Capacity cDNA reverse Transcription Kit (Applied Bioscience, Foster City USA). Individual samples in 20 µL reactions were reverse transcribed into cDNA using an Eppendorf Mastercycle epigradient thermal cycler (Eppendorf, Mississauga Canada). Quantitative Real-Time Polymerase Chain-Reaction (qRT-PCR) {#s4h} ---------------------------------------------------------- qRT-PCR was performed in 25 µL reactions using SYBR Green/Rox master mix (SuperArray Bioscience Corp., Frederick USA). Primers were custom made using published sequences ([table S1](#pone.0017392.s002){ref-type="supplementary-material"}). They were resuspened in 1× TE buffer (10 mM Tris-HCL, 0.11 mM EDTA) and frozen at −20°C until use. Using 0.2 mL PCR tubes (Axygen Inc., Union City USA) 12.5 µL of SYBR green, 2 µL forward primer, 2 µL reverse primer, 6.5 µL or 7.5 µL of H~2~O depending on the cDNA template volume, 1 µL or 2 µL was added depending on the amount necessary for the particular gene of interest (25 ng cDNA or 50 ng cDNA). qRT-PCR was performed using a Stratagene Mx3000P real-time PCR system (Stratagene, Santa Clara USA) and Stratagene MXPro QPCR Software Version 3.00 (Stratagene, Santa Clara USA). Changes in gene expression over time were expressed as fold changes ± SEM from pre values using the delta delta CT method [@pone.0017392-Schmittgen1] with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as a housekeeping gene. GAPDH expression was not different from PRE at any of the post-intervention time-points. Nuclear and Cytoplasmic Extraction {#s4i} ---------------------------------- Nuclear and cytoplasmic extraction was performed using a commercially available kit with minor modifications (NE-PER Nuclear and Cytoplasmic Extraction Reagents -- Thermo Fischer Scientific Inc., Wilmington USA). Between 20 mg and 60 mg of tissue were transferred into a 2 mL microcentrifuge tube while still frozen. 10× the volume in µL of the weight of the sample was added to the sample for CER I buffer (from kit with the addition of one protease and one phosphatase tablet dissolved in it). The muscle was then minced using mincing scissors in the 2 mL microcentrifuge tube for about 15 seconds. The sample was then homogenized four times for five seconds using a rotary homogenizer (PRO250, PRO Scientific Inc., Oxford USA). The homogenate was then transferred to a 1.5 mL microcentrifuge tube and vortexed for fifteen seconds. Following vortexing, 20 µL of CER II buffer was added per sample (from kit with the addition of half a protease and half a phosphatase tablet dissolved). The sample was then vortexed for five seconds and put on ice for one minute. The tube was again vortexed for five seconds and then centrifuged for five minutes at 16 000 g at 4°C. The supernatant (cytoplasmic extract) was immediately transferred to pre-chilled 1.5 mL microcentrifuge tubes and frozen down at −80°C. The remaining pellet was then washed with 400 µL of a PBS cocktail containing one protease and one phosphatase tablet dissolved, spun for five minutes at 16 000 g at 4°C. The liquid was decanted with the wash repeated three more times. On the last wash the pellet was resuspended prior to the final spin. After the final spin the pellet was suspended in 200 µL of ice-cold NER buffer (from kit with the addition of 50 µL of 10% SDS). The pellet was then broken up using Teflon pestles. The samples were then vortexed for 15 seconds and placed on ice for ten minutes. This was repeated three more times. The samples were then centrifuged at 16 000 g for ten minutes at 4°C. The supernatant (nuclear extract) was then immediately transferred to a pre-chilled 1.5 mL microcentrifuge tube and frozen down at −80°C. Bradford analysis was then conducted on the samples to obtain the concentration using a spectrophotometer (UltraSpec 300 pro, Biochrome Ltd., Cambridge UK). Western Blotting {#s4j} ---------------- Equal amounts (75 µg) of cytoplasmic or nuclear homogenate in 4× Laemlli buffer were boiled at 95°C for 5 minutes then loaded in the wells of at 7.5% gel. Nuclear samples were precipitated via the acetone precipitation [@pone.0017392-Wessel1] method to increase their concentrations so that 75 µg of nuclear protein could be loaded in each well. Briefly, samples were mixed with four volumes of ice cold acetone and incubated for 1 hour. Samples were spun at 13,000 g for 10 minutes and the resulting pellet was dried for 20 minutes. Dried pellets were reconstituted in 4× Laemlli buffer and ddH~2~O. Phosphorylated and total Jak2 were analyzed in the cytoplasm only, while phosphorylated and total Stat3 were analyzed in both the nucleus and cytoplasm. Nuclear p-STAT3 was run with cytoplasmic pSTAT3 from the same subject. Gels were run at 125 V for approximately 1 hour, and then transferred to polyvinylidene fluoride (PVDF; Millipore, Etobicoke, Canada) membranes at 70 V for 1 hour. Membranes were blocked with 5% non-fat powdered milk in PBS (10 mM, pH 7.4) for 1 hour at 4°C, then incubated in primary, phospho-specific antibody (phospho-Stat3 Tyr705, 1∶1000, and phospho-Jak2 Tyr1007/1008, 1∶500, Cell Signaling Technology, Boston, USA) overnight in 5% bovine serum albumin for cytosolic fractions and 5% non-fat dry milk in PBS for nuclear STAT3 blots (BSA, Santa Cruz Biotechnology, Santa Cruz, USA) at 4°C. After multiple washes, blots were incubated in goat anti-rabbit HRP (1∶50,000; Abcam Inc., Cambridge USA) in 5% BSA for 90 minutes at room temperature. After multiple washes, proteins were detected with ECL (SuperSignal West Dura; Thermo Fisher Scientific, Rockford, USA) using FluorChem SP (Alpha Innotech Corporation, San Leandro, USA). After detection of phosphorylated proteins, blots were washed and stripped with Restore Western Blot Stripping Buffer (Thermo Fisher Scientific, Rockford, USA) for 20 minutes at room temperature. Membranes were washed and re-probed with total-specific antibodies (Stat3, 1∶1000, and Jak2 D2E12 rabbit mAb, 1∶500, Cell Signaling Techonlogy, Boston, USA) in the same manner as phospho-specific protein detection. Following total protein detection, ponceau staining confirmed equal loading. Protein bands corresponding to the predicted molecular weight of Jak2 (∼125 kDa) and Stat3 (∼86 kDa) were quantified using the AlphaEase FC Software, Version 5.0.2 (Alpha Innotech Corporation, San Leandro, USA) with background correction. As we were interested in determining the changes in the amount of activated protein, the ratio of phosphorylated to total protein was determined. Immunohistochemistry {#s4k} -------------------- 7 µm muscle cross-sections were stained with antibodies against Pax7 (neat; cell supernatant from cells obtained from the DSHB, Iowa City USA); IL-6 (500 ng/mL, MAB 2061, R&D Systems, Minneapolis USA); p-STAT3 (1∶100, Cell Signaling Technologies Inc., Danvers USA) and laminin (1∶1000, L8271, Sigma-Aldrich, Oakville Canada). Secondary antibodies used were: Pax7 (AlexaFluor 488 or AlexaFluor 594, 1∶500, Invitrogen, Molecular Probes Inc., Camarillo USA or when using two mouse primary antibodies an immunoglobulin biotinylated secondary antibody, 1∶200, Dako Inc., Mississauga Canada; followed by a streptavidin-FITC fluorochrome, 1∶100, Carlsbad Biosource. USA); and Laminin (AlexaFluor 594, 1∶500, Invitrogen, Molecular Probes Inc., Camarillo USA). Histochemical methods were adapted from previously published methods from our lab [@pone.0017392-McKay3], [@pone.0017392-McKay1], [@pone.0017392-OReilly1], [@pone.0017392-McKay2]. Briefly, for co-immunofluorescent staining (Pax7 and IL-6, Pax7 and Laminin), sections were fixed with 2% paraformaldehyde (PFA, Sigma-Aldrich, Oakville Canada) for 10 min followed by several washes in PBS. Sections were then covered for 60 min in a blocking solution containing, 2% BSA, 5% FBS, 0.2% Triton-X 100, 0.1% sodium azide. Following blocking, sections were incubated in the primary antibody at 4°C overnight. After several washes, sections were then incubated in the appropriate secondary antibodies. Sections were then re-fixed in 2% PFA (Sigma-Aldrich, Oakville Canada) to prevent migration of the secondary antibodies and re-blocked in 10% GS in 0.01% Triton-X 100 (Sigma-Aldrich, Oakville Canada). The sections were then incubated in the second primary antibody, followed by incubation in the appropriate secondary antibody. Sections were then washed with PBS and 4′,6-diamidino-2-phenylindole (DAPI, 1∶20000) (Sigma-Aldrich, Oakville Canada) for nuclear staining. Staining was verified using the appropriate positive and negative controls to ensure specificity of staining. Stained slides were viewed with the Nikon Eclipse 90*i* Microscope (Nikon Instruments, Inc., Melville USA) and images were captured and analyzed using the Nikon NIS Elements 3.0 software (Nikon Instruments, Inc., Melville USA). For cMyc immunodetection, serial sections were used to visualize both Pax7 (with Laminin) and cMyc. A secondary-only control image has been included to demonstrate the specificity of the cMyc antibody ([Figure S1f](#pone.0017392.s001){ref-type="supplementary-material"}). Pax7/Laminin staining was performed as described above, and cMyc was stained as follows: slides were dried and fixed in acetone for 10 minutes. Slides were then washed several times in 1× PBS followed by quenching of endogenous peroxidases for 30 minutes using 0.3% H~2~O~2~ solution. After washing, sections were then blocked in 10% goat serum in a 0.2% Triton-X 100 (Sigma-Aldrich, Oakville Canada) solution for 30 minutes. cMyc antibody (cell supernatant, DSHB, Iowa City USA) was used at 1∶2 in a 1% BSA solution and was then incubated for 2 hours at room temperature. Slides were then washed several times in PBST. The secondary IGB goat anti-mouse (DAKO Inc., Mississauga Canada) was incubated at 1∶200 for an hour followed by incubation in the Vectastain Elite ABC kit (Vector Laboratories, Burlington Canada) according to the manufacturer\'s instructions for 30 minutes. Following several washes in PBST, slides were developed using the DAB kit (Vector Laboratories, Burlington Canada) according to the manufacturer\'s instructions. Slides were counterstained using Mayer\'s hematoxylin (Sigma-Aldrich, Oakville Canada). Immunohistochemical Analysis {#s4l} ---------------------------- Immunohistochemical quantification and enumeration was performed at all time points (PRE, T1, T3, T24) for IL6^+^/Pax7^+^ (n = 12) and pSTAT3^+^/Pax7^+^ (n = 9) and images were taken at 40× magnification. For cMyc^+^/Pax7^+^ stain 7 subjects were analysed (n = 7) at PRE and T24 with images being taken at 20× magnification using full muscle cross-section stitched images. Two separate blinded reviewers quantified the co-localization of Pax7 and cMyc. Slides were viewed and images captured with the Nikon Eclipse 90*i* Microscope (Nikon Instruments, Inc., Melville USA) and Nikon NIS Elements 3.0 software (Nikon Instruments, Inc., Melville USA). Immunohistochemical quantification {#s4m} ---------------------------------- Satellite cells were enumerated via double labelling with an anti-Pax7 antibody and DAPI. At least 300 myonuclei per timepoint were counted. Only those cells that were co-positive were counted as satellite cells. Furthermore, only cells associated with individual myofibers, that were not in the interstitial space, were counted as positive. IL6^+^/Pax7^+^ cells were only counted as positive if they were triple immunolabeled with DAPI and antibodies against IL-6 and Pax7. pSTAT3^+^/Pax7^+^ cells were only counted as positive if they were triple immunolabeled with DAPI and antibodies against pSTAT3 and Pax7. cMyc^+^/Pax7^+^ cells were only counted as positive if they were triple immunolabeled with Mayers Hemotoxylin and antibodies again cMyc and Pax7 with Laminin defining the SC niche. Non-fiber associated nuclei (interstitial nuclei) were not included in the quantification. All data are represented as a percentage of Pax7^+^ cells. Statistical Analysis {#s4n} -------------------- Statistical analysis and graphing were performed using Sigmstat 3.1.0 analysis software (Systat, SPSS Inc., San Jose USA) and Prism5 for Windows - version 5.01 (GraphPad Software Inc., La Jolla USA). mRNA, protein, IL-6 and creatine kinase plasma concentrations, IL-6^+^/Pax7^+^ and pSTAT^+^/Pax7^+^ enumeration were analysed using a 1-way repeated measures analysis of variance (ANOVA). cMyc^+^/Pax7^+^ enumeration was analysed via a two-tailed T-test. Statistical significance was set at P\<0.05. Tukey\'s HSD post hoc test was used to analyse main effects and significant interactions. Results are presented as mean ±SEM. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### (**S1a**) Pearson correlation between the serum concentrations of IL-6 (pg/mL) and cMyc (U/L); R^2^ = 0.3055; p\<0.001. The correlation is representative of the individual data points presented as mean values ± SD (error bars). (**S1b**) Representative image of nuclear and cytoplasmic preparations with the cytoplasmic marker LDH present only in the cytoplasm and the nuclear marker p84 present only in the nuclear fraction. (**S1c**) Pearson correlation between the mRNA regulation (fold change) of IL-6 and SOCS3; R^2^ = 0.5984, p\<0.001. The correlation is representative of the individual data points presented as mean values ± SD (error bars). (**S1d**) Pearson correlation between the mRNA regulation (fold change) of IL-6 and cMyc; R^2^ = 0.2876, p\<0.001. The correlation is representative of the individual data points presented as mean values ± SD (error bars). (**S1e**) Pearson correlation between the mRNA regulation (fold change) of SOCS3 and cMyc; R^2^ = 0.5406, p\<0.001. The correlation is representative of the individual data points presented as mean values ± SD (error bars). (**S1f**) Representative image of a muscle cross section stained with only the secondary and hematoxylin during the cMyc staining protocol. Note that there is no brown colour change indicating no false positivity caused from the addition of the secondary antibody. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### mRNA species that were analysed with their forward and reverse sequences, cDNA (ng) concentration used and annealing temperature (°C). (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to thank Todd Prior for his technical assistance. The Pax7 hybridoma cells developed by Dr. A. Kawakami and the cMyc cell supernatant were obtained from the Developmental Studies Hybridoma Bank (DSHB) developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biology, Iowa City, IA 52242 USA. **Competing Interests:**Dr. Gianni Parise is a PLoS ONE Academic Editor. The authors declare that no other competing interests exist. **Funding:**Funding for this study was provided by CIHR (86830: [www.cihr.ca](http://www.cihr.ca)) and NSERC (327073: [www.nserc.ca](http://www.nserc.ca)) grants awarded to G.P. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: KGT BRM GP. Performed the experiments: KGT BRM MDL JPL MAT GP. Analyzed the data: KGT BRM MDL JPL. Contributed reagents/materials/analysis tools: MDL GP. Wrote the paper: KGT BRM MDL GP. Manuscript review: KGT BRM MDL JPL MAT GP. Developed muscle fractionation protocol: KGT MDL JPL.

PubMed Central

2024-06-05T04:04:19.095959

2011-3-9

PMC3052299

Introduction {#s1} ============ The global energy sector is trending toward incorporation of increasing amounts of renewable energy, of which bioenergy---energy yielded from biological sources---is a growing component [@pone.0017222-Somerville1]. The United States (US) currently produces 4% (3.2 EJ) of its total energy from biomass [@pone.0017222-Lemus1], but has mandated 136 billion liters of renewable liquid transportation fuels by 2022, which may require up to 60 million additional hectares of land [@pone.0017222-Robertson1]. This additional cropland will not be evenly distributed across the US due to climatic variation, land availability, and resource requirements (e.g., irrigation). The US Department of Agriculture (USDA) estimates that nearly 50% of the biomass needed to meet the Renewable Fuel Standard will be grown in the Southeast, with an additional 43% in the Central-Eastern US [@pone.0017222-USDA1]. However, identifying crops capable of producing high yields on marginal lands or degraded soils with minimal inputs will be a tremendous challenge to the sustainability of the bioenergy industry globally [@pone.0017222-Lemus1]. Identifying regions and the climatic suitability of proposed biofuel species within targeted regions will aid selection of the most appropriate bioenergy crops that require the fewest inputs. For example, the highest recorded yields occur in the Amazon floodplain for *Echinochloa polystachya* (100 MT ha^−1^ yr^−1^) and *Pennisetum purpureum* (88 MT ha^−1^ yr^−1^) [@pone.0017222-Somerville1]. Perennial grasses using the C~4~ photosynthetic pathway---*Panicum virgatum* (switchgrass), *Miscanthus* spp., *Saccharum* spp. (sugarcane), and *Pennisetum* spp.---are intrinsically nutrient, light, and water use efficient, especially in the humid warm regions of the globe. Additionally, fast growing trees that are harvested or coppiced on short rotations have the potential to provide high quality biomass [@pone.0017222-Karp1]. Several studies have provided yield estimates or habitat suitability of select crops in certain parts of the world [@pone.0017222-Somerville1], [@pone.0017222-Heaton1], [@pone.0017222-Angelini1], [@pone.0017222-Sanderson1], [@pone.0017222-Evans1], which begins to address the need for choosing appropriate crops that require minimal inputs. However, no global assessment of large-scale suitability for a variety of herbaceous, grass, and woody species has been conducted. One additional complicating factor the bioenergy industry faces in achieving agronomic and economic goals is to prevent unintentionally introducing invasive species to susceptible natural or managed ecosystems [@pone.0017222-Barney1]. The desirable traits that bioenergy crops must possess---rapid growth rates, high annual yields with minimal inputs of pesticides, fertilizers, and irrigation, tolerance of poor growing conditions---typify the invasive species ideotype [@pone.0017222-Barney1], [@pone.0017222-Raghu1]. In fact, some of the taxa undergoing agronomic field trials are known invasive species in some portion of their introduced range [@pone.0017222-Low1], and exhibit an unknown risk to other environments [@pone.0017222-Barney1]. Some attempts have been made to evaluate the risk posed by some bioenergy crops in their target region by using the Pheloung Weed Risk Assessment (WRA) model [@pone.0017222-Barney1], [@pone.0017222-Crosti1], [@pone.0017222-Buddenhagen1], [@pone.0017222-Gordon1], which includes, in part, an assessment of climatic suitability [@pone.0017222-Pheloung1]. The authors have consistently found that the majority of the proposed bioenergy crops present an unacceptable level of invasion risk in their respective target regions according to this risk assessment. However, this WRA was designed as a pre-introduction evaluation for plants that are largely introduced for ornamental or horticultural purposes, and as such may be less relevant for bioenergy crops, or worse, may needlessly restrict adoption of "safe" crops due to misuse of an inappropriate risk assessment [@pone.0017222-Barney1]. Additionally, one critical, yet almost always overlooked aspect of a risk assessment, is the evaluation of suitable habitat [@pone.0017222-Barney1]. There is no possibility for invasion if the climate of the target region is unsuitable [@pone.0017222-Barney2]. Similarly, there is no possibility for agronomic production if the climate is not suitable. In the case of bioenergy crops, the climate niche represents both the region (possibly) suitable for agronomic production, as well as the regions (possibly) suitable for establishment outside of cultivation [@pone.0017222-Barney3]. For example, the Southeastern US was the focus of kudzu (*Pueraria montana* var. *lobata* (Willd.) Maesen & S. M. Almeida ex Sanjappa & Predeep) introduction for soil stabilization and forage in the early 20^th^ century, as this region was climatically suitable based on the native range in Japan [@pone.0017222-Forseth1]. The favorable climate of the Southeastern US did not provide a barrier to surviving outside cultivation [@pone.0017222-Mack1], while the originally desirable characteristics of rapid establishment and high growth rates contributed to the ultimate invasion of kudzu over 2.8 million hectares [@pone.0017222-Forseth1]. Therefore, comparing the climate niche of invasive species of agronomic origin with the climate niche of bioenergy crops may elucidate patterns and regions that could be the focus of screening and monitoring for escapes. This study aims to provide global climate niche estimates for the leading bioenergy crops, as well as for invasive species that have an agronomic origin (i.e., were introduced as a forage or agricultural crop). Our objectives were to: 1) evaluate the global climate niche for grass, herbaceous, and woody bioenergy crops, 2) compare the bioenergy crop climate niche with invasive species that were widely introduced for agronomic purposes, and 3) compare the ecoregional distribution of both bioenergy crops and invasive species in North America, especially the continental US. Materials and Methods {#s2} ===================== Species data {#s2a} ------------ Estimating the fundamental niche, or climate-driven range, can be performed using the native range, introduced range, or both with various advantages and disadvantages [@pone.0017222-Bradley1], [@pone.0017222-Beaumont1]. Using only the native range is the most restrictive estimate, but likely ignores the boundaries of the climate niche, which could only be elucidated by using the introduced range. Alternatively, the entire range (native+introduced) will best estimate a species fundamental niche at the risk of over-estimating the range potential in regions where the taxon is not yet introduced [@pone.0017222-Bradley1]. Since our goal was to provide a conservative estimate of the range potential for each species we chose to use the entire range (i.e., native and introduced) in our modeling. The Global Biodiversity Information Facility ([www.gbif.org](http://www.gbif.org)) hosts a data portal of natural history collections across the globe, which is available for download. We accessed the portal (February 2010) for each species in our study, which primarily comprises herbarium collections with label data, and used only those collections with geolocations. Population location files ranged from 93 to 58,115 records with widely introduced crops (e.g., *Medicago sativa*) and weeds (e.g., *Schedonorous phoenix = Festuca arundinacea*) having the most collections, and more recently introduced species (e.g., *Jatropha curcas*) having many fewer recorded populations. Collections included wild, cultivated, ornamental, and irrigated locations, which were taken into consideration while fitting the model (ie, cultivated locations were not used to guide fitting the climate niche as cultivation can mitigate environmental stochasticity [@pone.0017222-Mack1]). CLIMEX {#s2b} ------ We used the CLIMEX software to estimate the fundamental niche for each species, which utilizes the distribution and abundance of known populations to parameterize a climatic model [@pone.0017222-Sutherst1]. CLIMEX is flexible in allowing model parameterization by visually matching the output to conform to the known distribution, while also allowing basic biological information to drive parameter estimation [@pone.0017222-Barney3]. CLIMEX calculates a growth index where population growth is positive, and a stress index where population growth declines or is zero, each of which comprises sub-indices, based on the input parameters and climate [@pone.0017222-Sutherst1]. The Ecoclimatic Index (EI) is the synthetic measure of the growth and stress indices and ranges between 0 and 100. Regions with an EI≤10 are very stressful and unlikely to support a population, while an EI\>20 is favorable for population growth and an EI\>30 represents a region able to support substantial population densities [@pone.0017222-Olfert1], [@pone.0017222-Sutherst2], [@pone.0017222-Mika1]. For this study, model output was visually estimated to match the current distribution (i.e., high EI values where population density is highest, and low EI values where no known populations exist). Parameters were subsequently refined using biological information, if any existed, from the primary literature. The parameters were then adjusted iteratively to yield a model that most closely matches the distribution and abundance of both native and introduced populations globally, while always attempting to minimize overestimation. Therefore, we set a threshold of ≥80% of GBIF collections must occur within 'favorable' to 'very favorable' regions (i.e., EI≥20). Many species are current crops (e.g., turfgrasses, agronomic crops, ornamental plantings) that receive irrigation in some portion of their range, or species that occur in riparian areas (e.g., *Arundo donax*, *Phalaris aquatica*). Therefore, in most cases a second simulation was run with the "Permanent Water Scenario" by adding 9.6 mm day^−1^ to simulate the effects of agronomic irrigation or areas with a perennial source of water (e.g., streams, irrigation canals, and wetlands) [@pone.0017222-Barney3], [@pone.0017222-Pattison1]. Regions with an EI≥20 from the irrigation scenario were added to the final maps to exhibit regions suitable with a water subsidy. To create niche maps, CLIMEX results were exported to the geographic information system Manifold 8.0 (Carson City, NV) where Kriging was performed and contours generated for each EI level. The area of each contour was not calculated, as this value would be a gross overestimate of the actual range potential of each species because it represents the fundamental niche, which does not consider biotic interactions, edaphic conditions, disturbance regimes, land use, or trophic dynamics. One of our ultimate objectives relates to using the niche maps for risk assessment at sub-national boundaries. We have chosen ecoregions, which represent regions of repeating patterns of characteristic associations of soil and landforms that include the biota (including humans), geology, physiography, hydrology, and climate, at the scale of interest [@pone.0017222-Omernik1]. The International Commission for Environmental Cooperation ([www.cec.org](http://www.cec.org)) delineated three levels of ecoregions, of which we are using Level II, which comprises 50 types in North America and 20 in the Continental US, and best captures the desired level of spatial scale and utility. This ecoregional designation seems most appropriate given the scale, as well as being promoted by the US Environmental Protection Agency (EPA) ([www.epa.gov/wed/pages/ecoregions/na\_eco.htm](http://www.epa.gov/wed/pages/ecoregions/na_eco.htm)). CLIMEX niche maps were overlaid on the Level II ecoregions in Manifold and the number of GBIF populations contained within each ecoregion were calculated, as was the presence of 'very favorable' habitat (EI≥30) for each species. Results {#s3} ======= The CLIMEX model performed well, achieving ≥80% inclusion of global populations at an EI≥20 for all bioenergy crops and invasive species ([Tables 1](#pone-0017222-t001){ref-type="table"}, [2](#pone-0017222-t002){ref-type="table"}, [3](#pone-0017222-t003){ref-type="table"}). Model accuracy was positively correlated with the number of global population records (P = 0.072), especially for the forage species *Dactylis glomerata*, *Elytrigia repens*, *Medicago sativa*, *Phalaris arundinacea*, and *Schedonorus phoenix*, which had records ranging from 24,978 to 58,115. ::: {#pone-0017222-t001 .table-wrap} 10.1371/journal.pone.0017222.t001 Table 1 ::: {.caption} ###### CLIMEX parameters and values, number of records used in the analysis, and model accuracy for the eight perennial and one annual (*Sorghum bicolor*) grass biofuel feedstock crops. ::: {#pone-0017222-t001-1} Parameter *Arundo donax* *Miscanthus×giganteus* *Miscanthus sacchariflorus* *Miscanthus sinensis* *Panicum virgatum* *Pennisetum purpureum* *Phalaris arundinacea* *Saccharum officinarum* *Sorghum bicolor* ------------- ---------------- ------------------------ ----------------------------- ----------------------- -------------------- ------------------------ ------------------------ ------------------------- ------------------- DV0 10°C 8°C 10°C 10°C 10°C 15°C 5°C 15°C 5°C DV1 20°C 16°C 15°C 20°C 20°C 25°C 8°C 23°C 12°C DV2 35°C 30°C 28°C 30°C 30°C 40°C 27°C 33°C 34°C DV3 40°C 35°C 30°C 35°C 35°C 42°C 30°C 36°C 40°C SM0 0.1 0.19 0.1 0.1 0.1 0.2 0.1 0.35 0.01 SM1 0.2 0.3 0.2 0.2 0.2 0.7 0.4 7 0.1 SM2 2 1 1 1 1 1.5 2 1.5 0.6 SM3 10 10 10 10 10 2.5 10 10 10 TTCS 0°C 0°C −5°C 0°C 10°C 10°C \- 6°C −3°C THCS −0.0005 −0.0003 −0.0003 −0.0003 −0.00001 −0.0009 \- −0.01 −0.0005 TTHS 40°C 35°C 32°C 35°C 35°C 42°C 40°C 40°C 45°C THHS 0.002 0.01 0.05 0.01 0.01 0.0002 0.002 0.0002 0.005 SMDS 0.01 0.1 0.1 0.1 0.1 0.1 0.02 0.25 0.01 HDS −0.005 −0.02 −0.009 −0.02 −0.02 −0.0001 −0.005 −0.01 −0.0005 SMWS \- \- \- \- \- 2.5 \- \- \- HWS \- \- \- \- \- 0.002 \- \- \- TTHW 35°C 33°C 35°C 35°C \- \- \- \- \- MTHW 1 0.5 0.5 0.5 \- \- \- \- \- PHW 0.075 0.05 0.05 0.05 \- \- \- \- \- *N* 6819 \- 142 338 1714 593 49996 93 1096 EI≥30 98.7% \- 90.1% 90.2% 78.7% 85.0% 99.0% 77.4% 94.0% EI≥20 99.3% \- 97.9% 97.3% 87.4% 91.2% 99.4% 79.6% 97.2% EI≥20+water \- \- 97.9% 97.3% 96.6% \- 99.8% 87.1% \- ::: ::: {#pone-0017222-t002 .table-wrap} 10.1371/journal.pone.0017222.t002 Table 2 ::: {.caption} ###### CLIMEX parameters and values, number of records used in the analysis, and model accuracy for the four woody (first four) and three herbaceous dicots (last three) biofuel feedstock crops. ::: {#pone-0017222-t002-2} Parameter Description *Eucalyptus globulus* *Jatropha curcas* *Paulownia tomentosa* *Triadica sebifera* *Nicotiana tabacum* *Medicago sativa* *Pueraria montana* ----------- ----------------------------------- ----------------------- ------------------- ----------------------- --------------------- --------------------- ------------------- -------------------- DV0 Limiting low temperature 8°C 15°C 8°C 12°C 10°C 8°C 10°C DV1 Lower optimal temperature 14°C 20°C 12°C 24°C 12°C 15°C 16°C DV2 Upper optimal temperature 32°C 33°C 30°C 35°C 33°C 26°C 30°C DV3 Limiting high temperature 38°C 36°C 35°C 40°C 36°C 30°C 35°C SM0 Limiting low soil moisture 0.1 0.35 0.1 0.125 0.2 0.1 0.1 SM1 Lower optimal soil moisture 0.3 0.7 0.3 0.25 0.7 0.2 0.3 SM2 Upper optimal soil moisture 1.2 1.5 1 2 1 1 1 SM3 Limiting high soil moisture 2 2.5 2 3 2 2 2 TTCS Cold stress temperature threshold 0°C 2°C 0°C −3°C −4°C \- 0°C THCS Cold stress temperature rate −0.005 −0.0001 −0.0005 −0.007 −0.0003 \- −0.0005 TTHS Heat stress temperature threshold \- 37°C \- 42°C 40°C 35°C \- THHS Heat stress temperature rate \- 0.0002 \- 0.005 0.0002 0.01 \- SMDS Dry stress threshold 0.01 0.1 0.01 0.2 0.01 0.01 0.01 HDS Dry stress rate −0.003 −0.001 −0.007 −0.005 −0.001 −0.001 −0.007 SMWS Wet stress threshold \- 2.5 \- 2 4 \- \- HWS Wet stress rate \- 0.002 \- 0.002 0.002 \- \- TTHW Hot-wet degree-day threshold 30°C \- 30°C \- 36°C \- \- MTHW Hot-wet moisture threshold 1 \- 1 \- 0.7 \- \- PHW Hot-wet stress accumulation rate 0.075 \- 0.075 \- 0.075 \- \- Total number of records (*N*) 703 394 531 312 318 25,345 957 EI≥30 (% total) 98.4% 70.3% 96.2% 96.8% 82.1% 90.5% 90.7% EI≥20 (% total) 99.3% 97.7% 99.2% 96.8% 94.3% 95.6% 97.1% EI≥20 "water subsidy" (% total) \- 98.0% \- 98.1% \- 98.6% \- ::: ::: {#pone-0017222-t003 .table-wrap} 10.1371/journal.pone.0017222.t003 Table 3 ::: {.caption} ###### CLIMEX parameters and values, number of records used in the analysis, and model accuracy for the nine invasive species, including one dicot (*Cannabis sativa*) and eight perennial grasses. ::: {#pone-0017222-t003-3} Parameter *Cannabis sativa* *Cynodon dactylon* *Dactylis glomerata* *Elytrigia repens* *Imperata cylindrica* *Pennisetum clandestinum* *Phalaris aquatica* *Schedonorus phoenix* *Sorghum halepense* ------------- ------------------- -------------------- ---------------------- -------------------- ----------------------- --------------------------- --------------------- ----------------------- --------------------- DV0 5°C 12°C 5°C 5°C 10°C 10°C 8°C 5°C 5°C DV1 12°C 15°C 10°C 10°C 16°C 14°C 13°C 12°C 20°C DV2 27°C 33°C 26°C 28°C 30°C 31°C 27°C 27°C 30°C DV3 30°C 38°C 30°C 30°C 35°C 35°C 30°C 30°C 35°C SM0 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.01 SM1 0.4 0.2 0.2 0.2 0.3 0.4 0.4 0.4 0.1 SM2 1 1 1 1 1 1.5 1 1 1.5 SM3 10 10 10 1.5 2 10 10 4 2 TTCS 0°C 5°C \- \- 0°C 0°C 3°C −3°C 2°C THCS −0.0001 −0.0001 \- \- −0.005 −0.01 −0.001 −0.0002 −0.0001 TTHS 40°C \- 35°C 32°C \- 38°C \- 40°C 40°C THHS 0.002 \- 0.01 0.01 \- 0.0002 \- 0.002 0.005 SMDS 0.02 \- 0.01 0.01 0.01 0.02 0.02 0.02 0.01 HDS −0.005 \- −0.001 −0.001 −0.007 −0.0003 −0.05 −0.005 −0.005 SMWS \- \- \- 2.5 \- \- \- \- \- HWS \- \- \- 0.002 \- \- \- \- \- TTHW \- \- 30 32 \- 30°C \- \- \- MTHW \- \- 1 1 \- 0.8 \- \- \- PHW \- \- 0.07 0.002 \- 0.075 \- \- \- *N* 3163 8854 52242 58115 717 255 679 24978 2628 EI≥30 92.8% 84.6% 98.1% 97.7% 62.9% 80.0% 83.8% 98.6% 85.7% EI≥20 97.0% 92.6% 99.3% 99.1% 87.4% 91.8% 96.2% 99.2% 96.5% EI≥20+water 98.3% 97.8% 99.5% \- 96.2% 87.5% \- 99.8% 98.4% ::: The suite of potential bioenergy feedstocks we investigated demonstrates a vast range of potentially cultivatable land across the globe both with and without irrigation inputs ([Figs. 1](#pone-0017222-g001){ref-type="fig"}, [2](#pone-0017222-g002){ref-type="fig"}). With few exceptions (e.g., *S. officinarum*, *P. purpureum*), the bioenergy crops exhibit broad climatic tolerance, which allows tremendous flexibility in choosing crops, especially in areas with high summer rainfall and long growing seasons (e.g., southeastern US, Amazon Basin, eastern Australia). Unsurprisingly, the invasive species of agronomic origin have very similar global climate niche profiles as the proposed bioenergy crops ([Fig. 3](#pone-0017222-g003){ref-type="fig"}), also demonstrating broad climatic tolerance. The "perennial water scenario", which mimics both irrigation additions as well as access to a permanent water supply [@pone.0017222-Barney3], typically expanded the climate niche to regions that are arid during the growing season, but are otherwise suitable: western US, northern Africa, central and western Australia, and the Middle East ([Figs. 1](#pone-0017222-g001){ref-type="fig"}, [2](#pone-0017222-g002){ref-type="fig"}, [3](#pone-0017222-g003){ref-type="fig"}). ::: {#pone-0017222-g001 .fig} 10.1371/journal.pone.0017222.g001 Figure 1 ::: {.caption} ###### Climate suitability maps for nine grass candidate biofuel feedstocks. Some of the species (A, C, D, F, G) are known weeds of the US, others (B) are native, and some (E, H, I) are currently under cultivation. The colors represent the CLIMEX ecoclimatic index (EI) where gray (EI≤10) is 'unfavorable', light green (11\>EI\>20) is 'suitable', dark green (21\>EI\>30) is 'favorable', and blue (EI≥31) is 'very favorable'. The purple regions are those with an EI\>20 when a permanent water source is available. :::  ::: ::: {#pone-0017222-g002 .fig} 10.1371/journal.pone.0017222.g002 Figure 2 ::: {.caption} ###### Climate suitability maps for four woody and three herbaceous candidate biofuel feedstocks. Some of the species (A, C, D, G) are known weeds of the US, and some (B, E, F) are currently under cultivation. The colors represent the CLIMEX ecoclimatic index (EI) where gray (EI≤10) is 'unfavorable', light green (11\>EI\>20) is 'suitable', dark green (21\>EI\>30) is 'favorable', and blue (EI≥31) is 'very favorable'. The purple regions are those with an EI\>20 when a permanent water source is available. :::  ::: ::: {#pone-0017222-g003 .fig} 10.1371/journal.pone.0017222.g003 Figure 3 ::: {.caption} ###### Climate suitability maps for nine invasive species of agronomic origin. All taxa (A--I) are currently weedy species in the US. The colors represent the CLIMEX ecoclimatic index (EI) where gray (EI≤10) is 'unfavorable', light green (11\>EI\>20) is 'suitable', dark green (21\>EI\>30) is 'favorable', and blue (EI≥31) is 'very favorable'. The purple regions are those with an EI\>20 when a permanent water source is available. :::  ::: The ecoregional evaluation of bioenergy crops and known invasive species demonstrates tremendous overlap at both high (EI≥30) and moderate (EI≥20) climate suitability ([Figs. 4](#pone-0017222-g004){ref-type="fig"}, [5](#pone-0017222-g005){ref-type="fig"}). The southern and western US ecoregions support the greatest number of invasive species of agronomic origin, especially the Southeastern USA Plains, Mixed Woods Plains, and Mediterranean California ([Fig. 5B](#pone-0017222-g005){ref-type="fig"}). This differs only slightly for bioenergy crops with the Southeastern USA Plains, Mixed Woods Plains, and Western Sierra Madre Piedmont ecoregions supporting the most taxa ([Fig. 5A](#pone-0017222-g005){ref-type="fig"}). Bioenergy crops had a high climate match (EI≥30) in at least some part of 20--85% of US ecoregions without a permanent water supply, and 35--90% with an irrigation factor ([Table 4](#pone-0017222-t004){ref-type="table"}). The invasive species had a high climate match in 50--85% without a permanent water source, and 60--90% with permanent water source. ::: {#pone-0017222-g004 .fig} 10.1371/journal.pone.0017222.g004 Figure 4 ::: {.caption} ###### Ecoregion climatic suitability of biofuel crops and invasive species. Potential biofuel crops with (A) moderate and (B) high suitability in contrast to existing invasive species with (C) moderate and (D) high suitability. :::  ::: ::: {#pone-0017222-g005 .fig} 10.1371/journal.pone.0017222.g005 Figure 5 ::: {.caption} ###### Distribution of populations in ecoregions of the continental US. Proportion of (A) 15 proposed biofuel crops and (B) nine invasive species located in each ecoregion. The ecoregions begin with 5.2 Mixed Woods Shield at the 12 o\'clock position and proceed clockwise according to the legend. :::  ::: ::: {#pone-0017222-t004 .table-wrap} 10.1371/journal.pone.0017222.t004 Table 4 ::: {.caption} ###### Ecoregional distribution of biofuel crops and invasive species. ::: {#pone-0017222-t004-4} Species \% North American Ecoregions \% Continental US Ecoregions ----------------------------- ------------------------------ ------------------------------ ----- ----- **Biofuels** *Arundo donax* \- 62% \- 85% *Eucalyptus globulus* 50% \- 55% \- *Jatropha curcas* 38% 48% 20% 35% *Medicago sativa* 64% 66% 85% 90% *Miscanthus sacchariflorus* 56% \- 75% \- *M. sinensis* 56% 64% 60% 85% *Nicotiana tabacum* 60% \- 70% \- *Panicum virgatum* 56% 66% 55% 90% *Paulownia tomentosa* 58% \- 75% \- *Pennisetum purpureum* 40% 46% 25% 35% *Phalaris arundinacea* 80% 70% 80% 55% *Pueraria montana* 56% \- 65% \- *Saccharum officinarum* 32% 50% 20% 45% *Sorghum bicolor* 28% 36% 60% 75% *Triadica sebifera* 42% 44% 40% 60% **Invasives** *Cannabis sativa* 58% 54% 75% 70% *Cynodon dactylon* 56% 52% 80% 80% *Dactylis glomerata* 74% 72% 75% 80% *Elytrigia repens* 78% \- 85% \- *Imperata cylindrica* 50% 52% 50% 65% *Pennisetum clandestinum* 46% 38% 50% 35% *Phalaris aquatica* \- 48% \- 60% *Schedonorus phoenix* 54% 64% 70% 95% *Sorghum halepense* 62% \- 80% \- Percentage of ecoregions in North America (n = 50) or the Continental US (n = 20) that have some portion of the current or predicted (CLIMEX EI≥30) range of each species within its boundaries. ::: Discussion {#s4} ========== The climate niche for the bioenergy crops evaluated demonstrates that temperate to sub-tropical regions of the world that receive consistent summer rainfall and have a warm/hot summer and a long growing season will be most favorable, and will provide the greatest number of feedstock choices without the need for consistent summer irrigation. The most favorable regions include the southeastern and southcentral US, the Amazon basin, sub-Saharan and central Africa, western continental Europe, southeast Asia, and eastern Australia. In North America, the ecoregions that appear most suitable for bioenergy crops are the Southeastern USA Plains characterized by weakly developed soils, average annual temperatures of 13--19 C, and 1000--1600 mm of annual precipitation (23% of the databased populations), the Ozark Ouachita-Appalachian Forests characterized by weakly developed soils, average annual temperatures of 17--18 C, and 1000--2000 mm of annual precipitation (11% of databased populations), and the Mixed Woods Plains characterized by forest and fine textured soils, average annual temperatures 4--10 C, and 720--1200 mm of annual precipitation (9% of databased populations). The large number of collected populations in these ecoregions suggests that many of these bioenergy crops are already established, indicating high climatic suitability, as well as favorable abiotic and biotic conditions (locally at least). The USDA recently released an analysis demonstrating that the US Southeast will likely yield about 50% of the biomass needed to meet the Renewable Fuel Standard. Our analysis demonstrates that this region will have the greatest number of species from which to choose. The global climate niche distributions for the invasive species of agronomic origin were generally very similar to the bioenergy crops, except for the sub-tropical *Pennisetum clandestinum*. Many of these weedy species continue to be utilized as turfgrass (*S. phoenix*, *Cynodon dactylon*) or forages (*D. glomerata*, *E. repens*, *S. phoenix*), and may be under irrigation, which greatly expands their climate niche because cultivation generally reduces environmental stochasticity [@pone.0017222-Mack1]. Coincidently, the ecoregions that have the greatest number of invasive species populations are nearly identical to those for bioenergy crops, except for the Mediterranean region of California, which is one of the most heavily invaded regions of the US [@pone.0017222-DiTomaso1]. However, this arid environment is unlikely to be a major location for bioenergy crop production, due to the requirement for summer irrigation---currently a scarce resource in the western US [@pone.0017222-Schnoor1]. Broad climatic tolerance, or a large climate niche, is positively correlated with invasiveness, as this greatly increases the probability of surviving outside of cultivation in the multitude of possible environments that might be encountered. However, this is also a desirable character of crop plants by increasing the suitable agro-ecoregions for cultivation. Therefore, it is not surprising that the climate niche for plants with an agronomic origin are large, as breeders generally select for this characteristic, and often direct efforts to enhance cold, heat, or drought tolerance, which broadens the climate niche [@pone.0017222-Lewandowski1]. Additionally, the boundaries of the climate niche can be used to impose functional sterility on bioenergy crops, which increases yield while simultaneously reducing the escape potential by precluding seed production [@pone.0017222-Quinn1]. The fact that the bioenergy crops investigated here have similarly broad climatic tolerance as the invasive species in no way indicates eventual invasiveness. Nevertheless, this characteristic---broad climatic tolerance---should be considered when evaluating the risk of invasiveness for each bioenergy species [@pone.0017222-Barney1]. The climate niche of bioenergy crops must be accounted for when evaluating the invasion risk, and should not be assumed to be a high match as has previously occurred [@pone.0017222-Barney1], [@pone.0017222-Crosti1], [@pone.0017222-Buddenhagen1]. Additionally, evaluating the climate niche for introduced species should not occur at the continental or national geopolitical scale as is current practice in existing risk assessment frameworks [@pone.0017222-Pheloung1]. Large-scale assessments that cover vast geographic regions with diverse climates are prone to overestimating the risk of invasion because the probability of at least one propagule encountering one susceptible community is extremely high. Pheloung and colleagues recognized the importance of evaluating the climate match for target species [@pone.0017222-Pheloung1], but performed their assessments at the continental scale for Australia, which has regions varying from deserts to tropical rain forests, so the likelihood of at least one habitat having a high climate match for the country is nearly certain. An additional consideration is that populations of species are invasive, not the species themselves. For example, *Arundo donax* is a state-listed noxious weed in California and Texas where it dominates riparian habitat [@pone.0017222-DiTomaso1]. However, *Arundo* is only occasionally found in the Mid-Atlantic and Southeastern US where it has existed for many decades [@pone.0017222-Barney1]. Therefore, *Arundo* may be benignly cultivated in some areas of the US, while being a noxious weed in other areas of the US. This relates to the spatial context of a risk assessment in that invasiveness occurs on a spatial scale smaller than countries, and should not be restricted to ecologically arbitrary geopolitical boundaries. In an attempt to address the need for evaluating invasive risk at sub-national levels we incorporated an ecoregional assessment of the climate niche. There are several ecoregional designations for North America available that vary in spatial context: Level I contains 15 broad categories, Level II has 50 smaller categories, while Level III contains 182 categories. We chose the Level II designation as it provides 20 distinct ecoregions in the US that the species of interest occur, which captures sufficient variation in climate, ecosystems, and land use to be useful for stakeholders without being too general (Level I) or too specific (Level III). Some collected populations of both bioenergy crops and invasive species occurred in all 20 ecoregions, though the relative distribution of these populations was extremely unbalanced ([Fig. 5](#pone-0017222-g005){ref-type="fig"}), with the Southeastern US supporting the greatest number of populations. As the bioeconomy grows globally, especially in the southeastern US, which is estimated to support about 50% of the biomass to meet federal mandates [@pone.0017222-USDA1], precaution should be taken in large-scale introductions of potentially invasive bioenergy crops. This mistake has been made in the past by federally subsidized large-scale adoption of novel species that ultimately turn out costing orders of magnitude more taxpayer dollars to manage (eg, kudzu and johnsongrass). We would like to thank the anonymous reviewers who improved the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The authors have no support or funding to report. [^1]: Conceived and designed the experiments: JNB JMD. Performed the experiments: JNB. Analyzed the data: JNB. Contributed reagents/materials/analysis tools: JNB. Wrote the paper: JNB JMD.

PubMed Central

2024-06-05T04:04:19.100245

2011-3-9

PMC3052300

Introduction {#s1} ============ Human African trypanosomiasis (HAT), or sleeping sickness, is caused by the protozoan parasites *Trypanosoma brucei (T. b.) gambiense* or *T. b. rhodesiense*. These parasites, transmitted by the bite of an infected tse tse fly, cause a two-stage disease. The early hemolymphatic stage 1 corresponds to the proliferation of the parasite in the blood and lymph. This stage is followed by the meningoencephalitic stage 2 during which the parasites invade the central nervous system (CNS) [@pone.0016891-Buguet1]. If untreated, the disease evolves towards death in a deep cachectic condition. Nitric oxide (NO), a gaseous signaling molecule that acts as a messenger in the immune and nervous systems, has been involved in the physiopathological processes that occur in the rat model of HAT [@pone.0016891-Amrouni1]. NO is a highly reactive molecule possessing a very short half-life (\<0.5 second) and is synthesized from the amino acid L-arginine by three major isoforms of NO Synthase (NOS): the neuronal NOS (nNOS, type 1), the inducible NOS (iNOS, type 2) and the endothelial NOS (eNOS, type 3) [@pone.0016891-Alderton1]. The production of NO by iNOS takes place during inflammatory and infectious situations. Accordingly, it is a key element in counteracting the proliferation of trypanosomes through its cytotoxic properties [@pone.0016891-Gobert1], [@pone.0016891-AntoineMoussiaux1]. The biological availability of NO is achieved directly by the activity of the NOS enzymes. These enzymes are further submitted to a complex regulation involving arginase and N^G^, N^G^-dimethylarginine dimethylaminohydrolase (DDAH) enzymes ([Fig. 1](#pone-0016891-g001){ref-type="fig"}). Arginase, in catalyzing the last step of the urea cycle, leads to the production of L-ornithine, a synthesis precursor of L-proline, L-glutamate, L-glutamine, γ-aminobutyric acid (GABA) and polyamines ([Fig. 1](#pone-0016891-g001){ref-type="fig"}). Arginase exists in two isoforms: (i) arginase-1 (liver isoform) is strongly expressed in the cytosol of liver cells; (ii) while arginase-2 (extra-liver isoform, predominant in kidney) is localized into the mitochondrial membrane of cells. In the brain, arginase-1 and, to a lesser extent, arginase-2 have been identified in neurons but not in glial cells. ::: {#pone-0016891-g001 .fig} 10.1371/journal.pone.0016891.g001 Figure 1 ::: {.caption} ###### Simplified scheme of L-arginine pathways in mammalian host cells. L-arginine is a crossroad for multiple pathways since it is the common substrate of arginase and NOS that catalyzes the production of L-ornithine, L-citrulline and NO, respectively. ADMA, a product of the proteolysis of methylated proteins by PRMT-1, is a potent inhibitor of NOS. Its concentration is regulated by DDAH that catalyzes the production of L-citrulline and dimethylamine. L-ornithine and L-glutamate serve as precursors of polyamines and glutathione. These two metabolites will be further utilized by the trypanosomes for the synthesis of trypanothione, an essential element protecting the parasites from free radicals. Abbreviations: NO, nitric oxide; NOS, nitric oxide synthase; ADMA, asymmetric dimethylarginine; DDAH, N^G^,N^G^-dimethylarginine dimethylaminohydrolase; PRMT-1, Protein arginine N-methyltransferase 1; GABA, γ-aminobutyric acid. :::  ::: Arginase activity is particularly high in brain regions with strong nNOS expression [@pone.0016891-Yu1]. Arginase may contribute to reduce NOS activity, due to the use of the common substrate, L-arginine. Conversely, arginase may be inhibited by N-hydroxy-L-arginine, the intermediate product of the reaction catalyzed by NOS [@pone.0016891-Hecker1]. Another potent way regulating NO synthesis is mediated by DDAH. This enzyme catalyzes the hydrolysis of asymmetric dimethylarginine (ADMA) to dimethylamine and L-citrulline. ADMA is a product of the catabolism of arginine-methylated proteins by protein arginine N-methyltransferase 1 (PRMT-1) [@pone.0016891-Tsikas1] ([Fig. 1](#pone-0016891-g001){ref-type="fig"}). As for arginase, two isoforms of DDAH have been described, DDAH-1 and DDAH-2. It is reported that DDAH-1 is typically located in nNOS-containing tissues, whereas DDAH-2 predominates in iNOS-expressing tissues and in vascularized areas where eNOS is present [@pone.0016891-Tran1]. All NOS isoforms can be inhibited by ADMA, the substrate of DDAH. Under certain conditions, the S-nitrosylation of DDAH by NO may lead to a strong decrease in its activity, leading subsequently to an accumulation of ADMA and NOS inhibition. This feedback mechanism may thus contribute to the regulation of NO production [@pone.0016891-Leiper1]. In a recent study conducted in the rat infected with *T. b. brucei*, an experimental model of HAT, the production of NO was analyzed both in the brain (hypothalamus/thalamus) and the periphery (blood) [@pone.0016891-Amrouni1]. Changes occurring in NO production were dependent on iNOS activity. In the periphery, two main observations were recorded: (i) a marked decrease in blood NO production which likely limits the NO trypanocidal activity; and (ii) an increased activity of arginase favouring the synthesis of polyamines necessary for trypanosome growth [@pone.0016891-Duleu1]. In the brain, on the contrary, an increase in NO production occurred due to an enhanced iNOS activity in neurons and glial cells. Consequently, the iNOS activation observed in brain cells was primarily considered as a marker of deleterious inflammatory processes. To date, the brain mechanisms of this infection are part of the most severe aspects of the disease and remain unknown. The present work, an extension of the above investigations [@pone.0016891-Amrouni1], focused on the relative role of arginase and DDAH in the brain during the different stages of *T. b. brucei* infection. Measures were performed throughout the time course of the infection in the rat model infected with *T. b. brucei* to assess (i) the cerebral changes occurring in the enzymatic activities, mRNA and protein expression of arginase and DDAH; (ii) the determination of the immunohistochemical distribution and the morphometric parameters of the cells expressing the isoforms of DDAH (DDAH-1 and DDAH-2) within the diencephalon; and (iii) the changes occurring in the profile of the amino acids related to the NOS/arginase/DDAH pathways. Materials and Methods {#s2} ===================== Ethics statement {#s2a} ---------------- All animals were handled in strict accordance with good animal practice in compliance with the relevant decree of the French Agriculture Ministry (N°: 03-505). Animal procedure was approved by Departmental Direction of Veterinary Services of Rhône-Alpes. Animals and infection {#s2b} --------------------- Male Wistar rats (Janvier breeding, Le Genest Saint Isle, France) weighing 200--220 g, were kept under standard laboratory conditions: 12 h/12 h light/dark cycle, ambient temperature of 21 to 22°C, food and water *ad libitum*. After one week of adaptation, two groups of animals (control, n = 24; infected, n = 24) were constituted. The animals were infected with the AnTat 1.1 E clone of *T. b. brucei* (Institute of Tropical Medicine, Antwerp, Belgium). Before infection, the mobility and viability of trypanosomes were controlled under microscope. The intraperitoneal (i.p.) injection of 3,600 parasites defined the initial day (D0) of the experimental session. In parallel, control animals received an i.p. injection of the same volume of physiological saline solution. In the infected rats (HAT experimental model), body weight was measured and blood parasites counted every two days as previously described [@pone.0016891-Amrouni1]. Cerebrospinal fluid (CSF) samples were obtained from *cisterna magna* puncture achieved under chloral hydrate anesthesia (400 mg/kg). Parasite counts in CSF samples were performed every four days after infection (from D10 to D22). The biological signs characterizing the entry into the neurological state of the disease appeared, as previously described [@pone.0016891-Amrouni1], from D5 to D10 post-infection. At this time, however, the parasites were not yet observed in the CSF (hemolymphatic stage 1). From D10 to D22, while the body weight gain continue to declines, trypanosomes were observed in the CSF (neurological stage 2). The animals were sacrificed by decapitation at days D5, D10, D16 and D22 post-infection. Dissection and protein extracts from brain tissue {#s2c} ------------------------------------------------- In order to obtain protein extracts from cerebral tissues, the brain of each animal was quickly removed, briefly rinsed in ice-cold Tris-EDTA buffer (50 mM, pH 7.4), and dissected to collect the diencephalic (hypothalamus/thalamus) structure. Protein extracts from brain tissue were obtained as previously described [@pone.0016891-Amrouni1] and protein concentrations were determined in supernatant fractions using Bradford\'s method [@pone.0016891-Bradford1]. Plasma samples {#s2d} -------------- Blood samples were obtained from direct puncture of the heart with a heparinized syringe and centrifuged at 2,400 g (4°C, 10 min). Then, plasma supernatant was removed and samples kept at −80°C until use. Arginase and DDAH activity in brain tissue {#s2e} ------------------------------------------ ### Measurement of arginase activity {#s2e1} Arginase activity was determined by analyzing the conversion of L-arginine to urea with the colorimetric method described by Liu et al. [@pone.0016891-Liu1]. Briefly, for the enzyme activation, 15 µL of a solution containing 10 mM MnCl~2~ in 50 mM Tris-HCl (pH 7.5) were added to 15 µL of each diencephalic protein extract (about 75 µg of proteins). The mixture was incubated at 55°C for 10 min. Then, 30 µL of the substrate, 0.2 M L-arginine, were added and the new mixture incubated for 5 min at 37°C. The enzymatic reaction was stopped by adding 40 µL of 20% H~2~SO~4~. To quantify urea formation, 2.5 mL of a 20% H~2~SO~4~ solution containing 0.4% w/v antipyrin, 0.005% w/v iron (III)-sulfate hydrate and 2.5 mL of a 5% acetic acid solution containing 0.5 w/v diacethyl monoxim were added. After incubation at 100°C for 15 min, the absorbance of the colored product was measured at 450 nm. All assays were carried out in duplicates. The blank assay was performed in the same way, except that the enzymatic reaction was stopped with the acidic solution immediately after addition of the substrate to each diencephalic protein sample. The specific activity corresponding to nanomoles of urea formed per min and per mg of proteins (nmol/min/mg) was expressed as % of the values obtained in the non-infected condition. The arginase activity obtained from samples of rat liver was used as positive control. ### Measurement of DDAH activity {#s2e2} DDAH activity was measured by analyzing the conversion of ADMA to L-citrulline. After the addition of 225 µL of 5 mM ADMA in 100 mM phosphate buffer (pH 6.5) to 25 µL of each diencephalic protein extract (about 125 µg of proteins), the mixture was incubated at 37°C for two hours. The reaction was stopped by the addition of 250 µL of 10% tricloroacetic acid (ice-cold). After centrifugation (10 min at 3,000 g), 250 µl of a 20% H~2~SO~4~ solution containing 0.5% w/v antipyrin and 250 µl of a 5% acetic acid solution containing 0.8% w/v diacethyl monoxim, were added to each supernatant sample (about 490 µL) containing L-citrulline. The mixtures were then incubated at 100°C for 12 min before measuring the absorbance at 466 nm as described by Prescott and Jones [@pone.0016891-Prescott1]. All assays were carried out in duplicate. The blank assay was performed in the same way, except that the enzymatic reaction was immediately stopped (by the ice-cold tricloroacetic acid) just after addition of the buffered substrate solution to each diencephalic protein sample. The specific activities (picomoles of citrulline formed per min and per mg of protein (pmol/min/mg)) were expressed in % of the values obtained in the non-infected condition. The DDAH activity obtained from samples of rat kidney was used as positive control. Western-blotting {#s2f} ---------------- Forty µg of the total protein extract obtained from diencephalic samples was loaded and separated on NuPAGE® Novex® 10% Bis-Tris gels with MOPS running buffer (Invitrogen-Life Technologies, Cergy-Pontoise, France). Protein transfers to nitrocellulose membranes were performed using the iBlot™ DryBlotting device (Invitrogen) and iBlot™ Transfer stacks (program P3 for 7 min). After transfer, membranes were first saturated for one hour at room temperature in buffer A (Tris 50 mM pH 7.5, NaCl 166 mM, Tween 0.2%) containing 5% of non fat dry milk, followed by an incubation with the rabbit polyclonal anti-DDAH-1 (1/500, Orbigen, San Diego, USA), rabbit polyclonal anti-DDAH-2 (H-85 1/200, Santa Cruz Biotechnology, California, USA), rabbit polyclonal anti-arginase-1 (H-52, 1/200, Santa Cruz Biotechnology), rabbit polyclonal anti-arginase-2 (H-64, 1/200, Santa Cruz Biotechnology) or mouse anti-β actin antibody (1/1000, Santa Cruz Biotechnology) for 6 hours in buffer A containing 2.5% of non fat dry milk. After three washings in buffer A, membranes were incubated during one hour with corresponding biotinylated secondary antibody, goat anti-rabbit (1/2000, Vector Labs, ABCYS, Paris, France), or goat-anti mouse (1/1000, Vector Labs) in buffer A containing 2.5% non fat dry milk, washed again in buffer A, then incubated for two hours with avidin-peroxidase (1/1000, Vectastain kit, Vector Labs) in buffer A. Blots were then washed three times and the conjugates were visualized by enhanced chemiluminescence (ECL) (Super Signal West Pico, Pierce, Thermo Scientific, and Courtabeuf, France). The ECL signals were acquired with ChemidocXRS (BioRad, Marnes-la-Coquette, France) and quantified with Quantity One software (BioRad). Blots of anti-β actin were used as internal standard and samples of kidney and liver as positive controls. Quantification of the blots was performed using a BioRad ChemiDoc software. Mean values determined are given for control and experimental groups at different days post infection (D5, D10, D16 and D22). Real-time RT-PCR {#s2g} ---------------- As previously described [@pone.0016891-Meiller1], total RNA was prepared from diencephalic brain samples using a RNA-plus solution (Q-Biogen, Illkirch, France). For each gene, the sequence of the primer pairs and the annealing temperature are given in [Table 1](#pone-0016891-t001){ref-type="table"}. Results were normalized to 18S rRNA as endogenous control and expressed in percent of the corresponding values of the non-infected condition. ::: {#pone-0016891-t001 .table-wrap} 10.1371/journal.pone.0016891.t001 Table 1 ::: {.caption} ###### Real-time PCR analysis: sequence of the oligonucleotide pairs, annealing temperature (Ta) and product size. ::: {#pone-0016891-t001-1} Genes Primer pairs Ta (°C) Product size (bp) ------------ ------------------------------------------- --------- ------------------- **DDAH-1** ACTCACTCCAGCCTCTGTGT TCCACGTTCTCAGGACTCTC 57 153 **DDAH-2** AACTGAGGCAACGACTAGGT GATTAGAGCCGTGTCTCCTT 58 113 **ARG-1** GGCAGTGCCGTTGACCTTGT AGCAGCGTTGGCCTGGTTCT 59 158 **ARG-2** TGTCGGCTCTGGATCTTGTT AATGTGTCCGCCTTCTCTTG 57 131 **18S** TAGAGGGACAAGTGGCGTTC CGCTGAGCCAGTCAGTGTAG 64 100 ::: Immunohistochemistry of DDAH-1 and DDAH-2 {#s2h} ----------------------------------------- Animal perfusion, brain fixation and cutting were achieved as previously described [@pone.0016891-Amrouni1]. Saturation of the non specific binding sites of the antibody was performed by incubating the brain sections (25 µm thick) for 90 min in a phosphate buffered saline solution (PBS; 50 mM, pH 7.4) containing 1% of bovine serum albumin (BSA). Rabbit polyclonal anti-DDAH-1 antibody (1/500, Orbigen, San Diego, USA) and goat polyclonal anti-DDAH-2 antibody (1/500, Abcam, Paris, France) were used as primary antibody (2×24 h at 4°C). Biotinylated goat anti-rabbit (1/500, Vector Labs, ABCYS, Paris, France) and rabbit anti-goat (1/500, Vector Labs) were used as secondary antibody (12 h at 4°C). Three infected or un-infected rats were used for DDAH-1 and DDAH-2 immunohistochemistry. Rat kidney, expressing predominantly DDAH isoforms, was used as specific controls for the different antibodies employed. Morphometric analysis of DDAH-1 and DDAH-2 expressing cells {#s2i} ----------------------------------------------------------- The analysis was achieved on the ventral and posterior parts of the thalamus, an area where DDAH-1 and DDAH-2 immunoreactivity had been observed. The morphometric parameters employed were as follows: (1) Soma area (µm^2^); (2) Soma perimeter (µm); (3) Soma form factor (FF\* = 4π×Area/Perimeter, FF\* values equaled 1.0 for a perfect circle and \<1.0 for all ellipses). Measures were taken from 268 elements, i.e., 134 cells for each DDAH brain section isoform (DDAH-1 or DDAH-2). The external border of the soma was delimited by way of a computer mouse, then, the area of the soma, perimeter and form factor were automatically calculated by the Mercator software which was related to the Axioscope2 plus microscope and the QIMAGING camera. Amino acids (AA) assay by liquid chromatography tandem-mass spectrometry (LC-MS/MS) {#s2j} ----------------------------------------------------------------------------------- L-Arginine, L-glutamate, L-glutamine, L-proline and GABA were quantified in diencephalic brain structures as underivatized AA by the LC-MS/MS method previously described by Piraud et al. [@pone.0016891-Piraud1]. Only L-Arginine, L-glutamate and L-glutamine were quantified in blood samples by the same method. L-Citrulline, L-ornithine (in brain) and ADMA (in blood), were quantified as butyl esters derivatives in order to improve the detection limit. The method used is adapted from Schwedhelm et al. [@pone.0016891-Schwedhelm1] with some slight modifications of the chromatographic gradient and the internal standard. Derivatized AA assay {#s2k} -------------------- ### Standards and samples preparation {#s2k1} Fifty µL of calibration standard solutions (L-citrulline, L-ornithine or ADMA in increasing quantities: 0, 0.5, 1, 2, 5, 10, and 20 µM each), or 50 µL of brains extracts or 50 µL of plasma samples, were mixed with 20 µL of internal standard solution at 14 µM L-\[^2^H\] proline (Pro\*), L-\[^2^H\] ornithine (Orn\*) and L-\[^2^H\] arginine (Arg\*) (Cambridge Isotope Laboratories, Andover, USA). Protein and lipid precipitation was performed by adding 400 µL of ethanol and 400 µL hexane to all samples. The samples were then mixed for 2 min and stored for 10 min at room temperature. Afterwards, the supernatant was eliminated and the residual phase again precipitated by 400 µL of hexane. After centrifugation during 5 min at 17,500 g, dried supernatant were derivatized by butylation using 200 µL of 3N HCl in butanol (Interchim, Montluçon, France), followed by evaporation according to Schwedhelm et al. [@pone.0016891-Schwedhelm1]. Finally, dried samples were reconstituted in 50 µL of the mobile phase (formic acid/methanol, 1 mL/1 L) and fractions of 5 µL of brain extracts or 20 µL of plasma were injected into the LC apparatus. ### LC and gradient conditions {#s2k2} The derivatized AA separation was achieved using reversed-phase liquid chromatography (Waters Alliance 2795 LC set-up, Waters Corporation, Beverly, MA, USA) with a Varian analytical column (2.0×50 mm) packed with a Polaris C18-Ether (3 µm Bead size, Varian, Courtaboeuf France). Before use, the column was rinsed with methanol during one hour at a flow rate of 0.1 mL/min, then, overnight, with methanol/water (50/50, v/v). The chromatography was performed at 25°C with a flow rate of 0.2 mL/min. The gradient started with 10% mobile phase A (formic acid/methanol, 1 mL/1 L) and 90% mobile phase B (formic acid/water, 1 mL/1 L), then increased linearly during 2 min until 50% A, which is maintained during 5 min, and then back to 10% A during 5.5 min. Afterwards, 10% A was maintained during 15 min in order to re-equilibrate the column before a new injection. The LC set-up was connected to the electrospray source for ionization (ESI) of the tandem mass spectrometry. ### MS/MS {#s2k3} The AA identification was carried out on a Quattro micro triple quadruple tandem mass spectrometer (Waters Corporation). The electrospray source was used in positive ionization mode for all analyses. Nitrogen and argon were used as drying and collision gases, respectively. Source temperature was fixed at 120°C and desolvatation temperature at 350°C. The cone gas flow was fixed at 60 L/h, the desolvatation gas flow at 600 L/h, the voltage multiplier at 650 V and the dwell time at 0.05 s. All results were acquired with the MassLynx software version 4.0. Specific MS, LC parameters (cone voltage, CV in V; collision energy, CE in eV; retention time, RT in min) and internal standards (AA\*) used are given in [Table 2](#pone-0016891-t002){ref-type="table"}. ::: {#pone-0016891-t002 .table-wrap} 10.1371/journal.pone.0016891.t002 Table 2 ::: {.caption} ###### Specific parameters used for identification of amino acid in MS/MS positive ion mode. ::: {#pone-0016891-t002-2} Monitored transition (m/z) CV (V) CE (eV) AA\* RT (min) ---------------------------------- ---------------------------- -------- --------- --------- ---------- ***Derivatized (butylated) AA*** L-citrulline 232\>113 20 18 L-pro\* 3.11 L-ornthine 189\>70 12 18 L-orn\* 1.10 ADMA 259\>214 28 14 L-arg\* 2.16 ***Underivatized AA*** L-arginine 175\>70 20 32 L-arg\* 12.71 L-proline 116\>70 20 30 L-pro\* 2.38 L-glutamate 148\>84 15 24 L-glu\* 2.07 L-glutamine 147\>84 20 24 L-gln\* 2.02 GABA 104\>87 20 14 L-lys\* 9.16 AA: Amino acid. CV, CE correspond to the cone voltage (in V), and collision energy (in eV), respectively. AA\* corresponds to the stable labelled amino acid (AA) chosen as internal standard for the quantification. RT corresponds to the time of retention (in min) of each analyte under selected chromatographic conditions. ADMA: asymmetric dimethylarginine. GABA: γ-aminobutyric acid. ::: Underivatized AA assay {#s2l} ---------------------- The quantitative determination of underivatized AA was achieved as reported by Piraud et al. [@pone.0016891-Piraud1]. All MS, LC parameters and internal standards (AA\*) used for each AA are given in [Table 2](#pone-0016891-t002){ref-type="table"}. Statistical Analysis {#s2m} -------------------- Results are expressed as the mean ± standard error of the mean (SEM). Statistical analysis of the differences between rat groups (n = 6 animals in each group) at the different times post- infection (D5, D10, D16, D22) was performed by one-way analysis of variance (ANOVA). When ANOVA was significant at p\<0.05, a post hoc Fisher\'s least significant differences test was applied. Results {#s3} ======= Changes occurring in DDAH and arginase brain activity {#s3a} ----------------------------------------------------- Compared to healthy control rats, a significant increase in DDAH activity was observed at D10 (164.81±13.11%) and D16 (150.89±10.01%) in diencephalic brain structures. Afterwards, at D22 post-infection, the activity decreased towards control values ([Fig. 2A](#pone-0016891-g002){ref-type="fig"}). No significant variation was observed in arginase activity throughout the time course of the infection ([Fig. 2D](#pone-0016891-g002){ref-type="fig"}). ::: {#pone-0016891-g002 .fig} 10.1371/journal.pone.0016891.g002 Figure 2 ::: {.caption} ###### Relative changes in activity, mRNA and protein expression of DDAH and arginase in the brain during the course of infection in *Trypanosoma brucei brucei* (*T. b. b.*) infected rats. DDAH (A) and arginase (D) activities were measured in diencephalic (hypothalamus/thalamus) biopsies. Results (mean value ± SEM, n = 6 per experimental day) were normalized with corresponding values measured in control rats (n = 6 per experimental day). Control values (100%), delineated by the dotted line, correspond to 269.66±7.09 pmol/min/mg for DDAH and to 26.34±0.40 pmol/min/mg for arginase. Relative intensity of DDAH (B) and arginase (E) bands in western blot was calculated for each band and the results expressed as a mean value for control and different experimental days (β-actin was used as an internal standard). The significant up-regulation taking place for *DDAH-2* transcription (C) is not observed for *DDAH-1*(C), *ARG-1* or *ARG-2 (F)*. The transcript levels for *DDAH-1*, *DDAH-2*, *ARG-1 and ARG-2* were estimated by real-time quantitative RT-PCR and normalized to 18S rRNA as endogenous control. The graphs represent the mean ± SEM of 4 independent experiments (n = 4 per experimental day). Statistics: ANOVA followed by post-hoc Fisher\'s test (*\*p\<0.05*, \*\*p\<0.001 compared to healthy rats). Abbreviations: DDAH, N^G^, N^G^-dimethylarginine dimethylaminohydrolase; *DDAH-1*, DDAH isoform 1; *DDAH-2*, DDAH isoform 2; Ctrl: Control; *ARG-1*, arginase isoform 1; *ARG-2*, arginase isoform 2. :::  ::: Brain expression of the two isoforms of DDAH {#s3b} -------------------------------------------- ### Accumulation of DDAH-2 protein isoform in the brain of infected animals {#s3b1} Western blot analysis of hypothalamic and thalamic tissue confirms the presence of DDAH-1 and DDAH-2 in these brain structures ([Fig. 2B](#pone-0016891-g002){ref-type="fig"}). In agreement with data on DDAH activity, DDAH-2 protein levels increase in infected animals at D10 and D16 post-infection ([Fig. 2B](#pone-0016891-g002){ref-type="fig"}). Therefore, DDAH-1 did not differ between the two animal groups throughout the time course of the infection ([Fig. 2B](#pone-0016891-g002){ref-type="fig"}). ### Up-regulation of DDAH-2 transcription in the brain of infected animals {#s3b2} Compared to healthy control animals, a significant increase in mRNA transcripts of DDAH-2 was observed in the diencephalon of infected animals at D10 (2.76±0.64 fold) and D16 (1.87±0.64 fold) ([Fig. 2C](#pone-0016891-g002){ref-type="fig"}). No change was observed in the mRNA transcripts of DDAH-1. ### Brain distribution of DDAH-1 and DDAH-2 protein isoforms {#s3b3} DDAH-1 and, to a lesser extent, DDAH-2 immunoreactivities were observed in several brain areas (cortex, hippocampus, amygdala) and notably in the hypothalamic and thalamic groups of nuclei in healthy and infected rat (at D10 and D16 post-infection) groups. For DDAH-1, a cytosolic immunoreactivity was clearly identified among the interneurons of the perifornical (PeF) nucleus and the magnocellular part of the lateral hypothalamus. DDAH-1 immunoreactivity was also observed in the zona incerta (ZI), the habenular nuclei (Hb) and subincertal (SubI), reuniens (Re), central medial (CM), posterior (Po), ventral posterolateral (VPL), posteromedial (VPM), paracentral (PC), and laterodorsal (LD) thalamic nuclei ([Fig. 3A](#pone-0016891-g003){ref-type="fig"}). ::: {#pone-0016891-g003 .fig} 10.1371/journal.pone.0016891.g003 Figure 3 ::: {.caption} ###### Immunostaining of DDAH-1 in the diencephalon of *T. b. b.-*infected rats. \(A) Immunoreactivity of DDAH-1 observed in different hypothalamic and thalamic structures: perifornical nucleus (PeF), magnocellular nucleus of the lateral hypothalamus (MCLH); lateral hypothalamic area (LH); subincertal nucleus (SubI); zona incerta (ZI); reuniens thalamic nucleus (Re); central medial thalamic nucleus (CM); posterior thalamic nuclear group (Po); ventral posterolateral (VPL) and posteromedial (VPM) thalamic nucleus; paracentral thalamic nucleus (PC); laterodorsal thalamic nucleus (LD); and habenular nucleus (Hb). (B) Immunostaining of DDAH-1 (B-b, d) in kidney of healthy rats (control of specificity for anti-DDAH-1). For control for the specificity of the secondary antibody, kidney sections were incubated without anti-DDAH-1 (B-a, c). Number of animals, n = 3. The microscope used was a Nikon Eclipse E400 equipped with Sony DXC-390P camera (objective: X20, X40). Abbreviations: DDAH-1, N^G^, N^G^-dimethylarginine dimethylaminohydrolase isoform 1; see also [figure 2](#pone-0016891-g002){ref-type="fig"}. :::  ::: For DDAH-2, a cytosolic immunoreactivity was observed in the neurons of submedial (Sub), mediodorsal (MD), ventral posterolateral (VPL), posteromedial (VPM), laterodorsal (LD), paracentral (PC), posterior (Po), ventromedial (VM), ventrolateral (VL), and centrolateral (CL) thalamic nuclei ([Fig. 4A](#pone-0016891-g004){ref-type="fig"}). Unlike DDAH-1 and in agreement with western blot data ([Fig. 2 B](#pone-0016891-g002){ref-type="fig"}), our immunohistochemical study revealed an increase in DDAH-2 protein expression in infected rats in comparison to non-infected animals (see the example of the ventral and posterior part of the thalamus in [Fig. 4B](#pone-0016891-g004){ref-type="fig"}). ::: {#pone-0016891-g004 .fig} 10.1371/journal.pone.0016891.g004 Figure 4 ::: {.caption} ###### Immunostaining of DDAH-2 in the diencephalon of *T. b. b.-*infected rats. \(A) Immunohistochemistry for DDAH-2 observed in different thalamic structures: submedius thalamic nucleus (Sub); mediodorsal thalamic nucleus (MD); ventral posterolateral and posteromedial thalamic nucleus (VPL, VPM); laterodorsal thalamic nucleus (LD); paracentral thalamic nucleus (PC); posterior thalamic nuclear group (Po); ventromedial thalamic nucleus (VM); ventrolateral thalamic nucleus (VL); centrolateral thalamic nucleus (CL). (B) Comparison of the DDAH-2 protein expression shown by immunohistochemistry in the ventral posterior nucleus of the thalamus between infected animals and control healthy rats. (C) Immunostaining of DDAH-2 (B-b, d) in the kidney of healthy rats (control of specificity of anti-DDAH-2). The specificity of the secondary antibody was verified by incubating kidney sections without anti-DDAH-2 (B-a, c). Number of animals, n = 3. The microscope used was a Nikon Eclipse E400 equipped with Sony DXC-390P camera (objective: X20, X40). Abbreviations: DDAH-2, N^G^, N^G^-dimethylarginine dimethylaminohydrolase isoform 2; see also [figure 2](#pone-0016891-g002){ref-type="fig"}. :::  ::: The specificity of anti-DDAH-1 and anti-DDAH-2 antibodies used in the brain was verified by deploying the same immunohistochemical procedure in a rat kidney. DDAH was, indeed, predominantly expressed in this organ ([Fig. 3B](#pone-0016891-g003){ref-type="fig"}). For DDAH-1, the Immunostaining reactivity was exhibited in the cytoplasm of the proximal convoluted tubular cells ([Fig. 3B-b, d](#pone-0016891-g003){ref-type="fig"}). For DDAH2, the immunostaining was localized in the distal convoluted tubular cells and the cortical and inner medullary collecting ducts ([Fig. 4C-b, d](#pone-0016891-g004){ref-type="fig"}). A lack in staining was observed when either the anti-DDAH-1 or anti-DDAH-2 antibody were omitted in the immunohistochemical procedure ([Fig. 3B-a, c](#pone-0016891-g003){ref-type="fig"} and [4C-a, c](#pone-0016891-g004){ref-type="fig"}). ### Morphometric analysis of DDAH-1 and DDAH-2 diencephalic expressing cells {#s3b4} In the ventral and posterior parts of the thalamus, a Fisher\'s post-hoc analysis revealed that, in infected rats, the surface area of the soma and perimeter of cells expressing DDAH-2 were significantly larger *versus* those expressing DDAH-1 ([Table 3](#pone-0016891-t003){ref-type="table"}). These two groups of cells exhibited, however, a very close form factor (FF) ([Table 3](#pone-0016891-t003){ref-type="table"}). Considered as a whole, the analyses confirmed that DDAH-1 expressing cells belong to the interneuron cellular type. ::: {#pone-0016891-t003 .table-wrap} 10.1371/journal.pone.0016891.t003 Table 3 ::: {.caption} ###### Morphometric parameters for differentiation of DDAH-1 and DDAH-2 cell groups in *Trypanosoma brucei brucei*-infected Wistar rats. ::: {#pone-0016891-t003-3} Soma area (µm^2^) Soma perimeter (µm) Soma FF ----------------------------------------------------- ------------------- --------------------- ---------------------------------------- **DDAH-1 (** ***n = 134*** **)** 33.33±0.61 26.01±0.36 0.63±0.01 **DDAH-2 (** ***n = 134*** **)** 125.51±3.35 52.80±0.94 0,57±0,09 ***P*** **value** [\*](#nt106){ref-type="table-fn"} \<0.0001 \<0.0001 N.S[\*\*](#nt108){ref-type="table-fn"} \*The values given represent the mean ± standard deviation; a Fisher\'s post-hoc analysis was achieved. FF: The form factor. \*\*N.S: Non significant. ::: Expression of the two arginase isoforms {#s3c} --------------------------------------- Since both isoforms of arginase are expressed in the brain, we tried to confirm the lack of differences in arginase activity in the diencephalic brain structures (hypothalamus/thalamus) between control and infected rats. In this respect, protein (western blot) and mRNA (real-time RT-PCR) levels of arginase-1 and arginase-2 were analyzed in diencephalic brain structures during the time course of the infection. ### Western blot analysis of arginase-1 and arginase-2 {#s3c1} In agreement with data on arginase activity, arginase-1 and arginase-2 protein levels did not differ between control and infected rats throughout the time course of the infection ([Fig. 2E](#pone-0016891-g002){ref-type="fig"}). ### Analysis of the mRNA of arginase-1 and arginase-2 {#s3c2} As expected from data on arginase activity and western blot analysis, no significant changes were observed in diencephalic brain structures for mRNA transcripts of arginase-1 and arginase-2 ([Fig. 2F](#pone-0016891-g002){ref-type="fig"}). Changes occurring in the amino acids content throughout the time course of the disease {#s3d} -------------------------------------------------------------------------------------- In diencephalic brain structures, the main amino acids involved in the metabolic pathways catalyzed by NOS, DDAH and arginase were quantified from D5 to D22 after infection. In infected rats, the variations of relative concentrations of L-arginine and L-citrulline were in the same range. Compared to healthy control rats, their concentrations decreased significantly at D10 (L-arginine, 65.87±17.07%, p\<0.05; L-citrulline, 57.76±5.03%, p\<0.001; n = 6) and D16 (L-arginine, 82.33±4.39%, p\<0.05; L-citrulline, 70.52±8.17%, p\<0.05; n = 6). At D22 post infection, the relative concentrations of the two amino acids returned to control values ([Fig. 5](#pone-0016891-g005){ref-type="fig"}). No significant change was observed in the relative concentrations of L-glutamate, L-glutamine (data not shown), GABA, L-proline and L-ornithine in infected rats throughout the investigation ([Fig. 5](#pone-0016891-g005){ref-type="fig"}). The presence of ADMA in diencephalic brain structures was not detectable (data not shown). ::: {#pone-0016891-g005 .fig} 10.1371/journal.pone.0016891.g005 Figure 5 ::: {.caption} ###### Changes occurring in the amino acid content in the brain during the course of infection in *T. b. b.*-infected rats. The variations of the concentration of amino acids (mean value ± SEM) in the diencephalon (hypothalamus/thalamus) of infected rats were normalized with the corresponding values obtained in healthy control animals. 100% of the control value (delineated by the dotted line) corresponds to: 34±2 µM for L-arginine; 10±1 µM for L-citrulline; 892±36 µM for L-glutamate; 796±80 µM for L-GABA; 16±1 µM for L-proline and 5±0.4 µM for L-ornithine. Concentration of ADMA was undetectable. Statistics: ANOVA followed by post-hoc Fisher\'s test (*\*p\<0.05*, *\*\*p\<0.001* compared to healthy rats); n = 6 animals in each group at the different times post-infection: D5, D10, D16 and D22. Abbreviations: GABA, γ-aminobutyric acid; see also [figure 2](#pone-0016891-g002){ref-type="fig"}. :::  ::: Amino acids concentrations were also analyzed in the blood. In infected rats, the relative concentrations of L-arginine and L-citrulline decreased significantly at D5, D10 and D16 ([Fig. 6](#pone-0016891-g006){ref-type="fig"}). At D22, the concentration of L-citrulline remained lower than control values (73.87±13.67%, p\<0.05, n = 6), while that of L-arginine increased slightly above control values (128.76±14.37%, p\<0.05) ([Fig. 6](#pone-0016891-g006){ref-type="fig"}). L-glutamate ([Fig. 6](#pone-0016891-g006){ref-type="fig"}) and L-glutamine (data not shown) relative concentrations decreased significantly, from D10 (L-glutamate, 66.44±10.61%, p\<0.05; L-glutamine, 60.06±13.10%, p\<0.05; n = 6) to D16 (L-glutamate, 51.15±17.45%, p\<0.001; L-glutamine, 33.25±16.71%, p\<0.001; n = 6), and D22 (L-glutamate, 43.24±24.55%, p\<0.001; L-glutamine, 18.90±4.04%, p = 0.0001; n = 6). Plasma relative concentration of ADMA increased significantly throughout the time course of the illness, from D10 (155.96±30.44%, p\<0.05, n = 6) to D16 (182.51±37.99%, p\<0.001, n = 6), and D22 (207.91±48.58%, p\<0.001, n = 6), ([Fig. 6](#pone-0016891-g006){ref-type="fig"}). ::: {#pone-0016891-g006 .fig} 10.1371/journal.pone.0016891.g006 Figure 6 ::: {.caption} ###### Relative changes in the content of amino acid in blood during the course of infection in *T. b. b.*-infected rats. The changes in the concentration of amino acids (mean value ± SEM) in the blood of infected rats were normalized with the corresponding values obtained in healthy control animals. 100% of the control values (delineated by the dotted line) correspond to: 110±3 µM for L-arginine; 100±4 µM for L-citrulline; 95±4 µM for L-glutamate and 0.96±0.1 µM for ADMA. Statistics: ANOVA followed by post-hoc Fisher\'s test (*\*p\<0.05*, *\*\*p\<0.001*, *\*\*\*p\<0.0001* compared to healthy rats); n = 6 animals in each group at the different times post-infection: D5, D10, D16 and D22. Abbreviations: ADMA, asymmetric dimethylarginine; see also [figure 2](#pone-0016891-g002){ref-type="fig"}. :::  ::: Discussion {#s4} ========== First of all, we must emphasize the adequacy of the *T. b. brucei*-infected rat model in mimicking closely the two-stage disease in human *T. b. gambiense* infection. As previously reported in our rat model, the initiation of the neurological stage of the disease is characterized by a marked decrease in body weight gain and the presence of trypanosomes in the CSF 10 days post-infection [@pone.0016891-Amrouni1]. The neurological stage is also marked by an increase in cerebral iNOS activity, leading to an overproduction of NO in the brain. By use of double labeling, we have also shown that the iNOS immunoreactivity is located in glial elements (microglia and astrocytes) as well as in neurons, especially in brain structures involved in the sleep/wake regulation (lateral and posterior part of the hypothalamus, perifornical area) [@pone.0016891-Amrouni1]. The present investigation was designed to study the molecular changes occurring in arginase and DDAH in the brain at the key stages of the HAT infection. We suggest that DDAH most likely contributes to the regulation of iNOS activity in the brain, whereas arginase has a slight effect, if any, contrary to its role in the periphery. Arginase pathway {#s4a} ---------------- Concerning the arginase pathway, experimental data obtained in the brain indicate that, throughout the time course of the disease, no significant changes occurred in the total activity of arginase in the protein or in the mRNA levels of the two isoforms of the enzyme (arginase-1 and -2). This is also evident from the absence of any significant changes in the concentration of the major amino acids involved in the arginase pathway, such as L-ornithine, L-proline, L-glutamate, L-glutamine, and L-GABA. Since the arginase activity remained unchanged throughout the course of the experimental illness, the decrease in L-arginine content, observed at D10 and D16 post-infection, may be due to the utilization of the substrate through the iNOS pathway [@pone.0016891-Amrouni1]. Nevertheless, we cannot exclude a possible increase in the arginosuccinate synthetase (ASS) and arginosuccinate lyase (ASL) activities which would favor an increase in the concentration of L-arginine by recycling L-citrulline to L-arginine. Such a mechanism has been reported in rats subjected to brain excitotoxicity mediated by kainic acid. An increase in the production of NO in parallel to ASS and ASL activities takes place in the cerebral cortex, cerebellum and brain stem, while the arginase activity remains unchanged [@pone.0016891-Swamy1]. These mechanisms remain to be further investigated. In the present approach, we did not examine the changes occurring in arginase activity in the periphery, as such changes have already been addressed [@pone.0016891-Gobert1], [@pone.0016891-Duleu1]. In this respect, we suggested [@pone.0016891-Amrouni1] that arginase activity changes occurring in the periphery may reflect the strategy developed by trypanosomes to invade the host and insure their own survival and development through (i) enhancement of arginase activity mediating the synthesis of polyamines necessary to trypanosome growth, and (ii) limitation of trypanocidal properties carried out by the decrease in NO production. Arginase appears therefore to be a critical regulator of NO synthesis, through a competition with iNOS for L-arginine [@pone.0016891-Durante1]. In this respect, a significant decrease in plasma L-arginine concentration was observed in our infected rats. Such a change was also reported in *T. b. brucei*-infected mice [@pone.0016891-Gobert1] and *T. b. gambiense*-infected voles [@pone.0016891-Newport1]. A large proportion of the L-arginine in the blood might be metabolized by arginase since iNOS is inhibited during the time course of the infection [@pone.0016891-Amrouni1], reducing NO production and, consequently, trypanocidal pressure. In support of this interpretation, L-arginine supplementation in *T. b. brucei*-infected mice restores both NO production and NO trypanocidal activity [@pone.0016891-Gobert1], [@pone.0016891-Duleu1]. Furthermore, in the periphery, the arginase genes are up-regulated throughout the course of the infection. Arginase-1 and, to a lower extent, arginase-2 genes can be induced directly by the trypanosomes. Moreover, interleukin-4 (IL-4), IL-10 and Transforming Growth Factor-β (TGF-β), which are potent arginase inducers, are enhanced in experimental trypanosomiasis [@pone.0016891-Uzonna1]. Alternatively, arginase activation can also result from a direct interaction between trypanosome products, such as the trypanosome-derived lymphocyte-triggering factor (TLTF), and T lymphocytes [@pone.0016891-DeBaetselier1]. In any case, arginase activation enhances the production of polyamines that are essential for the synthesis of the trypanothione, a compound necessary for growth and survival of trypanosomes ([Fig. 1](#pone-0016891-g001){ref-type="fig"}). Trypanothione synthesis from glutathione and spermidine [@pone.0016891-Mller1] may account for the significant decrease in glutamate and glutamine concentration observed at the peripheral level in our infected rats. In the brain, the situation does not appear to be favorable for trypanosome multiplication and spread, since trypanocidal NO increases and arginase activity (necessary for trypanosome growth) remain stable. These mechanisms may act as additive protective elements for the brain, although trypanosomes will still be able to enter the central nervous system. The parasites pass through the blood-brain barrier (BBB) across or between the endothelial cells and the vessel basement membranes [@pone.0016891-Masocha1]. The trypanosomes are present in cerebrospinal fluid (CSF), especially in such areas where the BBB is weak, i.e., at the level of the circumventricular organ and around the ventricular cavities [@pone.0016891-Amrouni1], [@pone.0016891-Abolarin1], [@pone.0016891-Ormerod1]. Due to the large concentration of NO in the extracellular space of the brain parenchyma [@pone.0016891-Amrouni1], it is likely that trypanosomes cannot penetrate easily into deeper structures. Furthermore, trypanosomes present in the brain parenchyma are described as being damaged [@pone.0016891-VanMarck1], underlining potential difficulties to survive in such conditions. Intracellular localization of African trypanosomes may also protect the parasites against the trypanocidal properties of NO [@pone.0016891-Mattern1], [@pone.0016891-Stoppini1]. Due to the lack of changes observed in the arginase activity during the course of the infection, the energy necessary for trypanosome growth may be supplied by the high rate of the brain metabolism [@pone.0016891-Cespuglio1]. DDAH/ADMA pathway {#s4b} ----------------- Similar to our observations on the arginase pathway, the DDAH/ADMA pathway may also play a key role in the regulation of NO production since ADMA acts as a potent iNOS inhibitor [@pone.0016891-Dayoub1]. Indeed, we demonstrate that the cerebral activity of DDAH, mRNA and protein expression increase significantly in the later stage of the disease (at D10 and D16 after infection) concomitant with the activation of iNOS (activity and protein expression) and the enhancement of NO production [@pone.0016891-Amrouni1]. The tight correlation existing between these events supports the hypothesis of a contribution of DDAH in the control of NO production (particularly at D16) in the brain during the invasion by trypanosomes. In line with this view, it was suggested that, in inflammatory conditions (such as in the present case), the activation of DDAH may contribute to reduce the level of ADMA [@pone.0016891-Tran2].There are also reports indicating that ADMA concentration is significantly lower (−48%) in the CSF of patients suffering of Alzheimer\'s disease or in pathological aging with cognitive impairment [@pone.0016891-Abe1]. Trypanosomiasis might thus also contain insidious neurodegenerative processes and therefore represent a model of neurodegenerative diseases. In our experimental model, the cerebral increase in DDAH activity is mainly dependent on the DDAH-2 isoform which is transcriptionally regulated, a specific increase in DDAH-2 mRNA occurring throughout the infection. The DDAH-2 gene is located on chromosome 6 (6p21.3) in a region containing various genes involved in inflammatory processes linked with the autoimmune disease susceptibility [@pone.0016891-Tran1]. This localization and the wide expression of DDAH-2 in immune cells suggest that the DDAH-2 gene may represent a potential disease-susceptibility gene. In this sense, it has been reported that inflammatory cytokines such as IL-1β are capable of simulating simultaneously iNOS and DDAH in vascular smooth muscle cells [@pone.0016891-Ueda1]. This effect is accompanied by an increase in NO metabolites and a decrease in ADMA content in culture media. A chronic overexpression of proinflammatory cytokines (IL-1, TNF-α) and IFN-γ in the brains of *T. b.* -infected rats, might be responsible for the iNOS and DDAH-2 up-regulation, leading to a reduction of trypanosome proliferation, neuroinflammation and neurodegeneration [@pone.0016891-Quan1]. In diencephalic brain structures, DDAH-2 immunoreactivity was primarily observed in neurons. It was enhanced in infected rats as compared with healthy animals. Contrary to DDAH-2, no significant change was noticed with DDAH-1 immunoreactivity between infected and non-infected animals. It should be noted that DDAH-1 immunoreactivity is located in interneurons which express also nNOS [@pone.0016891-Bertini1], [@pone.0016891-Koliatsos1], an enzyme that remains unchanged throughout the *T. b. b.* infection process [@pone.0016891-Amrouni1]. Recently, Gow et al. [@pone.0016891-Gow1] reported that NO may S-nitrosylate a distinct subset of cellular proteins, DDAH being one of them. The cellular proteins may be regulated by S-nitrosylation through the NO derived from iNOS [@pone.0016891-Leiper1]. Under certain conditions, when NO increases, the S-nitrosylation of DDAH decreases its activity. This mechanism would lead to an accumulation of ADMA and consequently to the inhibition of the iNOS. The S-nitrosylation of DDAH provides thus a potential feedback mechanism for the regulation of NO production. This mechanism could explain why, in our experimental model of HAT, the activities of both iNOS and DDAH return simultaneously to the control level at D22 after infection. Such an effect may reflect the *pre-mortem* situation of our infected animals at that time of the infection course. In such extreme conditions, anti-inflammatory processes may be triggered to protect the brain as reported in autoimmune encephalitis [@pone.0016891-Cua1]. Peripheral ADMA concentration was elevated in our *T. b. brucei-*infected rats. These changes coincide with the entry into the second stage of the disease and correlate with the degree of the infection. The mechanism responsible for the increase in the ADMA level was not specifically investigated in this study. It may be explained, at least partly, by the existence of a decrease in the concentration of L-citrulline which occurs in parallel with the ADMA increase. It is also known that ADMA accumulation is associated with an endothelial dysfunction [@pone.0016891-Fliser1]. Recent *in vitro* and *in vivo* studies indicate that an enhanced ADMA synthesis coincides with an increased activity of the immune pathways, such as in HIV infected patients [@pone.0016891-Schroecksnadel1], [@pone.0016891-Kurz1]. In this respect, ADMA can be produced by the endothelial cells and the activated mononuclear blood cells [@pone.0016891-Kurz1]. The latter cellular elements release simultaneously Th1-type cytokine, interferon-γ (IFN-γ, an important NOS inducer) and ADMA [@pone.0016891-Schroecksnadel1]. It has also been reported that even modest changes in the ADMA level can have significant effects on NO synthesis [@pone.0016891-Dayoub1]. In our previous studies [@pone.0016891-Amrouni1], we reported that circulating NO concentration decreases significantly from D10 post-infection (in correlation with illness progression) and that this decrease is likely to depend on an impaired iNOS activity in macrophages. We suggest that ADMA may also contribute to the inhibition of NO synthesis in the periphery. ### Conclusion {#s4b1} In the *T. b. brucei-*infected rat model of human African trypanosomiasis, we demonstrated that the activity of cerebral arginase remains constant throughout the time course of the infection. Simultaneously, DDAH-2 isoform increased, contributing to an overproduction of NO, undoubtedly through a reduction of the ADMA level, a known iNOS inhibitor. These changes differed from those reported in the periphery in the same animal model ([Fig. 7](#pone-0016891-g007){ref-type="fig"}). They may constitute an additive protection against the trypanosomes that have entered the central nervous system, through the increase in the trypanocidal pressure carried by NO, while the arginase activity necessary for trypanosomes growth remained unchanged. The involvement of NO overproduction in the pathophysiology of the neurological stage of trypanosomiasis might also be considered. ::: {#pone-0016891-g007 .fig} 10.1371/journal.pone.0016891.g007 Figure 7 ::: {.caption} ###### Schematic representation of the enzymatic pathways involved in the control of trypanosome entry into the brain during the later stage of the disease. In the periphery (blood), the arginase is strongly expressed in the activated macrophages and competes with iNOS for a common substrate, L-arginine. This phenomenon, under control of cytokines, leads to a decrease in iNOS activity and consequently in the production of NO, a gaseous messenger possessing trypanocidal properties. Such a mechanism, favoring the growth and multiplication of the parasites, is strengthened by the accumulation of the endogenous inhibitor of iNOS, ADMA, as observed in the blood. An opposite mechanism is observed in the brain since the activity of iNOS increases in the three cell types: neurons, microglia and astrocytes. DDAH-2 is over expressed in neurons, likely by decreasing the ADMA pool, and may enhance iNOS activity and, consequently, NO production. Concomitantly, the activity and expression of the arginase remain unchanged. Abbreviations: NO, nitric oxide; iNOS, inducible nitric oxide synthase; ADMA, asymmetric dimethylarginine; DDAH, N^G^, N^G^-dimethylarginine dimethylaminohydrolase. Small arrows indicate an increase in the given product when pointing upwards and a decrease when pointing downwards. Hyphen is used to indicate that the amount in the given molecule is stable. :::  ::: We would like to thank Dr Lucienne Léger and the ANIPATH platform for help in immunohistochemical processing. We also thank Professor MW. Radomski, from the University of Toronto, who kindly accepted to review the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**University of Lyon. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: DA AM SGS MP BB PV AB RC. Performed the experiments: DA AM SGS RC. Analyzed the data: DA SGS AM PV BB AB RC. Contributed reagents/materials/analysis tools: DA RC. Wrote the paper: DA AM SGS RC BB PV AB MP. Administration attached to the manuscript submission: RC. Payment of the discounted fees: RC.

PubMed Central

2024-06-05T04:04:19.106491

2011-3-9

PMC3052301

Introduction {#s1} ============ "Barker Hypothesis" states that reduced foetal growth and development resulting from maternal malnutrition are associated with a number of chronic conditions in the later life of mammals [@pone.0017443-Barker1]. On the other hand, modern life in developed societies is not characterized by malnourishment, but conversely by overnutrition, associated with the development of an obese phenotype. It has been postulated that this should also result in health disturbance in the next generation. Observational human studies seem to confirm such postulates; but exact verifiable acquisitions of human data require decades and enormous work to reach statistically robust conclusions [@pone.0017443-Baker1], [@pone.0017443-Martorell1], [@pone.0017443-Reynolds1]. Consequently, customized animal studies with mice or rats may clarify such postulated connections. Very recently, Chang *et al.* [@pone.0017443-Chang1] described that maternal high-fat diet in rats results in long-term behavioural and physiological changes in offspring including increase of food intake, preference for fat, hyperlipidemia, and higher body weight. The mouse study presented here was initiated to determine the short- and long-term effects of a maternal high-protein diet in pregnancy or lactation on offspring\'s body weight, fat accumulation and cardiovascular parameters after weaning. As a result of our study we found the predicted, significant influence of the maternal nutrition on the development of offspring in their later life. However, much more dramatic are direct effects initiated by nutrition of pregnant animals reducing the number of live-born pups, and by the nutrition of dams during lactation significantly increasing the mortality of sucklings. From the latter we conclude that a high-protein diet during lactation strongly promotes the incidence of a sudden infant death-like syndrome. These findings point to a fundamentally new approach for the research on and prevention of redoubtable sudden infant death syndrome (SIDS) [@pone.0017443-Mitchell1], [@pone.0017443-Ostfeld1], [@pone.0017443-Prandota1]. Materials and Methods {#s2} ===================== Mice line {#s2a} --------- Outbred line DUK [@pone.0017443-Dietl1], [@pone.0017443-Timtchenko1] was bred at the Leibniz Institute for Farm Animal Biology (FBN) in Dummerstorf, Germany. Experiments on adult mice were performed in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85-23, revised 1996) and the Federal Law on the Use of Experimental Animals in Germany, and were approved by the local authorities (Reg.-Nr. LALLF M-V/TSD/7221.3-1.1-033/06; Reg.-Nr. G 0258/05). Sets of experimental mouse groups {#s2b} --------------------------------- Fifty male and female mice received control food (C, 21% protein) and 25 males and females were fed a high protein diet (HP, 42% protein) during mating ([**Figure 1**](#pone-0017443-g001){ref-type="fig"}). Mice were allocated to the two experimental diets in a random manner; experimental diets were fed already during conception. Males were withdrawn immediately after confirmation of pregnancy (appearance of a vaginal plug), which was denoted day 1 of pregnancy. After fertilisation/impregnation, females stayed on their respective diet until weaning. At birth, pups were allocated to standardized litters (8♂, 2♀) and put back to foster mothers (with delivery, mothers became foster mothers and were scaled down to 20 with C- and 10 with HP-diet). Consequently, the resulting experimental groups have been as followed: I. Offspring born to mouse dams fed control diet (C) during pregnancy was cross-fostered by dams that received the high protein (HP) diet during pregnancy and continued on the HP diet during lactation (C-HP), II. A group whose mothers were given the HP diet during pregnancy was cross-fostered by dams fed control diet throughout pregnancy and lactation (HP-C), and III. A control group (C-C) with mothers fed control diet in pregnancy was cross-fostered by dams fed control diet in pregnancy and lactation. ::: {#pone-0017443-g001 .fig} 10.1371/journal.pone.0017443.g001 Figure 1 ::: {.caption} ###### Scheme describing the generation of the different diet groups. Detailed description on the generation of the C-C, C-HP, and HP-C groups has been given in the [Materials and Methods](#s2){ref-type="sec"} and the Results parts. C  =  control diet, HP  =  high-protein diet. :::  ::: Pregnant and lactating dams were housed individually in standard rodent cages (Makrolon, Type II, EBECO, Castrop-Rauxel, Germany) in a controlled environment of 22°C with a 12∶12 h light/dark cycle. Water was available *ad libitum*. Diets {#s2c} ----- Two isoenergetic (16.3 MJ metabolizable energy/kg dry matter) semi-synthetic experimental diets with control and high protein level were fed. The two different diets consisted of casein (Molkereigesellschaft Lauingen mbH, Lauingen, Germany; C, 212 g/kg; HP 426 g/kg) supplemented with DL-methionine (4 g/kg; LAH GmbH & CO. KG, Cuxhaven, Germany), wheat starch (Ferdinand Kreutzer Sabamühle GmbH, Nürnberg, Germany; C, 443.9 g/kg; HP 225.9 g/kg), sucrose (Nordzucker GmbH, Hamburg, Germany; 160 g/kg), soya oil (Sedina ADM, Hamburg, Germany; 50 g/kg), microcellulose (50 g/kg), vitamin mixture (20 g/kg; SSNIFF Spezialdiäten GmbH, Soest, Germany), mineral mixture (60 g/kg; SSNIFF Spezialdiäten GmbH) and butylhydroxytoluene (0.1 g/kg; LAH GmbH & CO. KG). After weaning, a standard rodent diet was fed to all mice (21% protein, 0.4% L-Met, 55% starch, 5% sucrose, 5% fat; 5% cellulose, 2% vitamin and 6% mineral mixture; Altromin 1314, Altromin Spezialfutter GmbH & Co. KG, Lage, Germany). Glucose tolerance test {#s2d} ---------------------- At day 150, mice (n≥7 per group) were starved for 12 h. They were fed a glucose solution (1 g/kg body weight; concentration of glucose solution: 200 mg/ml) via oral gavage. Blood samples were taken from the tail-tip. Blood glucose measurement was performed with a Glucometer (Bayer Ascensia Elite®). Values were taken 30 min before feeding (fasted value) and again 5, 10, 15, 30, 60, 120, 180, 240, 300 and 360 min post glucose challenge. Cardiovascular characterization via tip-catheter {#s2e} ------------------------------------------------ At day 360, tip-catheter was inserted via Arteria carotis, Arcus aortae, and Valva aortae into the left ventricle as described previously (n≥7 per group) [@pone.0017443-Wang1]. Measurement included e.g. heart rate (HR), left ventricular pressure (LVP), contraction, systolic blood pressure (SBP) and diastolic blood pressure (DBP). Body weight and abdominal fat determination {#s2f} ------------------------------------------- Litters were weighted day 1 after birth, and single mice at weaning (day 21), day 180, and day 360. After culling the mice, abdomen was opened up, and fat pad connected to testicles was taken and weighted (n = 8 per group). Expression of whey acidic protein expression {#s2g} -------------------------------------------- Whey acidic protein (WAP) expression has been measured on mRNA level as described earlier (n≥8 per group) [@pone.0017443-Gors1]. Statistics {#s2h} ---------- Statistical calculations were performed using Prism or Instat software from GraphPad, San Diego (USA). Comparison of means was determined by student t tests. The variances of results were calculated by ANOVA tests. Significant differences between selected groups were calculated by Bonferroni posttests. To calculate the significances in the survival during lactation, the Friedman test has been used. Kaplan-Meier survival curves were used to compare the survival rate between the groups, and logrank test was performed to analyze the statistical difference. *P*\<0.05 values were considered to be statistically significant. Results and Discussion {#s3} ====================== To investigate the relation between maternal diet composition and offspring growth development and health status, we generated, as described in [Materials and Methods](#s2){ref-type="sec"}, three offspring groups depending on the diet of their mother during pregnancy or their foster mother during lactation. On initiation of the study, 50 (2×25) females at fertilization were fed with control diet, whereas 25 fertilized mice were on high-protein diet (isoenergetic to control diet; [**Figure 1**](#pone-0017443-g001){ref-type="fig"}). The body weight gain of the dams under high-protein diet during pregnancy (weight after delivery minus weight before pregnancy) was only half of that of the dams on a control diet ([**Figure 2a**](#pone-0017443-g002){ref-type="fig"}). Importantly, the impact on body weight was seen in both the pregnant animals and also in their offspring. As shown in [**Figure 2b**](#pone-0017443-g002){ref-type="fig"}, body weight of the newborns was significantly lower than that of newborns where mothers were on the control diet. Interestingly, the number of pups per litter was also significantly lower in the high-protein fed group ([**Figure 2c**](#pone-0017443-g002){ref-type="fig"}), implicating an effect of that diet on intrauterine development and mortality. This is in accord with previously reported studies, showing that higher maternal dietary protein intake in pregnancy is associated with lower ponderal index at birth [@pone.0017443-Andreasyan1]. ::: {#pone-0017443-g002 .fig} 10.1371/journal.pone.0017443.g002 Figure 2 ::: {.caption} ###### Effect of diets during pregnancy and birth. At day 1 after birth, mothers and litters were examined (group C \[n = 50\] as the base for C-C and C-HP, group HP \[n = 25\] is the base for HP-C). (**a**) Body weight gain of females during pregnancy as difference of weight after delivery vs. day of mating; Litter size (**b**) and birth weight (**c**) depending on diet of mother during pregnancy. Data are the means ± SEM, \**P*\<0.05, \*\*\**P*\<0.001 vs control food (C) using a Student\'s *t*-test. :::  ::: Immediately after birth, litters were standardised to 10 pups per dam (8 males, 2 females per litter). Thus, each of the three diet groups carried 80 male and 20 female offspring until weaning (21^st^ day of lactation; [**Figure 1**](#pone-0017443-g001){ref-type="fig"}). Although newborns from HP pregnancies were characterized by lower body weight ([**Figure 2b**](#pone-0017443-g002){ref-type="fig"}), this difference was abrogated until weaning due to the switch to a foster mother that was fed control diet during pregnancy and lactation ([**Figure 3a**](#pone-0017443-g003){ref-type="fig"}), suggesting that HP diet during pregnancy has a less impact on body weight gain at weaning than HP diet during pregnancy. This was further supported by the finding that the newborns from the control diet group during prenatal development, exposed to HP foster mothers only after delivery (C-HP), were characterized by massively lower body weight at weaning in comparison to the group exposed to HP diet during *in utero* life and with fosters, which received control diet during lactation ([**Figure 3a**](#pone-0017443-g003){ref-type="fig"}). ::: {#pone-0017443-g003 .fig} 10.1371/journal.pone.0017443.g003 Figure 3 ::: {.caption} ###### Effect of diets during lactation on body weight, survival and milk composition. Litters were cross-fostered and standardized (8♂, 2♀) at birth. Resulting from dietary treatments 80 males and 20 females counted per group (control-control: C-C, high protein-control: HP-C, and control-high protein C-HP). (**a**) Pup body weight at weaning (21^st^ day of life) depending on diet group (C-C: 98 pups; HP-C: 98 pups; C-HP: 83 pups). (**b**) Number of death in lactation in the three examined groups shown in % (of 100 pups) for period from birth till weaning and (**c**) Number of death in lactation in the same period subdivided in five-day steps; (**d**) mRNA levels of whey acidic protein in mammary glands was measured in lactating mice on control diet (C; n = 8) or high-protein diet (HP; n = 9). Data are the means ± SEM; group differences are calculated by ANOVA (**3a**), Friedman test (**3c**; *P* = 0.018 C-HP *vs* C-C and HP-C) or Student\'s t-test (**3d**), \**P*\<0.05 *vs* control food (C), \*\*\**P*\<0.001 *vs* control group (C-C). :::  ::: The most striking finding in the offspring group with high-protein diet during lactation was a significant association with sucklings\' survival (*P* = 0.018). During lactation, 17% of all C-HP pups died (17 pups), whereas only 2% of the pups suckled by control-fed dams died (2 pups, [**Figure 3b**](#pone-0017443-g003){ref-type="fig"}), independently of their dietary exposure during *in utero* life. Notably, prior to death, none of the sucklings was characterized by obvious signs of discomfort or illness. This data clearly points to a high risk for pups to die during lactation suckled by dams (foster mothers) fed with high-protein diet. This is further supported by detailed analyses of the occurrence of sucklings\' death during the lactation period. Divided in 5-day blocks, difference in survival occurred within the whole lactation phase, with highest difference within the second half ([**Figure 3c**](#pone-0017443-g003){ref-type="fig"}) where mortality further increased in the C-HP group while no further mortality has been observed in both other groups. Notably, the higher incidence for death has been distributed over all litters on high-protein diet. Since investigations on dead sucklings did not reveal morphological changes, we sought for explanations for the sudden death in the C-HP group. Whey acidic protein (WAP) is the major milk protein in certain mammals including mice. In an independent study, we measured mRNA concentration in mammary glands at peak lactation (day 14 of the lactation period) and found only less than 25% of WAP mRNA abundance in foster mothers receiving HP diet throughout pregnancy and lactation compared to mammary glands isolated from mothers on control diet ([**Figure 3d**](#pone-0017443-g003){ref-type="fig"}). Thus, our data implicates a clear relation between high-protein diet during pregnancy/lactation, lactogenesis, and sucklings\' mortality. To check for life-long effects of high-protein diet during pregnancy or lactation, we followed the remaining offspring of all 3 dietary groups (98 C-C; 98 HP-C; 83 C-HP) for one year and measured body weight after 180 and 360 days and biochemical and physiological parameters at the endpoint of the experiment (360 days). The distinctly lower body weight observed in the C-HP group after 21 days of nursing ([**Figure 3a**](#pone-0017443-g003){ref-type="fig"}) did not fully normalize after 5 months of normal diet, since the C-HP group at age of 6 months was still characterized by a lower body weight ([**Figure 4a**](#pone-0017443-g004){ref-type="fig"}). ::: {#pone-0017443-g004 .fig} 10.1371/journal.pone.0017443.g004 Figure 4 ::: {.caption} ###### Survival rate and physiological parameters in adulthood. (**a**) Body weight at day 180; (**b**) Oral glucose tolerance test at day 150 (C-C: n = 10; HP-C: n = 7; C-HP: n = 8); (**c**) Kaplan Meier curve from weaning until end of experiment (d360); (**d**) Cardiovascular characterization of mice at day 360 via Tip-catheter (C-C: n = 8; HP-C: n = 7; C-HP: n = 7). Heart rate, left ventricular pressure (LVP), contractility (dP/dtmax), and mean arterial pressure (MAP); Data are the means ± SEM; group differences are calculated by one way of ANOVA (**4a, 4d**) \**P*\<0.05 *vs* control group (C-C); ^\#^ *P*\<0.05 *vs* high protein-control (HP-C); or two way of ANOVA (**4b**), \*\**P*\<0.01; group differences (diet effects). For Kaplan-Meier analysis (**4c**) logrank calculation has been performed; ^\$^ *P*\<0.05. :::  ::: To investigate metabolic changes in these animals, we checked alterations in glucose tolerance at an age of 150 days. Fasting blood glucose levels under basal conditions were not significantly altered in either the C-HP or HP-C mice. However, the glucose tolerance test showed that HP exposure during pregnancy or lactation provides an advantage for insulin sensitivity ([**Figure 4b**](#pone-0017443-g004){ref-type="fig"}). This agrees with observations that a high-protein diet in adult rats improves glucose tolerance [@pone.0017443-Lacroix1]. In contrast, a glucose tolerance test revealed no remaining significant differences between the three treatment groups (at age 300 days, data not shown). To investigate whether the higher mortality is continued in the C-HP group we calculated survival until the endpoint of our experimental setting (360 days). Kaplan Meier curves revealed no further elevated mortality in the group exposed to maternal HP diet during lactation compared to animals receiving control diet during fetal development and lactation ([**Figure 4c**](#pone-0017443-g004){ref-type="fig"}). However, we found a significantly elevated mortality in animals that were born to dams fed a HP diet during their pregnancy, implicating that protein-rich diet during pregnancy might not only have an impact on survival *in utero* and on birth weight, but might influence survival in later life ([**Figure 4c**](#pone-0017443-g004){ref-type="fig"}). The reason for the unchanged survival in the C-HP group might relate to a balance between the negative effects of high-protein content during lactation and the beneficial effect of a lower body weight in adulthood [@pone.0017443-Greco1], [@pone.0017443-Mitchell2], [@pone.0017443-ThoneReineke1]. We also characterized cardiovascular parameters at the endpoint of the experimental settings. We found no significant differences for all measured parameters, as exemplarily shown for heart rate, contractility, left ventricular pressure (LVP), or mean arterial pressure (MAP) in [**Figure 4d**](#pone-0017443-g004){ref-type="fig"}. We also checked whether the remaining difference in body weight in the C-HP group after 180 days is still existent in an age of 1 year. While there was only a tendency to lower body weight ([**Figure 5a**](#pone-0017443-g005){ref-type="fig"}), the measurement of abdominal fat at 360 days showed significantly less fat accumulation in those animals that were exposed to maternal HP diet during lactation ([**Figure 5b and 5c**](#pone-0017443-g005){ref-type="fig"}). Thus, following the body weight development until one year of age (d21, d180, d360), the reduced body mass in the C-HP group was highly significant (*P*\<0.0001) whereby the difference became less prominent with aging. ::: {#pone-0017443-g005 .fig} 10.1371/journal.pone.0017443.g005 Figure 5 ::: {.caption} ###### Body weight and fat content at day 360. At the end of the experiment, the following parameters have been determined (C-C: n = 8; HP-C: n = 8; C-HP: n = 8): (**a**) Body weight at day 360; (**b**) Abdominal fat in mice; and (**c**) Abdominal fat/body weight ratio at day 360. Data are the means ± SEM, \**P*\<0.05, *vs* control group (C-C). :::  ::: Our data demonstrates that a high-protein diet during pregnancy or lactation has distinct and highly relevant effects. While the high protein during pregnancy may particularly affect embryonic lethality, birth weight, and survival in the second half of life, high-protein diet fed to the lactating mother might also have drastic direct effects on offspring during lactation. These direct effects might be responsible for a higher newborn mortality during lactation. Although our findings on diet effects cannot explain exclusively SIDS, because there is a significant amount of literature indicating higher rates of SIDS for mothers who are not breast feeding [@pone.0017443-Ip1], [@pone.0017443-McVea1], [@pone.0017443-Sullivan1]. Nevertheless, there is broad agreement that the causes for SIDS might be diverse. Therefore, our data offer a first trustable approach to explain the occurrence of SIDS in the group of nursed newborns. Interestingly, an intense debate about relations between maternal nutrition and the occurrence of SIDS has been recently started [@pone.0017443-Golding1], [@pone.0017443-McCowen1], [@pone.0017443-Raiten1]. Therefore, our findings could be a new and unique link between nutrition and the still incompletely understood human SIDS. Consequently, although offspring of lactating mothers on high-protein diet might have the advantage of lower abdominal fat within the second half of life, this benefit seems not to compensate the immense risk of an early sudden death during lactation. Taken together our data **may** implicate that both pregnant women and lactating mothers should not follow classical high-protein diets as Atkin\'s diet. We thank Steven Atkin (Hull, UK) for very helpful discussions. We thank Matthias Münzner (Berlin, Germany) for excellent technical assistance. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the Commission of the European Community, within the FP 6 priority 5.4.3.1, Food quality and safety (EARNEST, Food-CT-2005-007036); and by the Deutsche Forschungsgemeinschaft (DFG): WA1441/18 and SI483/8. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: TW W-ES CCM. Performed the experiments: TW ND ML MK. Analyzed the data: TW HH UR W-ES CCM. Wrote the paper: TW W-ES.

PubMed Central

2024-06-05T04:04:19.110857

2011-3-9

PMC3052302

Introduction {#s1} ============ Repair of mammalian chromosomal double-strand breaks (DSBs) entails use of one of two major repair pathways: non-homologous end-joining (NHEJ) and homologous recombination (HR) (reviewed in [@pone.0016501-Lieber1], [@pone.0016501-Paques1], [@pone.0016501-Sung1], [@pone.0016501-Hartlerode1]). These repair functions take place in relation to other chromosome processes such as transcription and DNA replication, and DNA replication alters the parameters governing DSB repair in several ways. First, the process of DNA replication itself is thought to be a major cause of endogenous DSBs. Second, replication generates a second copy of the chromosome in the form of a sister chromatid, which may favor DSB repair by homologous recombination, ("sister chromatid recombination" -- SCR) [@pone.0016501-GonzalezBarrera1], [@pone.0016501-Johnson1], [@pone.0016501-Kadyk1], [@pone.0016501-Nagaraju1]. Third, cell cycle-dependent variations in the activity of cyclin-dependent kinases affect the likelihood of engaging the DSB processing enzymes required for HR [@pone.0016501-Ira1]. In budding yeast, where HR is the major DSB repair pathway, repair of ionizing radiation (IR)-induced DSBs engages SCR in preference to interhomolog recombination [@pone.0016501-Kadyk1]. Similarly, in mammalian cells, intrachromosomal HR (of which SCR is likely a major component) is more efficient than interchromosomal HR [@pone.0016501-Johnson2]. In contrast to the constraints on HR, NHEJ can, in principle, operate at any stage of the cell cycle. Studies in which IR was used to induce chromosome breakage have supported the idea that HR and NHEJ contribute differently to DSB repair at different stages of the cell cycle [@pone.0016501-Saintigny1], [@pone.0016501-Rothkamm1], [@pone.0016501-Delacote1]. In the chicken lymphoblastoid cell line, DT40, cells defective for the HR gene *RAD54* reveal increased IR-sensitivity specifically in S phase [@pone.0016501-Takata1]. In contrast, DT40 cells lacking the NHEJ gene *KU70* are hypersensitive to IR during G1 and early S phase, consistent with a major role for NHEJ in repair of breaks generated prior to replication. Similar cell cycle dependencies of IR sensitivity and IR-induced HR were observed in mammalian cells, further supporting the notion that IR-induced HR is limited to the S/G2 cell cycle phases [@pone.0016501-Rothkamm1]. Indeed, in cells lacking the NHEJ factor XRCC4, IR-induced DSBs generated in G1 triggered elevated levels of HR in the subsequent S/G2. This suggests that the mechanism engaged to repair a DSB is a function of the cell cycle stage at which repair is executed, rather than the stage at which the DSB itself is induced [@pone.0016501-Saintigny1], [@pone.0016501-Delacote1]. The rare-cutting restriction endonuclease, I-SceI, has become a major tool for analyzing mechanisms of DSB repair [@pone.0016501-Rouet1]. I-SceI-induced breaks are likely limited to the canonical I-SceI target site and the DSB ends generated by I-SceI-mediated breakage are defined by the enzyme\'s known endonuclease activity. In contrast, IR-induced DSBs are distributed across the genome and the DSB ends may be chemically modified [@pone.0016501-Friedberg1]. For these reasons, the "rules" governing I-SceI-induced DSB repair might differ from those governing an IR-induced DSB. The ligand-binding domain (LBD) of the estrogen receptor (ER) has been used to regulate the activity of a variety of nuclear proteins, since fusion proteins containing the ER LBD are retained in the cytoplasm until activated by 17ß-estradiol [@pone.0016501-Feil1], [@pone.0016501-Zhang1], [@pone.0016501-Mattioni1]. Ligand binding facilitates correct folding of the released fusion protein, thereby activating the LBD-fused nuclear protein. The mouse ER LBD mutant G525R exhibits a ∼1000-fold reduction in 17ß-estradiol binding affinity compared to the wildtype protein, but retains normal affinity for the synthetic ER ligands and functional estrogen antagonists, tamoxifen (TAM) and 4-hydroxy-tamoxifen (4OHT) [@pone.0016501-Danielian1]. The corresponding human ER LBD mutant (G521R, termed "ER^T^") has similar properties [@pone.0016501-Feil1]. The 4OHT affinity of the human ER LBD mutant ER^T2^ (G400V/M543A/L544A) is approximately 4-fold higher than that of ER^T^ [@pone.0016501-Feil2]. Ectopic DNA enzymes such as Cre, Flp and I-SceI have successfully been controlled by fusion with ER^T^ [@pone.0016501-Feil2], [@pone.0016501-Hunter1], [@pone.0016501-Bennardo1], or with other nuclear hormone receptors [@pone.0016501-Soutoglou1]. In work described here, we characterize a set of I-SceI-ER^T^ or ER^T2^ fusion proteins and define optimal configurations for tight control of I-SceI activity, as measured by a sensitive I-SceI-inducible HR reporter. By enriching cells in specific phases of the cell cycle prior to activation of I-SceI, we have used this tool to study the impact of cell cycle stage on I-SceI-induced HR. Materials and Methods {#s2} ===================== Plasmids and RNAi {#s2a} ----------------- Expression plasmids for ER^T^ and ER^T2^ were described previously [@pone.0016501-Feil2]. New constructs described here were generated by standard cloning procedures. Control RNAi duplex against luciferase (5′-CGUACGCGGAAUACUUCGAdTdT-3′) and RNAi duplex against *BRCA1* (5′-AAUCACAGUGUCCUUUAUGUAdTdT-3′) were purchased from Dharmacon. Cell Lines and Cell Culture {#s2b} --------------------------- U2OS SCR \#18 cells were described previously [@pone.0016501-Puget1]. ER^T^ stable lines were maintained in phenol red-free DMEM supplemented with 10% charcoal-stripped fetal bovine serum (CS-FBS, Atlanta Biologicals), 0.2 ng/mL EGF (Invitrogen) and 100 U penicillin/100 mg streptomycin at 37°C and 6% CO~2~. To generate ER^T^ stable lines, the various ER^T^ expression constructs were transfected into U2OS SCR \#18 cells and 0.4 mg/mL G418 (Sigma-Aldrich) was added to the medium 2 days after transfection. After 2 weeks under continuous selection, G418-resistant colonies were isolated and screened by flow cytometry for 4OHT-induced HR. Antibodies and Immunoblotting {#s2c} ----------------------------- Cells were lysed in NP-40 lysis buffer (50 mM Tris-HCl \[pH 8.0\], 1% NP-40, 150 mM NaCl) supplemented with protease inhibitor cocktail (Roche). Cell lysates were resolved by SDS-PAGE on NuPAGE® Novex Bis-Tris Gels (Invitrogen) and analyzed by blotting with mouse monoclonal anti-Myc (9E10) antibody. Immunofluorescence Staining {#s2d} --------------------------- Cells on glass coverslips were treated with 1 µM 4OHT for the desired length of time (0--180 min), fixed in 3% paraformaldehyde/2% sucrose, permeabilized in Triton X-100 solution (0.5% Triton X-100, 20 mM HEPES \[pH 7.4\], 50 mM NaCl, 3 mM MgCl~2~, 300 mM sucrose), stained with anti-Myc (9E10) followed by goat anti-mouse IgG rhodamine-conjugated secondary antibody and imaged on a Zeiss microscope. Recombination Assays {#s2e} -------------------- Cells were plated at a density of 0.4×10^6^ cells/well on 6-well plates overnight prior to assay. Cells were treated with 1 µM 4OHT (Sigma-Aldrich) and GFP^+^ frequencies were measured 24--72 hr post-treatment by flow cytometry using an FC500 (Beckman Coulter). Cell Cycle Arrest {#s2f} ----------------- In order to arrest/synchronize cells at different stages of the cell cycle, several drug treatments were used. For G2/M synchronization, cells were treated for 16 hrs in 0.4 ng/mL nocodazole (Sigma-Aldrich), harvested by mitotic shake-off, and plated at a concentration of 0.4×10^6^ cells/well on 6-well plates with or without 1 µM 4OHT for 48 hrs. For G1 arrest by double drug treatment, cells were treated for 16 hrs in 0.4 ng/mL nocodazole (Sigma-Aldrich), harvested by mitotic shake-off, and plated at a concentration of 0.4×10^6^ cells/well on 6-well plates in 40 mM lovastatin (Axxora) for 48 hrs. 4OHT was added at a final concentration of 1 µM to induce I-SceI at the same time as cells were plated into lovastatin. For G1 arrest by single drug treatment, 40 mM lovastatin was added to cells with or without 1 µM 4OHT for 48 hrs. Cell Cycle Profiling {#s2g} -------------------- Cells were pulsed with 10 µM bromodeoxyuridine (BrdU, Sigma-Aldrich) for 20 min, collected, and fixed with ice cold 70% ethanol. For staining cells were treated with 2 M HCl/0.5% Triton X-100, neutralized with 0.1 M Na~2~B~4~O~7~ (pH 8.5), stained with mouse anti-BrdU primary antibody (Boehringer-Mannheim) and then goat anti-mouse IgG FITC-conjugated antibody (Jackson ImmunoResearch). Finally, cells were resuspended in 38 mM sodium citrate/69 µM propidium iodide/100 µg/mL RNaseA and propidium iodide/FITC staining levels were measured by flow cytometry using an LSR II (BD Biosciences). Southern Blotting {#s2h} ----------------- Genomic DNA was extracted from 5--10×10^6^ cells using the ArchivePure Cell/Tissue Kit (5 PRIME). Southern blotting for wt*GFP* was performed as described previously [@pone.0016501-Puget1]. Transfection {#s2i} ------------ 0.5×10^6^ U2OS SCR NEIE were plated overnight then transfected with 120 pmol siRNA using Lipofectamine™ 2000 (Invitrogen). GFP^+^ frequencies were measured 72 hr post-transfection by flow cytometry using an LSR II (Beckman Coulter). Results {#s3} ======= We used a previously described HR reporter containing two tandem mutant copies of the gene encoding the enhanced green fluorescent protein (*EGFP*, here termed *GFP*) ([**Figure 1A**](#pone-0016501-g001){ref-type="fig"}) [@pone.0016501-Puget1], [@pone.0016501-Xie1]. The first *GFP* copy (Tr-*GFP*) is truncated at the 5′ end and the second *GFP* copy (I-SceI-*GFP*) is full length, but is interrupted by the 18 base pair (bp) recognition site for the rare-cutting endonuclease, I-SceI, rendering this copy nonfunctional. Expression of I-SceI induces a site-specific DSB within the reporter and stimulates HR. Recombination between the two *GFP* copies by either intra- or inter-chromatid HR ([**Figure 1A**](#pone-0016501-g001){ref-type="fig"}), but not single strand annealing, can generate wild type *GFP* (wt*GFP*), which is readily quantified by flow cytometry. ::: {#pone-0016501-g001 .fig} 10.1371/journal.pone.0016501.g001 Figure 1 ::: {.caption} ###### Optimization of I-SceI-ER^T^ fusion proteins for controlled induction of chromosome breakage. A\) Repair of an I-SceI-induced DSB by intra- or interchromatid gene conversion. Double vertical lines: restriction site cut by I-SceI; Solid arrow: promoter sequence. B) I-SceI expression constructs used in this study. Each construct has an N-terminal triple-Myc tag and nuclear localization signal. ER^T^ represents the G521R mutant and ER^T2^ represents the mutant containing the G400V/M543A/L544A triple mutation of the human estrogen receptor ligand-binding domain. NI -- NLS-I-SceI; NIE -- NLS-I-SceI-ER^T^; NIE2 -- NLS-I-SceI-ER^T2^; NEIE -- NLS-ER^T^-I-SceI-ER^T^. C) Stable expression of I-SceI fusion proteins. Asterisk indicates endogenous Myc protein and can be used as a loading control. Arrowheads indicate predicted sizes of I-SceI fusion proteins. D) 4OHT-induced HR in NIE, NIE2, and NEIE cell lines. Error bars represent the standard error of the mean (S.E.M.) for triplicate samples. :::  ::: We fused cDNA sequences encoding ER mutants, ER^T^ or ER^T2^, in frame with the 3′ end of I-SceI in the I-SceI expression construct pcDNA3ß-myc-NLS-I-SceI, to generate "NIE" and "NIE2" constructs respectively ([**Figure 1B**](#pone-0016501-g001){ref-type="fig"}). We generated U2OS HR reporter cell lines (U2OS \#18 [@pone.0016501-Puget1]) stably transfected with either of these constructs and screened individual clones for the induction of HR upon addition of 4OHT (1 µM) to the medium. Steady state levels of the fusion proteins are shown in [**Figure 1C**](#pone-0016501-g001){ref-type="fig"}. 4OHT-responsive clones were identified for each construct; however, I-SceI-ER^T^ generated a higher proportion of responsive clones than I-SceI-ER^T2^. In clones that were found to be functionally responsive, as measured by induction of GFP^+^ events by 4OHT, the addition of 4OHT also produced the expected nuclear accumulation of myc-tagged I-SceI-ER^T^ or -ER^T2^ fusion protein. However, despite cultivation of the cells in phenol red-free medium and charcoal-stripped serum, every responsive clone revealed variable, weakly positive nuclear immunostaining for myc during passage in the absence of 4OHT. Further, all NIE or NIE^2^ 4OHT-responsive clones revealed an accumulation of GFP^+^ events in the absence of 4OHT at rates well above that observed in the parental clones lacking I-SceI-ER^T^ or -ER^T2^ expression ([**Figure 1D**](#pone-0016501-g001){ref-type="fig"} and data not shown). This indicates that the NIE and NIE^2^ fusion proteins were not perfectly controlled and that I-SceI-mediated breakage occurred even in the absence of the inducing 4OHT ligand. It was reported previously that the addition of more than one ER^T^ domain can improve the background of Cre fusion proteins [@pone.0016501-Zhang1]. We therefore generated an I-SceI-ER^T^ fusion protein in which a second ER^T^ domain was fused, in frame, between the NLS and I-SceI sequences of NIE, to generate a construct encoding a myc-NLS-ER^T^-I-SceI-ER^T^ fusion protein (here termed "NEIE" -- [**Figure 1B**](#pone-0016501-g001){ref-type="fig"}). Following stable expression of NEIE in U2OS HR reporter cell line \#18 ([**Figure 1C**](#pone-0016501-g001){ref-type="fig"}), we identified 4OHT-responsive clones and observed that, when cultivated in phenol red-free medium and charcoal-stripped serum, 4OHT-responsive NEIE clones appeared to exhibit improved stability in the absence of 4OHT than either NIE or NIE2 clones, while retaining equivalent responsiveness in the presence of the ligand ([**Figure 1D**](#pone-0016501-g001){ref-type="fig"}). The majority of 4OHT-responsive NEIE clones revealed a persistently low background frequency of GFP^+^ events, while none of the 4OHT-responsive NIE or NIE2 clones showed this pattern. Representative clones of each type are shown in [**Figure 1**](#pone-0016501-g001){ref-type="fig"}. While this work was in progress, Bennardo *et al.* reported similar observations using an independently generated system [@pone.0016501-Bennardo1]. To further characterize the performance of the NEIE fusion protein, we used immunostaining for myc followed by immunofluorescence microscopy to study the subcellular distribution of the NEIE fusion protein either in the absence of 4OHT or at defined times following addition of 4OHT to the medium ([**Figure 2A**](#pone-0016501-g002){ref-type="fig"}). Notably, in the absence of 4OHT, the NEIE fusion protein was observed in the cytoplasm and was fully excluded from the nucleus. Within 10 to 20 minutes of 4OHT addition, the NEIE protein was found to accumulate in the nucleus, and by 1 hour nearly every cell contained fully nuclear-localized NEIE protein. ::: {#pone-0016501-g002 .fig} 10.1371/journal.pone.0016501.g002 Figure 2 ::: {.caption} ###### Time course and dose dependence of HR induced by 4-hydroxytamoxifen in NEIE cells. A\) Immunofluorescence detection of 4OHT-induced NEIE nuclear accumulation over time. NEIE-6.46 cells were treated with 4OHT for the indicated times (in minutes) before fixation and staining. B) Southern blot detection of a 4OHT-induced site-specific DSB within the HR reporter. Arrowhead indicates the migration of PstI-restricted gDNA that has been cut by I-SceI. The uncut band is not shown in this figure. C) The effect of 4OHT concentration on NEIE-induced HR. The average raw percentage of GFP^+^ cells within the culture is shown. D) The effect of cell density on NEIE-induced HR. Shown is the average percentage of GFP^+^ cells, corrected for the background level of GFP^+^ cells. :::  ::: To test whether the freshly nuclear relocalized NEIE protein was active, we sought evidence of chromosome breakage within the HR reporter during the 4OHT response. We harvested genomic DNA (gDNA) from clone NEIE-6.46 either before the addition of 4OHT, or at defined time points following 4OHT addition, and analyzed the structure of the HR reporter by Southern blotting, as described in [Materials and Methods](#s2){ref-type="sec"}. Briefly, PstI-digested gDNA fragments that are probed with *GFP* reveal a single fragment of ∼5.9 kb if the reporter locus is intact. If the I-SceI site has been cut, a lower molecular weight species of ∼4.6 kb is observed ([**Figure 2B**](#pone-0016501-g002){ref-type="fig"}). Only a very small fraction (\<1%) of cells at any given time point following 4OHT addition revealed the I-SceI-cut ∼4.6 kb band ([**Figure 2B**](#pone-0016501-g002){ref-type="fig"}). The I-SceI-cut fragment was noted within 30 minutes following 4OHT addition and its intensity was approximately constant over time ([**Figure 2B**](#pone-0016501-g002){ref-type="fig"}). This suggests that the NEIE fusion protein is functionally active at the time of nuclear entry. Further, the low but consistent level of I-SceI-induced breakage observed at all time points following 4OHT addition suggests that there is a dynamic equilibrium between I-SceI-mediated chromosome breakage and DSB repair, with the balance tipped heavily in favor of repair in the cells examined here. This might reflect efficient religation of the I-SceI-induced DSB by precise NHEJ, which accurately reconstitutes the I-SceI site, provided canonical NHEJ is intact [@pone.0016501-Xie2], [@pone.0016501-Rass1]. However, in other experiments in which we examined breakage at the I-SceI site using Southern blotting in *Ku70* ^−/−^ or *XRCC4* ^−/−^ mouse embryonic stem cells, we failed to observe dramatic increases in steady state levels of I-SceI breakage, despite the loss of the canonical NHEJ pathway (data not shown). This might reflect cell type-specific effects or the engagement of alternative error-free DSB repair pathways. Further, the activity of the I-SceI endonuclease might be reduced in the context of the fusion protein. Importantly, removal of 4OHT from the culture medium of induced NEIE cells did not disrupt the nuclear distribution of the activated NEIE protein, and the strong nuclear myc signal persisted for up to 12 hours (data not shown). Therefore, although 4OHT can be used to activate I-SceI with precise timing, withdrawal of 4OHT appears not to immediately inactivate the enzyme. The above experiments were performed with 1 µM 4OHT calculated to saturate LBD receptor occupancy. To confirm that this dose was appropriate, we generated a dose-response curve for two of the NEIE stable cell lines, NEIE-6.39 and NEIE-6.46 ([**Figure 2C**](#pone-0016501-g002){ref-type="fig"}). I-SceI-induced HR in each test sample was assayed as the percentage of 4OHT-induced GFP^+^ cells. The resultant dose-response curves revealed median effective doses (ED~50~) of approximately 40 nM for each cell line - values similar to that reported previously for a Cre-ER^T^ fusion protein [@pone.0016501-Feil2]. 4OHT-induced HR in each cell line showed a modest dependence on the plating density of cells ([**Figure 2D**](#pone-0016501-g002){ref-type="fig"}). To determine the temporal relationship between DSB induction and HR, we examined the emergence of GFP^+^ cells as a function of duration of treatment of cell line NEIE-6.46 in 4OHT (1 µM). GFP^+^ cells first became detectable by flow cytometry within 10 to 12 hours of 4OHT addition ([**Figure 3A**](#pone-0016501-g003){ref-type="fig"}). Thereafter, GFP^+^ cells accumulated in 4OHT-treated populations at a fixed rate ([**Figure 3B**](#pone-0016501-g003){ref-type="fig"}). This rate of increase of the GFP^+^ fraction was remarkably constant at ∼0.05% per hour (∼1.2% per day), even when cells were passaged in 4OHT for a period of up to one month ([**Figure 3C**](#pone-0016501-g003){ref-type="fig"}). In contrast, cells passaged in the absence of 4OHT showed no increase in the GFP^+^ fraction over this one-month period, indicating that the NEIE cell line is tightly regulated when cultivated in medium lacking estrogenic compounds. ::: {#pone-0016501-g003 .fig} 10.1371/journal.pone.0016501.g003 Figure 3 ::: {.caption} ###### I-SceI activation induces a fixed rate of HR and GFP^+^ cell accumulation. A\) Initial emergence of GFP^+^ cells in NEIE-6.46 cells occurs after 10 hrs in 4OHT. B) The frequency of GFP^+^ cells increases in a linear fashion over the normal time course of experiments in this study (72 hrs). C) Linear increase in the frequency of GFP^+^ cells continues in cells exposed to 4OHT for up to 28 days. Open circle: no 4OHT. Black triangles: with addition of 4OHT. :::  ::: We asked whether I-SceI-induced HR is influenced by the cell cycle stage during which I-SceI is activated. To study this, we used combinations of lovastatin and nocodazole to synchronize cells at various points in the cell cycle. Lovastatin inhibits HMG-CoA reductase, causing cells to arrest in G1 [@pone.0016501-Rao1], while nocodazole is an antimitotic agent that disrupts microtubules and arrests the cell cycle at the G2/M phase [@pone.0016501-Jackman1]. Importantly, these drugs are not predicted to have a direct impact on DSB repair. Cell synchronization methods are depicted [**Figure 4A**](#pone-0016501-g004){ref-type="fig"}. In all treatment groups, cells were plated at the start of the experiment, were exposed to 4OHT 48 hours later, and were harvested for analysis of HR (measured by induction of GFP^+^ cells) and cell cycle distribution at the 96 hour time-point (i.e., 48 hours after initial exposure to 4OHT). Note that cells were exposed to 4OHT continuously throughout the second 48 hour period. Upon this framework we imposed one or more of the following two treatments: first, Nocodazole treatment for 16 hours, followed by release from nocodazole by mitotic shake-off, washing and replating, with the replating timed to coincide with the addition of 4OHT (the 48 hour time point). Second, lovastatin treatment, added to the medium at the same time as 4OHT (i.e., the 48 hour time point) and left in the medium until the time of harvesting (96 hour time point). Four treatment groups were analyzed: one received control diluents ("vehicle") but no cell cycle synchronizing drug; a second received nocodazole synchronization only; a third received lovastatin only; and a fourth received nocodazole+lovastatin ([**Figure 4A**](#pone-0016501-g004){ref-type="fig"}). As expected, lovastatin or nocozazole alone each diminished the percentage of cells in S phase at the end point of the experiment, and this effect was enhanced if the two drug treatments were combined (**Supplemental [Table S1](#pone.0016501.s001){ref-type="supplementary-material"}**). ::: {#pone-0016501-g004 .fig} 10.1371/journal.pone.0016501.g004 Figure 4 ::: {.caption} ###### BRCA1 depletion disrupts HR in S phase. A\) Cell synchronization methods. B) U2OS HR NEIE-6.39 and 6.46 cell lines were used to determine the cell cycle profile and frequency of HR after 48 hrs of 4OHT exposure in cultures synchronized at different stages of the cell cycle. Black circles represent data from several experiments involving different drug treatment protocols, as described in [Materials and Methods](#s2){ref-type="sec"}. These treatments include vehicle only, Lovastatin (Lov) alone, and cells treated with Nocodazole (Noc) and released into either vehicle or Lovastatin. A strong positive correlation is noted between the percentage of cells in S phase and the efficiency of HR, as indicated by the line of best fit (NEIE-6.39 *y* = 0.0244*x*+0.5395, R^2^ = 0.89503; NEIE-6.46 *y* = 0.0456*x*+0.3533, R^2^ = 0.85193). A negative correlation between the percentage of cells in G1 phase and the efficiency of HR is also noted (NEIE-6.39 *y* = −0.0357*x*+3.1208, R^2^ = 0.61909; NEIE-6.46 *y* = −0.0555*x*+4.588, R^2^ = 0.52032). The U2OS HR NEIE-6.46 cell line was transfected with control siRNA (siLuc -- siLuciferase) or BRCA1 siRNA (siBRCA1). 24 hrs post-transfection cells were replated in either 1 µM 4OHT or vehicle control, for a further 48 hours. Green squares indicate siLuc data points and red triangles indicate BRCA1 siRNA data points. C) Immunoblotting for BRCA1 in different siRNA treated groups. :::  ::: To determine whether cell cycle status at the time of harvest (i.e., the 96 hour time point) had any predictive value for the efficiency of HR measured at that same time point, we plotted the percentage of I-SceI-induced GFP^+^ cells against the percentage of cells in G1, S or G2 phases of the cell cycle, combining data from the four different treatment groups into one graph ([**Figure 4B**](#pone-0016501-g004){ref-type="fig"}). We generated regression lines for each of two independent clones (NEIE-6.39 and NEIE-6.46) and calculated the optimal fit of regression lines for each combination by the least squares method ([**Figure 4B**](#pone-0016501-g004){ref-type="fig"}). This analysis revealed a strong positive linear correlation (R^2^ = 0.90 and 0.85 for NEIE-6.39 and NEIE-6.46, respectively) between the percentage of cells in S phase and the level of 4OHT-induced HR ([**Figure 4B**](#pone-0016501-g004){ref-type="fig"}). We also observed a weak negative correlation (R^2^ = 0.62 and 0.52) between the percentage of cells in G1 and the efficiency of 4OHT-induced HR. These results show that I-SceI-induced HR in this system is strongly correlated with S phase, and that HR is inefficient in G1. This suggests that I-SceI induces a cell cycle-dependent pattern of HR equivalent to that deduced from previous analysis of IR-induced DSB repair [@pone.0016501-Saintigny1], [@pone.0016501-Rothkamm1], [@pone.0016501-Takata1]. However, some caution is needed in interpreting our results. First, the nature of the cell cycle synchronization is not absolute; it is to be expected that cell cycle distribution will vary over the 48 hours of 4OHT treatment, as cell cultures encounter the G1 block imposed by lovastatin or traverse G1 and enter S, in the case of the release from nocodazole. Second, we cannot exclude the possibilty that I-SceI-mediated DSB induction is itself cell cycle regulated. With these caveats in mind, the data does appear to support the notion that, operationally, I-SceI-induced HR is strongly correlated with S phase. BRCA1 is known to interact with Rad51 in discrete nuclear foci during the S and G2 phases of the cell cycle [@pone.0016501-Scully1] and to relocalize rapidly to sites of replication arrest in S phase cells treated with HU [@pone.0016501-Scully2]. In addition, *BRCA* mutant embryos contain chromosomes with "chromatid-type" structural aberrations, which are considered to arise from replication across a damaged template and from recombination errors in S phase [@pone.0016501-Patel1], [@pone.0016501-McPherson1]. These observations led to the proposal that BRCA1 regulates SCR during S phase [@pone.0016501-Scully3]. To test this hypothesis, we transfected U2OS HR NEIE cells with control siRNA or siRNA directed against human BRCA1, and plated cells directly into 4OHT 24 hours post-transfection, harvesting cells 48 hours later ([**Figure 4A and 4C**](#pone-0016501-g004){ref-type="fig"}). Depletion of BRCA1 reduced 4OHT-induced HR in U2OS HR NEIE cells, in comparison to control siRNA, but with minimal disruption of the cell cycle profile ([**Figure 4B**](#pone-0016501-g004){ref-type="fig"}). These results suggest that loss of BRCA1 disrupts the relationship between of S phase fraction and HR, and therefore support the long-standing hypothesis that BRCA1 has a major influence on HR during S phase. Discussion {#s4} ========== The development and characterization of a system to control I-SceI expression in a tight temporal fashion is an important research tool for the study of DSB repair. Its applications may include the study of cell cycle relationships, as discussed above, to potential use in chromatin immunoprecipitation studies [@pone.0016501-Savic1], real-time imaging, high-throughput screening and tight control of I-SceI activity in animal models. The accurate interpretation of siRNA experiments might also benefit from a tractable inducible DSB system such as that described here, where DSB induction could be initiated when siRNA knockdown is at its peak. The use of a recombination reporter is a stringent test of the activity of the enzyme -- probably more so than attempts to directly measure breakage at the locus. Notably, cultivation of cells in the presence of phenol red and endogenous estrogens in serum that was not charcoal stripped, triggered the gradual accumulation of GFP^+^ products of I-SceI-induced HR in the absence of 4OHT (data not shown). This suggests that, if residual estrogens are present in the culture medium, they can produce some basal activation of the fusion protein, even in its optimal "EIE" configuration. This suggests that the application of this technology in animal models may require additional levels of control. In this report, we have identified an optimal configuration of ER-LBD-I-SceI fusion protein that provides tight control of I-SceI activity in the absence of activating ligand. Our experience matches closely that described for ER LBD control of the Cre recombinase [@pone.0016501-Zhang1] and the results of Bennardo *et al.* in control of I-SceI [@pone.0016501-Bennardo1]. In each, the addition of ER LBD domains to both the N- and C-termini of the enzyme provide optimal control with retention of enzymatic activity. This result is also consistent with the crystal structure of I-SceI, in which the active site is predicted to be unaffected by modification of either the N- or C-termini of the polypeptide [@pone.0016501-Moure1]. We used this construct, stably expressed in U2OS HR reporter cells, to study the relationship between cell cycle stage and HR, by activating the I-SceI enzyme in cell cultures enriched at different cell cycle stages. By comparing the percentage of cells in a given cell cycle stage at the time of I-SceI activation against the HR outcome, we observed a strong positive correlation between the percentage of cells in S phase and the percentage of measured GFP^+^ HR events. As noted above, these experiments do not distinguish between S phase restriction of I-SceI-mediated breakage and S phase restriction of the HR pathway. However, given the known properties of I-SceI and precedent in the literature [@pone.0016501-Saintigny1], [@pone.0016501-Delacote1], it is perhaps more likely that it is the engagement of HR itself that exhibits the observed cell cycle restriction. It will be valuable to re-examine this question in other cell types that can be synchronized in other ways. Indeed, the availability of this tightly regulated form of I-SceI will provide new opportunities to study interactions between the cell cycle, cell differentiation and mammalian DSB repair. In view of the strong S phase linkage of I-SceI-induced HR in this system, treatments that strongly perturb HR without significantly affecting cell cycle distribution of the culture will likely affect HR during S phase. One such treatment, acute siRNA-mediated depletion of BRCA1, produced exactly these results, supporting the long-standing hypothesis that BRCA1 executes a major HR function during S phase [@pone.0016501-Nagaraju2]. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **Cell Cycle data and I-SceI-induced HR values for data shown in [Figure 4B](#pone-0016501-g004){ref-type="fig"}.** (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Scully lab members for stimulating discussions and a number of colleagues for helpful discussions during the development of this work -- especially Drs. Fred Alt, James Haber, Lewis Cantley and Stephen Elledge. We are also grateful to Dr. Pierre Chambon for providing mutant estrogen receptor cDNA plasmids. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**NIH R01CA095175; R01GM073894; R21CA144022. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: AJH RS. Performed the experiments: AJH SO IS JB RS. Analyzed the data: AJH SO RS. Contributed reagents/materials/analysis tools: AJH SO IS. Wrote the paper: AJH RS. [^2]: ¤a Current address: Constellation Pharmaceuticals, Cambridge, Massachusetts, United States of America [^3]: ¤b Current address: W. G. Hefner VA Medical Center, Salisbury, North Carolina, United States of America [^4]: ¤c Current address: Harvard Medical School, Boston, Massachusetts, United States of America

PubMed Central

2024-06-05T04:04:19.113459

2011-3-9

PMC3052303

Introduction {#s1} ============ Hazardous alcohol consumption is a significant public health problem, with an estimated 3.8% of all global deaths and 4.6% of global disability-adjusted life years lost attributable to alcohol [@pone.0014740-Rehm1]. The European Union (EU) is the heaviest drinking region of the world, drinking an average of 11 litres of pure alcohol per adult each year [@pone.0014740-Anderson1]. In the UK, deaths from cirrhosis are rising, and in some age groups the increase has been nearly 10 fold over one generation of 30 years [@pone.0014740-Chief1]. Despite the strong evidence supporting use of brief and less intensive interventions in people with alcohol use disorders, only a small minority actually receive help. Data from the Alcohol Needs Assessment Research Project indicates that in the UK fewer than 1 in 18 people with an alcohol misuse disorder access appropriate treatment, due to a combination of missed screening opportunities, limited availability of appropriate alcohol services, stigma associated with access and the wish to resolve problems alone [@pone.0014740-Drummond1]. Psychologically enhanced interventions delivered via the Internet could address all of these factors at minimal incremental cost, with potentially major public health implications. Population access to the Internet is increasing rapidly, and in 2009 penetration was estimated to be 77% in the UK, 64% in the EU as a whole, and 74% in the US [@pone.0014740-Internet1]. Psychologically enhanced web-based interventions make use of digital technologies to deliver a range of tailored behavioural techniques via the Internet, and have been shown to be associated with improved knowledge, self-efficacy, perceived social support, health behaviours and clinical outcomes [@pone.0014740-Murray1]. There is growing evidence about the use of the Internet to deliver smoking cessation interventions, where automated, self-help interventions tested in on-line randomized controlled trials have recruited large numbers of participants and yielded differences in abstinence rates ranging from 8% to 20% [@pone.0014740-Shahab1]. Despite evidence that large numbers of people with risky drinking behaviours access Internet based interactive interventions [@pone.0014740-Linke1], research in this area has been limited, with most studies employing brief normative feedback to college student samples recruited off-line [@pone.0014740-Bewick1], [@pone.0014740-Khadjesari1]. Additionally small trials of on-line interventions in adult populations have recruited through advertisements in newspapers, health related web-sites and telephone population surveys [@pone.0014740-Riper1]--[@pone.0014740-Cunningham2]. On-line trials can have major advantages over traditional face-to-face studies. Once the development costs have been met, they have minimal incremental running costs thus offering the ability to recruit very large numbers of participants. Different components of Internet technology allow rapid assessment, recruitment and randomisation, instantaneous collection of standardised and secure data, and delivery of on-line interventions in a controlled and uniform manner. Adoption of Internet based trial methods is increasing despite associated problems of high rates of attrition [@pone.0014740-Paul1], [@pone.0014740-Eysenbach1]. Studies have indicated that on-line trials are most suitable when the intervention is safe, the medical disorder can be confirmed by remote means and outcome measures assessed using electronically transmissible technologies [@pone.0014740-McAlindon1]. This paper reports the results of a large scale pragmatic on-line trial which satisfied all of these criteria. The aim of the trial was to compare the relative effectiveness and cost-effectiveness of an on-line, psychologically enhanced, interactive computer-based intervention (DownYourDrink, DYD) in reducing alcohol consumption with a flat, text-based information website in hazardous and harmful drinkers. The objectives were to: - Determine the effectiveness of DYD in enabling users to reduce their total alcohol consumption; - Determine the effectiveness of DYD in reducing alcohol related harm in users; - Determine the costs associated with the development and use of DYD; - Determine the cost-effectiveness of DYD as a public health intervention. Methods {#s2} ======= The protocol for this trial and supporting CONSORT checklist are available as supporting information; see [Checklist S1](#pone.0014740.s001){ref-type="supplementary-material"} and [Protocol S1](#pone.0014740.s002){ref-type="supplementary-material"}. Design {#s2a} ------ A two-arm individually randomised controlled trial for people with hazardous alcohol consumption was undertaken entirely on-line [@pone.0014740-Murray2]([Protocol S1](#pone.0014740.s002){ref-type="supplementary-material"}). It was conducted in three phases: pilot, main trial and main trial extension ([Figure 1](#pone-0014740-g001){ref-type="fig"}). There were only minor differences in design between each phase ([Box S1](#pone.0014740.s003){ref-type="supplementary-material"}), and as these were deemed unlikely to affect outcomes materially, analysis was undertaken on data pooled from all three phases. Ethics and data protection {#s2b} -------------------------- This study was conducted according to the principles expressed in the Declaration of Helsinki. Ethics approval for the study was granted by the University College London Research Ethics Committee, and all data were kept in accordance with provision of the UK Data Protection Act 1998. All patients provided written informed consent for the collection of data and subsequent analysis. Trial registration number {#s2c} ------------------------- SRCTN 31070347 Intervention and comparator websites {#s2d} ------------------------------------ For the duration of the trial, both the intervention and the comparator websites were located at a single website address: [www.downyourdrink.org.uk](http://www.downyourdrink.org.uk). The intervention website, hereafter known as DownYourDrink, or DYD, was a theoretically informed programme, based on brief intervention and psychological treatment principles. It offered three phases, each of which was divided into levels with different materials and associated exercises and tasks. If followed in order they provided a natural progression through three stages: decision making (Phase 1, *"It\'s up to you"*); implementing change (Phase 2, *"Making the change"*); and relapse prevention (Phase 3, *"Keeping on track"*). However, users were free to design their own route through the programme, and could use it as often or as seldom as they wished. Phase 1 was based on the principles of motivational enhancement therapy, phase 2 used computerised cognitive behavioural therapy and behavioural self control principles, and phase 3 was based on principles of relapse prevention. There were a number of interactive "e-tools" including a "thinking drinking diary" in which users could record their alcohol consumption along with emotional and behavioural triggers and responses. Further details about the development and content of the intervention are available elsewhere [@pone.0014740-Linke2]. The comparator website used a similar graphical design and style to present simple, text-based information about the harms caused by excess alcohol consumption. It did not contain any interactive components, and users did not have access to the e-tools. For the duration of the trial, this comparator website was also referred to as DownYourDrink so that participants were not aware whether they had access to the intervention or comparator site. Recruitment {#s2e} ----------- Participants were people who came across DownYourDrink while browsing the web. An earlier, simplified form of DYD had initially been launched in 2000 [@pone.0014740-Linke3] and by the start of the trial had accrued a large number of users [@pone.0014740-Linke1]. Most new users came to DYD from a web-search engine, such as Google or Yahoo, or from the home page of Alcohol Concern, the UK\'s largest alcohol charity. When users reached the home page they were invited to take a screening test (the three item Alcohol Use Disorders Identification Test or AUDIT-C [@pone.0014740-Bush1]). Users who scored 5 or more on the AUDIT-C were informed they were potentially at risk from their alcohol consumption, and invited to join the trial. They were informed that the trial was comparing different areas of the DownYourDrink website to see which was the most effective, and that for the duration of the trial, access to DYD was only available to trial participants. Eligible participants who consented to participate were asked to register, which included providing a user name, password and valid e-mail address. This e-mail address was used to send an automated link which gave participants access to the intervention or comparator site according to their randomised allocation. The AUDIT-C scores from users who did not consent to participation in the trial were discarded automatically for ethical reasons. Eligibility criteria were deliberately kept broad. Eligible participants were adults (aged 18 or over), scoring 5 or more on the AUDIT-C, who provided informed consent. Participants were required to have internet access. Participants who declared themselves unable to understand written English, or unwilling to complete follow-up questionnaires were excluded. People who were excluded from the trial, or who chose not to participate, were directed toward other on-line alcohol websites. Randomisation {#s2f} ------------- Randomisation occurred in two stages. The first randomisation occurred after completion of consent and core baseline data. At this point, participants were stratified by age and gender and randomised to one of four secondary outcome measures (see below). Once all baseline measures were completed, participants were randomised to either the intervention or the comparator website. This second randomisation marked the trial entry point. Both randomisation procedures were automated, using centrally-allocated computer-generated random numbers. Thus there was no possibility of any of the trial team influencing the allocation of participants and concealment of allocation was complete. Outcome measures {#s2g} ---------------- Reactivity to assessment, or the effect of measurement itself on alcohol consumption is a well-documented phenomenon in alcohol research [@pone.0014740-Kypri1], [@pone.0014740-McCambridge1]. For this reason, the total burden of assessment was kept to the minimum. All participants completed the primary outcome measure which was the TOT-AL [@pone.0014740-Khadjesari2]. The TOT-AL is a validated on-line measure which provided a drop-down menu for the selection of type, brand and size of beverage, and calculated the cumulative unit content of the drinks consumed over the previous 7 days (1 unit is equivalent to approximately 8 g ethanol). All participants also completed the 5 item quality of life measure, the EQ-5D [@pone.0014740-Rabin1] for the purposes of health economic analysis. We designed two single item measures to determine self-efficacy (confidence in one\'s ability to change behaviour) and intention, both important predictors of behaviour and intermediate variables along the pathway of change [@pone.0014740-Armitage1]. In addition, participants were asked to provide some basic demographic data at baseline (age, highest level of education attained, marital status, children, ethnicity and country of residence). Participants were randomly allocated to one of four secondary outcome measures, each of which addressed different domains of alcohol-related harm: the Alcohol Use Disorders Test (AUDIT) [@pone.0014740-Saunders1], the Leeds Dependence Questionnaire (LDQ) [@pone.0014740-Raistrick1], the Alcohol Problem Questionnaire (APQ) [@pone.0014740-Williams1], and the ten item Clinical Outcomes in Routine Evaluation (CORE) (a measure of mental health) [@pone.0014740-Evans1]. Data collection {#s2h} --------------- All data were collected on-line. At follow-up participants were sent an automated e-mail with an embedded hyperlink to the assessment questionnaires. Data collected at follow-up consisted of the primary outcome measure, the EQ-5D, single item measures of self-efficacy and intention, and the same secondary outcome measure completed at baseline. Up to three reminders were sent at 7 day intervals to non-responders, with the final reminder containing a request for participants to tell us their past week alcohol consumption only. The duration of follow-up varied in the three phases of the trial. During Phase 1 (pilot), follow up was at 1 and 3 months; in the main trial follow-up was at 3 and 12 months, and in the main trial extension, follow-up was at 3 months only ([Box S1](#pone.0014740.s003){ref-type="supplementary-material"}). The main reasons for extending the main trial were ethical concerns. The steady recruitment, combined with unsolicited free text emails from participants, suggested that DYD was meeting a need not met by alternative services. For this reason, we were reluctant to follow our original plan which had been to make DYD unavailable to new users once our target sample size had been achieved. Equally, we could not make the intervention freely available to new users for fear of contaminating the existing trial. Hence we decided to extend recruitment to the trial, but alter the consent and follow-up procedures so that follow-up was only requested at three months. After the end of Phase 3 (main trial extension), we made the control site freely available to new users for three months, and after all data collection had been completed, made the intervention site freely available to all users. Statistical methods {#s2i} ------------------- ### Sample size calculation {#s2i1} A 20% reduction in past week alcohol consumption, irrespective of initial level, is typical of non-internet brief interventions [@pone.0014740-Whitlock1]. In an earlier cohort study of DYD the observed mean reduction in alcohol consumption was 35% in men and 17% in women [@pone.0014740-Linke1]. In this study the standard deviation of weekly alcohol consumption was slightly less than the mean in both men and women at both baseline and follow-up. Making a conservative assumption that the standard deviation would be equal to mean, led to the calculation that 430 participants providing follow-up data at the principal end-point in each arm would be required to give 90% power at the 5% significance level to detect a 20% difference in the past week\'s reported alcohol consumption between intervention and control groups [@pone.0014740-Murray2]. ### Statistical analyses {#s2i2} Statistical analysis was carried out according to a pre-specified plan, comparing groups as randomised at each follow-up point. TOT-AL data were skewed and were therefore log-transformed (after adding 1 unit/week) before analysis. Means of the log-transformed data were transformed back to the original scale and are described as geometric means [@pone.0014740-Bland1]. For those unused to geometric means, the value of the geometric mean is very similar to the value of the median. To enable comparison of our data with other alcohol intervention trials we also report the arithmetic mean in the text, as this measure has often been used in reporting trial data despite the presence of skew in the data [@pone.0014740-Khadjesari1]. Adjusted analyses were performed using linear regression models of outcome on randomised group, adjusting for baseline values of the respective outcome measure, AUDIT-C, age, education, self-efficacy, log (TOT-AL+1), EQ5D and gender. Missing data were handled in three stages. First, primary analyses used all available results but without imputation of missing data. Second, alternative analyses used last observation carried forward (LOCF) and multiple imputation of missing outcomes from other outcomes and website use data. Third, sensitivity analyses for missing data assumed plausible arm-specific differences between responders and non-responders [@pone.0014740-White1]. Because the above analyses estimated only the effect of allocation to the intervention website, we additionally undertook a complier-average causal effect analysis to estimate the effect of compliance with the intervention [@pone.0014740-Dunn1]. This was initially performed defining compliance as more than 1 session or access to more than 10 pages within the first 3 months from randomisation, and subsequently assuming benefit to be proportional to number of page downloads and estimating the benefit of downloading 100 pages using instrumental variable methods [@pone.0014740-FischerLapp1]. Both these analyses used multiple imputation to handle missing outcome data. Health economics {#s2j} ---------------- Costs of the intervention included resources required in the original development of the DYD internet site and revisions undertaken for the trial by a development group comprising academics, clinicians and lay members and programmed by web consultants. Development of the control website was assumed to take a minimal proportion (5%) of overall costs. Care was taken to separate development of the intervention from research costs. Invoices for programming costs were separated into research, intervention and control costs, with 20% of the development group\'s time assumed to be concerned with research issues. All figures are at 2008 price levels. The primary outcome for economic evaluation was quality-adjusted life-years (QALYs) based on EQ-5D questionnaire responses valued by the UK Social Tariff valuations [@pone.0014740-Kind1]. Results {#s3} ======= Recruitment and follow-up {#s3a} ------------------------- The recruitment period was from February 2007 until May 2009 ([Figure 1](#pone-0014740-g001){ref-type="fig"}). Recruitment rates were maintained throughout, averaging around 65 participants per week ([Figure S1](#pone.0014740.s004){ref-type="supplementary-material"}). Of the 10,141 visitors consenting to take part in the trial, 7,935 (78%) completed baseline data collection and randomisation procedures to enter the trial. At 3 months, 1,592 (40%) of the intervention group completed the TOT-AL compared with 1,937 (49%) of controls (P\<0.001). Differential response rates were present across at all assessment points ([Figure 1](#pone-0014740-g001){ref-type="fig"}). ::: {#pone-0014740-g001 .fig} 10.1371/journal.pone.0014740.g001 Figure 1 ::: {.caption} ###### CONSORT diagram. :::  ::: Baseline assessment {#s3b} ------------------- Although the majority of participants were White British (84%) and resident in the UK (88%), there were some from ethnic minorities, and 73 countries were represented amongst respondents. Mean age was 38 years, 57% were women and 52% were educated to at least degree level. The participants were heavy drinkers (geometric mean past week\'s alcohol consumption at baseline 46.0 (SD 31.2) units), drinking most days, binge drinking, and regularly drinking above recommended limits ([Figure 2](#pone-0014740-g002){ref-type="fig"}, baseline), but reported little evidence of dependence. There were no differences between randomized groups for any baseline characteristic ([Table S1](#pone.0014740.s005){ref-type="supplementary-material"}). Arithmetic mean consumption at baseline was 49.1 units for women and 68.2 units for men. ::: {#pone-0014740-g002 .fig} 10.1371/journal.pone.0014740.g002 Figure 2 ::: {.caption} ###### Quantity and patterns of alcohol consumption and EQ5D scores by randomized group over time: means and 95% CIs. :::  ::: Website usage {#s3c} ------------- Participants in the intervention group made an average of 2.33 (SD 3.63) visits to the site and downloaded an average of 67 (SD 79) pages in the first month following recruitment. For the control group, the averages were 1.24 (SD 0.75) visits and 13 (SD 12) pages downloaded (p\<0.001 for both visit and page comparisons) ([Table S2](#pone.0014740.s006){ref-type="supplementary-material"}). Primary outcomes {#s3d} ---------------- At 3 months, there was a substantial reduction in mean reported alcohol consumption in the intervention group (46.3 to 26.4 units) and the controls (45.7 units to 25.6 units). The adjusted ratio of geometric means between the two groups at 3 months was 1.03 (CI 95% 0.97 to 1.10), providing no evidence of difference between groups. Similarly, no differences were shown at 1 month or 12 months, the confidence intervals effectively ruling out the possibility of a relative reduction in mean alcohol consumption of 15% or more ([Figure 2](#pone-0014740-g002){ref-type="fig"}, [Table 1](#pone-0014740-t001){ref-type="table"}). Similar reductions were seen in both groups at all assessment points in numbers of drinking days, days drinking above recommended limits and binge drinking occasions ([Figure 2](#pone-0014740-g002){ref-type="fig"}, [Table S3](#pone.0014740.s007){ref-type="supplementary-material"}). Arithmetic mean past week alcohol consumption for women at one, three and 12 month follow-up was 33.5, 33.1 and 27.9 units respectively, and for men intake was 48.6, 46.3 and 44.7 units at one, three and 12 months. Self-efficacy scores were higher for both groups at all follow-up assessments than at baseline. At 1 month, they were significantly higher in the intervention group than in controls, but this difference was small and not maintained at subsequent assessments. Intentions showed a slight decrease in both groups at all follow-up assessments. EQ5D scores showed little change in both groups at all assessment points ([Figure 2](#pone-0014740-g002){ref-type="fig"}, [Table S4](#pone.0014740.s008){ref-type="supplementary-material"}). ::: {#pone-0014740-t001 .table-wrap} 10.1371/journal.pone.0014740.t001 Table 1 ::: {.caption} ###### Reported alcohol consumption in last week (units)[\#](#nt101){ref-type="table-fn"} by randomised group. ::: {#pone-0014740-t001-1} Geometric mean (SD)[\*](#nt102){ref-type="table-fn"} Adjusted ratio (intervention: control) of geometric means (95%CI)[\$](#nt104){ref-type="table-fn"} ---------------------- ------------------------------------------------------ ---------------------------------------------------------------------------------------------------- --------------------- Baseline (n = 7,935) 46.3 (31.8) 45.7 (30.6) \- 1 month (n = 2,067) 27.1 (23.1) 27.1 (22.5) 0.98 (0.90 to 1.07) 3 months (n = 3,529) 26.4 (23.0) 25.6 (21.5) 1.03 (0.97 to 1.10) 12 months (n = 854) 22.0 (20.0) 23.5 (21.0) 0.99 (0.85 to 1.15) \# 1 unit = 8g of ethanol. \*Approximate SD back-calculated from the log scale. \*\*See [Figure 1](#pone-0014740-g001){ref-type="fig"} for the data contributing to each time point. \$ Adjusted for baseline alcohol consumption, AUDIT-C, age, sex, education, self efficacy and EQ5D. ::: Secondary outcome measures {#s3e} -------------------------- All measures showed improvements at all follow-up assessment points for participants in both the intervention and control groups but, with the exception of LDQ at 3 months, there were no significant differences between the groups for any measure ([Table S5](#pone.0014740.s009){ref-type="supplementary-material"}). Subgroup analyses {#s3f} ----------------- Analyses to determine impact of pre-specified baseline characteristics (sex, educational level, baseline consumption) on past week\'s alcohol consumption, showed no evidence of differential effects of the intervention (all interaction P values\>0.10, [Table S6](#pone.0014740.s010){ref-type="supplementary-material"}). Sensitivity analyses for missing data {#s3g} ------------------------------------- Results were little changed when missing data were handled using LOCF or multiple imputation ([Table S7](#pone.0014740.s011){ref-type="supplementary-material"}). Sensitivity analyses allowing for systematic differences between non-responders and responders indicated that equal differences in both arms of the trial would result in little change in results, but that asymmetrical differences could produce substantial changes ([Table S8](#pone.0014740.s012){ref-type="supplementary-material"}). Effect of website exposure {#s3h} -------------------------- In those complying with the intervention, the estimated average causal effect of allocation to intervention, expressed as a ratio of geometric means of past week\'s alcohol consumption, was 1.05 (95% CI 0.95 to 1.16) at 3 months. In those who downloaded 100 pages, the corresponding ratio was 1.06 (95% CI 0.94 to 1.19) ([Table S9](#pone.0014740.s013){ref-type="supplementary-material"}). Health economic analyses {#s3i} ------------------------ The total cost of development and delivery of the DYD intervention was £107,317 and the control site cost was £3,390. These costs are detailed in [Table S10](#pone.0014740.s014){ref-type="supplementary-material"}. With the exception of the web maintenance costs (a small proportion of the total), these costs do not differ according to numbers accessing the site; hence the incremental costs per participant are small. The average cost per participant in the trial is £27.02 for the intervention and 85p for the control, a difference of £26.17. No significant differences in EQ5D scores or variances were found and therefore no cost-effectiveness ratio was calculated. Discussion {#s4} ========== The psychologically enhanced, interactive computer-based intervention was not more effective in reducing alcohol consumption or related harms than a flat, text-based information website among hazardous and harmful drinkers. There were no differences in levels or patterns of alcohol consumption or secondary outcome measures between participants allocated to the intervention or control groups, at either the primary or secondary follow-up points. Participants in the intervention group made more use of the intervention than those in the control group, but we have no data on the relative satisfaction of the users in the two groups. Both groups showed evidence at all follow-up points of striking improvements from baseline values in levels and patterns of alcohol consumption and in all secondary outcome measures. There are various potential explanations for these findings. Although there is clearly no difference between the effectiveness of the two interventions, it is not clear whether both interventions were effective or both were ineffective. The improvements demonstrated by trial participants could be partly due to regression to the mean (where people are motivated to join a trial at the time that their problem is most severe and through the natural history of a waxing and waning condition show an improvement over time) or to the effects of the trial assessment procedures. The therapeutic effect of assessment on alcohol consumption in trials has been well documented [@pone.0014740-Kypri1] and even minimal assessment, such as completing the 10 item AUDIT has been shown to have an effect size of 0.23 (95% CI 0.01--0.45) at 2--3 months follow-up [@pone.0014740-McCambridge1]. Although we went to considerable lengths to reduce the burden of assessment it is still probable that completion of the primary outcome measure along with other aspects of study participation contributed to the observed reduction in alcohol consumption. The findings could also have been due in part to non-response bias, though this is not supported by the results of statistical analyses undertaken to deal with this anticipated aspect of the on-line trial performance. There was a marked differential in response rates between the intervention and control groups at 1 and 3 months, which had reduced but not vanished by 12 months. This differential response, with participants in the control group being more likely to respond than those in the intervention group has been seen in previous alcohol trials [@pone.0014740-Kaner1]. Our data cannot illuminate the reason for this differential, but it is possible that participants in the control group particularly welcomed the opportunity to undergo assessment, recognising this as an opportunity to reflect on their drinking behaviours. The annual maintenance costs of DYD intervention were estimated at £12,065. Even modest recruitment rates of 50 new entrants per week evidenced in the latter stages of the trial would yield a cost of only £4.64 per person. A mean improvement in health in terms of QALYs of only 0.01 over a 12 month period would make the intervention highly cost-effective (incremental cost-effectiveness ratio of £464 per QALY). As reduced drinking is also associated with a reduction in public sector spending and improved health, such interventions taken up by those not currently accessing services could well be cost neutral and potentially significantly cost saving. To our knowledge this is the largest pragmatic trial of an alcohol Internet intervention undertaken in the general population. It succeeded in attracting website visitors with hazardous alcohol consumption, recruiting numbers which substantially exceeded initial expectations. The study employed an innovative on-line methodology well suited to the nature of the Internet based intervention and control websites. This presented significant methodological challenges in relation both to the exclusive use of on-line assessment and to compliance with the intervention and follow-up. An extensive evidence base indicates that self reporting of alcohol consumption is at least as reliable as face to face, though uncertainty remains about the performance of these measures in on line trials [@pone.0014740-DelBoca1], [@pone.0014740-Whitford1]. Many on-line trials have experienced high rates of attrition from follow-up [@pone.0014740-Bull1] so we tested several methods to optimise response and employed a range of relevant statistical methods both to impute missing values and to estimate the effects of different levels of compliance with the intervention. Nonetheless, uncertainties remain, including the possibility of bias, as a result of the high rates of attrition from follow-up, and these need to be fully recognised in interpreting the findings. Our results differ from previous trials of online alcohol interventions and this may reflect differences in study populations, trial procedures and comparator interventions [@pone.0014740-Riper1]--[@pone.0014740-Cunningham2]. The trial population in the present highly naturalistic study were web-browsers, whereas other studies used at least some off-line recruitment procedures, either for obtaining consent [@pone.0014740-Riper1], or for initial identification of potential participants [@pone.0014740-Cunningham1], [@pone.0014740-Cunningham2]. This is likely to have implications for the study population. In this trial we used a non-interactive website which provided information about the harms of excessive alcohol consumption and advice on how to cut down. This contrasts with the Riper trial, where a pdf version of a psycho-educational brochure was used as a comparator [@pone.0014740-Riper1]. Our decision was made partly on ethical grounds so that all participants would receive something at least as good as widely available self-help sites, and partly on research grounds to ensure trial participants were not made aware of which arm they had been randomised to. The trial has indicated a potentially widespread and sustainable demand for Internet based interventions for people with hazardous alcohol consumption. Our findings do not provide any support for the hypothesis that psychologically enhanced interactivity confers additional benefit. However, the substantial improvement in quantity and patterns of alcohol consumption reported by participants in both arms of the trial suggests potential benefit from access to either website type, providing support for continued development and implementation of Internet applications of this kind. Supporting Information {#s5} ====================== Checklist S1 ::: {.caption} ###### CONSORT Checklist. (0.23 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Protocol S1 ::: {.caption} ###### Trial Protocol. (0.34 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Box S1 ::: {.caption} ###### Differences in design between phases of the trial. (0.04 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S1 ::: {.caption} ###### Cumulative recruitment to trial. (0.01 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Baseline data by randomised group. (0.04 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Use of intervention and comparator websites. (0.03 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### Patterns of reported alcohol consumption over time by randomised group. (0.04 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### Self-efficacy, intention and EQ5D scores over time. (0.04 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### Secondary outcome measures (assessed in 1:4 participants). (0.05 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S6 ::: {.caption} ###### Subgroup analyses adjusting for baseline values. (0.07 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S7 ::: {.caption} ###### Reported alcohol consumption in last week (units) by randomised group: alternative analyses allowing for missing outcome data. (0.04 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S8 ::: {.caption} ###### Effect of intervention on reported alcohol consumption in last week (units): sensitivity analyses allowing for missing data. (0.05 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S9 ::: {.caption} ###### Causal effects of using the intervention website on reported alcohol consumption in last week (units). (0.03 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S10 ::: {.caption} ###### Summary of costs incurred in developing the intervention and comparator (2008 costs). (0.03 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Alcohol Concern for their collaboration with the trial and gratefully acknowledge the support and guidance of the members of the DYD RCT Trial Steering Group: Colin Drummond, Jonathan Elford, Marlie Ferenczi, Andy Haines (Chair) and Fran Heron. We thank Harvey Linke of Net Impact and Richard McGregor of Codeface Ltd for their work in developing the intervention and comparator and trial websites, and Jo Burns and Orla O\'Donnell for project management. **Competing Interests:**Since June 2009, PGW has received payments from the charity Drinkaware in his capacity as its Chief Medical Advisor. PGW has no other competing interests and the authors confirm that this does not alter their adherence to all the PLoS ONE policies on sharing data and materials. **Funding:**This study was funded by the National Prevention Research Initiative, which includes the following funding partners: British Heart Foundation; Cancer Research UK; Department of Health; Diabetes UK; Economic and Social Research Council; Medical Research Council; Research and Development Office for the Northern Ireland Health and Social Services; Chief Scientist Office, Scottish Executive Health Department; and the Welsh Assembly Government. IRW and SGT are funded by the UK Medical Research Council (grant codes U.1052.00.006 and U.1052.00.001). The Alcohol Education and Research Council provided additional funding to assist with developing the intervention site. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: PGW EM JM ZK ST CG SL. Performed the experiments: ZK. Analyzed the data: IRW ST EK CG. Wrote the paper: PGW EM. Chief investigator and grant holder of the study: PGW. Wrote the paper originally submitted for review: PGW. Led on the design of the study and methodological issues: EM. Revised the paper in response to reviewer comments: EM. Contributed to the design of both the protocol and the intervention: JM. Contributed to the interpretation of the data and to the writing of the final report: JM ZK IW ST EK CG SL. Was the study research fellow and contributed to the development of the trial intervention and control websites: ZK. Was the health economist responsible for analysis and reporting of cost effectiveness: CG. Contributed to the study design and led the development of the intervention website: SL.

PubMed Central

2024-06-05T04:04:19.116045

2011-3-9

PMC3052304

Introduction {#s1} ============ High-throughput sequencing technologies have made a huge impact on microbiology, providing a rapid and cost-effective way of generating draft genomes and allowing metagenomic exploration of microbial diversity. Metagenomics, the survey of microbial or viral communities (and their encoded metabolic activities) from distinct environments, has been rapidly expanding over the past several years from its origins in environmental microbiology [@pone.0017288-Tringe1]--[@pone.0017288-Wooley1]. Recently, the National Institute of Health (NIH) roadmap Human Microbiome Project (HMP) initiative was jump-started to examine microbes associated with health and disease in several areas of the human body [@pone.0017288-Turnbaugh1], [@pone.0017288-Peterson1]. Metagenomics has been enabled by the advances in second-generation sequencing, with current sequencing machines generating reads that are shorter than those generated with gel-capillary technology. However, the amount of data produced is orders of magnitude greater than that generated by earlier techniques and can reach gigabases per machine day [@pone.0017288-Flicek1], [@pone.0017288-Metzker1]. The performance characteristics of high-throughput sequencing machines such as Roche/454\'s GS FLX, Illumina/Solexa\'s GA IIx, and Life Technologies SOLiD system are changing rapidly with respect to machine capacity, run time, read length, error profile, and cost per base. The immense amount of genomic and metagenomic data produced today requires an automated approach for data processing and analysis. A typical sequence processing pipeline includes several steps such as sequence cleaning, alignment to known reference sequences, and/or de novo assembly [@pone.0017288-Kunin1], [@pone.0017288-Metzker1]. The sequence cleaning step is an essential first step of the sequence processing pipeline before any further data processing in order to allow accurate downstream analysis. For most datasets, the sequence cleaning step usually includes filtering to remove read duplicates, low quality reads, contaminating sequences, and adaptor or barcode sequences. Sequences obtained from impure samples or nucleic acid preparations may contain DNA from sources other than the microbes in the sample. That sequence contamination is a serious concern: for the HMP all contaminating human genomic sequences must be removed from the sample prior to the data being made public; for other projects the quality of the data used for downstream analysis will be affected by contamination, possibly causing misassembly of sequence contigs and erroneous conclusions. In this paper we focus on identifying and removing human contamination from microbial metagenomes, such as those created under the auspices of the HMP. However, the methodology can be applied to any kind of sequence contamination. To detect human contamination, metagenomes need to be compared to the human genome. In addition to the public and private human genome sequencing efforts [@pone.0017288-Lander1], [@pone.0017288-Venter1], several individual human genomes were published in the last three years [@pone.0017288-Levy1]--[@pone.0017288-Li1]. Large-scale resequencing projects such as The 1000 Genomes Project (<http://www.1000genomes.org/>), the Cancer Genome Atlas [@pone.0017288-Collins1] and the Personal Genome Project (<http://www.personalgenomes.org/>) will also generate high-coverage human genomes. These projects provide the reference sequences that are used to detect human genome contamination in genomic and metagenomic datasets. Earlier-generation sequence alignment programs such as BLAST [@pone.0017288-Altschul1], [@pone.0017288-Altschul2] were designed to align DNA and protein sequences and to search through large databases to find homologous sequences. MegaBLAST was developed to speed up the alignment for query sequences that are highly similar to the reference sequences and was used to align large-scale sequencing data. Later, improvements on MegaBLAST were proposed such as database indexing methods to allow even faster alignments [@pone.0017288-Morgulis1]. The advances in sequencing technology over the last decade have brought new challenges in bioinformatics; consequently many new alignment programs that are much faster than BLAST have been published over the last few years. In general, the new alignment programs were developed to align DNA sequences to closely related reference genomes, especially long references such as mammalian genomes, with only few low quality alignments expected. For example, many short-read alignment programs were designed for reads 100 bp [@pone.0017288-Li2]--[@pone.0017288-Langmead1]. However, most next generation sequencing technologies already produce reads 100 bp (Illumina/Solexa), 400 bp (Roche/454), and 1,000 bp (Pacific Biosciences in early testing [@pone.0017288-Metzker1], [@pone.0017288-Eid1], [@pone.0017288-McCarthy1]). In a few years, long reads will likely dominate and programs for short reads will be less applicable. In contrast to short-read alignment algorithms that tend to maximize global alignments, longer-read alignment algorithms aim to find local matches because longer reads are more prone to structural variations and map over misassemblies in the reference sequence. Longer-read alignment programs must also be able to deal with alignment gaps since indels (insertions and deletions) occur more frequently in long reads and may be the dominant source of sequencing errors for some technologies such as Roche/454 and Pacific Biosciences [@pone.0017288-Li5]. The three approaches used by the currently available longer-read alignment programs are either hash table, suffix tree or Burrows-Wheeler Transform (BWT) [@pone.0017288-Flicek1], [@pone.0017288-Li6]. Hash table based algorithms basically follow the seed-and-extend paradigm and the idea of hash table indexing can be traced back to BLAST. The BLAST program keeps the position of each -mer subsequence of the query and scans the database sequences for -mer exact matches (called seeds) by looking up the hash table. BLAST then extends and joins the seeds and refines them by a Smith-Waterman alignment [@pone.0017288-Smith1]. The seeding step was accelerated by the idea of requiring multiple seed matches for an extension. This idea is implemented in SSAHA2 [@pone.0017288-Ning1] and BLAT [@pone.0017288-Kent1], which offer significantly faster alignment than BLAST for reads that are nearly identical to the reference database. Recently developed hash table based programs for longer reads such as Mosaik (<http://bioinformatics.bc.edu/marthlab/Mosaik>) build the hash table on the reference sequences and use it to scan for query subsequences. Hash tables are appropriate for DNA sequences, since they very likely contain repeats or duplicates and are unlikely to contain every possible combination of nucleotides. Depending on the size of the reference sequences, the size of the hash table may be very large (tens of GB) and take a lot of time or memory to build. Another group of algorithms rely on a representation of suffix/prefix trie. The advantage of using a trie is that an alignment to multiple identical copies of a substring in the reference is only done once since they collapse on a single path in the trie. It takes linear time to determine if a query has an exact match against a trie, but a trie takes quadratic space with respect to the reference length. A suffix tree achieves linear space while still allowing linear-time searching. The alignment program MUMmer [@pone.0017288-Kurtz1] is based on suffix tree and anchors the alignment with maximal unique matches (MUMs) and then joins these exact matches with gapped alignments. Most trie implementations require more than 10 bytes per nucleotide and make it impractical to hold the suffix tree of large reference genomes in memory. The FM-index proposed by Ferragina and Manzini [@pone.0017288-Ferragina1] was originally designed as a compressed data structure and is used by alignment programs to improve the memory usage (typically 0.5--2 bytes per nucleotide [@pone.0017288-Li6]). The FM-index data structure is basically a compressed suffix array, following the concept that a suffix array is much more efficient if it is created from the BWT sequence rather than from the original sequence. BWT implementations are widely used because of their small memory footprint and they are much faster than their hash-based alternatives at the same sensitivity level [@pone.0017288-Flicek1]. When comparing nucleotide sequences, even a unique query sequence can match a few million positions with a positive alignment score, with the majority being random matches or matches in short low-complexity regions. BWA-SW [@pone.0017288-Li5], an implementation of the Burrows-Wheeler aligner combined with a Smith-Waterman search uses heuristics to accelerate the alignment process. BWA-SW traverses the query prefix directed acyclic word graph (DAWG) in the outer loop and the reference prefix trie in the inner loop. From this, all the nodes in the reference prefix trie that match the query node with a positive score are found. Since the true alignment tends to have a high alignment score, it is possible to prune low-scoring matches at each node, and consequently restrict dynamic programming around good matches only. At each node in the DAWG, BWA-SW only keeps the top best-scoring nodes in the reference trie that match the node, rather than keeping all the matching nodes. This heuristic is referred to as -best strategy. Here, we selected longer-read alignment programs that are actively maintained and widely used and evaluated them on simulated datasets. These programs include BLAST, BLAST+ [@pone.0017288-Camacho1], Mosaik, NUCmer (from MUMmer package), and BWA-SW. Based on the evaluation results, we adopted BWA-SW for the removal of human sequence contamination from metagenomes and developed DeconSeq, a robust framework for the rapid, automated identification and removal of sequence contamination from longer-read datasets. DeconSeq is implemented in Perl and is freely available at <http://deconseq.sourceforge.net/>. Using DeconSeq, the amount of possible human DNA contamination in 202 metagenomes was investigated. Results {#s2} ======= Comparison of program performance {#s2a} --------------------------------- The programs Mosaik, NUCmer, BLAST, BLAST+ and BWA-SW were compared for their ability to perform the alignments of the simulated longer-read metagenomes against the human sequence database (see [Text S1](#pone.0017288.s001){ref-type="supplementary-material"} for details). Overall, BWA-SW performed with the lowest running time of approximately 22 minutes for the human simulated datasets and four minutes for the bacterial and viral simulated datasets (see [Text S1](#pone.0017288.s001){ref-type="supplementary-material"}, [Figure 1](#pone-0017288-g001){ref-type="fig"}). We did not include BLAT and SSAHA2 in our comparison as these programs were compared to BWA-SW previously and showed similar or worse sensitivity with much longer time spend for the computations [@pone.0017288-Li5]. ::: {#pone-0017288-g001 .fig} 10.1371/journal.pone.0017288.g001 Figure 1 ::: {.caption} ###### Alignment sensitivity of BWA-SW for human sequences. Query coverage and alignment identity values ranged from 90% to 100%. The sensitivity shows how many sequences could be aligned back to the reference. The simulated datasets contained 28,612,955 reads for 200 bp, 11,444,886 reads for 500 bp, and 5,722,210 reads for 1,000 bp. :::  ::: Based on the comparisons, we tentatively adopted BWA-SW to identify and remove human contamination from metagenomes. BWA-SW was the fastest algorithm to complete the identification of 100,000 human sequences using the available computational resources. However, speed and computational requirements are only one aspect of the identification of possible contaminating sequences. The program must also be able to accurately identify all of the contamination in the sample. We therefore assessed the sensitivity of the BWA-SW algorithm at identifying human DNA contamination. This test also identified which sequences the aligner commonly missed. Evaluation of alignment sensitivity {#s2b} ----------------------------------- There are known limitations to the alignment approach such as placing reads within repetitive regions in a reference genome. BWA-SW was evaluated for its ability to align simulated data containing sequences extracted from the human reference genome back to the reference genome. The simulated data contained 200 bp, 500 bp, or 1,000 bp long sequences. Errors were introduced at rates of 2% and 5%. The typical error rate for real data is approximately 0.5%, therefore this analysis provides a worst-case scenario [@pone.0017288-Huse1]. The human reference genome was used for this analysis, because it presents the only finished-grade human genome sequence available [@pone.0017288-Metzker1]. The human reference genome was constructed from multiple individuals, contains 2.86 Gbp, covers 99% of the human genome with 357 gaps and has an estimated error rate of 1 in every 100,000 bp (<http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/human/data/?build=37>). After computing the alignments, we filtered the results based on query coverage and alignment identity values ([Figure 1](#pone-0017288-g001){ref-type="fig"}). Using the default settings, longer sequences could be aligned to the correct region more often than shorter sequences independent of the error rates introduced. Without using alignment thresholds, more than 99.9% of all sequences could be aligned back to the reference. Of the simulated sequences that did not match the reference with the given thresholds, on average more than 56% of the sequences were derived from repeat regions of the human reference genome ([Figure 2](#pone-0017288-g002){ref-type="fig"}). Simple repeats and low complexity regions that represent the majority of the unaligned sequences with 0% error rate cover 0.84% and 0.55% of the human reference genome, respectively [@pone.0017288-Alexander1]. In contrast, the unaligned sequences were rarely from regions of the human genome that contained exons (on average less than 4%). ::: {#pone-0017288-g002 .fig} 10.1371/journal.pone.0017288.g002 Figure 2 ::: {.caption} ###### Repeats causing alignment problems for BWA-SW. The query coverage was set to 95%, with identity set to 99%, 97% and 94% for error rates of 0%, 2% and 5%, respectively. The numbers above the bars show the number of unaligned sequences of each category for the given thresholds. The values shown in parenthesis represent the percentage of unaligned sequences. The simulated datasets contained 28,612,955 reads for 200 bp, 11,444,886 reads for 500 bp, and 5,722,210 reads for 1,000 bp. :::  ::: The sequences that could not be aligned under the given thresholds were then aligned against the same human genome using higher -best values (ranging from two to ten) or additional human genome data. Increasing the -best value increased the number of unaligned sequences that could be aligned (see [Text S1](#pone.0017288.s001){ref-type="supplementary-material"}, [Figure 2](#pone-0017288-g002){ref-type="fig"}). However, using higher -best values almost linearly increases the runtime (see [Text S1](#pone.0017288.s001){ref-type="supplementary-material"}, [Figure 3](#pone-0017288-g003){ref-type="fig"}). Using additional human genome data as reference increased the number of unaligned sequences that could be aligned (see [Text S1](#pone.0017288.s001){ref-type="supplementary-material"}, [Figure 4](#pone-0017288-g004){ref-type="fig"}) using the default -best value. ::: {#pone-0017288-g003 .fig} 10.1371/journal.pone.0017288.g003 Figure 3 ::: {.caption} ###### DeconSeq web interface. Screenshots of the DeconSeq web interface at different steps of the data processing. The user can either input a data ID to access already processed data (A) or input a new sequence file and select the database (B). After processing the data, the results are shown including the input information (C), Coverage vs. Identity plots for "remove" databases (D) and "retain" databases (E), classification of input data into "clean", "contamination", and "both" (F), and download options (G). :::  ::: ::: {#pone-0017288-g004 .fig} 10.1371/journal.pone.0017288.g004 Figure 4 ::: {.caption} ###### Coverage vs. Identity plots generated by DeconSeq. The plots show the number of matching reads for different query coverage and alignment identity values. The size of each dot in the plots is defined by the number of matching reads with exactly this coverage and identity value. Red dots represent matching reads against the "remove" databases and blue dots against "retain" databases. The column and row sums at the top and right of each plot allow an easier identification of the number of sequences that match for a particular threshold value. The plots for matching reads against the "remove" databases do not show matching reads that additionally have a match against the "retain" databases (A). Results for reads matching against both databases are shown in a second plot where dots for a single read are connected by lines. If the match against the "remove" database is more similar, then the line is colored red, otherwise blue. In B, for example, the majority of sequences is more similar to the "retain" databases and in C the majority is more similar to the "remove" databases. :::  ::: Evaluation of DeconSeq accuracy {#s2c} ------------------------------- Alignments are scored based on their query coverage and alignment identity percentages. Only hits above both thresholds are considered as valid alignments in the following evaluation. The accuracy of DeconSeq was benchmarked using simulated metagenomic data because "real" metagenomes lack the correct annotation for all sequences. Metagenomes consist of sequence fragments derived from the available genomes in the sampled environment [@pone.0017288-Hugenholtz1]. To simulate metagenomes, we extracted sequences from completely sequenced genomes and simulated substitution and indel errors (see Methods). DeconSeq was additionally benchmarked using artificial microbial metagenomes obtained from the Joint Genome Institute [@pone.0017288-Mavromatis1]. The human genomes were used as "remove" databases and the bacterial and viral genomes as "retain" databases. The accuracy values were calculated for threshold values of 95% query coverage and varying alignment identity. For identity thresholds of 94% and 97%, more than 99.9% of each simulated metagenome were classified correctly ([Table 1](#pone-0017288-t001){ref-type="table"}). For an identity threshold of 99%, the human metagenomes were classified correctly with lower accuracy, caused by the lower number of possible matching sequences due to the introduced error rate above 1% using a 1% average error rate. Variation in read length did affect the accuracy of DeconSeq in identifying contaminating sequences, as mainly short sequences were misclassified. ::: {#pone-0017288-t001 .table-wrap} 10.1371/journal.pone.0017288.t001 Table 1 ::: {.caption} ###### Accuracy of DeconSeq for identifying human DNA contamination in simulated metagenomic datasets. ::: {#pone-0017288-t001-1} Metagenome group Accuracy (in %) for identity threshold of ------------------ ------------------------------------------- ------------------ ------------------ Virus 99.9997 (0.0027) 99.9994 (0.0054) 99.9990 (0.0060) Human 99.9834 (0.0086) 99.9293 (0.0177) 72.3199 (0.2389) Bacteria 100 (0.0000) 100 (0.0000) 100 (0.0000) Bacteria JGI 99.9999 (0.0008) 99.9999 (0.0008) 99.9999 (0.0008) The accuracy values are average values of ten viral, ten microbial and ten human datasets with 100,000 sequences each and three microbial simulated metagenomes from JGI [@pone.0017288-Mavromatis1]. The accuracy values are shown for threshold values of 95% query coverage and varying alignment identity. The low accuracy value for the human datasets and 99% identity threshold was caused by the lower number of matching sequences due to the introduced errors above 1%. ::: Standalone and web application {#s2d} ------------------------------ DeconSeq is publicly available as standalone version or through a user-friendly web interface ([Figure 3](#pone-0017288-g003){ref-type="fig"}). The interactive web interface facilitates navigation through the results, definition of threshold parameters, and allows the export of the results for subsequent offline analysis. The input page of DeconSeq provides a mechanism to import new datasets and to select the contamination databases. Users can choose between submitting and processing a new dataset or accessing already processed datasets using a unique identifier. The web interface additionally provides graphical visualizations of the alignment results and the number of reads classified as contamination. The coverage vs. identity plots ([Figure 4](#pone-0017288-g004){ref-type="fig"}) can guide the users in their threshold selection. The connected dots in these plots help to identify possible contaminant sequences from non-contaminant sequences that match against both the "remove" and "retain" databases. Identification of human contamination in 202 metagenomes {#s2e} -------------------------------------------------------- In an application example, DeconSeq was applied to 202 longer-read metagenomic datasets previously published and with a mean read length greater than 150 bp (see [Table S1](#pone.0017288.s002){ref-type="supplementary-material"}). Metadata was either retrieved with the data from NCBI or through manual literature search. No prior knowledge of the amount of human contamination was assumed. The FASTA files were provided as input and the human databases were selected as "remove" and bacterial and viral databases were selected as "retain" for microbial and viral metagenomes, respectively. The results of the human contamination identified are summarized in [Figure 5](#pone-0017288-g005){ref-type="fig"}. The human contamination was identified for up to 64% of the metagenomes using the thresholds of 95% query coverage and 94% alignment identity. The host-associated metagenomes showed the highest fraction of likely human contamination. Of all metagenomes, 145 (72%) contained at least one possible contamination sequence. The two mouse-associated metagenomes with 24% and 29% possible human contamination were further compared to the mouse reference genome C57BL/6J build 37 to investigate if the high amount of possible contamination is host-related or of human origin. The two metagenomes contained 56% and 57% mouse-like sequences, respectively. ::: {#pone-0017288-g005 .fig} 10.1371/journal.pone.0017288.g005 Figure 5 ::: {.caption} ###### Result of human DNA contamination identified in 202 metagenomes. All seven human genome sequences were used as "remove" databases and depending on the metagenome type (viral or microbial), the viral or bacterial genomes were selected as "retain" database. 145 (72%) of the metagenomes contained at least one possible contamination sequence using a threshold of 95% query coverage and 94% alignment identity. :::  ::: Discussion {#s3} ========== Sequence contamination is a serious concern to the quality of genomic and metagenomic data used for downstream analysis. Therefore, it is important to process sequence data before analyzing it. We presented a program available as either standalone or web-based application that implements features to improve the quality of sequence datasets by identifying and removing possible sequence contamination with high accuracy. The program is targeted towards longer-read datasets and able to process next-generation sequence datasets with gigabases of data. There are different approaches on how to identify sequence contamination in genomic and metagenomic datasets and the current methods have critical limitations. The dinucleotide relative abundance was used by Willner et al. [@pone.0017288-Willner1] to predict if a metagenome was contaminated. However, this approach only allows the identification of contamination in the whole dataset, not on the level of single sequences. Others used BLAST to compare metagenomes to the human reference genome [@pone.0017288-Willner2], [@pone.0017288-Turnbaugh2]. BLAST has a speed disadvantage that makes it a bottleneck for analyzing the huge amounts of data typical of current sequencing projects. The identification of possible contamination based on sequence alignments, however, seems to provide the only reliable option currently available to classify single sequences as contamination. A major limitation of the alignment approach is the lack of corresponding regions that do not exist in the reference genome(s), referred to as dark matter, which may result from gaps in the reference or the presence of structural variants in the genome(s) being analyzed [@pone.0017288-Frazer1], [@pone.0017288-Kidd1]. Li et al. [@pone.0017288-Li7] have found the presence of extensive novel sequences in recently sequenced human genomes that were absent from the human reference genome. Therefore, DeconSeq provides databases for the identification of human DNA contamination from the seven currently available human genomes. More genomes will be sequenced, for example, in large-scale resequencing projects such as The 1000 Genomes Project to enhance our understanding of how genetic differences affect health and disease. This will provide a resource for a more complete human decontamination reference database. The choice of the alignment program depends on the biological application and on the type of sequencing technology used to generate the data. The scalability of an alignment program for speed and memory usage is important as neither memory nor CPU power is growing as fast as sequencing capacity. The ability to align sequences of different lengths is an important factor as well, considering the rapidly evolving field of next-generation sequencing and the constantly increasing length of produced reads. As shown in this study, BLAST does not scale well for the identification of human DNA contamination in next-generation sequencing data. Using filters for low complexity or repeat regions may significantly reduce the resources consumed, but also decrease sensitivity. BWT-based alignment programs are more efficient on very long reference sequences such as mammalian genomes because their complexity is better than being linear to the reference length. On bacterial genomes however, these aligners might be slower than hash table based alignment programs. Furthermore, alignment programs such as BWA-SW can be used for sequences generated by methods ranging from pyrosequencing and Sanger sequencing to single molecule sequencing. Such programs allow the alignment of sequence reads of more than 1 Mbp. The ability to map sequence reads uniquely to the correct location is dependent on a number of factors such as the complexity of the reference data (highly polymorphic or repetitive regions), length of the sequence reads, error rates of the reads, and the diversity of the individual organism compared to the reference [@pone.0017288-Li2], [@pone.0017288-Turner1]. Wrong alignments may be caused by overlooking alignments with a similar score to the best reported alignment. Different alignment programs handle the issue of reporting unique hits or multiple hits differently. The implemented algorithm either randomly chooses one, reports all above a cutoff, or those with the best alignment score. When errors are introduced, reads might match better at a different locus than the original one, and therefore evaluations of programs with real data containing errors are challenging. The number of suboptimal hits may help to decide which alignments are reliable. In practice however, only the best alignment is used in the analysis [@pone.0017288-Meyer1]. To identify contamination, it is sufficient to find a single match above given thresholds without calculating all possible matches. Smith et al. [@pone.0017288-Smith2] found that using base quality scores improves alignment accuracy if the aligner uses lower penalties for an error-prone mismatch. However, accurate quality scores are not always available and the only program evaluated in this study that is able to incorporate quality scores into the alignment algorithm did not fulfill the system requirements. The algorithm implemented in BWA-SW does not make use of quality data, but the quality information could be exploited to estimate the confidence in an alignment. Not all sequences that should have been aligned might have been aligned using a given program. In some instances, the sequence read cannot be mapped to the reference. Most of these errors arise from failing to find a seed during the mapping step of the algorithm. Repeat regions are problematic for alignment algorithms and users tend to mask sequences before performing the alignments. However, not allowing seeding in matching regions of the reference sequence that are masked for repeats might result in unaligned query sequences. The human reference genome build 37 has 50.2% of the genome masked as repeat, reducing the number of possible seeding positions. It is more likely to find sufficient seeds from which to extend the alignment for longer reads. As read length increases, the mapping in repetitive regions will improve. We showed that the BWA-SW program used by DeconSeq has a high sensitivity (including repetitive regions) for sequences with low error rates or longer reads when aligning human DNA to the reference genome. Heuristics present another source of alignment errors especially for short queries, because only a few valid unique seeds may exist between the aligned sequences. BWA-SW, for example, tends to miss short alignments with high error rates, as it does not guarantee to find all local hits due to the heuristic acceleration. In contrast, BWA-SW might find seeds where other programs, such as BLAST, do not. BLAST uses identical seeds that might not work well for (short) query sequences that contain mismatches because there might be no seed sequence from which to extend the alignment. BWA-SW finds seeds by dynamic programming between two FM-indices and allows mismatches and gaps in the seeds. To achieve higher sensitivity, regions that do not align with a given program can be identified and aligned using more sensitive (and usually much slower) parameters or alternative programs. The default value for in BWA-SW is one. Increasing improved accuracy slightly for test datasets, but greatly reduced the alignment speed. The alignment of sequences against a reference is considered "embarrassingly" parallel, being easy to distribute the required computational work over the nodes of a compute cluster. However, parallelization alone does not always solve the problem of analyzing the huge amounts of data generated by next-generation sequencing machines and speed of the program stays an important factor when choosing programs. We showed that the identification of contaminating sequences done by BLAST+ in hours could be achieved by BWA-SW in minutes. Speed is gained in BWA-SW largely from the use of FM-indices and by reducing unnecessary extension for highly repetitive sequences [@pone.0017288-Li5]. However, the speed of alignments is largely determined by the error rate of the query sequences. The error rates will likely be reduced greatly using third-generation sequencing techniques, such as single-molecule techniques that are able to sequence the same template molecule more than once and produce a consensus read with reduced stochastic errors that may occur [@pone.0017288-Metzker1]. It is important that the limitations of the programs used for analysis be understood. Next-generation alignment programs were mainly designed for DNA alignments implementing a 2-bit representation of sequences. The 2-bit representation restricts the use of ambiguous bases such as N. SSAHA2 replaces ambiguous bases by base A and BWA-SW randomly chooses A, C, G or T as replacement. This can lead to false positive hits especially in long stretches of Ns in genomic sequences. To reduce the number of false positive, we removed long stretches of Ns in the genomes and modified BWA-SW to mismatch Ns in the query sequences during the Smith-Waterman alignment. There are also limitations in speed and accuracy of the BWA-SW program. BWA-SW can be used to align 100 bp reads, but it is slower than using BWA. BWA-SW is less accurate than SSAHA2 on 100--200 bp reads for error rates above 2% [@pone.0017288-Li5]. Most 454 libraries, however, have an average read length of 300--500 bp. Additionally, this and other studies [@pone.0017288-Li5], [@pone.0017288-Li6] show that BWA-SW is up to tens of times faster than existing programs. Next-generation alignment programs are under active development and the performance and feature set of each of these programs is likely to improve. If the loss of sequence data can be afforded, reads with high error rates (for example containing low base quality scores or ambiguous bases) and short reads should be filtered prior to using DeconSeq to ensure high accuracy of the contaminant classifications. The BWA-SW program was modified to fit the needs of DeconSeq. Those modifications do not change the default behavior of the algorithm and are only forced using additional parameters. The default SAM output contains data that is not needed for DeconSeq and usually generates huge output files for reference datasets with a large number of sequences. Furthermore, the Cigar string (a human readable alignment string) presented the only resource in the SAM output from BWA-SW that could be used to calculate coverage and identity values of the alignments. However, the Cigar string uses "M" for matching positions and replacement (mismatch) positions. This would require realigning the sequences in the regions specified by "M" to retrieve the number of replacements used for alignment identity calculations. The mapping quality in the SAM file did not present a sufficient value for the use as threshold. In any case where there are two or more equally likely alignments (multiple locations a query can map to), the mapping quality is zero. This may occur for the repeat-rich human genome and equally likely alignments can still represent contaminating sequences. DeconSeq uses coverage and identity thresholds to determine if a match is a possible contamination or not. This approach is based on the idea that looking for similar regions consists of grouping sequences that share some minimum sequence similarity over a specified minimum length. It is important that the limitations of this approach be understood. The approach invariably leaves out related regions that have degraded over time, so their similarity is below the threshold. Moreover, the thresholds chosen to group elements together often have no connection to evolutionary history and the underlying mechanisms of formation. For example, the operational definition of segmental duplications excludes ancient duplications that were formed by the same mechanisms long ago but that have since degraded below 90% sequence identity [@pone.0017288-Alexander1]. There is no "one-size-fits-all" solution and each user must make informed decisions as to the appropriate thresholds used for decontamination. Thresholds should not be set to 100% if errors are expected in the sequence reads. To our knowledge, DeconSeq is the first program optimized to automatically identify and remove sequence contamination from large sequence datasets. In order to avoid the classification of non-contaminating sequences as contamination, all possible contamination can be compared to a second set of databases and marked accordingly. This is especially useful for the identification of human contamination in viral metagenomes because there are a large number of viral or viral-like sequences hidden in the human genome. It is important to note that viral or bacterial sequences of unknown origin but highly similar to the human genome will be classified as contaminants due to the missing reference sequences in the second set of databases. We evaluated the classification of contamination using simulated datasets and showed that DeconSeq performed with very high accuracy. The highest levels of possible contamination in 202 previously published microbial and viral metagenomes were found in host-associated metagenomes suggesting DNA extraction issues rather than contamination introduced during sample processing. Next-generation sequencing data is available to most small laboratories. However, they do not always have access to the required computing resources. The web-based version of DeconSeq allows users to conduct decontamination using our in-house computing resources and provides additional visualizations such as coverage vs. identity plots to help users choose the best thresholds for their datasets. Furthermore, the web-based version provides the latest versions of a variety of datasets such as human genome sequence assemblies. Users can contact the authors and request additional databases for specific decontamination purposes. Design and Implementation {#s4} ========================= Reference data {#s4a} -------------- The human reference genome build 37, the Celera Genomics human genome assembly, the J. Craig Venter genome (HuRef), and The Center for Applied Genomics (TCAG) human chromosome 7 version 2 assembly were downloaded from National Center for Biotechnology Information (NCBI). The Korean male (Seong-Jin Kim; SJK) genome data was retrieved from KOBIC. The Asian male (Han Chinese individual; YH) genome data was retrieved from the YanHuang database. The unique James D. Watson sequences were downlaoded from NCBI and the unique Asian (YH) and unique Yoruban male (NA18507) sequences were downloaded from the supplemental material of Li et al. [@pone.0017288-Li7]. All unique sequences were filtered to remove sequence copies and only keep sequences with at least 300 bp. The bacterial genomes (1,116 genomes in 2,103 fasta files as of 06/06/2010) and viral genomes (3,642 genomic sequences as of 06/06/2010) data was retrieved from NCBI. The gene and repeat annotations for the human reference genome build 37 were downloaded from the UCSC Genome Browser [@pone.0017288-Kent2]. The amount of the genome that was repeat-masked was calculated based on all non-ambiguous bases. A more detailed description including links can be found in [Text S1](#pone.0017288.s001){ref-type="supplementary-material"}. Simulated metagenomes {#s4b} --------------------- The program Grinder version 0.1.8 (<http://sourceforge.net/projects/biogrinder/>) was used to create simulated human, bacterial and viral metagenomic sequences. Sequences were generated using an average error rate of 0.85% substitutions and 0.15% indels (-m 0.85 0.15), and normal distributed read lengths with a mean of 380 bp and standard deviation of 100 bp (−l 380 normal 100). Simulated sequences were then filtered using PRINSEQ [@pone.0017288-Schmieder1] to generate ten human, ten bacterial and ten viral datasets with 100,000 unique sequences containing no Ns and a read length of at least 100 bp. Additionally, three artificial microbial metagenomes with different complexity obtained from the Joint Genome Institute (JGI; <http://fames.jgi-psf.org/>) were used [@pone.0017288-Mavromatis1]. The JGI metagenomes were pre-processed using PRINSEQ to trim poly A/T tails longer than 10 bp and to remove reads shorter than 100 bp and exact sequence duplicates. The resulting three datasets contained 116,739, 97,479 and 114,430 sequences with a mean read length of 948.5 bp, 950.9 bp and 966.8 bp, respectively. Human reference datasets {#s4c} ------------------------ The Human reference genome build 37 was used to analyze the type and amount of unaligned sequences using BWA-SW. Datasets with sequences of 200 bp, 500 bp and 1,000 bp length were generated from the reference genome sequence using 50% overlap. All sequences that contained the ambiguous base N were discarded as N aligned to N is considered a mismatch and would alter the alignment identity for identical sequences. The resulting datasets contained 28,612,955 reads for 200 bp, 11,444,886 reads for 500 bp, and 5,722,210 reads for 1,000 bp. Error rates of exactly 2% and 5% (with 15% indels and 85% substitutions) were then simulated for each of the three datasets resulting in 6 additional datasets. Reference databases for web-based version {#s4d} ----------------------------------------- The web-based version offers pre-processed reference databases for a variety of complete genomes such as human, bacterial and viral genomes. The genome data was preprocessed before indexing using BWA. To reduce the number of false positive matches that might be introduced due to the long stretches of Ns that will be randomly replaced by A, C, G or T during database indexing, the genome sequences were split at stretches of 200 or more Ns. The separated sequences were then filtered for read duplicates to reduce redundancy in the sequence data and for short sequences that contained more than 5% of ambiguous bases (N). In its current version, BWA-SW fails to index the complete dataset from multiple human genomes (BWTIncConstructFromPacked error). However, the error was not a concern for the web-based version because to decrease the memory usage on the computing cluster, the genome data was split into smaller files that require a maximum of 1.5 GB of memory per chunk. The results for the split databases are automatically joined before generating the output for the web-based version of DeconSeq. Implementation and computational platform {#s4e} ----------------------------------------- DeconSeq was implemented as standalone and web-based version in Perl. The workflow of DeconSeq is shown in [Figure 6](#pone-0017288-g006){ref-type="fig"}. The DeconSeq web application is currently running on a web server with Ubuntu Linux using an Apache HTTP server to support the web services. The web interface provides a high level of compatibility with heterogeneous computing environments. The alignments are computed on a connected computing cluster with ten working nodes (each with 8 CPUs and 16 GB RAM) running the Oracle Grid Engine version 6.2. The input data is automatically split into chunks for optimized distribution of work over the working nodes. ::: {#pone-0017288-g006 .fig} 10.1371/journal.pone.0017288.g006 Figure 6 ::: {.caption} ###### Flowchart of DeconSeq for the identification of possible contaminant sequences. :::  ::: Modifications of BWA-SW {#s4f} ----------------------- The BWA-SW source code was modified to fit the requirements for DeconSeq. The file bwtsw2\_aux.c was modified to generate an alternative output, which presents a lightweight tab-separated output format containing only the necessary data required by DeconSeq (query identifier, reference identifier, query coverage and alignment identity). The file bwtsw2\_aux.c was additionally modified to force a mismatch when aligning the ambiguous base N in query sequences instead of randomly replacing it by A, C, G or T and possibly resulting in a match (BWA-SW default). The files stdaln.c, stdaln.h and bwtsw2\_aux.c were modified to include "R" for replacements in an extended version of the Cigar string, instead of using "M" for both match and replacement (mismatch). The files bwtsw2\_main.c and bwtsw2.h were modified to fix the double defined parameter -s (changed to -s and -S), and to add the new parameters -A (generate alternative output), -R (output extended version of Cigar string with replacements) and -M (force to mismatch Ns in query sequence). The modified version of BWA-SW is made available as part of the DeconSeq source. Input and output {#s4g} ---------------- The input for DeconSeq is FASTA formatted data containing the genomic or metagenomic reads. In addition to FASTA files, the user can submit FASTQ files (containing sequence and quality data) [@pone.0017288-Cock1] using the web interface. The BWA-SW algorithm does not make use of quality data during the alignment of sequences and does therefore not require quality data as input. The input data is checked to be a valid file with DNA data. If the input data fails the validation step, further processing is restricted. Sequence files can be of large size (several 100 MB), and therefore the web interface additionally allows the submission of compressed FASTA or FASTQ files to reduce the time of data upload (by approximately 70%) from the user machine to the web server. Currently, ZIP and GZIP compression algorithms are supported. If the compressed files contain more than one FASTA or FASTQ file, the single files will be joined into one dataset. The file formats and compression types are automatically detected and processed accordingly. There is no limit on the number of sequences or the size of the input file accepted by DeconSeq. The web-based version of DeconSeq offers several pre-processed databases to select from for the two categories of "remove" and "retain". Databases are available, for example, for the seven publicly available complete human genomes, as well as for groups of bacterial and viral genomes. The databases used for the web-based version are automatically updated on a regular basis. The user can download the results in the web interface in FASTA or FASTQ (if provided as input) format or its compressed version. The results can either be separated or joined files. This allows the user to further investigate the results separately. Results will be stored for the time selected by the user (either one day or one week), if not otherwise requested, on the web server using a unique identifier displayed during data processing and on the result page. This identifier allows the user to share the result with other researchers without having to re-submit and re-process the dataset. Filter and threshold parameters {#s4h} ------------------------------- The user can filter the data based on different parameters. Unlike the standalone version, the web-based program allows the user to define filter parameters based on the input data after the data is processed. This does not require an *a priori* knowledge of the best parameters for a given dataset and the parameter choice can be guided by the graphical visualization of the results. Sequences are classified as contamination if they have a match above the threshold values against any database selected for "remove". The thresholds are based on query coverage and alignment identity to allow an unbiased filtering for diverse datasets and databases, in contrast to using unnormalized E-values or alignment scores. Threshold values are rounded toward the lower integer (e.g. 99.95% is rounded to 99%). In order to avoid the classification of non-contaminating sequences as contamination, all possible contaminating sequences can be compared to alternative databases ("retain" databases) and matches above the thresholds are marked accordingly with "Hit to both". Analysis of 202 metagenomes {#s4i} --------------------------- The amount of possible human DNA contamination in 202 longer-read metagenomes (150 bp mean read length) was estimated using DeconSeq. These metagenomes were previously published and are publicly available from NCBI (<http://www.ncbi.nlm.nih.gov/>). The metagenomes used here represent viral and bacterial communities sampled from a diverse array of biomes and were categorized as one of the following: "aquatic", "terrestrial", and "host-associated". The metagenomes were further subdivided into their sampled environment, such as "human", "mouse", and "soil". Sampling, filtering, processing and sequencing methods differed among the compiled metagenomes. [Table 2](#pone-0017288-t002){ref-type="table"} provides a summary of the number of metagenomes from each type and biome (a more detailed list of the complete dataset can be found in [Table S1](#pone.0017288.s002){ref-type="supplementary-material"}). ::: {#pone-0017288-t002 .table-wrap} 10.1371/journal.pone.0017288.t002 Table 2 ::: {.caption} ###### Summary of metagenomes by type and biome used in this study. ::: {#pone-0017288-t002-2} Biome Number of viral metagenomes Number of microbial metagenomes ------------------------- ----------------------------- --------------------------------- Aquatic 1 58 Terrestrial 9 6 Host-associated (total) 65 63 Host-associated (human) 62 50 **Total** **75** **127** The metagenomes were previously published and available through NCBI. The metagenomes were not targeted to a single loci and the mean read length was above 150 bp after trimming and filtering. ::: The metagenomes used in this study were pre-processed prior to any processing with DeconSeq. UniVec build 5.2 (<http://www.ncbi.nlm.nih.gov/VecScreen/UniVec.html>) and cross\_match (<http://www.phrap.org/>) were used to screen for vector contamination in the metagenomes. TagCleaner [@pone.0017288-Schmieder2] was used to trim adapter and tag sequences. PRINSEQ [@pone.0017288-Schmieder1] was then used to filter exact sequence duplicates, sequences shorter than 50 bp or longer than 10,000 bp, sequences containing more than 5% of ambiguous base N after trimming Ns from the sequence ends, and sequences containing non IUPAC conform characters for DNA sequences. The resulting datasets were excluded from the study if the mean sequence length was below 150 bp or the dataset contained less than 1,000 metagenomic sequences. Metagenomes targeted to single loci such as 16S rRNA studies were excluded as well. For all metagenomes, DeconSeq was run using all human databases for "remove" and depending on the type (microbial or viral) the bacterial or viral genomes database was selected for "retain". The threshold values were set to 95% coverage and 94% identity. Calculation of sensitivity and accuracy {#s4j} --------------------------------------- Sensitivity (or true positive rate) was used to evaluate alignment performance for BWA-SW and was calculated for query coverage and identity thresholds ranging from 90% to 100%. Accuracy was used as measurement for the proportion of true classifications by DeconSeq and calculated for thresholds of 95% query coverage and varying alignment identity.Here, the reads that could be aligned back to the reference sequence were considered true positives (). Reads that could not be aligned were considered false negatives ().Here, reads that were human and that were classified as human were considered . Reads that were non-human and were not classified as human were considered true negatives (). Reads that were classified as "Hit to both" were considered for human reads and for non-human reads. The number of reads equals to the sum of true positives, false positives, true negatives and false negatives. Availability and Future Directions {#s5} ================================== The DeconSeq standalone version, test datasets, the documentation and the link to the web-based version are available at <http://deconseq.sourceforge.net/>. All further developments will be made available through this website. Future work will include interface improvements of the web-based version (additional visualizations and filter options) and non-redundant databases to account for the increasing amount of reference genomes containing only a small fraction of new sequence data. Supporting Information {#s6} ====================== Text S1 ::: {.caption} ###### **Additional methods and results.** This text includes a detailed guide to the retrieval of the reference data and benchmarked programs, the generation of benchmark data, the comparison of program performance for Mosaik, NUCmer, BLAST, BLAST+ and BWA-SW, as well as additional [Figures 1](#pone-0017288-g001){ref-type="fig"}, [2](#pone-0017288-g002){ref-type="fig"}, [3](#pone-0017288-g003){ref-type="fig"}, [4](#pone-0017288-g004){ref-type="fig"}. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Details of the 202 metagenomes used for the identification of possible human contamination by DeconSeq.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank John Edwards, Yan Wei Lim, and Forest Rohwer for insightful discussions and helpful suggestions on the manuscript. We also thank the DeconSeq users for their comments and suggestions. Finally, we thank the J. Craig Venter Institute for making metagenomes of the Global Ocean Sampling Phase II publicly available. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by grant DBI 0850356 Advances in Bioinformatics from the National Science Foundation (<http://www.nsf.gov/>). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: RS RE. Performed the experiments: RS. Analyzed the data: RS. Wrote the manuscript: RS RE.

PubMed Central

2024-06-05T04:04:19.119656

2011-3-9

PMC3052305

Introduction {#s1} ============ Insulin/IGF-I-like signaling (IIS) is a highly conserved pathway for promoting growth under replete conditions. Growth control is cell-autonomously regulated by IIS to determine cell and organ size in the affected tissue [@pone.0017369-Butler1], [@pone.0017369-Garofalo1]. In addition to this well-conserved cell-autonomous function, IIS can also have non-autonomous effects on other parts of the body. These actions of IIS have been uncovered in *C. elegans*, *Drosophila* and mice, through studies of mosaic animals with IIS restricted to specific tissues. In worms and flies, tissue-restricted IIS non-autonomously regulates development and adult longevity [@pone.0017369-Apfeld1], [@pone.0017369-Giannakou1], [@pone.0017369-Hwangbo1], [@pone.0017369-Libina1], [@pone.0017369-Wolkow1], [@pone.0017369-Iser1]. In mice, brain-specific insulin receptor deletion is associated with obesity and low fertility, likely reflecting hormonal disruptions [@pone.0017369-Bruning1]. Tissue-restricted IIS could confer these non-autonomous effects either specifically, through endocrine outputs, or non-specifically, through pleiotropic phenotypes resulting from tissue dysfunction due to inadequate growth. Thus, these findings raise new challenges for identifying the downstream pathways mediating non-autonomous effects of IIS [@pone.0017369-Broughton1]. This question can be investigated in *C. elegans*, for which the major IIS pathway components have been identified. These include *daf-2*, encoding the sole *C. elegans* insulin/IGF-I receptor-like protein, and *age-1*, encoding a p110 PI3K catalytic subunit that is the primary DAF-2/IR effector [@pone.0017369-Larsen1], [@pone.0017369-Kimura1], [@pone.0017369-Morris1], [@pone.0017369-Gottlieb1]. The major downstream target of *daf-2* and *age-1* is *daf-16*, which encodes a FOXO transcription factor antagonized by DAF-2 signaling [@pone.0017369-Ogg1], [@pone.0017369-Lin1]. In *C. elegans*, the *daf-2* pathway acts at both the cellular and organism level. At the cellular level, the *daf-2* pathway cell-autonomously regulates *sod-3* expression [@pone.0017369-Libina1], [@pone.0017369-Furuyama1], [@pone.0017369-Honda1]. A second cell-autonomous output of the *daf-2* pathway is the regulation of FIRE response sensitivity in intestinal cells [@pone.0017369-Iser1]. Two types of behavioral plasticity are also regulated cell-autonomously by *daf-2* [@pone.0017369-Tomioka1]. The non-autonomous outputs of *daf-2* regulate organismal phenotypes. The *daf-2* pathway promotes reproductive development and prevents dauer larval arrest under replete conditions [@pone.0017369-Gottlieb1], [@pone.0017369-Riddle1], [@pone.0017369-Thomas1], [@pone.0017369-Vowels1]. In adult animals, the *daf-2* pathway promotes wildtype longevity and normal stress resistance [@pone.0017369-Larsen1], [@pone.0017369-Gems1], [@pone.0017369-Johnson1], [@pone.0017369-Kenyon1], [@pone.0017369-Klass1], [@pone.0017369-Walker1]. Both dauer arrest and adult longevity are controlled non-autonomously by *daf-2* and *age-1* activity from several cell types [@pone.0017369-Apfeld1], [@pone.0017369-Wolkow1], [@pone.0017369-Iser1]. The downstream effectors for *daf-2* non-autonomous regulation of dauer arrest and adult longevity are not known. *daf-16*, the major cell-autonomous *daf-2* target, regulates longevity primarily from intestinal cells [@pone.0017369-Libina1]. A working model proposes that *daf-2* activity can regulate *daf-16* through both the cell-autonomous pathway, via *age-1* and *akt-1*, and non-autonomously, through unidentified pathways [@pone.0017369-Iser1]. The *daf-2* pathway\'s non-cell autonomous actions may reflect crosstalk with other signaling pathways that convergently regulate dauer arrest and adult longevity. One candidate is the heat-shock transcription factor, encoded by the *hsf-1* gene, which regulates lifespan, proteotoxicity and dauer arrest in collaboration with *daf-16* [@pone.0017369-Lin1], [@pone.0017369-Walker1], [@pone.0017369-Hsu1], [@pone.0017369-Morley1]. HSFs are highly conserved and direct the expression of heat-shock proteins in response to thermal stress. In *C. elegans*, *hsf-1* also promotes the expression of other, non-*hsp*, targets in *daf-2* mutants through both *daf-16*-dependent and independent mechanisms [@pone.0017369-Hsu1]. To identify factors mediating the non-cell autonomous effects of the *daf-2* pathway upon dauer arrest, we searched for transcriptional targets regulated non-autonomously by *age-1* and then analyzed factors directing their regulation in response to the *daf-2/age-1* pathway. Using microarrays, we examined gene expression in animals with *age-1* activity restricted to neurons or gut, and the results were compared with gene expression in wildtype animals and zygotically null *age-1* mutants (*m+z-*). This approach identified a collection of transcripts that were potentially regulated by *age-1* in a non-cell autonomous fashion. We characterized the *cis-* and *trans-*requirements for *daf-2*-dependent expression of on of these non-autonomous targets, *cyp-35B1/dod-13*. The findings suggest that *hsf-1* may be a component of pathways mediating *age-1* non-autonomous activities. Results {#s2} ======= Gene expression patterns in animals with tissue-restricted *age-1* activity {#s2a} --------------------------------------------------------------------------- In order to search for targets regulated non-autonomously by *age-1*, gene expression was analyzed in animals with tissue-restricted *age-1* activity. This analysis compared gene expression in zygotically null *age-1* mutants (*age-1(mg44)*(m+z-)) with that in *age-1(mg44)* animals carrying transgenes directing neuronally-restricted (CY251) or intestinally-restricted (CY262) *age-1* expression ([Fig. 1A](#pone-0017369-g001){ref-type="fig"}). Both neuronal and intestinal *age-1* expression rescued constitutive dauer arrest of *age-1(mg44)* [@pone.0017369-Iser1]. The extended lifespan of *age-1(mg44)* adults was also rescued by *age-1* expression in either tissue, although CY262 more strongly rescued adult longevity than CY251, consistent with a critical role for intestinal *daf-16* activity for extended lifespan [@pone.0017369-Libina1]. ::: {#pone-0017369-g001 .fig} 10.1371/journal.pone.0017369.g001 Figure 1 ::: {.caption} ###### Transcriptional microarrays were used to identify non-autonomous *age-1* target genes. Global gene expression was compared in three strains; zygotic null *age-1(mg44)* adults, CY251 (*age-1(mg44)* with neuronally-restricted *age-1* expression) and CY262 (*age-1(mg44)* with intestinally-restricted *age-1* expression). All 3 strains were compared to the wildtype (*age-1(+)*). The dauer-constitutive phenotype of *age-1(mg44)* is maternal-effect. Therefore, this analysis used zygotic *age-1(mg44)* homozygotes (z-) from heterozygous *age-1(mg44/+)* parents (m+), which developed to adulthood due to maternal *age-1* activity. Tissues with *age-1* activity are colored orange; tissues lacking *age-1* activity are outlined in black. A. Strategy for analyzing microarray results to screen for potential non-autonomous *age-1* target genes. Step (1): Transcript abundance in all 3 strains was compared to wildtype. Step (2): Transcript abundance in CY262 and CY251 was compared to *age-1(mg44)*. Transcripts altered in *age-1(mg44)* (p\<0.05 vs wildtype) and rescued in both CY262 and CY251 (p≥0.05 vs wildtype) were considered potential non-autonomous *age-1* targets. B. Heat map showing relative expression levels for 45 genes previously identified as *daf-2* targets [@pone.0017369-Murphy1]. The results show that elevated expression of 3 *hsp-16* genes in *age-1(mg44)* animals was rescued to low levels by tissue-restricted *age-1* activity, suggesting these may be non-autonomously regulated by *age-1*. However, *sod-3* levels remained elevated in the presence of tissue-restricted *age-1* activity, consistent with findings that the *daf-2* pathway cell-autonomously regulates *sod-3* in many tissues [@pone.0017369-Libina1]. :::  ::: Since *age-1* and *daf-2* mutants share many phenotypes, we expected that the *age-1* and *daf-2* transcriptomes would be similar. Therefore, we compared our results for *age-1(mg44)* and those of a previous study of gene expression in *daf-2* pathway mutants [@pone.0017369-Murphy1]. Of 113 *daf-2* targets which were also significantly changed in our experiment, 73% were changed concordantly in *age-1(mg44)* adults (fold-change p≤0.05, t-test) ([Table S1](#pone.0017369.s001){ref-type="supplementary-material"}). Considering the differences in reference pools and growth conditions, these results indicate high concordance of the *age-1* and *daf-2* transcriptomes, consistent with the fact that *age-1* and *daf-2* have similar mutant phenotypes [@pone.0017369-Gottlieb1]. These findings support the role of AGE-1/PI3K as the major effector for DAF-2 signaling. The goal of this analysis was to identify *age-1* target genes that could be regulated non-autonomously by the *age-1* pathway. We reasoned that non-autonomous targets would be rescued to wildtype levels in the strains CY262 and CY251, which express *age-1* only in the intestine or neurons, respectively. Using these criteria, we examined the effect of tissue-restricted *age-1* activity on the 82 transcripts whose expression was regulated concordantly in *age-1* and *daf-2* mutants. We found that 30% (25 targets) were rescued in both CY262 and CY251, one target was rescued only in CY251, three targets were rescued in only CY262 and 65% (53 targets) were not rescued in either strain or had inconclusive data ([Table S1](#pone.0017369.s001){ref-type="supplementary-material"}). The targets whose expression was rescued in both CY251 and CY262 are potential non-autonomous *age-1* targets ([Fig. 1B](#pone-0017369-g001){ref-type="fig"}, [Table S1](#pone.0017369.s001){ref-type="supplementary-material"}). These included genes involved in a variety of processes, such as antimicrobial defense (*clec-13*, *lys-7*), reproduction (*vit-2*, *-4* and *-5*), catalysis (*F09F7.7* dioxygenase) and metabolism (*acr-2*). Little expression data was available for these targets, although two are reportedly intestinal (*lys-7*, *lipl-4*) and four have complex expression patterns (*cdr-2*, *F09F7.7*, *acs-2* and *F552H3.5*). We then conducted this comparison for 791 transcripts that were significantly and reproducibly overrepresented in *age-1(mg44)* adults versus wildtype adults (at least 2-fold overrepresented, p≤0.05), without regard to their inclusion in the *daf-2* transcriptome ([Table S2](#pone.0017369.s002){ref-type="supplementary-material"}). We set two criteria for rescue of *age-1* targets in these strains. First, *age-1* targets were defined as being more than 2-fold overexpressed in comparison with wildtype levels with a p-value\<0.05 (t-test vs wildtype). For rescue in CY262 or CY251, expression was both insignificantly different from that in wildtype animals (p≥0.05) and significantly different from that in *age-1(mg44)* animals (p\<0.05). Within the group of 791 *age-1*-upregulated transcripts, 127 (16%) were rescued in both CY262 and CY251, and were potentially regulated non-autonomously by *age-1*. This group was composed of genes involved in a variety of biological processes, including defense or signaling (7 glutathione S-transferases, 5 cytochrome P450s, 6 lectins, 3 alcohol dehydrogenases, 4 glucuronosyltransferase and 3 nuclear hormone receptors). We expect that some non-autonomous *age-1* targets should be expressed outside of the *age-1*-expressing tissues in CY262 and CY251. Therefore, we surveyed the available expression data for these genes. Of the 127 potentially non-autonomous *age-1* targets, expression data was available for 18 (Wormbase, release 190). Eleven genes were reportedly expressed in the intestine, with 5 exclusively intestinal, although the significance of intestinal expression is unclear since this is a common site for promiscuous transgene expression in *C. elegans* [@pone.0017369-Iser1]. The targets that were not exclusively intestinal were expressed in a variety of tissues, including neurons (7), hypodermis (8), the gonad (2), epidermal seam cells (3), muscle (5) and the pharynx (5). This finding is consistent with the idea that *age-1* non-autonomous outputs might target many of the body\'s tissues. Using these rescue criteria, we also found that 12 targets were preferentially rescued in CY251 animals, while 37 were preferentially rescued in CY262 animals. The targets rescued preferentially in CY251 included one metalloprotease (*nas-9*), one NADH oxidase (*F17A9.5*) and one lectin (*clec-4*). Expression data was only available for *nas-9*, which is expressed in the hypodermis. The targets preferentially rescued in CY262 functioned in a variety of processes including stress resistance (3 glutathione S-transferases, 3 glucuronosyltransferases) and metabolism (1 fatty acid desaturase and 2 lipases). Expression data is available for 10 of the CY262-rescued targets. Eight of these targets are expressed intestinally, with 5 expressed exclusively in the intestine (*fat-5* fatty acid desaturase, *F09C8.1* phospholipase, *F54F3.3* lipase, *F42A10.6* unknown function and *lec-6* lectin). We note that the proportion of intestinal genes was the highest among the CY262-rescued targets (8/10, 80%) and was lower in the group of targets rescued in both CY251 and CY262 animals (5/11, 45%). We propose that the targets preferentially rescued in CY262 animals represent *age-1* targets that are autonomously regulated by intestinal *age-1* activity. *age-1* non-autonomously regulates a subset of stress response genes {#s2b} -------------------------------------------------------------------- Adult longevity and stress resistance in *daf-2* and *age-1* mutants results, at least in part, from transcriptional upregulation of stress-response genes [@pone.0017369-Walker1], [@pone.0017369-HalaschekWiener1], [@pone.0017369-Larsen2]. Therefore, we examined our microarray data to determine whether the stress-resistance genes regulated by the *daf-2* and *age-1* pathway were autonomously or non-autonomously regulated ([Table S3](#pone.0017369.s003){ref-type="supplementary-material"}). Interestingly, *sod-3* expression was not rescued in CY262 or CY251 animals, suggesting this target is regulated cell-autonomously and is not under endocrine control. The heat shock genes targeted by heat shock factor, HSF-1, are major contributors to *daf-2* longevity [@pone.0017369-Walker1], [@pone.0017369-Hsu1], [@pone.0017369-Morley1], [@pone.0017369-Garigan1]. Our array results revealed that overexpression of *hsp-16* genes and *hsp-17* was rescued in CY262 and CY251, suggesting that small heat-shock proteins may be non-autonomous *age-1* targets. We note that, although the *hsp-16* genes were robustly upregulated in one *age-1(mg44)* sample, the were less robustly upregulated in an *mg44* replicate sample, so that the average change was not statistically significant. Nevertheless, these findings indicate that a subset of the stress response genes induced in *age-1* animals are non-autonomously regulated by *age-1*. Expression of an endocrine *age-1* target, *cyp-35B1*, is *daf-16* and *hsf-1*-dependent {#s2c} ---------------------------------------------------------------------------------------- One of the putative endocrine targets of *age-1* that we identified was *cyp-35B1*, which had previously been described as a *daf-2* pathway target named *dod-13* [@pone.0017369-Murphy1]. The *cyp-35B1* transcript was 17.8-fold overexpressed in *age-1(mg44)* hermaphrodites compared to wildtype and was rescued in both CY262 and CY251 (1.64-fold and 2.73-fold overexpressed vs wildtype, respectively). The *cyp-35B1* gene encodes a cytochrome P450 enzyme implicated in xenobiotic detoxification [@pone.0017369-Budovskaya1]. The *cyp-35B1* mRNA declines with age in wildtype hermaphrodites and *cyp-35B1* is a target of the *elt-3* GATA transcription factor that initiates and maintains intestinal cell fates in *C. elegans* [@pone.0017369-Budovskaya1]. Our microarrays showed that the *age-1*-dependent endocrine pathway could rescue *hsp-16* overexpression in *age-1* mutants. This observation led us to hypothesize that the *hsf-1* heat-shock transcription factor which regulates *hsp-16* expression and can collaborate with *daf-16*, might be important for expression of other endocrine *age-1* targets [@pone.0017369-Hsu1],[@pone.0017369-Morley1]. We therefore examined whether *hsf-1* RNAi was also required for *cyp-35B1* expression in response to lowered *age-1* activity. Using semi-quantitative RT-PCR, we examined transcript levels for *cyp-35B1* and *sod-3* in *daf-2(e1370)* adult hermaphrodites treated with *daf-16* or *hsf-1* RNAi or an empty vector RNAi control ([Fig. 2A](#pone-0017369-g002){ref-type="fig"}). As expected, *sod-3* overexpression in *daf-2(1370)* adults was suppressed by *daf-16* RNAi, consistent with previous evidence showing that *sod-3* is a direct DAF-16 target [@pone.0017369-Furuyama1]. Treatment with *daf-16* RNAi also reduced *cyp-35B1* mRNA in *daf-2(e1370)*, consistent with the identification of *cyp-35B1* as a target of the *daf-2* and *age-1* pathway ([@pone.0017369-Murphy1]; this work). In contrast, only *cyp-35B1* mRNA was significantly reduced in *daf-2(e1370)* animals treated with *hsf-1* RNAi, while *sod-3* levels were unaffected. This result demonstrates that *hsf-1* activity is required for induction of *cyp-35B1*, but not *sod-3*, in the absence of *daf-2* and *age-1* activity. ::: {#pone-0017369-g002 .fig} 10.1371/journal.pone.0017369.g002 Figure 2 ::: {.caption} ###### Regulation of *cyp-35B1* expression by *daf-16* and *hsf-1* in *daf-2(e1370)*. A. RT-PCR analysis of endogenous *cyp-35B1* and *sod-3* mRNA levels in *daf-2(e1370)* adults under control conditions or under *daf-16* or *hsf-1* RNAi. Results are average from three independent experiments. T-test was used to determine significance. *cyp-35B1* mRNA was significantly reduced by *daf-16* RNAi and *hsf-1* RNAi (*p* = 0.003 and 0.046, respectively). *sod-3* mRNA was significantly reduced by *daf-16* RNAi (*p*\<0.001) but not by *hsf-1* RNAi (*p* = 0.55). B. Expression of *cyp-35B1:gfp* transcriptional reporter was low or undetectable in wildtype non-dauer larvae (*i*) and adults (*iii*), but was expressed in the intestine of dauer larvae (*ii*) and *daf-2(e1370)* adults (*iv*). In *daf-2(e1370)* adults, *cyp-35B1:*GFP expression was abrogated by *daf-16* RNAi (*v*) and substantially reduced by *hsf-1* RNAi (*vi*). The spot of GFP fluorescence in the head reflects the transformation marker, *gcy-7:*GFP, expressed in the ASE/L neuron. Bars, 100 µm. In some lines, *hsf-1* RNAi resulted in increased or broader intestinal *cyp-35B1*:GFP expression. Since this phenotype was not recapitulated by the endogenous *cyp-35B1* mRNA (panel A), we attribute this observation to mysregulation of the *cyp-35B1:gfp* transgene under conditions of reduced *hsf-1* activity. :::  ::: We next examined the tissue distribution of *cyp-35B1* by constructing a transcriptional reporter expressing GFP from a *cyp-35B1* promoter. A previously-described *cyp-35B1:gfp* reporter was expressed in an *elt-3*-dependent manner, suggesting the possibility of intestinal expression [@pone.0017369-Budovskaya1]. We identified the *cyp-35B1* promoter as a 660-bp sequence between the *cyp-35B1* initiating methionine and the next upstream gene, *cyp-35A5* ([Fig. 2B](#pone-0017369-g002){ref-type="fig"}). During reproductive development and in wildtype adults, *cyp-35B1:*GFP was either not detectable or expressed at very low levels in posterior hypodermal cells ([Fig. 2B*i*, *iii*](#pone-0017369-g002){ref-type="fig"}). In dauers and *daf-2(e1370)* adults, *cyp-35B1:*GFP was upregulated and expressed solely in the intestine ([Fig. 2B*ii*, *iv*](#pone-0017369-g002){ref-type="fig"}). Intestinal *cyp-35B1*:GFP levels in *daf-2(e1370)* adults were substantially reduced by *daf-16* RNAi ([Fig. 2B*v*](#pone-0017369-g002){ref-type="fig"}). Treatment with *hsf-1* RNAi also reduced levels of *cyp-35B1*:GFP in *daf-2(e1370)* adults, although not to the same extent as for *daf-16* RNAi ([Fig. 2B*vi*](#pone-0017369-g002){ref-type="fig"}). These findings together demonstrate that *cyp-35B1* expression under conditions of low *daf-2* pathway activity is regulated by coordinate action of *daf-16* and *hsf-1*. In some lines, we observed increased or broader intestinal expression of *cyp-35B1*:GFP in *hsf-1* RNAi-treated animals. We attribute these changes to misregulation of the GFP reporter transgene, as we never observed increased levels of the endogenous *cyp-35B1* mRNA under *hsf-1* RNAi conditions. Direct regulation of intestinal *cyp-35B1* expression by DAF-16 and HSF-1 {#s2d} ------------------------------------------------------------------------- Based on our results, we hypothesized that *cyp-35B1* may be a direct or indirect target of DAF-16 and/or HSF-1. To test these possibilities, we studied the *cyp-35B1* promoter by deletion using *gfp* reporters. Deletion of 150-bp from the 5′ end of the 0.6-kb *cyp-35B1* promoter abolished intestinal GFP in dauers, without significantly affecting the weak hindgut expression observed in developing larvae ([Fig. 3](#pone-0017369-g003){ref-type="fig"}). Inspection of the deleted region identified two sequences (TTAAACA & AAAACA) resembling the previously-identified DAF-16 binding element (DBE, GTAAAC/TA) [@pone.0017369-Furuyama1]. This suggested that *cyp-35B1* might be a direct DAF-16 target. Smaller deletions were made from this 150-bp region removing one or both of the DBEs. Deletion of sequences containing one DBE reduced, but did not eliminate, *cyp-35B1:*GFP in dauers (pWBI076), while deletion of both DBEs eliminated dauer expression (pWBI077). The DBE-containing region was not sufficient for intestinal *cyp-35B1:*GFP expression in dauers, however. In addition, maximal *cyp35B1:*GFP expression required an adjacent 100-bp sequence which lacked any DBE-like sequences (pWBI072). ::: {#pone-0017369-g003 .fig} 10.1371/journal.pone.0017369.g003 Figure 3 ::: {.caption} ###### Identification of *cis*-regulatory modules (CRM) for dauer *cyp-35B1:gfp* expression by deletion analysis. Deletions in the *cyp-35B1* promoter were constructed to identify *cis*-regulatory modules necessary for directing dauer-specific expression in the intestine. (A) Diagram of *cyp-35B1* promoter constructs; promoters are black bars, brown hatches designate potential DAF-16 binding elements (DBE, [@pone.0017369-Furuyama1]), pink boxes designate *cyp-35B1* exon 1 sequence fused to GFP (green bar). (B) Representative fluorescence images of *cyp-35B1*:GFP expression in dauer larvae carrying transgenes described in panel (A); bar, 100 µm. (C) GFP expression in dauer fore-, mid- and hindgut from indicated transgenes; black bars indicate % of dauers with medium or bright GFP fluorescence. GFP expression was scored in 20--59 dauers for each construct. :::  ::: To test whether DAF-16 and/or HSF-1 directly bind to the promoter regions shown to be required for *cyp-35B1:*GFP expression in dauers, we utilized the yeast 1-hybrid assay for DNA-protein interactions [@pone.0017369-Li1], [@pone.0017369-Lehming1], [@pone.0017369-Kim1], [@pone.0017369-Schena1]. We first constructed yeast expression plasmids for the two preys we wished to test, DAF-16 and HSF-1, utilizing a copper-inducible yeast expression system to drive expression of full-length *daf-16* or *hsf-1* cDNAs [@pone.0017369-Schena2], [@pone.0017369-Shostak1]. Next, we constructed bait plasmids by inserting the 150-bp dauer CRM or a dispensable downstream region in front of a minimal promoter driving ß-galactosidase [@pone.0017369-Shostak1]. Four bait plasmids were tested, containing either the dauer CRM or the downstream region in the sense or antisense orientation with respect to the ß-galactosidase reporter (pWBI084/085 and pWBI082/083, respectively) ([Fig. 4A](#pone-0017369-g004){ref-type="fig"}). Reporter expression was activated by HSF-1 in yeast cells containing pWBI084, which contains the dauer CRM in the sense orientation with respect to the ß-galactosidase reporter. In the opposite orientation (pWBI085), ß-galactosidase expression was enhanced in cell expressing DAF-16. In contrast, we did not detect any stimulation of ß-galactosidase expression by the DAF-16 or HSF-1 preys in cells containing baits with promoter fragments that were dispensible for dauer *cyp-35B1:gfp* expression (pWBI082 or 083). ::: {#pone-0017369-g004 .fig} 10.1371/journal.pone.0017369.g004 Figure 4 ::: {.caption} ###### Yeast 1-hybrid analysis to detect DAF-16 and HSF-1 binding to *cyp-35B1* dauer CRM. Yeast 1-hybrid assays were performed to detect DAF-16 and HSF-1 binding to *cyp-35B1* dauer CRM (A, B) and subfragments (C, D). DAF-12 binding was also examined for the *cyp-35B1* dauer CRM (A, B), but produced negative results. (E) As a positive control, HSF-1 could bind to the heat-shock element (HSE)-containing region in the *hsp-16* promoter in pWBI113. In DAF-16- or HSF-1-expressing cells, transcription factor binding to the promoter fragments was measured as beta-galactosidase reporter activity in yeast cell extracts, normalized for cell density. Charts show average beta-galactosidase activity, relative to vector (pRS424) controls. Number of trials: pWBI082, pWBI083, 4 trials; pWBI084, 9 trials; pWBI085, 8 trials; pWBI124, 6 trials; pWBI127, 5 trials; pWBI129, 6 trials; pWBI113, 2 trials. Statistical significance was determined by t-test; \**p* = 0.05; \*\**p* = 0.01. Arrows under plasmid names depict the number and orientation of the *cyp-35B1* promoter fragments within the 1-hybrid reporter constructs. (F) Sequences of the *cyp-35B1* promoter fragments within pWBI127 and pWBI129. Highlighted sequence in pWBI127 fragment depicts a possible DAF-16 binding site; highlighted sequence in pWBI129 fragment show possible HSF binding site containing imperfect inverted repeats of 5′-AGAAN-3″ sequences. :::  ::: To further delineate the binding regions for DAF-16 and HSF-1 binding within the promoter fragments in pWBI084/085, we subdivided this 160-bp fragment into three overlapping fragments of 60--65-bp each. In yeast expressing DAF-16, ß-galactosidase expression was stimulated in the presence of the pWBI127 bait plasmid, which contains the *cyp-35B1* promoter fragment with the 2^nd^ DBE ([Fig. 4C, D](#pone-0017369-g004){ref-type="fig"}). The adjacent promoter fragment, carried in pWBI129, stimulated beta-galactosidase reporter expression in yeast expressing HSF-1. HSF transcription factors bind as trimers to the heat shock element (HSE) 5′-AGAANNTTCTAGAAN-3′, consisting of three inverted repeats of the 5′-AGAAN-3′ monomeric sequence. Inspection of the *cyp-35B1* promoter element contained in pWBI129 identified a sequence on the bottom strand which resembles this consensus (5′-GAAcaTgCctGAAttgaCgtGAA-3′), but may be an imperfect inverted pentamer repeat of the 5′-AGAAN-3′ sequence ([Fig. 4F](#pone-0017369-g004){ref-type="fig"}). The presence of an HSE-like sequence in pWBI129 is consistent with the observation of HSF-1 binding in the 1-hybrid assay. Discussion {#s3} ========== The goal of this study was to identify potential endocrine targets of *age-1* activity that could be regulated non-autonomously from the nervous system and/or the intestine. Expression of wildtype *age-1* within neurons or intestinal cells rescues dauer arrest and lifespan phenotypes of *age-1(mg44)* animals [@pone.0017369-Wolkow1], [@pone.0017369-Iser1]. This evidence led to a working model whereby *age-1* activity within signaling tissues regulates an endocrine output that, in turn, can direct dauer morphogenesis and aging in target tissues [@pone.0017369-Iser1]. The major genetic target of *age-1* is *daf-16*, encoding a FOXO transcription factor [@pone.0017369-Ogg1]. All available evidence indicates that the major mechanism for AGE-1/PI3K regulation of DAF-16 is cell-autonomous, via phosphorylation by AKT kinases regulated by AGE-1/PI3K phospholipid products [@pone.0017369-Lin2], [@pone.0017369-Paradis1], [@pone.0017369-Lee1]. Thus, the finding that *daf-16* acts cell-autonomously in the intestine to promote longevity was an apparent contradiction to earlier mosaic and transgenic analyses showing that the upstream regulators, *daf-2* and *age-1*, regulate these processes non-autonomously [@pone.0017369-Libina1]. To resolve this conflict, we proposed that *age-1* can regulate *daf-16* activity in target tissues through convergent cell-autonomous and non-autonomous mechanisms [@pone.0017369-Iser1]. To investigate possible effectors through which *age-1* might non-cell autonomously regulate *daf-16*, we used transcriptional microarrays to identify *age-1* target genes that could be non-autonomously regulated by *age-1*. This search identified 127 potential non-autonomous *age-1* target genes. We found that one of these, *cyp-35B1/dod-13*, which is expressed in the intestine of dauer larvae and *daf-2(e1370)* adults, could be directly regulated by both HSF-1 and DAF-16. Furthermore, other HSF-1 target genes, the *hsp-16* genes, were non-autonomously regulated by *age-1*. Expression of a subset of *daf-2* target genes is known to be *hsf-1*-dependent, although HSF-1 has not been shown to directly regulate these targets [@pone.0017369-Hsu1]. These observations provide circumstantial evidence placing *hsf-1* in the pathway for *daf-2* and *age-1* non-autonomy. We propose that DAF-2/IR and AGE-1/PI3K signaling in non-intestinal cells regulates an endocrine output that affects HSF-1 activity in intestinal cells ([Fig. 5](#pone-0017369-g005){ref-type="fig"}). We hypothesize that HSF-1 and DAF-16 can interact within intestinal cells to optimize expression of HSF-1 and DAF-16 target genes that may extend lifespan. An alternative model is that HSF-1 and DAF-16 function independently to promote expression of prolongevity target genes in intestinal cells. Distinguishing between these possibilities will likely require characterization of other non-autonomous *age-1* targets and clearer descriptions of the interactions between *hsf-1* and *daf-16* in the regulation of *C. elegans* dauer arrest and longevity. ::: {#pone-0017369-g005 .fig} 10.1371/journal.pone.0017369.g005 Figure 5 ::: {.caption} ###### Regulation of intestinal *cyp-35B1/dod-13* expression by DAF-16 and HSF-1. Two intestinal DAF-16 target genes, *sod-3* and *cyp-35B1/dod-13*, are differentially regulated by HSF-1. Yeast 1-hybrid and promoter deletion mapping analyses suggest that DAF-16 and HSF-1 directly bind the *cyp-35B1* promoter. Quantitation of mRNA showed that *hsf-1* activity was required for *cyp-35B1* induction in dauers, but *hsf-1* was dispensible for *sod-3* induction. Microarray data indicated that *cyp-35B1* expression in *age-1* mutants could be regulated non-autonomously by *age-1* expression in neuronal cells (CY251), as well as cell-autonomously by *age-1* expression in the intestine (CY262). Thus, *cyp-35B1* is a possible endocrine target for *age-1* activity in neuronal cells. We propose that the *age-1* pathway may non-autonomously couple to *hsf-1* activity in intestinal cells to regulate *cyp-35B1* and other targets. :::  ::: Materials and Methods {#s4} ===================== Strains and growth {#s4a} ------------------ *C. elegans* strains were maintained at 15°C on NG agar medium with nystatin and streptomycin with OP50 bacteria as a food source [@pone.0017369-Brenner1]. The following strains were used: Bristol N2 (wildtype); SP75 (*sqt-1(sc13) age-1(mg44)/mnC1*); CY251 (*sqt-1(sc13) age-1(mg44)*; *bvIs2*); CY262 (*sqt-1(sc13) age-1(mg44)*; *bvIs1*); CY312 (*daf-16(mgDf50)*; *daf-2(e1370)*); DR40 (*daf-1(m40)*). *bvIs1* and *bvIs2* were previously described [@pone.0017369-Iser1]. Microarray analysis {#s4b} ------------------- Gene expression in CY262 and CY251 was compared to non-transgenic *age-1(mg44) m+z-*adults of similar age, and all comparisons were in reference to wildtype young adults. Synchronized populations of animals at the late-L4/young-adult stage were obtained from embryos isolated by bleach treating gravid hermaphrodites. Embryos were hatched overnight in S medium without food, causing developmental arrest at the first larval (L1) stage. Arrested L1s were plated with food, grown for 72 hours at 20°C and washed in M9. Total RNA was isolated with Absolutely RNA miniprep kit (Stratagene, La Jolla, CA). cDNA was labeled with the Quick Amp 2 Color Labeling Kit (Agilent Technologies, Santa Clara, CA) and hybridized to Agilent 2×22 k oligo microarrays containing probes for nearly the complete *C. elegans* genome. Slides were scanned using the Agilent Microarray Scanner (G2565B). Three biological replicates were examined per strain. After hybridization, probes with raw fluorescence signals \>100 in wildtype samples were selected for further analysis as being reliable signals. Expression ratios for each gene were calculated in all 3 strains with respect to wildtype controls at similar, or slightly later developmental stage. Expression ratios were calculated as z-scores for a statistical evaluation of relative expression, and as fold-changes for standardized displays of relative expression [@pone.0017369-Cheadle1]. To determine fold change, raw fluorescence values were first normalized by dividing the fluorescence intensity for a given sample by the average intensity of all samples for the array. Fold change ratios were then determined by dividing mean normalized value for all replicates of test versus control conditions. Statistical significance was judged using t-test to compare signal intensity measurements among biological replicates. Two replicates were performed for *age-1(mg44)* and 3 for CY251, CY262 and N2. MIAME-compliant raw microarray data is available from the NCBI Gene Expression Omnibus (Accession \#GSE18200). *cyp-35B1:gfp* reporter {#s4c} ----------------------- The *cyp-35B1* promoter was considered to be the sequences between the predicted translational start to the immediate upstream gene and was PCR-amplified from *C. elegans* genomic DNA using primers with unique restriction sites. Primer sequences were 5′-CAACAGAGGAGACAATGCCG-3′ and 5′-GGAAGAGAAACAGGTCCTGGTGGG-3′. PCR products were purified, digested and ligated into predigested pPD95.75, which contains GFP and the *unc-54* 3′UTR (Addgene, Cambridge, MA). Constructs were confirmed by DNA sequencing. Transgenic animals were created by microinjection transformation (100 ng/µl plasmid DNA with 50 ng/µl of the co-injection marker *gcy-7:gfp*). Stable transmitting lines were selected in the 2nd generation and analyzed. For *cyp-35B1* promoter deletions, 150-bp and 300-bp deletions were made in the *cyp-35B1* promoter in pWBI036, to create pWBI066 and pWBI067, respectively, by PCR using upstream primers that hybridized to internal sites within the cloned promoter. The PCR primers also contained unique BamHI and KpnI restriction sites, which were used to insert the PCR fragments into pPD95.75. To further define the region responsible for dauer expression, smaller deletions were made from the region deleted in pWBI066, using the same PCR strategy, to make pWBI074 (59-bp deletion), pWBI075 (109-bp deletion), pWBI076 (159-bp deletion) and pWBI077 (209-bp deletion). Next, a 161-bp fragment containing both putative DAF-16 binding sites was PCR amplified with primers containing unique BamHI and NheI restriction sites and the digested product was ligated to a basal promoter in pWBI067, consisting of 398-bp upstream from the translational start, creating pWBI072. GFP expression was compared in temperature-induced *daf-2(e1370)* dauers, starvation-induced wildtype dauers and *daf-2(e1370)* adults raised at 15°C and then held at 25°C for 24 hours. For microscopy, animals were either mounted on a 2% agar pad with levamisole and photographed with a Hamamatsu CCD camera (Hamamatsu Photonics, Japan) and OpenLab software (Improvision Inc., USA) on a Nikon E900 microscope (Nikon Corporation, Japan) or transferred onto an NGM plate spread with 20 µL of 10% sodium azide and photographed on a Nikon SMZ1500 stereodissecting microscope using a SPOT RT3 Slider camera and SPOT Advanced software (Diagnostic Instruments, USA). All images for a particular GFP reporter were collected using identical exposure times. For RNAi experiments, populations were obtained from synchronized egg lay on plates seeded with dsRNA-expressing bacteria or control bacteria. Larvae were raised for at 25°C (4 days) for dauers, or 15°C (5 days) for adults that were subsequently transferred to 25°C for 24 hours before examination. Yeast 1-hybrid assays {#s4d} --------------------- Plasmids for yeast 1-hybrid assays were kindly provided by K. Yamamoto (UCSF). pYSYE0002 is a prey plasmid containing the DAF-12 N500 gene expressed from the copper inducible *cup-1* promoter and pYSYR0002 is a bait plasmid contain a DAF-12 responsive element cloned upstream of P*cyc-1* driving lacZ expression [@pone.0017369-Shostak1]. To construct the *daf-16* and *hsf-1* prey vectors, the *daf-16* or *hsf-1* cDNAs were amplified with primers containing Xma I and Not I restriction sites. PCR fragments were cloned into corresponding restriction sites in the pYSYE0002, and the resulting expression clones, pWBI109 and pWBI108, were sequence confirmed. Bait plasmids were constructed containing *cyp-35B1* promoter fragments were constructed using PCR-amplified promoter fragments with added NotI restriction sites. These PCR products were inserted in the place of the DAF-12 responsive element in pYSYR0002 and the resulting plasmids were sequence confirmed. Yeast were transformed with both a bait and prey plasmid, or negative controls, using the Yeastmaker Yeast Transformer Kit (Clontech, USA) with selection on SC medium lack tryptophan and uracil. For ß-galactosidase assays, independent yeast colonies were inoculated into 3-mL of selective medium and grown overnight at 30°C. Prey protein expression was induced by inoculation with 100 µM CuSO4 for 4--5 hours at 30°C. ß-galactosidase activity was assayed from pelleted cells using the Yeast ß-Galactosidase Assay Kit (Thermo Scientific, USA). The parent vector for pYSYE0002 is pRS424, which was used as a negative control for bait auto-induction. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### mRNA expression levels of *daf-2* pathway class 1 and 2 target genes in *age-1* mutants with tissue-restricted *age-1* expression. Expression levels for class 1 and 2 *daf-2* pathway targets (Murphy et al. 2003) were obtained from microarray data for *age-1(mg44)*, CY262 and CY251 relative to wildtype. Relative expression levels are presented as fold-changed relative to wildtype. "Missing from array?", genes which were absent from the microarrays used in this study; "Our results", indicates whether previous target classifications were consistent in the current study; "Rescue category", indicates whether expression was rescued to wildtype or near-wildtype levels in CY262 and/or CY251 for targets that were congruent between the prior and current study; "Not rescued" indicates targets that maintained the mutant level of expression in CY262 and CY251; "Both" indicates targets whose expression was rescued to wildtype or near-wildtype levels in both CY251 and CY262. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### mRNA and anatomical expression data for genes upregulated by ≥2-fold in *age-1(mg44)* adults relative to wildtype. Expression data, as fold-change relative to wildtype and *age-1(mg44)*, is shown for each strain (*age-1(mg44)*, CY262 and CY251). "262 rescue", "251 rescue" and "Category" indicate whether upregulation in *age-1(mg44)* was determined to be rescued in CY262 or CY251 and the corresponding rescue category (262, 251, Both, None). Anatomical expression data were obtained from the curated expression pattern annotations for each gene in WormBase ([www.wormbase.org](http://www.wormbase.org)). (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### 10.1371/journal.pone.0017369.s003 Expression of stress response genes in *age-1(mg44)*, CY262 and CY251. mRNA expression levels were obtained from microarray data for genes annotated as heat-shock protein, glutathione S-transferase, catalase, superoxide dismutase, lysozyme or metallothionein. Expression levels in *age-1(mg44)*, CY262 and CY251 are shown as fold-change relative to wildtype. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to the following colleagues for providing reagents: Andrew Fire (Stanford University), Scott Alper (NIEHS/Duke), Patrick Hu (University of Michigan), Curtis Loer (USD) and Keith Yamamoto (UCSF). We thank Shailaja Kishan Rao for assistance with the microarray experiments. The *Caenorhabditis Genetics Center* and the Genome BC *C. elegans* Gene Expression Consortium provided several *C. elegans* strains [@pone.0017369-Garigan1], [@pone.0017369-Brenner1]. We thank Mark Mattson for helpful suggestions and the members of the Wolkow lab for critical review of the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was carried out with funding from the NIA Intramural Research Program (AG000320) and from an Ellison Medical Foundation New Scholar in Aging award ([www.ellisonfoundation.org](http://www.ellisonfoundation.org)). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: WBI MAW KB CAW. Performed the experiments: WBI MAW WHW CAW. Analyzed the data: WBI MAW CAW. Contributed reagents/materials/analysis tools: WBI MAW WHW KB CAW. Wrote the paper: CAW.

PubMed Central

2024-06-05T04:04:19.125147

2011-3-9

PMC3052306

Introduction {#s1} ============ The species composition of ecological communities is as varied as the biophysical conditions where they occur. Accordingly, there is a prevailing sentiment that a general understanding of mechanisms leading to patterns in communities will be difficult if not impossible because communities are riddled by complexity, context-dependency and idiosyncrasy. It is therefore quite remarkable that a comparatively simple species abundance distribution (SAD) model ably describes pattern in widely divergent communities, making it one of ecology\'s most enduring tenets. Put simply, the rank abundance of constituent species is dominated by many rare and a few highly abundant species, regardless of community type [@pone.0017342-Hubbell1], [@pone.0017342-McGill1], [@pone.0017342-Magurran1]. Given the generality of the pattern, dozens of statistical models have been fit to SAD data to identify the elusive "silver-bullet" mechanism(s) driving the pattern [@pone.0017342-McGill1], [@pone.0017342-Magurran2], [@pone.0017342-Preston1], [@pone.0017342-Williams1], [@pone.0017342-Whittaker1], [@pone.0017342-MacArthur1], [@pone.0017342-Magurran3]. Intense debate continues regarding which models fit best, how goodness-of-fit is measured and how to interpret successful or failed fits [@pone.0017342-McGill1]. Indeed, the use of SAD patterns to explain ecological patterns and evaluate theory is increasing ([Fig. 1](#pone-0017342-g001){ref-type="fig"}). The most recent incarnation of the debate centers on whether or not the unified neutral theory (UNT) [@pone.0017342-Hubbell1] sufficiently explains the pattern without the need to invoke non-neutral mechanisms. ::: {#pone-0017342-g001 .fig} 10.1371/journal.pone.0017342.g001 Figure 1 ::: {.caption} ###### ISI\'s Web of Science reports a considerable increase in scientific articles containing the search terms "species abundance distribution" or "relative species abundance." This increase represents the burgeoning debate about how ecological communities are structured. Shown is the percentage of articles with the search terms relative to all articles in the Biology and Ecology subject areas. :::  ::: We argue here that the ecological processes that structure natural communities cannot be determined by fitting models to SADs alone because they are generated from observational data, for which the underlying mechanisms are unknown [@pone.0017342-Clark1], [@pone.0017342-Leigh1]; That is, there is not a benchmark to know what pattern the mechanism should produce. SAD patterns may instead merely reflect the vagaries of sampling and statistical properties of data [@pone.0017342-McGill1], [@pone.0017342-Magurran3], [@pone.0017342-Nekola1], [@pone.0017342-May1] that may lead to spurious conclusions about underlying mechanisms. Indeed, researchers have shown that the methods used to test hypotheses and generate SAD patterns can produce similar patterns from non-ecological, apparently random, data [@pone.0017342-Nekola1], [@pone.0017342-Nee1]. These findings suggest that a universal mechanism structures ecological and non-ecological patterns, with the underlying mechanism unknown in both cases, or that the pattern is unrelated to underlying mechanisms. Methods {#s2} ======= Whereas previous researchers have generated SAD patterns from non-ecological data to illustrate SAD shortcomings, these have incorporated processes with mechanisms as cryptic as those in ecological systems (e.g., SAD patterns in stock prices) [@pone.0017342-Nekola1]. Instead, we ask whether the SAD pattern can be generated from a data set structured by a known mechanism -- not to infer the mechanics underlying ecological processes, but to examine the potential for SAD analysis to elucidate them. We do this by using the distribution of wins in college basketball games where the mechanism, "competitive exclusion," is understood [@pone.0017342-Miller1]. Competition in college basketball evolves from selection at all levels of organization. Universities invest heavily in salaries and facilities to attract top coaches and players, coaches invest long hours into rigorous recruiting and training the best players, and players invest many years toward improving skills and athleticism [@pone.0017342-Fizel1]. The end result is a community of teams with competitive edge skewed toward a few dominant teams that consistently win in head-to-head competition [@pone.0017342-Miller1], [@pone.0017342-Staudohar1]. Historically strong teams remain strong, whereas -- with some variance -- smaller schools in smaller conferences remain weak. Head-to-head competition structures the win-loss records of these teams, which play more often at regional scales with less common long-distance games. This competition creates a win-loss data set for which we know more about the structuring mechanism than for ecological community data, or for previous non-ecological data sets used to criticize SADs. For ecological analogy, we treat each team as a species and each win as an individual of that species occupying a site. We explore the patterning that emerges and relate it to classic SADs. College basketball provides little insight into ecological processes, but it does provide an intuitive framework to examine the universality of the SAD pattern and its connection with a known structuring mechanism. We analyze win-loss records for 327 NCAA Division I teams (years 2004--2008 for statistical replication). We consider each team a species, and each win an individual (total wins equals species abundance). College basketball data are consistent with assumptions outlined for the UNT [@pone.0017342-Hubbell1]. They follow a zero-sum gain as a win (*n*+1) by one team results in loss (*n*--1) by another (i.e. gain of an individual by one species results in loss of another species individual; and in college basketball teams cannot 'draw' a game). Further, there is a high species (team) richness and a high number of individuals (∼5,000 wins yr^−1^) competing on a single 'trophic' level. We rank abundance of wins per team (2004--2008, mean ±95% CI) creating a relative abundance distribution ([Fig. 2](#pone-0017342-g002){ref-type="fig"}). This is the classic method for empirically representing commonness and rarity in communities. We fit both the empirical and random data sets to a sigmoid curve using the nls() package [@pone.0017342-Bates1] in the R statistical program [@pone.0017342-R1]. ::: {#pone-0017342-g002 .fig} 10.1371/journal.pone.0017342.g002 Figure 2 ::: {.caption} ###### Rank abundance of college basketball wins by team. The abundance of wins in college basketball, a result of competition between teams of unequal abilities, creates the same pattern used by ecologists to infer mechanism from species abundance distributions (SADs). The log~10~ abundance of college basketball wins is ranked by team, just as the abundance of individuals is ranked by species for ecological communities. Mean wins (gray) across 2004 to 2008±95% CI are given along with random (*Normal*, µ = 16, σ = 6) wins (black), and these random and observed patterns are not significantly different (see text). :::  ::: Results and Discussion {#s3} ====================== The rank abundance distribution shows a classic left-skewed pattern interpretable as a community characterized by few abundant species and many rarer species ([Fig. 2](#pone-0017342-g002){ref-type="fig"}). We generate the same pattern (goodness-of-fit, *r* ^2^ = 0.99) using five random season of college basketball wins ([Fig. 2](#pone-0017342-g002){ref-type="fig"}). The estimated asymptote (asym), inflection point (xmid) and curve steepness (scal) for the empirical and random data were a significant (p\<0.001) fit to a sigmoid curve (asym/1+exp(-(xmid-data)/scal)). More importantly, the estimated parameters (±SE) for the empirical (asym: 27.3±0.3; xmid: 5.2±0.02; and scal: −0.7±0.02) and random (asym: 26.2±0.3; xmid: 5.3±0.02; and scal: −0.6±0.02) data did not differ significantly ([Fig. 2](#pone-0017342-g002){ref-type="fig"}). These results demonstrate that a non-ecological dataset (college basketball) with a known mechanism (competition), and where there is also some 'stochasticity' (i.e. the favorite does not always win), generates a pattern purported to arise for communities from underlying ecological processes [@pone.0017342-Magurran3], [@pone.0017342-Nekola1], [@pone.0017342-McGill2]. Moreover, randomly generated data produce the same pattern ([Fig. 2](#pone-0017342-g002){ref-type="fig"}). We can draw two important conclusions from these results. First, fitting niche or neutral models to this pattern -- or to deviations from this pattern in a null framework -- cannot deduce mechanism because even data with a known mechanism does not produce a SAD pattern that deviates from random. SAD patterns may be a universal product of large data sets and sampling artifacts, and this means they cannot truly be falsified -- making this approach uncertain for hypothesis testing and model fitting. Second, we know that mechanism matters in college basketball as the powerhouse teams from the top conferences typically dominate, and the top teams are predictable based on their "traits". In college basketball, such traits include the athletic department budget, the facilities, the coach\'s salary and the ability to attract top recruits [@pone.0017342-Miller1], [@pone.0017342-VanRiper1]. These traits are unequally distributed toward a few dominant teams, and these teams achieve a disproportionate number of wins. Yet, the outcome is a SAD pattern indistinguishable from random wins and most ecological communities. These results beg the question: what processes underlie patterns of species distributions where more biological complexity occurs? Neutral [@pone.0017342-Hubbell1] and niche [@pone.0017342-Clark1] based approaches are considered alternative theories in community ecology (but see [@pone.0017342-Adler1] for reconciliation), particularly because the universal nature of SAD patterning suggests that invoking niche differentiation is unnecessary in explaining community structuring [@pone.0017342-Hubbell1]. The frequency distribution of species reflects numerical abundance, but assumes all species utilize resources similarly, share the same body size and interact equally [@pone.0017342-MacArthur2], an assumption consistent with unified neutral theory [@pone.0017342-Hubbell2], but commonly violated in natural communities [@pone.0017342-Morlon1]. Instead, it appears that the rule of large numbers, as noted by May [@pone.0017342-May1], generates the SAD patterns for ecological communities and college basketball records. As we noted, college basketball teams do not share and utilize resources equally, and there is no reason to assume members of ecological communities do. Likewise, SAD patterns have been used as evidence of niche partitioning [@pone.0017342-MacArthur1], [@pone.0017342-Tokeshi1], but if they are generated by combinations of complex factors in large data sets -- or even many small random effects [@pone.0017342-Limpert1] -- rather than underlying biological mechanisms, they also provide no falsifiable evidence of niche apportionment. For example, the significant-digit (aka Benford\'s) law stipulates that the first digit of non-random data sets with numbers that span several orders of magnitude are biased toward lower values [@pone.0017342-Newcomb1], [@pone.0017342-Benford1]. As a result, data sets ranging from sports statistics to river size usually contain numbers that predominately begin with 1 (30%), followed by numbers that begin with 2 (18%), 3 (13%) down to numbers beginning with 9 (5%) \[proportion digit~x~ = log~10~ (digit~x+1~/digit~x~)\] [@pone.0017342-Benford1], [@pone.0017342-Hill1]. This prompted us to investigate the concordance between an empirical species abundance data set, 319 tropical trees \>10 cm DBH at Barro Colorado Island [@pone.0017342-Hubbell3], and Benford\'s law. A chi-square test examining the difference between observed species abundance digit distribution and the expected Benford distribution shows that the BCI data follow the significant-digit pattern (*χ^2^* = 3.22, *df* = 8, *p* = 0.920). Whereas Newcomb and Benford based their findings on empirical observations [@pone.0017342-Newcomb1], [@pone.0017342-Benford1], Hill [@pone.0017342-Hill1] offers a theoretical basis for the pattern. Essentially, the greater the complexity of interacting processes underlying a data set, the more the first digits converge to a logarithmic distribution as described by Benford\'s law [@pone.0017342-Hill1]. That BCI tree species abundance follows Benford\'s law indicates that the pattern may reflect multiple and complex ecological mechanisms, and this possibility further undermines SAD usefulness as substantiation for overarching theories. Ecological stalwarts such as MacArthur [@pone.0017342-MacArthur3] and May [@pone.0017342-May1] long ago questioned the use of SAD patterns in ecological analysis, even going so far to call it an "obsolete approach to community ecology" [@pone.0017342-MacArthur3]. Substantial evidence suggests the SAD pattern represents statistical and sampling artifacts equally as well as any structuring mechanism [@pone.0017342-Nekola1], [@pone.0017342-Nee1], [@pone.0017342-McGill2], making their ecological validity difficult to assess ([Fig. 2](#pone-0017342-g002){ref-type="fig"}). Whilst we do not provide nor posit a proof that SAD fitting fails to adequately represent ecological communities, we provide sufficient evidence that SAD patterns may (1) derivate from purely statistical or sampling processes and/or (2) oversimplify and obfuscate complex ecological dynamics. The escalation in the use and analysis of SAD patterns may represent a substitution of statistical elegance for ecological relevance. After almost 80 years of attempts to explain SADs, with equivocal results [@pone.0017342-McGill1], [@pone.0017342-Magurran3], this pattern fitting persists within the ecological milieu and even has increased in recent decades ([Fig. 1](#pone-0017342-g001){ref-type="fig"}), e.g., [@pone.0017342-Rosindell1]. Our findings do not resolve niche vs. neutral debates, nor do we shed light on the mechanisms underlying ecological processes that structure communities, but we do highlight the critical need for field research rather than SAD patterning to test competing hypotheses explaining community patterns [@pone.0017342-Magurran3], [@pone.0017342-Clark1], [@pone.0017342-Clark2]. This may require acceptance that ecological systems are cryptic and complex and not easily synthesized to fit simple overarching models [@pone.0017342-Clark1], [@pone.0017342-Clark2]. This approach requires improved integration of empirical and theoretical ecology with direct experimental evidence of putative structuring mechanisms to evaluate the niche and/or neutral processes structuring ecological communities. We thank Volker Bahn and Jeffrey Lake for comments on initial versions of this project, Thomas Bell and an anonymous reviewer for helpful comments on this manuscript, and Center for Tropical Forest Science of the Smithsonian Tropical Research Institute for making available the BCI data. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by funds from the National Science Foundation (award DEB-0823293) to the Coweeta Long Term Ecological Research Program and by the Yale School of Forestry and Environmental Studies. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: RJW MAB. Performed the experiments: RJW. Analyzed the data: RJW DKS OJS MAB. Contributed reagents/materials/analysis tools: RJW. Wrote the paper: RJW.

PubMed Central

2024-06-05T04:04:19.128617

2011-3-9

PMC3052307

Introduction {#s1} ============ Ubiquitin (Ub) is a small protein that consists of 76 amino acids about 8.5 kDa. Ubiquitin conjugation sites of protein (Ubiquitylation), which is an essential post-translational modification, is a sequential process that involves a group of enzymes known as E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase). The ubiquitylation system is well-known for the selective degradation of serveral short-lived proteins in eukaryotic cells [@pone.0017331-Hershko1]. The attachment of a ubiquitin or poly-ubiquitin chains to proteins influences serveral cellular processes, including transcriptional regulation, signal transduction, development, apoptosis, endocytosis, and cell proliferation [@pone.0017331-Ou1]. Ubiquitin is mostly conjugated on the lysine residue of a protein by Ub-ligating (E3) enzymes. The E3 ligase must be sufficiently specific and must act only on a defined subset of cellular targets to ensure signal fidelity [@pone.0017331-Hicke1]. Another enzyme, E4, has that can stabilize and extend a poly-ubiquitin chain, has also been found [@pone.0017331-Gilon1]. With the development of high-throughput tandem mass spectrometry-based proteomics, the number of studies of the comprehensive identification ubiquitylated proteins and their conjugated sites is increasing [@pone.0017331-Peng1]. UbiProt [@pone.0017331-Chernorudskiy1] identified all experimentally verified ubiquitin-conjugated sites from the publicly literature. Some of the entries include information about enzyme data obtained by enzyme purification and isolation. The entries supply annotations of the ubiquitin-conjugated and the exact positions of the ubiquitin-conjugation sites. UniProtKB/Swiss-Prot [@pone.0017331-Boeckmann1] is a comprehensively annotated protein database. Both experimentally validated and putative ubiquitin-conjugated annotations can be obtained from the post-translation modification annotations in the database. Experimental identifications of ubiquitin-conjugation sites on ubiquitylated proteins *in vivo* and *in vitro* are the foundation for understanding the mechanisms of ubiquitination dynamics. However, these experiments are commonly time-consuming, labor-intensive and expensive. the *in silico* prediction of ubiquitin-conjugated sites with high predictive performance could be promising for preliminary analyses and could greatly reduce the number of potential targets that require further *in vivo* or *in vitro* confirmation. UbiPred [@pone.0017331-Tung1] used an algorithm for mining informative physicochemical properties from protein sequences to train SVM-based ubiquitylation site prediction system. Based on leave-one-out cross-validation, the SVM model that is trained with 31 physicochemical properties was evaluated. It was found to improve the predictive accuracy from 72.1% to 84.4%. Recently, Radivojac *et al.* [@pone.0017331-Radivojac1] have investigated that the sequence biases and structural preferences around known ubiquitination sites are similar to those of intrinsically disordered protein regions. Additionally, Radivojac *et al.* developed a random forest predictor of ubiquitination sites, UbPred, that could reach a balanced accuracy of 72%. Given the importance of ubiquitin conjugation in biological processes, this investigation presents a method, UbSite, in which an efficient radial basis function (RBF) network is utilized to identify protein ubiquitin conjugation (ubiquitylation) sites. The experimentally verified ubiquitylated proteins and ubiquitylation sites are collected from UbiProt [@pone.0017331-Chernorudskiy1] and UniProtKB/Swiss-Prot [@pone.0017331-Boeckmann1]. Not only amino acid composition, but also structural characteristics, physicochemical properties, and evolutionary information of amino acids around the ubiquitylation (Ub) sites are investigated. With reference to the pathway of ubiquitin conjugation, which is by sequential process that involves a group of enzymes, E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase), the distant sequence features of ubiquitylation sites for E3 recognition are investigated. A position specific scoring matrix (PSSM), which is generated by PSI-BLAST [@pone.0017331-Altschul1] search against a non-redundant database of protein sequences, is utilized to study the evolutionary information surround the ubiquitin conjugation sites. The constructed PSSM is regarded as a measure of residue conservation in a window of a given length. Based on the measurement of F-score in a large window size (−20∼+20), the statistically significant amino acid composition and evolutionary information, which are mainly located at positions distant from the ubiquitylation sites, can be utilized effectively to differentiate ubiquitylation sites from non-ubiquitylation sites. An evaluation of the trained models based on five-fold cross-validation revealed, that the prediction sensitivity, specificity and accuracy were 65.5%, 74.8%, and 74.5%, respectively. The independent test demonstrates shows that UbSite outperform previous ubiquitylation prediction tools. Materials and Methods {#s2} ===================== As shown in [Fig. 1](#pone-0017331-g001){ref-type="fig"}, the proposed approach, UbSite, is composed of three major analytical steps - data collection and preprocessing, feature extraction, and model training and evaluation. This investigation comprehensively analyzes the structural characteristics and physicochemical properties that surround the ubiquitin conjugation sites. The details of the analysis are described as follows. ::: {#pone-0017331-g001 .fig} 10.1371/journal.pone.0017331.g001 Figure 1 ::: {.caption} ###### The analytic flowchart of UbSite. :::  ::: Data collection and preprocessing {#s2a} --------------------------------- Experimentally confirmed ubiquitin conjugation sites are collected from UbiProt [@pone.0017331-Chernorudskiy1] and UniProtKB/Swiss-Prot [@pone.0017331-Boeckmann1]. UbiProt consists of 158 experimentally confirmed ubiquitin-conjugation sites. Then, we extracted the sequences from release 57.0 of UniProtKB/Swiss-Prot if the sequences are annotated as 'ubiquitin' in the 'MOD\_RES' fields. We also removed the sites that are annotated as "by similarity", "potential" or "probable". A total of 337 entries are annotated as ubiquitin-conjugated proteins in UniProtKB/Swiss-Prot, and they include 416 ubiquitylation sites. After removing the redundant data from UbiProt and UniProtKB/Swiss-Prot, a total of 442 experimental ubiquitylation sites associated with 350 ubiquitylated proteins are obtained. In this work, the 442 experimental ubiquitylation sites are regarded as the positive dataset. To prevent overestimation of the predictive performance, homologous sequences are removed from the training data by using a window size of 2*n*+1 for ubiquitylation sites. With reference to the reduction process in MASA [@pone.0017331-Shien1], two ubiquitylated protein sequences with more than 30% identity were defined as homologous sequences. Then, two homologous sequences were specified to re-align the fragment sequences using a window length of 2*n*+1, centered on the ubiquitylation sites using BL2SEQ [@pone.0017331-Tatusova1]. For two fragment sequences with 100% identity, when the ubiquitylation sites in the two proteins are in the same positions, only one site was kept. The homologous negative data were also reduced by using the same approach. With respect to classification, the predictive performance of the trained models may be overestimated because of the over-fitting of a training set. To estimate the real predictive performance, the experimental ubiquitylation sites, whose annotated dates are after April 4 2006, are selected as the independent test set. As shown in [Table 1](#pone-0017331-t001){ref-type="table"}, the data in the non-homologous training set include 385 positive sites (ubiquitylation) and 12582 negative sites (non-ubiquitylation) in 301 ubiquitylated proteins. The data of the non-homologous independent test set include 57 positive sites and 3502 negative sites in 49 ubiquitylated proteins. Following the evaluation by five-fold cross-validation, the trained model with the highest accuracy was further evaluated based on independent test data. The independent test sets were utilized to test not only the proposed method but also the previously proposed ubiquitylation prediction tools, UbiPred [@pone.0017331-Tung1] and UbPred [@pone.0017331-Radivojac1]. ::: {#pone-0017331-t001 .table-wrap} 10.1371/journal.pone.0017331.t001 Table 1 ::: {.caption} ###### The statistics of non-homologous training data and independent test data for ubiquitylation and non-ubiquitylation sites. ::: {#pone-0017331-t001-1} Training data[a](#nt101){ref-type="table-fn"} Independent data[b](#nt102){ref-type="table-fn"} ------------------------------------- ----------------------------------------------- -------------------------------------------------- Number of proteins 301 49 Number of ubiquitylated lysines 385 57 Number of non-ubiquitylated lysines 12,582 3,502 a Training data: the annotation date of experimental ubiquitylation site is before April 4 2006. b Independent data I: the annotation date of experimental ubiquitylation site is between April 4 2006 and January 1 2008. ::: Feature extraction and coding {#s2b} ----------------------------- ### Coding of amino acid sequences {#s2b1} Fragments of amino acids are extracted from positive and negative training sets using a window of length 2*n*+1 that is centered on ubiquitylation sites. Various values of *n* are used to determine the optimal window length. The BLOSUM62 matrix is adopted to represent the protein primary sequence information as the basic feature set for learning radial basis function networks. A matrix of *m*×*n* elements is used to represent each residue in a training dataset, where *n* denotes the window size and *m* = 21, which elements comprise 20 amino acids and one terminal signal. Each row of the normalized BLOSUM62 matrix is adopted to encode one of 20 amino acids. ### Compositions of amino acids and amino acid pairs {#s2b2} A total of *n* vectors {*x~i~*, *i* = 1, ..., *n*} were used, to represent all *n* proteins in the training data. Each vector is labeled with the group of its corresponding protein (e.g. ubiquitylated or non-ubiquitylated). The vector *x~i~* has 20 elements for the amino acid composition and 400 elements for the amino acid pair composition. The 20 elements specify the numbers of occurrences of 20 amino acids normalized with the total number of residues in the protein, and the 400 elements specify the numbers of occurrences of 400 amino acid pairs normalized with the total number of residues in the protein. In this investigation, amino acid composition and amino acid pair composition are combined, yielding, 420 elements in each vector. ### Position Specific Scoring Matrix Profiles {#s2b3} In the point of view of structure, several amino acid residues of a protein can be mutated without changing its structure, and two proteins may have similar structures with different amino acid compositions. Position Specific Scoring Matrix (PSSM) profiles, which have been extensively utilized in protein secondary structure prediction, subcellular localization and other bioinformatics problems are adopted herein with significant improvement [@pone.0017331-Jones1], [@pone.0017331-Xie1], [@pone.0017331-Ou2]. The PSSM profiles were obtained by PSI-BLAST against non-redundant fragment sequences of Ub sites. [Figure 2](#pone-0017331-g002){ref-type="fig"} displays in detail how to generate the 400D PSSM features for each ubiquitylation and non-ubiquitylation site. The matrix of *m*×20 elements has rows centered on ubiquitylation or non-ubiquitylation site, extracted from the PSSM profile, where *m* represents the window size and 20 represents the position specific scores for each type of amino acid. Thereafter, the *m*×20 matrix is transformed into a 20×20 matrix by summing up the rows that are associated with the same type of amino acid. Finally, every element in 20×20 matrix is divided by the window length *m* and then is normalized using the formula: . ::: {#pone-0017331-g002 .fig} 10.1371/journal.pone.0017331.g002 Figure 2 ::: {.caption} ###### The detailed process of generating position specific scoring matrix (PSSM) and encoding the fragment of amino acid sequence by generated PSSM. :::  ::: ### Structural characteristics {#s2b4} Since most of the experimental ubiquitylated proteins do not have corresponding protein tertiary structures in PDB [@pone.0017331-Berman1], an effective tool, RVP-Net [@pone.0017331-Ahmad1], was used to compute the ASA value based on the protein sequence. The computed ASA value is the percentage area of each amino acid on the proteins that is accessible to the solvent. RVP-net applies a neural network to predict real ASA values of the residues based on neighborhood information, with a mean absolute error of 18.0--19.5%, defined as the absolute difference between the predicted and experimental values of relative ASA per residue [@pone.0017331-Ahmad1]. Full-length protein sequences with experimental ubiquitylated sites are input to RVP-Net to compute the ASA value for all residues. The ASA values of amino acids that surround the ubiquitylation site were extracted and scaled from zero to one. PSIPRED [@pone.0017331-McGuffin1] was utilized to compute the secondary structure that surrounds the ubiquitylation sites from the protein sequence. PSIPRED is a simple and reliable method for predicting secondary structure, which applies two feed-forward neural networks to analyze the output obtained from PSI-BLAST [@pone.0017331-Altschul2]. PSIPRED 2.0 achieved a mean Q~3~ score of 80.6% across all 40 submitted target domains without obvious sequence similarity with structures that are present in PDB; accordingly, PSIPRED has been ranked as the best of 20 evaluated methods [@pone.0017331-Bryson1]. The output of PSIPRED is "H," "E" or "C", which stand for helix, sheet and coil, respectively. The full-length protein sequences with ubiquitylation sites are inputted to PSIPRED to determine the secondary structure of all residues. The orthogonal binary coding approach is adopted to transform the three terms that specify the secondary structure into numeric vectors. For instance, helix is encoded "100;" sheet is encoded "010," and coil is encoded "001." ### F-score measurement {#s2b5} To study further the specificity of the substrate sites, the features that statistically differ between ubiquitylation sites and non-ubiquitylation sites are identified, based on a statistical measurement of F-score [@pone.0017331-Lin1]. The F-score of the *i*th feature is defined as,where , and are the average value of the *i*th feature in whole, positive, and negative data sets, respectively. denotes the number of positive data, denotes the number of negative data, denotes the *i*th feature of the *k*th positive instance, and denotes the *i*th feature of the *k*th negative instance [@pone.0017331-Lin1]. F-score supports a simple approach for measuring features that are more discriminative. If the *i*-th feature has a high F-score, then this feature effectively discriminates between positive and negative datasets. Training and evaluation of model {#s2c} -------------------------------- In this work, the QuickRBF package [@pone.0017331-Ou3] has been employed to construct radial basis function network (RBFN) classifiers. As presented in [Fig. S1](#pone.0017331.s001){ref-type="supplementary-material"} (See Supplementary Materials), the general architecture in an RBFN consists of three layers, namely the input layer, the hidden layer, and the output layer. The input layer broadcasts the coordinates of the input vector to each of the nodes in the hidden layer. Each node in the hidden layer then produces an activation based on the associated radial basis kernel function. Finally, each node in the output layer computes a linear combination of the activations of the hidden nodes. The general mathematical form of the output nodes in RBFN is as follows:where denotes the function corresponding to the *j*-*th* output node and is a linear combination of *k* radial basis functions with center m*~i~* and bandwidth s*~i~*; Also, *w~ji~* denotes the weight associated with the correlation between the *j*-*th* output node and the *i*-*th* hidden node. In this work, we adopted a fixed bandwidth (σ) of five, and used all input nodes as centers (*k* = n). With its several bioinformatics applications, classification based on radial basis function network has been extensively adopted to predict factors such as the cleavage sites in proteins [@pone.0017331-Yang1], inter-residue contacts [@pone.0017331-Zhang1], protein disorder [@pone.0017331-Su1], discrimination of β-barrel proteins [@pone.0017331-Ou2], and identification of O-linked glycosylation [@pone.0017331-Chen1]. Predictive performance of the constructed RBFN classifier is evaluated by performing *k*-fold cross validation. The original data (training data in [Table 1](#pone-0017331-t001){ref-type="table"}) is divided into *k* subgroups by splitting each dataset into *k* approximately equal sized subgroups. In one round of cross-validation, a subgroup is regarded as the test set, and the remaining *k*-1 subgroups are regarded as the training set. The cross-validation process is repeated *k* rounds, with each of *k* subgroups used as the test set in turn. Then, the k results are combined to produce a single estimation. The advantage of *k*-fold cross-validation is that all original data are regarded as both training set and test set, and each data is used for test exactly once [@pone.0017331-Ron1].The following measures of predictive performance of the trained models are defined. Precision (Pr) = TP/(TP+FP), Sensitivity (Sn) = TP/(TP+FN), Specificity (Sp) = TN/(TN+FP), and Accuracy (Acc) = (TP+TN)/(TP+FP+TN+FN), where TP, TN, FP and FN represent the numbers of true positives, true negatives, false positives and false negatives, respectively. Results and Discussion {#s3} ====================== Amino acid composition of ubiquitin conjugation sites {#s3a} ----------------------------------------------------- This investigation focuses on the analysis of ubiquitin conjugated lysine. In ubiquitin conjugation, the region of the ubiquitin-conjugated lysine residues is in directly contact with the E3 ligase catalytic center. Since E3 ubiquitin ligase enzymes have a substrate binding specificity, whether the region of ubiquitin-conjugated lysine conserved amino acid motifs for E3 ubiquitin ligase recognition must be explored. After the duplicated sequences of experimental ubiquitylation sites are removed, as shown in [Fig. 3](#pone-0017331-g003){ref-type="fig"}, the amino acids composition that flanked the ubiquitin-conjugated lysines (ubiquitylation site centered at position 0) are graphically visualized as a 41-mer sequence logo. WebLogo [@pone.0017331-Crooks1], [@pone.0017331-Schneider1] is adapted to generate the graphical sequence logo for the relative frequency of the amino acid at each position around the ubiquitylated sites. The conservation of amino acids around the ubiquitylation sites can then be easily investigated. Based on the sequence logo representation, the most abundant residues of the ubiquitylation sites are the charged and polar amino acids, including Aspartic acid (D) and Glutamic acid (E). The amino acids around the modified sites are not obviously conserved, a slight difference between the preferences of amino acids for the ubiquitylation and non-ubiquitylation sites. ::: {#pone-0017331-g003 .fig} 10.1371/journal.pone.0017331.g003 Figure 3 ::: {.caption} ###### The position-specific amino acid composition, accessible surface area and secondary structure of ubiquitin conjugated lysines and non-ubiquitin conjugated lysines. :::  ::: Since the representation of sequence logos involves different preferences of amino acids for ubiquitylated and non-ubiquitylated sites, the statistical difference in the distribution of amino acids around ubiquitylated (Ub) and non-ubiquitylated (non-Ub) lysines is calculated. [Figure S2](#pone.0017331.s002){ref-type="supplementary-material"} (See Supplementary Materials) displays the compositional differences between Ub and non-Ub sites. The more abundant amino acids at the Ub sites are Alanine (A), Aspartic acid (D), Glycine (G) and Isoleucine (I), and the depleted hydrophobic residues around these sites include Cysteine (C) and Leucine (L) around Ub sites. Moreover, the Lysine (K) and Serine (S) are less abundant around Ub sites. The amino acid sequences around the ubiquitin-conjugated sites can be alternatively grouped by various methods to generalize the sequence feature because amino acid classification is hierarchical. As presented in [Table S1](#pone.0017331.s004){ref-type="supplementary-material"} (See Supplementary Materials), the three-class grouping method and the eight-class grouping method are used to 20 amino acids into subgroups that capture their chemical properties. Three-class grouping methods can be based on hydrophobicity [@pone.0017331-Chothia1], polarity [@pone.0017331-Grantham1], normalized van der Waals volume [@pone.0017331-Fauchere1] and polarizability [@pone.0017331-Charton1]. Additionally, a Two Sample Logo [@pone.0017331-Vacic1] of 41-mer compositional biases around Ub conjugation sites compared to non-Ub conjugation sites is presented in [Fig. S3](#pone.0017331.s003){ref-type="supplementary-material"} (See Supplementary Materials). The amino acid residues that significantly enriched and depleted (*P*-value \<0.05; t-test) around Ub conjugation sites are shown. With the investigation of position-specific difference of amino acid composition in 41-mer window length, [Figure S3](#pone.0017331.s003){ref-type="supplementary-material"} indicates the positions that are distant from Ub sites have statistically significant differences of amino acid composition. Structural characteristics of ubiquitin conjugation sites {#s3b} --------------------------------------------------------- A side-chain of an amino acid that undergoes post-translational modification preferentially accesses the surface of a protein [@pone.0017331-Pang1]. To investigate the preference of the solvent accessible surface area [@pone.0017331-Hu1] that surrounds ubiquitin conjugation sites in protein tertiary structures, the experimentally identified ubiquitylation sites are mapped to the corresponding positions of the protein entries in the Protein Data Bank (PDB) [@pone.0017331-Berman1]. The preference of the secondary structure around the ubiquitylation sites is also considered. Since most of the experimentally confirmed ubiquitylated proteins do not have corresponding protein tertiary structures in PDB [@pone.0017331-Berman1], RVP-Net [@pone.0017331-Ahmad1] and PSIPRED [@pone.0017331-McGuffin1] are adopted to compute the ASA value and secondary structure, respectively, from the protein sequence. [Figure 3](#pone-0017331-g003){ref-type="fig"} presents the sequence logo of the secondary structure and the average percentage of ASA in the 41-mer window (−20∼+20) of the ubiquitylation (Ub) and non-ubiquitylation (non-Ub) sites. In the investigation of secondary structure around the Ub sites, Catic *et al.* [@pone.0017331-Catic1] has found the preference for coil structure. In this work, the observations reveal that Ub ligase (E3) prefers to recognize the regions that are located in coil (loop) or helix structures. In contrast to Ub sites, non-Ub sites don not have an obviously preferred secondary structure. In the study of solvent accessibility, most of the Ub or non-Ub lysines are located in the highly accessible surface area. However, the mean solvent-accessible surface area that surrounds the Ub sites slightly exceeds that around non-Ub sites. Investigation of distant sequence features for ubiquitylation sites {#s3c} ------------------------------------------------------------------- Owing to the direct interaction between the enzyme and the substrate site, most of the proposed PTM prediction methods investigate amino acid sequences close to the modified sites. The ubiquitin conjugation pathway, which involves a sequential process with a group of enzymes known as E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase) [@pone.0017331-Hershko1] motivates an investigation of the distant sequence features that are distant from ubiquitylation sites. In [Fig. 4](#pone-0017331-g004){ref-type="fig"}, a graphical model represents the hypothesis that contains a substrate site that is distant from ubiquitylated lysine for E3 recognition. Ubiquitin is mostly conjugated on the lysine residue of a protein by substrate recognition of Ub-ligating (E3) enzymes [@pone.0017331-Ou1]. E3 enzymes function as the substrate recognition modules of the system and are capable of interaction with both E2 and substrate. Thus, the E3 ligase must be sufficiently specific and must act only on a defined subset of cellular targets to ensure signal fidelity [@pone.0017331-Hicke1]. Based on the measurement of the F-score in a large window size (−20∼+20), [Fig. 5](#pone-0017331-g005){ref-type="fig"} displays the statistically significant composition of amino acids at positions −16, −10, −3, −1, +1, +5, +10, +13, and +17. The surrounding positions that have high F-scores are (significant for differentiating the ubiquitylation sites from the non-ubiquitylation sites. Additionally, [Tables S2](#pone.0017331.s005){ref-type="supplementary-material"} and [S3](#pone.0017331.s006){ref-type="supplementary-material"} (See Supplementary Materials) present the significant amino acids and di-peptides in the surrounding region (−20∼+20), which have a higher F-score value. ::: {#pone-0017331-g004 .fig} 10.1371/journal.pone.0017331.g004 Figure 4 ::: {.caption} ###### The hypothetic model of identifying the distant sequence features for E3 recognition. :::  ::: ::: {#pone-0017331-g005 .fig} 10.1371/journal.pone.0017331.g005 Figure 5 ::: {.caption} ###### The statistically significant composition of amino acids for each position in the window length from −20 to +20. Based on the measurement of F-score, the positions −16, −10, −3, −1, +1, +5, +10, +13, and +17, containing higher value of F-score, are significant for differentiating the ubiquitylation sites from non-ubiquitylation sites. :::  ::: Position specific scoring matrix (PSSM), which is generated by PSI-BLAST [@pone.0017331-Altschul1] search against a non-redundant database of protein sequence, is utilized to obtain evolutionary information about amino acids around the ubiquitin conjugation sites. The constructed PSSM includes the probability that each amino acid is present at each position. Therefore, PSSM is regarded as a measure of residue conservation in a window of a particular length. [Figure 6](#pone-0017331-g006){ref-type="fig"} displays statistically significant evolutionary information concerning amino acids at each position in the window from −20 to +20. Based on the measurement of F-scores, the positions −19, −17, −15, −12, −10, −4, −1, +5, +9, +13, +15 and +18, where the F-scores are highest, are significant for differentiating the ubiquitylation sites from the non-ubiquitylation sites. In the investigation of distant sequence features that are distant from ubiquitylation sites, the length of the training data window for learning the predictive model is set to 41-mer (−20∼+20). ::: {#pone-0017331-g006 .fig} 10.1371/journal.pone.0017331.g006 Figure 6 ::: {.caption} ###### The statistically significant evolutionary information of amino acids for each position in the window length from −20 to +20. Based on the measurement of F-score, the positions −19, −17, −15, −12, −10, −4, −1, +5, +9, +13, +15 and +18, containing higher value of F-score, are significant for differentiating the ubiquitylation sites from non-ubiquitylation sites. :::  ::: To demonstrate the distant sequence features are informative for the identification of ubiquitylation sites, herein, five-fold cross-validation is performed to evaluate the models trained with various window sizes 2*n*+1, where *n* varies from five to twenty. The predictive models (RBFN classifiers) are trained with the feature of amino acid composition. [Figure 7](#pone-0017331-g007){ref-type="fig"} presents the sensitivity, specificity, and accuracy of cross-validation based on various window lengths. As different window sizes from 11-mer to 41-mer are applied, the predictive accuracy improves from 63.1% to 73.7%, the sensitivity and specificity increase as well. Especially for the window length which is longer than 35-mer, the predictive power is apparently improved with the accuracy that is higher than 70.0%. As the investigation of distant sequence features in [Figures 5](#pone-0017331-g005){ref-type="fig"} and [6](#pone-0017331-g006){ref-type="fig"}, the model that was trained with a large window length performs better than that without the distant sequence features. ::: {#pone-0017331-g007 .fig} 10.1371/journal.pone.0017331.g007 Figure 7 ::: {.caption} ###### The predictive performance of the models trained with different window length varying from 11-mer to 41-mer. :::  ::: Predictive performance of cross-validation using various training features {#s3d} -------------------------------------------------------------------------- Most predictive models are based on the features of amino acid sequences. To determine which features can be utilized to construct models that differentiate between ubiquitylation sites and non-ubiquitylation sites, various features, including the sequence of amino acids, amino acid composition, accessible surface area, and physicochemical properties are evaluated by *k*-fold cross-validation. The amino acids (AA) and accessible surface area (ASA) around the ubiquitylated sites are encoded using a BLOSUM62 matrix and the RVP-Net-computed ASA values, respectively. [Table 2](#pone-0017331-t002){ref-type="table"} presents the predictive performance achieved using various training features, based on five-fold cross-validation. Of the models trained using individual features, those that are trained using amino acid composition slightly outperform those that are trained using amino acids, ASA, the secondary structure, or PSSMs. In particular, the model trained with the PSSM profile of non-redundant ubiquitylated protein sequences achieves an accuracy of 70%. However, the model that is trained with the secondary structure underperforms prediction based on ubiquitylation sites. According to the F-score of distant sequence features, the amino acid composition and evolutionary information (PSSM) at several flanking positions are statistically differently distributed between ubiquitylation sites and non-ubiquitylation sites. Therefore, the effects of combining informative features are evaluated. As presented in [Table 2](#pone-0017331-t002){ref-type="table"}, the model that is trained using the combination of amino acid composition and PSSM profile of non-redundant fragment sequences of Ub sites performs best, with the best-balanced predictive sensitivity and specificity. ::: {#pone-0017331-t002 .table-wrap} 10.1371/journal.pone.0017331.t002 Table 2 ::: {.caption} ###### The predictive performance of cross-validation using various training features. ::: {#pone-0017331-t002-2} Training features Sensitivity (%) Specificity (%) Accuracy (%) ------------------------- ----------------- ----------------- -------------- AA (Blosum62) 54.3 67.9 67.5 AA composition 63.6 74.1 73.7 AA pair composition 59.2 74.4 74.0 Accessible Surface Area 59.3 69.7 68.6 Secondary structure 58.4 59.7 59.1 PSSM 1 60.0 66.2 66.0 PSSM 2 50.9 69.3 68.7 PSSM 3 54.3 68.9 68.5 AA composition + PSSM 1 62.3 73.5 73.1 AA composition + PSSM 2 61.8 74.6 74.2 AA composition + PSSM 3 **65.5** **74.8** **74.5** AA: amino acid; PSSM 1: The PSSM profiles were obtained by using PSI-BLAST against UniProt NR database; PSSM 2: The PSSM profiles were obtained by using PSI-BLAST against non-redundant Ub protein sequence database; PSSM 3: The PSSM profiles have been obtained by using PSI-BLAST against non-redundant fragments of Ub site sequ2ences. ::: Predictive performance of independent testing {#s3e} --------------------------------------------- To determine whether the models (are over-fitted to their training data, independent sets of data concerning Ub sites and non-Ub sites are constructed and used to test the model that was trained with the combination of amino acid composition and the PSSM profile of non-redundant fragment sequences of Ub sites, which have the highest predictive accuracy. Independent test sets include 57 ubiquitylation sites and 3502 non-ubiquitylation sites, According to [Table 3](#pone-0017331-t003){ref-type="table"}, are used to determine the predictive sensitivity, specificity, and accuracy of the proposed method, which were 57.9%, 72.4%, and 72.2%, respectively. Generally, the performance in an independent test approaches that of cross-validation. Whereas cross-validation outperforms independent testing, the performance of the trained model may be overestimated. The independent test establishes that the constructed RBF model does not over-fit the training data. The independent test sets were used to test other ubiquitylation predictors. The predictive sensitivity and specificity of UbiPred [@pone.0017331-Tung1] were 52.6% and 52.6%, respectively, indicating balanced predictive performance. However, the independent test also indicates that UbiPred does not perform as well as its developers claimed. The predictive sensitivity and specificity of UbPred [@pone.0017331-Radivojac1] were 42.1% and 68.7%, respectively, indicating poor sensitivity for an independent test set. In UbiPred and UbPred, the data source of experimental ubiquitylation sites is collected from UbiProt [@pone.0017331-Chernorudskiy1], which mainly stored the yeast ubiquitylation data. However, UbSite integrates the experimental ubiquitylation sites from UbiProt and UniProtKB/Swiss-Prot [@pone.0017331-Boeckmann1], which accumulated the ubiquitylation data from multiple species. This could partially explain the low sensitivities of UbPred and UbiPred on these independent test data which come from multiple species. ::: {#pone-0017331-t003 .table-wrap} 10.1371/journal.pone.0017331.t003 Table 3 ::: {.caption} ###### Comparison between our method (UbSite) and other ubiquitylation prediction tools. ::: {#pone-0017331-t003-3} Tools UbSite UbiPred UbPred ---------------------- ------------------------------- ---------------------------- -------------------------------------------------------------- ------- Materials UbiProt + Swiss-Prot UbiProt UbiProt Method Radial basis function network Support vector machine Random forest Training features AAC + PSSM Physicochemical properties AAC + PSSM + disordered regions + physicochemical properties Training data Number of positive data 385 151 272 Number of negative data 12,582 3,424 4,651 Window length −20∼+20 −10∼+10 −12∼+12 Proposed performance Sensitivity (%) 65.5 83.44 \- Specificity (%) 74.8 85.43 \- Accuracy (%) 74.5 84.44 72.0 Independent test Sensitivity (%) **57.9** 52.6 42.1 Specificity (%) **72.4** 52.6 68.7 Accuracy (%) **72.2** 52.6 68.3 Abbreviation: AAC, amino acid composition; PSSM, position-specific scoring matrix. ::: Conclusion {#s3f} ---------- This investigation proposes a method, UbSite, which incorporates the efficient radial basis function (RBF) network to identify ubiquitin conjugation sites on protein sequences. Not only the amino acid composition but also the structural characteristics, physicochemical properties, and evolutionary information of amino acids around the ubiquitylation (Ub) sites are explored. With reference to the pathway of ubiquitin conjugation, which involves a sequential process with a group of enzymes known as E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase), the substrate sites for E3 recognition, which are distant from ubiquitylation sites, are examined. According to the measurement of F-score in a large window size(−20∼+20), most of the statistically significant amino acids and evolutionary information (PSSM), which can be used effectively to differentiate Ub sites from non-Ub sites, are located at large distances from Ub sites. To prevent any overestimation of predictive performance, duplicated sequences are removed using a window size determined by the collected data sets. Although the amino acid sequences around the ubiquitin conjugated sites do not contain a conserved motif, cross-validation results indicate that the integration of the evolutionary information around the sites can improve predictive performance. [Table 3](#pone-0017331-t003){ref-type="table"} compares proposed method with other ubiquitylation prediction tools, in terms of materials, method, training features, number of training data, window length, and proposed predictive performances. Furthermore, the independent test demonstrates that UbSite can outperform other ubiquitylation prediction tools. Although the proposed method is accurate and robust, according to independent tests, some issues remain to address in the future. Firstly, the structural preferences of ubiquitin conjugation sites preferred structures at ubiquitin conjugation sites must be examined in greater detail because flanking residues are not conserved. In addition to the solvent-accessible surface area and the secondary structure, the B-factor, intrinsically disordered region, protein linker region, and other factors should be explored experimentally at ubiquitylation sites in the protein regions with PDB entries. Following work done previously on phosphorylation [@pone.0017331-Zanzoni1], the local 3D structure of ubiquitylation sites may be extracted for further analysis. Secondly, with reference to the pathway of ubiquitin conjugation, the ubiquitylated proteins may contain a substrate site is distant from ubiquitylation lysine and useful in E3 recognition. Therefore, the distant sequence features of ubiquitylation sites should be investigated in more detail. For instance, using the motif discovery tool, like MEME [@pone.0017331-Bailey1], to explore the significant motifs which may be the substrate sites of E3 recognition. Supporting Information {#s4} ====================== Figure S1 ::: {.caption} ###### **The general architecture of RBFN consisting of input layer, hidden layer, and output layer.** (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **The compositional differences of amino acids around ubiquitylation sites compared to non-ubiquitylation sites.** (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **A Two Sample Logo of the compositional biases around Ub conjugation sites compared to non-Ub conjugation sites.** The amino acid residues that significantly enriched and depleted (*P*-value\<0.05; t-test) around Ub conjugation sites are shown. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **The graphical representation of chemical properties surrounding ubiquitylation sites using different grouping method.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **F-score of amino acid composition for 40 positions around Ubi site.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Top 20 di-peptides with high value of F-score in the 41-mer window size (−20∼+20) around Ub site.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank the reviewers for their constructive comments, which improved the quality of this manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was partially supported by National Science Council (NSC) of Taiwan, NSC99-2320-B155-001 to TYL, and NSC99-2221-E155-073 to YYO. No additional external funding was received for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: TYL SAC HYH YYO. Performed the experiments: TYL SAC HYH YYO. Analyzed the data: TYL SAC HYH YYO. Contributed reagents/materials/analysis tools: TYL SAC HYH YYO. Wrote the paper: TYL SAC YYO.

PubMed Central

2024-06-05T04:04:19.130713

2011-3-9

PMC3052308

Introduction {#s1} ============ Consider an economic decision paradigm with two options. The first option (A) is constant and consists of winning euros after month with % of probability, while the second option (B) can consist, for instance, of winning the same amount of money after months with % of probability. Some people would prefer the first -closer in time but riskier- option and some others would prefer the second -delayed in time but safer- option. When varying the probability and the time of option B, one could find a task configuration where both options are evaluated as highly similar in terms of attractiveness. This kind of situation gives rise to a decision conflict. Different task configurations might produce heterogeneous decision patterns covering from a predominance of option A to a predominance of option B. Briefly, task configurations that produce a predominant answer (either option A or B) can be characterized by a low uncertainty, while task configurations with a balanced number of A and B outcomes can be characterized by a high uncertainty. The variability of the outcomes comes from within- and inter-subject variabilities. The former happens when decisions of a subject for certain configuration are not self-consistent and the latter happens when different subjects provide opposed decisions. How can the level of conflict in a decision be evaluated? This question has received increasing attention in the last decade. Prediction paradigms, where participants have to anticipate an outcome have been the norm. In such paradigms, the level of ambiguity of the experiment is controlled, manipulating either the information the subject used to correctly make the prediction [@pone.0017408-Volz1], [@pone.0017408-Volz2] or the probability of success [@pone.0017408-Volz2]--[@pone.0017408-Huettel1]. Consequently in these studies the ambiguity level was proposed a priori during the design stage. However, it has been shown that sometimes participants behavior does not necessarily correspond to that inferred from the probability of success [@pone.0017408-Paulus1]. In two of the earliest studies, participants had to advance the color or the suit of a card [@pone.0017408-Elliott1] or whether the next card was bigger or lower than the previous one [@pone.0017408-Critchley1]. This permitted the comparison between low and high difficulty guessing. Prefrontal areas, but also the anterior cingulate, were more related to trials with high difficulty. In other studies [@pone.0017408-Paulus1], [@pone.0017408-Paulus2] participants predicted the appearance of stimuli. Prefrontal, parietal and thalamic areas were active during such task. Volz et al. [@pone.0017408-Volz1], [@pone.0017408-Volz2], [@pone.0017408-Volz3] presented pairs of alien comic figures and subjects had to infer which figure would win in a fictional fight. In one of the experiments there was an unknown probability of winning for each pair of figures that had to be learned as the experiment advanced. In the other experiment there were a set of rules that marked which figure won each time. The level of uncertainty of the experiment was manipulated by varying the degree of knowledge of the winning rules provided to the participants. Although there were minor differences in brain activation between the two paradigms, a fronto-median cluster correlated with the degree of uncertainty independently of the paradigm used. Huettel et al. also found a frontomedian activation when processing uncertainty in a paradigm where visual cues helped to predict the following answer [@pone.0017408-Huettel1]. In a more recent article [@pone.0017408-Pochon1], male subjects were required to discriminate attractiveness between pairs of women faces. Each picture had been rated previously by another group of participants, allowing to estimate and control the level of decision conflict. While all these studies associate pre-frontal and or fronto-median areas to the processing of conflict, a role in uncertainty management has also been assigned to the cerebellum [@pone.0017408-Blackwood1]. As shown above, the concepts of certainty/uncertainty have been commonly used in decision making studies and most of their quantifications have been represented by either theoretical probability distributions or by empirical relative frequencies. Interestingly, an uncertainty descriptor that can be quantified from any probability distribution is the central measurement of information theory. In information theory, Shannon entropy [@pone.0017408-Shannon1] (denoted by ) measures the amount of information or uncertainty contained in a message (usually measured in bits). Its use in decision making tasks has been scarce [@pone.0017408-BischoffGrethe1], [@pone.0017408-Fiorillo1] and mostly focused on the uncertainty of task-related probabilistic events [@pone.0017408-Aron1]--[@pone.0017408-Harrison1] and not on the decisions of the subjects. However, the close relationship between Shannon\'s concept of information and the psychological concept of uncertainty has been pointed out [@pone.0017408-Huettel1]. Briefly, for random variables with possible values has two main properties. First, is bits if and only if all the values contained in the message are the same (i.e, the outcome is completely certain). Second, is maximum when the frequency of values in the message is equal, resulting in bits. Intuitively, a sequence of flipping a perfect coin would have maximum entropy ( bit) while a two-tails coin would have the minimum entropy ( bits). Going back to our economic decision task, let us consider that we aim to transmit within a message () the decisions of all participants for a certain task configuration (i.e. specifying the probability and time of option B). Such message will be formed by a finite sequence of symbols with values A or B indicating the options selected. What would be the uncertainty of the message? On the one hand, a message with the decisions for a very easy task configuration would be constant (either or ) and thus the uncertainty associated to it would be bits. On the other hand a message formed by the decisions for a very difficult task configuration would be, for instance, and thus the uncertainty associated to it would be bit. Messages obtained from other task configurations would produce intermediate values of uncertainty within the range . The aim of this study is to introduce the concept of Shannon entropy in decision making paradigms as a decision uncertainty descriptor of the task and to map the functional fingerprint of such uncertainty using an economic decision task under different configurations of probability and time. To achieve this, decision outcomes and fMRI BOLD data were analyzed in three steps. Firstly, Shannon entropy concept was used to characterize the decision uncertainty associated to each task configuration in terms of within- inter- and pooled-variabilities. Multi-linear regression analyses revealed that pooled-entropy was the best predictor of the response times and was used to characterize the uncertainty associated to each task configuration. Secondly, these pooled-entropy values were used as a neural correlate with BOLD activity in order to obtain brain areas codifying uncertainty. Thirdly, the psycho-physiological interactions (PPI) paradigm and a conjunction analysis were employed to study the functional integration of the uncertainty codification, i.e., which brain areas gained functional connectivity as the entropy associated to the task configurations increased. Results {#s2} ======= Behavioral results {#s2a} ------------------ During the scanning sessions performed, participants answered times to each of the different task configurations presented in a pseudo-random order. Each task was constituted by two options and each answer consisted of making a binary choice between them. For out of the task configurations, there was a constant option (A) which consisted of winning euros after month with of probability. In those cases the alternative option (B) was different at each task configuration by varying the time from to months and the reward probability from to . Two additional configurations were included in which both options A and B varied (see [methods](#s4){ref-type="sec"} for a detailed explanation). In our decision making experiment, uncertainty associated to a task configuration is intrinsically related to the variability of the decisions reported for such task. The decision sets of each task configuration (i.e. the collection containing all the decisions reported by the subjects for a specific task configuration) contain two different sources of variability that might contribute to quantify the level of uncertainty. On the one hand, a high within-subject variability reveals lack of self-consistency during the responses made by a participant for certain task configuration. On the other hand, a high inter-subject variability reveals the existence of opposed preferences among individuals. Both factors add evidence of a task being difficult, i.e. a task with associated high uncertainty. Furthermore we hypothesized that the entropy of the *pooled* variability containing both inter-subject and within-subject variabilities might be an appropriate descriptor of such task uncertainty level. Hence we quantified the entropy of each task in terms of within-subject variability (), inter-subject variability () and pooled variability (). Individual entropy maps of can be seen at [Figure S1](#pone.0017408.s001){ref-type="supplementary-material"}, where the presence of highly consistent subjects (such as , , and ) and lowly consistent subjects (such as , and ) can be observed. [Tables S1](#pone.0017408.s005){ref-type="supplementary-material"} and [S2](#pone.0017408.s006){ref-type="supplementary-material"} summarize the entropy values corresponding to and respectively. These two descriptors show a qualitative similar behavior. Values corresponding to displayed in [Figure 1C](#pone-0017408-g001){ref-type="fig"} may be presented as an interpolated entropy map ([Figure 2](#pone-0017408-g002){ref-type="fig"}) which summarizes the effects of probability and time dimensions on the uncertainty of the decisions. Results in a numerical format are shown in [Table S2](#pone.0017408.s006){ref-type="supplementary-material"}. Axis and respectively determine the time (months) and the reward probability of every option . Therefore each point unequivocally represents one task configuration. An entry at row and column of [Table S2](#pone.0017408.s006){ref-type="supplementary-material"} describes the pooled uncertainty of the task configuration whose options presented were ,, month and ,,, i.e., the constant and alternative options respectively. Covering the range of a value delimited variable (Shannon entropy in our case) is specially relevant in order to have accurate results when using it as a neural correlate. The set of task configurations selected for our experiment produced a heterogeneous set of uncertainty values of within the range . For instance, we identified low-uncertainty configurations such as and , intermediate uncertainty configurations such as and and high uncertainty configurations such as and . ::: {#pone-0017408-g001 .fig} 10.1371/journal.pone.0017408.g001 Figure 1 ::: {.caption} ###### Overview of the decision making paradigm. **A. Visual presentation.** Example of visual presentation with two options shown to the participants during the decision-making task. This presentation corresponds to the task configuration . **B. Presentation design.** The decision-making trials were presented in blocks of three and were interleaved alternatively with one of the different controls (C1, C2), which also appeared in blocks of three. **C. Shannon entropy.** Continuous line stands for the entropy model with respect to the probability (dichotomous variable) and squares refer to experimental entropy values of the pooled decisions, , made at different task configurations, . Note that entropy model is symmetric to the probability. It reaches low values when the variable under study takes most of the times either one value or the other, and reaches its maximum when the random variable takes each of the possible values with of probability. :::  ::: ::: {#pone-0017408-g002 .fig} 10.1371/journal.pone.0017408.g002 Figure 2 ::: {.caption} ###### Shannon entropy map () associated to the decision task . Reward probability () and time () axis characterize option B { euros,, } and define each task configuration , since option A is constant { euros, % month}. Vertices formed by dotted lines correspond to actual evaluated task configurations and intermediate points are the result of a bi-dimensional interpolation process. Two additional task configurations with no constant option are not included in this figure. :::  ::: One of the most common indicators of difficulty is the response time (RT). The average RT per subject per task was used as the dependent variable in three multi-linear regression analyses including as independent variables the average response time of each subject and one of the entropies at a time (within- inter and pooled-entropies). Results shown in [Table 1](#pone-0017408-t001){ref-type="table"} revealed that, being the three entropies significant factors, was the best predictor of response times and hence was used to characterize the entropy associated to each task. Additional evidence of the appropriateness of this approach was obtained an analogous the multi-linear regression analysis performed only on consistent decisions (see [Table 1](#pone-0017408-t001){ref-type="table"}). We selected, for each subject and for each task configuration, only those sets of 5 responses that had been fully consistent (i.e. always A or always B). This corresponds to zero entropy values at the individual entropy maps of [Figure S1](#pone.0017408.s001){ref-type="supplementary-material"}. This subset of decisions was used in the last model in [Table 1](#pone-0017408-t001){ref-type="table"} to show that, even in this consistent data subset, significantly contributed to predict . ::: {#pone-0017408-t001 .table-wrap} 10.1371/journal.pone.0017408.t001 Table 1 ::: {.caption} ###### Multi-linear regression analyses. ::: {#pone-0017408-t001-1} Model ------------------------------------------------ ------- ------- ------- 0.613 0.319 0.485 0.619 0.322 0.487 0.619 0.357 0.510 Model (consistent decisions only, i.e. where ) 0.639 0.322 0.505 Evaluation of the three entropy measurements (within-subject, inter-subject and pooled) with respect to average response times per subject per task . Models include the average response time of each subject during the experiment and a constant . Beta values correspond to the standardized coefficients. reflects the fraction of variance of explained by the model. Results indicate that, when is fixed, is the best descriptor in order to explain . This model containing the term had also the highest . Even when analyzing only the consistent decisions per subject per task, had a significant predictive capacity and the model was able to explain about half of the variability of . ::: Neural correlate between BOLD signal and uncertainty {#s2b} ---------------------------------------------------- We evaluated the set of uncertainties , ..., , one from each task configuration, as a neural correlate. This allowed us to test which brain areas showed during every task a BOLD activity that codifies the decision uncertainty quantified for each task. The regression analysis yielded two clusters (see [Table 2](#pone-0017408-t002){ref-type="table"}). The largest one ( voxels, [Figure 3](#pone-0017408-g003){ref-type="fig"}) includes parts of the right middle cingulate cortex (MCC), of pre-supplementary motor area (pre-SMA, bilateral) and a small part of the left superior medial gyrus. A second cluster ( voxels) was found in the left thalamus. It includes part of mediodorsal (MD) nucleus and ventral anterior (VA) that project to the pre-frontal cortex [@pone.0017408-Behrens1]. These two clusters showed also greater activation during the decision making tasks (DM) than during the motor action control (C2). The DMC2 t-contrast map ([Figure S2](#pone.0017408.s002){ref-type="supplementary-material"}) shows those areas with a higher activation during DM with respect to C2. [Figure S3](#pone.0017408.s003){ref-type="supplementary-material"} shows coronal and sagital views of the two clusters. ::: {#pone-0017408-g003 .fig} 10.1371/journal.pone.0017408.g003 Figure 3 ::: {.caption} ###### Cluster codifying uncertainty. This is cluster 1 at [Table 2](#pone-0017408-t002){ref-type="table"}. It includes parts of the right MCC and of the pre-SMA(bilateral) obtained by using Shannon entropy () of decision sets as a neural correlate ( uncorrected, ). MNI Coordinates indicated by intersecting blue lines are \[12 18 42\] and correspond to global maxima correlation, which is located at the right MCC. A second cluster located at left thalamus was also found. [Figure S3](#pone.0017408.s003){ref-type="supplementary-material"} shows coronal and sagital views of the two clusters. :::  ::: ::: {#pone-0017408-t002 .table-wrap} 10.1371/journal.pone.0017408.t002 Table 2 ::: {.caption} ###### Neural substrate of uncertainty. ::: {#pone-0017408-t002-2} Lobe Anatomical area Side MNI coordinates t-value Cluster (size) ---------- ------------------------- ------ ----------------- --------- ---------------- ------ --------- Frontal Middle Cingulate Cortex R 12 18 42 7.72 1 (674) pre-SMA L,R −8 8 52 6.07 1 (674) Superior Medial Gyrus L −6 24 38 5.45 1 (674) Thalamus MD nucleus and VA L −10 −8 −2 4.04 2 (10) Height threshold: t-value = , uncorrected. Extent threshold: voxels. Clusters codifying uncertainty of the economic decision task, i.e., with BOLD signal positively correlated with the entropy of the decisions produced at each task configuration. ::: Functional integration of uncertainty {#s2c} ------------------------------------- Enhanced connectivity was detected by means of PPI analyses seeded in those areas of the largest cluster that codified uncertainty. This cluster includes MCC(right) and pre-SMA(left and right). A conjunction analysis was applied to the pre-SMA PPIs in order to obtain common increased connectivities to both pre-SMAs. These analyses (see [methods](#s4){ref-type="sec"} for a detailed explanation) allowed us to identify areas that gained connectivity with MCC(right) or with pre-SMA(bilateral) as the decision uncertainty increased in the economic decision making task. The PPI seeded in the MCC(right) identified clusters located in *frontal lobe* -left middle frontal gyrus, left superior frontal gyrus, superior medial gyrus (bilateral) and left middle orbital gyrus-, and *temporal lobe* -middle temporal gyrus (bilateral)-. A summary of areas is listed in [Table 3](#pone-0017408-t003){ref-type="table"} and shown in [Figure 4A](#pone-0017408-g004){ref-type="fig"}. Clusters are listed in [Table S3](#pone.0017408.s007){ref-type="supplementary-material"}. ::: {#pone-0017408-g004 .fig} 10.1371/journal.pone.0017408.g004 Figure 4 ::: {.caption} ###### Functional integration of uncertainty. A. Clusters that increment functional connectivity with right MCC as entropy increases (PPI analysis). B. Clusters that increment functional connectivity with pre-SMA(bilateral) as entropy increases. This is the result of a conjunction analysis of the PPIs seeded in left and right pre-SMAs. The psychological variable was the Shannon entropy ( of the decisions associated to each task configuration. :::  ::: ::: {#pone-0017408-t003 .table-wrap} 10.1371/journal.pone.0017408.t003 Table 3 ::: {.caption} ###### Functional integration of uncertainty focused on right MCC. ::: {#pone-0017408-t003-3} Lobe Anatomical area Side MNI coordinates t-value ---------- ------------------------ ------ ----------------- --------- ----- ------ Frontal Middle Frontal Gyrus L −24 24 46 5.35 Superior Medial Gyrus L,R 10 62 2 5.20 Middle Orbital Gyrus L −8 54 −2 5.06 Superior Frontal Gyrus L −16 30 40 4.68 Temporal Middle Temporal Gyrus L,R 62 −6 −24 5.91 Height threshold: t-value = , uncorrected. Extent threshold: voxels. MNI coordinates of seed at right MCC: \[12 18 42\]. Areas showing functional connectivity with right MCC as the entropy increases (PPI analysis). The psychological variable was the Shannon entropy of the decisions associated to each task configuration. ::: The conjunction analysis of PPIs seeded in left and right preSMAs identified areas located in *frontal lobe* -MCC (bilateral), paracentral lobule (bilateral), left middle frontal gyrus, left anterior cingulate cortex (ACC), right superior medial gyrus, right precental gyrus and left superior frontal gyrus -, *temporal lobe* - right temporal pole, middle temporal gyrus (bilateral) and right superior temporal gyrus -, *parietal lobe* - right precuneus, right superior parietal lobule and right postcentral gyrus -, *subcortical structures* - left caudate nucleus and left anterior thalamus -, *cerebelum*- left VIII -, and *insular* - right posterior insula -. A summary of areas is listed in [Table 4](#pone-0017408-t004){ref-type="table"} and shown in [Figure 4B](#pone-0017408-g004){ref-type="fig"}. Clusters are listed in [Table S4](#pone.0017408.s008){ref-type="supplementary-material"}. ::: {#pone-0017408-t004 .table-wrap} 10.1371/journal.pone.0017408.t004 Table 4 ::: {.caption} ###### Functional integration of uncertainty focused on bilateral pre-SMA. ::: {#pone-0017408-t004-4} Lobe Anatomical area Side MNI coordinates t-value --------------- --------------------------- ------ ----------------- --------- ----- ------ Frontal Superior Medial Gyrus L,R 10 60 4 4.12 Precentral Gyrus R 46 −14 62 4.11 Middle Cingulate Cortex L,R −2 −28 46 3.81 Anterior Cingulate Cortex L −4 50 0 3.83 Paracentral Lobule R 14 −32 50 3.64 Middle Frontal Gyrus L −20 22 44 3.62 Superior Frontal Gyrus L −18 14 48 3.62 Temporal Temporal Pole R 62 4 −2 4.18 Middle Temporal Gyrus R 60 −12 −20 3.94 Superior Temporal Gyrus R 58 −60 22 3.80 Parietal Precuneus R 12 −46 62 4.51 Superior Parietal Lobule L −26 −46 66 3.84 Postcentral Gyrus L −26 −32 72 3.82 Insular Posterior Insula R 40 −6 −8 4.34 Cerebellum VIII L −20 −46 −56 5.03 Basal ganglia Caudate Nucleus L −6 14 −4 4.31 Thalamus Pulvinar L −14 −34 2 3.80 Height threshold: t-value = , uncorrected. Extent threshold: voxels. MNI coordinates of seed at left pre-SMA: \[−8 8 52\]. MNI coordinates of seed at right pre-SMA: \[6 12 54\]. Areas showing functional connectivity with pre-SMA as the entropy increases. This is the result of a conjunction analysis of the PPIs with seeded in left and right pre-SMAs. The psychological variable was the Shannon entropy of the decisions associated to each task configuration. ::: These results indicate that, as the entropy associated to the decision making task increases, clusters mostly belonging to the associative cortex and located at frontal, temporal and parietal lobes get involved in the process by means of an increased coupling with right MCC or with pre-SMA(bilateral); areas whose activity is codifying the uncertainty. Tables with individual seed coordinates based on the Montreal Neurological Institute (MNI) template can be found in [Table S5](#pone.0017408.s009){ref-type="supplementary-material"} for MCC(right), [Table S6](#pone.0017408.s010){ref-type="supplementary-material"} for pre-SMA(left) and [Table S7](#pone.0017408.s011){ref-type="supplementary-material"} for pre-SMA(right). Discussion {#s3} ========== In this work, Shannon entropy was obtained from the decision outcomes of a group of subjects during an economic decision task under different configurations parametrized by reward probability and time before getting a monetary reward. Among the three different entropy descriptors evaluated, which combines inter and intra-variabilities was the better predictor of the response times. Hence this descriptor was chosen as neural correlate and identified two clusters codifying uncertainty. Although it could be argued that entropies based on either only inter- () or only within- () subject variability of decisions would be more intuitive neural correlates, our analyses indicated that response times were slightly better predicted by this pooled approach. Furthermore, this descriptor had a significant contribution to explain response times even for the case of consistent responses, where is necessarily zero. This finding indicates that some subjects are consistent in their decisions within task configurations even when the decision becomes difficult and a longer response time is required. In this sense, the desirable tendency to keep self-consistency in responses would be a plausible explanation for this behavior which prevents analysis based on individual outcomes to be the the most appropriate option. Two clusters whose BOLD activity correlated with the uncertainty associated to each task configuration were found. In particular, a positive linear correlation was found between the activity of these clusters and the Shannon entropy of the pooled decisions reported at each task configuration. A large cluster containing parts of pre-SMA (bilateral) and right MCC and a small cluster located at thalamus (L) were found (see [Table 2](#pone-0017408-t002){ref-type="table"}). These results provide evidence that pre-SMA may cooperate with MCC in codifying and processing uncertainty in decision making. Previous studies have associated medial and or anterior parts of cingulate cortex to decision conflict monitoring and processing. This was obtained by means of activity contrasts between tasks of high and low difficulty guessing [@pone.0017408-Elliott1], high and low conflict measured at the group level [@pone.0017408-Pochon1] or high and low congruency [@pone.0017408-Fan1] tasks. Our study contributes to better define the modulation of MCC activity in decision making. Rather than obtaining an increased activity in high uncertainty task configurations with respect to low ones, we found that uncertainty is codified within the activity of a cluster that includes right MCC and pre-SMA. In our experiment, the DMC2 contrast showed a significantly higher activity in this cluster (see [Figure S2](#pone.0017408.s002){ref-type="supplementary-material"}). This result indicates that the activity codifying uncertainty in both right MCC and pre-SMA(bilateral) is not reflecting motor actions. Furthermore, the connectivity analysis at pre-SMA(bilateral) mainly identified associative areas. Pre-SMA has been implicated in the resolution of conflict, most commonly characterized as an interference between competing motor plans [@pone.0017408-Isoda1], [@pone.0017408-Nachev1]. There is a remarkable difficulty to differentiate motor conflict and decision conflict contributions of this area. While Pochon et al.[@pone.0017408-Pochon1] aimed to uncouple decision conflict from motor conflict and identified a cluster (with similar coordinates to our cluster in [Table 2](#pone-0017408-t002){ref-type="table"}), Fortsmann et al. [@pone.0017408-Fortsmann1] reported that both right pre-SMA and right anterior striatum facilitate fast actions during a decision-making under time pressure. A second cluster was found at the left anterior thalamus. According to the thalamic connectivity atlas [@pone.0017408-JohansenBerg1], is likely to be connected with the pre-frontal cortex (the reported probability was ). Two connectivity analyses were carried out to search those areas that gained functional connectivity with right MCC and pre-SMA(bilateral) as the entropy of the task configuration increases. MCC showed functional connectivity with clusters located at the associative cortex within areas ( at the frontal lobe and at the temporal lobe). Functional connectivity between the cingulate cortex and frontal and motor areas in a experiment of high versus low congruency has been pointed out [@pone.0017408-Fan1]. We report here functional connectivity of Pre-SMA(bilateral) with right MCC and with out of its functionally connected areas. In addition, pre-SMA(bilateral) was functionally connected with clusters located in parietal, occipital, and subcortical areas, including well known decision making areas such as MCC and ACC. The insular lobe is also considered to play a key role in emotional decision-making, by means of its reciprocal connectivity with the vmPFC [@pone.0017408-Augustine1], [@pone.0017408-Ongur1], and with the ventral striatum and amygdala [@pone.0017408-Reynolds1]. In particular, the posterior insula cortex together with the left caudate nucleus and with the left putamen activity has been associated to choosing delayed relative options instead of immediate rewards [@pone.0017408-Wittmann1]. Right middle temporal gyrus is functionally connected with both MCC (cluster ) and pre-SMA (cluster ). This area has been associated to the action of finding an insight solution to a problem [@pone.0017408-AzizZadeh1], [@pone.0017408-JungBeeman1]. Left cerebellum VIII was functionally connected to pre-SMA, which has been associated to sensorimotor representation and control [@pone.0017408-Stoodley1]. Activity at the dorsolateral and orbital prefrontal cortices have been associated to ventral striatum and thus related to reward and impulsiveness. In particular, they have been implicated in solving decisions under uncertainty [@pone.0017408-Huettel1]--[@pone.0017408-Paulus2]. While activity of these areas was not correlated with Shannon entropy of the decisions, in our experiment they were functionally connected to right MCC and pre-SMA(bilateral). Two possible interpretations can be extracted from a PPI analysis. On the one hand, a psycho-physiological interaction can be seen as a change in the contribution of one area to another due to a change in the psychological variable or context. On the other hand, it can be interpreted as a differential response of an area to the psychological variable which depends on the contribution of a second area [@pone.0017408-Friston1]. In our case the latter possibility would mean that the more active MCC and pre-SMA are, the more sensitivity of associative areas to depict uncertainty. The relationship between probability and Shannon entropy is non linear ([Figure 1C](#pone-0017408-g001){ref-type="fig"}), being entropy more sensitive as probability reaches its minimum or maximum values. Single-neuron recordings in macaque studies have provided preliminary evidence that such a property may better fit the behavior of dopaminergic neurons under uncertainty paradigms [@pone.0017408-Fiorillo1], [@pone.0017408-McCoy1]. In our experiment entropy was used as an uncertainty modulator based on the relative frequency of decision outcomes. This approach has permitted to identify the brain areas that codify decision uncertainty and their functional connectivity with other (mainly associative) areas. Therefore Shannon entropy and other information theory measurements should be taken into account as suitable descriptors in cognitive experiments where magnitudes such as conflict, difficulty or uncertainty are aimed to be quantified. Materials and Methods {#s4} ===================== Subjects {#s4a} -------- Fifteen undergraduate or graduate students from the University of Navarra were recruited as volunteers for the study. There were seven males and eight females and the mean age was years old (SD ). In order to exclude subjects with a current episode or with history of neurological or psychiatric illness, all the volunteers were assessed using The Mini International Neuropsychiatric Interview (MINI) [@pone.0017408-Sheehan1] and interviewed about their clinical history by a psychiatrist. Ethics statement {#s4b} ---------------- The protocol was approved by the ethical committee of the University of Navarra Hospital. Subjects provided written informed consent before entering the scanner. Experimental setup {#s4c} ------------------ Participants laid supine head first inside the scanner with a four button response box on their abdomen. The middle and index fingers of the right hand and their corresponding buttons were used to choose answers. Experimental stimuli were projected to a mirror over the subject\'s eyes. Stimulus presentation and response collection were controlled using Cogent 2000 (Wellcome Department of Imaging Neuroscience, UCL, London, UK) and Matlab (The Mathworks, Natick, MA). Task {#s4d} ---- Inside the scanner participants had to repeatedly choose between two options (A and B), which were visually presented. This experiment, including the participants recruitment and the task configurations, was designed according to the probability-time trade-off model within a particular range of paired configurations of probability and time [@pone.0017408-Baucells1]. Each option consisted of an amount of money ( euros) to be received some time in the future with a specific probability. Time () and reward probability () were varied in option B from to months and from to , in intervals of month and respectively. Option A was maintained in all cases as { euros,, month} (see [Figure 1A](#pone-0017408-g001){ref-type="fig"}). Two additional task configurations, with option A not being constant, were formed by { euros,,*now*},{ euros,,*now*} and { euros,, month},{ euros,, months} respectively. Therefore different task configurations were shown to each participant. Subjects had to choose the option that they considered more attractive using the button box. They were instructed that there were no correct or incorrect answers, and they were not explicitly asked to minimize the time to answer. The time limit to make the decision was fixed to seconds and a white cross was presented one second before the end. There were two control tasks which were interleaved in the presentation (see [Figure 1B](#pone-0017408-g001){ref-type="fig"}). They were designed to use almost identical sensory stimulation and required the same motor activity as the decision-making tasks (DM). There were two control tasks. In the attentional control task (C1), subjects had to sum the numbers in each of the options and press the button for the option with the highest result. In the motor control task (C2), options had no numbers but symbols instead (see [Figure S4](#pone.0017408.s004){ref-type="supplementary-material"}). Subjects were asked to press alternatively one of the buttons every time this control task appeared. C1 was used to check the attentional level of participants and C2 was used to produce a DMC2 activation map to see areas with activity significantly higher at DM, i.e., activity involved in the decision making task that is not due to motor actions (see [Figure S2](#pone.0017408.s002){ref-type="supplementary-material"}). The three conditions (DM,C1 and C2) were grouped in blocks of three trials (see [Figure 1B](#pone-0017408-g001){ref-type="fig"}), alternating the experimental task with one of the two control tasks. Three -minute scanning sessions were carried out. Sessions consisted of the different experimental choices which were repeated or times in a pseudo-random order and the same number of control tasks. Therefore there were twice the number of experimental presentations than of control tasks. Across the sessions every task configuration was presented times. Subjects were awarded with a fixed payment of euros. Additionally, they were told that after the experiment one of their choices inside the scanner would be randomly selected and they would have the possibility of receiving the euros with the elected probability and delay. This extra reward was given in order to motivate participants for the task. Scanning procedure {#s4e} ------------------ The fMRI protocol was carried out with a Tesla MR imager (Siemens TRIO, Erlangen, Germany) with a twelve channel head coil. volumes were acquired in every session using T2\*-weighted gradient echo-planar imaging (EPI) sequences ( axial slices; slice thickness =  mm; ; ms; ms; image resolution =  ; FOV =  ; ). Each time series comprised or repetitions of the decision-making condition and or repetitions of each control condition (C1, C2). The anatomical image was isotropic. A T1-weighted MPRAGE sequence (, , , , , slices) was used for its acquisition. Data processing {#s4f} --------------- Data were analyzed using Statistical Parametric Mapping program software (SPM), version (Wellcome Department of Imaging Neuroscience, UCL, London, UK) in Matlab. For each subject, all EPI volumes were realigned to the first volume of the time series, corrected for differences in the image acquisition time, co-registered with the structural image and spatially normalized into the Montreal Neurological Institute (MNI) template. Finally, a Gaussian smoothing kernel of full-width at half maximum was applied to the EPI images. Response times {#s4g} -------------- Overall, responses consisted of the subjects answering times to each of the task configurations. Response times (RT) of each of these answers were measured in milliseconds. The average response time per subject was used as a speed response descriptor of each of the participants. The average response time per subject per task was used to characterize the mean time required by each participant to answer each task configuration and thus contains values. Shannon entropy {#s4h} --------------- In information theory, Shannon entropy [@pone.0017408-Shannon1] is a measure of the uncertainty associated with a random variable, usually expressed in bits. Its rationale is based on quantifying the amount of information contained in a message. In a more general perspective, the entropy of a discrete random variable with possible values is where is the probability of being exactly equal to (the probability mass function). In the case of empirical data, can be estimated by the percent of times that the discrete random variable equaled . The entropy range of values for any discrete random variable goes from up to and hence to when is used. In the case of a binary variable, the maximum entropy is and corresponds, for example, to the entropy of the sequence of outcomes expected when flipping a perfect coin. Strictly based on the responses obtained, three different entropies , and were measured focused on the within-, inter- and pooled-variabilities of the decisions respectively. measured, for each task configuration and for each subject, the uncertainty of the 5 decisions given. measured, for each task configuration, the uncertainty of the preferred decision of subjects (i.e. the most common response provided among the responses given). Finally, measured, for each task configuration, the uncertainty produced when merging both within- and inter-subject decisions, i.e., when measuring the entropy of the decisions reported for each task configuration. Every task configuration produces a sequence of binary outcomes that contain the choices made between the constant option A and the alternative option B. Let us define as a task configuration with option A constant and with option B defined by probability and time . In the particular case of , all the decisions made by the participants for each task configuration constitute a dichotomous random variable . The relative frequencies of choices made by participants at each are denoted by and respectively. Hence, according to the equation proposed by C. Shannon, the entropy of each decision variable associated to a task configuration can be defined as . The entropy of a dichotomous random variable is maximal at and is highly non linear, being very sensitive to small probability changes near the extremes (when is close to or to ). In the case of , every decision set is a random variable consequence of decisions, since every presentation was shown for times to each of the participants. However, in some task configurations where one decision was not made, the entropy was measured according to decisions. The entropy model for dichotomous random variables and the values of obtained for the task configurations are shown in [Figure 1C](#pone-0017408-g001){ref-type="fig"}. Multi-linear regression analyses {#s4i} -------------------------------- Three multi-linear regression models were evaluated in order to select which of the three entropies (, and ) would be used for the neuroimaging analyses. The dependent variable was in all cases . One of the independent variables was , which controlled the possible effects of faster/slower participants. The second independent variable was one of the entropies on each model. For each model, the standardized coefficients of each dependent variable (, ) and the statistic were used to evaluate the goodness of fit were measured. It was hypothesized that, for certain task configurations, a subject could choose every time the same option not only due to a low level of difficulty found but also due to factors such as maintaining self-coherence along the experiment. In this sense, self-coherent answers could still be masking a high level of cognitive conflict. In order to prove such hypothesis one further analysis was carried out analyzing only the of those answers that conformed consistent decisions per subject per task. Influence of independent variables was considered to be significant with . Neural correlate analysis of uncertainty {#s4j} ---------------------------------------- Individual task-related activation was evaluated in a first step using a general linear model. Considering that RT distribution was , each condition (DM, C1 and C2) was evaluated as event related using a delta function convolved with the hemodynamic response function (canonical HRF). The entropy based modulator () was introduced in the analysis in a later step. This regression model was used to test the areas which showed a positive linear correlation between their BOLD signal and the regressor. Finally, in order to make inferences at the population level, individual contrast images were incorporated into a random effects model [@pone.0017408-Strange2], [@pone.0017408-Penny1]. The statistical significance was set at (uncorrected for multiple comparisons). Areas were named according to atlas provided by the SPM anatomy toolbox [@pone.0017408-Eickhoff1]. Functional integration analysis of uncertainty {#s4k} ---------------------------------------------- Analysis of functional connectivity assesses the hypothesis that activity in one brain region can be explained by an interaction between the presence of a cognitive process and activity in another part of the brain. In particular, we used the psycho-physiological interactions (PPI) method [@pone.0017408-Friston1] to estimate functional connectivity with three sources or seeds (MCC(right), pre-SMA(left) and preSMA(right)) during a decision making task whose configurations were labeled by their Shannon entropy (). The PPI method is an exploratory multi-regression analysis [@pone.0017408-Stephan1] which includes 4 terms. The psychological variable (Shannon entropy of each task configuration in our case) is the task regressor, the time series of a region (seed) is the physiological variable, a bilinear term formed by the element-by-element product of the task regressor and the seed time series compound the PPI regressor and finally a constant fourth term. The analysis procedure was performed based on [@pone.0017408-Stephan2]. For each subject, three local maxima corresponding to pre-SMA(left and right) and MCC(right) were determined using the individual map obtained from the DMnull contrast (coordinates are shown in [Tables S6](#pone.0017408.s010){ref-type="supplementary-material"}, [S7](#pone.0017408.s011){ref-type="supplementary-material"} and [S5](#pone.0017408.s009){ref-type="supplementary-material"} respectively). The individual time series for each seed region were obtained by extracting the first principal component from the raw BOLD time series in a spherical ROI ( radius) centered on the coordinates of each subject specific local maximum. In a later step, individual level analyses with a separate condition for each task configuration were performed. Within this design, the interaction term (PPI regressor) was estimated. It was computed as the element-by-element product of the time series (for each seed separately) and a Shannon entropy vector coding the uncertainty associated to each task configuration (task regressor). The PPI regressor, the task regressor (psychological variable), the seed time series (physiological variable) and the constant term were introduced as regressors in a first level analysis. At the individual level a -contrast was created using the PPI regressor exclusively. These contrast images were entered into a random effects model [@pone.0017408-Strange2], [@pone.0017408-Penny1], followed by a one-sample t-test. The resulting maps were thresholded at and . In the case of pre-SMA, a one-way within subject ANOVA with the factor seed, pre-SMA(left) and pre-SMA(right), was performed. Subsequently a conjunction analysis with a conjunction null hypothesis was carried out to find areas common to the two connectivity group maps, i.e., pre-SMA(bilateral). Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Individual entropy maps of** **.** Entropy produced by the answers reported by each subject to each task configuration. X-axis and Y-axis respectively denote the reward probability and the time to wait of option B. A bilinear interpolation process was applied to the the actual time and probability values evaluated. Color gradient represents the entropy values from (dark blue) to (red). Those maps of subjects with more areas in dark blue correspond to highly self-consistent participants along the whole experiment (e.g. subjects , and ). (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **contrast task activation.** Blue solid lines indicate MCC(right) with MNI coordinates \[12 18 42\]. The cluster involving this location contains the largest cluster found to codify decision entropy (cluster at [Table 1](#pone-0017408-t001){ref-type="table"} in the manuscript). Therefore neither the activity magnitude nor the activity modulation (correlate with decision entropy) are explained by motor actions. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Areas positively correlated with the entropy values of** **(** **,** **).** Top. Coronal views. Bottom. Sagital views. The four anatomical regions involved were: middle cingulate cortex (right), pre-supplementary motor area (bilateral), superior medial gyrus (right) and thalamus (left). (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Left. Slide corresponding to motor control (C2) events.** Subjects were asked to press alternatively one of the buttons every time this control task appeared. **Right Example of the decision making task (DM).** Subjects were asked to choose the economic option considered more attractive. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **: Shannon entropy (bits) for each task configuration** **.** Values are based on the inter-subject variability of the decisions. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **: Shannon entropy (bits) for each task configuration** **.** Values are based on the pooled variability (inter- and within-subject) of the decisions. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Clusters showing functional connectivity gain with MCC(right) as the entropy increases (PPI analysis).** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### **Clusters showing functional connectivity gain with pre-SMA(bilateral) as the entropy increases.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### **Individual seeds for PPI analysis at the MCC(right).** This table specifies the MNI coordinates used for each subject at MCC(right) and their individual t-values in the DMC2 contrast. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S6 ::: {.caption} ###### **Individual seeds for PPI analysis at the Pre-SMA (left).** This table specifies the MNI coordinates used for each subject at Pre-SMA(left) and their individual t-values in the DMC2 contrast. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S7 ::: {.caption} ###### **Individual seeds for PPI analysis at the Pre-SMA (right).** This table specifies the MNI coordinates used for each subject at Pre-SMA(right) and their individual t-values in the DMC2 contrast. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to acknowledge Rafael Franco, Federico Villagra and Marta Vidorreta for useful discussions and comments. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded by FIMA of the University of Navarra, IESE Business School of the University of Navarra and by CIBERNED. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MAS MFS FH MAP. Performed the experiments: MAS FRL MFS. Analyzed the data: JG MAS GA. Contributed reagents/materials/analysis tools: JG MAS GA MFS FH MAP. Wrote the paper: JG GA MAS MAP. Coupled an information theory paradigm with the decision making experiment: JG.

PubMed Central

2024-06-05T04:04:19.134370

2011-3-9

PMC3052309

Introduction {#s1} ============ To deter herbivores, plants have evolved a broad range of defense mechanisms that can be generalized into two categories: pre-formed constitutive defenses and inducible defenses [@pone.0017195-Chen1]. Constitutive defenses include the physical and chemical barriers that exist before insects attack, whereas induced defenses includes direct and indirect defenses. Direct defenses are plant traits that by themselves affect the susceptibility of host plants to insect attacks. Indirect defenses, on the other hand, include plant traits that by themselves do not affect the susceptibility of host plants, but can serve as attractants to natural enemies of attacking insects [@pone.0017195-Maffei1]--[@pone.0017195-Kessler1]. Insects may respond to plants by choosing different feeding sites, by altering their consumption rates or by induction of physiological/detoxification enzymes [@pone.0017195-Scriber1]. Approximately 90% of herbivorous insects have narrow host ranges, feeding on plants within a single taxonomic family, and many species are confined to a single host species [@pone.0017195-Strong1]. The variability in fitness on different host plant species favors behavioral genotypes that restrict feeding to the most suitable hosts; on the other hand, maintaining mechanisms that nullify the disparate defensive adaptations of many different plant species is too costly for generalist herbivores [@pone.0017195-Rausher1]. Thus, the key to understanding why certain herbivores remain specialized lies in the observation that specialization involves both behavioral and physiological adaptation [@pone.0017195-Scriber1], [@pone.0017195-Cogni1], [@pone.0017195-Smiseth1]. The defense strategy in aromatic plants like *Mentha aquatica* is a direct defense, through the constitutive production of terpenoids in specialized tissues known as the glandular trichomes [@pone.0017195-Maffei1]. These plants may have chemical barriers to potential herbivore colonists, and they appear to be accessible to relatively few insect lineages, which may be pre-adapted to chemically similar or related host plants [@pone.0017195-Farrell1]. As some insects become adapted to these metabolites, interactions between the two groups of organisms occasionally lead to highly specific relationships, as in the case of *M. aquatica* and the herbivore *Chrysolina herbacea*. *M. aquatica*, or watermint, is a perennial plant belonging to the Lamiaceae. It produces leaf glandular trichomes secreting volatile organic compounds (VOCs) of varying chemical composition [@pone.0017195-Malingr1], [@pone.0017195-Jerkovic1]. The presence of the oxygenated monoterpenes (+)-pulegone and (+)-menthofuran contributes to the plant\'s toxicity. Both viridiflorol from the essential oil and (S)-naringenin from an ethanolic extract have been isolated by bioassay-guided fractionation with binding to the GABA-benzodiazepine site. Furthermore, *M. aquatica* contains psychoactive compounds that display both monoamine oxidase-inhibitory activity and mitochondrial respiration uncoupling [@pone.0017195-Jager1]--[@pone.0017195-Mucciarelli1]. At least 24 species of insect herbivores have been observed feeding on *M. aquatica* [@pone.0017195-Warren1], [@pone.0017195-Horwood1], and among these, *C. herbacea*, also known as the mint beetle, is quite diffuse in mint fields. The feeding behavior of this beetle has been described recently [@pone.0017195-Bienkowski1]. Both larvae and adult beetles attack the leaves. Leaf beetles like *C. herbacea* are also known for their ability to import structurally distinct allelochemicals (reviewed by [@pone.0017195-Burse1]). The ability to produce deterrents to natural enemies from plant-derived compounds is typical of some *Chrysolina* species [@pone.0017195-Pasteels1]--[@pone.0017195-Laurent2]. Because herbivore feeding alters the aromatic profile of essential oil-producing plants like *M. aquatica*, the issue is both ecologically and economically relevant [@pone.0017195-Banchio1]--[@pone.0017195-Valladares1]. In this work we describe the chemical interaction between *M. aquatica* and *C. herbacea* by evaluating the ability of the herbivore to locate and recognize plant chemical cues and the capacity of the host plant to respond to herbivory by emitting deterrent molecules. To this end, plant VOC emissions we analyzed before and after herbivore feeding, and the ability of the emitted molecules to attract or deter *C. herbacea* was tested by bioassay. Results {#s2} ======= *C. herbacea* responds to *M. aquatica* VOC emission {#s2a} ---------------------------------------------------- Successful co-adaptation between plants and insects requires that plants produce specific compounds in response to herbivory, and that insects respond adequately to molecules emitted by plants. To look for possible relationships between *M. aquatica* VOC emissions and herbivore responses, the behavior of *C. herbacea* was first evaluated in Y-tube olfactometry tests. *C. herbacea* was found to be attracted to undamaged plants, with respect to pure air. When the choice was between infested plants, undamaged plants or air, the insect was found to be deterred by infested plants ([Fig. 1](#pone-0017195-g001){ref-type="fig"}). Furthermore, the insects were found to lay eggs on undamaged plants and they produced larvae capable of surviving successive instars up to the adult phase (see Supporting [Fig. S3](#pone.0017195.s003){ref-type="supplementary-material"}). Preliminary studies demonstrated that the qualitative response of *M. aquatica* to larvae feeding was not significantly different from that of adult insects. For this reason, the study focused on adult insects. ::: {#pone-0017195-g001 .fig} 10.1371/journal.pone.0017195.g001 Figure 1 ::: {.caption} ###### Response of *C. herbacea* in a Y-tube olfactometer when offered to *M. aquatica* cuttings of undamaged and *C. herbacea*-infested leaves versus pure GC-grade air. A χ^2^ test served to evaluate differences from a 50∶50 distribution over two olfactometer arms. Insects that did not reach the end of either olfactometer arm within 10 min (NC: no choice) are indicated by white bars. The asterisks indicate significant (P\<0.05) differences. :::  ::: *M. aquatica* reacts to *C. herbacea* herbivory by emitting specific deterrent molecules {#s2b} ---------------------------------------------------------------------------------------- The reason why *C. herbacea* was differentially attracted to undamaged and infested plants was examined. VOCs emitted by untreated *M. aquatica* were analyzed by SPME, which revealed the presence of two major compounds, (+)-pulegone and (+)-menthofuran ([Fig. 2a](#pone-0017195-g002){ref-type="fig"}), along with other minor terpenoids including the monoterpenes (−)-limonene, (−)-menthone and α-terpineol, and the sesquiterpene, (*E*)-β-caryophyllene ([Fig. 2b](#pone-0017195-g002){ref-type="fig"}). However, it is important to note that the major compounds were emitted at much higher levels (0.13--0.75 µg g^−1^ f. wt) than the minor compounds in the range of 3--35 ng g^−1^ f. wt. The feeding activity of *C. herbacea* significantly changed the quantitative VOC composition of *M. aquatica* VOC emissions by dramatically increasing the content of (+)-menthofuran and decreasing the content of (+)-pulegone ([Fig. 2a](#pone-0017195-g002){ref-type="fig"}). The levels of all minor compounds, with particular reference to myrcene, were also elevated significantly upon herbivory; with the sole exception of *p*-cymene, which declined significantly ([Fig. 2b](#pone-0017195-g002){ref-type="fig"}). Since herbivory leads to the rupture of glandular trichomes, which are the main storage tissues of terpenoids in the Lamiaceae [@pone.0017195-Maffei1], [@pone.0017195-Maffei2], whether mechanical injury alone was able to increase the VOC emissions in *M. aquatica* was evaluated. Surprisingly, leaves damaged mechanically by a pattern wheel and having the same extent of herbivore damage had lower emissions of both major and minor compounds in comparison to control plants, particularly in comparison to herbivore wounded leaves ([Figs. 2a and 2b](#pone-0017195-g002){ref-type="fig"}). ::: {#pone-0017195-g002 .fig} 10.1371/journal.pone.0017195.g002 Figure 2 ::: {.caption} ###### VOC emission by *M. aquatica* in undamaged plants (control), in response to mechanical damage caused by a pattern wheel (MD) and after herbivory by *C. herbacea* (HW). **a**, Content of the major components; **b**, content of the minor components. Bars indicate the standard error over the mean of at least three biological replicates. Asterisks indicate significant differences with respect to controls (P\<0.05). :::  ::: *C. herbacea* responds differentially to specific monoterpenes emitted by *M. aquatica* {#s2c} --------------------------------------------------------------------------------------- Following assessment of the chemical composition of the major VOCs released by undamaged and infested *M. aquatica*, the responses of *C. herbacea* to the main monoterpenes (+)-menthofuran and (+)-pulegone were tested. *C. herbacea* was significantly attracted to (+)-pulegone when the choice was limited to pure air, infested plants or (+)-menthofuran, whereas the insect preferred undamaged plants when offered with (+)-pulegone. On the contrary, (+)-menthofuran was always found to repel the insects, no matter which choice test was performed ([Fig. 3](#pone-0017195-g003){ref-type="fig"}). To evaluate whether the synthetic mixture composed of both major and minor compounds identified from leaf volatiles affected *C. herbacea* behavior, insect preference was determined by performing choice tests with several comparisons. *C. herbacea* was significantly attracted to the synthetic mixture typical of uninfested plants when compared to both air and a synthetic mixture typical of infested plants ([Fig. 3](#pone-0017195-g003){ref-type="fig"}). When only minor compounds of synthetic mixtures were compared to either air or other mixtures, no significant difference could be found ([Fig. 3](#pone-0017195-g003){ref-type="fig"}). ::: {#pone-0017195-g003 .fig} 10.1371/journal.pone.0017195.g003 Figure 3 ::: {.caption} ###### Response of *C. herbacea* in a Y-tube olfactometer when offered to either Sigma-grade (+)-pulegone, (+)-menthofuran, synthetic mixtures of compounds with the same content of typical intact plant emissions (Synt Mix Plant), synthetic mixtures of compounds with the same content of typical infested plant emissions (Synt Mix Infested Plant), synthetic mixtures of only minor compounds with the same content of typical intact plant emissions (Synt Minor Plant), synthetic mixtures of only minor compounds with the same content of typical infested plant emissions (Synt Minor Infested Plant) or GC-grade air. A χ^2^ test served toevaluate differences from a 50∶50 distribution over two olfactometer arms. Insects that did not reach the end of either olfactometer arm within 10 min (NC: no choice) are indicated by white bars. Asterisks indicate significant differences (P\<0.05). :::  ::: Feeding of *C. herbacea* induces terpenoid gene expression in *M. aquatica* {#s2d} --------------------------------------------------------------------------- Herbivory was found to affect the percentage of some monoterpenes emitted by *M. aquatica*. For this reason, gene expression involved in the biosynthetic pathway leading to the bioactive monoterpenes (+)-pulegone and (+)-menthofuran was evaluated. Previous studies have established the biochemical pathway that in mints that leads to the production of these two important monoterpenes ([Fig. 4](#pone-0017195-g004){ref-type="fig"}) [@pone.0017195-Croteau1]. We considered early genes such as *Dxs* and *Ippi*, which are involved in the formation and isomerization of the precursors isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP), respectively. *Dxp* showed no regulation by either herbivory or mechanical damage, whereas *Ippi* was down-regulated by mechanical damage and up-regulated by herbivory. *Gpps*, the gene coding for the enzyme that condenses one unit each of IPP and DMAPP into the monoterpene precursor geranyl pyrophosphate (GPP), was up-regulated almost 3-fold by herbivory and to a lesser extent by mechanical damage. *Ls*, coding for the enzyme conducting cyclisation of the universal precursor GPP to the parent olefin (−)-limonene, showed the same trend as *Ippi*, being up-regulated by herbivory and down-regulated by mechanical damage ([Fig. 5](#pone-0017195-g005){ref-type="fig"}). Several genes were always up-regulated by herbivory and showed no regulation after mechanical damage ([Fig. 5](#pone-0017195-g005){ref-type="fig"}). The genes included *L3oh*, which codes for the enzyme responsible for the NADPH- and O~2~-dependent hydroxylation of (−)-limonene to (−)-*trans*-isopiperitenol, *Ipd*, which codes the operationally soluble, NAD-dependent isopiperitenol dehydrogenase that catalyzes allylic oxidation to the α,β-unsaturated ketone (−)-isopiperitenone, *Ipr*, which codes the soluble enzyme (−)-isopiperitenone reductase that catalyzes the stereospecific, NADPH-dependent reduction of (−)-isopiperitenone to (+)-(1R,4R)-cis-isopulegone, and *Mfs*, which codes for menthofuran synthase, an enzyme responsible for the transformation of (+)-pulegone to (+)-menthofuran ([Fig. 4](#pone-0017195-g004){ref-type="fig"}). Isopulegone isomerase, which catalyzes double bond migration and causes the isopropenyl double bond of (+)-(1R,4R)-cis-isopulegone to yield (+)-pulegone, has not yet been isolated and was not assayed. Finally, *Pr*, which codes for pulegone reductase, the enzyme responsible for NADPH-dependent reduction of the conjugated double bond of the terpenone to yield (−)-menthone, was down-regulated by herbivory and showed no significant regulation upon mechanical damage ([Figs. 4](#pone-0017195-g004){ref-type="fig"} and [5](#pone-0017195-g005){ref-type="fig"}). ::: {#pone-0017195-g004 .fig} 10.1371/journal.pone.0017195.g004 Figure 4 ::: {.caption} ###### Schematic representation of the monoterpene biochemical pathway in *M. aquatica*. *Dxs*, 1-deoxy-D-xylulose-5-phosphate synthase; *Gpps* indicates geranyl diphosphate synthase; *Ls* indicates (−)-limonene synthase; *L3oh* indicates (−)-limonene-3-hydroxylase; *Ipd* indicates (−)-(3S,4R)-*trans*-isopiperitenol dehydrogenase; *Ipr* indicates (−)-isopiperitenone reductase; *Mfs* indicates (+)-menthofuran synthase; and *Pr* indicates (+)-pulegone reductase. :::  ::: ::: {#pone-0017195-g005 .fig} 10.1371/journal.pone.0017195.g005 Figure 5 ::: {.caption} ###### qPCR terpenoid gene expression of *M. aquatica* in mechanically damaged (MD) and herbivore damaged (HW) leaves, with respect to control leaves. Bars indicate the standard error over the mean of at least three biological replicates. Asterisks indicate significant differences with respect to controls (P\<0.05). :::  ::: Discussion {#s3} ========== Over the past two decades it has been documented that plants produce blends of volatile compounds in vegetative tissues in response to damage and herbivore attack [@pone.0017195-Dicke2], [@pone.0017195-Howe1], [@pone.0017195-Dudareva1]--[@pone.0017195-Heil2], suggesting that these substances act in plant defense [@pone.0017195-Unsicker1]. Several lines of evidence indicate that VOCs released from vegetative tissues act as direct repellents against herbivores [@pone.0017195-Holopainen1] and that the release of VOCs can result from the bursting of pre-existing structures in which volatiles are stored, such as glandular trichomes [@pone.0017195-Par1]. Plant VOCs can also attract natural enemies of attacking herbivores, such as parasitic wasps, flies, predatory mites or birds that can protect the signaling plant from further damage [@pone.0017195-Kessler1], [@pone.0017195-Dudareva1], [@pone.0017195-Mithfer1], [@pone.0017195-Turlings1], [@pone.0017195-Mntyl1] The results of this work show that *C. herbacea* is perfectly adapted to the blend of terpenoids emitted by undamaged *M. aquatica* and it uses this blend as a cue to locate plants. The fact that the insect lays eggs on undamaged plants is further evidence of such adaptation. In the case of the lepidopteran *Plutella xylostella*, the insect does not normally lay eggs on *Chrysanthemum morifolium*, because of the repellence of the monoterpene volatiles emitted from undamaged plants [@pone.0017195-Wang1]. In the attracting blend of terpenoids produced by *M. aquatica*, (+)-pulegone was found to be the major compound in undamaged leaves and a potent attractant to *C. herbacea* in olfactometer bioassays. Volatiles emitted from plants can stimulate the behavioral or antennal responses of herbivores [@pone.0017195-Zhao1]. As a response to herbivore feeding, *M. aquatica* activates genes for terpenoid biosynthesis, diverting most of the terpene production toward the synthesis of (+)-menthofuran, which was found to repel *C. herbacea* in bioassay tests. Over the past decade, evidence that vegetative volatile compounds function to directly repel herbivores has begun to accumulate [@pone.0017195-Unsicker1], [@pone.0017195-Scharf1]. However, the minor compounds emitted by uninfested and infested leaves were not effective in attracting or deterring *C. herbacea*. An open question is why *M. aquatica* produces a herbivore attractant. One possible explanation is that the emission of (+)-pulegone is exploited by plants because of the antimicrobial [@pone.0017195-Mkaddem1], [@pone.0017195-Mora1], nematocidal [@pone.0017195-Ntalli1], acaricidal [@pone.0017195-Ribeiro1], antifeedant [@pone.0017195-Dancewicz1] and mitochondrial respiration uncoupling [@pone.0017195-Mucciarelli1] properties of this compound. Preliminary studies have identified a *C. herbacea* egg parasite and studies on its behavior are under way. This work suggests that constitutive plant defense can be modulated by interactions with herbivorous insects. The latter can trigger plant terpenoid gene expression and synthesis in a way that simple mechanical damage cannot. Usually, mechanical damage to plant foliage elevates VOC release in case of artificial damage carried out by researchers [@pone.0017195-Holopainen1]; however, mechanically damaged *M. aquatica* does not exhibit emissions as intense or with the same compositional pattern as after herbivory. Most genes directly involved in the biosynthesis of *p*-menthane monoterpenes in mints are transcriptionally regulated in a coordinated fashion [@pone.0017195-McConkey1] and it seems likely that the expression of these genes is controlled by a common transcription factor [@pone.0017195-Mahmoud1]. Herbivory had no effect on the expression of *M. aquatica Dxs*, a gene involved in the early steps of terpenoid biosynthesis for the mevalonate-independent (MEP)-pathway gene, the product of which is considered to catalyze one of the rate-limiting steps of this pathway [@pone.0017195-Estevez1]. On the contrary, herbivory up-regulated almost all other genes involved in the pathway. Over-expression of *Gpps*, *Ls* and *L3oh* would be expected to increase production of both GPP and other key monoterpenes [@pone.0017195-Mahmoud1]. The synthesis of (−)-limonene, providing the first committed intermediate of the pathway, is a possible rate limiting step of monoterpene production in mints [@pone.0017195-Croteau1]. Thus, the over-expression of both *Gpps* and *Ls* justifies the increase in the precursor (−)-limonene and the end-products of the pathway (particularly (+)-menthofuran). The up-regulation of *Mfs* parallels the increase in (+)-menthofuran; the latter was found to be a competitive inhibitor of Pr [@pone.0017195-RiosEstepa1]. In transgenic lines with increased expression of *Mfs* and more (+)-menthofuran in the essential oil, (+)-pulegone amounts were greater than in controls. This finding led to the hypothesis that the metabolic fate of (+)-pulegone is controlled by (+)-menthofuran-mediated transcriptional down-regulation of Pr levels [@pone.0017195-Mahmoud2]. In *M. aquatica*, herbivory down-regulated the expression of *Pr* and up-regulated *Ipr*; however, the amounts of (+)-pulegone were never significantly different from those in the controls. This might indicate the presence of other factors that may exert post-translational control over the enzyme activity of Mfs or the regulation of the isopulegone isomerase, which has not yet been isolated. In general, following herbivore attack, plants release green leaf volatiles (GLV), six-carbon aldehydes, alcohols, and esters, that are considered typical wound signals [@pone.0017195-Dicke2], [@pone.0017195-Dudareva1], [@pone.0017195-Frost1]--[@pone.0017195-Matsui1]. By contrast, significant amounts of green-leaf volatiles were not emitted by *M. aquatica* during herbivory. This may be related the high amount of terpenoids produced in the glandular trichomes; by comparison plants that do not accumulate these compounds in secretory tissues (e.g., *Arabidopsis* or Lima bean) usually released smaller amounts of terpenoids [@pone.0017195-Maffei4], [@pone.0017195-Aharoni1]. In some cases it has been suggested that oxidative damage of membranes is one of the primary factors inducing GLV emissions [@pone.0017195-Arimura1]. The function of VOCs in alleviating oxidative stresses has been shown to be related to the high reactivity of certain monoterpenes [@pone.0017195-Holopainen1], suggesting that the release of VOCs upon herbivory in *M. aquatica* might have other functions in addition to deterrence. The proportions of emitted monoterpenes (VOCs) may differ from those in the plant oil of glandular trichomes with regard to chemical composition [@pone.0017195-Banchio1], [@pone.0017195-Gershenzon1]. Nonetheless, molecular data support the hypothesis that terpenoid biosynthesis is modulated by *C. herbacea* herbivory. In conclusion, *C. herbacea* attacks undamaged *M. aquatica*, but it avoids herbivore-infested *M. aquatica*. Upon herbivory, *M. aquatica* produces repellent compounds, thus reducing the damage from further insect attacks ([Fig. 6](#pone-0017195-g006){ref-type="fig"}, see also Supporting [Fig. S4](#pone.0017195.s004){ref-type="supplementary-material"}). ::: {#pone-0017195-g006 .fig} 10.1371/journal.pone.0017195.g006 Figure 6 ::: {.caption} ###### The behavior of *C. herbacea* and *M. aquatica* before and during herbivore feeding. **a**, Undamaged plants emit (+)-pulegone, which acts as an attractant for *C. herbacea*. **b**, Feeding activity induces gene expression and increases content of the deterrent compound (+)-menthofuran, along with the emission of the attractant compound (+)-pulegone; as a result, fewer insects are attracted to the plants. **c**, Intense feeding induces a reduction in (+)-pulegone content and a dramatic increase in the repellent compound (+)-menthofuran; *C. herbacea* avoids over-fed plants and moves towards undamaged plants. :::  ::: Materials and Methods {#s4} ===================== Plant material and growth conditions {#s4a} ------------------------------------ Stolons of *Mentha aquatica* L. were collected from wild populations growing in Cambiano (Turin province, Italy, alt 240 m a.s.l.) and San Secondo di Pinerolo (Turin province, Italy, alt 413 m a.s.l.). Stolons were surface sterilized with 70% ethanol (Sigma-Aldrich, St. Louis, MO, USA) for 20 s and with sodium hypochlorite (1% v/v available chlorine) (Sigma-Aldrich) for 5 min. Stolons were then rinsed three times with sterile distilled water. Plants were grown in plastic pots with sterilized peat and vermiculite (V/V 4∶1) at 23°C and 60% humidity using daylight fluorescent tubes at 270 µE m^−2^ s^−1^ with a photophase of 16 h. Insect collection and rearing {#s4b} ----------------------------- Adults of *Chrysolina herbacea* (Duftschmid 1825) (Coleoptera, Chrysomelidae, Chrysomelinae) were collected by hand picking from infested *M. aquatica* fields. After collection, beetles were reared at 22°C in ventilated glass chambers and fed to *M. aquatica* cuttings. The beetles were starved for 24 h. prior the experiments. Collection of plant volatiles, gas chromatography and mass spectrometry {#s4c} ----------------------------------------------------------------------- Experiments were conducted in 4 l glass desiccators by using non-flowering *M. aquatica* five-node cuttings, placed in 100 ml Erlenmeyer\'s flasks filled with 60 ml tap water and sealed with aluminium foil to prevent fall of insects into the water during experiments. Three cuttings per flask in a single desiccator where used. Plants were illuminated with fluorescent light bulbs generating about 50 µmol m^−2^ s^−1^ with a photophase of 16 h, the temperature inside desiccators was about 24°C and the relative humidity about 70%. Glass desiccators were connected to a GC-grade air generator (HPZA-3500-220, Parker Balston, Cleveland, OH, USA) through a cork plug with two openings allowing gases to go in and out. Air was pumped into the jars at a flow rate of 300 ml min^−1^. A clean Pasteur glass pipette was inserted in the outlet of the cork plug and VOCs were sampled with a Carboxen/Polydimethylsiloxane (CAR/PDMS) Supelco (Bellefonte, PA, USA) solid-phase micro-extraction (SPME) fibre (model 57334-U). Before use, SPME fibres were always conditioned at 250°C, according to manufacturer\'s instructions (see also Supporting [Fig. S1](#pone.0017195.s001){ref-type="supplementary-material"} for more details). Undamaged plants, leaves mechanically damaged with a pattern wheel and plants infested for 6 h were assayed for VOC emission. All experiments were standardized at 6 h, because the presence of eight herbivores for 6 h was found to cause about 30% of leaf damage. SPME fibres, which were placed in various paths to adsorb VOCc for 6 h (see Supporting [Figs. S1](#pone.0017195.s001){ref-type="supplementary-material"} and [S2](#pone.0017195.s002){ref-type="supplementary-material"}), were desorpted and VOCs were analyzed by gas-chromatography mass spectrometry (GC-MS 6890N-5973A, Agilent Technologies, Santa Clara, CA, US). The desorpted compounds were separated on a ZB-5MS Zebron (7HG-G010-11, Phenomenex, Torrance, CA, US) capillary column (stationary phase: polydimethylsiloxane - 5% diphenyl, 30 m length, 250 µm internal diameter, and 0.25 µm film thickness) with a temperature program of 60°C (kept for 5 min) followed by a temperature rise at a rate of 3°C min^−1^ to 270°C (kept for 5 min). Working conditions were: injector 250°C, transfer line to MSD 280°C, oven temperature: start 60°C, hold 5 min, programmed from 60°C to 270°C at 3°C min-1, hold 5 min; carrier gas was He under a constant flow of 1 ml min-1; in SPME desorption and subsequent analysis the fibre was exposed in the injection port during the entire GC run; the injector was maintained in splitless mode during the desorption phase; ionization energy: EI 70 eV; acquisition parameters: scanned m/z 50--250 amu. Separated compounds were identified by pure standard comparison, by comparison of their mass spectra and retention indexes (Kováts indexes) with those of reference substances and by comparison with the NIST mass spectral search software v2.0 using the libraries NIST 98 library and Adams [@pone.0017195-Adams1] library. Different concentraitions of (+)-menthofuran, (+)-pulegone, myrcene, *p*-cymene, (−)-limonene, (−)-menthone, α-terpineol and (*E*)-β-caryophyllene were used to create a standard curve used as an external standard for SPME quantitative measurements. Reference compounds were mixed in relative proportions similar to those that were quantified in the plant samples. Total RNA extraction and quantitative Real Time-PCR (qPCR) {#s4d} ---------------------------------------------------------- After each experiment, leaves were collected and immediately frozen in liquid nitrogen. One hundred mg of frozen control, herbivore damaged, and mechanically damaged leaves were ground in liquid nitrogen with mortar and pestle. Total RNA was isolated using Qiagen RNeasy Plant RNA kit and RNase-Free DNase set (Qiagen, Hilden, Germany). Sample quality and quantity was checked by using the RNA 6000 Nano kit and the Agilent 2100 Bioanalyzer (Agilent Technologies) according to manufacturer\'s instructions. Quantification of RNA was also confirmed spectrophotometrically by using a NanoDrop ND-1000 (Thermo Fisher Scientific, Waltham, MA, US). First strand cDNA synthesis was accomplished with 2 µg of total RNA and random primers using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, US), according to the manufacturer\'s recommendations. Primers for real-time PCR were designed on *Mentha piperita* available sequences using the Primer 3 software [@pone.0017195-Rozen1]. qPCR was done on an Mx3000P Real-Time PCR System (Stratagene, La Jolla, CA, US). The reaction was performed with 25 µl of mixture consisting of 12.5 µl of 2× Maxima™ SYBR Green qPCR Master Mix (Fermentas International, Inc, Burlington, ON, Canada), 0.5 µl of cDNA and 100 nM primers (Integrated DNA Technologies, Coralville, IA, US). Relative RNA levels were calibrated and normalized with the level of two housekeeping genes: actin and 18S ribosomal mRNA. PCR conditions were determined by comparing threshold values in dilution series of the RT product, followed by non-template control for each primer pair. Relative expression levels of genes were calculated by using the Pfaffl method [@pone.0017195-Luciano1]. A suitable melt curve analysis was always performed. PCR conditions were the following: *18S*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 58°C, and 30 s at 72°C; *Actin*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 57°C, and 30 s at 72°C; *Dxs*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 58°C, and 30 s at 72°C; *Ippi*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 57°C, and 30 s at 72°C; *Gpps*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 58°C, and 30 s at 72°C; *Ls*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 57°C, and 30 s at 72°C; *L3oh*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 57°C, and 30 s at 72°C; *Ipd*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 58°C, and 30 s at 72°C ; *Ipr*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 57°C, and 30 s at 72°C; *Pr*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 57°C, and 30 s at 72°C; *Mfs*: initial polymerase activation of 10 min at 95°C; and 40 cycles of 15 s at 95°C, 30 s at 58°C, and 30 s at 72°C. Primers used for qPCR were the following: 18S, (NCBI GenBank accession no. NR\_022795), forward primer 5′-ATGATAACTCGACGGATCGC-3′, reverse primer 5′-CTTGGATGTGGTAGCCGTTT -3′; actin, (NCBI GenBank accession no. AW255057), forward primer 5′-GCTCCAAGGGCTGTGTTCC-3′, reverse primer 5′- TCTTTCTGTCCCATGCCAAC-3′ [@pone.0017195-Lange1]. *Dxs*, (NCBI GenBank accession no. AF019383), [@pone.0017195-Lange2], forward primer 5′-CCACCAGGCTTACCCACACAA-3′, reverse primer 5′-GCCACCGCCATCCCTAAAC-3. *Ippi*, (NCBI GenBank accession no. AW255524), [@pone.0017195-Lange1], forward primer 5′-CTCTTGGGGTGAGAAATGCT-3′ reverse primer 5′-CATCTGAGGGGGCTTTGTA-3. *Gpps*, (NCBI GenBank accession no. EU108696), [@pone.0017195-Burke1], forward primer 5′-ATGATAAGCGGGCTGCATAG-3′ reverse primer 5′-CCGAAATTCCTCAGCTTCTG-3′. *Ls*, (NCBI GenBank accession no. AW255536), [@pone.0017195-Lange1], forward primer 5′-CGGTGGTGGAGAAATACTGGGTTT-3′, reverse primer 5′-CCGTAATCAGAGCGTGACTTTGC-3′. *L3oh*, (NCBI GenBank accession no. AF124817), [@pone.0017195-Lupien1], forward primer 5′-CCCCATCACCACCAACTCCA-3′, reverse primer 5′-GCTCCGCCAGCACCCATAG-3′; *Ipd*, (NCBI GenBank accession no. AY641428), [@pone.0017195-Ringer1], forward primer 5′-GAGCTTCTATGGGCAGGTCA-3′, reverse primer 5′- GGCCACGAATGGTAAACACT-3′. *Ipr*, (NCBI GenBank accession no. AY300162), [@pone.0017195-Ringer2], forward primer 5′-AGCCAATGGAGAAATGATCG-3′, reverse primer 5′- GAGAGGAATGAGGGCTTGTG-3′. *Pr*, (NCBI GenBank accession no. AAQ75423), [@pone.0017195-Ringer2], forward primer 5′-ACAGCCTGAAGCAGCCTGAA-3′, reverse primer 5′-CGGCAGAACCATCTCAAGGA-3′. *Mfs*, (NCBI GenBank accession no. AF346833), [@pone.0017195-Bertea1], forward primer 5′-GCAGAACGAGGTGCGAGAAG -3′, reverse primer 5′-TGCGAAAGGTGGATGTAGGC-3′. The length of PCR products was from 98 to 200 bp. Bioassays {#s4e} --------- Experiments were conducted in a glass Y-tube olfactometer connected to the glass jars. GC-grade air was pumped into the jars at a flow rate of 300 ml min^−1^. The Y-tube was housed in a blackened box with a diffused fluorescent lamp giving a constant light directly above the centre point of the Y-tube. Thirty *C. herbacea* were monitored for up to 7 min by recording "choice" or "no-choice" events. All glassware was carefully washed to remove any contaminating substances and air-dried at 120°C for 4 h to remove volatile compounds. Test temperature was maintained at about 24°C and 70% humidity. (+)-Menthofuran, (−)-pulegone, myrcene, *p*-cymene, (−)-limonene, (−)-menthone, α-terpineol and (*E*)-β-caryophyllene pure standards were purchased from Sigma-Aldrich and were used in chemical tests (see also Supporting [Fig. S2](#pone.0017195.s002){ref-type="supplementary-material"}). The concentration of pure standards in Y-tube tests was used at the same concentrations as found in leaf volatile emissions. Statistical analysis {#s4f} -------------------- Initial and final choice data were analyzed using a Chi-squared test. Yates\' correction was applied to adjust for the data with only one degree of freedom. Data were also statistically processed using a log-likelihood test (G-test) and by ANOVA by using the statistical program SPSS (version 16.0,SPSS Inc., Chicago, IL, USA). For genomic and chemical analyses, the overall data sets are expressed as mean values of at least three biological replicates each one repeated three times (technical replicates). Metric bars indicate SD. Significance of differences observed in data sets was tested by ANOVA using the SYSTAT 10 software. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Left panel, glass desiccators containing the plants are connected to the GC-grade air generator. The cork plug has two holes: one for allowing the GC-grade air to enter the jar and the other hosts a glass Pasteur pipette. In the right figure the arrow indicates the SPME fibre that adsorbs the VOCs exiting from the jar. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### Left panel, a glass Y-tube olfactometer is connected to the jars where a flux of GC-grade air blows the VOCs produced by undamaged and infested leaves. Arrows indicate the presence of the SPME fibre which is located just before the olfactometer arms, into the air path. Right upper panel shows *C. herbacea* making a choice. The lower right panel shows the flow-meter used to standardize the air flow blowing from the jars and the timer used during the choice tests. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### *C. herbacea* was found to lay eggs on undamaged *M. aquatica* plants. The left panel shows clutches of eggs laid on a young *M. aquatica* leaf. The right panel shows young *C. herbacea* larvae feeding on *M. aquatica* leaves. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### *M. aquatica* plants observed in the wild. The upper left panel shows an undamaged *M. aquatica* in full bloom. The upper right panel shows male and female *C. herbacea* mating on partially damaged *M. aquatica* leaves. The two lower panels show the typical level of damage inferred by *C. herbacea* on wild *M. aquatica* leaves. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors are grateful to A. Bienkowski for helpful suggestions. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Financial support came from the doctorate school of Pharmaceutical and Biomolecular Sciences of the University of Turin. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MEM SAZ. Performed the experiments: SAZ CMB SB AO GG. Analyzed the data: MEM SAZ GG. Contributed reagents/materials/analysis tools: MEM. Wrote the paper: MEM SAZ.

PubMed Central

2024-06-05T04:04:19.139900

2011-3-9

PMC3052310

Introduction {#s1} ============ Probiotics are defined as live, non-pathogenic bacteria that confer health benefits beyond their nutritional value [@pone.0016953-Schrezenmeir1]. A growing body of evidence, accumulated in the last decade, suggests that probiotics can be useful in the treatment of inflammatory bowel disease (IBD), especially in patients with ulcerative colitis (UC) and pouchitis [@pone.0016953-Kruis1]--[@pone.0016953-Zocco1]. In spite of this clinical evidence, our understanding of the biological processes involved in the beneficial effects of probiotics is still limited. Several mechanisms have been postulated to contribute to the anti-inflammatory effect of probiotics in the gut, including competitive exclusion of pathogens, production of antimicrobial agents and organic acids, enhancement of the epithelial barrier function, increase of mucosal IgA secretion, and modulation of lymphocyte and dendritic cell function [@pone.0016953-Dotan1]--[@pone.0016953-Sturm1]. Abnormalities of T cell function are recognized to play a central role in IBD pathogenesis [@pone.0016953-Xavier1]. Among them, resistance of mucosal T-cells to undergo apoptosis is believed to be a critical event by altering the equilibrium between cell death and proliferation resulting in an excessive accumulation of T-cells in the gut [@pone.0016953-Boirivant1]--[@pone.0016953-Sturm2]. The pathogenic importance of defective T-cell apoptosis is supported by the observation that various drugs effective in the treatment of CD promote immune cell apoptosis [@pone.0016953-Doering1]--[@pone.0016953-vandenBrande2]. This has led to the notion that induction of apoptosis in the mucosal immune compartment is a key step for successful treatment of IBD [@pone.0016953-Mudter1]. Apoptosis is induced by multiple factors and mediated by various mechanisms [@pone.0016953-Krammer1]. One of them involves sphingomyelinases, a group of enzymes responsible for the conversion of sphingomyelin into ceramide, a powerful second messenger which mediates all forms of apoptosis, including receptor-mediated, stress-induced, and cell detachment [@pone.0016953-Morales1]. Three main families of sphingomyelinases have been described, acidic (ASMase), alkaline and neutral (NSMase), according to their optimal pH [@pone.0016953-Marchesini1], [@pone.0016953-Duan1]. In relation to probiotics, recent observations described that VSL\#3, a probiotic that contains a mixture of eight different strains of bacteria, stimulated mucosal alkaline sphingomyelinase activity, although the molecular downstream targets contributing to the therapeutic effect of VSL\#3 were not elucidated [@pone.0016953-Soo1]. NSMAse is also found in bacteria, yeast and mammalian cells, and is a membrane-bound in mammalian cells but a soluble protein in bacterial cells, with great variations in NSMAse concentrations among different bacterial strains [@pone.0016953-DiMarzio1]. Since the role of NSMase in the therapeutic benefits of probiotics has not been previously reported and because a large number of bacteria produce NSMase [@pone.0016953-Clarke1] and probiotics can induce apoptosis [@pone.0016953-Salinas1], we hypothesised that the probiotic strains used in the management of IBD might exert an anti-inflammatory effect by inducing immune cell death. If present in adequate concentrations, probiotic NSMase could induce ceramide formation and trigger signalling pathways leading to apoptosis of lamina propria mononuclear cells (LPMC) in patients with IBD, counterbalancing the defective immune cell death. Therefore, the specific aim of this study was to investigate the capacity of probiotics to produce NSMase, and induce ceramide-mediated immune cell apoptosis. Our data indicate that bacterial NSMase-mediated ceramide generation induced immune cell apoptosis via JNK activation and ROS overgeneration, two known targets of ceramide action. Results {#s2} ======= Increased NSMase activity in probiotic bacteria {#s2a} ----------------------------------------------- Considering that several strains of bacteria can produce different amounts of NSMase [@pone.0016953-Clarke1], we initially measured the NSMase activity in each one of the probiotic and non-probiotic sonicates used in our study. *S. thermophilus and L Brevis* exhibit well documented probiotic properties and are considered probiotic bacteria. S. thermophilus is one of the components of the probiotic mixture VSL\#3, which has been used in the treatment of pouchitis and ulcerative colitis, among other conditions [@pone.0016953-Kruis1], [@pone.0016953-Bibiloni1]. As measured by HPTLC analysis, the sphingomyelinase activity of the two probiotic bacteria, *L. brevis* and *S. thermophilus*, was approximately ten-fold higher (p\<0.01) than that of the two non-probiotic commensal bacteria *E. coli* and *E. faecalis* ([**Fig. 1**](#pone-0016953-g001){ref-type="fig"}). Determination of ASMase activities indicated similar levels of activation in sonicates of probiotic and non-probiotic commensal bacteria (data not shown). ::: {#pone-0016953-g001 .fig} 10.1371/journal.pone.0016953.g001 Figure 1 ::: {.caption} ###### Neutral SMAse activity of bacterial sonicates. Neutral SMAse activity was quantified in *L. brevis*, *S. thermophilus*, *E. coli* and *E. faecalis* sonicates. SMAse activity of each bacterial sonicate is expressed as relative SMAse activity, assigning an arbitrary value of 100 to the mean SMAse activity of *L. brevis* sonicates. Each bar represents 3 separate experiments. \*p\<0.01 vs. *E. coli* and *E. faecalis*. :::  ::: Induction of LPMC apoptosis by probiotic but not commensal enteric bacteria {#s2b} --------------------------------------------------------------------------- In view of the markedly higher levels of NSMAse activity generated by the two probiotic compared to the two non-probiotic bacterial sonicates, we next tested the ability of each bacterium to induce apoptosis of mucosal immune cells. In the absence of bacteria, LPMC exhibited a low rate of spontaneous apoptosis, with no differences among control, UC and CD cells ([**Fig. 2a, b**](#pone-0016953-g002){ref-type="fig"}). Exposure of LPMC to the probiotic *L. brevis* sonicate resulted in an increased rate of apoptosis (p\<0.001) both in control and IBD cells, the increase being significantly (p\<0.05) greater for UC and CD compared to control LPMC ([**Fig. 2a**](#pone-0016953-g002){ref-type="fig"}). Similarly, *S. thermophilus* also resulted in an increased LPMC apoptosis (p\<0.01), however, this increase was significant (p\<0.05) only in UC and CD, but not control cultures ([**Fig. 2a**](#pone-0016953-g002){ref-type="fig"}). When the same control, UC and CD cells were exposed to the non-probiotic, non-pathogenic gut commensal *E. coli* or *E. faecalis* sonicates, no significant effect on LPMC apoptosis was found ([**Fig. 2a, b**](#pone-0016953-g002){ref-type="fig"}). The pro-apoptotic effect of *L. brevis* and *S. thermophilus* was confirmed in an additional set of experiments in which apoptosis was assessed by double staining with annexin V and propidium iodide ([**Fig. 2c**](#pone-0016953-g002){ref-type="fig"}). ::: {#pone-0016953-g002 .fig} 10.1371/journal.pone.0016953.g002 Figure 2 ::: {.caption} ###### Induction of LPMC apoptosis by bacterial sonicates. A\) Control, UC, and CD LPMC were left alone or exposed to *L. brevis*, *S. thermophilus*, *E. coli* or *E. faecalis* sonicates for 24 hours, after which the percentage of apoptotic cells was assessed by PI staining and flow cytometric analysis. Each bar represents 23 control, 7 UC, and 10 CD separate LPMC isolates. \* p\<0.01 and \*\* p\<0.001 vs. no bacteria; \# p\<0.05 vs. control LPMC. B) Flow cytometric analysis of control (upper panel) and CD (lower panel) PI-stained LPMC. Figure is representative of 23 control, and 10 CD separate LPMC isolates. C) Flow cytometric analysis of control (left panels) and CD (right panels) Annexin V and PI-stained LPMC. Figure is representative of 4 control, and 4 CD separate LPMC isolates. :::  ::: Influence of the cell activation on probiotic-induced immune cell apoptosis {#s2c} --------------------------------------------------------------------------- The observation that the magnitude of probiotic-induced apoptosis was significantly greater for UC and CD than control LPMC suggests that mucosal cell activation, as found in IBD, may be involved in the process of probiotic-induced apoptosis. To test this hypothesis, we activated freshly isolated control LPMC with anti-CD3/CD28 antibodies for increasing periods of time prior to their exposure to bacterial sonicates. Engagement of the CD3/CD28 pathways resulted in a marked, time-dependent, and significant enhancement of both *L. brevis* and *S. thermophilus*-induced LPMC apoptosis compared to cultures with no bacteria (p\<0.01--0.001) and unstimulated LPMC (p\<0.05--0.01) ([**Fig. 3a**](#pone-0016953-g003){ref-type="fig"}). Interestingly, *E. coli* was unable to upregulate apoptosis of activated LPMC ([**Fig. 3a**](#pone-0016953-g003){ref-type="fig"}). To ascertain whether the ability of undergoing probiotic-induced, activation-dependent apoptosis was a distinctive property of LPMC, freshly isolated PBMC were used as an alternate source of immune cells. After activation with anti-CD3/CD28, the pattern of PBMC pro-apoptotic response to both *L. brevis* and *S. thermophilus* resembled that observed with LPMC (p\<0.05--0.01) ([**Fig. 3b**](#pone-0016953-g003){ref-type="fig"}), confirming that cell activation has a definitive role in probiotic-induced immune cell apoptosis. ::: {#pone-0016953-g003 .fig} 10.1371/journal.pone.0016953.g003 Figure 3 ::: {.caption} ###### Effect of cell activation on probiotic-induced mononuclear cell apoptosis. A\) Control LPMC were left untreated or activated with combined anti-CD3/CD28 for 24, 48 or 72 hours. Subsequently, cells were exposed to *L. brevis*, *S. thermophilus*, or *E. coli* sonicates for 24 hours, after which the percentage of apoptotic cells was assessed. Each bar represents 4 separate experiments. \* p\<0.01 and \*\* p\<0.001 vs. no bacteria; \# p\<0.05 and \#\# p\<0.01 vs. unstimulated LPMC. B) Control PBMC were left untreated or activated with combined anti-CD3/CD28 for 24, 48 or 72 hours. Subsequently, cells were exposed to *L. brevis*, *S. thermophilus*, or *E. coli* sonicates for 24 hours, after which the percentage of apoptotic cells was assessed. Each bar represents 4 separate experiments. \* p\<0.01 and \*\* p\<0.001 vs. no bacteria; \# p\<0.05 and \#\# p\<0.01 vs. unstimulated PBMC. :::  ::: Induction of apoptosis by exogenous ceramide and NSMAse {#s2d} ------------------------------------------------------- Having detected marked differences in the NSMase activity between probiotic and non-probiotic bacterial sonicates, we next undertook a series of experiments aimed at ascertaining whether the addition of exogenous ceramide to immune cells could reproduce the pro-apoptotic effect observed with probiotic bacteria. Exposure of control, UC and CD LPMC to a pre-determined optimal ceramide concentration (30 µg/mL) resulted in a significant increase of the number of apoptotic cells in all groups (p\<0.01--0.001) ([**Fig. 4a**](#pone-0016953-g004){ref-type="fig"}). Of note, the degree of apoptosis was significantly greater (p\<0.05) in UC and CD than control LPMC ([**Fig. 4a**](#pone-0016953-g004){ref-type="fig"}). Dihydroceramide, the inactive precursor of ceramide, failed to induce apoptosis in LPMC regardless of their source either from control, UC or CD patients ([**Fig. 4a**](#pone-0016953-g004){ref-type="fig"}). A similar effect was observed when LPMC were exposed to a pre-determined optimal concentration of *Bacillus cereus*-derived NSMAse (data not shown). The pro-apoptotic effect of NSMase required Mg^2+^ supplementation. Ceramide also induced apoptosis in activated LPMC and PBMC (p\<0.05) compared to unstimulated cells (see below). Thus, these findings underscore that exogenous ceramide and NSMase reproduce the effects of probiotic sonicates in activated mucosal cells. ::: {#pone-0016953-g004 .fig} 10.1371/journal.pone.0016953.g004 Figure 4 ::: {.caption} ###### Induction of mononuclear cell apoptosis by exogenous ceramide. A\) Control, UC, and CD LPMC were left untreated or cultured with ceramide or dihydroceramide for 24 hours, after which the percentage of apoptotic cells was assessed. Each bar represents 23 control, 7 UC, or 10 CD separate LPMC isolates. \* p\<0.01 and \*\* p\<0.001 vs. no bacteria; \# p\<0.05 vs. control LPMC. B) Control LPMC were left untreated or activated with combined anti-CD3/CD28 for 24, 48 or 72 hours. Subsequently, cells were cultured with ceramide or dihydroceramide for 24 hours, after which the percentage of apoptotic cells was assessed. Each bar represents 4 separate experiments. \* p\<0.05 and \*\* p\<0.01 vs. no bacteria; \# p\<0.05 vs. unstimulated LPMC. C) Control PBMC were left untreated or activated with combined anti-CD3/CD28 for 24, 48 or 72 hours. Subsequently, cells were cultured with ceramide or dihydroceramide for 24 hours, after which the percentage of apoptotic cells was assessed. Each bar represents 4 separate experiments. \* p\<0.05 and \*\* p\<0.01 vs. no bacteria; \# p\<0.05 and \#\# p\<0.01 vs. unstimulated PBMC. :::  ::: Effect of the cell activation state on ceramide-induced apoptosis {#s2e} ----------------------------------------------------------------- As demonstrated above, cell activation enhanced the degree of probiotic-induced apoptosis. To further corroborate that this effect was mediated by the NSMse/ceramide pathway, we evaluated the effect of cell activation on ceramide-mediated apoptosis following incubation with CD3/CD28 antibodies for 24--72 hours. Anti-CD3/CD28 activation resulted in a marked and time-dependent enhancement of ceramide-induced LPMC apoptosis ([**Fig. 4b**](#pone-0016953-g004){ref-type="fig"}), mimicking the previously observed increase in apoptosis of anti-CD3/CD28-activated LPMC by *L. brevis* and *S. thermophilus* (p\<0.05--0.01). The pro-apoptotic effect of ceramide was also noted with activated PBMC (p\<0.05--0.01) ([**Fig. 4c**](#pone-0016953-g004){ref-type="fig"}). Activation of JNK and production of ROS in probiotic-triggered LPMC apoptosis {#s2f} ----------------------------------------------------------------------------- Because ceramide activates stress-activated protein kinases, like c-jun N-terminal kinase (JNK) [@pone.0016953-Ruvolo1], we investigated whether exposure of control LPMC to *L. brevis* resulted in JNK activation. In a first series of experiments we found that exposure to *L. brevis* sonicates increased JNK phosphorylation in treated compared to untreated LPMC ([**Fig. 5a**](#pone-0016953-g005){ref-type="fig"}). In contrast, no increase in JNK phosphorylation was observed when LPMC were exposed to sonicates of the non-probiotic bacteria *E. coli*. JNK was not detected in *L. brevis* or *E. coli* sonicates ([**Fig. 5a**](#pone-0016953-g005){ref-type="fig"}). In a second series of experiments, we tested the effect of exogenous NSMase, and we also observed a time-dependent phosphorylation of JNK in LPMC that mimicked the results obtained with *L. brevis* sonicates ([**Fig. 5b**](#pone-0016953-g005){ref-type="fig"}). Together, these results further support the involvement of the NSMAse/ceramide pathway in probiotic-induced LPMC apoptosis. Treatment of *L. brevis* sonicates with the specific JNK inhibitor SP-600125 did not prevent the induction of LPMC apoptosis by *L. brevis* sonicates (data no shown), suggesting that other nSMAse-mediated, signalling pathways are able to forward the pro-apoptotic signal. ::: {#pone-0016953-g005 .fig} 10.1371/journal.pone.0016953.g005 Figure 5 ::: {.caption} ###### *L. brevis*-induced JNK phosphorylation by LPMC. A\) *L. brevis*- but not *E.coli*-induced phosphorylation of JNK in LPMC. Control LPMC were left alone or exposed to *L. brevis* or *E. coli* sonicates for 30 minutes or 2 hours, after which JNK phosphorylation was assessed. JNK is not detected in *L. brevis* or *E. coli* sonicates (two last lanes on the right). B) Exogenous nSMAse-induced phosphorylation of JNK in LPMC. Control LPMC were left alone or exposed to *L. brevis* or recombinant neutral SMAse for 15, 30 or 60 minutes, after which JNK phosphorylation was assessed. LB = *L. brevis* sonicates; EC = *E. coli* sonicates; nS = nSNAse. Figure is representative of 3 separate experiments. :::  ::: Another hallmark of ceramide-mediated cell signalling is the production of reactive oxygen species (ROS) [@pone.0016953-GarciaRuiz1], [@pone.0016953-Gudz1], [@pone.0016953-QuilletMary1]. Exposure of control LPMC to *L. brevis* sonicates resulted in a 3-fold increase (p = 0.02) in ROS production compared to untreated cells ([**Fig. 6a**](#pone-0016953-g006){ref-type="fig"}). On the contrary, treatment of the same LPMC with sonicates of the non-probiotic bacteria *E. coli* failed to significantly change LPMC ROS production compared to untreated cells ([**Fig. 6a**](#pone-0016953-g006){ref-type="fig"}). Panels b, c and d in [Fig. 6](#pone-0016953-g006){ref-type="fig"} are representative flow cytometry analyses of control LPMC incubated with no bacteria, *L. brevis* or *E. coli* sonicates, respectively. Similar results were obtained when control LPMC were exposed to bacterial sonicates for 60 minutes (data not shown). These results show that ROS are generated during LPMC apoptosis mediated by probiotic nSMAse, further supporting the involvement of the NSMAse/ceramide pathway in this process. Treatment of *L. brevis* sonicates with the phenolic antioxidant butylated hydroxyanisole (BHT), which prevents ROS production, did not influence the induction of LPMC apoptosis by *L. brevis* sonicates (data no shown), suggesting that other NSMAse-mediated, signalling pathways are able to forward the pro-apoptotic signal. ::: {#pone-0016953-g006 .fig} 10.1371/journal.pone.0016953.g006 Figure 6 ::: {.caption} ###### *L. brevis*-induced reactive oxygen species production by LPMC. Left panel (A): control LPMC were left alone or exposed to *L. brevis* or *E. coli* sonicates for 30 minutes, after which production of reactive oxygen species by LPMC was assessed. Data are expressed as mean fluorescence intensity of LPMC. Each bar represents 5 separate experiments. \*p = 0.02 vs. no bacteria; \#p = 0.04 vs. *E. coli*. Right panel: Flow cytometric analysis of control LPMC. Cells were left alone (B) or exposed to *L. brevis* (C) or *E. coli* (D) sonicates for 30 minutes, after which production of reactive oxygen species by LPMC was quantified and expressed as mean fluorescence intensity. Figures are representative of 5 separate experiments. :::  ::: Blockade of probiotic-induced cell death by inhibition of NSMAse activity {#s2g} ------------------------------------------------------------------------- The results so far point to NSMAse as the mediator responsible for the pro-apoptotic effects of probiotic bacterial sonicates. To confirm this notion, we performed a series of blocking experiments using GSH, which is known to inhibit NSMase [@pone.0016953-Liu1]. When control LPMC were exposed to sonicates of either *L. brevis* or *S. thermophilus* previously preincubated with GSH, a complete abrogation of the sonicates\' pro-apoptotic effect was observed (p\<0.05) ([**Fig. 7a**](#pone-0016953-g007){ref-type="fig"}). A similar effect was also observed with CD LPMC, with a significant (p\<0.05) reduction in the percentage of apoptotic cells with both *L. brevis* and *S. thermophilus* GSH-treated sonicates ([**Fig. 7b**](#pone-0016953-g007){ref-type="fig"}). In contrast, preincubation of *E. coli* or *E. faecalis* sonicates with GSH had no significant modulatory effect on control or CD LPMC apoptosis ([**Fig. 7a,b**](#pone-0016953-g007){ref-type="fig"}). GSH also inhibited the pro-apoptotic effect of *L. brevis* and *S. thermophilus* on resting and CD3/CD28-activated PBMC ([**Fig. 7c**](#pone-0016953-g007){ref-type="fig"}). Although the capacity of GSH to abrogate the activity of NSMase is well established, GSH could have other effects on bacterial sonicates that might affect cell apoptosis. Therefore, we undertook additional experiments using GW4869, a specific inhibitor of NSMase [@pone.0016953-Luberto1]. Preincubation of *L. brevis* sonicates with various concentrations of GW4869 resulted in a clear dose-dependent inhibition of NSMase activity (p\<0.05--0.01) ([**Fig. 8a**](#pone-0016953-g008){ref-type="fig"}). When an optimal inhibitory dose of GW4869 was added to *L. brevis* sonicates a marked and significant (p\<0.05--0.01) reduction of their capacity to induce LPMC apoptosis was observed, more so for UC and CD than control LPMC ([**Fig. 8b**](#pone-0016953-g008){ref-type="fig"}). The combined results of the blocking experiments confirm the essential role of the NSMAse/ceramide pathway in mediation of probiotic-induced LPMC apoptosis. ::: {#pone-0016953-g007 .fig} 10.1371/journal.pone.0016953.g007 Figure 7 ::: {.caption} ###### Inhibition of *L. brevis*-induced cell apoptosis by glutathione. A\) Control LPMC were left alone, exposed to *L. brevis*, *S. thermophilus*, *E. coli* or *E. faecalis* sonicates, or bacterial sonicates previously treated with 5 mM *glutathione* for 24 hours, after which the percentage of apoptotic cells was assessed. B) CD LPMC were left alone, exposed to *L. brevis*, *S. thermophilus*, *E. coli* or *E. faecalis* sonicates, or bacterial sonicates previously treated with 5 mM *glutathione* for 24 hours, after which the percentage of apoptotic cells was assessed. C) Control PBMC were left unstimulated or activated with combined anti-CD3/CD28 for 24, 48 or 72 hours. Subsequently, cells were exposed to *L. brevis or S. thermophilus* sonicates, or to bacterial sonicates previously treated with 5 mM *glutathione* for 24 hours, after which the percentage of apoptotic cells was assessed. Each bar represents 4 separate experiments. \* p\<0.05 and \*\* p\<0.01 vs. no bacteria; \# p\<0.05 vs. no *glutathione*. :::  ::: ::: {#pone-0016953-g008 .fig} 10.1371/journal.pone.0016953.g008 Figure 8 ::: {.caption} ###### Specific inhibition of *L. brevis*-induced cell apoptosis by GW4869. A\) *L. brevis* sonicates were exposed to increasing concentrations of GW4869. SMAse activity of each bacterial sonicate is expressed as relative SMAse activity, assigning an arbitrary value of 100 to the mean SMAse activity of *L. brevis* sonicates. Each bar represents 3 separate experiments. \* p\<0.05 and \*\* p\<0.01 vs. no GW4869. B) Control LPMC were left alone or exposed to *L. brevis* sonicates previously treated with different doses of GW4869 for 24 hours, after which the percentage of apoptotic cells was assessed. Each bar represents 4 separate LPMC isolates. \* p\<0.05 and \*\* p\<0.01 vs. no bacteria; \# p\<0.05 and \#\# p\<0.01 vs. no GW4869. :::  ::: Discussion {#s3} ========== The molecular mechanisms of the beneficial effects of probiotics remain to be fully elucidated. Recent observations showed that VSL\#3, a probiotic containing a mixture of bacteria, stimulated mucosal alkaline sphingomyelinase and reduced inflammation [@pone.0016953-Soo1]. However, the underlying mechanisms and downstream targets for the putative role of alkaline sphingomyelinase in mediating the therapeutic effects of probiotics were not characterized [@pone.0016953-Soo1]. The present study shows that *L. brevis* and *S. thermophilus*, two probiotic bacterial sonicates, can induce apoptosis of mucosal immune cells through the SMAse/ceramide pathway. This effect is not shared by *E. coli* and *E. faecalis*, the two non-probiotic enteric bacterial sonicates studied, suggesting that the ability to kill immunocytes is restricted to bacterial strains containing sufficient quantities of the key enzyme NSMase. We show that extracts of the probiotic *L. brevis* or *S. thermophilus* induced apoptosis of systemic and mucosal immune cells. Not only was this effect greater in magnitude than that of the non-probiotic *E. coli* or *E. faecalis*, but was also greater for LPMC from IBD than control non-IBD mucosa. Because IBD cells are in an enhanced state of activation, activation or differentiation may be a requirement to render immune cells susceptible to the pro-apoptotic effect of probiotics. This appears to be the case, as CD3/CD28-stimulated LPMC and PBMC displayed increased apoptosis upon exposure to *L. brevis* or *S. thermophilus* sonicates. The degree of cell activation, as determined by the duration of CD3/CD28 stimulation, correlated closely with the amount of mononuclear cell apoptosis, underscoring the importance of cell activation for probiotic-mediated cell death. These results are in keeping with previous evidence showing a tight relationship between immune cell activation and apoptosis. Activation-induced cell death is a complex physiological event aimed at deleting an excess of activated lymphocytes to prevent autoimmunity or uncontrolled immune reactivity [@pone.0016953-Krammer1], [@pone.0016953-Zhang1]. The relevance of this mechanism to autoimmunity prevention and maintenance of tolerance is exemplified by the development of autoimmune diseases in mice and humans with inherited defects in Fas or FasL [@pone.0016953-Rieuxlaucat1], [@pone.0016953-Fisher1]. Because probiotics are live bacteria with a vast array of components and functions, the exact mechanisms explaining the beneficial effects are complex and difficult to dissect. If the capacity to induce immune cells apoptosis is an exclusive property of some probiotic bacteria, as found in the present study, certain products with the ability to induce apoptosis should be predominantly expressed by these probiotics. The striking quantitative difference in NSMAse activity between the two probiotic and the two commensal bacterial sonicates included in our study clearly points towards this enzyme as the likely culprit for the pro-apoptotic action of *L. brevis* and *S. thermophilus*. To test this possibility a series of experiments was carried out. First, *L. brevis* and *S. thermophilus* but not *E.coli* or *E. faecalis* sonicates exhibited enhanced NSMase activity. Further, we demonstrated that exposure of resting and activated PBMC and LPMC, to either exogenous ceramide or recombinant NSMAse results in a marked enhancement of cell apoptosis, mimicking the effect seen with the probiotic strains *L. brevis* and *S. thermophilus*. As negative control, the inactive sphyngolipid dihydroceramide, failed to induce apoptosis. Corroborating evidence was obtained by showing that the level of ceramide-induced apoptosis was also dependent of the degree of immune cell activation. Second, we investigated some of the ceramide-triggered pathways during probiotic-induced mononuclear cell apoptosis. Signalling downstream of ceramide is extremely complex, but two of the best-known pathways involve JNK activation [@pone.0016953-Ruvolo1], [@pone.0016953-Sanchez1] and ROS production [@pone.0016953-GarciaRuiz1], [@pone.0016953-Won1]. As we hypothesised, phosphorylation of JNK was observed after exposure of LPMC to *L. brevis*, but not to the commensal *E. faecalis*. Of note, exposure of LPMC to recombinant NSMAse also induced JNK activation, and to a similar degree than *L. brevis* sonicates. Likewise, *L. brevis*, but not *E. faecalis*, induced a four-fold increase in ROS production by LPMC. The selectivity of *L. brevis* effects on JNK activation and ROS formation bolster the notion that NSMAse, present in large quantities in probiotics, and its product ceramide are intimately involved in the process of probiotic-induced immune cell apoptosis. The fact that neither JNK nor ROS production blockade separately were able to prevent probiotic sonicate-induced LPMC apoptosis reflects the complexity of signalling downstream of ceramide, suggesting the existence of redundant pathways to forward the pro-apoptotic signal. For instance, in addition to JNK/ROS, ceramide (independently of its mechanism of generation) can target other cell death regulators, such as cathepsin D or protein phosphatase PP1 or PP2, among others [@pone.0016953-Morales1]. Third, we investigated the consequences of NSMAse inhibition, which should result in a reduction of the probiotic pro-apoptotic capacity. Preincubation of *L. brevis* and *S. thermophilus* sonicates with optimal concentrations of GSH, a potent inhibitor of NSMAse activity [@pone.0016953-Liu1], markedly reduced apoptosis. A potential limitation to the interpretation of these results is that GSH, besides nSMAse inhibition, could influence the ability of probiotic sonicates to induce apoptosis by modifying the redox state [@pone.0016953-Martin1]. To address this issue, we used the highly specific nSMAse inhibitor GW4869, which does not influence the redox state [@pone.0016953-Luberto1]. GW4869 almost completely abrogated in a dose-dependent fashion the nSMAse activity of *L.brevis*, the strain with the highest SMAse activity. Additionally, pre-incubation of *L. brevis* sonicates with optimal doses of GW4869 obviously reduced *L. brevis*-induced LPMC apoptosis, an inhibitory effect more evident for IBD than control LPMC. Our results cannot be extrapolated to all probiotic agents described to date, since only two of them, *L. brevis* and *S. thermophilus*, were evaluated in our study. The fact that other probiotic microorganisms might have low or absent NSMAse activity, a key molecule for probiotic-induced immune cell apoptosis, as shown in our study, implies that the assessment of NSMAse activity may have clinical applications, helping the investigators to select the best probiotics to be tested in IBD. The relevance of our findings mostly rests on the concept that resistance of mucosal immune cells to apoptosis is considered an essential component of IBD pathogenesis (17), probably in combination with an excessive cell proliferation [@pone.0016953-Sturm3], [@pone.0016953-Sturm4]. The clinical relevance of immune cell apoptosis in IBD is supported by evidence showing that induction of apoptosis is a mechanism common to most IBD therapies, including aminosalycilates (12), steroids (13), immunosupressants [@pone.0016953-Frustaci1], and infliximab [@pone.0016953-vandenBrande3]. In conclusion, the ability of *L. brevis* and *S. thermophilus* to promote apoptosis may compensate the deficient cell death and thus contribute to re-establish mucosal immune homeostasis and decrease inflammation in IBD-involved intestine. Materials and Methods {#s4} ===================== Bacterial strains and preparation of bacterial sonicates {#s4a} -------------------------------------------------------- Bacterial strains used in this study included the probiotic bacteria *L. brevis* (CD2; DSM\#11988; Deutsche Sammlung von Mikroornanisgmen und Zellkulturen, Braunschweig, Germany) and *S. termophilus* (CD8; DSM\#14667), and the non-probiotic non-pathogenic bacteria *E. coli* (ATCC\#49420; ATCC, Manassas, VA) and *E. faecalis* (ATCC\#700802; ATCC). Bacterial suspensions were prepared from cultures of each individual strain. *L. brevis* and *S. thermophilus* were grown in De Man Rogosa Sharp medium (MRS; Difco, Detroit, MI). *E.coli* and *E. faecalis* were grown in a 1% tryptone peptone medium (Difco), 0.5% NaCl, and 0.5% yeast extract (Difco). Bacteria were harvested at the end of logarithmic phase by centrifugation, suspended in four times their wet weight in desintegration buffer containing (7.8 g/l NaH~2~PO~4~, 7.1 g/l Na~2~HPO~4~, 0.247 g/l MgSO~4~×7H~2~O, 1 mM EDTA, 1 mM DTT, proteases inhibitors, and triton) and, after three freeze-thaw cycles, sonicated (30 min, alternating 10 s of sonication and 10 s of pause) with a Vibracell sonicator (Sonic and Materials, Danbury, CT). Isolation of human LPMC and PBMC {#s4b} -------------------------------- Lamina propria mononuclear cells (LPMC) were isolated from surgical specimens obtained from patients admitted for bowel resection for UC and CD, while malignant and nonmalignant conditions provided histologicaly normal control tissue. All diagnoses were confirmed by clinical, radiologic, endoscopic, and histologic criteria. LPMC were isolated as previously described [@pone.0016953-West1]. Blood samples were collected from patients with UC and CD, as well as healthy subjects. Peripheral blood mononuclear cells (PBMC) were isolated from heparinized venous blood, using a Ficoll-Hypaque density gradient, as previously reported [@pone.0016953-Sturm5]. This project was approved by the Institutional Review Board of University Hospitals of Cleveland and Hospital Clinic i Provincial of Barcelona. Flow-cytometric characterization of mononuclear cell apoptosis {#s4c} -------------------------------------------------------------- A series of preliminary studies were carried out to ascertain the optimal concentration of bacterial sonicates and time of exposure of mononuclear cells to induce apoptosis. In subsequent experiments, bacterial sonicates were added to mononuclear cell cultures at a concentration of 1 mg protein of bacterial sonicate/10^6^ cells/ml, for 24 h. Apoptosis and cell cycle phase distribution were analyzed by propidium iodide staining followed by flow cytometry, as previously described [@pone.0016953-Sturm3]. Where indicated, LPMC and PBMC were pre-stimulated with CD3 monoclonal antibody (mAb) (OKT3; Ortho Diagnostic System INC., Raritan, NJ) and CD28 mAb (Pharmingen Biosciences, San Diego, CA, USA) for different periods of time (24, 48 or 72 hours) before exposure to the bacterial sonicates. Similarly, where indicated, *L. brevis* sonicates were preincubated for 30 min with 5 mM glutathione (GSH) (Calbiochem) or various concentrations of the specific NSMAse Inhibitor GW4869 (Sigma, St. Louis, MO), before addition to the cell cultures. In an additional set of experiments, apoptosis was assessed by combined annexin V/propidium iodide staining (Dako, Glostrup, Deenmark), according to the manufacturer\'s instructions. In some cases, LPMC and PBMC were incubated with various concentrations of C2-ceramide (Biomol International, Plymouth Meeting, PA), the inactive sphingolipid dihydroceramide (Biomol International), or nSMase from *Bacillus cereus* (Sigma) and examined for apoptosis as described above. Quantification of the NSMase activity of bacterial sonicates {#s4d} ------------------------------------------------------------ Lyophilized bacteria (10 mg) were resuspended and sonicated as described above. Protein concentration was determined by the microbicinchoninic acid assay kit (Pierce, Rockford, IL) with bovine serum albumin standards. Different amounts of bacterial sonicates (1, 0.5, and 0.25 mg) in 50 mM Tris·HCl buffer, pH 7.4, containing 5 mM MgCl~2~ and \[*N*-*methyl*-14C\]sphingomyelin (0.28 µCi/ml, specific activity 55 mCi/mmol; Amersham, Buckinghamshire, UK), were used for determination of NSMase activity as described before [@pone.0016953-Coll1]. Alternatively, NSMase activity was also determined by incubating extracts with C6-7-nitro-2,1,3-benzoxadiazol-4-yl (NBD)-sphingomyelin analyzing NBD-ceramide by TLC [@pone.0016953-Loidl1]. Determination of ceramide levels was performed by the diacylglycerol (DAG) kinase assay as described [@pone.0016953-GarciaRuiz1]. In some instances, ceramide levels were analyzed by HPLC after deriatization of the sphingoid base with O-phthaldehyde following deacylation of ceramide as described previously [@pone.0016953-GarciaRuiz2]. Where indicated, the bacterial preparations were preincubated for 30 min with 5 mM GSH [@pone.0016953-Liu1] or various concentrations of GW4869 to inhibit nSMase activity, as previously described [@pone.0016953-Luberto1]. JNK activation in LPMC {#s4e} ---------------------- At the end of 24 hr culture with bacterial sonicates, cells were washed, lysed and 40 µg of protein were run on a 10% SDS-PAGE and transferred onto a nitrocellulose membrane at 140 mA in 25 mM Tris-HCl, pH 8.3, 192 mM glycine, and 20% methanol. To study the activation of JNK, western blot analysis was performed using an anti-phosphorylated JNK antibody, dilution 1∶1000 (Pharmingen Biosciences, San Diego, CA), according to the manufacturer\'s instructions. In additional experiments, we also assessed the functional impact of JNK blockade on probiotic sonicates-induced LPMC apoptosis. The specific JNK inhibitor SP-600125 was purchased from Calbiochem (San Diego, CA, USA) and dissolved in dimethyl sulfoxide (DMSO, Sigma-Aldrich). Stock solutions of at least 20 mM were made. *L. brevis* sonicates were incubated with SP-600125 for 30 minutes, before its addition to LPMC cultures. LPMC apoptosis and cell cycle phase distribution were analyzed by PI staining followed by flow cytometry, as described above. Reactive oxygen species (ROS) production in LPMC {#s4f} ------------------------------------------------ Hydrogen peroxide was measured by cytometry using chloromethyl-2′-7′-dichlorofluorescein diacetate (DCF; Invitrogen Molecular Probes, Eugene, OR). LPMC were incubated with the fluorescent probe at 2 mM, in the absence or presence of bacterial sonicates. Fluorescence was determined at 529 nm for emission and 503 nm for excitation, with slit widths of 10 and 5 nm, respectively, after 30 and 60 minutes of LPMC exposure to bacterial sonicates [@pone.0016953-GarciaRuiz3], [@pone.0016953-Cathcart1]. Fluorescence of DCF was correlated with increasing concentrations of hydrogen peroxide allowing determination of hydrogen peroxide, as described [@pone.0016953-GarciaRuiz3]. In additional experiments, we also assessed the functional impact of ROS production blockade on probiotic sonicates-induced LPMC apoptosis. To prevent ROS production the phenolic antioxidant butylated hydroxyanisole (BHT) was obtained from Sigma (St. Louis, MO, USA) and added to bacterial sonicates for 30 minutes, before its addition to LPMC cultures. LPMC apoptosis and cell cycle phase distribution were analyzed by PI staining followed by flow cytometry, as described above. Statistical Analysis {#s4g} -------------------- Data were analyzed by Stat View software (SAS Institute Inc., Cary, NC) using the non-parametric Kruskall-Wallis and Mann-Whitney tests. Repeated measures for the same subject were analyzed by using the Student paired *t* test. Values are expressed as the mean ± SEM. Statistical significance was set at a P value of less than 0.05. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The funding sources included grants from Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Ministerio de Ciencia e Innovación (SAF2005-00280 and SAF2008-03676 to MS, FIS2009-00056 to AM, SAF2009-11417 to JCF), Fundación Ramón Areces (to MS), the National Institutes of Health (DK30399 and DK50984 to CF) and the Research Center for Liver and Pancreatic Diseases funded by the United States National Institute for Alcohol Abuse and Alcoholism (P50 AA 11999 to JCF). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JCFC AM CF MS. Performed the experiments: SA AM LL MCM JP JV. Analyzed the data: SD JP CD JCFC CF MS. Contributed reagents/materials/analysis tools: NP MGC CDS SD JV. Wrote the paper: JCFC MS.

PubMed Central

2024-06-05T04:04:19.143070

2011-3-9

PMC3052311

Introduction {#s1} ============ Gap junction channels are formed when connexin (Cx) subunits situated in adjacent plasma membranes dock together. These transmembrane channels allow the passage of metabolites, ions and second messengers and nucleotides up to 1 kDa in size [@pone.0014746-Evans1]. Undocked connexins, or hemichannels have also been identified and they provide a means to contact the extracellular environment. For example, they allow the passage of low molecular weight molecules into the cell [@pone.0014746-Windoffer1], [@pone.0014746-Paul1] and are gated by the cations Na^+^, K^+^ and Ca^2+^ [@pone.0014746-Verselis1], [@pone.0014746-DeVuyst1]. Gap junction communication has been shown in glial and neuronal cells and recently Cx43 was shown to negatively modulate neuronal differentiation [@pone.0014746-Condorelli1]. Cx36 is also hypothesized to play a role in neuronal development, because its expression peaks in the inferior olive, cerebellum, striatum, hippocampus and cerebral cortex during the first 3 postnatal weeks, a period that coincides with extensive inter-neuronal coupling [@pone.0014746-Belluardo1]. Importantly, during development Cx36 expression becomes restricted to neuronal cells, while Cx43 expression becomes restricted to astrocytes [@pone.0014746-Gulisano1], [@pone.0014746-Prime1]. Cx36 is expressed dynamically during murine embryonic development and it is switched on earlier than other Cxs. Expression is evident at E9.5 in the forebrain and expands into the midbrain as neurogenesis occurs. By E12.5 its expression pattern matches that of major morphogenetic boundaries within the brain and this elevated pattern of expression continues until P14. This bimodal pattern of neural expression correlates with two major periods of circuit formation and further indicates a role for Cx36 in the fine-tuning of neural development [@pone.0014746-Gulisano1], [@pone.0014746-Prime1]. Neural stem cells have been identified in both the developing and adult nervous systems [@pone.0014746-AlvarezBuylla1]-[@pone.0014746-Luskin1]. These cells are self-renewing and can give rise to neurons, astrocytes and oligodendrocytes in the central nervous system (CNS). Functional gap junction proteins (including Cx43 and Cx36) have been identified in neural and embryonic stem cells and they are thought to play an important role in cell survival and differentiation [@pone.0014746-Huettner1]. Current data suggests that widespread Cx expression is required for synchronizing and fine-tuning developing populations of cells and their expression is both spatially and temporally regulated during mammalian CNS development [@pone.0014746-Elias1]. In this study we investigated the role Cx36 plays during neuronal differentiation from neural stem cells and lentiviral over expression and knockdown strategies demonstrated that neural differentiation is positively influenced by Cx36. Materials and Methods {#s2} ===================== Construction of Cx36 shRNA lentiviral vectors {#s2a} --------------------------------------------- Rat (NCBI accession number Y16898) Cx36 shRNA (5′CAGATTGCTTAGAGGTTAA 3′) and scrambled control (5′GGCACATTAGGAACCATACAT 3′) sequence primers were designed and sub-cloned into pCR2.1-TOPO (Invitrogen) as previously described [@pone.0014746-Harper1]. The resulting plasmid was digested with EcoRI and the U6.Cx36.shRNA sequence inserted into the lentiviral backbone plasmid pRRLsincppt.CMV.EGFP.wpre via MfeI sites ([figure 1A](#pone-0014746-g001){ref-type="fig"}). ::: {#pone-0014746-g001 .fig} 10.1371/journal.pone.0014746.g001 Figure 1 ::: {.caption} ###### Validation of lentiviral expression cassettes. Lentiviral expression cassettes (A). Cx36 levels in neuronal cultures were assessed by western blotting following transduction with a lentivirus expressing an anti-Cx36 shRNA (B) and one overexpressing Cx36 (C). Rat neurosphere cultures transduced with a lentivirus expressing EGFP (D). Scale bar = 100 µm. :::  ::: Construction of Cx36 expressing lentiviral vectors {#s2b} -------------------------------------------------- The Cx36 coding sequence was excised from pcDNA3.2/V5/GW/D\_rat Cx36 plasmid (GlaxoSmithKline) via SacI and AgeI restriction sites and sub cloned into the pRRLsinppt CMV.IRES.EGFP plasmid, between the CMV and IRES regions using SacI and XmaI restriction sites. The Cx36.IRES.EGFP sequence was excised from the resulting plasmid using SpeI and NotI restriction sites and the EGFP sequence was removed from the lentiviral backbone plasmid using XbaI and SalI restriction sites. The resulting plasmid construct is illustrated in [figure 1A](#pone-0014746-g001){ref-type="fig"}. Lentiviral vectors were prepared as previously described [@pone.0014746-Howarth1]. Animals and primary culture {#s2c} --------------------------- All animals were treated ethically in accordance with the rules of Bristol University. Time-mated E18 and E14 rats were obtained from Harlan. Animals were terminally anaesthetized and embryos were removed from the uterus, then transferred into dissecting medium (Hank\'s balanced salt solution \[HBSS, Gibco\]). Neuronal cultures {#s2d} ----------------- E18 rat hippocampus was removed and dissociated by incubation in 10% \[v/v\] trypsin-EDTA/growth media for approximately 15 minutes followed by trituration to obtain a single cell suspension. Cells were plated in poly-D-lysine (0.1 mg/ml; Sigma) coated 24 well plates (7.5×10^4^ cells/well) in 500 µl growth media (Neurobasal medium \[Gibco\] supplemented with B27 \[2%, Gibco\] and penicillin/streptomycin \[250U\]). Cells were fed every 2--3 days and transductions were done at 3 days *in vitro* (DIV). Cells were maintained for a further 7 DIV when harvested for Western blot analysis. Striatal neurons were also grown in media containing DMEM∶Ham\'s F12 (2∶1); B27 (2%); penicillin/streptomycin (250 U); foetal calf serum (1% \[v/v\]) and were fed every 2--3 days. Striatal neurosphere culture {#s2e} ---------------------------- The striatum was removed at E14 from the embryos in HBSS and incubated for 10 minutes in accutase at 37 °C. After trituration, cells were seeded in polyhema (120 mg/ml, Sigma) coated 96 well plates at a density of 1×10^5^ cells per well in 100 µl media (DMEM∶Ham\'s F12 \[3∶1\] Gibco; B27 (2%); penicillin/streptomycin (250U); heparin (5 µg/ml); FGF2 (20 ng/ml); EGF (20 ng/ml). Transductions were performed whilst cells were in single cell suspension at a multiplicity of infection (MOI) of 5. Media was doubled the following day and at 3--4 days, neurospheres were transferred to a T25 flask where media added resulted in a five times dilution of the cells. Neurospheres were passaged every 7 days by re-seeding single cells at a density of 100 K/ml. At each passage samples were collected for RNA analysis by snap freezing in liquid nitrogen and storing at −80 °C and for differentiation studies. Briefly, cells were treated with accutase and triturated to obtain a single cell suspension and were spotted on poly-D-lysine coated coverslips in 24 well plates at a density of 5×10^4^ cells in 30 µl differentiation medium (DMEM∶Ham\'s F12 \[3∶1\]; B27 (2%); penicillin/streptomycin (250U) per well. Cells were maintained for 7 days before fixing in 4% paraformaldehyde. Western Blot Analysis {#s2f} --------------------- Transduced E18 primary neurons were washed once with PBS then incubated in RIPA lysis buffer (50 mM Tris-HCL pH 7.4; 150 mM NaCl; 1% Triton X-100; 1% sodium deoxycholate; 0.1% sodium dodecyl sulphate supplemented with protease inhibitors (Complete Mini, Roche)). Cells were harvested and homogenized with a needle and syringe prior to incubation on ice (20 minutes). The lysates were then centrifuged (\>13,000×g, 20 minutes) and supernatants retained. Protein concentration was determined using the BCA protein assay (Pierce). Membranes were incubated with polyclonal rabbit anti-Cx36 (1∶500; Zymed 51-6300) or monoclonal mouse anti-alpha tubulin (1∶1000; Sigma) as the loading control. Horseradish peroxidase conjugated secondary antibodies were used (1∶10000; Amersham) and bound antibody was detected using ECL plus Western Blotting Detection Reagent (Amersham). Autoradiograph film (Amersham) was exposed to the membrane and developed using a Kodak film processor. Fluorescence Recovery After Photobleaching (FRAP) {#s2g} ------------------------------------------------- E18 primary striatal neurons were plated in 35 mm glass-bottomed culture dishes (Iwaki) at a density of 4×10^5^ cells per dish in 2 ml media. Cells were transduced with either Cx36 shRNA or scrambled control lentiviruses (MOI 5) at 3 DIV and experiments performed 7 days post-transduction. Media was supplemented with 25 mM Hepes (Sigma) prior to performing FRAP experiments. Cells were incubated in the presence of 5-(6)-carboxyfluoresceine-diacetate (10 µM, CFDA; Invitrogen) diluted in HBSS for 5 minutes at room temperature followed by washing 3 times with HBSS and returned to their original media prior to FRAP measurements. Briefly, experiments were performed on a confocal microscope using the Leica SP2 system. Images were obtained using a 40× oil objective with zoom. A single cell within a cluster of cells was selected for bleaching for 5 frames for a total of 30 seconds and recovery recorded at 20 second intervals for a total of 15 minutes (45 frames). Frame sequences from each experiment were analyzed in ImageJ and the fluorescence intensity (F~i~) of each bleached cell and its subsequent recovery was analyzed. A non-bleached cell was also analyzed in order to correct for a loss of fluorescence during the acquisition process. The following formulae were used to calculate fluorescence level [@pone.0014746-Phair1]: F~i~ of bleached cell = (F~t~−F~0~)/(F~pre~−F~0~), whereby Ft =  fluorescence at each recovery time point, F~0~ =  post-bleach intensity and F~pre~ =  pre-bleach fluorescence. To correct for bleaching during image acquisition: F~Norm(i)~ = Ft/Fpre, whereby FNorm(i) = F~i~ of unbleached cell. The final corrected value is calculated by F~i~/F~Norm(i)~. Data was analysed by repeated measures ANOVA and Bonferroni post-hoc test. Reverse Transcription and Polymerase Chain Reaction {#s2h} --------------------------------------------------- RNA was isolated from frozen neurospheres using Trizol. Isolated RNA was DNaseI treated (Roche) and the preparation cleaned up using the RNeasy minelute kit (Qiagen). RNA quality and quantity was determined using spectrophotometry and RNA was reverse transcribed into cDNA using the Superscript III First Strand Synthesis Kit (Invitrogen) according to manufacturer\'s instructions. PCR primers for Cx36 and GAPDH were used as previously described [@pone.0014746-LauxFenton1]; [@pone.0014746-Fraga1]. PCR reactions were performed following the manufactures instructions. BrdU incorporation and Immunocytochemistry {#s2i} ------------------------------------------ Cells were pulsed with 5-bromo-2′-deoxyuridine (BrdU; 0.2 µM, Sigma) for 24 hours prior to differentiation. Antigen retrieval was performed following fixation of cells with 4% paraformaldehyde. Cells were incubated on ice in HCL (1 M, 10 mins) followed by HCl at 37 °C (2 M, 20 mins). The cells were then washed in borate buffer (0.1 M, 12 mins at room temperature) and incubated with sheep anti-BrdU (Abcam; 1∶500) overnight. Alexafluor secondary antibodies were used at 1∶500. Immunostaining using O4 was performed on live cells. Briefly, cells were incubated with anti O4 (diluted 1∶4, derived from hybridoma cell lines, ECACC, Cambridge, UK) for 30 minutes at 37 °C. Cells were washed 3 times with DMEM before fixing. Cells were blocked and incubated with secondary anybodies as normal. For the remaining immunostaining, cells were fixed in 4% PFA and rinsed with phosphate buffered saline (PBS) before blocking and incubating with primary antibodies overnight at 4 °C. Primary antibodies and dilutions used were mouse monoclonal anti-βIII tubulin (1∶500, Covance), rabbit polyclonal anti-glial fibrillary acidic protein \[GFAP\] (1∶500, DAKO) and rabbit polyclonal anti-Cx36 (1∶400, Zymed). Cells were washed with PBS/Triton-X (0.1%) and incubated in the appropriate Alexafluor secondary antibodies (1∶500, Invitrogen). Coverslips were mounted using Fluorsave reagent (Calbiochem). Cell counts were performed on at least 3 fields per coverslip with at least 3 coverslips per treatment group 4 independent experiments were performed for each treatment group. All counts are expressed as a percentage of total cells. Cell body area, neurite branching and length of longest neurite were analyzed using the ImageJ plug-in NeuronJ. Statistical analysis was performed by one-way ANOVA followed by a Bonferroni post-hoc test. Results {#s3} ======= Lentiviral vectors mediate the effective overexpression and knockdown of Cx36 {#s3a} ----------------------------------------------------------------------------- To test the efficacy of Cx36 knockdown we initially used primary hippocampal neuronal cultures transduced with vectors encoding Cx36 or Cx36 shRNA ([Figure 1A](#pone-0014746-g001){ref-type="fig"}). Transduction levels of these cells were highly efficient, reaching levels of ∼100%. Cx36 protein levels were analysed by western blotting and the Cx36 shRNA was found to mediate highly effective knock down ([Figure 1B](#pone-0014746-g001){ref-type="fig"}) while the Cx36 vector increased expression in a concentration dependent manner ([Figure 1C](#pone-0014746-g001){ref-type="fig"}). To validate Cx36 knockdown a functional FRAP model was used to monitor gap junctional intracellular communication (GJIC). To achieve this Cx36 shRNA transduced primary neuronal cultures were loaded with the GJ-permeable dye CFDA and this was subsequently photobleached in a single cell. Transfer of dye from cell to cell occurs via GJs and this can be monitored by real time by microscopy. It was found that knockdown of Cx36 reduced dye recovery in photobleached cells whereas recovery was unaltered in control cells ([Figure 2A](#pone-0014746-g002){ref-type="fig"}, Movie S1and Movie S2). Cellular fluorescence intensity was pooled together and quantified ([Figure 2B](#pone-0014746-g002){ref-type="fig"}) and a statistically significant difference between recovery of fluorescence in Cx36 shRNA and control cells was observed (p\<0.0001; Repeated measures 2-way ANOVA and Bonferroni post-hoc test). ::: {#pone-0014746-g002 .fig} 10.1371/journal.pone.0014746.g002 Figure 2 ::: {.caption} ###### Cx36 shRNA- knockdown of gap junction intercellular communication. GJIC was assessed by fluorescence recovery after photobleaching in shRNA-transduced striatal neurons loaded with 6-(5) carboxyfluorescein diacetate. Arrow marks cell of interest (A). Pooled data showing relative fluorescence intensity from 3 separate cultures is shown (B). Scale bar = 20 µm. Error bars represent standard error of the mean. :::  ::: During mammalian development, Cx36 is expressed in the striatum and in the mature brain Cx-36 mediated neuronal coupling remains. We therefore chose to study Cx-36 function in an E14 striatal neurosphere model and showed that our vectors mediated highly effective transduction, as assessed by EGFP expression ([Figure 1D](#pone-0014746-g001){ref-type="fig"}). Neurospheres formed after 3--4 days post-dissection all expressed EGFP. This expression persisted for up to 21 days in neurosphere culture and following differentiation, EGFP expression was maintained in ∼100% of cells ([Figure 3B](#pone-0014746-g003){ref-type="fig"}). ::: {#pone-0014746-g003 .fig} 10.1371/journal.pone.0014746.g003 Figure 3 ::: {.caption} ###### Expression of Cx36 in neurosphere cultures. Cx36 mRNA expression progressively decreases in neurosphere cultures as time *in vitro* increases (A). The lenti-Cx36-EGFP vector sustains expression throughout the culture period. Scale bar = 40 µm (B). Cx36 Immunostaining of differentiated neural cells following viral transduction (C). Scale bar = 20 µm. :::  ::: Expression patterns of Cx36 in neurosphere cultures {#s3b} --------------------------------------------------- Striatal NPCs were maintained as neurospheres and passaged every 7 days in order to allow expansion of progenitors. The effect of this on Cx36 expression was examined by RT-PCR. Levels of Cx36 mRNA were initially high in primary E14 striatal and cortical tissue. The maintenance of neurospheres in culture resulted in reduced Cx36 expression when measured every 7 days and it became undetectable by 21 DIV ([Figure 3A](#pone-0014746-g003){ref-type="fig"}). Transduction of these cells with a lentivirus expressing Cx36-EGFP or an anti-Cx36 shRNA followed by differentiation resulted in an increase or reduction in Cx36 mRNA levels respectively (as assessed by immunocytochemistry). EGFP expression was observed by auto fluorescence in differentiated cells from cultures up to 21 DIV ([Figure 3B](#pone-0014746-g003){ref-type="fig"}), indicating that the vector had successfully integrated into the genome of the host cell and was still expressing the coding cassette. Double staining for Cx36 and βIII tubulin revealed sustained over expression of Cx36 protein in neurons up to 21 DIV ([Figure 3C](#pone-0014746-g003){ref-type="fig"}) although this was found in fewer cells than at 7DIV. At 7 DIV Cx36 expression is mostly located within the cell body and, as time in culture increases, Cx36 assemblies had been trafficked to the dendrites as well. Boxed areas of [figure 3C](#pone-0014746-g003){ref-type="fig"} illustrate the dendritic location of Cx36. Cx36 regulates neuronal differentiation {#s3c} --------------------------------------- The shRNA mediated knockdown of Cx36 resulted in a significant decrease (p = 0.0001) in the number of βIII tubulin positive cells ([Figure 4](#pone-0014746-g004){ref-type="fig"}) differentiating 7 days post-transduction (29.05±1.65%) compared to cells transduced with the scrambled control (36.63±0.63%). The data also show there is no change in cell counts at the 14 DIV or 21 DIV time points. In parallel with this data, over expression of Cx36 resulted in a significant increase (p\<0.01) in the number of βIII tubulin positive cells (42.13±0.71%) compared to the EGFP control (35.65±1.77%). Again there were no significant changes at 14 and 21 day time points ([Figure 4A](#pone-0014746-g004){ref-type="fig"}). However, oligodendrocyte differentiation was found to be altered following Cx36 overexpression. A significant increase (p = 0.0001, one-way ANOVA with Bonferroni post-hoc test) in the number of O4 positive cells observed at 7 DIV following Cx36 over expression (Cx36: 12.31±0.84% and EGFP control: 9.20±0.75%). Cx36 shRNA treated cells also show a modest decrease in the number of O4 positive cells yet this did reach statistical significance ([Figure 4B](#pone-0014746-g004){ref-type="fig"}). Finally, the number of GFAP positive cells was analyzed and no significant changes were observed in cells transduced with the Cx36 lentivirus compared with the control ([figure 4C](#pone-0014746-g004){ref-type="fig"}). Knockdown of Cx36 produced a significant increase in the number of GFAP positive cells (p\<0.01; 26.07±2.87%) compared with the control group (17.3±2.27%) at the 7DIV time point. Over 95% of neurons that spontaneously differentiated were GABA positive ([Figure S1](#pone.0014746.s001){ref-type="supplementary-material"}). To observe any early differential changes (prior to 7DIV) caused by viral transduction βIII tubulin counts were carried out after 4 (and 7) days expansion ([Figure 5](#pone-0014746-g005){ref-type="fig"}). The results showed that βIII tubulin counts were increased significantly at 4 (and again at 7) days following Cx36 transduction and were decreased significantly following transduction with the Cx36 shRNA ([Figure 5](#pone-0014746-g005){ref-type="fig"}). Furthermore, there was no change in trypan blue (total) counts after 4 or 7 expansion (data not shown). The proliferation of neuronal precursors at 7 DIV was assessed by BrdU incorporation ([Figure 6](#pone-0014746-g006){ref-type="fig"}). This indicted the proliferation profile of precursors prior to differentiation and, whilst a small number of cells will continue to proliferate following growth factor removal, the comparison between proliferation of neuronal precursors in Cx36 transduced and control cells was analysed. Co-localization of the neuronal marker βIII tubulin and BrdU did not significantly differ between cells transduced with Cx36 (5.39±1.53%) and the control (4.67±1.56%) (Student\'s T-test) suggesting that there was no difference between the proliferating neuronal precursors. ::: {#pone-0014746-g004 .fig} 10.1371/journal.pone.0014746.g004 Figure 4 ::: {.caption} ###### Cx36 influences cell fate. The lentiviral mediated expression of Cx36 increased neuronal differentiation, whereas shRNA mediated Cx36 knockdown decreased neuronal differentiation (A). Cx36 increased oligodendrocytes number whereas Cx 36 shRNA transduction has no effect on differentiation (B). No effect on astrocyte differentiation was observed (C). Maintenance of spheres in culture for longer periods had no effect on neural differerentiation. \*\*p≤0.01; \*\*\*p≤0.001, one-way ANOVA with Bonferroni post-hoc test. Error represents standard error of the mean. :::  ::: ::: {#pone-0014746-g005 .fig} 10.1371/journal.pone.0014746.g005 Figure 5 ::: {.caption} ###### Cx36 influences cell fate after 4 and 7 days expansion. Cx36 expression increased neuronal differentiation following 4 and 7 days expansion, whereas Cx36 knockdown decreased it. \*P\<0.05 versus relevant control (by one-way ANOVA with Bonferroni post-hoc test). Control for Cx36 is GFP and control for Cx36-shRNA is scrambled control labelled as control in the figure. :::  ::: ::: {#pone-0014746-g006 .fig} 10.1371/journal.pone.0014746.g006 Figure 6 ::: {.caption} ###### Analysis of cell proliferation in neuronal precursor cultures. Double-labelling for BIII tubulin and BrdU (A). Scale bar 40 um. There was no significant difference in proliferation of neuronal precursor cells (BIII tubulin/BrdU^+^) when Cx36 is over expressed compared with control (B). Error bars represent standard error of the mean. :::  ::: To ensure any early differential changes (prior to 7DIV) in cell viability/proliferation were observed we counted apoptotic cells using Tunel staining ([Figure 7](#pone-0014746-g007){ref-type="fig"}) and again measured BrdU incorporation after 4 and 7 days expansion ([Figure S1](#pone.0014746.s001){ref-type="supplementary-material"}). The results showed there was no change in the number of apoptotic (Tunel stained) cells after 4 or 7 days expansion when control cells were compared with shCx36 or Cx36 transduced cells ([Figure 7](#pone-0014746-g007){ref-type="fig"}). Furthermore, a similar comparison showed was there was no change in BrdU incorporation in total cells or in neurons ([Figure S1](#pone.0014746.s001){ref-type="supplementary-material"}) after 4 (or 7) days. Statistical analyses of the results showed that βIII tubulin counts were increased significantly at 4 (and 7) days following Cx36 transduction and were decreased significantly following transduction with the Cx36 shRNA ([Figure 5](#pone-0014746-g005){ref-type="fig"} and [Figure S1](#pone.0014746.s001){ref-type="supplementary-material"}). These data together with the βIII tubulin counts show that viral-mediated manipulation of Cx36 in neurosphere cultures had no effect on cell growth kinetics and cultures continued to expand at a similar rate in all conditions. ::: {#pone-0014746-g007 .fig} 10.1371/journal.pone.0014746.g007 Figure 7 ::: {.caption} ###### Tunel assay. Tunel+ nuclei (red) of neural precursor cells following 4 days expansion. Dapi (blue). Scale bar = 100 um. (A). Percentages of Tunel positive cells showed no significant difference in the number of apoptotic cells within the four different conditions at 4 and 7 days expansion (B). :::  ::: No significant changes were observed in cell body area or length or longest neurite in any of the treatment conditions. Though, overexpression of Cx36 caused a significant increase in the total number of neurites (one way ANOVA p\<0.05) at 14 DIV, (5.55±0.42) compared to the control (4.86±0.16). At all other time points and in the Cx36 shRNA-treated cells there was no difference in neurite branching ([Figure S2](#pone.0014746.s002){ref-type="supplementary-material"}). In order to investigate whether Cx36 had the ability to alter neuronal fate in established cultures, cells that had already been maintained for 7 days were transduced with Cx36. The Cx36-mediated increase in neuronal differentiation was only observed when cells were transduced at early time-points as there was no change in the number of differentiating neurons when cells were transduced following 7 days in culture, when endogenous Cx36 levels had declined, suggesting that the cultures had lost the ability to respond to Cx36 over expression ([Figure S3](#pone.0014746.s003){ref-type="supplementary-material"}). Discussion {#s4} ========== In the developing brain, neurons and their positions are established and then glial and other cells are matched to the architecture of the network [@pone.0014746-Miller1]. Foetal neurospheres maintained in culture contain a mixture of multipotent and lineage restricted cell types that can be differentiated to give rise to neurons, astrocytes and oligodendrocytes [@pone.0014746-Svendsen1], [@pone.0014746-Caldwell1]. Differentiation in the brain and in neurosphere models is known to be governed via intrinsic pathways and via responses to extracellular cues and by cell-cell contact [@pone.0014746-Gage1], [@pone.0014746-Ostenfeld1], all of which can be regulated by Cxs. However, Cx36 is the only Cx identified to be expressed in neurons [@pone.0014746-Belluardo1], [@pone.0014746-Condorelli2]. In this study we showed that shRNA vectors significantly reduce Cx36 protein expression and also inhibit GJIC. Additionally, altering Cx36 expression does not affect the growth of neurosphere cultures. Cx36 knock down resulted in significantly reduced neuronal differentiation while the overexpression of Cx36 increased neurogenesis. Furthermore, BrdU and Tunel experiments showed that the results were not a reflection of increased neuroblast proliferation prior to differentiation or differential apoptosis respectively as there were no differences between any of the experimental groups. The increased expression of Cx36 during development may provide enhanced cell-cell contact between neuronal progenitors and hence promote neuronal differentiation. A possibility supported by the dependence on Cx36 GJIC for neuronal coupling (also observed in this study). Interestingly, the knockdown of Cx36 resulted in a decrease in the number of neurons and an increase in the number of GFAP-positive cells, which could suggest the decrease in neuronal differentiation, elicits a compensatory rise in astrocyte differentiation. The effects of Cx36 manipulation on neurogenesis were only observed in cells that had been expanded for 7 days or less suggesting that the cells have already become committed and no longer respond to Cx36 over expression or knock down. This suggestion fits with the observed decrease in endogenous Cx36 expression that occurs in NPCs after 14--21 days and with Cx36 transduction of 7 day old cultures having no effect on neuronal differentiation. To explore further the involvement of Cx36-dependent gap junction communication in intra-neuronal communication and development experiments reflecting the in vivo composition of cells could be undertaken. This could be achieved using conditional Cx-36 transgenic mice and/or the stereotactic injection of viruses. Significantly following the lentiviral mediated overexpression of Cx36 in the intact hippocampus CA3 region of adult rats increased gamma oscillatory activity was measured (unpublished). Results which further support Cx36 playing a significant physiological role in intra-neuronal gap junction communication in intact networks. In addition to the increase in the number of neurons following Cx36 over expression, we found the number of oligodendrocytes was also significantly increased. Oligodendrocytes are the myelinating cells of the CNS and act to insulate neuronal axons transmitting electrical impulses. It is also known that oligodendrocytes provide trophic factors that promote neuronal survival and hence there may be a cooperative relationship between neurons and oligodendrocytes during the differentiation process [@pone.0014746-Wilkins1]. The rise in oligodendrocytes observed may hence be due increased need for myelination and trophic factor support. Studies on transgenic mice lacking both Olig genes revealed that differentiation of motorneurons and oligodendrocytes was replaced with differentiation of interneurons and astrocytes [@pone.0014746-Lu1], [@pone.0014746-Zhou1]. It may also be the case that an increase in the number of neurons provides a more supportive environment for the differentiation of oligodendrocytes and may explain the changes observed. Gap junctions have primarily been studied as an intracellular channel that connects one cell to another yet they have recently been described as having a number of other functions. Cxs have been shown to be necessary for radial glial migration in the developing neocortex by acting as dynamic anchoring points for the migrating cells [@pone.0014746-Elias2]. Indeed, neuronal differentiation from mouse and human stem cells was induced by cell-cell contact and cells deficient in Cx43 show aberrant specification [@pone.0014746-Parekkadan1], [@pone.0014746-Santiago1]. Cxs may also function in embryonic development to coordinate coupled populations and as an adhesive contact between migrating cells [@pone.0014746-Elias1]. Extensive coupling has been identified within hESC colonies [@pone.0014746-Huettner1], [@pone.0014746-Wong1] and several other reports demonstrate that GJIC is required to maintain NPC populations in a proliferative state [@pone.0014746-Todorova1], [@pone.0014746-Cheng1]. During neuronal differentiation of NPCs, the expression profile of Cxs is dynamic [@pone.0014746-Cai1]. Progenitor cell populations respond to growth factors in the culture media and Cxs themselves can be regulated by them [@pone.0014746-Rozental1]. Alterations in Cx36 expression may change the way in which NPCs respond to their extracellular environment leading to transformed neural fate decisions. The identification of hemichannels has also implicated Cxs in paracrine signalling. Purinergic signalling plays a major role in neural development, particularly during later developmental stages. Spontaneous radial glial Ca^2+^ waves have been recorded in the proliferative cortical ventricular zone (VZ). These waves are initiated by Cx hemichannels and disruption of this signalling reduces VZ proliferation [@pone.0014746-Weissman1]. This may be important in the activation of intracellular signalling pathways and upregulation of neural genes. Expression of Cx36 may allow the assembly of greater numbers of hemichannels and offer a greater reactive potential to purinergic signals and thereby allow a neuronal fate to be determined. Conclusions {#s4a} ----------- Gap junctions are highly expressed in stem cell populations and they play important roles in mediating intra- and inter-cellular communication. However, the role that specific connexins play in cell specification has not been investigated. Here we report that lentiviral mediated Cx36 expression in NPC cultures mediated an increase in neuronal and oligodendrocyte differentiation. Importantly, knockdown of Cx36 significantly reduced the number of differentiated neurons and increased the number of differentiated astrocytes, suggesting an increase in glial progenitor proliferation following a failure in neuronal differentiation. The effects that we observed occurred shortly after NPC expansion suggesting that Cx36 may regulate early neurogenesis during development, perhaps to fine tune the establishment of neural networks within the brain. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### A. GABAergic neurons Double staining for BIII tubulin (red) and GABA (green). Neurons that spontaneously differentiated were primarily GABAergic. Scale bar 40 um. B. Analysis of cell proliferation in neuronal precursor cultures following 4 days expansion. Neural precursor cells expanded for 4 days in the presence of BrdU showed no significant difference in BrdU incorporation in total cells or in neurons (BrdU+/BIII Tubulin+) when Cx36 was over expressed or knocked down compared with respective controls (A). Control for Cx36 is GFP and control for CX36-shRNA is scrambled control labeled as control in the figure. The results were analyzed by ANOVA followed by post-hoc Newman-Keuls multiple comparison test. There was no difference between any of the BrdU treatment groups. When the BrdU+/BIII Tubulin+results were compared: \*\*P\<0.01 when Cx36 treated cells were compared to its GFP controls; \*\*\*P\<0.001 when Cx36-shRNA treated cells were compared to the scrambled controls. These results reflect the increase in neuronal BIII Tubulin+cells in the Cx36 treated group and decrease in BIII Tubulin+ cells in the Cx36-shRNA treated group. C. Analysis of cell proliferation in neural precursor cultures when BrdU was administered during the 7 day differentiation phase. Neural precursor cells expanded for 7 days in the absence of BrdU but differentiated for 7 days in the presence of BrdU showed no significant difference in proliferation rate when Cx36 was over expressed or knocked down compared with respective controls (A). Control for Cx36 is GFP and control for CX36-shRNA is scrambled control labeled as control in the figure. (1.46 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### Neurite analysis. The number of neurites was analyzed in Cx36 shRNA versus scrambled control treated cells (A) and also in Cx36 versus EGFP control treated cells (B). Primary, secondary, and tertiary branching was analyzed along with total neurites. \*p = ≤0.05, Student\'s T Test. (1.05 MB EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Effect of Cx36 on differentiation of older NPC cultures. Neurospheres were transduced with Cx36 when their endogenous Cx36 mRNA levels had declined. There was no significant difference in the percentage of cells differentiating into neurons in Cx36-transduced cells compared to the control. Error bars represent standard error of the mean. (0.43 MB EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Movie S1 ::: {.caption} ###### Frap experiment: Bleach in the presence of anti-Cx36-shRNA (1.09 MB MP4) ::: ::: {.caption} ###### Click here for additional data file. ::: Movie S2 ::: {.caption} ###### Control FRAP experiment. (1.10 MB MP4) ::: ::: {.caption} ###### Click here for additional data file. ::: **Competing Interests:**GlaxoSmithKline provided some funding toward the studentship. However, this does not alter the authors\' adherence to all the PLoS ONE policies on sharing data and materials. **Funding:**The MRC and GSK provided the studentship that funded EH. Grants from the Wellcome Trust and BBSRC to JBU also helped support this work. These funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Funding for the research was also provided by Neurosciences CEDD GlaxoSmithKline to Colin Glover who is an employee of Neurosciences CEDD GlaxoSmithKline. This funder therefore had a role in the study design, data collection and analysis, decision to publish, and preparation of the manuscript. [^1]: Conceived and designed the experiments: EMH CPG MC JBU. Performed the experiments: EMH FR. Analyzed the data: EMH MC JBU. Contributed reagents/materials/analysis tools: FR LFW CPG JBU. Wrote the paper: EMH MC JBU. [^2]: ¤ Current address: Oxford Parkinson\'s Disease Centre, Department of Physiology, Anatomy and Genetics, University of Oxford, Oxford, United Kingdom

PubMed Central

2024-06-05T04:04:19.145880

2011-3-9

PMC3052312

Introduction {#s1} ============ Sporadic Creutzfeldt-Jakob disease is a uniformly fatal prion disorder of humans that results from refolding of the normal prion protein (PrP^C^) to a β-sheet rich pathogenic conformation called PrP-scrapie (PrP^Sc^) [@pone.0016804-Prusiner1], [@pone.0016804-Aguzzi1]. Currently, the only definitive diagnostic test for sCJD is based on detection of PrP^Sc^ in the brain tissue obtained by biopsy or at autopsy [@pone.0016804-Aguzzi1]. Available pre-mortem diagnostic tests have provided inconsistent results, making it difficult to use therapeutic options where available [@pone.0016804-Green1]--[@pone.0016804-VanEverbroeck1]. Moreover, undiagnosed cases of sCJD are likely to transmit the disease to healthy individuals, raising significant public health concerns [@pone.0016804-Aguzzi1], [@pone.0016804-Weissmann1]. Consequently, significant effort has gone into the identification of sensitive and specific biomarkers for sCJD [@pone.0016804-VanEverbroeck1], [@pone.0016804-Otto2], [@pone.0016804-Geschwind2]. Of these, analysis of CSF proteins has proved promising due to its reasonable reliability and low cost. Commonly tested CSF proteins include 14-3-3, total tau (T-tau), S100B, and neuron-specific enolase (NSE) [@pone.0016804-Hsich1]--[@pone.0016804-Pennington1]. Several large scale studies, some across centers, have explored the accuracy of these biomarkers in the context of variables such as clinical stage of disease, valine/methionine polymorphism in the prion protein gene, and sub-type of sCJD [@pone.0016804-Castellani1]--[@pone.0016804-Zerr1]. Based on these and other considerations, the World Health Organization approved a combination of CSF 14-3-3 and EEG for the pre-mortem diagnosis of clinically suspected cases of sCJD [@pone.0016804-Hsich1], [@pone.0016804-Zerr1], [@pone.0016804-World1]. Subsequent studies reported high sensitivity but relatively low specificity of this test, resulting in its replacement by a combination of CSF 14-3-3 and T-tau for cases of dementia suspected of sCJD [@pone.0016804-Green1]--[@pone.0016804-Aksamit1], [@pone.0016804-SanchezJuan1], [@pone.0016804-Pennington1], [@pone.0016804-Wiltfang1]. However, subsequent reports indicated varying sensitivity and specificity for 14-3-3 and T-tau when used singly or in combination, re-kindling the debate and dissent concerning the diagnostic accuracy of this test [@pone.0016804-Green1]--[@pone.0016804-Aksamit1], [@pone.0016804-SanchezJuan1], [@pone.0016804-Pennington1], [@pone.0016804-Satoh1]--[@pone.0016804-Sssmuth1]. Major factors contributing to the inconsistent performance of 14-3-3 and T-tau combination include: 1) variable false positives since 14-3-3 and T-tau are elevated in several other dementias besides sCJD [@pone.0016804-Franz1]--[@pone.0016804-Sssmuth1], 2) experimental variability of 14-3-3 that is semi-quantitative at best [@pone.0016804-Kurt1]--[@pone.0016804-Aksamit1], and 3) lack of comparative analysis of biomarker accuracy within and across studies using a rigorous statistical approach [@pone.0016804-Hsich1]--[@pone.0016804-Zerr1]. To address these weaknesses, it is critical to identify a biomarker that is altered by sCJD associated pathology and therefore specific, abundant in the CSF and easily quantifiable, and superior in diagnostic accuracy relative to the current biomarkers 14-3-3 and T-tau using uniform benchmarks of analysis. One such potential biomarker is Tf that is increased in sCJD and prion disease affected animal brains [@pone.0016804-Singh1]--[@pone.0016804-Kell2]. The effect of this change on CSF levels of Tf, however, is not known. In this study, we analyzed levels of CSF Tf in 99 autopsy confirmed cases of sCJD (CJD+) and 75 cases of dementia of non-CJD origin (CJD-) to evaluate the potential of CSF Tf and its isoforms Tf-1 and Tf-β2 as 'new' diagnostic biomarkers of sCJD. The diagnostic accuracy of new biomarkers was compared with the 'current' biomarker T-tau using common standards of evaluation and rigorous statistical modeling. Finally, the diagnostic accuracy of optimal 'new' biomarker and T-tau combination was compared with the 14-3-3 and T-tau combination currently used for the pre-mortem diagnosis of sCJD. Methods {#s2} ======= Ethics statement {#s2a} ---------------- Human tissue and CSF samples were obtained from the National Prion Disease Pathology Surveillance Center (NPDPSC) at Case Western Reserve University, Ohio. All samples are from deceased subjects, and personal information is limited to age, sex, symptoms, neuropathology, and classification of prion disease. The use of these samples has been granted waiver of informed consent by the Case Western Reserve University Institutional Review Board since the protocol meets criteria for exemption under Federal regulations 45 CFR 46.101 (b). ### Human samples {#s2a1} To avoid sample bias, pre-mortem CSF from autopsy-confirmed cases of sCJD (CJD+) and dementia of non-CJD origin (CJD-) received between 2006 and 2008 were included in this study. The samples consisted of 99 CJD+ and 75 CJD- cases, amounting to a total of 174 cases. Specific diagnosis of CJD- cases included Alzheimer\'s Disease (AD) (17), encephalitis (3), meningoencephalitis (2), cortical dysplasia (1), cortical angiopathy (1), angiitis with microinfarcts (1), infarct (1), cerebral vasculopathy (1), Lewy body dementia (1), Lewy body variant of Alzheimer\'s disease (1), anoxic and Wernicke\'s encephalopathies (1), meningeal carcinomatosis (1), Alexander disease (1), leukoencephalopathy (1), hippocampal sclerosis (1), frontotemporal dementia (1), coccidioidomycosis meningoencephalitis (1), CNS lymphoma (1), chronic meningoencephalitis (1), epilepsy or encephalopathy (1), glioma (1), perivenous encephalomyelitis (1), CNS lymphoma (1), military metastases (1), leptomeningeal lymphomatosis/leukemia (1), mitochondrial encephalopathy (1), lymphoma (1), and non-CJD dementia of uncertain diagnosis (29). The age at autopsy ranged from 37--85 years for CJD+, and 47--84 years for CJD- cases. All CSF samples were stored at −80°C until use. Human brain tissue from the frontal cortex of CJD+ and age-matched cases of dementia (CJD−) was obtained from the NPDPSC and stored at −80°C (brain tissue and CSF samples are from different cases). Tissue homogenates (10%) were prepared in lysis buffer and analyzed by Western blotting following standard protocols. ### Western blot and ELISA {#s2a2} Equal volume of CSF from CJD+ and CJD- cases was fractionated by SDS-PAGE and proteins transferred to PVDF membranes were probed with anti-transferrin antibody (Genetex Inc., Cat \# GTX 21223) followed by HRP-conjugated secondary antibody and visualization of reactive bands with ECL (Amersham). Several procedural and statistical precautions were taken to reduce error in comparing multiple samples. Procedurally, a similar protocol was followed for all Western blots, including exposure times. In addition, strips of PVDF membrane representing specific molecular weights were exposed to X-ray film simultaneously in large cassettes to obtain similar exposure. Quantification of immunoreactive bands was performed with UN-SCAN-IT software (version 6.1, Silk Scientific Inc., Utah, USA) using three exposures from a single membrane showing exponential increase in intensity. Statistically, results from different Western blots were analyzed simultaneously in logistic regression by treating these as clustered observations. The statistical software Stata allows estimation of regression coefficients after controlling for clustering to produce unbiased standard errors. Two Stata procedures were used to do this: the 'vce cluster' option and svy option. Each logistic regression model was first tested without adjustment for clustering and then tested with each of the 2 clustering options. Both had the effect, in general, of increasing standard errors and p-values for the coefficient estimates as expected. However, the significance of the coefficients of the biomarkers did not change; all their coefficients remained significant at the p\<.05 level. T-Tf (ELISA) was estimated with Human Transferrin ELISA kit (Alpha diagnostic international Inc., Cat \# 1210) following the manufacturer\'s instructions. T-tau was determined by Human Tau (total) ELISA Kit (Invitrogen, Camarillo, CA; KBH0042) as directed by the manufacturer. Iron saturation of CSF Tf was determined by radiolabeling with ^59^Fe followed by fractionation on native gradient gels and autoradiography. Statistical analysis {#s2b} -------------------- The following new biomarkers were evaluated for their diagnostic accuracy: total transferrin by Western blot (T-Tf (WB)) and ELISA (T-Tf (ELISA)) techniques, Tf-1, Tf-β2, and iron saturation of Tf (^59^Fe-Tf). Of these, T-Tf, Tf-1, and Tf-β2 offered greater promise, and were analyzed further either alone or in combination with T-tau, a current biomarker that is quantifiable and is superior to other CSF biomarkers used for the diagnosis of sCJD [@pone.0016804-SanchezJuan1], [@pone.0016804-Pennington1], [@pone.0016804-SanchezJuan2], [@pone.0016804-Baldeiras1]. A combination of univariate and multivariable statistical methods were used to evaluate the relative diagnostic accuracy of individual biomarkers. Since the biomarkers were expected to evidence skewed distributions, nonparametric statistical procedures were used as these are subject to fewer distributional assumptions and provide less biased estimators when data are non-normal. For relative analysis of individual biomarkers, the nonparametric Mann-Whitney *U* test was used for focused comparisons to test if the levels of new and current biomarkers differ significantly between CJD+ and CJD-, and between CJD+ and AD cases. In addition, correlation analysis was performed using Spearman\'s rank correlation (ρ), which is also nonparametric. Finally, logistic regression, which makes no assumptions regarding the distributional properties of data (e.g., non-normal) [@pone.0016804-Tabachnick1] was used to parameterize the diagnostic accuracy of biomarkers after controlling for age and duration (interval between CSF collection and death). Because the data were clustered (four sets of Western blots), logistic regression was conducted in Stata 10 with the option to correct standard errors for clustering. This option is known to correct for underestimates of coefficient standard errors, resulting in more conservative results. Using logistic regression results, an analytic expression for the risk of CJD was derived for each individual biomarker. Area under the receiver operating characteristic (ROC) curve and the Aikake Information Criterion (AIC) were obtained. A ROC curve graphically shows the trade-offs between sensitivity and specificity for different cut-offs used to discriminate between positive and negative cases (i.e., CJD+ and CJD- cases). The area under the ROC curve (AUC) can be understood as an estimate of the probability that the biomarker being tested correctly ranks a CJD+ case higher than a CJD- case and, therefore, indexes the discriminating power of the biomarker. The AIC is used to compare different logistic regression models for different biomarkers (individually or in combination). A model with a relatively lower AIC is considered superior due to its better fit to data and its parsimony. For each model, estimates of specificity and positive and negative likelihood ratios (LR) were obtained given a baseline sensitivity of 85%, which has precedence in the literature as a reasonable cut-off level for biomarker comparison of Alzheimer\'s disease (AD) [@pone.0016804-Consensus1], [@pone.0016804-Mattsson1]. Positive LR is defined as sensitivity/(1-specificity) and represents the increase in likelihood of a positive test result if the disease is present compared to its being absent. The negative LR is defined as (1-sensitivity)/specificity and represents the decrease in likelihood of a negative test result if the disease is present compared to it being absent. Positive and negative LR, like specificity, depends on the chosen cut-off for sensitivity (set at 85% in this study). Confidence intervals for proportions were calculated according to the efficient-score method and corrected for continuity [@pone.0016804-Newcombe1]. Confidence intervals for LRs were calculated as proposed in a previous report [@pone.0016804-Simel1]. All statistical tests were conducted at the 95% confidence level. To identify an optimal combination of biomarkers, all two way combinations of new and current biomarkers were also entered into a logistic regression model with controls for age, duration, and clustered data (as above). Combining new and current biomarker T-tau was intended to take advantage of the differential pathogenic mechanisms represented by these biomarkers. We focused on two-way combinations for reasons of practicality. However, a specific combination involving T-Tf and 14-3-3 could not be evaluated because the readouts for 14-3-3 are not precisely quantifiable, and this biomarker has demonstrated relatively low specificity in previous studies [@pone.0016804-Kurt1]--[@pone.0016804-Aksamit1]. All models were compared using three criteria: AUC, specificity, and AIC [@pone.0016804-Altman1]. Model comparisons were based on multiple indicators while recognizing that specificity at a given level of sensitivity is a key discriminating factor. Models with higher specificity were preferred. The optimal biomarker or combination of biomarkers was expected to ideally have the highest AUC and specificity, and the lowest AIC. We also conducted power analysis as proposed in a previous report [@pone.0016804-Kapaki1]. We determined that the new biomarkers would possess a discriminating power at least as good as T-tau; thus we based our power calculation on the expected effect size of T-tau. From previously published data on the power of T-Tau to discriminate between CJD+ cases and other forms of dementia [@pone.0016804-Otto4]--[@pone.0016804-VanEverbroeck3], we calculated an expected standardized difference of at least 1.7. Based on this, our study yields an estimate of statistical power exceeding 0.99 for logistic regression as well as the Mann-Whitney *U* test. Results {#s3} ======= Tf levels are decreased in the CSF of sCJD cases {#s3a} ------------------------------------------------ Evaluation of CSF Tf by Western blot (WB) reveals two distinct bands representing Tf-1 and Tf-β2, the latter representing deglycosylated Tf specific to the brain and CSF ([Figure 1](#pone-0016804-g001){ref-type="fig"} A, lanes 1--13). Total Tf (T-Tf (WB)) comprising of Tf-1 and Tf-β2 collectively, and individual subunits i.e. Tf-1 and Tf-β2 individually are decreased in the CSF of CJD+ relative to CJD- cases ([Figure 1](#pone-0016804-g001){ref-type="fig"} A, lanes 1--13). In contrast, brain homogenates show a relative increase in T-Tf (WB), Tf-1, and Tf-β2 in CJD+ relative to CJD- cases ([Figure 1](#pone-0016804-g001){ref-type="fig"} A, lanes 14--17) (cut lanes were re-aligned for comparison. Original blot is shown in [Figure S1](#pone.0016804.s001){ref-type="supplementary-material"}). ::: {#pone-0016804-g001 .fig} 10.1371/journal.pone.0016804.g001 Figure 1 ::: {.caption} ###### Tf is decreased in the CSF and increased in the brain of CJD+ cases. (**A**) Immunoblotting for Tf shows reduced levels of Tf-1 and Tf-β2 in the CSF of CJD+ relative to CJD- samples (lanes 1--13), and increased levels of both isoforms in the brain tissue of CJD+ relative to CJD- samples (lanes 14--17) (55). (**B**--**D**) Box and whisker plots of T-Tf (WB) indicating the median, 25--75^th^ percentiles, and outliers. The top and bottom of the "box" represent 75^th^ and 25^th^ percentile respectively. The "whiskers" represent the highest and lowest values. Outliers are values over 1.5 times the interquartile range and are shown as circles away from the whiskers. Significance of biomarker differences between CJD- vs. CJD+ and AD vs. CJD+ was calculated by the M-W test. :::  ::: Quantitative comparison shows a decrease in CSF T-Tf (WB) by 49% (z = 8.73, p\<.001) and an increase in brain T-Tf (WB) by 39% (z = 2.76, p\<.008) in CJD+ relative to CJD- cases ([Figure 1](#pone-0016804-g001){ref-type="fig"} B, C). Comparison with AD cases shows a decrease in CSF T-Tf (WB) by 39% in CJD+ relative to AD cases (z = 4.92, p\<.001) ([Figure 1](#pone-0016804-g001){ref-type="fig"} D). Estimation of CSF T-Tf by ELISA shows a decrease of 30% in CJD+ relative to CJD- (z = 6.53, p\<.001), and 26% relative to AD cases (z = 3.02, p  = .003) ([Figure 2](#pone-0016804-g002){ref-type="fig"} A, B). The difference in T-Tf measured by WB and ELISA methods is due to optimization of the ELISA kit for serum Tf that lacks Tf-β2, and is therefore less accurate in estimating brain T-Tf. ::: {#pone-0016804-g002 .fig} 10.1371/journal.pone.0016804.g002 Figure 2 ::: {.caption} ###### Quantitative comparison of CSF Tf between CJD+ and other dementias by the M-W *U* test. (**A--F**) Box and whisker plots of different biomarkers indicating the median, 25--75^th^ percentiles, and outliers as in [Figure 1](#pone-0016804-g001){ref-type="fig"} B--D. Significance of biomarker differences between CJD- vs. CJD+ and AD vs. CJD+ was calculated by the M--W test. :::  ::: Comparison of individual subunits of Tf shows a decrease in CSF Tf-1 by 35% (z = 7.59, p\<.001) and Tf-β2 by 61% (z = 6.96, p\<.001) in CJD+ relative to CJD- samples ([Figure 2](#pone-0016804-g002){ref-type="fig"} C, E), and a decrease of 28% in Tf-1 (z = 4.44, p\<.001) and 59% in Tf-β2 (z = 3.96, p\<.001) relative to AD samples ([Figure 2](#pone-0016804-g002){ref-type="fig"} D, F). To compare the iron content of CSF Tf between CJD- and CJD+ cases, first the iron saturation of CSF Tf from a relatively normal CJD- case was determined by competing radiolabeled ^59^FeCl~3~ with decreasing concentrations of unlabeled FeCl~3~ and fractionating ^59^Fe-Tf on a native gel for quantification ([Figure S2](#pone.0016804.s002){ref-type="supplementary-material"} A, B). These results demonstrate that CSF Tf is not fully saturated with iron ([Figure S2](#pone.0016804.s002){ref-type="supplementary-material"} B). Subsequently, CSF samples representing equal concentration of Tf (determined by ELISA) from CJD- and CJD+ cases were incubated with equal counts of ^59^FeCl~3~ to radiolabel Tf, dialyzed to remove free ^59^FeCl~3~, and fractionated on a native gel ([Figure S2](#pone.0016804.s002){ref-type="supplementary-material"} C). Quantification of ^59^Fe-Tf shows equivalent saturation of CSF Tf from CJD-, CJD+, and AD samples, indicating normal iron binding capacity CSF Tf regardless of the underlying disease ([Figure S2](#pone.0016804.s002){ref-type="supplementary-material"} D, E). Biomarker levels and differential diagnostics of T-Tf as a new biomarker for sCJD {#s3b} --------------------------------------------------------------------------------- Comparison of different biomarkers using Mann-Whitney *U* test reveals that the new biomarkers T-Tf (WB), Tf-1, and Tf-β2 are more sensitive differentiators of the CJD+ vs. CJD- groups in comparison to the current biomarker T-tau ([Table 1](#pone-0016804-t001){ref-type="table"}, [Figures 2](#pone-0016804-g002){ref-type="fig"} A--F and 3 A, B). Relative to T-tau, the absolute z-values for T-Tf (WB) and Tf-1 are higher by at least 2 and 1 standard deviation (SD) respectively ([Table 1](#pone-0016804-t001){ref-type="table"}). Noting that differential diagnosis is meaningful if the biomarker discrimination exceeds at least 2 SD, the current use of T-tau biomarker provides equivalent diagnostic accuracy as Tf-β2 (absolute z = 6.37 and 6.96 respectively) and Tf-1 (Δz = 1.22), but significantly poorer than T-Tf (WB) (Δz = 2.36) ([Table 1](#pone-0016804-t001){ref-type="table"}, [Figures 2](#pone-0016804-g002){ref-type="fig"} A, C, E and 3 A). However, the diagnostic accuracy of new biomarkers in differentiating CJD+ from AD does not yield comparable superiority over T-tau. Rather, these biomarkers evidence similar discrimination. T-tau has the highest z-value (z = −5.17, p\<.001), but its diagnostic accuracy is equivalent to T-Tf, Tf-1 and Tf-β2 (Δz = 0.25, 0.73 and 1.21 respectively) ([Table 1](#pone-0016804-t001){ref-type="table"}, [Figure 1](#pone-0016804-g001){ref-type="fig"} D, [Figures 2](#pone-0016804-g002){ref-type="fig"} B, D, F and 3 B). (In hypothesis testing, at a confidence level of 95%, any difference in z-values greater than 1.96 is significant. Based on this we reasoned that when z-values corresponding to individual biomarkers differ by 2 or more, the biomarker with greater z-value provides significantly greater discrimination at p = .05). ::: {#pone-0016804-g003 .fig} 10.1371/journal.pone.0016804.g003 Figure 3 ::: {.caption} ###### Quantitative comparison of CSF T-tau between CJD+ and other dementias by the M--W *U* test. (**A & B**) Box and whisker plots of T-tau comparing CJD- vs. CJD+ and AD vs. CJD+ cases. Plot shows median, 25 to 75^th^ percentiles, and outliers as in [Figures 1](#pone-0016804-g001){ref-type="fig"} and [2](#pone-0016804-g002){ref-type="fig"}. :::  ::: ::: {#pone-0016804-t001 .table-wrap} 10.1371/journal.pone.0016804.t001 Table 1 ::: {.caption} ###### Individual biomarker performance by the Mann-Whitney *U* test. ::: {#pone-0016804-t001-1} Comparison test New Biomarkers Current Biomarker ---------------------- ---------------- ------------------- ------------- ------------ --------------- **CJD+ vs. CJD-** CJD+ median (n) 1257093 (99) 1017031 (87) 236243 (87) 26.46 (99) 11530.87 (95) CJD- median (n) 2177893 (67) 1555881 (67) 606359 (67) 38.90 (64) 1266.97 (64) M--W U Test \* Z-statistic* 8.73 7.59 6.96 6.53 −6.37 \* p-value* \<.001 \<.001 \<.001 \<.001 \<.001 **CJD+ vs. AD only** CJD+ median (n) 1257093 (99) 1017031 (87) 236243 (87) 26.46 (99) 11530.87 (95) AD median (n) 2009259 (17) 1413527 (17) 577539 (17) 36.89 (15) 876.40 (15) M--W U Test \* Z-statistic* 4.92 4.44 3.96 3.02 −5.17 \* p-value* \<.001 \<.001 \<.001 0.003 \<.001 Note: The n ranges from 87 to 99 for biomarkers in the CJD+ group and from 64 to 67 for the CJD- group due to insufficient CSF volume. To ensure that varying n does not introduce bias, an additional analysis was performed using cases with only non-missing values for all biomarkers (87 for CJD+ and 64 for CJD-). The results did not differ noticeably from those reported above. ::: After controlling for patient age and duration, T-tau showed significant correlation with duration in CJD+ (ρ = −0.35, p\<.001), but not in CJD- cases (ρ = −0.14, p = 0.28). None of the new biomarkers showed a significant correlation with duration in the CJD+ or CJD- group. (Correlation of T-Tf (WB), Tf-1, and Tf-β2 in CJD+ group ρ = −0.08, p = 0.41; ρ = −0.02, p = 0.89; ρ = 0.11, p = 0.29, respectively, and in CJD- group ρ = 0.06, p = 0.62; ρ = 0.01, p = 0.96; ρ = 0.06, p = 0.64, respectively). These results indicate that T-tau changes as sCJD progresses, while the new biomarkers remain fairly stable in CJD+ and CJD- cases. None of the biomarkers showed any correlation with age. For comparative analysis of new and the current biomarker T-tau, logistic regression analysis was performed. Only T-Tf (WB), not T-Tf (ELISA) results are considered for reasons mentioned above. When individual biomarkers are tested for their diagnostic accuracy, the new biomarkers T-Tf (WB), Tf-1, and Tf-β2 are superior to the current biomarker T-tau ([Table 2](#pone-0016804-t002){ref-type="table"}). T-Tf (WB) and Tf-1 have significantly higher AUC (0.90 and 0.86 respectively) than T-Tau (0.78). Tf-β2 also has a higher AUC (0.82) than T-tau, but the difference is not significant. T-Tf (WB), Tf-1, and Tf-β2 yield significantly higher specificities of 71.6%, 65.7% and 64.2% respectively, while T-tau shows a relatively low specificity of 48.4% (see discussion below). Finally, the AICs for T-Tf (WB), Tf-1 and Tf-β2 of 0.81, 0.96 and 1.03 respectively are all lower than the AIC for T-Tau (1.23) ([Table 2](#pone-0016804-t002){ref-type="table"}). ::: {#pone-0016804-t002 .table-wrap} 10.1371/journal.pone.0016804.t002 Table 2 ::: {.caption} ###### Comparison of 'new' and 'current' biomarkers by logistic regression. ::: {#pone-0016804-t002-2} Statistic New Biomarkers (n) Current biomarker (n) New + Current ------------------------ -------------------- ----------------------- ------------------- ------------------- ------------------- ------------------- Area under ROC (95%CI) 0.90 (0.85--0.94) 0.86 (0.80--0.92) 0.82 (0.76--0.89) 0.80 (0.73--0.88) 0.78 (0.71--0.85) 0.93 (0.89--0.97) Sensitivity (95%CI) 85.9 (77.1--91.8) 85.1 (75.4--91.5) 85.1 (75.4--91.5) 85.9 (77.1--91.8) 85.3 (76.2--91.4) 85.3 (76.2--91.4) Specificity (95%CI) 71.6 (59.1--81.7) 65.7 (53.0--76.6) 64.2 (51.5--75.3) 64.1 (51.0--75.4) 48.4 (35.9--61.2) 87.5 (76.3--94.1) Positive LR (95%CI) 3.0 (2.1--4.5) 2.5 (1.8--3.5) 2.4 (1.7--3.3) 2.4 (1.7--3.3) 1.7 (1.3--2.1) 6.8 (3.5--13.1) Negative LR (95%CI) 0.2 (0.1--0.3) 0.2 (0.1--0.4) 0.2 (0.1--0.4) 0.2 (0.1--0.4) 0.3 (0.2--0.5) 0.2 (0.1--0.3) PPV (%) (95%CI) 81.7 (72.7--88.4) 76.3 (66.4--84.1) 75.5 (65.6--83.4) 78.7 (69.6--85.8) 71.1 (61.7--79.0) 91.0 (82.6--95.8) NPV (%) (95%CI) 77.4 (64.7--86.7) 77.2 (63.8--86.8) 76.8 (63.3--86.6) 74.5 (60.7--84.9) 68.9 (53.2--81.4) 80.0 (68.4--88.3) AIC 0.81 0.96 1.03 1.06 1.23 0.70 Accuracy 80.1 76.6 76 77.3 70.4 86.2 Note: Accuracy was defined as true positives + true negatives/total number tested. Cut-off was chosen to achieve a sensitivity of ∼85% [@pone.0016804-Consensus1], [@pone.0016804-Mattsson1]. Positive/negative LR results are robust to variations in prevalence rates between the study and overall population, and are provided in addition to PPV/NPV that are highly dependent on the population tested. ::: Among the new group of biomarkers, T-Tf (WB) has by far the highest diagnostic accuracy relative to other individual biomarkers. Thus, T-Tf (WB) yields an AUC of 0.90 that is significantly higher than corresponding values for other biomarkers with the exception of Tf-1. The AIC associated with the logistic regression model using T-Tf (WB) is lowest across all other models. Finally, the specificity of T-Tf (WB) is significantly higher than that of T-Tau and equivalent to that of Tf-1 and Tf-β2. However, T-Tf (ELISA) does not have the same superior characteristics: the model has a lower AUC (0.81), higher AIC (1.06), and poorer specificity of 64.1% for reasons mentioned above. Optimal prediction of sCJD {#s3c} -------------------------- Past research shows that relative to any single biomarker, using two biomarkers in combination generally enhances the prediction of individuals likely to suffer from AD or sCJD [@pone.0016804-Skinningsrud1], [@pone.0016804-Mattsson1]. Accordingly, different combinations of new biomarkers and T-tau were evaluated for their diagnostic accuracy. An optimal predictive model is achieved when T-Tf (WB) is used in conjunction with T-Tau ([Figure 4](#pone-0016804-g004){ref-type="fig"}). This combination yields an AUC of 0.93 (highest of all combinations except for T-Tf (WB) where the difference is insignificant), specificity of 87.5% (highest of all combinations), and AIC of 0.7 (lowest of all combinations) ([Figure 4](#pone-0016804-g004){ref-type="fig"}, panel A, [Table 2](#pone-0016804-t002){ref-type="table"}). The addition of a second predictor variable makes the model less parsimonious compared to models including only a single biomarker, which normally, all else being equal, increases the AIC. The fact that this model has the lowest AIC is therefore remarkable; it indicates that the gain in fit to the data compared to other models more than makes up for the decrease in parsimony. The superiority of this combination over individual biomarkers is also clear in its positive LR (6.8) which is significantly higher than all other biomarkers, positive predictive value of 91%, and higher accuracy of 86.2% relative to each biomarker individually ([Table 2](#pone-0016804-t002){ref-type="table"}, [Figure 4](#pone-0016804-g004){ref-type="fig"}). Finally, an equation was constructed based on logistic regression analysis for the combination of T-Tf (WB) and T-tau such that the sensitivity for CJD+ would approximate 85% [@pone.0016804-Consensus1], [@pone.0016804-Mattsson1]. To visualize the discriminative power of T-Tf (WB) (y) and T-tau (x) combination, this equation was plotted as a line in the scatter plot of CJD, AD and other dementia cases ([Figure 4](#pone-0016804-g004){ref-type="fig"}, panel C). ::: {#pone-0016804-g004 .fig} 10.1371/journal.pone.0016804.g004 Figure 4 ::: {.caption} ###### Potential of CSF T-Tf alone or in combination with T-tau as a diagnostic test for CJD+. (**A & B**) ROC curve of T-Tf (WB) and T-Tf (WB) in combination with T-tau. (**C**) Scatter graph shows separation of CJD+ from cases of AD and other dementias. Reference line shows the cut-off equation derived to achieve a sensitivity of ∼85%. :::  ::: Discussion {#s4} ========== This report describes the accuracy of new biomarkers CSF T-Tf (WB) including its two isoforms Tf-1 and Tf-β2 in identifying CJD+ from dementia of other causes. Using rigorous statistical modeling and analysis of new and currently used biomarker T-tau, we demonstrate that each of the new biomarkers is superior to T-tau, and the combination of T-Tf (WB) and T-tau is more accurate in identifying sCJD than the currently used combination of 14-3-3 and T-tau. Below, we discuss distinctive features of the methodology used for comparing new and current biomarkers, possible reasons for high specificity of new biomarkers when used singly and significant improvement in diagnostic accuracy when combined with T-tau, and limitations of our study leading to directed avenues in the identification of accurate CSF biomarker(s) for sCJD. A rigorous statistical approach was used to examine the significance (nonparametric M--W test), quantitative comparison using uniform criteria (logistic regression), and sensitivity-specificity interdependence (ROC) of different biomarkers. In contrast to specific cut-off values for individual biomarkers, a preset sensitivity of 85% was applied to all biomarkers to provide a common quantitative baseline for comparison [@pone.0016804-Consensus1], [@pone.0016804-Mattsson1]. Though this method has been used in the past for evaluation of biomarkers of AD [@pone.0016804-Mattsson1], to our knowledge, this is the first report where a similar uniform approach has been applied to compare different biomarkers of sCJD. By this approach, the new biomarkers T-Tf (WB) and its isoforms Tf-1 and Tf-β2 performed better than the current biomarker T-tau in all parameters tested, i.e. sensitivity, specificity, AUC, AIC, and accuracy. Comparison among the new biomarkers revealed that T-Tf (WB) is superior to Tf-1 and Tf-β2 in all of the above parameters. This is probably because the ratio of Tf-1 and Tf-β2 varies between different cases, and the sum of these two isoforms represented by T-Tf provides a more accurate estimate of CSF Tf levels. The relative superiority of new biomarkers over T-tau is probably due to their association with sCJD associated pathology, in particular imbalance of brain iron homeostasis that has emerged as a common feature of sCJD and prion disease affected experimental models [@pone.0016804-Singh1]--[@pone.0016804-Kell2]. Diseased brains show a phenotype of iron deficiency in the presence of increased brain iron, suggesting sequestration of iron in biologically unavailable complexes [@pone.0016804-Singh1]. Reflection of this abnormality as altered levels of CSF Tf therefore provides a disease associated biomarker, a significant improvement over the surrogate biomarkers currently used for the diagnosis of sCJD [@pone.0016804-Green1]--[@pone.0016804-VanEverbroeck1], [@pone.0016804-Geschwind2]--[@pone.0016804-Castellani1]. However, instead of the compensatory increase in CSF Tf in response to brain iron deficiency as observed in cases of Restless Leg Syndrome [@pone.0016804-Mizuno1]--[@pone.0016804-ClardyS1], CSF Tf is significantly reduced in sCJD cases. A simplistic explanation for this observation is increased uptake by the iron starved brain tissue, reducing levels of Tf in the brain interstitial fluid and CSF that form a common compartment. However, this is difficult to comprehend given the high metabolic demand of brain cells for iron and the consequent tight regulation of brain iron metabolism [@pone.0016804-Moos1], [@pone.0016804-MacKenzie1]. It is likely that accumulated iron in PrP^Sc^-protein complexes mitigates the signal for iron deficiency, reducing Tf secretion [@pone.0016804-Singh1], [@pone.0016804-Moos1], [@pone.0016804-MacKenzie1]. We did not note a significant difference in iron saturation of Tf between the two groups, ruling out dysfunction of brain Tf as the underlying cause of brain iron deficiency. It is surprising that age and duration had insignificant effects on CSF T-Tf levels. This result is counter-intuitive given the increase in brain Tf with disease progression [@pone.0016804-Singh1], [@pone.0016804-Duguid1]. Our inability to detect a change in CSF Tf with duration may simply be due to the late stage in the incubation period when CSF samples were collected. A progressive decrease in CSF Tf may become apparent if a single patient is followed over time. Regardless of the underlying cause, the difference in CSF Tf between CJD+ and CJD- cases is noted much before end-stage disease, providing a useful pre-mortem diagnostic biomarker for sCJD. When combined with the surrogate biomarker T-tau, the diagnostic accuracy of T-Tf (WB) and T-tau improves significantly, and is superior to the reported accuracy of 14-3-3 and T-Tau combination [@pone.0016804-Jesse1], [@pone.0016804-Pennington1], [@pone.0016804-Satoh1]. The dramatic improvement in diagnostic accuracy of T-Tf and T-tau combination is probably due to the distinct pathological processes represented by each biomarker; T-Tf representing brain iron status, and T-tau the extent of neuronal death [@pone.0016804-Otto2]. Since brain iron deficiency is likely to induce neuronal death, the two processes may be related, altering these biomarkers in unison and therefore complementing their diagnostic capability. Contrary to the norm, a specific cut-off value for T-Tf or T-tau was not identified to calculate the specificity and sensitivity of these biomarkers. For T-Tf we thought it pre-mature to decide on such a value due to the limited sample size. For T-tau we did not use the conventional cut-off of 1200 or 1300 pg/ml for three reasons: 1) a consistent parameter was essential for comparing T-Tf and T-tau singly or in combination, 2) majority of CJD- samples were collected in the last month before death when T-tau is likely to be released into the CSF from damaged neurons, and 3) inclusion of conditions such as AD and brain inflammation that are associated with increased levels of T-tau in the CSF [@pone.0016804-Franz1]--[@pone.0016804-Sssmuth1]. The latter two conditions also explain the low specificity of T-tau observed in our study relative to published reports [@pone.0016804-Pennington1], [@pone.0016804-Skinningsrud1], [@pone.0016804-Satoh1]. However, our samples represent a typical CJD- group presenting for diagnostic testing, and partly explain the wide variation in the specificity of T-tau reported by different centers [@pone.0016804-Heinemann1], [@pone.0016804-Wiltfang1]. This is a major challenge since most dementias associated with brain inflammation are curable if identified early, and underscores the need for a diagnostic test of high specificity to avoid misdiagnosis as sCJD for which there is currently no treatment. It is encouraging that by combining T-Tf (WB) with T-tau the specificity increases from 71.6% for T-Tf and 48.4% for T-tau to 87.5% for the T-Tf and T-tau combination, indicating that these markers complement each other in their diagnostic capability. In addition to its superior sensitivity and specificity, CSF Tf by itself and in combination with T-tau offers several additional advantages as a biomarker for sCJD: 1) CSF T-Tf reflects prion disease associated brain iron imbalance and is therefore likely to be more specific [@pone.0016804-Singh1], [@pone.0016804-Piubelli1], 2) significant decrease in CSF Tf is noted \>12 months before end-stage disease, providing an opportunity for early diagnosis, 3) Tf is relatively resistant to limited degradation by proteinase-K [@pone.0016804-Singh1], ensuring consistent results even in poorly preserved CSF samples, 4) since Tf-β2 displays significant discriminatory power, blood contaminated CSF samples can be used for testing, 5) levels of CSF Tf are in the µg/ml range relative to 14-3-3 and T-tau that are in the pg/ml range, allowing accurate detection from a small sample volume without the necessary step of pre-absorption of albumin and immunoglobulins, and 6) CSF T-Tf is quantifiable. In conclusion, reduced levels of CSF Tf reinforce previous reports indicating the association of brain iron dyshomeostasis with prion disease pathology, and offer promise as a pre-mortem diagnostic test for sCJD. Although alteration of CSF Tf in sCJD cases has been described previously [@pone.0016804-Piubelli1], [@pone.0016804-Brechlin1], the small number of cases analyzed precludes direct comparison of these reports with our results. Nevertheless, these studies provide confidence that CSF Tf holds promise as biomarker of sCJD. Evaluation of additional CSF samples from sCJD and other forms of prion disorders and comparison with cases of rapid onset dementia will validate these observations further, and probably lead to the optimization of current automated procedures for quantifying serum Tf to CSF Tf and provide a quick and sensitive pre-mortem diagnostic test for sCJD and perhaps other human and animal prion disorders. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Tf is increased in the brain tissue of CJD+ cases.** Original immunoblot of brain Tf from CJD+ and CJD- cases. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Iron saturation of CSF Tf is comparable in CJD+ and CJD- cases.** (**A**) Competition of ^59^FeCl~3~ with graded concentrations of unlabeled FeCl~3~ for binding to CSF Tf from a CJD- case. (**B**) Standard curve demonstrating iron saturation of CSF Tf from a CJD- case. (**C**) Level of ^59^Fe-Tf in the CSF of CJD+ and CJD- cases is similar. (**D & E**) Difference between iron saturation of CSF Tf from CJD+, CJD-, and AD cases is not significant. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Jagdip Singh (Case Western Reserve University) for help and guidance with statistical analysis and critical evaluation of the manuscript, Shreya Nayak for assistance, and Janis Blevins and Kay Edmonds from the NPDPSC for providing CSF samples and patient data. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was supported by funds from National Institutes of Health grant R21AG109106 to NS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: NS. Performed the experiments: AS. Analyzed the data: AJB AS. Wrote the paper: NS ′AJB AS.

PubMed Central

2024-06-05T04:04:19.149213

2011-3-9

PMC3052313

Introduction {#s1} ============ Since emerging in 1997, highly pathogenic avian influenza (HPAI) H5N1 viruses have been associated with over 500 human infections with an unprecedented fatality rate exceeding 60%. The significant virulence of these viruses, their continual evolution in birds, and the co-circulation of the pandemic H1N1 virus lead to concerns that the H5N1 viruses pose pandemic threats. The severe disease associated with HPAI H5N1 infections in humans and animals could result from several factors including dissemination of the virus beyond the respiratory tract, higher and prolonged viral replication leading to increased viral cytolytic damage, differences in the host response induced by the H5N1 viruses, or a combination of all these factors. Although host responses are clearly complex, the clinical data and animal models suggest that the innate immune responses differ in H5N1 infected individuals (reviewed in [@pone.0017377-Peiris1], [@pone.0017377-Maines1]). As compared to seasonal influenza infection H5N1 infected patients have elevated serum levels of several chemokines and cytokines [@pone.0017377-deJong1], [@pone.0017377-Peiris2], [@pone.0017377-To1]. Similar results were observed in animal models where H5N1 infection is associated with elevated cytokine and chemokine levels [@pone.0017377-Maines1], [@pone.0017377-Salomon1], [@pone.0017377-Tumpey1], [@pone.0017377-Szretter1], [@pone.0017377-VanHoeven1], enhanced recruitment of macrophages and neutrophils into the lungs leading to acute lung inflammation [@pone.0017377-Perrone1], and premature apoptosis of dendritic cells [@pone.0017377-Baskin1]. This aberrant host response is reminiscent of 1918 influenza virus infected animals [@pone.0017377-Perrone1], [@pone.0017377-Kobasa1]. A major component of the innate immune response that has not been evaluated during HPAI H5N1 infections is complement. The complement system is comprised of more than 30 proteins responsible for recognizing and eliminating pathogens while stimulating early and late cellular functions (reviewed in [@pone.0017377-Daha1]). Three biochemical pathways activate the complement system: the classical complement pathway, the alternative complement pathway, and the mannose-binding lectin pathway [@pone.0017377-Janeway1]. The hydrolysable C3 protein is the converging point for all three complement activation pathways, making it the central player in the complement cascade [@pone.0017377-Zipfel1]. Upon activation, C3 is cleaved into C3a and C3 convertase, which supports the further cleavage of C5 into C5a. C3a and C5a function similar to chemokines promoting localized attraction and activation of immune cells including neutrophils, which serve an important role in early and late defense against pathogens including influenza virus [@pone.0017377-Hartshorn1], [@pone.0017377-Hartshorn2], [@pone.0017377-Salvatore1], [@pone.0017377-Tumpey2]. A recent study by Boon et al demonstrated a protective role for complement C5 in H5N1 influenza pathogenesis [@pone.0017377-Boon1]. Thus, the goal of these studies was to fill our gap in knowledge by determining if mice infected with H5N1 influenza virus differed in complement C3 activation as compared to a seasonal or the pandemic 2009 H1N1 influenza virus and if C3 was required for protection from HPAI H5N1 influenza infection. Results {#s2} ======= H5N1 influenza virus increases C3 and C5a lung levels as compared to seasonal or pandemic influenza strains {#s2a} ----------------------------------------------------------------------------------------------------------- To quantitate C3 and C5a protein levels during influenza infection, C57BL/6 mice were lightly anesthetized and intranasally inoculated with PBS (uninfected control), seasonal H1N1 A/Puerto Rico/8/1934 (PR/8), pandemic 2009 H1N1 A/California/7/2009 (CA/09), or HPAI H5N1 A/Vietnam/1194/04 (VN/1194) and bronchioalveolar lavage (BAL) collected on days 1, 3, and 6 post-infection (pi). Using a mouse-specific C3 sandwich ELISA we found that within 3 days post-infection (dpi) C3 levels were \>3-fold higher in the H5N1 VN/1194 virus infected mice as compared to controls and continued to increase to \>4-fold by 6 dpi ([Fig. 1A](#pone-0017377-g001){ref-type="fig"}). This is in contrast to mice infected with seasonal (PR/8) or pandemic (CA/09) viruses where C3 levels never increased above control. A similar trend was observed for C5a. H5N1 VN/1194 infected mice had increased C5a levels within 3 dpi (∼2-fold) increasing to \>4-fold by 6 dpi as compared to uninfected controls ([Fig. 1B](#pone-0017377-g001){ref-type="fig"}). There was no significant increase in PR/8 and CA/09 virus infected mice. ::: {#pone-0017377-g001 .fig} 10.1371/journal.pone.0017377.g001 Figure 1 ::: {.caption} ###### Complement levels in BAL during influenza infection. On days 1, 3, and 6 post-infection, BAL was collected from mice inoculated with PBS (control) or infected with PR/8, CA/09, or VN/1194 influenza virus and analyzed for complement C3 (A) or C5a levels (B) by sandwich ELISA. Error bars represent standard deviation. (C) C3 processing in the BAL of uninfected and influenza-infected mice at different times post-infection were analyzed by western blot analysis and results quantitated by densitometry (D). Results are representative of 2 separate experiments. :::  ::: To monitor C3 processing in the BAL, immunoblotting was performed under reducing conditions with a polyclonal anti-C3 antibody and protein loading determined by anti-actin antibody ([@pone.0017377-Morrison1]; [Fig. 1C](#pone-0017377-g001){ref-type="fig"} and [Fig. S1](#pone.0017377.s001){ref-type="supplementary-material"}). To quantitate differences, densitometry was performed as compared to actin and results expressed as the fold change over uninfected controls ([Fig. 1D](#pone-0017377-g001){ref-type="fig"}). The anti-C3 antibody detected 2 major proteins at ∼65 and ∼43 kDa, which are likely βC3 and iC3b respectively. Mouse/rat C3 byproducts run at slightly different sizes than human [@pone.0017377-Sahu1], [@pone.0017377-Kang1]. Similar to the C3 ELISA data, H5N1 VN/1194 infected animals had increased levels of both βC3 and iC3b in BAL at 3 and 6 dpi ([Fig. 1C and D](#pone-0017377-g001){ref-type="fig"}). Mice infected with seasonal PR/8 virus had levels at or below control uninfected throughout the course of infection. Comparable results were observed with the pandemic CA/09 infected mice except for a dramatic increase in iC3b levels at 3 dpi only. The reason for this is unclear and no increase was seen in the ELISA. Parallel trends were seen in BALB/c mice and mice infected with other strains of HPAI H5N1 virus suggesting that these results are not mouse or H5N1 VN/1194 strain specific (data not shown). In summation, these results suggest that H5N1 virus elicits higher levels of C5a and C3 in the lungs of infected mice. Increased morbidity and exacerbated inflammation in the lungs of C3^−/−^ infected mice {#s2b} -------------------------------------------------------------------------------------- Increased complement has been associated with enhanced inflammation and tissue destruction [@pone.0017377-Morrison2]. Thus, we hypothesized that the elevated complement levels in the HPAI H5N1 infected mice could be involved in the prominent inflammatory response associated with these infections. To evaluate this, wild-type (WT) and C3 knockout mice (C3^−/−^) mice were intranasally inoculated with PBS (control), CA/09, or H5N1 VN/1194 virus, lungs collected at 3 ([Fig. 2](#pone-0017377-g002){ref-type="fig"}) and 6 dpi ([Fig. 3](#pone-0017377-g003){ref-type="fig"}) and histopathology performed. To evaluate the infiltrating cell population, tissues were stained for macrophages and neutrophils and quantitated by ImageScope. By 3 dpi, both the CA/09 ([Fig. 2H](#pone-0017377-g002){ref-type="fig"}) and VN/1194 ([Fig. 2N](#pone-0017377-g002){ref-type="fig"}) infected C3^−/−^ mice had enhanced inflammation as compared to WT ([Fig. 2G and 2M](#pone-0017377-g002){ref-type="fig"}) with the VN/1194 being more severe. In the CA/09 infected C3^−/−^ mice, this was associated with increased numbers of macrophages ([Fig. 2J](#pone-0017377-g002){ref-type="fig"}) as compared to WT ([Fig. 2I](#pone-0017377-g002){ref-type="fig"}). Quantitative analysis showed an increase from ∼17% positive cells in WT to ∼34% in C3^−/−^ mice ([Fig. 4A](#pone-0017377-g004){ref-type="fig"}). There was no change in neutrophil levels ([Fig. 2K and L](#pone-0017377-g002){ref-type="fig"}). Tissues from uninfected control animals are shown in [Fig. 2A--F](#pone-0017377-g002){ref-type="fig"}. ::: {#pone-0017377-g002 .fig} 10.1371/journal.pone.0017377.g002 Figure 2 ::: {.caption} ###### Increased inflammation and innate immune cells of influenza infected C3^−/−^ mice at 3 dpi. At 3 days post-infection, lung tissue was collected, formalin-fixed, and paraffin-embedded. Sections (4 µm thick) were stained with hematoxylin and eosin (A, B, G, H, M, N) or for macrophages (C, D, I, J, O, P), or neutrophils (E, F, K, L, Q, R). Representative pictures of each group are shown at 20× magnification. :::  ::: ::: {#pone-0017377-g003 .fig} 10.1371/journal.pone.0017377.g003 Figure 3 ::: {.caption} ###### Increased inflammation and innate immune cells of influenza infected C3^−/−^ mice at 6 dpi. At 6 days post-infection, lung tissue was collected, formalin-fixed, and paraffin-embedded. Sections (4 µm thick) were stained with hematoxylin and eosin (A, B, G, H, M, N) or for macrophages (C, D, I, J, O, P), or neutrophils (E, F, K, L, Q, R). Representative pictures of each group are shown at 20× magnification. :::  ::: ::: {#pone-0017377-g004 .fig} 10.1371/journal.pone.0017377.g004 Figure 4 ::: {.caption} ###### Increased morbidity in influenza infected C3^−/−^ mice. \(A) Digital images of the slides was obtained and the percentage of positive nuclei in four random sections of the lung for each animal were determined with ImageScope using a nuclear-based algorithm. (B) C57BL/6 (WT) or C3^−/−^ mice were inoculated i.n. with equivalent MID~50~ units of CA/09 (10^5^ TCID~50~, n = 14) or VN/1194 (10^4^ TCID~50~, n = 14) influenza virus and weights monitored every 48 hpi (A). Error bars represent standard error deviation. :::  ::: Although inflammation was more apparent in the H5N1 VN/1194 infected C3^−/−^ mice, this was not associated with significant differences in the numbers of macrophage or neutrophil as compared to WT ([Fig. 2O--R](#pone-0017377-g002){ref-type="fig"}). Both strains of mice had minimal numbers of macrophages and ∼10% staining for neutrophils ([Fig. 4A](#pone-0017377-g004){ref-type="fig"}). These numbers were significantly lower than those seen in CA/09 infected animals; regardless of the strain. By 6 dpi, there was increased inflammation in the C3^−/−^ mice as compared to WT; including uninfected C3^−/−^ mice ([Fig. 3B versus 3A](#pone-0017377-g003){ref-type="fig"}). The CA/09 infected WT mice exhibited mild bronchiolitis and perivasculitis ([Fig. 3G](#pone-0017377-g003){ref-type="fig"}) comprised of both macrophages ([Fig. 3I](#pone-0017377-g003){ref-type="fig"}) and neutrophils ([Fig. 3K](#pone-0017377-g003){ref-type="fig"}). In contrast, the CA/09 infected C3^−/−^ mice had moderate to moderately severe bronchitis, bronchiolitis, and vasculitis, with perivasculitis, consolidation ([Fig. 3H](#pone-0017377-g003){ref-type="fig"}) and increased numbers of macrophages ([Fig. 3J](#pone-0017377-g003){ref-type="fig"}). Quantitative analysis showed an increase from ∼17% positive cells in WT to ∼42% in C3^−/−^ mice ([Fig. 4A](#pone-0017377-g004){ref-type="fig"}). What was most evident was the decrease in neutrophil numbers in the C3^−/−^ mice from ∼22% in WT to ∼10% in C3^−/−^ mice ([Fig. 4A](#pone-0017377-g004){ref-type="fig"}). Although the pathologic changes were similar, the most prominent difference between VN/1194 infected WT and C3^−/−^ mice was the extent of the damage. VN/1194 infected WT mice had increased inflammation and damage in localized areas of the lung, primarily at the edges of the bronchi ([Fig. 3M](#pone-0017377-g003){ref-type="fig"}). In contrast, infected C3^−/−^ mice displayed a more diffuse, non-focal pathology with increased interstitial involvement ([Fig. 3N](#pone-0017377-g003){ref-type="fig"}). This was associated with an increase in neutrophils ([Fig. 3R](#pone-0017377-g003){ref-type="fig"} versus 3Q). Neutrophil levels increased from ∼25% in WT to ∼40% in C3^−/−^ mice ([Fig. 4A](#pone-0017377-g004){ref-type="fig"}). To determine if the changes in inflammation played a role in pathogenesis, infected mice were monitored for weight loss ([Fig. 4B](#pone-0017377-g004){ref-type="fig"}). By 8 dpi, the influenza infected C3^−/−^ mice had lost \>25% of their day 0 weights (CA/09 *p*\<0.009, VN/1194 *p*\<0.05) and had to be humanely euthanized. In contrast, the infected WT mice never lost more than 8% of their day 0 weights (CA/09 *p*\<0.005, VN/1194 *p*\<0.0001). Delayed viral clearance in C3^−/−^ infected mice {#s2c} ------------------------------------------------ Finally to better understand the mechanism of C3-mediated protection, viral titers were measured in the lungs on days 3, 6, and 8 pi ([Fig. 5](#pone-0017377-g005){ref-type="fig"}). In CA/09-infected WT mice, lung titers were highest 3 dpi (10^6.5^ TCID~50~/ml) and decreased to approximately 10^3.5^ TCID~50~/ml by 8 dpi ([Fig. 5](#pone-0017377-g005){ref-type="fig"}). In contrast, the C3^−/−^ infected mice had significantly higher viral titers at days 6 (*p*\<0.04) and 8 pi (*p*\<0.03) suggesting a delay in viral clearance. Similar trends were observed in the VN/1194-infected C3^−/−^ mice, although these differences were not significant until 8 dpi when the viral titers in WT mice were 10^3.5^ TCID~50~/ml as compared to 10^5.75^ TCID~50~/ml (*p*\<0.02) in C3^−/−^ infected mice. Overall, these studies suggest that C3 is an important host response during influenza infection; potentially by aiding in viral clearance and regulating lung inflammation. ::: {#pone-0017377-g005 .fig} 10.1371/journal.pone.0017377.g005 Figure 5 ::: {.caption} ###### Delayed viral clearance in influenza infected C3^−/−^ mice. C57BL/6 (WT) or C3^−/−^ mice were inoculated i.n. with equivalent MID~50~ units of CA/09 (10^5^ TCID~50~, n = 14) or VN/1194 (10^4^ TCID~50~, n = 14) influenza virus. At 3, 6, and 8 days post-infection, lung homogenates were monitored for viral titers by TCID~50~ analysis on MDCK cells. Error bars represent standard deviation and asterisk (\*) indicates significant increase in viral titers as compared with WT infected mice. :::  ::: Discussion {#s3} ========== Here, we present data that the complement component C3 is required for protection from pandemic 2009 H1N1 and HPAI H5N1 influenza virus infections by aiding in viral clearance and regulating lung inflammation. Our work complements previous studies demonstrating a protective role for C3 against mouse-adapted strains of influenza virus [@pone.0017377-FernandezGonzalez1], [@pone.0017377-Jayasekera1], [@pone.0017377-Kopf1]. These studies demonstrated that complement C3 was important for T-cell priming and migration to the lung [@pone.0017377-Kopf1] and promoting the expansion of CD8+ and CD4+ T cells during infection [@pone.0017377-Nakayama1]. The lack of T cell priming, expansion and migration in the C3^−/−^ mice could explain the delayed viral clearance. The complement system is a major component of innate immunity and consists of both soluble factors and cell surface receptors that interact to sense and respond to invading pathogens. Three general activation pathways, referred to as the classical, alternative, and lectin pathways, converge on C3 (composed of an alpha and beta chain), the central component of the complement system. Formation of C3 convertases leads to cleavage of C3 to its active fragments, the anaphylatoxin C3a and the opsonin C3b. This cleavage event exposes a reactive thioester that allows covalent attachment of C3b to target surfaces. C3b can be further cleaved into the signaling fragments iC3b, C3dg, and C3d which regulate phagocytosis and a variety of other immune cell effector functions. In addition, C3b binds the C3 convertases resulting in a substrate change from C3 to C5, which is cleaved to the anaphylatoxin C5a and the initiator of the membrane attack complex C5b. Although it was clear that complement component C3 was required for host defense, the levels of C3 and complement activation fragments in BAL differed amongst the different influenza virus infections. We failed to detect any significant increase in total C3 levels in the BAL of PR/8 or 2009 pandemic CA/09 virus infected mice as compared to uninfected controls using a mouse-specific C3 sandwich ELISA. In contrast, increased levels of total C3 were detected in the BAL of mice inoculated with H5N1 VN/1194. Increased levels of C5a, a marker of complement activation and a potent anaphylatoxin, were also detected in the BAL of H5N1 VN/1194 inoculated mice, but not mice inoculated with either PR/8 or CA/09. To further investigate activation of the complement system by the different influenza virus infections, immunoblot analyses of BAL were performed. Consistent with the C3 and C5a ELISA data, increased levels of the C3 beta chain, which is not proteolytically processed and serves as a marker for total C3 levels, and iC3b, which is derived from the proteolysis of the C3 alpha chain and serves as a marker of complement activation, were detected in BAL collected from mice inoculated with H5N1 VN/1194. In addition, immunoblot analyses detected an increase in iC3b levels at 3 dpi in BAL collected from CA/09 virus infected mice. Taken together, these results suggest that CA/09 and H5N1 VN/1194 are more powerful activators of the complement system compared to PR/8 infection. The reasons for the differences amongst the viruses and the viral genes involved are under investigation. Increased complement activation can be associated with enhanced viral pathogenesis. For example, compared to WT animals, complement-depleted mice infected with Sindbis virus had less morbidity and mortality despite increased viral replication and spread, suggesting a potentially pathologic role for complement [@pone.0017377-Hirsch1], [@pone.0017377-Hirsch2]. Similar results occurred with the arthritogenic alphavirus Ross River Fever virus in which complement-deficient mice exhibited far less severe disease signs and tissue damage than WT mice despite similar viral titers [@pone.0017377-Morrison2], [@pone.0017377-Morrison3], [@pone.0017377-Morrison4]. Because of this, we hypothesized that the increased complement levels in the HPAI H5N1 infected mice could be associated with the enhanced inflammation associated with these infections. Pathological examination demonstrated that the H5N1 VN/1194 infected WT mice had enhanced inflammation as compared to the pandemic 2009 H1N1 CA/09 infection within 3 dpi. However, the inflammation was exacerbated in the infected C3^−/−^ animals suggesting that C3 plays a protective role during influenza infection. In summary, the results of our studies suggest that complement contributes to protection against 2009 pandemic H1N1 and HPAI H5N1 influenza infection in mice. Given that complement polymorphisms are common and can be associated with increased susceptibility to several infectious diseases [@pone.0017377-Coffey1], [@pone.0017377-Mollnes1], [@pone.0017377-Schneider1], [@pone.0017377-Thio1], it would be interesting to determine the state of complement activation during influenza infection in humans and assess the role in disease severity. Materials and Methods {#s4} ===================== Ethics statement {#s4a} ---------------- All procedures involving animals were approved by the University of Wisconsin-Madison School of Medicine and Public Health, and the St. Jude Children\'s Research Hospital IACUC\'s and was in compliance with the Guide for the Care and Use of Laboratory Animals. These guidelines were established by the Institute of Laboratory Animal Resources and approved by the Governing Board of the U.S. National Research Council. Laboratory facilities {#s4b} --------------------- All experiments using H5N1 viruses were conducted in a Biosafety level 3 enhanced containment laboratory [@pone.0017377-Richmond1]. Investigators were required to wear appropriate respirator equipment (RACAL, Health and Safety Inc., Frederick, MD). Mice were housed in HEPA-filtered, negative pressure, vented isolation containers. Viruses and cells {#s4c} ----------------- A/Puerto Rico/8/34 (PR/8) and A/California/04/2009 (CA/09, provided by Dr. Jacco Boon, St. Jude Children\'s Research Hospital) H1N1 viruses were propagated in the allantoic cavity of 10-day-old specific pathogen-free embryonated chicken eggs at 37°C. Allantoic fluid was harvested, clarified by centrifugation and stored at −70°C. A/Vietnam/1194/2004 (VN/1194) (provided by Dr. Alexander Klimov; CDC, Atlanta, Georgia) was propagated in Madin Darby canine kidney (MDCK) cells as described [@pone.0017377-Jones1]. MDCK cells were cultured in Eagle\'s minimum essential medium supplemented with 4.5 g glucose per liter, 2 mM glutamine (Mediatech), and 10% FBS (Gemini BioProducts, West Sacramento, CA) and grown at 37°C under 5% CO~2~. Viral titers {#s4d} ------------ Viral titers were determined by tissue culture infectious dose 50 (TCID~50~) assays on MDCK cells as described [@pone.0017377-Jones1] and quantitated by Reed and Muench analysis [@pone.0017377-Reed1]. Results are expressed as the mean log~10~ TCID~50~/ml. Animal experiments {#s4e} ------------------ 10--12 week of age female C57BL/6J (WT) or C3^−/−^ (B6.129S4-*C3^tm1Crr^*/J) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). For infection, mice were lightly anesthetized using isofluorane and intranasally (in) inoculated with PBS or equivalent mouse infectious dose 50 (MID~50~) of PR/8 (1×10^4^ TCID~50~), CA/09 (1×10^5^ TCID~50~), or VN/1194 (1×10^4^ TCID~50~) in 25 µl PBS. Mice were monitored daily for clinical signs of infection [@pone.0017377-Morton1] and weighed every 48 hpi. At different times post-infection, two control and three infected mice were anesthetized, terminally bled, and lungs harvested. The large lung lobe was washed and immediately stored in 10% buffered formalin for histological analysis. The remaining lobe was homogenized in 1 ml PBS and viral titers determined by TCID~50~ analysis. Histopathology and immunohistochemical staining {#s4f} ----------------------------------------------- Tissues were fixed in 10% neutral buffered formalin solution, processed, and paraffin embedded. Four micron thick sections were stained with hematoxylin and eosin or for macrophages (Mac2/Galectin 2, 1∶1000, Abcam, Cambridge, MA) or neutrophils (NeuN, 1∶1000, Abcam) by the St. Jude Veterinary Pathology Core Facility. Digital images were obtained with ScanScope (Aperio, Vista, CA) and the percentage of positive nuclei in four random sections of the lung for each animal were determined with ImageScope using a nuclear-based algorithm. Complement component C3 and C5a ELISA {#s4g} ------------------------------------- Protein levels were determined in BAL by BCA Protein Assay (Pierce, Rockford, IL) and C3 and C5a levels were quantitated by sandwich ELISAs (GenWay Biotech Inc, San Diego, CA and R&D Systems, Minneapolis, MN respectively) using equivalent protein concentrations following manufacturer\'s instructions. Western blot {#s4h} ------------ Equivalent volumes (30 µl) of BAL were separated by SDS-PAGE under reducing conditions and transferred to a nitrocellulose membrane using the iBlot Dry Blotting system (Invitrogen, Carlsbad, CA). The membrane was blocked with 5% dry milk in TBS containing 0.1% Tween-20 (TTBS), and probed with goat anti-mouse C3 antibody (1∶1000; Cappel) or goat anti-actin antibody (1∶1000 Santa Cruz) overnight at 4°C. After extensive washes, membranes were incubated with a donkey anti-goat HRP-conjugated antibody (1∶10,000; Southern Biotechnology, Birmingham, AL) for 1 h at room temperature, and signal was detected using the SuperSignal West Pico Chemiluminescent Substrate (Pierce). Densitometry was performed on scanned immunoblot images using the ImageJ gel analysis tool. The gel analysis tool was used to obtain the absolute intensity (AI) for each experimental complement band and corresponding actin control band. The ratios for each experimental band was calculated by dividing the corresponding control AI by the experimental AI to determine the fold change as compared to the uninfected control samples. Statistical analyses {#s4i} -------------------- Statistical significance of data was determined by using analysis of variance (ANOVA) or Student\'s *t*- test on GraphPad Prism (San Diego, CA). All assays were run in triplicate and are representative of at least 2 separate experiments. Error bars represent standard deviation and statistical significance was defined as a *p* value of less than 0.05. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Complement activation products in BAL during influenza infection.** On days 1, 3, and 6 post-infection, BAL was collected from mice inoculated with PBS (control) or infected with PR/8, CA/09, or VN/1194 influenza virus and analyzed for complement C3 activation products by western blot analysis. Results are representative of 2 separate experiments. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors would like to thank members of the Schultz-Cherry lab including Lindsey Moser, Erik Settles, and Crystal Tobin for critical review of this manuscript and Pam Freiden, Brad Seufzer, Jeremy Jones, Sean Cherry, and Alexandra Rivera for expert technical assistance. We gratefully acknowledge the expertise and assistance of Drs. Charles Czuprynski, Zsuanna Fabry, Annette Gendron-Fitzpatrick, Lisa Fox-Brown, Subbu Hegde, John Mansfield, and Chris Olsen with microscopy, histology and critical analysis of experimental procedures. We also thank the St. Jude Children\'s Research Hospital\'s Veterinary Pathology Core and Cell and Tissue Imaging Center for immunohistochemical staining and microscopy assistance. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by a Howard Hughes Medical Institute Gilliam Fellowship, an Advanced Opportunity Fellowship (AOF) from the University of Wisconsin-Madison, and a Sigma Xi Grants-in-Aid-of-Research grant to K. O\'Brien, NIH grant AR047190 to M.T. Heise, and NIH AI059049, and in part by the NIH NIAID contract number HHSN266200700005C and the American Lebanese Syrian Associated Charities (ALSAC) to S. Schultz-Cherry. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: KBB TEM MTH SSC. Performed the experiments: KBB DYD SSC. Analyzed the data: KBB TEM DYD MTH SSC. Contributed reagents/materials/analysis tools: MTH. Wrote the paper: KBB SSC.

PubMed Central

2024-06-05T04:04:19.152424

2011-3-9

PMC3052314

Introduction {#s1} ============ Antiretroviral treatment (ART) has saved millions of lives by transforming HIV infection from a fatal into a chronic disease [@pone.0017518-World1] and the vast majority (97%) of patients in sub-Saharan Africa receive a non-nucleoside reverse transcriptase (NNRTI) based regimen as first-line treatment [@pone.0017518-Harries1]. Given the costs associated with second-line protease inhibitors, the long-term sustainability of ART in many of low- and some middle-income countries depends on finding feasible ways for early detection of treatment failure to maintain patients on first-line regimens [@pone.0017518-Simon1], [@pone.0017518-Kuritzkes1]. Since viral load (VL) monitoring is not currently accessible in most resource-limited high-endemic contexts [@pone.0017518-Harries1], [@pone.0017518-Calmy1], patients are often continued on first-line ART until the emergence of clinical symptoms or until any of the World Health Organization (WHO) criteria for immunologic failure are met. Although virologic failure and drug resistance usually precede immunologic failure, these measures are not always well correlated at clinical follow-up [@pone.0017518-Rewari1], [@pone.0017518-Kantor1], [@pone.0017518-Hosseinipour1]. To routinely assess the effectiveness of ART at HIV treatment clinics and to minimize preventable HIV drug resistance (HIVDR), WHO recommends using available site-based data from medical and pharmacy records, e.g. on-time adherence to monthly ART drug pick-up and clinic appointment-keeping [@pone.0017518-World2], [@pone.0017518-Hedt1], as an early warning indicator (EWI) of inconsistent drug exposure. Failure to pick up drugs on time serves as a proxy for treatment interruption and suboptimal drug concentrations, which are associated with virologic failure and the evolution of drug resistance [@pone.0017518-Nachega1], [@pone.0017518-Cambiano1], [@pone.0017518-Parienti1]. Earlier reports from resource-limited settings have defined virologic failure as a VL of \>400, \>1,000 or \>5,000 copies/ml at one or two repeated visits [@pone.0017518-Barth1], [@pone.0017518-ElKhatib1]. However, drug resistance mutations can emerge at lower VL levels [@pone.0017518-Hatano1] and in high-income countries, monitoring guidelines recommend using VL \>50 copies/ml as an indicator of virologic failure for patients on ART [@pone.0017518-Hammer1]. As more robust, sensitive and lower cost assays are developed, ART programmes in low- and middle-income may be able to adopt lower threshold values for virologic failure. Thus, we assessed the proportion of repeated VL \>50 copies/ml, immunologic failure and median CD4 cell count gains in a cohort of long-term, first-line recipients in Soweto, who had been on NNTRI-based regimens for up to eight years. We also assessed the relationship between cumulative adherence to drug-refill visits and treatment failure. Methods {#s2} ======= Setting and study population {#s2a} ---------------------------- At the time of the study, more than 1,500 patients were receiving ART at this clinic. Aiming for a precision of +/− 5% for a proportion of 50% using 95% confidence interval, a sample size of 384 patients was required [@pone.0017518-Rothman1]. Adding a margin to the minimum required sample size, a cohort of 458 participants on long-term ART (range 12 to 99 months) were enrolled during March-September 2008 [@pone.0017518-ElKhatib2] for a cross-sectional assessment and a retrospective review of their medical charts. The inclusion criteria were: being on NNRTI-containing regimen for ≥12 months; consent to an interviewer-mediated questionnaire; medical record review; and withdrawal of 10ml of blood. The study, named South African Virologic Evaluation (SAVE) [@pone.0017518-ElKhatib1], was performed at the Perinatal HIV Research Unit (PHRU) adult HIV clinic at the Chris Hani Baragwanath Hospital, in the Soweto township outside Johannesburg, South Africa. The study site [@pone.0017518-Martinson1] and a cross-sectional description of viremia and drug resistance have been described in depth previously [@pone.0017518-ElKhatib1], [@pone.0017518-ElKhatib3]. Data collection {#s2b} --------------- Patients were interviewed at study enrolment to obtain demographics, socioeconomic characteristics, reasons for missing doses and the strategies used to remember taking their pills on time. The retrospective review included: year of HIV diagnosis; pre-ART initiation characteristics (VL, CD4 cell count, pre-exposure to single-dose nevirapine (sdNVP) or other antiretroviral drugs (ARVs); current and previous TB therapy; dates for drug refill visits; and treatment interruptions. VL and CD4 counts had been measured every six months on average. For two of the patients, medical records were not available and they were excluded from the longitudinal analysis, leaving 456 patients for data analyses. A survey form, called Medical Records-SAVE (MR.SAVE©), was developed, piloted and modified [@pone.0017518-ElKhatib4] for data collection. EpiData [@pone.0017518-ElKhatib5], [@pone.0017518-EpiData1] was used for data entry. Definitions {#s2c} ----------- The clinic uses a private diagnostic laboratory service that provides VL measures with the Versant HIV-1 RNA 3.0 (Siemens Deerfield, IL, USA) based on bDNA technology, with a lower limit of detection of 50 copies/ml. Virologic failure was defined as two repeated VL \>50 copies/ml at any time after three months on ART. Immunologic failure was defined according to the WHO guidelines as having either (i) a CD4 cell count of \<100 cells/µl post six months on ART, (ii) a CD4 cell count of less or equal to CD4 pre-ART after six months on ART or (iii) \>50% reduction from the on-ART peak CD4 cell count [@pone.0017518-World3]. Patients were dispensed ART monthly and at scheduled doctor\'s visits. They were also given projected monthly pharmacy refill dates and the date of the next doctor\'s appointment. At each refill visit, pharmacy staff dispensed pills and recorded the date. To estimate adherence, we calculated the total number of days that the patient was late for the drug-refill visits divided by the total duration on ART. The formula was: \[The number of days late for drug-refill visit  =  (Date when the patient came for drug refill - Date of the pre-scheduled appointment indicated on the patient\'s medical record)\]. The results were then summarized for repeated refill visits to obtain the cumulative number of days coming late per client. To estimate adherence, the following formula was used \[The cumulative number of days coming late ×100)/Total number of days the patient was assumed to be exposed to ART given the dispensed number of pills\]. Appointment dates were censored after the date of virologic failure. Incomplete versus complete adherence was defined as a cumulative adherence to drug refill visits of \< and ≥95% respectively. Treatment interruption was defined as a reported and/or planned history of stopping and resuming therapy, identified through chart review. Statistical analysis {#s2d} -------------------- Descriptive analyses including median (IQR) for numerical variables, frequencies and proportions for categorical variables were performed. Bivariate analyses to assess risk factors for virologic and immunologic failure were performed using Pearson Chi Square and Fisher\'s exact tests. Thereafter variables with a p-value ≤0.10 were added into a multivariate logistic regression model and those with a p-value \<0.05 were considered significant in the final multivariate model, calculating odds ratios (OR) and 95% confidence intervals (CIs). However, the variables sex and age were always maintained in the final multivariate models to account for possible remaining confounders. Kaplan Meier survival analysis was done using months as the time unit in order to assess time to virologic and immunologic failure on ART among all patients. Known pre-ART risk factors, exposure to sdNVP [@pone.0017518-Paredes1], [@pone.0017518-Datay1], [@pone.0017518-Coovadia1] or any type of ARVs, CD4 cell count and age [@pone.0017518-Paredes1], [@pone.0017518-Ford1], [@pone.0017518-Boulle1] were adjusted using Cox regression analysis for virologic failure. For immunologic failure, Cox regression analysis was done by adjusting for the same above-mentioned variables and any virologic failure. Due to collinearity between sex and pre-ART CD4 count, sex was not included in the final survival model. Finally, we assessed the median gain in CD4 cell count during ART on a six-monthly basis (range +/−3 months) among 1) patients with incomplete vs. complete cumulative drug-refill adherence, and 2) patients with virologic failure vs. suppression up to 36 months on ART only, due to data availability. Stata/SE College Station, Texas (version 10.1) [@pone.0017518-StataCorp1] and Graphpad Prism (version 4.0c) [@pone.0017518-Prism1] were used for data analysis. In the survival analysis, the command sts graph in STATA was used. Adjusting for covariates, the command fits separate Cox regression models for each group, and the separately calculated baseline survivor was then retrieved. Ethical approval for this study was obtained from the research ethics committees at the University of the Witwatersrand in Johannesburg, South Africa, and the Regional Medical Ethics Board in Stockholm, Sweden. Written informed consent was obtained from all patients. Results and Discussion {#s3} ====================== Patients\' characteristics {#s3a} -------------------------- A total of 456 patient records were reviewed. Most patients (79%) were diagnosed with HIV between 2001-2004 (median 2003), 14% before 2000 and the remaining 7% between 2005 and 2008. Pre-ART initiation, 51% (222/434) had a CD4 cell count of ≤100 cell/µl and 41% (172/421) had a VL of ≥100 000 copies/ml ([Table 1](#pone-0017518-t001){ref-type="table"}). Overall, the median time on ART was 44 months (IQR 38--48; 1,510 person-years) and 77% (349/456) were women ([Table 1](#pone-0017518-t001){ref-type="table"}). Eighteen percent (80/445) had been exposed to ARVs previously; 15% of the women (52/349) had received sdNVP for the prevention of mother-to-child transmission (PMTCT) with a median time before ART initiation of 15 months, and 6% of all patients (28/446) had received ART before starting the current ART regimen. Approximately half (48%, 199/414) had been treated for tuberculosis (TB) before ART initiation. At the time of study enrolment, the patients spent a median of 40 minutes (IQR 30--60) travelling to the clinic by mini-van (90%), walking (5%) and using their own car (4%). ::: {#pone-0017518-t001 .table-wrap} 10.1371/journal.pone.0017518.t001 Table 1 ::: {.caption} ###### Demographics, socioeconomic and clinical characteristics and bivariate analysis for the association with a) virologic and b) immunologic failure among 456 patients on ART in Soweto, South Africa. ::: {#pone-0017518-t001-1} Demographics and socio-economic characteristics a\) Virologic failure∧ b\) Immunologic failure∧∧ -------------------------------------------------------------------- ------------------------ --------------------------- ------ ---------- --------------------------------------- **Sex** Women 349 (77%) 68 (19%) 59 (17%) Men 107 (23%) 20 (19%) 0.87 28 (26%) 0.03 **Age** 18--24 21 (5%) 5 (24%) 0.42 3 (14%) 0.62 25--34 235 (51%) 45 (19%) 46 (20%) 35--44 147 (32%) 24 (16%) 25 (17%) ≥45 53 (12%) 14 (26%) 13 (24%) **Marital status** Have partner 265 (58%) 49 (18%) 50 (19%) Single/no partner at all 191 (42%) 39 (20%) 0.61 37 (19%) 0.89 **Born in South Africa** No 17 (4%) 4 (23%) 2 (12%) Yes 439 (96%) 84 (19%) 0.65 85 (19%) 0.43 **Education level** No education or primary schooling 50 (11%) 11 (20%) 0.61 4 (8%) Secondary or tertiary education level 406 (89%) 77 (19%) 83 (20%) 0.03[\*](#nt101){ref-type="table-fn"} **Year diagnosed with HIV** ≤2000 63 (14%) 10 (16%) 0.79 10 (16%) 0.66 2001--2004 348 (79%) 68 (20%) 70 (20%) 2005--2008 31 (7%) 6 (19%) 5 (16%) **Any type of work** No 283 (65%) 59 (21%) 60 (21%) Yes 154 (35%) 28 (18%) 0.51 23 (15%) 0.11 **Clinical characteristics** [\*\*](#nt102){ref-type="table-fn"} **CD4 cell count -- pre-ART** ≤50 122 (28%) 21 (17%) 0.59 26 (21%) 0.52 51--100 100 (23%) 24 (24%) 21 (16%) 101--249 199 (46%) 37 (19%) 31 (16%) ≥250 13 (3%) 2 (15%) 3 (23%) **Viral load level -- pre-ART (RNA copies/ml)** \<5,000 22 (5%) 2 (9%) 0.54 1 (5%) 0.26 5,000--29,999 110 (26%) 23 (21%) 24 (22%) 30,000--99,999 117 (28%) 20 (17%) 24 (21%) ≥100,000 172 (41%) 35 (20%) 30 (17%) **Pre-exposure to sdNVP or other type of ARVs** Treatment naïve 365 (82%) 64 (18%) 72 (20%) Single dose-nevirapine (sdNVP) 52 (12%) 13 (25%) 0.19 7 (13%) 0.28 Had any ARVs pre-ART initiation[\#\#](#nt106){ref-type="table-fn"} 28 (6%) 9 (32%) 0.06 6 (21%) 0.83 **Treatment experienced (sdNVP or any ARVs)** 80 (18%) 22 (28%) 0.04 13 (16%) 0.47 **TB therapy** Not treated before ART 215 (52%) 36 (17%) 38 (18%) Treated before ART 137 (33%) 27 (20%) 0.48 33 (24%) 0.14 Was on TB therapy when started on ART 62 (15%) 11 (18%) 0.85 10 (16%) 0.78 **Disclosed HIV status pre-ART** No 35 (8%) 4 (11%) 8 (23%) Yes 400 (92%) 80 (20%) 0.22 77 (19%) 0.61 **Cumulative adherence to on-time drug refill** 95-100% 430 (94%) 79 (18%) 78 (18%) \<95% 26 (6%) 9 (35%) 0.04 9 (35%) 0.04 **Any treatment interruption** No 324 (71%) 66 (20%) 58 (18%) Yes[\$](#nt107){ref-type="table-fn"} 132 (29%) 22 (17%) 0.36 29 (22%) 0.32 \*2-sided Fisher exact test; \*\*All patients were initiated on an NNRTI-based regimen (the majority had efavirenz); \# Two patients (2/458) were excluded due to missing longitudinal data; ∧Two repeated VL \>50 copies/ml post-three months on ART; ∧∧WHO criteria for immunologic failure; \#\# Two patients were exposed to sdNVP and ART, pre-ART initiation, but virologically suppressed. \$ Treatment interruption: once for 129 patients and twice for 3 patients. ::: Virologic failure {#s3b} ----------------- Overall, 19% (88/456) met the criteria for virologic failure. In bivariate analysis, being exposed to any type of ARVs including sdNVP prior to ART initiation or incomplete adherence was significantly associated with virologic failure (p = 0.04) ([Table 1a](#pone-0017518-t001){ref-type="table"}). These two factors remained significant after adjustment for confounding by age and sex in the multivariable analysis model. The odds ratio of virologic failure among patients with incomplete adherence almost tripled (adjusted OR 2.8, 95%CI 1.2--6.7) and doubled among those exposed to any type of ARVs prior to ART initiation (adj. OR 2.1, 95%CI 1.2--3.9). However, exposure to sdNVP alone did not reach statistical significance. Neither the level of CD4 cell count nor VL at pre-ART initiation was significantly associated with subsequent virologic failure on ART in bivariate analyses, and therefore was not included in the multivariable model. In a separate analysis, there was no significant association between incomplete adherence and any of the patients\' demographic, socioeconomic or clinical data (not shown). Planned treatment interruptions were not part of the clinical guidelines at this clinic. Of the 456 patients, 132 (29%) had a reported history of treatment interruption registered in their medical charts, 98% of these (129/132) only once. Three patients were reported to have experienced treatment interruption twice but were not found to fail virologically during the study period. The median duration of treatment interruption was longer among patients with, compared to those without, virologic failure; 45 days (IQR 27--97) vs. 36 days (IQR 19--63) respectively, although this difference was not statistically significant (p = 0.12). Analysing the risk of virologic failure over time using Kaplan Meier survival analysis, the virologic failure rate was 23% up until month 99 ([Figure 1a](#pone-0017518-g001){ref-type="fig"}). There was a significant difference in time to virologic failure between patients with complete vs. incomplete adherence. By month 12 on ART, the failure rate was similar (7% vs 8% among patients with complete and incomplete adherence respectively) but by month 48, the difference in failure rates had reached statistical significance between the groups, 19% vs. 37% respectively (logrank p value = 0.02) ([Figure 1a](#pone-0017518-g001){ref-type="fig"}). Following adjustment for CD4 cell count, age and exposure to any ARVs pre-ART initiation in a Cox regression analysis, there was a significant difference in virologic failure rates already at month 12 on ART; 2% vs 11%, and at month 48. This difference was even more pronounced, 18% vs. 43%, among patients with complete vs incomplete adherence respectively (logrank p value \<0.01) ([Figure 1b](#pone-0017518-g001){ref-type="fig"}). ::: {#pone-0017518-g001 .fig} 10.1371/journal.pone.0017518.g001 Figure 1 ::: {.caption} ###### (a) Kaplan-Meier survival analysis for time to virologic failure by level of cumulative adherence to drug refill visits. \(b) Cox regression analysis after adjustment for confounding by CD4 cell count, age and being exposed to sdNVP or any ART pre-ART initiation. :::  ::: Immunologic failure {#s3c} ------------------- Overall, 87/456 (19%) of the patients met one or more of the definitions of immunologic failure based on CD4 cell count [@pone.0017518-World4]. In bivariate analysis, immunologic failure was associated with incomplete adherence (p = 0.04), gender (p = 0.03), and low education level (p = 0.03) ([Table 1](#pone-0017518-t001){ref-type="table"}). However, none of these variables remained significant in the final multivariate logistic regression model. More than one third (37%; 32/87) of patients failing immunologically were also found to be viremic. Among those with immunologic failure, there was no significant difference in CD4 cell count or VL, pre-ART initiation, between the 32 viremic and 55 non-viremic patients (data not shown). Kaplan Meier survival analysis demonstrated an overall immunologic failure rate of 27% by month 48 ([Figure 2a](#pone-0017518-g002){ref-type="fig"}). The risk of immunologic failure was 41% vs 19% among those with incomplete and complete adherence respectively after 48 months on ART (logrank p value = 0.02). After adjustment for CD4 cell count, age and exposure to any ARVs pre-ART initiation and for virologic failure in the Cox regression analysis, patients with incomplete adherence had an immunologic failure rate of 90% at month 48, while the corresponding figure among patients with complete adherence was only 11% ([Figure 2b](#pone-0017518-g002){ref-type="fig"}) (logrank p value\<0.01). ::: {#pone-0017518-g002 .fig} 10.1371/journal.pone.0017518.g002 Figure 2 ::: {.caption} ###### (a) Kaplan-Meier survival analysis for time to immunologic failure, by level of cumulative adherence to drug refill visits. \(b) Cox regression analysis, adjusted for confounding by virologic failure, CD4 cell count, age and being exposed to sdNVP or any ART pre-ART initiation :::  ::: Immunologic response to ART {#s3d} --------------------------- There was a significant difference in CD4 cell count gain during ART between patients with complete vs incomplete adherence at almost all of the clinical visits (p≤0.04) ([Figure 3](#pone-0017518-g003){ref-type="fig"}). As expected, patients initiating ART with a CD4 cell count of \>100 cells/µl had a significantly higher median CD4 cell count in comparison to patients started on ART at a CD4 cell count of ≤100 cells/µl throughout the follow-up period ([Figure 4](#pone-0017518-g004){ref-type="fig"}). The median CD4 cell count was also significantly higher among patients with continuous virologic suppression patients as compared to those with virologic rebound up to at least three years on ART (months 18, 24, 30 and 36, p≤0.03) ([Figure 5](#pone-0017518-g005){ref-type="fig"}). ::: {#pone-0017518-g003 .fig} 10.1371/journal.pone.0017518.g003 Figure 3 ::: {.caption} ###### Median CD4 (IQR) for patients initiated on ART, by adherence \< or ≥95% based on drug-refill visits. :::  ::: ::: {#pone-0017518-g004 .fig} 10.1371/journal.pone.0017518.g004 Figure 4 ::: {.caption} ###### Median CD4 (IQR) for patients initiated on ART, with CD4 ≤ or \>100 cells/µl. :::  ::: ::: {#pone-0017518-g005 .fig} 10.1371/journal.pone.0017518.g005 Figure 5 ::: {.caption} ###### Median CD4 (IQR) for patients who showed virologic failure vs. those who remained suppressed. :::  ::: Barriers and facilitators for adherence {#s3e} --------------------------------------- When all patients were asked about the reasons for missing pills, the three main reasons stated were: being away from home (32%); simply forgetting (20%); and being busy with other things (10%) ([Figure 6](#pone-0017518-g006){ref-type="fig"}). Eleven other reasons were reported but all with a prevalence of ≤6%. ::: {#pone-0017518-g006 .fig} 10.1371/journal.pone.0017518.g006 Figure 6 ::: {.caption} ###### Proportion of patients (%) with self-reported reasons for not taking any of their pills in general at study enrolment (N = 458). :::  ::: Patients used a combination of methods to remember their medication; mobile phone alarms (49%), relying on their own memory (49%), a close friend/relative (20%) or a partner (8%) to remind them, or pill-boxes (1%). All interviewees denied any intentional non-adherence in order to maintain eligibility for disability support. This was despite 39% of the patients reporting that their primary source of financial support was the social welfare disability grant provided in South Africa for HIV-infected patients with a CD4 cell count below 200 cells/µl. Antiretroviral therapy is a life-long undertaking and finding feasible and affordable means for early detection of treatment failure is crucial to sustain first-line therapy effectiveness. Our study found that the estimated proportion of patients failing virologically was 2--3 times higher among patients who were late for their drug-refill visits compared to those with an adherence to drug refill above 95%. This provides evidence that failure to collect ART may serve as a proxy for reduced drug exposure over time. According to WHO [@pone.0017518-World2], [@pone.0017518-Hedt1], monitoring the extent to which ART sites function through EWIs such as adherence to on-time drug refills is of high priority in order to minimize preventable HIVDR. The usefulness and importance of the on-time drug refill indicator became most obvious in the Cox regression analysis over time. It showed a significant difference in first-year failure rates (2% vs 11%) among patients with complete adherence vs incomplete adherence to drug refill already after 12 months on treatment. This difference became even more visible after 4 years on ART (18% vs 43% virologic failure rate respectively) adjusting for CD4 cell count at ART initiation, age and exposure to any ARVs pre-ART initiation. The cumulative proportion with virologic failure in our study is comparable to similar cohorts of long-term ART-recipients in urban South African townships including other studies from an urban site in Johannesburg [@pone.0017518-Sanne1], and from Khayelitsha [@pone.0017518-Boulle1]. Similarly high failure rates have also been reported from Switzerland [@pone.0017518-Keiser1], England [@pone.0017518-Lampe1], [@pone.0017518-Lampe2], France, Spain, Germany and Canada [@pone.0017518-Lampe2]. Apart from \<95% adherence to drug refills, previous exposure to other ARVs was the only significant risk factor associated with an increased risk of virologic failure in our multivariable analysis, supporting ample existing evidence that sdNVP [@pone.0017518-Phillips1], [@pone.0017518-Datay2], [@pone.0017518-Jackson1], [@pone.0017518-Eshleman1] or other ARVs [@pone.0017518-Madge1] may predispose to virologic failure and the emergence of HIVDR mutations among women treated with NNRTI-based ARVs. During the first year on ART, CD4 T-cell restoration can be slow [@pone.0017518-Taiwo1], but we found that the rate of CD4 cell count gain up to 36 months on ART was the same regardless of the initial level of CD4. Thus, patients starting ART with advanced immuno-suppression and very low CD4 cell counts (\<100 cells/µl) maintained a significantly lower CD4 cell count level throughout the study period. This put them at risk of increased morbidity for a number of years post-ART initiation [@pone.0017518-Boulle1], [@pone.0017518-Egger1], and provides further justification to initiate ART earlier, as embodied in the recent WHO recommendations [@pone.0017518-World4]. The gain in CD4 cell count was significantly dependent on high adherence to drug refill and viral suppression supporting the findings by Bisson and co-workers [@pone.0017518-Bisson1]. However, in our study, nearly two thirds of patients failing immunologically were virologically suppressed at clinical assessment. Thus CD4 cell count is a poor predictor of virologic outcomes [@pone.0017518-Rewari1], [@pone.0017518-Kantor1], [@pone.0017518-Mee1], [@pone.0017518-Badri1] and the use of immunological criteria only for monitoring treatment responses may jeopardize clinical management. All study participants in this study had been on ART for at least 12 months, and over 90% had achieved a CD4 cell count of \>200 cells/µl. There is an ongoing debate in South Africa where some have suggested that HIV patients may keep their CD4 counts at \<200 cells/µl by intentionally missing doses [@pone.0017518-Kagee1], [@pone.0017518-Nattrass1] to retain eligibility for disability grants. However, this was firmly denied by our participants, who rather ascribed missing doses to being away from home or simply forgetting to take their pills. The interesting fact that half of our patients used their mobile phones as a medication reminder opens up future opportunities of systematic adherence support through text reminders since cell phone access and usage are rapidly increasing throughout sub-Saharan Africa [@pone.0017518-Richard1]. Additionally, dispensing ART for more than 30 days at a time may also reduce the risk of missed doses. VL measurements are rarely available in most resource-limited settings and where measured, virologic failure has often been defined as VL \>400, \>1,000 or even \>5,000 copies/ml [@pone.0017518-ElKhatib1], [@pone.0017518-Ford1], [@pone.0017518-Boulle1]. In this study, a more sensitive and conservative definition for virologic failure was used, \>50 copies/ml. While the fact that some patients with intermittent viremia (blips) were included may cause some concern, 85% (75/88) were confirmed viremic in two sequential assessments \>50 copies/ml. In this study, we used \<95% as a conservative cut-off point for incomplete adherence. In early adherence studies, including unboosted protease inhibitors (PIs), a 95% adherence level was shown to be associated with high virologic suppression [@pone.0017518-Paterson1], a greater increase in CD4 cell count, and lower hospitalization rates [@pone.0017518-LowBeer1], [@pone.0017518-Nachega2]. More recent studies have indicated that moderate adherence levels (70--90%) may be enough to achieve acceptable virologic suppression with antiretroviral regimens containing boosted PIs or NNRTIs. However, at the 95% level we can be quite certain that a very small proportion of individuals on NNRTI are likely to fail virologically [@pone.0017518-Shuter1], [@pone.0017518-Nachega3], [@pone.0017518-Martin1]. The aim of the current study was to look at treatment failure among long-term recipients of ART, i.e. excluded those with \<12 months of ART [@pone.0017518-ElKhatib1]. However, our retrospective assessment of medical charts went back and traced all virologic failures, including the first year on treatment, enabling us to assess the risk of failure from treatment start-up to a maximum follow-up of over 8 years. However, with this design, we missed those who first failed virologically and then dropped out of the programme before reaching 12 months on treatment. Given that a substantial proportion of patients enrolled in ART in similar urban settings are expected to drop out early [@pone.0017518-Unge1], this would lead to an underestimation of the true virologic failure rate among NNRTI recipients in the current assessment. Finally, the lack of a significant association between longer documented treatment interruptions and virologic failure is likely due to incomplete clinical data since we only had access to reported interruptions recorded in the medical charts. In summary, there was a strong association between cumulative reduced adherence to drug-refill visits and both virologic and immunologic failure. One in five failed virologically, the majority within the first two years on ART. On-time ART pick-up appears to be a feasible tool to identify individuals at risk, and if followed by prompt and targeted interventions, it could be used to reduce the rate of virologic failure. We would like to thank Ms. Albertina Dambuza, Belinda Dambuza and Johanna Ledwaba for assisting in reviewing medical records; Ms Ntshebo Mirriam Moeketsi and Alina Malebye for conducting interviews, and Noreen Boikanyo, Rebecca Phofa, Catherine Lephoto, Agnes Ramashiga and Fikile Mbatha for study logistics. We are grateful to all patients for sharing their experiences with us. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was funded by the Swedish International Development Cooperation Agency to National Institute for Communicable Diseases (NICD) and Karolinska Institutet; The Karolinska Institutet (PhD student grant to ZEK and senior research grant to AME); Sven Gard\'s Fund for Virology Research, African Programme for Training in HIV/TB Research Fogarty International Center/NIH 2 U2R TW006878 and the Freeman-Spogli Institute at Stanford University to ZEK. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: ZE-K DK LM MP FL LM AME. Performed the experiments: ZE-K DK FL MP LM. Analyzed the data: ZE-K DK GM MP AME. Contributed reagents/materials/analysis tools: ZE-K FL LM MP DK LM AME. Wrote the paper: ZE-K DK GM LM MP FL LM AME.

PubMed Central

2024-06-05T04:04:19.154788

2011-3-9

PMC3052315

Introduction {#s1} ============ Cells in the developing embryo undergo step-wise progression toward particular fates. Understanding the details of this progression program is dependent upon marking and identifying the emerging cellular populations. In the hematopoietic system, specific cell surface markers for each developmental step have been highly successful at elucidating these stages [@pone.0017536-DeRosa1], [@pone.0017536-Appay1]. The ability to classify other developmental lineages in this rigorous manner would be a significant advance for developmental biology and for regenerative medicine, which greatly depends upon understanding and selecting pure populations of precise cellular types. Human embryonic stem cells (hESCs) can differentiate into cells reflective of early germ layers, including mesoderm, endoderm and ectoderm [@pone.0017536-Murry1]. While the derived cell types express batteries of markers of the *in vivo* situation, the homogeneity of these cells remains unexamined. The ability to separate subpopulations of these particular lineages is critical for developing more targeted methods for specific tissue engineering. In the case of endoderm, for example, the ability to isolate and characterize a FOXA1, FOXA2 and HNF-4α positive population, might allow the more efficient development of cultured hepatocytes [@pone.0017536-Lee1], [@pone.0017536-Li1]. Despite much investigation, comprehensive cell surface markers have been difficult to identify in embryonic lineages, and thus teasing apart the stepwise progression of these lineages using Fluorescence Activated Cell Sorting (FACS) has remained difficult. Although cell surface markers have not been well characterized in these emerging cell types, transcription factors are known to specifically mark cellular lineages [@pone.0017536-Lee1]--[@pone.0017536-Afouda1]. To date using nuclear proteins to examine cellular phenotypes has not been feasible due to limitations in technology [@pone.0017536-Preffer1]. In this report, we present a methodology that uses lineage-specific transcription factors to purify specific cellular populations by multi-channel FACS. This technology, which we term tfFACS, produces intact RNA that can be further examined to deduce the molecular nature of the cells. We applied multichannel tfFACS to examine the cellular populations that emerge upon endoderm differentiation in hESCs. Materials and Methods {#s2} ===================== Cell Culture {#s2a} ------------ Undifferentiated hES cells (H9) were maintained on irradiated mouse embryonic fibroblast (MEF) feeders as previously described [@pone.0017536-Chiao1]. Briefly, the H9 hES cell line was obtained from WiCell Research Institute (Madison, WI, <http://www.wicell.org/>). Cells were cultured in DMEM/F12 medium supplemented with 20% KnockOut serum replacement, 0.1 mM nonessential amino acids (NEAA), 2 mM L-glutamine, 0.55 mM 2-mercaptoethanol (all from Invitrogen, Carlsbad, CA, <http://www.invitrogen.com>) and 8 ng/ml recombinant human FGF2 (Peprotech, Rocky Hill, NJ, <http://www.peprotech.com>). Cultures were passaged with 200 units/ml collagenase IV (Invitrogen) at a 1∶3 split ratio every 4 days. Definitive endoderm differentiation was induced from hESCs by using activin A as previously described [@pone.0017536-DAmour1]--[@pone.0017536-Brunner1]. For differentiation, hESCs were first passaged onto dishes coated with growth factor reduced Matrigel (BD Biosciences, San Diego, CA, <http://www.bdbiosciences.com>) and cultured in hESC media conditioned overnight on primary MEF (CM) for 2 days. Differentiation was carried out in RPMI 1640 (Invitrogen) supplemented with Glutamax, penicillin/streptomycin and varying concentrations of defined FBS (Thermo Scientific HyClone, Logan, UT, <https://www.thermoscientific.com>). Before initiating differentiation, hESCs were given two brief washes in PBS (Invitrogen). In differentiation experiments, FBS concentrations were 0% for the first 24 h, 0.2% for the second 24 h, and 2.0% for subsequent days of differentiation. Recombinant human activin A (R&D Systems Inc., Minneapolis, MN, <http://www.rndsystems.com>) was added to the differentiation medium at 100 ng/ml, and cells were treated for 5 days, with medium changed once at day 3. RT-quantitative PCR {#s2b} ------------------- Total RNA was isolated from triplicate samples using RNeasy Plus Mini Kit (Qiagen, Duesseldorf, Germany, <http://www.qiagen.com>) or the Ambion Recover All nucleic acid extraction kit (optimized for fixed cells) (Applied Biosystems Ambion). The RNA concentration and purity were measured by NanoDrop (Thermo Scientific, Wilmington, DE, <http://www.nanodrop.com>). Only the samples with the OD A260/A280 ratio and the OD A260/A230 ratio close to value of 2.0, which indicates that the RNA is pure, were analyzed. 1 µg RNA was used for reverse transcription with random hexamers in a 20 µl reaction using SuperScript ΙΙΙ First-Strand cDNA synthesis kit (Invitrogen). PCR reactions were run using 1/20 of the cDNA per reaction, and 500 nM forward and reverse primers with iQ SYBR Green Supermix (Bio-Rad, Hercules, CA, <http://www.bio-rad.com>). Real-time PCR was performed using the Bio-Rad iCycler. Cycling was performed as follows: 94°C for 5 min followed by 40 cycles consisting of denaturation (95°C, 30 s), annealing (56°C, 30 s), and extension (72°C, 30 s), with a final incubation at 72°C for 10 min. Relative quantification was calculated using the comparative threshold cycle (CT) method and relative quantified values were normalized against that of housekeeping gene *cyclophilin G* (*CYCG*) [@pone.0017536-DAmour1]. PCR was performed in triplicate for each sample, and 3 independent experiments were carried out. The means and standard derivations were calculated and reported here using data from one representative experiment. Primer sequences are listed in [Table S1](#pone.0017536.s006){ref-type="supplementary-material"}. FACS Cell Fixation and CXCR4 Antibody Staining {#s2c} ---------------------------------------------- Cells were dissociated using 0.05% trypsin-0.53 mM EDTA (Invitrogen) at 37°C for 3 min followed by neutralization in hESCs medium with serum. After washing three times in Staining Buffer \[bovine serum albumin (BSA) or fetal bovine serum (FBS)\] (BD Biosciences, San Diego, CA, <http://www.bdbiosciences.com>), 1.25×10^5^ cells were aliquoted for each antibody staining. Cells were resuspended in 200 µl of the same buffer and first Fc-blocked by treatment with 50 µl human serum supplement (Irvine Scientific, Santa Ana, CA, <http://www.irvinesci.com>) for 15 minutes at room temperature or on ice. Excess blocking serum should not be washed from this reaction. 1.25×10^5^ pelleted cells were fixed in 100 µl of 4% paraformaldehyde (PFA) (BD Biosciences) PBS solution at 4°C for 15 minutes. Cells were washed twice in Staining Buffer (BD Biosciences). The Fc-blocked cells were then labeled with 5 µl of anti-human CXCR4-PE antibody (with direct fluorophore conjugation, R&D Systems Inc.) and incubated for 30 min on ice. Live cells without fixation were stained directly for comparison. As a negative control for analysis, cells in a separate tube were treated in parallel with PE-labeled mouse IgG2A antibody. The results showed comparable staining for fixed and unfixed cells for the cell surface markers we have used in our experiments, including CXCR4 ([Fig. S1A](#pone.0017536.s001){ref-type="supplementary-material"}). This is consistent with a previous study in which when methanol was used to fix cells for CD surface marker staining [@pone.0017536-Szaniszlo1]. GATA4 and SOX17 Direct Fluorophore Antibody Conjugation and Two-Channel FACS Antibody Staining {#s2d} ---------------------------------------------------------------------------------------------- For SOX17 and GATA4, direct fluorophore-conjugated antibodies were not commercially available. Goat anti-human SOX17 and GATA4 (both from R&D systems Inc.) were used, but the common serotype of these primary antibodies meant that secondary fluorescent antibodies would not distinguish between them. We therefore conjugated these primary antibodies directly to fluorophores using the Molecular Probe Zenon® antibody labeling kit as follows: Cells were fixed and blocked as described above. Cells were then permeablized using Cytofix/Cytoperm containing 1% sapanin (BD Biosciences) at room temperature or on ice for 20 minutes. During penetration, label transcription factor antibodies with different fluorescence dyes: goat anti-human GATA4, Goat anti-human SOX17 Abs were conjugated with Alexa 488 and 647 respectively by using Zenon® Goat IgG Labeling Kit from Molecular Probes, according to the manufacturer\'s instructions (Invitrogen). Following conjugation, each labeled antibody was titrated based on the quantitative result of two-step single staining with secondary antibody. For the formal experiment, cells were then incubated on ice for 30 min with both titrated Alexa 488 conjugated anti-human GATA4 and Alexa 647 conjugated anti-human SOX17 antibodies. Each of the Isotype- Goat IgG was also labeled and stained as a negative control. Three-Channel GATA4, SOX17 and CXCR4 FACS Staining {#s2e} -------------------------------------------------- For three-way multichannel FACS with the transcription factor-GATA4, SOX17 and cell surface marker CXCR4, staining was performed as follows: After fixation and blocking, cells were labeled with mouse anti-human CXCR4-PE antibody. Cells were then washed, permeablized, and stained with Alexa 488 conjugated anti-human GATA4 and Alexa 647 conjugated anti-human SOX17 according to the staining protocol indicated as above. As negative controls, PE-conjugated normal mouse IgG (for anti-human CXCR4) and Goat IgGs (for GATA4 and Sox17) were also stained in the same manner as the corresponding antibodies. Compensation samples were prepared by staining fixed hESCs with APC-conjugated mouse anti-human SSEA4 antibody, PE-conjugated mouse anti-human SSEA4 antibody (both from R&D Systems Inc.) and Alexa 488-conjugated mouse anti-human OCT4 antibody (eBioscience, San Diego, CA, <http://www.ebioscience.com>) for each of the 3 channels. The cell surface marker SSEA4 and transcription factor OCT4 were stained the same as for CXCR4 and endodermal transcription factor markers-GATA4 and SOX17, respectively. To exclude nonspecific staining signals from the dead cells, cells were co-stained with LIVE/DEAD® Fixable Dead Cells Stain single-color dye (Molecular Probes, Invitrogen), in parallel with antibody staining. Compared with live cells, dead cells have 50-fold higher intensity with near-IR fluorescent reactive dye. We performed nuclear transcription factor marker staining with fixable dead cell dyes and found that dead cells produced very low signal (\<10%) ([Fig. S1B](#pone.0017536.s001){ref-type="supplementary-material"}). Since the fluorescence signals came mainly from live cells, we concluded that contamination by dead cells was not a concern. Cells were washed twice in Staining Buffer and were analyzed using LSR 1 or LSRII (BD Bioscience) in the Stanford Shared FACS Facility. Data were analyzed using the Flowjo software (Tree Star, Inc., Ashland, Oregon, <http://www.treestar.com>). RNA Quality Optimization {#s2f} ------------------------ Four procedures will affect the intact RNA quality: fixation, staining, sorting and RNA extraction. We harvested the stained cells at different stages to check RNA quality using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, <http://www.home.agilent.com>). Total RNA was isolated using Ambion Recover All nucleic acid extraction kit (optimized for fixed cells) (Applied Biosystems Ambion). Before checking RNA quality, the RNA concentration and purity were measured by NanoDrop described as above in the section of RT-quantitative PCR analysis. When we used the standard FACS protocol and extracted RNA from the sorted cells, the RNA from fixed and stained cells appeared to be of very poor quality, and even before sorting ([Fig. S2A](#pone.0017536.s002){ref-type="supplementary-material"}), consistent with previous reports in the literature [@pone.0017536-Appay1], [@pone.0017536-Krutzik1]--[@pone.0017536-Ravo1]. Since the fixation process may be a cause of the RNA degradation, we varied the fixation duration to see how it affected the RNA. The results showed that fixation was not a primary cause of RNA damage ([Fig. S2B](#pone.0017536.s002){ref-type="supplementary-material"}). Next we investigated the staining process. We stored cells in the regular staining buffer for different durations of time after fixation. As shown in [Figure S2C](#pone.0017536.s002){ref-type="supplementary-material"}, the RNA quality becomes increasingly poor as the storage period increases. This suggested that when cells were dead and penetrated, the exposed RNAs might be gradually degraded by the staining buffer, perhaps due to trace amounts of RNase. Therefore we modified the staining procedure in several ways to eliminate RNase activities: instead of using serum, cells were blocked and stained in staining buffer with BSA (100 µg/ml), RNase Inhibitor (100 U/ml), and DTT (5 mM) added. We also used RNase free water to make stain solution, and maintained very low temperature (on ice or 4°C) throughout the whole procedure. Using our new protocol, we could obtain RNA of high quality. This is demonstrated in [Figure S2D](#pone.0017536.s002){ref-type="supplementary-material"} where clean peaks for 18S and 28S rRNA are still evident after fixation, staining, and sorting. The fixatives which are used for intracellular marker staining, either for flow cytometry or laser capture microdissection studies, include precipitive-type fixatives such as methanol, acetone, ethanol, and cross-linking fixative-neutral-buffered formalin and paraformaldehyde (PFA). According to current studies, to both fix the intracellular proteins and keep the RNA intact, methanol, acetone, and ethanol are preferred over 4% PFA [@pone.0017536-Goldsworthy1]--[@pone.0017536-Medeiros1]. These three fixatives have been successfully used in FACS staining for intracellular phosphorylated signaling proteins [@pone.0017536-Krutzik1], [@pone.0017536-Lamoreaux1]. Conversely, for tfFACS staining, we found that 4% PFA provides higher quality results. FACS {#s2g} ---- When cells are prepared for sorting, two way or three way tfFACS staining was performed following the protocols above using the improved RNA conditions. d5CXCR4^+^ sorting was performed on live cells. Isotype controls were used to gate the cells ([Fig. 1A, B](#pone-0017536-g001){ref-type="fig"} and [Fig. S1C](#pone.0017536.s001){ref-type="supplementary-material"}). Sorting was performed using Aria (BD Bioscience) in the Stanford Shared FACS Facility. Sorting was done at 4°C. Cells were collected into tubes with RNase free PBS. We performed the purity checking of the sorted cells immediately after FACS separation ([Fig. 1C](#pone-0017536-g001){ref-type="fig"}). All the cells either from sorted populations or from the presorted mixtures were centrifuged at 14,000 rpm, 2 min, 4°C to get cell pellets. Total RNA was isolated using Ambion Recover All nucleic acid extraction kit (optimized for fixed cells) (Applied Biosystems Ambion, Austin, TX, <http://www.ambion.com>). ::: {#pone-0017536-g001 .fig} 10.1371/journal.pone.0017536.g001 Figure 1 ::: {.caption} ###### Endodermal subpopulations emerging after activin A treatment using tfFACS. (**A**) A representative experiment using two-channel FACS analysis of GATA4 and SOX17. Compared with the isotype negative control (bottom panels), three distinct cellular populations: SOX17^+^GATA4^−^, SOX17^+^GATA4^+^, and SOX17^−^GATA4^+^ are emerging gradually upon differentiation: at day 1, 13% are SOX17^+^GATA4^+^, increasing to 23% by day 3. Another significant population consists of 18% SOX17^−^GATA4^+^ at day 1 and 25% at day 3. (**B**) After 5 days of differentiation, using three-way multichannel FACS analysis for SOX17, GATA4, and CXCR4, we found that the SOX17^+^GATA4^+^ population dominates the culture (62%) and CXCR4 is expressed in 49% of the cells, most of which are SOX17^+^GATA4^+^CXCR4^+^ (41%). There are also approximately 27% GATA4^+^CXCR4^−^ cells, which comprises the population of SOX17^+^GATA4^+^CXCR4^−^ cells (21%). (**C**) Post sorting, FACS analysis demonstrated that 97% of day 5 SOX17^+^GATA4^+^CXCR4^+^ cells were positive for GATA4, 88% were SOX17 positive, and 95% were CXCR4 positive. This was consistent over 5 separate experiments. (**D**) Expression analysis using RT-qPCR demonstrates that day 5 SOX17^+^GATA4^+^CXCR4^+^ and day3 SOX17^+^GATA4^+^ cells have higher level of expression of SOX17, GATA4 and CXCR4 than unsorted fixed cells or day 3 SOX17^−^GATA4^−^ (d3SOX17negGATA4neg) cells. :::  ::: Microarray Analysis {#s2h} ------------------- Samples collected after 5 days of differentiation included SOX17^+^GATA4^+^ CXCR4^+^ cells, unfixed CXCR4^+^ cells, and unsorted fixed cells. Samples collected after 3 days of differentiation included SOX17^+^GATA4^+^ cells, SOX17^−^GATA^−^ cells, and unsorted fixed cells. As controls we also collected fixed, stained hESCs using the same SOX17 GATA4 CXCR4 three-channel protocol, but without sorting. Unfixed hESCs Exon array data using the same protocol were also analyzed together [@pone.0017536-Chiao1]. All of these samples contained biological replicates, triplicates or quadruplicates. Total RNA was extracted using the Ambion Recover All nucleic acid extraction kit (optimized for fixed cells) (Applied Biosystems Ambion). Probes for the Affymetrix human Exon Array ST 1.0 were prepared and hybridized to the array using the GeneChip Whole Transcript Sense Target Labeling Assay (Affymetrix) according to the manufacturer\'s suggestions [@pone.0017536-Chiao1]. Briefly, for each sample, 1.5 g of total RNA was subjected to ribosomal RNA reduction. Following rRNA reduction, double-stranded cDNA was synthesized with random hexamers tagged with a T7 promoter sequence. The double-stranded cDNA was used as a template for amplification with T7 RNA polymerase to create antisense cRNA. Next, random hexamers were used to reverse transcribe the cRNA to produce single-stranded sense strand DNA. The DNA was fragmented and biotin labeled. The probes of all samples (H9 passages 40--55) were hybridized to the Affymetrix Exon Array ST 1.0 microarrays and scanned. Expression Data Processing {#s2i} -------------------------- We computed gene expression indices for all the samples analyzed using the GeneBASE software [@pone.0017536-Kapur1]. Specifically, correction for background noise was performed for every core probe using the adapted MAT model of background probes in Affymetrix Exon Arrays. The background-corrected intensities were normalized across arrays by core-probe-scaling so that the median intensity of core probes in each sample was equal to 100. The normalized probe intensities were then summarized to gene level expression indices based on the dChip model [@pone.0017536-Li2]. The gene expression indices across arrays were quantile-normalized to generate the final gene expression profiles. The clustering heatmap was generated by dChip using the default setting, i.e, the "1-correlation" distance metric and the centroid linkage method. The raw data files have been deposited in the Gene Expression Omnibus (GEO) database with accession number GSE24135. Results {#s3} ======= tfFACS Allows Isolation of Cells Expressing Combinations of SOX17 and GATA4 {#s3a} --------------------------------------------------------------------------- hESCs can differentiate into endodermal cells by dosing with high levels of the NODAL signaling pathway, but it remains unknown whether this differentiation results in several endodermal cell sub-types or a single homogeneous population. We sought to isolate and characterize these resulting endodermal cells. To this end, we differentiated hESCs into endoderm using activin A in low serum conditions [@pone.0017536-DAmour1]--[@pone.0017536-Brunner1]. Over the five days of differentiation, consistent with the observations of others, we found that markers of mesendoderm, including *BRACHYURY* are transiently expressed at 24 hours, and markers of endoderm, including *SOX17* and *GATA4*, become highly expressed at 3 and 5 days post-differentiation ([Fig. S3](#pone.0017536.s003){ref-type="supplementary-material"}) [@pone.0017536-KanaiAzuma1]--[@pone.0017536-Afouda1], [@pone.0017536-Liu1], [@pone.0017536-McLean1]. The expression of these transcription factors, allowed us to develop multichannel tfFACS using antibodies against SOX17 and GATA4. To this end, hESCs derived endodermal cells 5 days post differentiation were fixed, processed and examined for RNA quality. While multiple conditions were investigated, most of these led to massive RNA degradation, consistent with previous reports ([Fig. S2A](#pone.0017536.s002){ref-type="supplementary-material"}) [@pone.0017536-Goldsworthy1]--[@pone.0017536-Ravo1]. We found that the single most influential factor was not extent of fixation, but the amount of time the sample is stored following fixation (See [Materials and Methods](#s2){ref-type="sec"} for details and [Fig. S2B, C](#pone.0017536.s002){ref-type="supplementary-material"}). Briefly, cells were fixed with 4% paraformaldehyde at 4°C for 15 min, and stained using both anti-human GATA4 and anti-human SOX17 antibodies conjugated with the fluorescence dyes-Alexa 488 and 647, respectively. As negative controls for analysis, normal goat IgG antibody was also conjugated with Alexa 488 and 647. Stained cells were then analyzed using two-channel FACS. We found three distinct cellular populations in hESC derived endoderm after 5 days of differentiation: SOX17^+^GATA4^−^, SOX17^+^GATA4^+^ and SOX17^−^GATA4^+^ ([Fig. 1A, B](#pone-0017536-g001){ref-type="fig"} and [2](#pone-0017536-g002){ref-type="fig"}). This observation demonstrates that treatment with activin A causes hESCs to differentiate into molecularly distinct subpopulations of endoderm. ::: {#pone-0017536-g002 .fig} 10.1371/journal.pone.0017536.g002 Figure 2 ::: {.caption} ###### Venn diagram cartoon summarizing data obtained from 4 independent experiments which were averaged. The color key is represented on the lower right. :::  ::: tfFACS Can Be Used with Combinations of Antibodies Against Transcription Factors and Cell Surface Proteins {#s3b} ---------------------------------------------------------------------------------------------------------- To further investigate the extent of heterogeneity in the endodermal culture, we followed the subpopulations through the differentiation time course by adding an additional marker, CXCR4 [@pone.0017536-McGrath1]. We chose CXCR4 as the third marker because it is one of the few cell surface markers used to isolate definitive endoderm from mouse and human ESCs [@pone.0017536-DAmour1], [@pone.0017536-Yasunaga1]. We examined hESC-derived endoderm after a 1, 3 or 5 days of differentiation using three-way multichannel FACS analysis for SOX17, GATA4 and CXCR4, or two-way multichannel FACS analysis for SOX17 and GATA4. FACS analysis immediately following sorting to check the purity showed that 95% of the day 5 SOX17^+^GATA4^+^CXCR4^+^ cells were positive for GATA4, 90% were positive for SOX17 and more than 95% were positive for CXCR4, suggesting efficacy of the sorting protocol ([Fig. 1C](#pone-0017536-g001){ref-type="fig"}). To further validate the sorted populations, we performed marker analysis using RT-qPCR for *GATA4*, *SOX17* and *CXCR4*. Compared to day 3 and day 5 fixed cells, which are highly heterogeneous mixtures of differentiating cells, the day 5 SOX17^+^GATA4^+^CXCR4^+^, and the day 3 SOX17^+^GATA4^+^ express these transcripts at a much higher level, consistent with an increase of purity ([Fig. 1D](#pone-0017536-g001){ref-type="fig"}). Overall, we found that, during the first 24 hours of differentiation, GATA4^+^ cells increase substantially, and approximately 13% of these are also SOX17^+^. However, by day 3, the double SOX17^+^GATA4^+^ population becomes the predominant marked population ([Fig. 1A](#pone-0017536-g001){ref-type="fig"}, [2](#pone-0017536-g002){ref-type="fig"}) and dominates the culture by day 5 (\>50%) ([Fig. 1B](#pone-0017536-g001){ref-type="fig"}, [2](#pone-0017536-g002){ref-type="fig"}). SOX17^+^GATA4^−^ cells are rare throughout the timecourse, strongly suggesting that if a cell is SOX17^+^, GATA4^+^ will also be present. By day 5, CXCR4 is expressed in approximately 43% of the cells. Interestingly, this population does not entirely overlap with that of SOX17^+^GATA4^+^ ([Fig. 1B](#pone-0017536-g001){ref-type="fig"}, [2](#pone-0017536-g002){ref-type="fig"}), suggesting that the diversity of cells after treatment with activin A is greater than previously thought. This indicates that experiments using CXCR4 to isolate definitive endoderm may have missed the SOX17^+^GATA4^+^CXCR4^−^ cells, which comprise about 17% of the total population. tfFACS Does Not Substantially Alter Gene Expression {#s3c} --------------------------------------------------- In order to further elucidate the molecular nature of these endodermal populations, we first needed to show that tfFACS does not alter gene expression due to the fixation protocol. Initially, we examined both hESCs and derived endoderm, either fixed or unfixed for the expression of lineage specific markers. No difference in expression levels of *OCT4* (hESCs) or *SOX17*, *GATA4*, *or CXCR4* (derived endoderm) were observed between fixed and unfixed cells ([Fig. S2E](#pone.0017536.s002){ref-type="supplementary-material"}). We next measured global gene expression using microarray technology on cells sorted using tfFACS. Samples collected after 5 days of differentiation included SOX17^+^GATA4^+^ CXCR4^+^ cells, unfixed CXCR4^+^ cells, and unsorted fixed cells. Samples collected after 3 days of differentiation included SOX17^+^GATA4^+^ cells, SOX17^−^GATA^−^ cells, and unsorted fixed cells. As controls, we analyzed both unfixed hESCs and fixed hESCs [@pone.0017536-Chiao1]. All samples contained biological duplicates, triplicates or quadriplicates ([Fig. S4](#pone.0017536.s004){ref-type="supplementary-material"}). We then performed hierarchical clustering to demonstrate whether cellular fixation alone could change gene expression. We based this analysis on 1647 transcript clusters with coefficient of variation \>0.5 across the samples and expression values \> = 500 in at least 2 out of the 21 samples. We found that the degree of distortion due to fixation is small particularly when compared between samples and stages. Two illustrations of this are that, first, fixed and unfixed cells cluster together based upon differentiation stage, not based upon degree of fixation, second, even though hESC and d5CXCR4^+^ are unfixed, unstained samples, they do not cluster together. Instead, each is clustered with the fixed samples that are biologically similar: hESCs with fixed hESC cells, and d5 CXCR4^+^ cells with fixed day 5 samples ([Fig. S4](#pone.0017536.s004){ref-type="supplementary-material"}). SOX17^+^GATA4^+^CXCR4^+^ is enriched for Definitive Endodermal Transcripts {#s3d} -------------------------------------------------------------------------- Our tfFACS analysis, showing discrete subpopulations with defined markers, strongly suggests that hESC derived endoderm comprises cells already specified toward particular endodermal fates. Since tissue engineering of endodermal organ systems is still in its infancy, our aim was to determine the endodermal character of each isolated population and then examine whether these subpopulations represented more specialized endodermal tissue types. To this end, we first sought to determine whether the subpopulations could be classified as definitive endoderm. Because a reliable set of human definitive endodermal marker genes has not been established, we compiled "gold-standard" definitive endoderm gene sets: one from the Mouse Genome Informatics (MGI Set) database based on RNA in situ hybridization or immunohistochemistry evidence in E7.0--8.0 mouse (<http://www.informatics.jax.org>; 22 genes) and another from Sherwood et al., (Melton Set) based upon microarray profiling of E8.25 mouse definitive endoderm (51 genes, see [Table 1](#pone-0017536-t001){ref-type="table"}) [@pone.0017536-Sherwood1]. To determine whether these 'gold-standard genes' are present in the subpopulations at a level significantly higher than reference, we employed the GSEA algorithm [@pone.0017536-Subramanian1]. We first compared the SOX17^+^GATA4^+^CXCR4^+^ isolated from day 5 with all the other samples, with the exception of SOX17^+^GATA4^+^ cells from day 3 and CXCR4^+^ cells from day 5, which would have extensive overlap. *As* shown in [Figure 3A--C](#pone-0017536-g003){ref-type="fig"}, the MGI gene set is highly enriched in the SOX17^+^GATA4^+^CXCR4^+^ day 5 sorted cells in multiple comparisons (d5 SOX17^+^GATA4^+^CXCR4^+^ vs hESC: P\<0.0002; d5 SOX17^+^GATA4^+^CXCR4^+^ vs Unsort1: P = 0.0304; and d5 SOX17^+^GATA4^+^CXCR4^+^ vs Unsort2: P = 0.0013. The Unsort1 represents d3Fix+d3 SOX17^−^GATA^−^+d5Fix.1+d5Fix.3, and Unsort2 represents d5Fix.2+d5Fix.4). We repeated the GSEA analysis on the Melton gene set. Again, this gene set is enriched in d5 SOX17^+^GATA4^+^CXCR4^+^ in all comparisons ([Fig. 3D--F](#pone-0017536-g003){ref-type="fig"}). ::: {#pone-0017536-g003 .fig} 10.1371/journal.pone.0017536.g003 Figure 3 ::: {.caption} ###### GSEA analysis of the definitive endoderm (DE) gene sets for the day 5 SOX17^+^GATA4^+^CXCR4^+^ group. As shown in (**A--C**), the MGI gene set is highly enriched in the d5 SOX17^+^GATA4^+^CXCR4^+^ cells in multiple comparisons. d5SOX17^+^GATA4^+^CXCR4^+^ vs. hESC (unfixed hESCs+fixed hESCs): P\<0.0002 (**A**); d5SOX17^+^GATA4^+^CXCR4^+^ vs. Unsort1 (d3Fix+d3 SOX17^−^GATA^−^+d5Fix.1+d5Fix.3): P = 0.0304 (**B**); and d5 SOX17^+^GATA4^+^CXCR4^+^ vs Unsort2 (d5Fix.2+d5Fix.4): P = 0.0013 (**C**). We repeated the GSEA analysis on the Melton gene set. This gene set is enriched in d5SOX17^+^GATA4^+^CXCR4^+^ cells in all comparisons (**D--F**). :::  ::: ::: {#pone-0017536-t001 .table-wrap} 10.1371/journal.pone.0017536.t001 Table 1 ::: {.caption} ###### Definitive endoderm (DE) gene sets used in the analyses. ::: {#pone-0017536-t001-1} Gene set Number of genes Gene Name ---------- ----------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ MGI 22 CER1, GALNAC4S-6ST, CLDN4, CPM, DKK1, EDA, EFNA1, EMB, FOXA2, HHEX, HNF1B, ITGA3, JARID1B, LAMA1, PRDM1, SDC1, SHH, TMEM46, SOX17, TES, TMPRSS2, TRH Melton 51 SOX17, FOXC1, GATA3, PAX6, FOXA1, EVX1, IRX3, ZHX2, PAX1, DLX5, HOXB9, RIPK4, SP6, ISL1, IRX5, SOX21, DMRTA1, PAX8, SIX3, HOXD9, PAX9, MEOX1, HOXC4, HOXA9, FOXC2, HOXB2, T, HOXB3, PAX3, PKNOX2, DLX3, DLX2, SIX1, TPBG, HOXC8, HOXD8, RFX3, CDX4, HOXA3, SOX9, HOXB1, ARNT2, HOXD1, HOXA1, FOXG1, GLI3, SOX11, IRX2, HEY2, SSBP2, PBX1 ::: We then asked whether the SOX17^+^GATA4^+^CXCR4^+^ day 5 cells and day 3 SOX17^+^GATA4^+^ were more enriched for 'gold-standard' endodermal genes than the CXCR4^+^ day 5 population, which has generally been used to isolate hESC-derived endoderm [@pone.0017536-DAmour1]. To this end, we performed GSEA analysis to compare SOX17^+^GATA4^+^CXCR4^+^, SOX17^+^GATA4^+^, and CXCR4^+^ to the control group, which were all other samples combined. While both the MGI set and Melton set are enriched in both SOX17^+^GATA4^+^CXCR4^+^ and CXCR4^+^, we observed higher enrichment levels in the SOX17^+^GATA4^+^CXCR4^+^ in both comparisons (MGI: P\<0.0002 and P = 0.0038, respectively; Melton: P = 0.0057 and P = 0.0105, respectively) ([Fig. S5](#pone.0017536.s005){ref-type="supplementary-material"}). Furthermore, day 3 SOX17^+^GATA4^+^ cells have a similar enrichment in the MGI set as day 5 SOX17^+^GATA4^+^CXCR4^+^ (P\<0.0002). Importantly, the above analyses suggest that triple selection using SOX17^+^GATA4^+^CXCR4^+^ and double selection of SOX17^+^GATA4^+^ produce a more homogenous population of DE cells than selection using CXCR4 alone, which may deduce that protocols using a single FACS channel with CXCR4 are mixed with other lineages, or missing a valuable population of definitive endodermal cells. Additionally, we showed that day 3 SOX17^+^GATA4^+^ cells can be obtained at a much earlier stage of differentiation, prior to expression of CXCR4^+^ cells, providing an important new tool to isolate this earlier definitive endoderm subtype. Isolated Populations Are Associated With Biological Processes {#s3e} ------------------------------------------------------------- To determine whether these endodermal subpopulations were indeed already fated toward specific endodermal fates, we sought to identify functional signatures using GO [@pone.0017536-Dennis1]. To this end, we selected high value representative genes from each sorted cellular population (upregulated with fold change \>3 and expression values difference \>100 compared to control samples). With these criteria, we selected 331 genes from the SOX17^+^GATA4^+^CXCR4^+^ day 5 cells, 442 from the CXCR4^+^ day 5 cells and 197 from the SOX17^+^GATA4^+^ day 3 cells. DAVID analysis on these groups yielded similar annotations consisting of significant biological process terms including terms "pattern specification process", and "gastrulation" ([Table S2](#pone.0017536.s007){ref-type="supplementary-material"}, [S3](#pone.0017536.s008){ref-type="supplementary-material"}, [S4](#pone.0017536.s009){ref-type="supplementary-material"}, [S5](#pone.0017536.s010){ref-type="supplementary-material"}, [S6](#pone.0017536.s011){ref-type="supplementary-material"}, [S7](#pone.0017536.s012){ref-type="supplementary-material"}). As these annotations are shared between the sorted populations, we asked whether they arose from overlap between the sets. Comparing the gene lists from the SOX17^+^GATA4^+^CXCR4^+^ and CXCR4^+^ day 5 cells, we found that 197 genes are shared, demonstrating that overlap between these populations is extensive. DAVID analysis of these shared 197 genes is again significantly enriched in biological processes such as pattern specification process and cell morphogenesis ([Table S8](#pone.0017536.s013){ref-type="supplementary-material"}, [S9](#pone.0017536.s014){ref-type="supplementary-material"}). Unexpectedly, the genes unique to d5 CXCR4^+^ (241 genes) annotate as being significant for blood vessel morphogenesis and nervous system development ([Table S10](#pone.0017536.s015){ref-type="supplementary-material"}, [S11](#pone.0017536.s016){ref-type="supplementary-material"}) whereas those unique to SOX17^+^GATA4^+^CXCR4^+^ (129 genes) annotate as being significant only for cell adhesion ([Table S12](#pone.0017536.s017){ref-type="supplementary-material"}, [S13](#pone.0017536.s018){ref-type="supplementary-material"}). This data suggests that while there overlap as may be expected due to the use of CXCR4 in each sort, there are also distinct differences between these populations. Discussion {#s4} ========== While the transcriptome of whole organisms, organ systems and culture regimes, have been described, the extent of the molecular similarities of cells within these complex groups is far from understood. This distinction is critical, as differentiating cellular populations must contain rapidly diversifying cellular types. Distinguishing between these subtle varieties of cell types is central toward a more complex biological investigation of single cell differences within these larger systems. For example, based upon transcriptional profiling it is clear that Human embryonic stem cells can differentiate into definitive endodermal cells, but based upon what we understand from the embryo these cells are unlikely to be a purely homogeneous population [@pone.0017536-Murry1], [@pone.0017536-DAmour1], [@pone.0017536-Liu1], [@pone.0017536-McLean1]. For regenerative medicine and for a developmental understanding, it is important that these subtypes be isolated and characterized further. Definitive Endoderm cells show a remarkable versatility in serving as the precursor to a multitude of cell types that constitute the visceral organs [@pone.0017536-Murry1], [@pone.0017536-KanaiAzuma1], [@pone.0017536-Sherwood1]. Using the technology described in this report, transcription factors can now be used to define populations emerging from human Embryonic Stem Cells, filling an urgent need to classify intermediate steps of differentiation. While tfFACs represents a new methodology to isolate and characterize similar cellular types from a complex mixture, it does not allow continued growth of sorted cells and thus their lineage specific commitments cannot be readily assessed. Regardless, this new method does provide a means to examine new subtypes genomically, opening up the potential for discovery of new cell surface markers and for elucidating previously uncharacterized cellular populations. As the approach has the potential to scale up to 11 channels, it could prove an unparalleled means to define cellular populations [@pone.0017536-DeRosa1]. Using this approach, we find that definitive endoderm derived from hESCs is not a homogeneous population of cells, but rather diverse. We find cells within the differentiating cellular population express SOX17, GATA4 and CXCR4 together or in all possible combinations, suggesting that differing lineage potentials exist within the culture of endoderm. Overall, this represents an advance in FACS technology that can be used to evaluate specific subpopulations and avoids the *a priori* need for lineage-specific cell surface markers, an unfulfilled need that has limited our understanding of lineage differentiation from embryonic stem cells as well as in a multitude of other disciplines, including cancer biology. The use of tfFACS to characterize lineage commitment in a systematic step wise fashion will provide inroads into understanding the molecular nature of *in vitro* derived cellular populations. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Methods to identify whether fixation will affect cell surface marker staining, whether to exclude nonspecific dead cell signals from fixed cells, and how the cell sorting was performed.** (**A**) To test if fixation distorts cell surface marker staining, live and 4% paraformaldehyde (PFA) fixed (4°C, 15 min) day 5 differentiating cells were stained with PE-conjugated anti-human CXCR4 antibody, based on its negative isotype control mouse IgG (blue histogram), comparable CXCR4 staining result was detected (red histogram). Day 5 CXCR4^+^ sorting was performed on live cells. (**B**) To exclude nonspecific fluorescence from dead cell, we performed nuclear TF SOX17 staining with fixable dead cell dyes. By comparison to the isotype negative control GtIgG (bottom panel), we found that dead cells produced very low signal when sorted for Sox17 (5.47%, upper right quadrant), while the vast majority of Sox17 positive signals are from live cells (53%, lower right quadrant). (**C**) According to isotype controls, day 5 CXCR4^+^ (orange), CXCR4^−^ (purple), and SOX17^+^GATA4^+^ (box in bottom panel) cells were gated. Based on CXCR4^+^ and CXCR4^−^ subsets, day 5 SOX17^+^GATA4^+^CXCR4^+^ (blue) and SOX17^+^GATA4^+^CXCR4^−^ (green) populations were selected respectively. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **The tfFACS method used produce intact RNA following fixation, nuclear staining and FACS sorting.** (**A**) When we used the standard FACS protocol, extracted and amplified RNA from the sorted cells, the RNA from fixed and stained cells appeared to be of very poor quality measured by Agilent bioanalyzer, compared with unfixed and unstained cells. (**B**) When we varied the fixation duration from 5 min to 10 min or 15 min, we found that fixation was not a primary cause of RNA damage. Relatively intact RNA can be obtained from cells fixed by 4% paraformaldehyde at 4°C for 15 min at a level similar to that of cells fixed for 5 min and 10 min. (**C**) We stored the cells in the regular staining buffer for different amount of time after fixation. The RNA quality becomes increasingly poor as the storing period increases from 24 hours to 4 months at 4°C. (**D**) After modifying the staining procedure in several ways, we could obtain intact RNA which has clean peaks for 18S and 28S rRNA after fixation, staining and sorting. (**E**) Fixed and unfixed samples were examined by RT-qPCR analysis to determine expression levels of *OCT4* (hESCs) and *SOX17*, *GATA4*, *and CXCR4* (day 5 endoderm). (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Molecular examination of endodermal differentiation from hESCs over the course of 5 days.** RT-qPCR analysis showed that markers of endoderm, including *SOX17*, *GATA4*, and *CXCR4* become highly expressed at day 3 and day 5 post-differentiation, while *BRACHYURY (BRACH)*, a mesendodermal marker, is expressed transiently at day 1. hESCs have very low expression of endodermal genes. The cells are not expressed *SOX1*, a neuroectoderm marker throughout the timecourse. X-axis indicates days of endodermal differentiation by activin A; numbers on the Y-axis indicate relative gene expression level, normalized to that of *cyclophilin*G (*CYCG*). qPCR was performed using triplicates for each sample, and 3 independent experiments were carried out. Error bars indicate standard derivations which were calculated and reported here using data from one representative experiment. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Hierarchical cluster shows that fixatives do not substantially change expression of cell types.** We performed hierarchical clustering and found that fixed and unfixed cells cluster together based upon cellular character, and not due to methodology. For example, hESC and d5CXCR4^+^, which have not been processed, do not cluster together, but clustered with the fixed samples that are biologically similar: hESCs with fixed hESC cells, and d5 CXCR4^+^ cells with fixed day 5 samples. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Comparing the definitive endoderm (DE) gene set expression in SOX17^+^GATA4^+^CXCR4^+^ day 5 cells, SOX17^+^GATA4^+^ day 3 cells and day 5 CXCR4^+^ cells using GSEA analysis.** We performed GSEA analysis to compare these three populations to the control group, which are all the combined rest samples. While both the MGI DE set and Melton DE set were enriched in both d5 SOX17^+^GATA4^+^CXCR4^+^ and d5 CXCR4^+^ cells, we observed higher enrichment levels in the d5SOX17^+^GATA4^+^CXCR4^+^ population in both comparisons. MGI: P\<0.0002 (**A**) and P = 0.0038 (**C**); Melton: P = 0.0057 (**D**) and P = 0.0105 (**F**). Interestingly, d3SOX17^+^GATA4^+^ cells have similar DE gene sets enrichment to d5 SOX17^+^GATA4^+^CXCR4^+^ cells (**B, E**). (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Primers used for RT-qPCR analysis.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Enrichment of top gene categories in the d5 SOX17^+^GATA4^+^CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Genes in each enriched category from d5 SOX17^+^GATA4^+^CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### **Enrichment of top gene categories in the d5 CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### **Genes in each enriched category from d5 CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S6 ::: {.caption} ###### **Enrichment of top gene categories in the d3 SOX17^+^GATA4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S7 ::: {.caption} ###### **Genes in each enriched category from the d3 SOX17^+^GATA4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S8 ::: {.caption} ###### **Enrichment of top gene categories in the overlapping 197 genes from the d5 SOX17^+^GATA4^+^CXCR4^+^ and d5 CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S9 ::: {.caption} ###### **Genes in each enriched category with overlapping 197 genes from the d5 SOX17^+^GATA4^+^CXCR4^+^ and d5 CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S10 ::: {.caption} ###### **Enrichment of top gene categories in the unique 241 genes from the d5 CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S11 ::: {.caption} ###### **Genes in each enriched category with the unique 241 genes from d5 CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S12 ::: {.caption} ###### **Enrichment of top gene categories in the unique 129 genes from the d5 SOX17^+^GATA4^+^CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S13 ::: {.caption} ###### **Genes in each enriched category with the unique 129 genes from the d5 SOX17^+^GATA4^+^CXCR4^+^ cells.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We wish to thank Dr. Matthew P. Scott for critical reading of the manuscript. We would like to thank the members of the Baker, Wong and Scott laboratories for valuable discussions and comments throughout the course of this work, especially Si Wan Kim for the assistance with the experiments, Se-Jin Yoon for the helpful discussions, Andrea Elizabeth Wills for the generous help with the manuscript reading. We also thank Jonathan Van Dyke, Catherine Carswell-Crumpton, and Elizabeth T. Zuo for their technical support. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the California Institute of Regenerative Medicine (CIRM) Grants RC1-00100 (J.C.B., W.H.W.) and RC1-00133 (W.H.W.), National Institutes of Health grant 1U01HL100397 (W.H.W.). Z.O. was supported by a CIRM predoctoral fellowship. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: YP WHW JCB. Performed the experiments: YP. Analyzed the data: YP ZO WHW JCB. Contributed reagents/materials/analysis tools: YP ZO WHW JCB. Wrote the paper: YP ZO WHW JCB.

PubMed Central

2024-06-05T04:04:19.157947

2011-3-9

PMC3052316

Introduction {#s1} ============ While studies are few, genetic diversity is believed to facilitate successful biological invasions [@pone.0017524-Lockwood1]. In animals, for example higher levels of genetic diversity increased population persistence and colonization success [@pone.0017524-Ahlroth1] and may contribute to increased range expansion [@pone.0017524-Porter1], [@pone.0017524-Krieger1]. In invasive plants however, genetic diversity is high in some species, but many successful invaders have little or no genetic diversity (for a review see [@pone.0017524-Ward1]). Aphids are valuable study systems for investigating the roles of genetic variation and phenotypic plasticity on population, ecological, and evolutionary dynamics [@pone.0017524-Brisson1], [@pone.0017524-Powell1], [@pone.0017524-Loxdale1]. The Oleander aphid, *Aphis nerii* Boyer de Fonscolombe, is a pest of several plant families including Apocynaceae (*Nerium* and *Vinca*), Asclepiadaceae (*Asclepias*, *Calotropi*, and *Gomphocarpus*), Asteraceae, Convolvulaceae, Euphorbiaceae, and Rutaceae [@pone.0017524-Blackman1]. This aphid, along with its principal host plant (Oleander), is thought to be Mediterranean in origin. *Aphis nerii* has since become a common invasive species in warm temperate and tropical regions of the world [@pone.0017524-Blackman1]. In the United States, the Oleander aphid commonly infests two plant families, Apocynaceae and Asclepiadaceae [@pone.0017524-Stoetzel1]. Oleander, *Nerium oleander* (Apocynaceae), is a common ornamental plant in southern and coastal states and frequently grows along US highways [@pone.0017524-Hall1]. Native and ornamental Milkweed (Asclepiadaceae) is the other common US host. Milkweed distribution overlaps with that of Oleander, but extends into northern and central states [@pone.0017524-Groeters1], [@pone.0017524-Groeters2]. Although these two host plant types are ecologically different, both are patchily distributed, contain cardiac glycosides which *A. nerii* sequesters for defense, and are unable to support aphid populations throughout the year [@pone.0017524-Blackman1], [@pone.0017524-Hall1]. Aphid fitness tradeoffs among host plant species results in selection for host fidelity which can inhibit gene flow and result in the development of host races; i.e., host-associated population genetic structure [@pone.0017524-Powell1], [@pone.0017524-Via1]. Increased use of molecular markers to study aphid populations has revealed that host races are common [@pone.0017524-VanlerbergheMasutti1], [@pone.0017524-RuizMontoya1], [@pone.0017524-Miller1]. There is currently no evidence of differential fitness for *A. nerii* inhabiting species of milkweed that differed in the amount cardiac glycosides that they possess [@pone.0017524-Groeters2]. Fitness tradeoffs in *A. nerii* among different host plant families such as Milkweed and Oleander, to our knowledge, have not been examined. Aphid species vary in their mode of reproduction from obligate to cyclical parthenogens, but less than 3% of species are strictly clonal [@pone.0017524-Simon1]. *Aphis nerii* is believed to be an obligate parthenogen; males have never been found in natural populations [@pone.0017524-Blackman1]. Males, and sexual reproduction, have been induced in laboratory lines under short-day conditions [@pone.0017524-Takada1], [@pone.0017524-Takada2]. These laboratory-induced sexuals, however, had low fecundity and the extent to which sexual reproduction occurs in nature is unclear [@pone.0017524-Takada1], [@pone.0017524-Takada2]. Sexual reproduction and recombination increases variation and sets the stage for selection and adaptation [@pone.0017524-Simon1], [@pone.0017524-Simon2]. Conversely, asexual reproduction may limit genetic variation and adaptive potential, but provides reproductive assurance in stable environments and during colonization events [@pone.0017524-Simon1], [@pone.0017524-Simon2], [@pone.0017524-Halkett1] (see also [@pone.0017524-Lushai1]). The genetic structure of aphid populations is shaped spatially and temporally by habitat distribution, dispersal capabilities, and life-cycle. Many aphids, including *A. nerii*, produce winged forms in response to overcrowding and/or decreasing host plant quality [@pone.0017524-Groeters3], [@pone.0017524-Dedryver1]. Dispersal range is unknown for most species, but reports of less than one to hundreds of kilometers when habitat is continuous, are not uncommon [@pone.0017524-Loxdale2]. *Aphis nerii*, however, is found in patchily distributed habitats throughout the United States, and both Oleander and Milkweed are unable to support aphid populations year round except in the southernmost latitudes of the US [@pone.0017524-Hall1]. Consequently, most populations are characterized by frequent extinction events followed by re-colonization from unknown source population(s). Spatial and temporal genetic variation is driven by the magnitude of population bottlenecks, the number of founding individuals during re-colonization, and the genetic variation of the source population(s). The aim of this study was to characterize genetic variation within and among southern US populations of *A. nerii* with the goal of gaining insight into the population dynamics, life history, and ecology of this well-established invasive species. In so doing, we asked the following questions: 1) Is there any evidence of sexual reproduction in US populations of *A. nerii*?; 2) Do aphids inhabiting different hosts comprise host races?; 3) Does the patchy distribution of suitable habitat result in population genetic subdivision over small and/or large geographic ranges?; and 4) Is there temporal variation in population genetic structure? To address these questions, we used microsatellite markers to assess genetic diversity over a two year period within and among three populations separated by roughly 100 km and 3,700 km on two host plant species. Here, we report no evidence of sexual reproduction or the existence of host races in *A. nerii*. Whilst genetic variation was extremely low within and among populations, the genetic composition of one population was found to change drastically over time. Our findings suggest that *A. nerii* is an efficient colonizer that demonstrates true metapopulation dynamics. Methods {#s2} ======= Population Sampling {#s2a} ------------------- Aphids were collected from Oleander (*N. oleander*) at two locations in Georgia and one location in central California, June-August 2008, 2009. This "overwintering" period between samples allowed us to assess temporal genetic variation, and determine if sexual reproduction occurred. Sampling locations were: Statesboro, Georgia (SGO; 32°24′N, 81°46′W); Tybee Island, Georgia (TIGO; 31°59′N, 80°50′W); and Concord, California (CCO; 37°57′N, 121°56′W). Sampling areas consisted of multiple patches of Oleander plants in close proximity but intermittently up to 2.5 km apart. To survey genetic diversity within each population, aphids were collected from different parts of the same plant and from as many different plants as possible. To assess genetic variation among aphids inhabiting different host plant species, aphids were collected from Milkweed (*Asclepias amplexicaulis*) at a second site in Statesboro, Georgia (SGM; 32°25′N, 81°47′W) in June 2008. Microsatellite analyses {#s2b} ----------------------- Individual aphids were genotyped at five microsatellite loci (Ago24, Ago66, Ago69, Ago89, and Ago126) using primers originally designed for the Cotton/Melon aphid, *Aphis gossypii* [@pone.0017524-VanlerbergheMasutti2]. These loci have been used previously to study at least three different *Aphis* species [@pone.0017524-VanlerbergheMasutti2], [@pone.0017524-Michel1]. No linkage disequilibrium among loci has been detected in any species suggesting that they are unlinked in *A. nerii* as well [@pone.0017524-VanlerbergheMasutti2], [@pone.0017524-Michel1]. These five loci were surveyed in 50 individuals from each population for each year except for SGM where 19 individuals were genotyped (n = 319 total aphids). DNA was extracted by macerating individual aphids in 70 µl cell-lysis/proteinase-K buffer (10 mM of Tris, 50 mM KCl, 0.5 tween, 0.2 mg/ml proteinase-K, pH. 8.0) followed by incubation for 1 h at 65°C and 15 min at 99°C (Lee and Frost 2002). PCR reactions were carried out in 20 µl final volumes with 2.5 mM MgCl~2~. Reactions consisted of a 4 min denaturation at 94°C, followed by 35 cycles of 30 s at 94°C, 35 s at 58°C, and 45 s at 72°C, and a final 10 min extension step at 72°C. PCR products were run on small (15 cm×17 cm×0.8 mm), non-denaturing TAE (tris-acetate-EDTA, pH 8.0) buffered polyacrylamide gels of either 9% or 10% concentration, with the lower portion of the gel supplemented with EnhanceIT polymer (Elchrom Scientific, Switzerland). Gels were run for 2 h at 30 mA, stained with ethidium bromide, and visualized on a UV light box. Allele sizes at each locus were estimated using the M3 size standard (Elchrom Scientific, Switzerland). Observed and expected heterozygosities were calculated for each locus in each population. Deviations from Hardy-Weinberg expectations were tested according to Guo and Thompson [@pone.0017524-Guo1] and estimates of the inbreeding coefficient *F* ~IS~ [@pone.0017524-Weir1] were estimated using FSTAT v. 2.9.3 [@pone.0017524-Goudet1]. Molecular subdivision among populations and over time were calculated by estimating both *F* ~ST~ and *R* ~ST~ using the program FSTAT v. 2.9.3 [@pone.0017524-Goudet1]. The significance of *F*-statistics was tested using the randomization procedure available in FSTAT using 5000 permutations. Estimates of genetic distance between multilocus genotypes (MLGs, see below) (δμ^2^ and D~SW~) were calculated according to Shriver et al. [@pone.0017524-Shriver1]. A genotypic diversity index was calculated as the ratio of the number of distinct genotypes out of the total number of samples (G/N ratio). We also compared the ratio of observed multilocus genotypic diversity (G~O~) to that expected under conditions of sexual reproduction (G~E~), as described by Stoddart and Taylor [@pone.0017524-Stoddart1]. The presence of null alleles in the genotype data was estimated using Microchecker [@pone.0017524-VanOosterhout1]. Results {#s3} ======= All five microsatellite loci used in this study were polymorphic in *A. nerii*. The number of alleles was low; 3 to 4 alleles per locus. Three of the five loci each had an observed heterozygosity (H~O~) of 0.143, the two loci an H~O~ of 1. Null alleles were not detected for any loci in any populations sampled. Extremely low levels of genotypic variation were observed in *Aphis nerii*. Only 2 multi-locus genotypes (MLGs) or "clones" were detected among the 319 individuals assayed across all populations and both years ([Table 1](#pone-0017524-t001){ref-type="table"}). MLG 1 was dominant both spatially and temporally, comprising 84.3% of the samples. The remaining 15.7% of the samples consisted of MLG 2. MLG 2 was found in one population (TIGO) in 2009 only. The MLG\'s differed in levels of heterozygosity, with 2 of 5 loci (40%) heterozygous in MLG 1 and 5 of 5 (100%) in MLG 2. ::: {#pone-0017524-t001 .table-wrap} 10.1371/journal.pone.0017524.t001 Table 1 ::: {.caption} ###### Multi-locus genotypes among genotypes and across years. ::: {#pone-0017524-t001-1} Year 2008 2009 --------------------- ------ ----------------------------------------- --------- --------- --------- --------- --------- --------- Ago24 140/148[2](#nt102){ref-type="table-fn"} 140/148 140/148 140/148 140/148 134/138 140/148 Ago66 160/160 160/160 160/160 160/160 160/160 156/166 160/160 Ago69 100/100 100/100 100/100 100/100 100/100 90/94 100/100 Ago89 171/171 171/171 171/171 171/171 171/171 155/161 171/171 Ago126 169/175 169/175 169/175 169/175 169/175 171/179 169/175 "Clone" designation Clone 1 Clone 1 Clone 1 Clone 1 Clone 1 Clone 2 Clone 1 1 SGO = Statesboro, Georgia - Oleander; SGM = Statesboro, Georgia - Milkweed; TIGO = Tybee Island, Georgia - Oleander; CCO = Concord, California -- Oleander. 2 Numbers indicate estimated allele sizes for each locus. ::: Despite the low level of genotypic diversity observed, the genotypes of the two MLGs were divergent. That is, the two MLGs did not share any alleles, clearly indicating that they were not the product of recombination through sexual reproduction. Estimates of genetic distance, based on the stepwise mutation model, were large (δμ^2^ = 15.35; D~SW~ = 15.35) suggesting that the MLGs do not share a close genealogical relationship [@pone.0017524-Calabrese1]. No evidence of sexual or mitotic recombination was found. First, no homozygous allelic arrangements were found at any of the heterozygous loci ([Table 1](#pone-0017524-t001){ref-type="table"}). Second, levels of heterozygosity were high, ranging from 0.40 to 1.00 ([Table 2](#pone-0017524-t002){ref-type="table"}), and there were significant deviations from Hardy-Weinberg expectations at polymorphic loci within all populations due to heterozygote excess. Third, the ratio of the number of observed genotypes (G) to the number of individuals sampled (N) ranged from 0.02 to 0.05 (0.024±0.011, mean±SD), and the ratio of the observed multi-locus genotypic diversity to that expected under sexual reproduction (G~O~/G~E~) from 0.141 to 0.048 (0.128±0.035, mean ± SD). Fourth, estimates of F~IS~ were −1.0 for all populations. ::: {#pone-0017524-t002 .table-wrap} 10.1371/journal.pone.0017524.t002 Table 2 ::: {.caption} ###### Comparison of genetic parameters among populations and across years. ::: {#pone-0017524-t002-2} Year 2008 2009 ----------------------- ------ -------- -------- -------- -------- -------- -------- -------- Number of MLGs 1 1 1 1 1 1 1 "Clone" Designation 1 1 1 1 1 2 1 Mean \# alleles/locus 1.400 1.400 1.400 1.400 1.400 2.000 1.400 G~O~/G~E~ 0.141 0.141 0.141 0.141 0.141 0.048 0.141 G/N 0.020 0.050 0.020 0.020 0.020 0.020 0.020 H~O~ 0.400 0.400 0.400 0.400 0.400 1.000 0.400 H~E~ 0.202 0.205 0.202 0.202 0.202 0.505 0.202 F~IS~ (multi-locus) −1.000 −1.000 −1.000 −1.000 −1.000 −1.000 −1.000 1 SGO = Statesboro, Georgia - Oleander; SGM = Statesboro, Georgia - Milkweed; TIGO = Tybee Island, Georgia - Oleander; CCO = Concord, California - Oleander. ::: To test for a non-random distribution of genetic variation among host plant species, we sampled *A. nerii* from two common host plants (Oleander and Milkweed) from Statesboro, GA (SGO and SGM, respectively). All samples from both hosts consisted of MLG 1, indicating that there is no host associated subdivision at this location (*F* ~ST~ = 0, *R* ~ST~ = 0) ([Table 3](#pone-0017524-t003){ref-type="table"}). ::: {#pone-0017524-t003 .table-wrap} 10.1371/journal.pone.0017524.t003 Table 3 ::: {.caption} ###### Pairwise molecular subdivision *F* ~ST~ (below diagonal) and *R* ~ST~ (above diagonal) between populations and across years. ::: {#pone-0017524-t003-3} Year 2008 2009 ------ ------ ----------- ----------- ----------- ----------- ----------- ----------- ----------- SGO \- 0.000 0.000 0.000 0.000 **0.787** 0.000 2008 SGM 0.000 \- 0.000 0.000 0.000 **0.758** 0.000 TIGO 0.000 0.000 \- 0.000 0.000 **0.787** 0.000 CCO 0.000 0.000 0.000 \- 0.000 **0.787** 0.000 SGO 0.000 0.000 0.000 0.000 \- **0.787** 0.000 2009 TIGO **0.650** **0.609** **0.650** **0.650** **0.650** \- **0.787** CCO 0.000 0.000 0.000 0.000 0.000 **0.650** \- 1 SGO = Statesboro, Georgia - Oleander; SGM = Statesboro, Georgia - Milkweed; TIGO = Tybee Island, Georgia - Oleander; CCO = Concord, California -- Oleander. **Bold values** indicate p\<0.0005. ::: There was no genotypic variation within any population for either 2008 or 2009, suggesting that each population was composed of a single genotypic "clone" ([Table 1](#pone-0017524-t001){ref-type="table"}). The geographic distribution of MLGs differed between years. Samples collected in 2008 from Tybee Island, Georgia; Statesboro, Georgia; and Concord, California comprised a single MLG, indicating no population subdivision (*F* ~ST~ = 0, *R* ~ST~ = 0) ([Table 3](#pone-0017524-t003){ref-type="table"}). The Tybee Island, Georgia samples differed significantly from both the Statesboro, Georgia and Concord, California samples in 2009 (*F* ~ST~ = 0.650, *R* ~ST~ = 0.787) ([Table 3](#pone-0017524-t003){ref-type="table"}). During the 2009 sampling period, the Tybee Island population consisted solely of MLG 2 individuals which were not found in any other population. Between 2008 and 2009 there was, interestingly, a complete change in the genetic composition of the Tybee Island population from MLG 1 to MLG 2. Discussion {#s4} ========== Small, genetically-uniform populations are subject to ecological and evolutionary forces (i.e., genetic bottlenecks and genetic drift) which threaten population persistence, in both native and introduced habitats [@pone.0017524-Lockwood1]. In animals, higher genetic diversity is often associated with an increased ability to establish viable populations in novel environments [@pone.0017524-Lockwood1]. This is not necessarily true for plants, however; some species are very successful with little or no genetic diversity, particularly those that are clonally-reproducing, self-pollinating, or apomictic [@pone.0017524-Ward1]. Here, we analyzed the genetic patterning of the Oleander aphid, *A. nerii* to better understand the ecology, life history, and population dynamics of this well established invasive species. We found that Oleander aphids are remarkably invasive throughout the southern United States, with extremely low genetic diversity. Reproduction and Life History {#s4a} ----------------------------- *Aphis nerii* is believed to be obligately parthenogenetic, based on the complete absence of males under natural conditions [@pone.0017524-Blackman1]. Laboratory lines of the aphid derived from populations in Kyoto, Japan produced males when exposed to short-day conditions [@pone.0017524-Takada1], [@pone.0017524-Takada2], suggesting that the ability to sexually reproduce is retained in at least some asexual lineages from some populations. From genotypic data, we found no evidence for sexual reproduction in any of the populations we examined, although, of course, further sampling may yet reveal sexual forms. Obligate parthenogenesis is supported by the lack of expected recombinant genotypes observed over the two year sampling period. A high level of heterozygosity between genotypes is consistent with expected genotypic patterns for long term asexual populations; i.e., the "Meselson effect" [@pone.0017524-Welch1]. That is, in long term asexual populations, heterozygosity is expected to increase because allelic pairs within a genome will continue to diverge over time while meiotic recombination and segregation do not occur to mix and purge alleles [@pone.0017524-Balloux1], [@pone.0017524-DeMeeus1], [@pone.0017524-Meselson1]. This genotypic pattern has been observed in several primarily parthenogenetic taxa [@pone.0017524-Simon3], [@pone.0017524-Butlin1], [@pone.0017524-Corrie1], [@pone.0017524-Delmotte1], [@pone.0017524-Muller1]. Our results are consistent with this pattern, suggesting that sexual reproduction is rare or non-existent in natural populations of *A. nerii* in the southern US. Geographic Genetic Variation {#s4b} ---------------------------- The most striking pattern observed in this study was the low level of genotypic diversity within and among populations over a large geographic area. We observed only two MLGs and a maximum estimate of genotypic diversity (G/N) within any population of 0.05. Several similar studies of other aphid species using comparable numbers of loci (4 to 7) have revealed higher levels of genetic variation than that observed in *A. nerii*. For example, long term asexual populations of the pea aphid, *Acyrthosiphon pisum* (Harris) in Japan were composed of five to seven MLGs, and cyclically parthenogenetic populations harbored much greater diversity than did asexual populations [@pone.0017524-Kanbe1]. Genotypic diversity (G/N) estimates in this species range from 0.10 to 0.69 [@pone.0017524-Kanbe1]. Several other aphid species show similar or greater levels of genotypic diversity as *A. pisum*, both within and among populations [@pone.0017524-Dedryver1], [@pone.0017524-Fuller1], [@pone.0017524-Delmotte2], [@pone.0017524-Massonnet1], [@pone.0017524-Vorburger1]. Aphid populations are sometimes composed of a small number of dominant genotypes ("clones") and many low frequency (rare) genotypes [@pone.0017524-Corrie1], [@pone.0017524-Vorburger1], [@pone.0017524-Sunnucks1], [@pone.0017524-Haack1], [@pone.0017524-Peccoud1]. The term "superclone" has been used to describe genotypes that comprise 40--60% of a population in a region [@pone.0017524-Vorburger2]. For example, Peccoud *et al.* [@pone.0017524-Peccoud1] sampled recently introduced asexual populations of *A. pisum* in Chile. Among the 432 individuals sampled over a 570 km range, 16 MLG\'s were identified with three MLG\'s comprising≈90% (47%, 29%, and 14%) of the diversity. Compared to previous studies, the pattern we observe in *A. nerii* is extreme even for "superclones". Within any sampling year, each population consisted of a single MLG. In 2008, all populations sampled from both California and Georgia showed only a single MLG (MLG 1). In 2009, populations from Statesboro, Georgia and California comprised the same MLG (MLG 1), whilst a population from Tybee Island, Georgia comprised a single but entirely different MLG (MLG 2). Variation of morphometric and life history traits in *A. nerii* has been assessed within and among populations from California, Iowa, and Puerto Rico [@pone.0017524-Groeters1]. Significant variation was observed within populations for traits such as maturation time, fecundity, and wing length while only the proportion of winged offspring differed among populations [@pone.0017524-Groeters1]. Assuming low genetic diversity in this population, these data suggest that phenotypic plasticity and maternal effects may be of utmost importance in shaping the population dynamics of this parthenogenetic species [@pone.0017524-Lushai1], [@pone.0017524-Lushai2], [@pone.0017524-Wilson1]. The genetic uniformity observed in *A. nerii* likely results from successive genetic bottlenecks or founder events. Subsequent clonal propagation and rapid spread of genotypes throughout the host range would be followed by rapid clonal competition/selection resulting in a few geographically-widespread and dominant clones (but see also [@pone.0017524-Loxdale3]). These processes would occur during the initial introduction of this species into the US and/or during annual extinction and recolonization events. A study of *A. nerii* on *N. oleander* in California clearly showed annual colonization and extinction cycles [@pone.0017524-Hall1], corresponding with our own observations. Patches of host plants are typically colonized by aphids in late Spring (May or June) followed by rapid increases in population sizes which peak in mid-summer. Aphid numbers decline rapidly in early Fall (September or October), resulting in population extinction [@pone.0017524-Hall1]. In the southernmost US latitudes, *A. nerii* has the potential to overwinter as adults, as freezing temperatures are uncommon. Host plant quality, however, substantially decreases in Fall and Winter likely driving the population decline [@pone.0017524-Hall1]. Milkweed (*Asclepias* spp.) is the most common host of *A. nerii* in the northern US where *N. oleander* grows infrequently [@pone.0017524-Blackman1]. *Asclepias* species are perennial plants that sprout in Spring, bloom in early Summer, and then set seed and die-back in the Fall. The life history of this host plant would require *A.nerii* to have a secondary host (though this has not previously been reported) or to produce sexuals to produce overwintering eggs (and as previously noted, has only been noted under laboratory conditions). On both Oleander and Milkweed, a regular pattern of colonization, rapid increase, and population extinction would be expected. Host Associated Genetic Variation {#s4c} --------------------------------- Groeters [@pone.0017524-Groeters2] found no evidence of fitness trade-offs in *A. nerii* when feeding on different species of milkweed, but differences between Oleander and Milkweeds were not examined. Prior to this study, it was not known if *A. nerii* populations inhabiting Oleander and Milkweed were genetically different. Several aphid species have "host races" or show non-random distribution of genetic variation among host plants, including the Pea aphid [@pone.0017524-Via1], [@pone.0017524-Peccoud1], the Grain aphid, *Sitobion avenae* (F.) [@pone.0017524-Lushai2], [@pone.0017524-Vialatte1], the Cotton /Melon aphid, *Aphis gossypii* Glover [@pone.0017524-VanlerbergheMasutti1] and others [@pone.0017524-RuizMontoya1], [@pone.0017524-Miller1], [@pone.0017524-Akimoto1]. This phenomenon undoubtedly results from habitat choice/host fidelity inhibiting interpopulation gene flow [@pone.0017524-Via1], [@pone.0017524-Via2]. If the same process of habitat choice/host fidelity applied to *A. nerii* populations, we would have expected to see genetic variation between populations inhabiting Oleander and those inhabiting Milkweed, regardless of geographic proximity. Our findings suggest that there is no selection for host specificity and that *A. nerii* is a polyphagous, i.e. generalist species, although more data are required to confirm this contention. This pattern is consistent with Lynch [@pone.0017524-Lynch1] who suggested that obligate parthenogenetic species evolve to be ecological generalists (selection favors clones that can survive in all environments). In summary, populations of *A. nerii* surveyed in this study show the genetic signature of obligate parthenogenetic reproduction, supporting previous reports that the species is indeed obligately asexual. Within any sampling period, each population was composed of a single MLG. This level of variation is remarkably low compared to the variation observed in other aphid species presumed to be largely or completely asexual [@pone.0017524-Corrie1], [@pone.0017524-Vorburger1], [@pone.0017524-Peccoud1]. Furthermore, only two MLGs were identified among all populations, and populations separated by as far as 3,600 km were genetically homogenous. Temporal variation occurred in one population, one MLG was completely replaced by another between years. There was no correlation between host plant and MLG. In an ecological context, our results suggest that *A. nerii* is a generalist species with strong dispersal capabilities. As patches of host plants are colonized by few individuals that reproduce rapidly through parthenogenesis, founder individuals likely come from a source population(s) characterized by low genetic diversity. Despite having extremely low genetic diversity Aphis nerii is a well established invasive species. Understanding the temporal and geographic genetic variation of *A. nerii* provides great insight for the management of invasive pests and the population dynamics of clonal organisms. We thank Carlie Smith and Hillary Williams for laboratory and field support and Hugh Loxdale for his valuable comments on an earlier version of this manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by Georgia Southern University and Advisement and Scholarship Promoting Inquiry-based Research Experiences in STEM (ASPIRES) funded through NSF-STEP grant (DUE-0622460). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JSH EBM. Performed the experiments: JSH EBM. Analyzed the data: JSH EBM. Contributed reagents/materials/analysis tools: JSH EBM. Wrote the paper: JSH EBM.

PubMed Central

2024-06-05T04:04:19.161122

2011-3-9

PMC3052317

Introduction {#s1} ============ Among the different advantages commonly linked to an increase in body size [@pone.0017628-Sander1], a widespread concept is that of an increasing digestive efficiency in larger herbivores. Based on the observation that energetic requirements of animals scale to metabolic body mass (i.e., M^0.75^) but gut capacity scales linearly with body mass (M^1.0^) in mammalian herbivores, Bell [@pone.0017628-Bell1] and Jarman [@pone.0017628-Jarman1] deducted that at larger M, more gut capacity was available per unit energy requirement/food intake. This so-called 'Jarman-Bell principle' [@pone.0017628-Geist1] was further refined subsequently [@pone.0017628-Parra1]--[@pone.0017628-Illius1] and has found widespread application in ecology [@pone.0017628-Fleming1]--[@pone.0017628-McNab1]. This attractive concept provides an intuitive reason for the observation that larger-bodied herbivores usually ingest food of lower nutritional quality [@pone.0017628-Codron1], [@pone.0017628-OwenSmith1]. However, recent findings do not support the notion that digestibility [@pone.0017628-PrezBarbera1], [@pone.0017628-Clauss1] or ingesta retention [@pone.0017628-Clauss2] increase systematically with body mass in mammals, and also not in herbivorous reptiles [@pone.0017628-Franz1]. Among potential disadvantages, ingesta particle size -- one of the factors influencing digestive efficiency -- increases with body mass [@pone.0017628-Fritz1], [@pone.0017628-Fritz2], and it has been suggested that energetic losses due to methane production are also higher in larger animals [@pone.0017628-Clauss3]. Methane production has been mainly measured in domestic herbivores to address the issue of feed energy use or, more recently, methane mitigation to reduce greenhouse gas emissions [@pone.0017628-Martin1]. Studies on methane production of non-domestic species have mainly been to complete national or global methane budgets [@pone.0017628-Crutzen1]. In contrast, comparative investigations on methane production with respect to herbivore physiology are rare. Methane production has been demonstrated in faeces of captive specimens of nearly all herbivorous terrestrial herbivores, including reptiles [@pone.0017628-Hackstein1], and methanogenes have been demonstrated by fluorescence microscopy in land and marine iguanas [@pone.0017628-Mackie1]. In vivo methane production has not been investigated in reptiles to our knowledge. Recently, Franz et al. [@pone.0017628-Franz2], [@pone.0017628-Franz3] presented data collections that suggest that methane production scales linearly with M in ruminant and nonruminant mammalian herbivores. The implication of this finding is that because food intake scales to M^0.75^, energetic losses due to methane increase per unit ingested food with increasing body size. Thus, methane energy losses could become a serious constraint in species with large body size. Similarly, allometric relationships were the basis of the investigation of Smith et al. [@pone.0017628-Smith1] who found that the body mass distribution in a herbivore fauna will impact this fauna\'s contribution to the global methane budget. Apparantly, methane production scales differently than metabolic requirements or rates. In order to test the concept of disproportionately increasing methane losses with increasing herbivore M with an original dataset, we chose herbivores of another clade, tortoises. In tortoises, a large range of *M* is available with minimal differences in digestive anatomy and physiology. Scaling of food intake, gut capacity or digesta retention with M is generally similar in herbivorous reptiles and mammals [@pone.0017628-Fritz2], [@pone.0017628-Franz4]. The aim of our study was to test whether, in tortoises, voluntary food intake scales to M^0.75^, and methane production scales linearly with M. Materials and Methods {#s2} ===================== This study was performed in accordance with Swiss animal welfare legislation (approved by the Cantonal Veterinary Office Zurich under experimental licence number 192/2006). We performed intake and respiration chamber measurements in 24 individual tortoises of the species *Testudo graeca* (n = 5, 1.16±0.95 kg, range 0.52--2.83 kg), *T. hermanni* (n = 6, 1.28±0.36 kg, range 0.91--1.72 kg), *G. nigra* (n = 2, 5.50±0.28 kg, range 5.30--5.70 kg), *Geochelone sulcata* (n = 8, 27.8±18.0 kg, range 7.2--50.0 kg), *Dipsochelys dussumieri* (n = 3, 141±38 kg, range 104--180 kg). Animals were kept individually for 30 days at 27--30°C for intake measurements after an adaptation period of one week. The diet consisted of grass hay and salad in varying proportions; details on intake and digestibility measurements were described previously [@pone.0017628-Franz1]. Water was available ad libitum at all times. Feed offered and left over was quantified, and faeces were collected completely. Representative subsamples were used to determine dry matter (DM), crude protein, gross energy (GE) and neutral detergent fibre (NDF) concentrations using standard methods [@pone.0017628-AOAC1]; these data allowed the calculation of the apparent digestibility of DM, GE and NDF [@pone.0017628-Robbins1]. Experimental conditions or sample size did not always allow all analyses to be performed for all individuals (cf. [Table 1](#pone-0017628-t001){ref-type="table"}). The ingested diets contained crude protein at 130±18 g kg DM^−1^ (range 95--170) and NDF at 488±107 g kg DM^−1^ (296--662). ::: {#pone-0017628-t001 .table-wrap} 10.1371/journal.pone.0017628.t001 Table 1 ::: {.caption} ###### Allometric scaling relationships for tortoises (T), mammalian nonruminants (NR) and ruminants (R) for daily methane production with body mass (M) according to the equation *y* = *a* M*^b^*. ::: {#pone-0017628-t001-1} Herbivore group *y* unit n[\*](#nt102){ref-type="table-fn"} *a* *95% CI a* *b* *95% CI b* *r^2^* *p* ----------------- --------- ----------------- ------------------------------------ ------- -------------- ------ -------------- -------- --------- T Methane L d^−1^ 24 0.014 0.009--0.023 1.03 0.84--1.22 0.85 \<0.001 NR 41 0.181 0.144--0.227 0.97 0.92--1.02 0.98 \<0.001 R 62 0.661 0.420--1.040 0.97 0.88--1.07 0.87 \<0.001 T L (kg DMI)^−1^ 22 3.02 2.07--4.40 0.33 0.18--0.47 0.52 \<0.001 NR 25 3.34 2.63--4.26 0.16 0.10--0.22 0.59 \<0.001 R 45 16.58 12.17--22.60 0.12 0.06--0.18 0.25 \<0.001 T L (kJ GEI)^−1^ 21 0.70 0.47--1.05 0.29 0.139--0.446 0.46 0.001 NR 25 0.79 0.63--0.99 0.15 0.093--0.204 0.57 \<0.001 R 44 3.53 2.52--4.94 0.13 0.058--0.195 0.25 \<0.001 T L (kJ DEI)^−1^ 16 0.91 0.51--1.60 0.32 0.13--0.51 0.45 0.003 NR 31 1.48 1.21--1.81 0.17 0.13--0.21 0.71 \<0.001 R 35 7.87 5.13--12.06 0.09 −0.001--0.18 0.11 0.053 T L (g dNDFI)^−1^ 21 10.1 6.6--15.5 0.30 0.13--0.46 0.43 0.001 NR 23 11.1 9.1--13.5 0.17 0.12--0.22 0.70 \<0.001 R 17 57.4 26.3--125.2 0.11 −0.05--0.27 0.12 0.170 DM dry matter, GE gross energy, DE digestible energy, dNDF digestible neutral detergent fibre, I intake tortoise data from this study; ruminant data collection from Franz et al. [@pone.0017628-Franz2], nonruminant data collection from Franz et al. [@pone.0017628-Franz3]. \*sample sizes vary between measurements because for tortoises, not all measurements could be performed due to logistic reasons, and because for mammals, data available from the literature varied between sources. ::: After 30-day intake measurements, tortoises were transferred to open circuit respiration chambers constructed and operated as described in Soliva and Hess [@pone.0017628-Soliva1] for two consecutive 22.5 h periods (temperature 29±1°C, constant humidity 60%, pressure 987±8 hPa; chambers for M from 0.5--10 kg: volume 0.85 m^3^, air flow 1.09±0.08 m^3^ h^−1^; chambers for M from 20--180 kg: volume 4.55 m^3^, air flow 6.08±2.77 m^3^ h^−1^). Animals were measured individually except for the tortoises \<5 kg; after pilot measurements, two groups of five individuals between 0.5--2 kg and one group of three individuals between 2--3 kg were measured together, and results divided by the number of animals. Animals had access to feed and water in the respiration chambers. All gas volumes were corrected for standard conditions (1013 hPa, 0°C, 0% relative humidity). Methane concentrations were measured by Binos 1001 (infra-red; Fisher-Rosemount, Baar-Walterswil, Switzerland). Following various conventions in the scientific literature, daily methane production was not only expressed in absolute terms, but also in relation to DM, GE, digestible energy (DE) and digestible NDF (dNDF) intake. Data were analysed after ln-transformation using regression analysis with PSAW 18.0 (SPSS Inc., Chicago, IL), indicating 95% confidence intervals (*95%CI*) according to y = *a* M*^b^* or ln~y~ = ln*~a~*+*b* ln~M~. Results {#s3} ======= Mean dry matter intake (in kg d^−1^) of the tortoises scaled to 0.005 (*95%CI* 0.004--0.007) M^0.75\ (*95%CI*\ 0.64--0.87)^ (*n* = 22, *r^2^* = 0.90, *p*\<0.001) and mean daily gross energy intake (in kJ d^−1^) to 86.1 (*95%CI* 64.5--114.7) M^0.77\ (*95%CI*\ 0.66--0.88)^ (*n* = 21, *r^2^* = 0.92, *p*\<0.001). In contrast, mean daily methane production scaled linearly to M ([Table 1](#pone-0017628-t001){ref-type="table"}, [Fig. 1](#pone-0017628-g001){ref-type="fig"}). During measurements in the respiration chamber, it was noted that methane production was not constant throughout the day but occurred in distinct bursts ([Fig. 2](#pone-0017628-g002){ref-type="fig"}). ::: {#pone-0017628-g001 .fig} 10.1371/journal.pone.0017628.g001 Figure 1 ::: {.caption} ###### Relationship between body mass and absolute daily methane production; data for ruminants (dark grey regression line; data collection from Franz et al. [@pone.0017628-Franz2]), nonruminant mammalian herbivores (light grey regression line; data collection from Franz et al. [@pone.0017628-Franz3]) and for tortoises in this study. :::  ::: ::: {#pone-0017628-g002 .fig} 10.1371/journal.pone.0017628.g002 Figure 2 ::: {.caption} ###### Example of methane production in an open circuit respiration chamber in a *Geochelone sulcata* (10.5 kg) for one uninterrupted measurement period of 22 hours. :::  ::: When expressed in relation to intake of digestible energy and fibre, methane losses scaled to M^0.32^ and M^0.30^, respectively ([Table 1](#pone-0017628-t001){ref-type="table"}, [Fig. 3](#pone-0017628-g003){ref-type="fig"} and [4](#pone-0017628-g004){ref-type="fig"}). The *95%CI* of scaling exponent *b* overlapped between tortoises, nonruminant mammals, and ruminants where data had been obtained in previous assessments [@pone.0017628-Franz2], [@pone.0017628-Franz3], except for the scaling exponent when methane was related to digestible energy (not significant in ruminants). The *95%CI* of factor *a* was invariably higher in ruminants than in the other two groups ([Table 1](#pone-0017628-t001){ref-type="table"}). ::: {#pone-0017628-g003 .fig} 10.1371/journal.pone.0017628.g003 Figure 3 ::: {.caption} ###### Relationship between body mass and methane energy losses in % of daily digestible energy intake; data for ruminants (dark grey regression line; data collection from Franz et al. [@pone.0017628-Franz2]), nonruminant mammalian herbivores (light grey regression line; data collection from Franz et al. [@pone.0017628-Franz3]) and for tortoises in this study. :::  ::: ::: {#pone-0017628-g004 .fig} 10.1371/journal.pone.0017628.g004 Figure 4 ::: {.caption} ###### Relationship between body mass and methane energy losses related to the daily intake of digestible cell wall (neutral detergent fibre); data for ruminants (dark grey regression line; data collection from Franz et al. [@pone.0017628-Franz2]), nonruminant mammalian herbivores (light grey regression line; data collection from Franz et al. [@pone.0017628-Franz3]) and for tortoises in this study. :::  ::: Discussion {#s4} ========== The results of this study suggest that in herbivores, methane production scales linearly with body mass, and the proportional losses of energy from feed ingested due to methane output increase with increasing body mass. Although the existing data must still be considered scarce, the parallel findings in ruminant and nonruminant mammalian herbivores and herbivorous tortoises strongly suggest a general scaling pattern. Similar scaling patterns in reptiles and mammals have been found for other parameters such as field metabolic rate [@pone.0017628-Nagy1], [@pone.0017628-Bennett1], feed intake [@pone.0017628-Clauss2], [@pone.0017628-Franz1], [@pone.0017628-Meienberger1], or ingesta particle size [@pone.0017628-Fritz2] -- although on different levels; whilst some other measures appear relatively similar between herbivorous reptiles and mammals, such as the proportion of the gut contents of total body mass [@pone.0017628-Franz1], [@pone.0017628-Franz4] or the achieved digestibilities [@pone.0017628-Hatt1], [@pone.0017628-Karasov1]. Generally, it is assumed that energy metabolism in reptiles is roughly a tenth of that observed in mammals [@pone.0017628-Kirkwood1]. The difference in the intercept *a* of the regression equation describing dry matter intake in the tortoises of this study (0.005) compared to the intercept of 0.047 found in herbivorous mammals in general [@pone.0017628-Clauss2] fits this pattern, as does the difference in the intercept describing the absolute methane output (0.014 in tortoises vs. 0.181 in nonruminant mammals, [Table 1](#pone-0017628-t001){ref-type="table"}). Consequently, when methane production is expressed per unit intake, there is no significant difference in the intercept *a* between tortoises and nonruminant mammals ([Table 1](#pone-0017628-t001){ref-type="table"}). This finding indicates a common adaptation of the gastrointestinal fauna between ectotherms and endotherms. Other similarities between the microbial faunas of herbivorous reptiles and mammals have been reported, such as the number of gut bacteria and the presence of protozoa [@pone.0017628-McBee1]--[@pone.0017628-Fenchel1], cellulase activity [@pone.0017628-Nagy2], or the concentration of fermentation products [@pone.0017628-Foley1]--[@pone.0017628-Bjorndal1]. A relatively similar methane production per unit food intake in reptiles and mammals means that the processes of microbial fermentation must be similar even though the microbial faunas of reptiles and mammals will vary distinctively in their temperature sensitivity. The findings suggest that methane production is a more or less constant, unavoidable by-product of microbial fermentation in herbivores. Because of the well-documented differences in ingesta retention times between herbivorous reptiles (230±140 h [@pone.0017628-Franz1], [@pone.0017628-Hailey1]) and mammals (40±25 h [@pone.0017628-Clauss2]), the similarity in methane scaling between reptiles and mammals also indicates that retention time as such is not the main factor influencing the scope of methane production, even if it may be relevant when comparing data within species [@pone.0017628-PinaresPatio1], [@pone.0017628-Okine1]. Our results also suggest that the increase in methane production with increasing body size is not only due to an increase in fibre digestibility at higher body sizes; when expressed per unit of digestible fibre intake, the effect of an increasing methane production remains and scales similarly with M as when expressed in relation to other intake measures ([Table 1](#pone-0017628-t001){ref-type="table"}). Prins and Kreulen [@pone.0017628-Prins1] and Van Soest [@pone.0017628-VanSoest1] suggested that a different group of methanogenes -- slower-growing archeae with a generation time of about 4 days that produce methane from acetate in sewers, for example -- may actually limit body size in herbivores. They considered ingesta retention a function of body mass [@pone.0017628-Demment1], [@pone.0017628-Illius1], [@pone.0017628-Clauss2] and hypothesized that when retention times surpass 4 days, energetic losses due to acetate-based methanogenesis would become prohibitve for the host. In herbivorous reptiles retention times beyond 96 h are common [@pone.0017628-Hailey1], [@pone.0017628-Hatt2] which indicates that other factors than retention time must limit the occurence of slow-growing archeae in herbivores. An interesting question is could methane production by the fast-growing archeae be a constraint on the evolution of body size? This has been suggested for ruminants, due to the high proportion of energetic methane losses in this group [@pone.0017628-Franz2]; for nonruminant mammals, these losses might become limiting at extrapolated body masses of 100 metric tonnes [@pone.0017628-Franz3] -- a putative constraint that might apply conceptually for the largest dinosaurs [@pone.0017628-Sander1]. Reptiles never reached such proportions. When the regression equation from tortoises is directly applied to the largest known chelonian, *Archelon ischyros*, a marine turtle with an estimated maximum M of 5000 kg [@pone.0017628-Anonymous1], extrapolated methane energy losses per unit of digestible energy intake (14%) approach those found in large ruminants. Note that this similarity to ruminants, in spite of the general similarity in scaling between tortoises and nonruminant mammals, is due to the determined exponent *b* of 0.32, which is numerically higher than the one calculated for nonruminant mammals (0.17), though overlapping in its confidence interval. Differences in exponent should be considered with caution when extrapolations beyond the M range are performed that served to generate the regression equation [@pone.0017628-Franz4]. Why herbivores apparently did not evolve to avoid methane losses is a fundamental question. Intervention studies in domestic ruminants have shown that functional digestion can be maintained in the absence or near-absence of Archeae and without methane production [@pone.0017628-McCrabb1]--[@pone.0017628-Sawyer1]. An alternative view of methanogenes could be that they are among the prerequisites for herbivory. Pimentel et al. [@pone.0017628-Pimentel1] showed that, in a models with dogs and guinea pigs, methane slowed intestinal passage by decreasing intestinal contractile activity. In humans, methane production is associated with increased digesta retention times [@pone.0017628-Soares1]--[@pone.0017628-Stephen1], and is positively correlated with constipation and negatively with diarrhoea [@pone.0017628-Chatterjee1], [@pone.0017628-Hwang1]. Reduction of methane production by oral antibiotic treatment leads to a reduction of constipation [@pone.0017628-Pimentel2], [@pone.0017628-Low1]. While offering new insights into potential therapeutical interventions against human irritable bowel syndrome, these results also give rise to the speculation that the presence of methane, and its passage-delaying effect, was an important component of the evolution of physiological adaptations to herbivory, which requires long passage times. However, confirmation of this hypothesis requires much further research. Our study shows that methane losses not only occur in mammalian but also in reptilian herbivores, and that they scale linearly with body mass, thus representing proportionally increasing losses at increasing body size. Therefore, differences in the proportion of ingested energy lost to methane, according to the body size composition of any mammal or reptile herbivore fauna should be considered when reconstructing trophic energy fluxes in ecosystems, or contributions of these ecosystems to changes in the composition of the atmosphere [@pone.0017628-Smith1]. Further studies combining in vivo measurements and microbiological analyses should unravel the fundamental principles behind the link between microbial fibre fermentation in vertebrate herbivores and methane production. We thank the Zoological Garden of Zurich, Peter Sandmeier, Ruth Huber and Otti Steck for providing the experimental animals. This is contribution no. 99 of the DFG Research Unit 533 "The Biology of Sauropod Dinosaurs". **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was funded by DFG (German Research Foundation) grant CL 182/5-1. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MC JH CRS MK J-MH. Performed the experiments: RF SF MC. Analyzed the data: RF JH MC. Contributed reagents/materials/analysis tools: CRS MK SF J-MH. Wrote the paper: RF MK MC.

PubMed Central

2024-06-05T04:04:19.163253

2011-3-9

PMC3052318

Introduction {#s1} ============ The phlebotomine sand fly *Lutzomyia longipalpis*, Lutz and Neiva 1912 is the best studied and most important vector of American Visceral Leishmaniasis (AVL) [@pone.0017486-Lainson1], [@pone.0017486-Soares1]. It is widely found in Latin America, from the South of Mexico to the North of Argentina [@pone.0017486-Grimaldi1], [@pone.0017486-Romero1]. *Lu. longipalpis* is a permissive vector to *Leishmania* infections [@pone.0017486-Myskova1] and this together with its wide distribution in urban environments [@pone.0017486-Costa1] makes this sand fly species an ideal model to study phlebotomine physiology to help develop future vector control methods for the disruption of *Leishmania* transmission. Previous studies in other blood-feeding insect vectors have shown that reactive oxygen species (ROS) play an important role in both reproductive output [@pone.0017486-DeJong1] and survival [@pone.0017486-Ha1], [@pone.0017486-MagalhaesDEB1], [@pone.0017486-GracaSouza1], [@pone.0017486-Citelli1], [@pone.0017486-Walshe1]. Biological damage related to ROS production has also been implicated in the process of ageing in dipterans like *Drosophila melanogaster*, previous work done on this species showed that oxidative stress increases with age, while antioxidant enzyme activity decreased over time [@pone.0017486-Das1], [@pone.0017486-Sohal1], [@pone.0017486-Ferguson1]. ROS are regularly generated by mitochondrial electron transport [@pone.0017486-Loft1]. Partially reduced and highly reactive metabolites of O~2~ such as superoxide anion (O~2~ ^−^·) and hydrogen peroxide (H~2~O~2~) are formed during cellular respiration. These partially reduced metabolites of O~2~ are often referred to as reactive oxygen species due to their higher reactivity in relation to molecular oxygen [@pone.0017486-Thannickal1]. Excessive release of ROS damages lipids, proteins, and DNA [@pone.0017486-Freeman1] which leads to oxidative stress, loss of cell function, and programmed cell death [@pone.0017486-Nordberg1]. ROS are also actively released as a response against bacterial and parasitic pathogens in different insect species [@pone.0017486-Ha1], [@pone.0017486-Leto1], [@pone.0017486-MolinaCruz1]. To regulate oxidative stress, the eukaryotic cell produces different ROS-scavenging enzymes, such as superoxide dismutase (which reduces O~2~ ^−^· to H~2~O~2~), glutathione peroxidase and catalase (which reduces H~2~O~2~ to H~2~O) [@pone.0017486-Thannickal1]. Although many studies have been published regarding mechanisms to resist oxidative stress in *Leishmania* parasites [@pone.0017486-Miller1], [@pone.0017486-Levick1], [@pone.0017486-Mandal1], [@pone.0017486-Mehta1] there is little information on ROS-scavenging molecular mechanisms on sand flies, despite the fact that putative antioxidant enzymes have been found to be upregulated in two different species of phlebotomine sand flies upon *Leishmania* infection [@pone.0017486-Jochim1], [@pone.0017486-Dillon1]. In order to investigate the biological role of ROS-scavenging in fecundity and survival, we analysed the expression of catalase in three different age groups of female sand flies. Fecundity and catalase expression decreased with age. Catalase was incriminated as an important component in the loss of fecundity using RNAi. Dietary supplementation with an exogenous ROS-scavenger was found to partially reverse the differences in fecundity and increase mortality with a concomitant activation of the phenoloxidase (PO) cascade. These results show that both fecundity and survival are affected by endogenous and exogenous ROS-scavenging in female *Lutzomyia longipalpis*. Results {#s2} ======= Age-related decrease of fecundity {#s2a} --------------------------------- To evaluate the effect of ageing in fecundity, females of different age groups were blood fed and dissected to examine difference in developing oocyte numbers. Female *Lu. longipalpis* from the older age group showed a decrease in the number of developing oocytes dissected five days after blood feeding in comparison to younger sand flies([fig. 1](#pone-0017486-g001){ref-type="fig"}). Female *Lu. longipalpis* that were bloodfed 3 and 6 days post-emergence (PE) showed no significant difference in oocyte numbers. However, sand flies bloodfed at 9 days PE showed a significant decrease in oocyte numbers after dissecting 5 days after blood feeding ([fig. 1](#pone-0017486-g001){ref-type="fig"}; P\<0.005, ANOVA). ::: {#pone-0017486-g001 .fig} 10.1371/journal.pone.0017486.g001 Figure 1 ::: {.caption} ###### Effect of age at blood feed on subsequent fecundity of female *Lu. longipalpis*. Bars represents average number of oocytes dissected 5 days after blood meal ± SEM. Sand flies were blood-fed at 3, 6 and 9 days Post-Emergence. Asterisk indicates statistical difference at P\<0.005 (ANOVA). Results represent two independent biological replicates. :::  ::: ROS-scavenging reverses age related loss of fecundity {#s2b} ----------------------------------------------------- To evaluate the role of ROS scavenging in age-related decrease of fecundity, 9 day old female *Lu. longipalpis* were fed a sucrose meal supplemented with a ROS-scavenger upon emergence until end of the experiment. Sand flies were offered a 70% sucrose solution supplemented with 20 mM ascorbic acid and subsequently blood-fed on day 9 PE. 20 mM ascorbic acid was chosen after evaluating mortality of sand flies when offered 100, 50, 20, 10 and 5 mM ascorbic acid in 70% sucrose (data not shown). The number of developing oocytes dissected 5 days after blood feeding was significantly higher ([fig. 2](#pone-0017486-g002){ref-type="fig"}; P\<0.0001, t-test) in sand flies that received a sugar meal supplemented with 20 mM ascorbic acid in comparison to control sand flies fed on 70% sucrose solution. This suggests that exogenous ROS-scavenging can reverse age-related loss of fecundity in sand flies blood fed 9 days PE. ::: {#pone-0017486-g002 .fig} 10.1371/journal.pone.0017486.g002 Figure 2 ::: {.caption} ###### Effect of ascorbic acid supplementation on fecundity in *Lu. longipalpis*. Flies were blood-fed 9 days Post-Emergence and bar chart represents average number of oocytes dissected 5 days after blood meal ± SEM (combined samples derived from 2 independent experiments). Sand flies fed on 20 mM ascorbic acid-supplemented 70% sucrose solution show significantly higher oocyte numbers in comparison to control sand flies (P\<0.0001, *t*-test). :::  ::: Catalase activity is reduced in developing oocytes of older flies and ROS scavengers reverse catalase depletion {#s2c} --------------------------------------------------------------------------------------------------------------- Flies were assayed at 24 h and 48 h to find out if catalase accumulated in developing oocytes. Ovaries of *Lu. longipalpis* dissected 6 days PE contained higher catalase enzymatic activity at 48 h compared to 24 h after blood feeding ([fig. 3A](#pone-0017486-g003){ref-type="fig"}; P\<0.0001 , t-test). Moreover, mRNA expression of catalase increased with oocyte development from 12 to 48 hours after blood feeding ([fig. 3B](#pone-0017486-g003){ref-type="fig"}). ::: {#pone-0017486-g003 .fig} 10.1371/journal.pone.0017486.g003 Figure 3 ::: {.caption} ###### Changes in catalase in the developing oocyte of *Lu. longipalpis*. \(A) Catalase activity of developing oocytes after blood feeding. Six day old female *Lu. longipalpis* were blood-fed and dissected at 24 and 48 hours. Enzymatic activity in the developing oocytes was significantly higher at 48 hours compared to 24 hours after blood feeding (*P*\<0.0001 , *T*-test). Bar charts represent mean ± SE of combined samples from 2 independent experiments. (B) Relative expression of catalase LlongKat1 mRNA in developing oocytes dissected at 12, 24 and 48 hours from 6 days-old blood-fed female *Lu longipalpis*, (*n* = three groups of 20 females each). Asterisk indicates statistical difference at P\<0.05 (ANOVA). Bar charts represent mean ± SEM of combined samples from 2 independent experiments. (C) Age-related decrease of catalase mRNA relative expression in developing oocyte. Flies were blood-fed at 3, 6 and 9 days Post-Emergence (*n* = three groups of 15 females each) and whole ovaries were dissected 48 hours after blood feeding. Relative expression was statistically different in all 3, 6 and 9 days old flies (P = 0.001, ANOVA). A 4^th^ group (n = 15 females) fed on an ascorbic acid-supplemented sugar solution upon emergence (9-AscA) showed catalase relative expression levels similar to groups of younger flies fed on 70% sucrose solution, and statistically higher than the non-treated, 9 DPE group (P\<0.002, t-test). Bar charts represent mean ± SEM of combined samples from 2 independent experiments. :::  ::: To further understand the role of endogenous ROS-scavenging and ageing, flies from different age groups were assayed for catalase LlonKat1 expression. Flies from different age groups (3, 6 and 9 days PE) showed a decrease in expression in ovaries dissected at 48 hours after blood feeding ([fig 3c](#pone-0017486-g003){ref-type="fig"}; P = 0.001, ANOVA). Interestingly, when 9 day old sand flies were fed with a 20 mM ascorbic acid supplemented sugar meal, catalase LlonKat1 mRNA expression was significantly higher compared to flies of the same age fed on sucrose only ([fig 3c](#pone-0017486-g003){ref-type="fig"}; P\<0.002, t-test). The results show that a) catalase accumulates in the developing oocyte as shown by increase in enzymatic activity and relative expression, b) catalase expression is age-dependant and is lower in older flies and c) the dietary supplementation with an exogenous ROS-scavenger increases catalase expression in older flies. *Lutzomyia longipalpis* catalase sequence was already described [@pone.0017486-Jochim1], [@pone.0017486-Dillon1] and was retrieved from the GenBank (ABV60342.1). It codes for a protein (named LlonKat1 in this study) with molecular mass of 57682 Da and isoelectric point of 8.28, without a signal peptide and mitochondrial or peroxisomal targeting sequences of types 1 and 2. LlonKat1 has high identity (ranging from 46--73%) to catalase sequences from other insects, crustaceans, yeast and mammals and lower identity to the bacterial catalase from *Pseudomonas syringae* ([fig. 4](#pone-0017486-g004){ref-type="fig"}). LlonKat1 sequence contains the conserved residues His73 and Asn147 (catalytic), Ser113, Val115, Phe152, Phe160, Leu298, Met349, Arg353, Tyr357 (heme binding/coordination) and His193, Arg202, Ile301, Gln304 (putative NADPH binding pocket) ([fig. 4](#pone-0017486-g004){ref-type="fig"}). ::: {#pone-0017486-g004 .fig} 10.1371/journal.pone.0017486.g004 Figure 4 ::: {.caption} ###### Amino acid sequence alignment of selected catalases. Sequences were retrieved from GenBank (GB), Protein Data Bank (PDB) or from Peroxibase (PB). The listed proteins are respectively from *Lutzomyia longipalpis* (GB:ABV60342.1), *Aedes aegypti* (PB:5267), *Anopheles gambiae* (PB:5269), *Bombyx mori* (PB:5266), *Drosophila pseudoobscura* (PB:5273), *Haemonchus contortus* (PB:5270), *D. melanogaster* (GB:NP\_536731.1), *Glossina morsitans morsitans* (GB:ADD20421.1), *Culex quinquefasciatus* (GB:XP\_001848573.1), *Penaeus vannamei* (PB:5278), *Saccharomyces cerevisiae* (PDB:1A4E), *Bos taurus* (PDB:8CAT), *Pseudomonas syringae* (PDB:1M7S). Conserved residues in catalases are with black background, consensus alternatives are shaded. The symbols ▾, +, and \* mark catalytic, heme binding and NADPH binding residues, respectively. The symbol \# mark residues that define heme orientation. All sequences are from clade 3 of monofunctional catalases, with the exception of Psyr, which is a clade 2 enzyme. In catalases from clade 2 (Psyr numbering), heme orientation (His-IV) is defined by residues 301 (never Leu) and 350 (frequently Leu). In catalases from clade 3, these positions are commonly occupied by Leu and non-Leucine residues, respectively. NADPH binding catalases have the signature (Btau numbering) His 193, Arg 202, Val 301 and His 304, which is not present in catalases from clades 1 (not shown) and 2 (Psyr). Insect catalases share some of the NADPH binding residues, but not all. However, catalytic residues and heme binding residues are fully conserved in all sequences. :::  ::: Catalase gene RNAi mediated depletion leads to a decrease in sand fly fecundity {#s2d} ------------------------------------------------------------------------------- The gene sequence of *Lu. longipalpis* catalase was obtained from a cDNA library constructed from sand fly whole bodies (NSFM-142e04.q1k [@pone.0017486-Dillon1]) and aligned with a previous described catalase obtained from *Lu. longipalpis* midguts (GenBank Accession number: EU124624.1) , showing a high level of identity (99%) and similarity (99%) ([fig. S1](#pone.0017486.s001){ref-type="supplementary-material"}). As any other sequence from *Lu. longipalpis* was identified as putative catalase, either in the whole body or midgut cDNA libraries, this gene is probably a single copy gene in *Lu. longipalpis*. To confirm the role of endogenous ROS-scavenging in fecundity catalase was depleted using RNAi. Flies injected with 144 ng of dsRNA for catalase (dsCAT) showed a dramatic decrease in oocyte number dissected 48 hours after blood feeding ([fig. 5A](#pone-0017486-g005){ref-type="fig"}; P\<0.005, ANOVA) compared to sand flies injected with a non-related dsRNA (dsGFP) and uninjected sand flies. A change in appearance of ovaries was observed during dissections with matured ovaries, in sand flies injected with dsRNA for catalase, appearing underdeveloped in comparison to both mock-injected and uninjected controls ([fig. 5B](#pone-0017486-g005){ref-type="fig"}). A dsRNA-mediated significant reduction in catalase expression in whole flies was observed by RT-PCR ([fig. 5C](#pone-0017486-g005){ref-type="fig"}), with no effects on the catalase expression in the midgut (data not shown). These results confirm that endogenous ROS-scavenging in developing oocytes plays a major role in female *Lu. longipalpis* fecundity. ::: {#pone-0017486-g005 .fig} 10.1371/journal.pone.0017486.g005 Figure 5 ::: {.caption} ###### RNAi-mediated depletion of catalase LlonKat1 in female *Lu. Longipalpis* and its effect on fecundity. \(A) Average number of developing oocytes dissected 48 hours days after blood meal ± SEM (combined samples derived from 2 independent experiments). Asterisk indicates statistical significance at P\<0.005(ANOVA). (B) Relative development of female *Lu. longipalpis* ovaries observed upon catalase gene knockdown by RNAi, in comparison to mock-injected and uninjected control sand flies. Bar = 1 mm. (C) Relative expression of catalase LlongKat1 mRNA in whole fly homogenates from dsRNA-injected catalase knock-down sand flies. Bar charts represent mean ± SEM of combined samples from at least 2 independent experiments. Asterisk indicates statistical difference at P\<0.05 (ANOVA). :::  ::: Effect of ROS-scavenging in the survival of sand flies {#s2e} ------------------------------------------------------ To evaluate the role of exogenous ROS scavenging in survival, female *Lu. longipalpis* were fed with an antioxidant-supplemented sugar meal upon emergence. Mortality was recorded from day 1 PE up to day 7 PE. Survival curves depict an increase in mortality due to exogenous ROS-scavenging by an exogenous antioxidant ([fig. 6A](#pone-0017486-g006){ref-type="fig"}). In order to assess whether the higher mortality rate was related to an effect on sand fly immune homeostasis , phenoloxidase (PO) activity was measured in control and antioxidant-supplemented females. Spontaneous PO is defined as the activity measured upon reaction with 3,4 dihydroxy-DL-phenylalanine (DOPA), and corresponds to the enzyme that is already activated in physiological conditions and total activity was the activity observed after *in vitro* activation of the enzyme, by preincubating the sample with bovine trypsin. Sand flies fed on ascorbic acid-supplemented sucrose showed a significant increase in spontaneous PO ([fig. 6B](#pone-0017486-g006){ref-type="fig"}; P\<0.05 , t-test) but no difference in total PO activity. ::: {#pone-0017486-g006 .fig} 10.1371/journal.pone.0017486.g006 Figure 6 ::: {.caption} ###### Effect of dietary supplementation of ascorbic acid on mortality of sugar fed *Lu. longipalpis*. (**A**) Female sand flies were offered a 70% sucrose solution supplemented with 20 mM ascorbic acid or a non-supplemented sucrose solution. Experimental flies (sucrose +20 mM ascorbic acid) exhibited a significantly lower survival rate compared to control flies, (*p*\<0.001, Kaplan-Meier, Log Rank χ^2^ test). (**B**) Spontaneous and total phenoloxidase (PO) activity in *Lu. longipalpis* females after 7 days of feeding with 70% sucrose solution or 70% sucrose solution supplemented with 20 mM ascorbic acid. Spontaneous PO activity in ascorbic acid supplemented flies was significantly higher than control flies (P\<0.05 , *T*-test). Results are mean ± SEM from 2 independent experiments with 10 sand flies per experiment. :::  ::: To further investigate if ROS-scavenging was implicated in increased mortality, catalase LlonKat1 was depleted via RNAi injection in female *Lu. longipalpis* and mortality was recorded from day 1 PE up to day 7 PE. Mortality rates was higher in knocked down (dsCAT) sand flies ([fig. 7](#pone-0017486-g007){ref-type="fig"}), compared to flies injected with a non-related dsRNA (dsGFP) and non-injected flies. These results show that ROS-scavenging by either endogenous or exogenous antioxidants play an important role in female *Lu. longipalpis* survival. ::: {#pone-0017486-g007 .fig} 10.1371/journal.pone.0017486.g007 Figure 7 ::: {.caption} ###### Survival in female *Lu. Longipalpis* after RNAi-mediated depletion of catalase LlonKat1. Experimental group (dsCAT) exhibits a significantly lower survival rate compared to both dsGFP and pricked control groups, (P\<0.0001, Kaplan-Meier, Log Rank χ^2^ test). Results represent mean ± SEM of 3 independent biological replicates. :::  ::: Discussion {#s3} ========== The present study suggests for the first time that catalase-mediated ROS scavenging has a significant impact on female *Lu. longipalpis* fecundity and survival. Female *Lu. longipalpis* from different age groups showed differences in developing oocytes numbers, with the oldest (9 days PE) presenting the lowest number of oocytes ([fig. 1](#pone-0017486-g001){ref-type="fig"}). The age-related loss of fecundity could be reversed with dietary supplementation of a potent exogenous ROS-scavenger ([fig. 2](#pone-0017486-g002){ref-type="fig"}). This underlines the importance of catalase in the reproductive success of blood sucking dipterans. Evidence from other dipterans show that aging results in increase of oxidative stress and loss of enzymatic antioxidant efficiency [@pone.0017486-Das1], [@pone.0017486-Sohal1], [@pone.0017486-Yan1]. Moreover, inactivation or silencing of catalase in *Drosophila melanogaster* [@pone.0017486-Mackay1], *Musca domesti*ca [@pone.0017486-Allen1], *Rhodnius prolixus* [@pone.0017486-Paes1]and *Anopheles gambiae* [@pone.0017486-MagalhaesDEB1] led to increased mortality due to increase in ROS levels. It is likely that accumulation of ROS in older flies could account for the decrease of female sand fly fecundity due to an increase in oxidative stress, loss of antioxidant enzymatic efficiency or both. In *An gambiae*, fecundity of female mosquitoes declined with age, with reduction of number of eggs oviposited and number of larvae hatched per female [@pone.0017486-DeJong1]. We did not measure differences in fecundity in terms of larval development but it is likely that the age-related differences in fertility would have resulted in less viable larvae being produced from older flies, as they would be presumably exposed for a longer periods to oxidative damage. Catalase enzymatic activity as well as catalase LlonKat1 mRNA relative expression increased in the ovaries of older female sand flies (6 days PE) after the blood feeding ([figs. 3a and b](#pone-0017486-g003){ref-type="fig"}). Protein expression and accumulation increased upon blood feeding in maturing ovaries of mosquitoes due to nutrient allocation for egg production [@pone.0017486-Wheeler1], [@pone.0017486-Ahmed1]. It has been shown in different insect species that antioxidant activity increases in the ovaries to protect the embryo from oxidative damage [@pone.0017486-Logullo1], [@pone.0017486-Freitas1]. It is conceivable that such accumulation of catalase in sand fly ovaries also provides the means to protect developing eggs from oxidative damage. Additional support for this hypothesis was given by the dramatic decrease in developing oocyte numbers upon successful catalase gene depletion by RNAi in female sand flies ([fig. 5A and b](#pone-0017486-g005){ref-type="fig"}). Interestingly, oral delivery of ascorbic acid seemed to stimulate catalase LlonKat1 mRNA expression in older flies to levels similar to that of younger flies ([fig. 3C](#pone-0017486-g003){ref-type="fig"}). It has been shown that age-related accumulation of ROS/oxidative stress leads to loss of efficiency in cellular processes [@pone.0017486-Das1], [@pone.0017486-Sohal1], [@pone.0017486-Ferguson1], [@pone.0017486-Yan1], [@pone.0017486-Sohal2], therefore it is possible that ROS-scavenging by an exogenous antioxidant slowed or lowered such deleterious effects in either catalase LlonKcat1 mRNA or in other molecules involved in its upregulation. On the other hand, it has been shown that ascorbate is a potent inhibitor of catalase [@pone.0017486-Orr1], the inhibition being independent of substrate concentration and pH and strongly influenced by temperature. Furthermore, catalase incubation with ascorbate leads to degradative changes to the catalase molecule [@pone.0017486-Orr2]. In our experiments, the increase in catalase gene expression might reflect a compensation response to replenish normal catalase levels in the sand fly body after catalase was degraded, by an unknown mechanism, during ascorbic acid supplementation with the sugar meal. Catalase LlonKcat1 does not have a signal peptide or targeting sequences to mitochondria or peroxisomes. These features suggest a cytosolic location but this needs confirmation. Based on the identities with other catalases retrieved from Peroxibase [@pone.0017486-Koua1], LlonKat1 seems to belong to the monofunctional clade 3 of catalases, which includes sequences from bacteria, archaebacteria, protists, fungi, plants and animals. These enzymes have small subunits with molecular mass ranging from 43--75 kDa [@pone.0017486-Zamocky1], which is consistent with LlonKat1 monomer predicted molecular mass (57.7 kDa). All conserved catalytic and heme binding residues are present in LlonKat1 sequence, suggesting a full catalytic activity, and the presence of residues Leu298, Met349 indicate that His70 is above the ring III of the heme molecule (His-III orientation), as seen in other clade 3 catalases [@pone.0017486-Chelikani1]. ROS-scavenging by dietary supplementation of ascorbic acid ([fig. 6a](#pone-0017486-g006){ref-type="fig"}) led to a reduction in sand fly survival. When antioxidants were provided to a susceptible strain of *Anopheles gambiae* to *Plasmodium* infection, a similar but more drastic effect was observed with female mosquitoes [@pone.0017486-DeJong1]. Magwere *et al.* [@pone.0017486-Magwere1] observed that antioxidant supplementation did not extend the lifespan of wild type *Drosophila*. Similarly, Bayne *et al.* [@pone.0017486-Bayne1] showed that overexpression of MnSOD and catalase, despite protecting *Drosophila* from oxidative stress, were detrimental for lifespan and physical fitness of the insects. Kang *et al.* [@pone.0017486-Kang1] observed a reduction in the lifespan of *Anopheles stephensi* when the mosquitoes were bloodfed with the antioxidant MnTBAP in comparison with the buffer control. It has been hypothesized that a minimal level of ROS might be required to maintain the balance of the gut microbiota and that a baseline level of ROS activity might be crucial for basic midgut physiology. Previous studies done with other dipteran species had showed that ROS release constitutes a first line of defence against pathogens in the midgut [@pone.0017486-Hoffmann1]. Experiments in *D. melanogaster* have demonstrated the existence of a midgut-specific active ROS releasing system against orally delivered bacteria [@pone.0017486-Ha1]. In the present study, higher activities of spontaneous PO were recorded and this might be due to an increase in microbial infection associated with sand flies that fed on an ascorbic acid-supplemented sugar meal. In insects, PPO activation is often related to bacterial or fungal infections [@pone.0017486-Hoffmann1]. Since only the soluble form of PO was measured (see Material and Methods), it is more likely that the activity was related to the immune response rather than to the melanisation of the adult cuticle or egg shell. PO has already been described in gut tissues or adhered hemocytes in other dipterans [@pone.0017486-Gillespie1]. It is possible that mortality in our experimental group fed with sucrose supplemented with ascorbic acid may be due to a decrease in ROS production inside the midgut and that ROS activity, similar to the events in certain strains of mosquitoes, may play a role in sand fly immunity towards opportunistic microbes or be involved in important cellular signalling pathways [@pone.0017486-Morey1], [@pone.0017486-Kamata1]. There is evidence of other antioxidant enzymes with catalase-like functions found in the sand fly midgut, such as peroxiredoxins [@pone.0017486-Jochim1], [@pone.0017486-Dillon1]. These are a family of thioredoxin-dependent peroxidases, found in several insect species [@pone.0017486-Wang1], [@pone.0017486-Hu1], [@pone.0017486-Radyuk1], [@pone.0017486-Kim1], that function as ROS-scavengers as well as other cellular processes. However their efficiency in converting H~2~O~2~ was found to be significantly lower compared to catalase [@pone.0017486-Wood1]. We are currently investigating the role of these antioxidant enzymes in ROS detoxification and fecundity and survival of female sand flies, as well as studies to confirm if proliferation of bacteria due to ROS reduction could account for the differences in sand fly survival. Recent studies on transgenic *Anopheles stephensi* (the leading malaria vector in India and parts of Asia and the Middle East) overexpressing the protein kinase AKT gene increased the insulin signalling in the mosquito midgut, significantly reducing mosquito lifespan and inhibiting *P. falciparum* development [@pone.0017486-CorbyHarris1]. The role of genes involved in stress responses in *Plasmodium* survival within the mosquito midgut was investigated by Jaramillo-Gutierrez et al. [@pone.0017486-JaramilloGutierrez1]. RNAi gene knockdown of the OXR1 gene (oxidation resistance gene) in *Anopheles gambiae* showed that this gene regulates the basal levels of catalase and glutathione peroxidase expression and that OXR1 gene knockdown decreased *Plasmodium berghei* oocyst formation. The finding of a *Lu. longipalpis* OXR1 gene homologue (unpublished) will shed some light into the regulation of ROS production within the sand fly gut and will also help to understand how ROS production impacts *Leishmania* development in the sand fly midgut. Current vector control strategies rely on spraying of residual insecticides to control vector population. Insect transgenesis and paratransgenesis are novel strategies that aim at reducing insect vectorial capacity by using genetic manipulation of disease vectors, rendering them incapable or less efficient to transmit a given pathogen [@pone.0017486-CoutinhoAbreu1] or even reducing the longevity and fecundity of a given insect vector. This study shows that catalase is a key gene in determining survival and fecundity of phlebotomine sand flies and future developments may warrant this gene being included as a potential target to reduce female sand fly fitness and reproductive capacity in the field. Materials and Methods {#s4} ===================== Insects {#s4a} ------- All experiments were carried out using insectary-reared *Lu. longipalpis* from a colony first started with individuals caught in Jacobina, Brazil. Insects were kept under standard laboratory conditions [@pone.0017486-Modi1]. Insects were fed with 70% w/v sucrose solution in cotton wool (unless stated differently in experiments), kept under a photoperiod of 8 hours light/16 hours darkness, temperature of 27°C (±2) and a relative humidity of \>80% inside the rearing cages. Rabbit blood feeding was via a Hemotek membrane feeder (Discovery Workshops, UK) at 37°C. All procedures involving animals were performed in accordance with UK Government (Home Office) and EC regulations. Fecundity assays {#s4b} ---------------- Female *Lu. longipalpis* were allowed to mate under regular rearing conditions and fed with rabbit blood at three, six and nine days post-emergence (DPE). A batch of \>500 flies was released into a large (20 m^3^) rearing cage and groups of ∼100 individuals were transferred to medium sized cages (5 m^3^) at 3, 6 and 9 DPE and blood-fed as above. Fifteen fully-engorged females were then transferred to a new medium rearing cage. Insects were dissected five days later to count developing oocytes. ROS-scavenger feeding {#s4c} --------------------- Fecundity assays were carried out as described above with female *Lu. longipalpis* fed on a 70% sucrose solution supplemented with 20 mM ascorbic acid and blood-fed at 9 DPE. Supplemented sucrose-meal was freshly changed daily and continued after blood-feeding. A 9 DPE control group was reared under the same conditions but fed with a 70% sucrose solution. Only fully engorged insects from both groups were selected for the experiments. Ovarian Catalase Activity {#s4d} ------------------------- Ovaries were collected from 5 female sand flies at 24 and 48 hrs post blood feeding (PBF). Samples were homogenised in 50 ul of 0.15 M NaCl solution, kept on ice and transferred to a −80°C freezer until needed. Before assays, samples were centrifuged at 5000 RPM for 2 minutes and 1 µl of the supernatant was diluted in 24.9 µl of 0.15 M NaCl solution. Catalase activity was determined using Amplex Red Catalase Assay Kit (Invitrogen Ltd) following the manufacture\'s protocol. Enzyme-specific activities were expressed as units/mg of protein. One unit of catalase activity was defined as 1 µM of H~2~O~2~ consumed per minute. All assays were carried out in triplicate. Fluorescence was measured using a Varioskan fluorescence spectrometer (Thermo Electron) with an excitation wavelength of 560 nm and an emission wavelength of 590 nm. Ovarian catalase activity was normalised using the total amount of protein in the whole body (minus dissected ovaries) using the BIORAD® Protein assay reagent following the manufacturer\'s protocol and using bovine serum protein as standard. Endpoint absorbance was measured at 595 nm in a 96 well plate with a microplate reader (VersaMax Microplate Reader, Molecular Devices Inc.). Ovarian Catalase Expression {#s4e} --------------------------- Six DPE sand flies were blood fed and ovaries from 10 sand flies (two pools of 5 flies) were dissected at 12, 24 and 48 hours PBF, homogenised in 50 µl of TRI Reagent® (Ambion, Austin, TX) and kept at −80°C until needed. RNA was extracted following the manufacturer\'s protocol. Total RNA was quantified using a Nanodrop®(NanoDrop Technologies, Wilmington, USA) and normalised to 10 ng/µl. RT-PCR was carried out with SuperScript® III One-Step RT-PCR System with Platinum® Taq DNA Polymerase Kit (Invitrogen, San Diego, CA) performing 19 cycles and following the manufacture\'s protocol (primers listed on [table 1](#pone-0017486-t001){ref-type="table"}). Relative expression of catalase was normalised using a housekeeping gene (AM088777, 60S ribosomal protein L3). RT-PCR products were analysed by 1.5% agarose/ethidium bromide gel electrophoresis and reduction in catalase expression was determined by densitometric measurement of bands using the softwares GeneSnap/GeneTools (Syngene, UK). ::: {#pone-0017486-t001 .table-wrap} 10.1371/journal.pone.0017486.t001 Table 1 ::: {.caption} ###### Oligonucleotides for dsRNA synthesis and Reverse Transcriptase PCR. ::: {#pone-0017486-t001-1} Oligonucleotide 5′-3′sequence Size (bp) ------------------------ -------------------------------------------------- ----------- **dsCAT484 Forward** **TAATACGACTCACTATAGGG**TGTTGCAGGGACGTCTCTTTGCC 524 **dsCAT484 reverse** **TAATACGACTCACTATAGGG**AGGTTGGAGCACTTCTTGCGTTCG **dsGFP Forward** **TAATACGACTCACTATAGGG**ACGTAAACGGCCACAAGTTC 693 **dsGFP Reverse** **TAATACGACTCACTATAGGG**CTTGTACAGCTCGTCCATGCC **RT CAT484 Forward** TGTTGCAGGGACGTCTCTTTGCC 484 **RT CAT484 Reverse** AGGTTGGAGCACTTCTTGCGTTCG **RT Ribo60S Forward** TCTCATCGGAAGTTTTCTGC 850 **RT Ribo60 Reverse** GGCTTGTGACACCCTTGAAT ::: Age-related expression of ovarian catalase {#s4f} ------------------------------------------ To measure catalase LlongKat1 mRNA expression levels in different age groups, 3, 6 and 9 DPE sand flies were blood fed and ovaries from 10 sand flies (two pools of 5 flies) were dissected at 48 hours PBF. Additionally, to evaluate the effect of feeding a ROS-scavenger in age-related expression of ovarian catalase, a group of 9 days old sand flies was fed with ascorbic acid-supplemented sucrose solution as described above, blood fed and dissected at 48 hours. RNA was extracted and checked for catalase relative expression as above. RNAi-mediated catalase knockdown {#s4g} -------------------------------- Sense and anti-sense catalase-specific primers flanked by the T7 promoter site ([Table 1](#pone-0017486-t001){ref-type="table"}) PCR amplified a 484 bp product from a plasmid obtained from a whole body Lu. longipalpis normalised cDNA library [@pone.0017486-Dillon1] that was used as template for double-stranded RNA synthesis dsRNA. Transcription reactions and column purification were carried out using the Megascript RNAi Kit (Ambion®) following the manufacturer\'s protocol. dsRNA purity was assessed by 1.5% agarose/ethidium bromide gel electrophoresis and dsRNA was quantitated using a Nanodrop ND-1000 Spectrophotometer (LabTech, UK). dsRNA was eluted with nuclease-free water at 65°C, concentrated to 4.5 µg/µL with a Christ® RVC 2--25 rotational vacuum concentrator and stored at −80°C until needed. Enhanced Green Fluorescent protein (eGFP) dsRNA was produced from a 653 bp amplicon of the pEGFP-N1 expression plasmid (Clontech) and used as a 'mock' injected control. RNAi was achieved by dsRNA injections as previously described [@pone.0017486-SantAnna1]. After injections, sand flies were transferred to cages and kept with access to 70% sucrose solution *ad libitum*. Developing oocytes were dissected and counted 48 hours after blood feeding. Non-injected flies of the same age and kept under the same conditions were used as second control. Three pools of three whole sand flies were collected from each group to evaluate knockdown by RT-PCR. Survival assays {#s4h} --------------- To assess sand fly survival mediated by ROS-scavenging related to catalase activity, RNAi-mediated catalase knock down was carried out in a group of 50 sand flies. Flies were injected with dsRNA for catalase (dsCAT) as described above. To exclude wound-related mortality, all dead flies at 24 hrs post-injections were removed and were not included in the experiment. Dead sand flies were counted and removed from the cage daily from day 2 to 7 after injection. Flies injected with dsRNA for GFP (dsGFP) and a needle-pricked group were used as controls. To assess exogenous ROS-scavenging related survival, 50 female *Lu. longipalpis* were collected upon emergence and sugar fed on a 70% w/v sucrose solution supplemented with 20 mM ascorbic acid. Dead sand flies were counted and removed from the cage every day until day seven. A group of sand flies fed with 70% sucrose was used as a control. Phenoloxidase assays {#s4i} -------------------- Phenoloxidase activity was determined by measuring the production of dopachrome from 3,4 dihydroxy-DL-phenylalanine (DOPA) [@pone.0017486-Pomerantz1], [@pone.0017486-Genta1]. Briefly, single flies were homogenized in 60 µL of PBS and centrifuged at 25,000 g for 5 min at 4°C to recover the soluble fraction. 20 µL of supernatant was mixed with 10 µL of PBS (spontaneous PO) or trypsin solution (for total PO activity; 1 mg/mL in PBS, FLUKA cat. no. 93614), incubated for 20 min at 37°C followed by the addition of 20 µL of a saturated solution of DOPA (4 mg/mL in PBS) and absorbance (490 nm) measured by kinetic assay for 1 h at 5 minutes intervals in a microplate reader at 30°C. The PO activity was measured to ensure that activity was proportional to protein concentration and incubation time. Independent experiments showed that the PO activity was stable in the conditions above. Controls with no enzyme or no substrate were included. One unit of enzyme (U) is defined as the amount that produces 0.001 unit of absorbance/min. Sequence analysis {#s4j} ----------------- The coding sequence of LlonKat1 was analyzed using the algorithms pI/Mw tool [@pone.0017486-Walker1] , signal IP [@pone.0017486-Emanuelsson1], PTS1 Predictor [@pone.0017486-Neuberger1], PeroxiP [@pone.0017486-Emanuelsson2], TargetP [@pone.0017486-Emanuelsson1] based at the EXPASY Proteomics Server (<http://expasy.org/>). Selected amino acid sequences of catalases were aligned with catalase LlonKat1 using the ClustalW Multiple Alignment tool in BioEdit Sequence Alignment Editor (<http://www.mbio.ncsu.edu/BioEdit/BioEdit.html>). Alignment was generated using Boxshade (<http://www.ch.embnet.org/software/BOX_form.html>). Statistical analysis {#s4k} -------------------- Comparisons between means of two independent groups were carried put using a pair-wise t-test. Multiple comparisons were done by one-way ANOVA. Survival curves were analyzed with the Kaplan-Meier Log Rank χ^2^ test. Significance was considered when P\<0.05. All data were analysed with the use of the SPSS Data Editor software (version 17.0, SPSS Inc). Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Structure-based alignment of the aminoacid sequence of** ***Lutzomyia longipalpis*** **catalase, translated from a whole body (GeneDB NSFM-142e04.q1k** [@pone.0017486-Dillon1] **) and a midgut-specific (GenBank Accession number: EU124624.1** [@pone.0017486-Jochim1] **) cDNA library.** Sequences show a 99% identity and a 99% similarity. \> Represents the targeted region for dsRNA-mediated gene silencing. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Davina Moor for her technical assistance. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded by Consejo Nacional de Ciencia y Tecnología ([www.conacyt.mx](http://www.conacyt.mx)) Fellow ref 207646; The Leverhulme Trust ([www.leverhulme.co.uk](http://www.leverhulme.co.uk)) ref F/00 808/C. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: HDA MRVS FAG RJD. Performed the experiments: HDA RM FAG. Analyzed the data: HDA RM FAG. Contributed reagents/materials/analysis tools: RJD. Wrote the paper: HDA MRVS FAG RJD.

PubMed Central

2024-06-05T04:04:19.165663

2011-3-9

PMC3052319

Introduction {#s1} ============ Systemic sclerosis (SSc) is a connective-tissue disease characterized by excessive collagen deposition in the dermis and internal organs, and by vascular hyper-reactivity and obliterative microvascular phenomena [@pone.0017551-Tamby1]. SSc is responsible for diminished life expectancy, related to skin extent and visceral involvement [@pone.0017551-Steen1]. SSc is also responsible for tendon, joint, and vessel damage, leading to disability, handicap, and impaired health-related quality of life (HRQoL) [@pone.0017551-Poole1]. In addition, psychiatric symptoms, including anxiety and depression, have been reported as a consequence of disease chronicity in SSc patients, with a prevalence of depressive symptoms ranging from 18% to 65% [@pone.0017551-Angelopoulos1], [@pone.0017551-BenrudLarson1], [@pone.0017551-Haythornthwaite1], [@pone.0017551-Hyphantis1], [@pone.0017551-Legendre1], [@pone.0017551-Matsuura1], [@pone.0017551-Roca1], [@pone.0017551-Straszecka1], [@pone.0017551-Thombs1]. Consistent with other auto-immune diseases, SSc is predominant among females, with a ratio of females to males of 1∶1 to 14∶1 [@pone.0017551-Chifflot1], along with gender differences in disease activity and incidence. Such differences have been explained by genetic and hormonal factors and lifestyle [@pone.0017551-Oliver1], [@pone.0017551-Simeon1]. Male gender is usually considered a factor of poor prognosis in SSc [@pone.0017551-PetersGolden1], [@pone.0017551-Wynn1]. A cohort of 91 SSc patients (10% males) from Spain revealed clinical and immunological differences between the genders; arthralgias were more often encountered in females, whereas myositis and nucleolar antinuclear antibodies were more frequent in males [@pone.0017551-Simeon1]. More recently, as compared with female SSc patients, males were found to more often exhibit renal failure, increased blood pressure, arrythmia and inflammatory myopathy and less often sicca syndrome and anti-centromere antibodies. Causes of death and mortality also differed between the sexes [@pone.0017551-Joven1]. In a large European cohort of 1180 patients with early SSc (19% males), features of diffuse disease were significantly more frequent in males [@pone.0017551-Carreira1]. Recently, Hudson *et al.* found that the time to diagnosis was longer for women than men after the onset of Raynaud\'s phenomenon, and suggested that there may be possible biologic differences in the progression of disease or in the health care trajectories of men and women with early SSc [@pone.0017551-Hudson1]. Although gender differences in disease-related clinical manifestations are well established, few studies have compared HRQoL, disability, and psychiatric symptoms between male and female patients with SSc. In the present study, we aimed to assess the association of gender with clinical expression, HRQoL, disability, and self-reported symptoms of depression and anxiety in patients with SSc. Methods {#s2} ======= Study design {#s2a} ------------ We performed a cross-sectional survey of 381 patients. Patients with SSc were prospectively included during 7 consecutive annual meetings of the French SSc patients\' association, the "Association des Sclérodermiques de France" (ASF), between 2003 and 2009, or during their hospitalization in the internal medicine departments of Cochin (between January 2006 and June 2009) or Claude Huriez (between January and June 2009) hospitals. Since some patients were evaluated during several ASF annual meetings, only the most recent assessment of each patient was considered. Patients had to complete self-administered questionnaires first and then to undergo an interview with a physician to check for unanswered question, fully complete questionnaires, and gather clinical data. Patients {#s2b} -------- To be eligible for the study, patients had to fulfil the American College of Rheumatology [@pone.0017551-Preliminary1] and/or the Leroy and Medsger [@pone.0017551-LeRoy1] criteria for SSc. Patients from the ASF were assessed within 48 hr during spring (temperature 20°C). Parameters recorded were age; sex; age at disease onset; disease duration; body mass index (BMI); disease subset (limited SSc \[lSSc\], limited cutaneous SSc \[lcSSc\] or diffuse cutaneous SSc \[dcSSc\]); mouth opening (inter-incisor distance measured in millimetres); skin involvement; telangiectasia; Raynaud\'s phenomenon; pitting scars; digital ulcers; calcinosis; gastrointestinal tract, joint and/or muscle involvement; dyspnoea (assessed by the New York Heart Association \[NYHA\] 4-point scale); ILD; echocardiography systolic pulmonary artery pressure \[PAP\]\>35 mmHg); and renal crisis. History of esophagus, gastrointestinal, joint, muscle and/or heart involvement; ILD; echocardiography PAP\>35 mmHg; and renal crisis was obtained from detailed clinical charts for hospitalized patients and self-reports for ASF members. Health status {#s2c} ------------- Health status was assessed by the KPS score, the scale ranging from 0 (dead) to 100 (normal no complaints; no evidence of disease) [@pone.0017551-Mouthon1]. Originally developed for cancer patients, because it strongly predicted cancer outcome [@pone.0017551-Andrews1], [@pone.0017551-Evans1], the KPS score has been shown to provide clinical estimates of patient\'s physical state, performance, and prognosis and to be associated with social status in patients with SSc [@pone.0017551-Mouthon1], [@pone.0017551-Nguyen1]. Health-related quality of life {#s2d} ------------------------------ HRQol was assessed by the French version of the Medical Outcomes Study 36-Item Short Form Health Survey (SF-36) [@pone.0017551-Ware1], a self-administered questionnaire covering 8 areas: physical function, physical role, bodily pain, general health, vitality, social function, emotional role, and mental health. For each area, scores range from 0 (poorer health status) to 100 (better health status). Scores can also be summarized in 2 global scores: physical component score (PCS) and mental component score (MCS). Disability {#s2e} ---------- ### Global disability {#s2e1} Global disability was assessed by use of the standard disability index of the Health Assessment Questionnaire (HAQ-DI) that contains 20 items (each scored ranging from 0 \[no disability\] to 3 \[maximal disability\]), divided into 8 domains [@pone.0017551-Merkel1]. ### Patients\' perceived disability {#s2e2} Patients\' perceived disability was assessed by the McMaster Toronto Arthritis Patient Preference Disability Questionnaire (MACTAR) [@pone.0017551-Tugwell1]. Patients were asked to select the 3 situations among activities of daily living (ADL) that caused them maximal trouble [@pone.0017551-Mouthon1]. Each item is scored on an 11-point quantitative scale (range 0--10). The global score ranges from 0 (no disability) to 30 (maximal disability). This score has been validated in SSc [@pone.0017551-Mouthon1], [@pone.0017551-Nguyen2]. ### Hand disability {#s2e3} Hand disability was assessed by the Cochin Hand Function Scale (CHFS) [@pone.0017551-Duruoz1], a questionnaire administered by the physician that contains 18 items related to ADL. Each question is scored on a scale of 0 (performed without difficulty) to 5 (impossible to do). The total score is obtained by adding the scores of all items (range 0--90). This questionnaire has been validated in SSc [@pone.0017551-Rannou1]. ### Mouth disability {#s2e4} Mouth disability was assessed by the Mouth Handicap In Systemic Sclerosis (MHISS) scale, a questionnaire administered by the physician that contains 12 items concerning difficulties in performing ADL. Each question is scored on a scale of 0 (never) to 4 (always) [@pone.0017551-Mouthon2]. The total score is obtained by adding the scores of all items (range 0--48). Anxiety and depression {#s2f} ---------------------- Self-reported anxiety and depression symptoms were assessed by the Hospital Anxiety and Depression scale (HADS). This scale has 7 questions for the anxiety dimension (HADa) and 7 for the depression dimension (HADd) [@pone.0017551-Zigmond1]. Each item is scored on a scale of 0 to 3, the total score ranging from 0 (no depression, no anxiety) to 21 (maximal depression, maximal anxiety). Scores of 0--7 in subscales are considered normal, 8--10 borderline and ≥11 pathologic cases [@pone.0017551-Bjelland1]. The definition of clinical anxiety and/or depression was based on the HADS score cutoff ≥8 found to be relevant in patients with autoimmune diseases [@pone.0017551-Honarmand1]. Aesthetic impairment {#s2g} -------------------- Aesthetic impairment was assessed on an 11-point quantitative scale, the total score ranging from 0 (no aesthetic impairment) to 10 (maximal aesthetic impairment). Statistical analysis {#s2h} -------------------- Data analysis involved use of Systat 9 (SPSS Inc., Chicago, IL, USA). Quantitative variables were described with means ± standard deviations (SD) and qualitative variables with frequencies and percentages. For bivariate analysis, parametric tests were used since all parameters met criteria for normal distribution. Comparisons between male and female groups involved the Pearson chi-square test for qualitative variables and two-sample *t* test for quantitative data. Bonferroni adjustment was used for multiple comparisons (43 comparisons); therefore a *p* value less than 0.001 was considered statistically significant. Multivariate analysis was used to determine the association of gender and SSc-related variables. Backward stepwise regression all-inclusive analysis was run, including all dependent variables, with values of 0.20 to enter and 0.10 to stay in the model. Adjustment for age and type of recruitment from either patient association or hospitalization was performed. Odds ratios (OR) and 95% confidence intervals were calculated. Ethics statement {#s2i} ---------------- This survey was conducted in compliance with the protocol of Good Clinical Practices and Declaration of Helsinki principles. Patients gave their consent to participate after being orally informed about the study protocol. In accordance with European regulation (Directive 2001/20/EC of the European Parliament and of the Council of 4 April 2001 on the approximation of the laws, regulations and administrative provisions of the Member States relating to the implementation of good clinical practice in the conduct of clinical trials on medicinal products for human use; Directive 95/46/EC of the European Parliament and of the Council of 24 October 1995 on the protection of individuals with regard to the processing of personal data and on the free movement of such data), French observational studies from data obtained without any additional therapy or monitoring procedure, do not need formal approval of an Institutional Review Board or an Independent Ethics Committee, and a formal written consent from the patients is not required for this kind of project. Results {#s3} ======= Demographic and clinical data {#s3a} ----------------------------- Overall, 381 patients were included. One-hundred-and-forty-three of them were recruited during their hospitalization in the internal medicine departments of Cochin (n = 127) or Claude Huriez (n = 16) hospitals, and the remaining 238 patients were recruited during ASF annual meetings from 2003 to 2009. The proportion of patients from the ASF who agreed among those who were asked to participate were 51 among 80 (63.8%) (44 females) in 2003, 50 among 80 (62.5%) (44 females) in 2004, 71 among 98 (72.4%) (59 females) in 2005, 70 among 95 (73.7%) (55 females) in 2006, 70 among 101 (69.3%) (55 females) in 2007, 86 among 130 (66.1%) (74 females) in 2008 and 2009 alltogether. Of the 381 patients, 62 were males (16.4%), with a female to male ratio of 5∶1. The mean age at the time of evaluation was 55.9 (13.3) years, and mean disease duration was 9.5 (7.8) years. A total of 149 (40.4%) patients had dcSSc, 187 (50.7%) had lcSSc, and 34 (9.2%) had lSSc. The mean KPS was 77.6 \[11.7\] (range 50--100) ([Table 1](#pone-0017551-t001){ref-type="table"}). ::: {#pone-0017551-t001 .table-wrap} 10.1371/journal.pone.0017551.t001 Table 1 ::: {.caption} ###### Demographic and clinical characteristics of patients with SSc[\*](#nt101){ref-type="table-fn"}. ::: {#pone-0017551-t001-1} ----------------------------------------- ---------------- Age, years, mean (SD) 55.9 (13.3) Age at disease onset, years, mean (SD) 46.2 (12.9) Male sex 62/379 (16.4) Patient association 62/191 (32.5) Disease duration, years, mean (SD) 9.5 (7.8) Body mass index, kg/m^2^, mean (SD) 23.4 (4.5) Diffuse cutaneous SSc 149/369 (40.4) Limited cutaneous SSc 187/369 (50.7) Limited SSc 34/369 (9.2) KPS (0--100), mean (SD) 77.6 (11.7) Inter-incisor distance, mm, mean (SD) 35.9 (9.3) Skin involvement 339/370 (91.6) Telangiectasia 253/347 (72.9) Raynaud\'s phenomenon 369/377 (97.9) Pitting scars 221/376 (58.8) Digital ulcers 170/375 (45.3) Calcinosis 105/312 (33.7) Gastrointestinal tract involvement 304/375 (81.1) Arthralgia 254/375 (67.7) Myalgia 209/375 (55.7) Dyspnea, NYHA classification, mean (SD) 2.1 (0.8) Interstitial lung disease 163/373 (43.7) Echocardiography systolic PAP\>35 mmHg 48/375 (12.8) Scleroderma renal crisis 34/375 (9.1) ----------------------------------------- ---------------- \*Values are number/number of patients for whom the data is available (%), otherwise indicated in parenthesis. KPS: Karnofsky performance status; NYHA: New York Heart Association; PAP: pulmonary artery pressure; SD: standard deviation; SSc: systemic sclerosis. ::: Association of gender with SSc clinical expression {#s3b} -------------------------------------------------- Males and females were comparable in age at the time of evaluation and at disease onset, disease duration and health status as assessed by the KPS score. For other clinical variables, some differences were observed between males and females, without reaching statistical significance. Indeed, BMI was higher in males than females (25.0 \[4.1\] *vs* 23.0 \[4.5\], p = 0.013) ([Table 2](#pone-0017551-t002){ref-type="table"}). DcSSc was more frequent in males (54.8 *vs* 37.2%, p = 0.010), whereas lSSc was more frequent in females (10.7% *vs* 1.6%, p = 0.024). Regarding visceral involvement, males more often exhibited ILD and echocardiography PAP\>35 mmHg than did females (54.8% *vs* 41.2%, p = 0.048; and 24.2% *vs* 10.5%, p = 0.003, respectively), and females more often calcinosis than males (36.0% *vs* 21.4%, p = 0.036). On multivariate logistic regression, gender was significantly associated with BMI (OR 1.12, 95% confidence interval \[CI\] 1.01--1.24) and echocardiography PAP\>35 mmHg (OR 0.23, 95% CI 0.07--0.76). ::: {#pone-0017551-t002 .table-wrap} 10.1371/journal.pone.0017551.t002 Table 2 ::: {.caption} ###### Association of gender with clinical manifestations in SSc[\*](#nt103){ref-type="table-fn"}. ::: {#pone-0017551-t002-2} Malesn = 62 Femalesn = 319 p-value^†^ -------------------------------------------- -------------- ---------------- ------------ Age, years (mean \[SD\]) 55.7 (14.5) 55.9 (13.1) 0.924 Age at disease onset, years (mean \[SD\]) 46.7 (14.4) 46.1 (12.7) 0.749 Disease duration, years (mean \[SD\]) 8.9 (6.8) 9.7 (8.1) 0.577 Body mass index, kg/m^2^ (mean \[SD\]) 25.0 (4.1) 23.0 (4.5) 0.013 Diffuse cutaneous SSc 34/62 (54.8) 115/309 (37.2) 0.010 Limited cutaneous SSc 27/62 (43.5) 160/309 (51.8) 0.237 Limited SSc 1/62 (1.6) 33/309 (10.7) 0.024 KPS (0--100) (mean \[SD\]) 78.0 (11.6) 77.5 (11.7) 0.755 Inter-incisor distance, mm (mean \[SD\]) 37.0 (9.4) 35.7 (9.3) 0.359 Skin involvement 59/62 (95.2) 280/310 (90.3) 0.221 Telangiectasias 46/60 (76.7) 207/289 (71.6) 0.426 Raynaud\'s phenomenon 59/62 (95.2) 310/317 (97.8) 0.237 Pitting scars 41/62 (66.1) 180/316 (57.0) 0.180 Digital ulcers 28/61 (45.9) 142/316 (44.9) 0.890 Calcinosis 12/56 (21.4) 93/258 (36.0) 0.036 Gastrointestinal tract involvement 50/62 (80.6) 254/315 (80.6) 0.999 Arthralgias 39/62 (62.9) 215/315 (68.3) 0.411 Myalgias 32/62 (51.6) 177/315 (56.2) 0.507 Dyspnea, NYHA classification (mean \[SD\]) 2.2 (0.8) 2.0 (0.9) 0.190 Interstitial lung disease 34/62 (54.8) 129/313 (41.2) 0.048 Echocardiography systolic PAP\>35 mmHg 15/62 (24.2) 33/315 (10.5) 0.003 Scleroderma renal crisis 8/62 (12.9) 26/315 (8.3) 0.243 \*Values are number/number of patients for whom the data is available (%), otherwise indicated in parenthesis. SSc: systemic sclerosis; SD: standard deviation; KPS: Karnofsky Performance Status; NYHA: New York Heart Association; PAP: pulmonary artery pressure. ::: Association of gender with SSc HRQoL and disability {#s3c} --------------------------------------------------- HRQol assessed by the SF-36 was comparable in both groups. PCS and MCS were similar, and lower than 40 out of 100, for both males and females (34.2 \[10.0\] and 35.9 \[9.6\], p = 0.240, and 35.1 \[12.3\] and 34.3 \[13.0\], p = 0.667, respectively). Consistently, regarding global, patient-perceived, and location-specific disability as assessed by the HAQ, MACTAR, and CHFS and MHISS, respectively, we found no gender differences within each of these variables. Both groups exhibited similar aesthetic impairment ([Table 3](#pone-0017551-t003){ref-type="table"}). ::: {#pone-0017551-t003 .table-wrap} 10.1371/journal.pone.0017551.t003 Table 3 ::: {.caption} ###### Association of gender with health-related quality of life and disability in SSc[\*](#nt105){ref-type="table-fn"}. ::: {#pone-0017551-t003-3} All patientsn = 381 Malesn = 62 Femalesn = 319 p-value^†^ ----------------------------- --------------------- ------------- ---------------- ------------ KPS (0--100) 77.6 (11.7) 78.0 (11.6) 77.5 (11.7) 0.755 SF-36 (0--100) • Physical functioning 35.7 (24.8) 31.4 (24.7) 36.6 (24.7) 0.152 • Physical role 19.4 (24.8) 15.8 (29.5) 20.1 (31.5) 0.350 • Bodily pain 22.5 (25.7) 19.6 (24.3) 23.1 (25.9) 0.349 • General health perception 24.1 (20.2) 23.9 (22.4) 24.1 (19.8) 0.932 • Vitality 21.9 (20.7) 21.7 (22.6) 21.9 (20.3) 0.973 • Social functioning 30.0 (32.5) 30.4 (34.3) 30.0 (32.2) 0.923 • Emotional role 24.2 (36.5) 24.5 (38.0) 24.1 (36.3) 0.945 • Mental health 32.6 (25.1) 31.6 (27.5) 32.8 (24.7) 0.753 • PCS 34.9 (14.6) 34.2 (10.0) 35.9 (9.6) 0.240 • MCS 38.5 (36.9) 35.1 (12.3) 34.3 (13.0) 0.667 HAQ (0--3) 1.1 (0.8) 1.0 (0.8) 1.1 (0.8) 0.237 MACTAR (0--30) 18.5 (8.2) 18.8 (7.1) 18.4 (8.4) 0.753 CHFS (0--90) 20.1 (19.3) 19.8 (20.3) 20.2 (19.2) 0.890 MHISS (0--48) 19.0 (11.6) 20.4 (15.0) 18.7 (10.7) 0.512 Aesthetic impairment 4.6 (2.6) 4.5 (3.4) 4.6 (2.4) 0.833 \* Values are the mean (standard deviation). SSc: systemic sclerosis; KPS: Karnofsky Performance Status Scale; SF-36: Medical Outcomes Study 36-Item Short Form Health Survey; PCS: Physical Component Score; MCS: Mental Component Score; HAQ: Health Assessment Questionnaire; MACTAR: McMaster-Toronto Arthritis Patient Preference Disability Questionnaire; CHFS: Cochin Hand Function Scale; MHISS: Mouth Handicap In Systemic Sclerosis Scale. ::: Association of gender with SSc self-reported symptoms of depression and anxiety {#s3d} ------------------------------------------------------------------------------- On bivariate analysis, some differences were observed between males and females for self-reported symptoms of depression and anxiety, but without reaching statistical significance. Indeed, self-reported symptoms of anxiety, as defined by HADa subscale score ≥8 were more frequent in females than males (62.3% *vs* 43.5%, p = 0.006), whereas absence of self-reported symptoms of both depression and anxiety, as defined by HADa and HADd subscale scores \<8 was more often encountered in males than females (46.8 *vs* 31.6%, p = 0.021) ([Table 4](#pone-0017551-t004){ref-type="table"}). Males and females did not differ in depression symptoms. On multivariate analysis, gender was associated with anxiety only (OR 5.50, 95% CI 1.12--27.04). ::: {#pone-0017551-t004 .table-wrap} 10.1371/journal.pone.0017551.t004 Table 4 ::: {.caption} ###### Association of gender with depression and anxiety in SSc[\*](#nt107){ref-type="table-fn"}. ::: {#pone-0017551-t004-4} All patientsn = 381 Malesn = 62 Femalesn = 319 p-value^†^ ---------------------------- --------------------- -------------- ---------------- ------------ HADa (0--21) (mean \[SD\]) 9.2 (4.5) 8.3 (5.1) 9.4 (4.4) 0.088 • HADa≥8 224/378 (59.3) 27/62 (43.5) 197/316 (62.3) 0.006 HADd (0--21) (mean \[SD\]) 6.6 (4.2) 6.5 (4.6) 6.6 (4.1) 0.781 • HADd≥8 154/378 (40.7) 25/62 (40.3) 129/316 (40.8) 0.957 • HADa and HADd\<8 129/378 (34.1) 29/62 (46.8) 100/316 (31.6) 0.021 \*Values are number/number of patients for whom the data is available (%), otherwise indicated in parenthesis. SSc: systemic sclerosis; n: number; HADa: Hospital Anxiety and Depression scale for Anxiety; HADd: Hospital Anxiety and Depression scale for Depression. ::: Discussion {#s4} ========== In the present study of 381 patients with SSc, we found a ratio of females to males of 5 to 1, which is in agreement with previous studies [@pone.0017551-Oliver1]. Some differences were observed between males and females for clinical symptoms and self-reported symptoms of depression and anxiety, however without reaching statistical significance. Indeed, dcSSc, echocardiography PAP\>35 mmHg and ILD were more often encountered in males, whereas lSSc and calcinosis were more often encountered in females. Females were more frequently found with self-reported symptoms of anxiety. Conversely, we found no association with gender regarding perceived health status, HRQoL and reported global and location-specific disability. On multivariate analysis, BMI, echocardiography PAP\>35 mmHg, and anxiety were the variables most closely associated with gender. The prevalence of dcSSc in male patients was high and reached 54.8%. DcSSc was more frequent than lcSSc in males. The exact opposite was observed in females and was more consistent with previous reports of epidemiology studies of both male and female SSc patients. In two large US and German studies, the prevalence of lcSSc and dcSSc among SSc patients was 66.2% and 33.8%, and 45.5% and 32.7%, respectively [@pone.0017551-Hunzelmann1], [@pone.0017551-Mayes1]. In 3 cohorts of 1,012 Italian, 249 Swedish and 185 Canadian patients, dcSSc was more frequent in males than females (range from 37% to 67%) [@pone.0017551-AlDhaher1], [@pone.0017551-Ferri1], [@pone.0017551-Hesselstrand1]. Conversely, 2 studies from Spain comparing male and female SSc patients found no gender differences by disease type [@pone.0017551-Simeon1], [@pone.0017551-Joven1]. Finally, from a recent retrospective French survey of 121 SSc patients, dcSSc was more frequent in males than in females (22% *vs* 5%) [@pone.0017551-Gaultier1]. In these last 3 studies, the male sample sizes were rather small (n = 9, n = 26 and n = 36, respectively). We found BMI significantly lower in females than in males. This finding might be of clinical relevance despite lack of clear explanatory reports. Indeed, in a prospective multiethnic cohort of 250 SSc patients, low BMI was among the 7 independent variables predictive of mortality. The authors even hypothesized that strong association of low BMI with mortality could be an objective and/or complete surrogate for generalized deconditioning or gastro-intestinal involvement [@pone.0017551-Assassi1]. Females were also more likely to have calcinosis, which is a frequent manifestation of SSc and found in about 25% of patients [@pone.0017551-Robertson1]. Calcinosis mainly affects the extremities, at sites of recurrent microtrauma such as the forearms, elbows or fingers [@pone.0017551-Boulman1]. It occurs predominantly at a late stage of disease and is not restricted to patients with the lSSc [@pone.0017551-Black1]. However, clinical features associated with calcinosis remain poorly described. Recently, we provided evidence that calcinosis is an independent factor associated with digital ulcers (OR 2.33, 95% CI 1.04--5.19) [@pone.0017551-Mouthon3]. In agreement with previous studies, we found that men were more likely than women to have echocardiography PAP\>35 mmHg and ILD, for prevalences of 24.2% and 54.8%, respectively. Lung involvement is common in the course of SSc, and together, ILD and pulmonary hypertension are considered the 2 main causes of death in this disease [@pone.0017551-Steen2]. ILD is more frequent in male SSc patients at the time of diagnosis and during follow-up [@pone.0017551-Gaultier1]. In addition, male gender is associated with pulmonary hypertension during follow-up [@pone.0017551-Gaultier1]. In 1180 SSc patients (19% men) studied at early stages of the disease, men more often than women were found to have lung fibrosis and lower diffusing lung capacity for carbon monoxide than women [@pone.0017551-Carreira1]. Thus, lung involvement in male SSc patients requires special attention and specific care because of its frequency and association with poor prognosis. Interestingly, we found gender differences for both self-reported symptoms of depression and anxiety in SSc. Females more often exhibited self-reported symptoms of anxiety, whereas men were more often free of self-reported symptoms of both anxiety and depression. In a recent cross-sectional survey of 108 patients visiting a rheumatology outpatient department, the only factor significantly associated with psychiatric symptoms was gender [@pone.0017551-Azad1]. Conversely, in another cross-sectional study of 111 patients visiting a rheumatology clinic, 9% with SSc, gender had no effect on the frequency of anxiety and depression [@pone.0017551-Waheed1]. Finally, in a study designed to assess psychological adjustment of 112 patients with early polyarthritis, female gender was found to be associated with high levels of depression and anxiety [@pone.0017551-Ramjeet1]. Substantial evidence indicates that females report greater fear and are more likely to have anxiety disorders than are males. Complex processes underlie gender differences in anxiety. Individual differences in etiological factors of anxiety and fear are moderated by socialization processes that prescribe gender-specific expectations for expression of anxiety and the acceptable means of coping with anxiety [@pone.0017551-McLean1]. Finally, we found no differences in depression symptoms by gender (40.3 *vs* 40.8%, p = 0.96), which is consistent with recent findings by Thombs *et al* [@pone.0017551-Thombs1]. Remarkably, despite our finding of gender differences in clinical expression in SSc, males and females experienced comparable loss of function, global and location-specific disability, and HRQoL impairment, as evidenced by similar HAQ, CHFS, MHISS, MACTAR and SF-36 scores. Gender may not be a major determinant of perceived disability and impaired HRQoL in patients with SSc, and functional and social issues should be considered as severe in males as in females. Consistently, we recently found in a cohort of 87 SSc that employment status was strongly associated with perceived disability and health status but not with gender [@pone.0017551-Nguyen1]. In addition, using the World Health Association Disability Assessment Schedule II to assess HRQoL, Hudson *et al* found that clinical correlates of HRQoL did not include gender [@pone.0017551-Hudson2]. HRQoL and functional disability may be associated with the meaning that SSc patients ascribe to their condition, which may be comparable for both males and females, rather than with its severity or its organ manifestation. Our work has some limitations. Our sample of males was small, and our inability to demonstrate statistically significant differences between the two groups might be due to the lack of statistical power. Another limitation was the procedure used to recruit patients. Since all patients belonged to the French association of patients or were hospitalized in tertiary care units, they may not be representative of the whole French SSc population. Patients had longstanding disease, which could imply more symptoms. HAQ scores were high but remained comparable to those reported from previous studies conducted in tertiary care settings [@pone.0017551-Georges1]. Moreover, patients recruited from the patient association may have had more severe SSc than hospitalized patients [@pone.0017551-MestreStanislas1]. Further studies conducted in other cohorts are required to confirm the gender differences we observed. Finally, our study was not designed to explore the reasons for the observed gender differences. One can only hypothesize about the associated etiological factors, which may involve hormonal influences; genetics such as X-chromosome inactivation and monosomy, or microchimerism; as well as lifestyle (e.g., the debated connection with silicone implants) [@pone.0017551-Oliver1]. In conclusion, we confirm the association of gender and clinical manifestations in patients with SSc. Diffuse disease and lung involvement are more frequent in males, whereas females more often exhibit calcinosis and self-reported symptoms of anxiety. Despite SSc patients displaying gender-related clinical differences, the disease impact on perceived health status, HRQol and disability, is comparable in both groups. Studies comparing male and female patients living in different countries, with different occupational, lifestyle and medical exposure, would also be of interest to further clarify the role of environmental factors in such gender differences. We thank the patients from the ASF for their participation in the study. We also thank members of the ASF for their logistical help. Luc Mouthon, Alice Bérezné, and Loïc Guillevin are members of the *Groupe Français de Recherche sur la Sclérodermie*. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The authors have no support or funding to report. [^1]: Conceived and designed the experiments: CN TB AB AP CM-S LM LG SP MR FR SM-D. Performed the experiments: CN TB AB AP CM-S LM LG SP MR FR SM-D. Analyzed the data: CN TB AB AP CM-S LM LG SP MR FR SM-D. Contributed reagents/materials/analysis tools: CN TB AB AP CM-S LM LG SP MR FR SM-D. Wrote the paper: CN TB AB AP CM-S LM LG SP MR FR SM-D.

PubMed Central

2024-06-05T04:04:19.168661

2011-3-9

PMC3052320

Introduction {#s1} ============ Hitherto, gammaretroviruses have never been found in human populations In mice, for example, the murine leukemia viruses (MLVs) which belong to this genus of retroviruses can be both exogenous and endogenous. Classically, endogenous MLVs are categorised on the basis of their envelope gene sequence and their cognate receptor polymorphism into ecotropic (replicating only in murine cells), xenotropic (replicating in multiple mammalian cells, including human but not murine cells), polytropic and modified polytropic (able to grow in murine and nonmurine cells) [@pone.0017592-Stoye1]. The discovery by microarray technology of a retrovirus in human prostates that was most closely related in its sequence to the xenotropic MLVs explains its designation as xenotropic murine leukemia virus-related virus (XMRV) [@pone.0017592-Urisman1]. Several reports have since confirmed the XMRV association with prostate cancer in the US [@pone.0017592-Schlaberg1]--[@pone.0017592-Danielson1], but not in Europe with the exception of one study in Germany study in which it was found in one patient and one control [@pone.0017592-Hohn1]--[@pone.0017592-Fischer1]. One Mexican study described the identification of XMRV sequences in one healthy control, but not in prostate tumor [@pone.0017592-MartinezFierro1]. Interest in XMRV heightened when a study from the Whittemore Peterson Institute, published in *Science* by Lombardi et al [@pone.0017592-Lombardi1] identified XMRV in 67% of patients suffering from chronic fatigue syndrome (CFS) and in 3.7% of healthy controls. One [@pone.0017592-Lo1] of the many independent studies that failed to find XMRV in a variety of CFS patient cohorts [@pone.0017592-Erlwein1]--[@pone.0017592-Henrich1] amplified instead genetic sequences closely related to the polytropic MLVs. We [@pone.0017592-Erlwein1] among others failed to confirm the presence of XMRV in CFS patients tested not only in Europe [@pone.0017592-vanKuppeveld1], [@pone.0017592-Groom1] and China [@pone.0017592-Hong1] but also in the US [@pone.0017592-Switzer1], [@pone.0017592-Henrich1] where the original study was carried out. It has been suggested that these negative results could have arisen because of a failure to duplicate the experimental conditions described in the original publication by Lombardi et al [@pone.0017592-Lombardi1]. To remove any element of doubt, we have repeated our genetic analysis using the same oligonucleotide primer sets described in [@pone.0017592-Lombardi1] and, in addition, we have assayed sera taken from 130 of our CFS patients and 30 normal healthy subjects (NHS) for antibodies to the New Zealand Black (NZB) xenotropic virus and its envelope protein, gp70, a virus that shares more than 94% overall homology with XMRV. Materials and Methods {#s2} ===================== Patient samples {#s2a} --------------- All patients gave written informed consent for the use of their DNA to test aetiological theories of CFS, and the study was approved by the South London and Maudsley NHS Trust Ethics Committee. The patient samples have been described in detail previously [@pone.0017592-Erlwein1]. Briefly, patients were recruited into our study from consecutive referrals to the CFS clinic at King\'s College Hospital, London and constitute a well-characterised and representative sample of routine clinic attenders. All patients had undergone medical screening to exclude detectable organic illness, including a minimum of physical examination, urinalysis, full blood count, urea and electrolytes, thyroid function tests, liver function tests, 9 a.m. cortisol and ESR. Patients were interviewed using a semi-structured interview [@pone.0017592-Sharpe1]; all included patients met the CDC international consensus criteria for CFS [@pone.0017592-Fukuda1]. Patients with the Fukuda-specified exclusionary psychiatric disorders, or somatisation disorder (as per DSM-IV), were not included. The use of standardised scales revealed high levels of fatigue and disability (see [@pone.0017592-Erlwein1] and supplementary correspondence for further details). Overall we are confident that the patients in this study are typical of those seen in secondary care and/or CFS clinics around the world [@pone.0017592-Wilson1]. PCR amplification of XMRV sequences {#s2b} ----------------------------------- DNA extraction has been described previously [@pone.0017592-Erlwein1]. Briefly, DNA was extracted from whole blood collected in EDTA. As a control for DNA integrity, a 124 nucleotide (nt) section of the human beta-globin gene was amplified. Amplification of XMRV *gag* sequences was carried out using primers, 419F and 1154R, and of *env* sequences using primers, 5922F and 6273R [@pone.0017592-Lombardi1]. Reactions were carried out in a volume of 25 µl which contained 0.5 units TaqGold (Applied BioSystems, Warrington, UK), 1× TaqGold reaction buffer (Applied BioSystems), 1.5 mM Mg++, 200 mM each dNTP, 2.5 pmol each primer to which 5 µl DNA extract or control (see below) was added. The PCR conditions were as follows; one cycle at 94°C for 8 minutes, 45 cycles at 94°C for 30 seconds, 55°C for 30 seconds 72°C for 1 minute and a final annealing cycle of 72°C for 7 minutes. A 5 µl aliquot of the PCR product was applied to a 1% agarose gel, stained with Ethidium Bromide and subjected to electrophoresis. Each PCR included six no-template (water) controls and a positive control, plasmid pXMRV, containing the full-length XMRV isolate, vp62, for which we are grateful to R. Silverman. Serology {#s2c} -------- Sera from 130 CFS patients and 30 normal healthy subjects (NHS) as controls were assayed blind and in duplicate for antibody responses to the xenotropic NZB retrovirus envelope protein, gp70 by two independent ELISAs, based on antibodies against the NZB virus and the ecotropic Rauscher MLV. For the assays, a CNBr-activated Sepharose 4B column coupled with gp70-enriched serum glycoproteins was used to affinity-purify either goat anti-NZB xenotropic virus antibodies, or goat anti-Rauscher MLV gp70 antibodies, as previously described [@pone.0017592-Izui1]. gp70 antigen capture ELISA {#s2d} -------------------------- Microtiter plates were coated with 83A25 rat IgG2a anti-retroviral gp70 mAb (10 µg/ml), which recognizes xenotropic, polytropic, ecotropic and amphotropic murine leukemia viruses [@pone.0017592-Evans1]. Serum samples were diluted 1∶100 in PBS containing 1% BSA and 0.05% Tween 20 and incubated overnight at 4°C. Virus was detected following a 5-hour incubation with affinity-purified goat anti-Rauscher MLV gp70 or affinity-purified goat anti-NZB xenotropic virus antibodies, labelled with alkaline phosphatase. Results are expressed as absorbance values at 405 nm (A~405~). In these assays, 2-fold serially diluted NZB virus, obtained from ViroMed Biosafety Laboratories, Camden, NJ, was used as a positive control. XMRV antibody response {#s2e} ---------------------- Serum IgG anti-XMRV responses were determined by ELISA. Microtiter plates were coated with whole inactivated NZB retrovirus (10 µg/ml), and incubated with 1∶100 diluted serum samples overnight at 4°C. The assay was developed following a 5 hour incubation with alkaline phosphatase-labelled goat anti-human IgG antibody (Cappel Laboratories, Durham, NC). The A~405~ values are shown. As there is no defined XMRV-positive human antiserum available, we used 603, a murine anti-Xeno gp70 monoclonal antibody [@pone.0017592-Portis1] as a positive control. Statistical analysis {#s2f} -------------------- Data were analysed by unpaired Student\'s T test comparison of absorbance values between CFS patients and NHS controls. The ELISA cut-off was calculated as the mean + 3 standard deviations from the A~405~ value of the NHS sera. Results {#s3} ======= We examined 48 of the previously described [@pone.0017592-Erlwein1] 186 samples by PCR using primers derived from the *gag* and *env* regions of the XMRV genome previously published by Lombardi et al [@pone.0017592-Lombardi1]. No amplicons were observed in either reaction ([**Figures 1a and 1b**](#pone-0017592-g001){ref-type="fig"}). ::: {#pone-0017592-g001 .fig} 10.1371/journal.pone.0017592.g001 Figure 1 ::: {.caption} ###### PCR amplification for XMRV in CFS patients. (**A**) Amplification products from nested PCR using XMRV *env* primers are shown. The *env* specific product which is 352 nt long was not amplified from CFS patients 37--48, as shown in lanes 1--12, but was amplified from the plasmid pXMRV, lane 13. The no-template controls are shown in lanes 14--19 and the DNA marker in lane 20. (**B**) The *gag* specific product, 736 nt in length, was not amplified from the same CFS patient DNA samples, lanes 1--12. The DNA marker is shown in lane 13, the plasmid pXMRV in lane 14 and the water controls in lanes 15--20. :::  ::: Sera from 130 CFS patients and 30 NHS controls were assayed by ELISA for antibodies to MLV Env protein and in an antigen capture ELISA for the presence of MLV Env itself. Results are shown in [**Figures 2**](#pone-0017592-g002){ref-type="fig"} **and** [**3**](#pone-0017592-g003){ref-type="fig"}. ::: {#pone-0017592-g002 .fig} 10.1371/journal.pone.0017592.g002 Figure 2 ::: {.caption} ###### Antibody ELISA using sera from CFS patients and normal health subjects (NHS). The absorbance A~405~ values for 130 sera from CFS patients and NHS are shown. For the antibody ELISA whole NZB xenotropic virus was coated to the microtitre plate following incubation with serum samples overnight. Antibodies were detected by incubation with anti-human IgG antibodies labelled with alkaline phosphatase. The solid lines represent the mean. The BSA background of 0.071 is represented by the dotted line. :::  ::: ::: {#pone-0017592-g003 .fig} 10.1371/journal.pone.0017592.g003 Figure 3 ::: {.caption} ###### Antigen-capture ELISA using sera from CFS patients and NHS. (**A**) Goat anti-Rauscher MLV gp70 (lane a) and goat anti-NZB xenotropic virus (lane b) antibodies were tested for their reactivity against NZB xenotropic virus in 2-fold serial dilutions starting with 10 µg/ml. The A~405~ values are given. For the antigen capture ELISA, xenotropic virus envelope was captured onto the microtitre plate using 83A25 rat monoclonal antibody [@pone.0017592-Evans1]. Detection was then carried out using goat anti-Rauscher antibodies or goat anti-NZB xenotropic MLV gp70 antibodies. (**B**) Results using goat anti-Rauscher MLV gp70 antibodies. The solid lines represent the mean and the dotted line indicates the BSA background of 0.161. (**C**) Results shown for goat anti-NZB xenotropic virus antibodies. The solid line represents the mean and dotted line indicates the BSA background of 0.095. :::  ::: For the antibody ELISA, NZB virus was coated onto microtiter plates and patient sera incubated at a dilution of 1∶100. The reaction was detected by alkaline phosphatase-labelled polyclonal goat anti-human IgG antibody. The A~405~ value obtained with BSA as background was 0.071. The cut-off was at 0.289. No specific serological reaction could be detected (*p* = 0.336, [**Figure 2**](#pone-0017592-g002){ref-type="fig"}). For the antigen capture ELISA, 83A25 rat IgG2a antibody raised against murine leukaemia virus envelope protein and able to recognise xenotropic, polytropic, ecotropic and amphotropic MLV [@pone.0017592-Evans1], was coated onto microtiter plates and the serum added at a dilution of 1∶100. On Western blots this antibody recognised recombinant XMRV envelope protein, gp70, produced in 293 T cells (data not shown). To detect bound Env protein, affinity-purified polyclonal antibodies, goat anti-Rauscher MLV gp70 or goat anti-NZB xenotropic virus labelled with alkaline phosphatase were employed. These antibodies had demonstrated reactivity to NZB virus in experiments using 2-fold serial dilutions ([**Figure 3a**](#pone-0017592-g003){ref-type="fig"}). When goat anti-Rauscher MLV gp70 antibody was used as an XMRV detection reagent on sera from CFS patients and NHS ([**Figure 3b**](#pone-0017592-g003){ref-type="fig"}), the background A~405~ value obtained with BSA was 0.161. The cut-off value was 0.296. Four CFS samples and one NHS sample gave A~405~ values above the cut-off range, but the reactivity was not significant (*p* = 0.632). When using goat anti-NZB xenotropic virus antibody, the A~405~ value of BSA was of 0.095. Twenty CFS sera were significantly above the cut-off value of 0.222, *p* value of 0.017 ([**Figure 3c**](#pone-0017592-g003){ref-type="fig"}). However, the reactivity was not demonstrated in both ELISAs ([**Table 1**](#pone-0017592-t001){ref-type="table"}). For example, the CFS sera 122 and 186, which displayed the highest A~405~ values of 0.730 and 0.371, respectively, in the goat anti-NZB virus antibody ELISA, had A~405~ values of 0.159 and 0.138 in the goat anti-Rauscher gp70 MLV ELISA, a value below the BSA background of 0.161. Moreover, CFS sera 261 and 61, which showed the highest absorption of o.492 and 0.419 in the goat anti-Rauscher gp70 ELISA, gave values of 0.181 and 0.167 (both below the cut-off value of 0.222) in the anti-NZB virus antibody ELISA. ::: {#pone-0017592-t001 .table-wrap} 10.1371/journal.pone.0017592.t001 Table 1 ::: {.caption} ###### Absorbance values of CFS sera in the two antigen-capture ELISAS used. ::: {#pone-0017592-t001-1} ELISA A~405~ ----- -------------- ------- 17 0.304 0.299 61 0.419 0.167 261 0.492 0.181 185 0.327 0.278 122 0.159 0.730 186 0.138 0.371 α-Rauscher, goat anti-Rauscher MLV antibody. α-NZB, goat anti-NZB antibody. The values are the mean of two independent experiments. α -Rauscher, goat anti-Rauscher gp70 MLV antibody; α-NZB, goat anti-NZB virus antibody. ::: Only sera 17 and 185 produced A~405~ values of 0.304 and 0.327, slightly above the cut-off of 0.296 in the goat anti-Rauscher MLV antibody ELISA. These sera also had a modest reaction of 0.299 and 0.278 in the goat anti-NZB virus antibody ELISA (cut-off being 0.222). However, as other samples showed similar absorbance values in this assay that did not significantly react in the goat anti-Rauscher gp70 MLV ELISA, the reactions are considered to be non-specific. Taken together, these experiments suggest that neither XMRV nor a specific serological response against this family of viruses was found in CFS patients. Discussion {#s4} ========== The connection between XMRV and CFS remains highly controversial. The initial report by Lombardi et al [@pone.0017592-Lombardi1] identified XMRV in 67% of CFS patients and 3.7% of health control subjects. Subsequent to this, using the same experimental protocol an independent study by Lo et al. [@pone.0017592-Lo1] detected four classes of MLV-related *gag* sequences in 86.5% of CFS patients and blood donors. In this study [@pone.0017592-Lo1] it was suggested that MLV-related sequences found in CFS cases vary to the extent that they may have remained undetected by our original PCR primers which targeted the leader *gag* region and the *pol* region of XMRV. We have now used the primers described in the original paper [@pone.0017592-Lombardi1] which bind to the XMRV *gag* and *env* open reading frames. We have again failed to detect XMRV or MLV-related sequences in 48 of our CFS patients, demonstrating that our failure to find XMRV in CFS tissue is not a reflection of the primers used in the amplification process. Indeed, the original primer sets have been exploited in several studies which have failed to amplify XMRV sequences in CFS patient samples [@pone.0017592-Groom1], [@pone.0017592-Switzer1], [@pone.0017592-Henrich1]. We have extended our study to investigate any possible serological response to XMRV in our CFS patients, since there appears to be a mis-match between PCR data and serology [@pone.0017592-Lombardi1], [@pone.0017592-Mikovits1]. In the absence of a specific XMRV serological assay, we made use of an ELISA in which Goat anti-Rauscher MLV gp70 and goat anti-NZB xenotropic virus antibodies were employed to detect gp70 Env protein in sera of CFS patients. In control experiments these antibodies clearly recognised recombinant XMRV gp70 Env expressed on dog thymus D17 cells (data not shown). Both assays highlighted samples with A~405~ values above the cut-off, however, for the goat anti-Rauscher MLV gp70 ELISA, this result was not significant. Using the goat anti-NZB xenotropic virus ELISA, twenty sera were above the cut-off, but most of these sera cluster outside the linear range with low absorbance values at the limit of detection of the assay as indicated by the flat curve b in [**Figure 3a**](#pone-0017592-g003){ref-type="fig"}. Considering further the fact that these sera did not behave consistently in both ELISAs, we believe that no specific serological reactions could be detected in the sera of the CFS patients. Non-specific serological reactions have also been described by Switzer et al in an ELISA using recombinant XMRV Env protein [@pone.0017592-Switzer1]. In this case, one of 51 sera from CFS patients and one from 53 healthy controls were slightly above the cut-off, but both sera were negative by further testing by immunofluorescence assays and PCR. Groom et al [@pone.0017592-Groom1] detected some neutralisation activity against MLV particles, pseudotyped with XMRV Env. This neutralisation activity was also present in healthy blood donors and showed no correlation with CFS [@pone.0017592-Groom1], nor could XMRV specific sequences be detected by PCR in these samples. It would not be inconceivable that our failure to replicate was due to patient differences. Clinical details of patients in the original study [@pone.0017592-Lombardi1] remain sparse. By contrast, we provide evidence confirming that our patients not only had CFS, but had considerable disability, and were not contaminated by either exclusionary psychiatric or medical diagnoses. While it is true that there will always be some variation based on local referral practices, differences in health care provision and so on, and while it is also true that CFS remains a difficult and elusive construct, even when international consensus criteria are applied, we doubt that this could explain the full range of differences in prevalence of XMRV published to date. Our serological data are consistent with the idea that XMRV may be present in the human population at some level [@pone.0017592-Groom1], [@pone.0017592-Switzer1]. However, this study confirms our previous report that a connection of XMRV or MLV-related viruses with UK CFS patients could not be substantiated. We thank Prof. Leonard Evans for providing the rat monoclonal antibody 83A25. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the NIHR Biomedical Research Centre. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: OE SI MOM SK. Performed the experiments: OE GW MJR SI. Analyzed the data: OE MJR MOM SK SI JW. Contributed reagents/materials/analysis tools: SW AC DC SI. Wrote the paper: MOM OE MJR.

PubMed Central

2024-06-05T04:04:19.171194

2011-3-9

PMC3052321

Introduction {#s1} ============ The primate brain effectively associates or integrates information from different modalities in order to establish robust representations of the outer world [@pone.0017499-Ernst1], [@pone.0017499-Welch1]. It has been considered that multisensory information processing is more closely related and mutually interactive than classical views had assumed [@pone.0017499-Driver1]. With regard to audiovisual interaction in motion perception, the effect of visual motion information on auditory motion perception has been mainly reported. For example, the adaptation made in response to moving visual stimuli induces a motion aftereffect in the auditory modality [@pone.0017499-Kitagawa1]. Moving visual stimuli also capture the perceived motion direction of the auditory stimulus [@pone.0017499-SotoFaraco1]--[@pone.0017499-SotoFaraco3]. These findings suggest that there are common neural substrates to motion perception between the visual and auditory modalities [@pone.0017499-SotoFaraco4]. The modulatory effect of auditory information on visual motion perception has been also reported. A transient sound disambiguates bistable visual motion perception [@pone.0017499-Sekuler1], [@pone.0017499-Watanabe1] by capturing the temporal positional information of a moving visual stimulus [@pone.0017499-Freeman1]. However, the inducing or driving effect of auditory information had not yet been reported. The effect of auditory motion information on visual motion direction perception was found to be absent [@pone.0017499-SotoFaraco2], very weak [@pone.0017499-Alais1], or indistinguishable from a response bias [@pone.0017499-Meyer1]. The lack of an inducing or driving effect of auditory motion information on visual motion perception was interpreted based on the reliability-based concept of multimodal interaction [@pone.0017499-Ernst1]; visual systems are usually superior to auditory systems in spatial processing so that auditory information has no effect on vision in motion processing [@pone.0017499-SotoFaraco3]. However, it has recently been demonstrated that the alternation of sound location can induce illusory visual motion perception to a static stimulus [@pone.0017499-Hidaka1], [@pone.0017499-Teramoto1]. This phenomenon is called as sound-induced visual motion (SIVM). In SIVM, a blinking visual stimulus at a fixed location was perceived to be laterally moving when it was synchronized with an alternating left-right sound source. SIVM was clearly observed when the visual stimulus was presented in the peripheral visual field (more than 10 deg). In line with the study on spatial localization [@pone.0017499-Alais2] and the reliability-based concept [@pone.0017499-Ernst1], the findings regarding SIVM suggest that auditory information becomes relatively more reliable in motion perception when visual information is vulnerable or degraded in the peripheral visual field. Whereas SIVM provides strong evidence demonstrating the inducing or driving effect of auditory information on visual motion perception, the underlying mechanisms are still unclear. It is possible that an alternating left-right sound captures the positions of the visual stimuli, resulting in visual motion perception (auditory positional capture) [@pone.0017499-Alais2], [@pone.0017499-Radeau1]. It is also possible that the auditory motion signals could directly contribute to visual motion perception, for example, when an alternating left-right sound source is perceived to be moving (i.e., apparent motion) [@pone.0017499-Briggs1], the moving sound source directly triggers motion perception of a static stimulus (c.f. *visual* motion capture) [@pone.0017499-Ramachandran1], [@pone.0017499-Ramachandran2]. The aim of the current study was to investigate the direct contributions of auditory motion signals on visual motion perception. In the first experiment (Experiment 1), we investigated how SIVM occurred in a situation where the spatiotemporal position of the flashes was located in the middle of the trajectory of a lateral auditory motion provided by a virtual stereo noise source smoothly shifting in a horizontal plane (either left to right or right to left) ([Figure 1B](#pone-0017499-g001){ref-type="fig"}). In this situation, the lateral position of the sound gradually changes from side to side, so that we can present a flash at the moment the sound is located near the flash in lateral position. Because there was little discrepancy in lateral position between the sound and flash, auditory positional information would have little influence on the perceived position of visual stimuli. If SIVM was observed in this situation, we could assume that auditory motion information directly contributes to visual motion perception. ::: {#pone-0017499-g001 .fig} 10.1371/journal.pone.0017499.g001 Figure 1 ::: {.caption} ###### Schematic illustrations of the experimental design and results of Experiment 1. \(A) Visual stimuli. (B) Time course of the presentation of auditory and visual stimuli. (C) Results. The vertical axis denotes the proportion of motion perception to the static visual stimuli. The horizontal axis denotes the retinal eccentricities of the visual stimuli. The error bars denote the standard error of the means. :::  ::: The second experiment (Experiment 2) further investigated the effect of an auditory motion signal on visual motion perception without one-to-one correspondence between the auditory and visual stimuli. Together with the lateral auditory motion, we presented a global visual motion display in which different localized motion signals, contained in multiple visual stimuli, were combined to produce a coherent motion perception [@pone.0017499-Williams1] ([Figure 2A](#pone-0017499-g002){ref-type="fig"}). There was no clear one-to-one correspondence between the auditory stimuli and each visual stimulus. If the auditory stimuli containing motion information affected integrated visual motion information and its perception, this would provide strong evidence for a direct interaction between auditory and visual modalities in motion processing. ::: {#pone-0017499-g002 .fig} 10.1371/journal.pone.0017499.g002 Figure 2 ::: {.caption} ###### Schematic illustrations of the experimental design and results of Experiment 2. \(A) Global motion display containing multiple local motion vectors. (B) Time course of the presentation of auditory and visual stimuli. (C) Psychometric curves as the proportion of motion direction perception consistent with sounds against the motion coherence. In the horizontal axis, the positive values indicate the situation where visual motion direction and alternate direction of sound (lateral-shift condition) or the presented location (one-sided condition) of the sounds was consistent. The inconsistent situation is represented by negative values. (D) The point of subjective equality (PSE) and (E) Slope of psychometric functions (JND) obtained in each retinal eccentricity. The error bars indicate the standard error of the means. :::  ::: Methods {#s2} ======= Ethics Statement {#s2a} ---------------- Written consent was obtained from each participant prior to experiments. The experiments were approved by the local ethics committee of Graduate School of Arts and Letters at Tohoku University. Participants and Apparatus {#s2b} -------------------------- Seven volunteers participated in Experiments 1 and 2; both experimental groups included two of the authors (S.H. and W.T.). All participants were experienced observers in psychophysical experiments, and five participants were naive to the purpose of the experiments. All participants had normal or corrected-to-normal vision and normal hearing. We presented the visual stimuli on a CRT display (Sony Trinitron GDM-FW900, 24 inch) with a resolution of 1600×1200 pixels and a refresh rate of 60 Hz. Sounds were presented through an audio interface (Roland EDIROL FA-66) and headphones (Sennheiser HDA 200). A customized PC (Dell-Dimension 8250) and MATLAB (The Mathworks, Inc.) with the Psychophysics Toolbox [@pone.0017499-Brainard1], [@pone.0017499-Pelli1] were used to control the experiments. We confirmed that the onset of the visual and auditory stimuli was synchronized using a digital oscilloscope (Iwatsu TS-80600). All of the experiments were conducted in a dark room. Stimuli {#s2c} ------- ### Experiment 1 {#s2c1} A red circle (0.4 deg in diameter; 17.47 cd/m^2^) was presented as a fixation point on a black background (0.4 cd/m^2^). A sequence of white bars (0.2 deg×3 deg; 4.99 cd/m^2^) was presented as visual stimuli at an eccentricity of either 2.5, 5, 10, or 20 deg along the horizontal plane ([Figure 1A](#pone-0017499-g001){ref-type="fig"}). A white noise was presented as an auditory stimulus for 400 ms with a cosine ramp of 5 ms at the onset and offset. The sampling frequency was 22050 Hz. The white noise was created per trial. There were three sound conditions: lateral-shift, one-sided, and no-sound. The lateral smooth movement of a virtual sound source was generated by cross-fading a pair of two white noises of 400 ms duration between the left and right ears (lateral-shift condition). For the cross-fading, each white noise was initially presented at the sound pressure level (SPL) of 85 dB and was then faded to null. The cross-fading white noise was presented two times without interstimulus interval (ISI) so that a virtual sound source was simulated to move from one side to the other and return to the original side, while the total sound power was kept constant. In the one-sided condition, a sound with a constant SPL of 85 dB (400 ms duration) was presented two times either to the left or right ears without ISI. In these conditions, the visual stimulus was presented for 200 ms in the middle of the cross-fading sound (lateral-shift condition) or the constant (one-sided condition) sound with 200 ms of ISI ([Figure 1B](#pone-0017499-g001){ref-type="fig"}). In each trial of these conditions, each white noise was presented 6 times and the visual stimulus was presented 5 times in total. In the no-sound condition, only the visual stimulus was presented 5 times. Each trial began with the presentation of the fixation point for 500 ms. The visual stimuli were presented in the participant\'s dominant eye field. ### Experiment 2 {#s2c2} A global motion display containing 200 white (4.99 cd/m^2^) dots was presented as visual stimuli. The diameter of each dot was 0.1 deg, and each dot was randomly located within 5 deg in diameter of an invisible circular window ([Figure 2A](#pone-0017499-g002){ref-type="fig"}). The global motion display was presented at an eccentricity of either 5, 10, or 20 deg. At the beginning of each trial, each dot moved in a random direction from 0 to 360 deg for 166 ms (10 frames) to 332 ms (20 frames). Then, the target motion signal was presented for 400 ms (24 frames), during which 6, 12, 24, or 48% of the dots moved either 0 deg (left) or 180 deg (right) as a target direction ([Figure 2B](#pone-0017499-g002){ref-type="fig"}). The other dots moved in a random direction except for the target motion direction. The lifetime of each dot was 2 frames. The velocity of each dot was 8 deg/s. The lateral-shift and one-sided conditions were both tested. The cross-fading sound or the constant sound was presented for 400 ms with 0 ms of ISI. The onset of the sounds was synchronized with that of the target motion signal ([Figure 2B](#pone-0017499-g002){ref-type="fig"}). Except for these variations, the stimulus parameters were identical to those of Experiment 1. Procedure {#s2d} --------- ### Experiment 1 {#s2d1} We asked the participants to place their heads on a chin rest, fixate on a red circle, and judge whether a flash (white bar) was perceived to be moving or not. This experiment consisted of a training session and a main experimental session. In the training session, the participants were asked to discriminate between static and moving visual stimuli for 80 trials: visual stimuli (2: static/moving) × eccentricities (4) × repetitions (10). The white bar was displaced back and forth by 0.2 deg in the horizontal direction when it moved. The training session was repeated until the discrimination performance reached above 75% for each eccentricity. The main session consisted of 240 trials where visual stimuli were always static: eccentricities (4) × auditory stimuli (3) × repetitions (20). In the lateral-shift condition, the first sound was delivered to the right ear for one-half of the trials and to the left ear for the other half. In the one-sided condition, the sounds were delivered to the right ear for one-half of the trials and to the left ear for the other half. The presentation order of the conditions and the location of the first sound (left/right) were randomized in the main session. Additionally, 96 filler trials where the white bar was actually displaced by 0.2 deg in the horizontal direction were randomly introduced in the trials of the main session: eccentricities (4) × auditory stimuli (3) × repetitions (8). In the filler trials, the white bar was physically displaced to the right for the rightward auditory motion and to the left for the leftward auditory motion because our preliminary observation confirmed that the perceived motion direction was congruent between the auditory and visual stimuli in the lateral-shift condition. In the one-sided condition, the initial onset position was consistent between the auditory and visual stimuli. The initial onset position of the visual stimuli (and that of the sounds) was randomized in the lateral-shift and one-sided conditions, and was randomly assigned in the no-sound condition. ### Experiment 2 {#s2d2} The participants were asked to fixate on a red circle and to report the perceived visual motion direction (left or right). This experiment contained only the main session. The main session consisted of 960 trials: auditory conditions (2: lateral-shift/one-sided) × auditory motion direction or location (2: left (ward) /right (ward)) × visual target motion direction (2: leftward/rightward) × coherence (4: 6, 12, 24, and 48%) × eccentricities (3: 5, 10, and 20 deg) × repetitions (10). The motion direction of the visual target and the auditory stimuli (lateral-shift condition) or the presented location (one-sided condition) of the sounds was either consistent or inconsistent. For example, leftward visual motion was presented with leftward (lateral-shift condition) or left-sided (one-sided condition) auditory stimulus in the consistent situation, while leftward visual motion with rightward (lateral-shift condition) or right-sided (one-sided condition) auditory stimulus in the inconsistent situation). The presentation order of the conditions and the target motion directions were randomized. Results {#s3} ======= We confirmed that the data without the authors\' responses and those including them were not statistically different (see [Figure S1](#pone.0017499.s001){ref-type="supplementary-material"}). Thus, we included the authors\' data in later analyses. Experiment 1 {#s3a} ------------ The proportion of motion perception to the static visual stimuli was calculated ([Figure 1C](#pone-0017499-g001){ref-type="fig"}). Then, we conducted a repeated measures analysis of variance (ANOVA) with 2 within-participant factors: eccentricities (2.5, 5, 10, and 20 deg) and auditory conditions (lateral-shift, one-sided, and no-sound). The ANOVA revealed a significant main effect of eccentricities (*F* ~3,\ 18~ = 5.87, *p*\<.01) and auditory conditions (*F* ~2,\ 12~ = 19.54, *p*\<.001). An interaction effect between these factors was also significant (*F* ~6,\ 36~ = 7.16, *p*\<.001). Regarding the significant simple main effect of auditory conditions (5 deg: *F* ~2,\ 48~ = 5.78, *p*\<.01; 10 deg: *F* ~2,\ 48~ = 14.88, *p*\<.001; 20 deg: *F* ~2,\ 48~ = 35.54, *p*\<.001), the post-hoc test (Tukey\'s HSD, *p*\<.05) revealed that the proportion of motion perception was higher in the lateral-shift condition than the other conditions for 5, 10, and 20 deg of eccentricity. In contrast, there were no significant differences between the one-sided and no-sound conditions. The results of Experiment 1 clearly showed that the lateral auditory motion of a single sound image smoothly shifting in the horizontal plane induced motion perception to the static visual stimulus when the spatiotemporal position of the visual stimuli was in the middle of the auditory motion trajectory; the flashes at a fixed location appeared to be moving laterally in the lateral-shift condition. This effect was clearly observed at the eccentricities larger than parafovea (5, 10, and 20 deg), and the effect appeared to become obvious as retinal eccentricities increased. This indicates that the degraded reliability of the visual stimuli increased the effect of sounds on visual motion perception [@pone.0017499-Hidaka1], [@pone.0017499-Teramoto1]. In contrast, the absence of a significant effect of sound in the one-sided condition suggested that the results in the lateral-shift condition were inattributable to the effect of the sound presentation itself. It would certainly be possible that the moving sound might induce response or decisional bias; the participants might expect that they perceived visual stimuli to move whenever the moving sound was presented. Thus, we conducted an additional experiment in order to test this possibility. The lateral-shift, one-sided, and no-sound conditions were presented as the auditory stimuli. The visual stimuli were either static or moving (0.2 deg) and presented at 10 deg of retinal eccentricity. Each block consisted of 144 trials (auditory conditions (3) × visual conditions (2) × repetitions (24)). To reduce the bias to the maximum possible degree, all of the participants (N = 10) were newly recruited and naïve to the purpose of the experiment; moreover, we provided them visual feedback (words such as "Correct!") to inform whether or not their judgments were correct in each trial. Except for these variations, the stimulus parameters and procedures were identical to those in Experiment 1. On the basis of the proportional data, we calculated d-prime as the index of sensitivity, which can be separated from the index of criterion or response bias (β) [@pone.0017499-Macmillan1] ([Figure 3](#pone-0017499-g003){ref-type="fig"}). The responses of perceiving static stimuli were regarded as a "hit" for the static trials and as a "false alarm" for the moving trials in the visual conditions. With regard to d-prime, a one-way repeated measures ANOVA revealed a significant main effect of the auditory conditions (*F* ~2,\ 18~ = 4.49, *p*\<.05). A post-hoc test (Tukey\'s HSD, *p*\<.05) revealed that the d-prime value in the lateral-shift condition was smaller than that in the other conditions. In contrast, there was no significant main effect regarding β (*F* ~2,\ 18~ = 1.67 *p* = .22). When the visual stimulus was in motion, the participants correctly perceived the motion in all of the conditions. In contrast, the static visual stimulus was illusorily perceived as moving only when the sounds traveled in the lateral direction. These phenomenal aspects were clearly shown in the decrement of d-prime in the lateral-shift condition. We further confirmed that the effect of the lateral shifts in sounds was so strong that the participants could not distinguish between actual and illusory motion even if they were given feedback, and that the changes in sensitivity (d-prime) were independent from those in criterion (β). These results suggest that the continuous left-right shifts of the virtual sound source actually changed the sensitivity to visual motion perception, and that the findings in Experiment 1 could not be simply explained by the response or the decisional bias. ::: {#pone-0017499-g003 .fig} 10.1371/journal.pone.0017499.g003 Figure 3 ::: {.caption} ###### D-prime data. \(A) Proportion of motion perception, (B) d-primes, and (C) response criterion (β). The horizontal axis denotes the types of auditory and visual conditions. The error bars denote the standard error of the means. :::  ::: The results of Experiment 1 demonstrated that the SIVM occurred even when auditory laterality information could have little influence on the perceived position of visual stimuli. We also confirmed that this was true even for discrete shifts of sound and the resulting auditory apparent motion [@pone.0017499-Briggs1], which was used in the previous studies [@pone.0017499-Hidaka1], [@pone.0017499-Teramoto1] ([Figure 4](#pone-0017499-g004){ref-type="fig"}). The mechanism underlying the SIVM would be the direct interaction between the motion signal contained in the lateral shifts in sounds (leftward or rightward) and visual motion perception rather than the auditory capture on visual localization [@pone.0017499-Alais2], [@pone.0017499-Radeau1]. ::: {#pone-0017499-g004 .fig} 10.1371/journal.pone.0017499.g004 Figure 4 ::: {.caption} ###### Data for discrete shifts of sound source. \(A) Time course of the presentation of auditory and visual stimuli. The horizontal shift of a sound (lateral-shift condition) was demonstrated by presenting the sound alternately to the left and right ears. In the one-sided condition, the sound was presented to either the left or the right ear. In these conditions, the sound was presented 6 times for 200 ms each with 200 ms of ISI, and the visual stimulus of 200 ms in duration was presented in between 2 successive sounds; that is, the visual stimulus was presented 5 times with 200 ms of ISI. In the no-sound condition, only the visual stimulus was presented 5 times. Except for these sound manipulations, the stimulus parameters and procedures were consistent with those of Experiment 1. (B) Results (N = 7). The participants were the same as those of Experiment 1. The vertical axis denotes the proportion of motion perception to the static visual stimuli. The horizontal axis denotes the retinal eccentricities of the visual stimuli. The error bars denote the standard error of the means. A repeated measures analysis of variance (ANOVA) with eccentricities (2.5, 5, 10, and 20 deg) and auditory conditions (lateral-shift, one-sided, and no-sound) revealed a significant main effect of auditory conditions (*F* ~2,\ 12~ = 23.03, *p*\<.001). An interaction effect between these factors was also significant (*F* ~6,\ 36~ = 2.56, *p*\<.05). Regarding the significant simple main effect of the auditory conditions (5 deg: *F* ~2,\ 48~ = 9.70, *p*\<.001; 10 deg: *F* ~2,\ 48~ = 14.03, *p*\<.001; 20 deg: *F* ~2,\ 48~ = 18.83, *p*\<.001), the post-hoc test (Tukey\'s HSD, *p*\<.05) revealed that the proportion of motion perception was higher in the lateral-shift condition than the other conditions for 5, 10, and 20 deg of eccentricity. (C) D-prime data (N = 8, all of which were newly recruited naïve participants). In another experiment, we estimated d-prime and β values for the discrete sounds (see the section of Experiment 1 in the Results part for details). With regard to d-prime, a one-way repeated measures ANOVA revealed a significant main effect of the auditory conditions (*F* ~2,\ 14~ = 8.14, *p*\<.005). The post-hoc test (Tukey\'s HSD, *p*\<.05) revealed that the d-prime value in the lateral-shift condition was smaller than that in the other conditions. In contrast, the β value in the one-sided condition was higher than that in the other conditions (ANOVA: *F* ~2,\ 14~ = 9.08, *p*\<.005; post-hoc test: *p*\<.05). This tendency was inconsistent with that of d-prime so that the changes in sensitivity could be assumed to be independent from those in criterion. :::  ::: Experiment 2 {#s3b} ------------ We depicted psychometric curves as the proportion of visual motion direction perception consistent with the lateral shifts in sounds as a function of the visual motion coherence in each eccentricity and each auditory condition ([Figure 2C](#pone-0017499-g002){ref-type="fig"}). Then, we estimated the point of subjective equality (PSE) as the 50% threshold of the psychometric functions by fitting a cumulative Gaussian distribution function to each participant\'s data using a maximum likelihood method ([Figure 2D](#pone-0017499-g002){ref-type="fig"}). A repeated measures ANOVA with 2 within-participant factors, auditory conditions (lateral-shift, one-sided) and eccentricities (5, 10, and 20 deg), was conducted. This revealed that the main effect of the auditory conditions was significant (*F* ~1,\ 6~ = 7.52, *p*\<.05); the PSE in the lateral-shift condition shifted toward the inconsistent direction as compared to that in the one-sided condition. The main effect of the eccentricities (*F* ~2,\ 12~ = .53, *p* = .60) and the interaction between the factors (*F* ~2,\ 12~ = 1.11, *p* = .36) were not significant. The results showed that the lateral auditory motion provided by smooth shifts of a single sound image in horizontal plane altered the motion direction perception of a global visual motion signal. Specifically, the lateral auditory motion perceptually cancelled out the opposite lateral visual motion information and induced consistent motion perception to the visual stimuli. Contrary to the previous studies [@pone.0017499-Alais1], [@pone.0017499-Meyer1], the current results clearly demonstrate that continuous lateral shifts of sound can induce visual motion direction perception consistent with auditory movement in a global motion display. With respect to the effect of an auditory motion signal on a global visual motion display, it was reported that auditory motion information affected the judgments of perceived visual motion direction only when the coherence of the visual local motion signal was considerably low [@pone.0017499-Meyer1]. This result indicates that auditory motion information was utilized for making decisions only when the visual motion direction was ambiguous and hard to discriminate alone. If this decisional biasing effect existed in Experiment 2, the slope of the psychometric functions would become less steep especially at the center against the visual stimuli with lower coherences. In order to confirm this possibility, we calculated the slope of psychometric functions as just noticeable differences (JND) by the following formula: (75% threshold -- 25% threshold)/2 ([Figure 2E](#pone-0017499-g002){ref-type="fig"}). A repeated measures ANOVA revealed that the main effect of the auditory conditions (*F* ~1,\ 6~ = 2.16, *p* = .19), that of the eccentricities (*F* ~2,\ 12~ = 1.67, *p* = .23), and the interaction between these factors (*F* ~2,\ 12~ = 1.11, *p* = .36) were not significant. We confirmed that the slopes of each psychometric function were consistent between the auditory conditions, indicating that the decisional criteria of motion direction perception were consistent between the auditory conditions. We, therefore, could consider that the decisional bias was unattributable to the main factor. The results of Experiment 2 indicate that the auditory motion information contained in continuous lateral shifts of a sound image can directly alter visual motion perception extracted from different localized motion vectors of multiple visual stimuli. Therefore, we could consider that motion processing and perception directly interact between auditory and visual modalities. Discussion {#s4} ========== We found that a lateral auditory motion provided by a pair of cross-fading white noises smoothly shifting along a horizontal trajectory induced illusory visual motion perception (SIVM) even when the flash was presented in the middle of the trajectory of the sound shifts; the spatial position of the virtual sound source was perceived around the visual stimulus at the moment the flash was presented, and the laterality information of the sound (left or right) could have little influence on the visual stimuli (Experiment 1). It was also revealed that the lateral auditory motion altered the visual motion direction perception in a global motion display (Experiment 2); different localized motion signals of multiple visual stimuli were combined to produce a coherent visual motion perception so that one-to-one correspondence between the auditory and visual stimuli was hard to be established. These findings suggest that there exists direct audio-visual interaction in motion processing, and that there might be common neural substrates for auditory and visual motion processing. Eye movements might be induced by the left-right shifts of sound. However, we confirmed that the SIVM occurred without eye movements (see [Figure S2](#pone.0017499.s002){ref-type="supplementary-material"}). In Experiment 2, the lifetime of each dot in our global motion display was only two frames, making it difficult to associate eye movements with each visual stimulus. We, therefore, could conclude that eye movements did not play a decisive role in the results of the present study. The involvement of response or decisional bias might be also suspected. However, in Experiment 1, we found that the d-prime in the lateral-shift condition was lower than that in the other conditions, whereas the β values did not differ among the conditions ([Figure 3](#pone-0017499-g003){ref-type="fig"}; see also [Figure 4C](#pone-0017499-g004){ref-type="fig"}). These results suggest that the left-right shifts of virtual sound source indeed change the sensitivity of motion perception in SIVM. Moreover, we also confirmed that the JNDs for global visual motion display were consistent between the auditory conditions in Experiment 2; instead, the PSEs changed by the lateral auditory motion generated by continuous left-right shifts of the virtual sound source ([Figure 2D and E](#pone-0017499-g002){ref-type="fig"}). This means that the auditory motion signal could perceptually cancel out the opposite visual motion signal and induce consistent motion perception to the visual stimuli in almost all of the coherences, namely even when there were relatively sufficient visual motion signals. These results indicate that the current findings could not be simply explained by the biases. Based on these findings, we can consider that the audio-visual interaction involved in motion processing could explain the current phenomenon. In the previous studies [@pone.0017499-Hidaka1], [@pone.0017499-Teramoto1], two possible mechanisms were considered regarding the auditory inducing effect on visual motion perception. One was the direct interaction of motion information between the auditory and visual modalities (c.f. *visual* motion capture) [@pone.0017499-Ramachandran1], [@pone.0017499-Ramachandran2]. Another mechanism is auditory capture on visual localization [@pone.0017499-Alais2], [@pone.0017499-Radeau1] in which the auditory spatial information (left or right) simply modulates visual inputs in a spatial domain. The current research demonstrated that SIVM occurred even when auditory laterality information could have little influence on the perceived position of visual stimuli. We also found that the continuous shifts of the virtual sound source altered the perception of global visual motion where there was no clear correspondence between the auditory stimuli and each of the visual stimuli. These findings indicate that auditory motion information can directly trigger or induce visual motion perception. Some previous studies showed that auditory motion affected visual motion perception. For example, auditory motion could direct an ambiguous, bistable motion perception to an unambiguous one [@pone.0017499-Sanabria1]. This finding indicates that auditory motion information could modulate visual motion perception. In contrast, our current findings demonstrated that auditory motion information triggered motion perception to static visual stimuli and drove motion perception against global visual motion signals. It was also reported that auditory motion affected the perception of a global visual motion display [@pone.0017499-Meyer1]. However, this effect could be primarily explained by response or decisional bias because the auditory effect was dominant only when the visual motion signal was highly ambiguous [@pone.0017499-Alais1], [@pone.0017499-Sanabria1]. On the contrary, as mentioned above, our results showed that auditory motion information affected global visual motion perception even when the visual signal contained highly coherent motion signals, and this effect could be distinguishable from response bias. We, therefore, could consider that the current findings are unique in that they demonstrate the driving and inducing effects of auditory motion on visual motion perception. In Experiment 1, the effect of eccentricity was observed; SIVM frequently occurred at the eccentricities larger than parafovea (5, 10, and 20 deg), and the effect of auditory motion appeared to become more obvious with increasing retinal eccentricities ([Figure 1C](#pone-0017499-g001){ref-type="fig"}). In line with the previous study [@pone.0017499-Hidaka1], [@pone.0017499-Teramoto1], we could assume that the auditory effect became obvious when the visibility or reliability of visual inputs degraded in the peripheral vision. This reliability-based theory would be consistent with the concept regarding the manner in which multimodal integration occurs [@pone.0017499-Ernst1]. In contrast, the effect of the eccentricities seemed not to be observed in Experiment 2; the effect of auditory motion was almost identical among the eccentricities ([Figure 2D](#pone-0017499-g002){ref-type="fig"}). The global motion display consisted of multiple visual stimuli with different localized motion signals. In the experiment, the coherence of the motion varied from trial to trial. Moreover, since the lifetime of each stimulus was only 2 frames, it was hard to discriminate or identify each stimulus. Since these manipulations alone could considerably degrade the reliability of the visual stimuli, the left-right sound source would induce the robust auditory effect on visual motion information among the parafoveal and peripheral visual fields. Contrary to the current research, the previous studies showed that the effect of auditory motion information on visual motion perception was not obvious [@pone.0017499-Alais1] or indistinguishable from biases [@pone.0017499-Meyer1] in a global visual motion display. The discrepancy might be considered in terms of the eccentricity of visual stimuli. In a previous study [@pone.0017499-Meyer1], the fixation point was presented at the center of the participants\' global motion display (16 deg ×16 deg) so that the participants received the visual information from both the left and right visual field, including the fovea and parafovea (±8 deg). In another study [@pone.0017499-Alais1], participants were presented with relatively large stimuli (50 deg ×38 deg) as a global motion display. However, a fixation point was not presented so that the participants could scan the visual stimuli with their eye movements during the presentation (670 ms). On the contrary, our global motion display was relatively small (5 deg in diameter) and was presented only in the participants\' dominant eye field together with the fixation point. In the previous studies, therefore, the visibility or reliability of the visual stimuli might be kept higher than that of the auditory stimuli so that the effect of auditory motion on visual motion perception would not be manifested. It was also notable that whereas the previous studies presented auditory stimuli through loud speakers, we used headphones for the presentation of sounds. Thus, the factor of spatial co-localization between the visual and auditory stimuli [@pone.0017499-Calvert1] also might be different between the previous and current studies. A detailed investigation regarding these issues is beyond the purposes of the current research and should be addressed in future research. In summary, the current study demonstrates that auditory motion signals can drive or induce visual motion perception consistent with auditory motion perception. The effect of auditory motion signals becomes obvious when the reliability of visual inputs is degraded. We have confirmed that the current results were not explained by auditory position capture effect, eye movements, or biases. The evidence of our study suggests the existence of direct interactions and common neural substrates between the auditory and visual modalities in motion processing and motion perception. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Data with and without the authors\' responses.** In order to compare the data without the authors\' responses and those including them, we conducted a mixed-design ANOVA by adding the factor of authors (2; with/without the authors\' data) as a between-subjects variable to main analyses (with regard to the main analyses, see the Result part for details) (A) Experiment 1. A main effect and interaction effects related to the factor of authors were not significant for continuous (authors: *F* ~1,\ 10~ = .08, *p* = .79; authors×auditory conditions: *F* ~2,\ 20~ = .44, *p* = .65; authors×eccentricities: *F* ~3,\ 30~ = .44, *p* = .94; authors×auditory conditions×eccentricities: *F* ~6,\ 60~ = .18, *p* = .98) and discrete (authors: *F* ~1,\ 10~ = .09, *p* = .77; authors×auditory conditions: *F* ~2,\ 20~ = .36, *p* = .71; authors×eccentricities: *F* ~3,\ 30~ = .16, *p* = .93; authors×auditory conditions×eccentricities: *F* ~6,\ 60~ = .17, *p* = .98) shifts of sound source. (B) Experiment 2. A main effect and interaction effects related to the factor of authors were not significant for point of subjective equality (authors: *F* ~1,\ 10~ = .47, *p* = .51; authors × auditory conditions: *F* ~1,\ 10~ = .11, *p* = .75; authors × eccentricities: *F* ~2,\ 20~ = .36, *p* = .70; authors × auditory conditions × eccentricities: *F* ~2,\ 20~ = .07, *p* = .93) and just noticeable difference (authors: *F* ~1,\ 10~ = .70, *p* = .42; authors×auditory conditions: *F* ~1,\ 10~ = .65, *p* = .44; authors × eccentricities: *F* ~2,\ 20~ = .19, *p* = .89; authors × auditory conditions×eccentricities: *F* ~2,\ 20~ = .01, *p* = .99). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Eye movement data.** We conducted a control experiment of Experiment 1 in which eye movements were recorded (continuous block). We also collected data for the discrete sounds (discrete block) (see [Figure 4](#pone-0017499-g004){ref-type="fig"}). The lateral-shift and no-sound conditions were presented as the auditory stimuli. The visual stimuli were presented at 10 deg of retinal eccentricity. Each block consisted of 80 trials of the main session with a static flash (auditory conditions (2)×repetitions (40)) and 32 trials of the filler session with a moving (0.2 deg) flash (auditory conditions (2)×repetitions (16)). The participant\'s eye position was recorded from the left eye at a sampling rate of 60 Hz with EMR-9 (NAC Image Technology, Inc.). Except for these variations, the stimulus parameters and procedures were identical to those of Experiment 1 or the additional experiment for discrete sounds. Trials in which eye position deviated by more than 1 deg of visual angle in the horizontal direction from the center of the fixation point during the stimulus presentation were discarded from the analysis. Whereas 12.1±4.8 (SEM) % and 9.2±3.0 (SEM) % of trials were excluded in the continuous block, 12.1±5.6 (SEM) % and 21.1±6.4 (SEM) % of trials were excluded in the discrete block in each auditory condition (lateral-shift and no-sound), respectively. (A) Proportion of visual motion perception without eye movements (N = 6, including 2 of the authors (S.H. and W.T.)). The error bars denote the standard error of the means. A paired two-tailed *t* test confirmed that the reliable amount of motion perception occurred in the lateral-shift condition in each block (continuous block: *t*(5) = 5.25, *p*\<.005; discrete block: *t*(5) = 3.11, *p*\<.05). We, therefore, could assume that eye movement was not a decisive factor of the result for SIVM. (B) Examples of eye movement recording data for a participant. The upper and lower data show the time course of eye position for the lateral-shift and no-sound conditions in each block, respectively. The data for all trials are shown, except for those in which the eye deviation was more than 1 deg. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to Jiro Gyoba and Yukio Iwaya for their suggestions. We also thank Junichi Takahashi for his support with data collection. We appreciate the helpful comments and suggestions by Melissa Coleman and two anonymous reviewers. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by a Grant-in-Aid for Specially Promoted Research (19001004) of the Ministry of Education, Science, and Culture, Japan. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SH WT Y. Sugita YM SS Y. Suzuki. Performed the experiments: SH WT Y. Sugita YM. Analyzed the data: SH WT Y. Sugita. Contributed reagents/materials/analysis tools: Y. Sugita SS Y. Suzuki. Wrote the paper: SH WT Y. Sugita YM SS Y. Suzuki.

PubMed Central

2024-06-05T04:04:19.173051

2011-3-9

PMC3052322

Introduction {#s1} ============ Many Gram-negative bacterial pathogens produce proteinaceous pathogenic factors that are secreted and injected into the host cell via the type III secretion system (T3SS) during the infective process [@pone.0017614-Blocker1], [@pone.0017614-He1], [@pone.0017614-Buttner1]. A great deal of focus has been aimed at understanding the T3SS of phytopathogenic *Xanthomonas* species that infect a wide variety of plant hosts, many of which are of great economic importance [@pone.0017614-Blocker1], [@pone.0017614-Rong1], [@pone.0017614-Alegria1], [@pone.0017614-Buttner2], [@pone.0017614-daSilva1], [@pone.0017614-Kay1], [@pone.0017614-VanGijsegem1], [@pone.0017614-Furutani1], [@pone.0017614-Marguerettaz1], [@pone.0017614-Szczesny1], [@pone.0017614-Song1], [@pone.0017614-Moreira1], [@pone.0017614-Pieretti1]. The phytopathogen *Xanthomonas citri* subsp. citri (*Xanthomonas axonopodis* pv citri strain 306; *Xac*) is the causal agent of citrus canker, a disease that threatens citrus crops world-wide [@pone.0017614-Brunings1]. The *Xac hrp* locus (*hrp*: "hypersensitive response and pathogenicity") encompasses a group of 25 genes that code for a T3SS. Some products encoded by these genes are conserved in all T3SS, including core flagellar secretory components, while others are proteins of unknown function but whose homologs are essential for T3SS function in other *Xanthomonas* species [@pone.0017614-Alegria1], [@pone.0017614-daSilva1], [@pone.0017614-Dunger1]. The *Xac* T3SS is required for the development of disease symptoms in susceptible citrus plants as well as for the hypersensitive response (HR) in resistant plants [@pone.0017614-Dunger1], [@pone.0017614-Swarup1]. Deletions in the *hrpB* and *hrpD* operons and deletions of the *hrpF* gene in *Xac* failed to produce canker in citrus plants or hypersensitive response (HR) in cotton [@pone.0017614-Dunger1]. Furthermore, a specific T3SS substrate, PthA (a member of the AvrBs3 family), has been shown to contribute significantly to T3SS-dependent development of disease symptoms by *Xac* in citrus and the introduction of the *pthA* gene into strains of *X. phaseoli* and *X. campestris* pv. *malvacearum* (neither pathogenic in citrus) resulted in the elicitation of HR in their respective hosts, bean and cotton [@pone.0017614-Swarup1]. A PthA homolog coded by the *hssB3.0* gene was found to be required for virulence of *Xac* KC21 on *Citrus grandis* cultivars [@pone.0017614-Hiroshi1]. Other possible T3SS-related factors have been identified in the *Xac* genome by bioinformatics analysis [@pone.0017614-daSilva1] but have not been studied at the genetic or protein level. We have previously identified protein-protein interactions involving components, substrates and regulators of the T3SS of *Xac* strain 306 [@pone.0017614-Alegria1] whose genome has been sequenced [@pone.0017614-daSilva1]. One of the interactions identified was that involving HrpB2 and HrcU. HrpB2 is a small protein found associated with the T3SS of only a few phytopathogenic bacteria (*Xanthomonas* spp., *Ralstonia solanacearum, Acidovorax avenae*) and of *Burkholderia* spp that can infect animals and plants. In *Xanthomonas campestris* pv. vesicatoria (*Xcv*), HrpB2 is secreted and is essential for the secretion of the AvrBs3 virulence protein by the T3SS [@pone.0017614-Rossier1] and has been shown to interact with HpaC, a protein required for the efficient secretion of other effectors proteins [@pone.0017614-Lorenz1]. These observations have led to the suggestion that HrpB2 may play a role in controlling the hierarchy of a stepwise secretion process [@pone.0017614-Rossier1], [@pone.0017614-Lorenz1]. HrcU homologs are found in all known T3SSs and flagellar systems and are made up of an N-terminal domain containing several transmembrane helices and a cytoplasmic C-terminal domain. In *Xanthomonas campestris* pv. *glycines* 8ra, the HrcU homolog is not required for HR induction on non-host plants, pepper and tomato, or for the multiplication of bacteria in the host plant, but was required for the pathogenic symptoms on soybean [@pone.0017614-Oh1]. On the other hand, insertion mutagenesis in the *Xcv hrpC* operon, which codes for both HrcU and HrcV, resulted in nonpathogenic mutants that exhibited significantly reduced growth in pepper leaves and lost the ability to induce HR in resistant host plants and in non-hosts [@pone.0017614-Bonas1]. HrcU is a paralog of the flagellar protein FlhB. The 173-residue C-terminal domain of FlhB from *Salmonella* is specifically cleaved between Asn-269 and Pro-270 within a NPTH motif [@pone.0017614-Minamino1] via an autocatalytic process [@pone.0017614-Ferris1]. This NPTH motif is conserved in all FlhB homologs, including those found in T3SS of animal and plant pathogens and a similar cleavage has been observed in the homolog YscU from the T3SS of *Yersinia pseudotuberculosis* [@pone.0017614-Lavander1]. In flagellar systems, mutations that abolish cleavage in FlhB also abolish the secretion of flagellin and other late export extracellular components but not early export proteins such as FlgD [@pone.0017614-Fraser1]. Cleavage of YscU does not however seem to be essential for the secretion of virulence factors by the *Yersinia* T3SS [@pone.0017614-Lavander1] and thus appears to discriminate between translocator and effector proteins [@pone.0017614-Sorg1]. Substitutions of N263 abolish autocleavage of YscU while P264 and H266 showed partial cleavage [@pone.0017614-Wiesand1]. Structural studies of YcsU [@pone.0017614-Wiesand1] and its homologs EscU from enteropatogenic *E. coli*, SpaS from *Salmonella typhimurium* [@pone.0017614-Zarivach1] and Spa40 from *Shigella flexneri* [@pone.0017614-Deane1] reported similar structural and functional data. In this report, we have characterized the interaction between HrpB2 and the C-terminal domain of HrcU of *Xac* using purified recombinant proteins. We show that when expressed in *E. coli*, HrcU~XAC~ suffers a cleavage within the NPTH motif in a manner similar to that observed for the HrcU homologs FlhB and YscU and that the HrpB2~XAC~ binding site on HrcU~XAC~ corresponds to the region C-terminal to the cleavage site. Deletion mutations in the *hrcU* and *hrpB2* genes (Δ*hrcU* and Δ*hrpB2*) resulted in a total loss of virulence *in planta* and pathogenicity could be regained by the expression of HrcU~XAC~ and HrpB2~XAC~ from extrachromosomal plasmids. Furthermore, citrus canker symptoms could be observed in infections of the Δ*hrcU* mutant expressing a HrcU~XAC~ variant in which the NPTH site has was abolished. We also show that HrpB2~XAC~ is secreted in a manner that depends on HrcU~XAC~ but is only partly dependent on HrcU~XAC~ cleavage. Expression of HrpB2~XAC~ variants in a Δ*hrpB2* background showed that the last seven amino acids are essential for HrpB2~XAC~ function in the development of canker disease symptoms. Results {#s2} ======= Expression of the cytosolic domain of HrcU~XAC~ (HrcU~XAC\_207--357~) in *E. coli* produces a 7 kDa polypeptide {#s2a} --------------------------------------------------------------------------------------------------------------- The C-terminal domain of HrcU corresponding to residues 207--357 (HrcU~XAC\_207--357~, sequence shown in [Fig. 1A](#pone-0017614-g001){ref-type="fig"}) was expressed in *E. coli* BL21(DE3) cells. The expression of the recombinant protein was expected to produce a 158 residue, 17 kDa polypeptide. However, SDS-PAGE analysis failed to detect a 17 kDa fragment but instead a 7 kDa fragment appeared in Coomassie-stained gels after induction with IPTG (data not shown). This fragment was subsequently purified ([Fig. 1B](#pone-0017614-g001){ref-type="fig"}, lane 1). This result was obtained after expression in a variety of different *E. coli* strains including BL21(DE3), BL21(DE3)RP, BL21(DE3)RIL, BL21(DE3)pLysS, BL21(DE3)CY, BL21(DE3)SI and BL21(DE3)Star (data not shown). ::: {#pone-0017614-g001 .fig} 10.1371/journal.pone.0017614.g001 Figure 1 ::: {.caption} ###### Expression of HrcU~XAC~ C-terminal fragments. **A**) Primary sequence of the C-terminal domain (residues 207-357) of HrcU~XAC~. Residues 207-264 are in *italic* and residues 277-357 are shown in *bold*. The underlined sequence was shown to interact with HrpB2~XAC~ in yeast two-hybrid assays [@pone.0017614-Alegria1]. In HrcU~XAC\_207-264~ and HrcU~XAC\_207-357(AAAH)~, residues Q207 and H208 were replaced with Met and Asp residues respectively. The highly conserved NPTH sequence is double-underlined and the cleavage site between N264 and P265 is indicated with an asterisk. The two tryptophan (W209 and W340) residues are indicated with a dot above their letter symbols. **B**) Coomassie-stained SDS-PAGE of purified recombinant HrcU fragments. Purified HrcU~XAC\_208-264~ (lane 1), HrcU~XAC\_207-357AAAH~ (lane 2) and HrcU~XAC\_His277-357~ (lane 3). Molecular mass markers (M) are shown on the left with masses in kDa. **C**) Western blots of purified HrcU~XAC~ fragments (lanes 3, 6, 9) and of *E. coli* cell lysates before (lanes 2, 5, 8) and after (lanes 1, 4, 7) expression using the polyclonal antiserum raised against HrcU~XAC\_207-357AAAH~. HrcU~XAC\_207-357~ (lanes 1-3), HrcU~XAC\_207-357AAAH~ (lanes 4-6), HrcU~XAC\_His277-357~ (lanes 7-9). :::  ::: Mutation of the conserved NPTH site results in the production of a full-length 17 kDa polypeptide {#s2b} ------------------------------------------------------------------------------------------------- HrcU orthologs and paralogs all possess a conserved Asn-Pro-Thr-His (NPTH) sequence (residues 264--267 in HrcU~XAC~, [Fig. 1A](#pone-0017614-g001){ref-type="fig"}) which has been shown to be a site of auto-cleavage in the flagellar protein FlhB [@pone.0017614-Minamino1], [@pone.0017614-Ferris1]. To test the hypothesis that a similar cleavage was occurring in HrcU~XAC\_207--357~, we mutated residues 264--266 to alanine and expressed the polypeptide in *E. coli*. As shown in [Figure 1B](#pone-0017614-g001){ref-type="fig"} (lane 2), expression and purification of HrcU~XAC\_207--357(AAAH)~ produced a protein of the expected size (17 kDa). To test whether the 7 kDa fragment was in fact derived from HrcU~XAC\_207--357~ we used purified HrcU~XAC\_207--357(AAAH)~ to obtain polyclonal antiserum against the HrcU~XAC~ C-terminal domain. Western blot assays against lysates of *E. coli* cultures obtained before and after IPTG-induced expression of HrcU~XAC\_207--357~ and HrcU~XAC\_207--357(AAAH)~ showed that the antibody recognizes the 17 kDa HrcU~XAC\_207--357(AAAH)~ fragment ([Fig. 1C](#pone-0017614-g001){ref-type="fig"}, lanes 4 and 5) as well as the 7 kDa fragment ([Fig. 1C](#pone-0017614-g001){ref-type="fig"}, lanes 1 and 2). The antibody also recognized the purified 7 kDa fragment ([Fig. 1C](#pone-0017614-g001){ref-type="fig"}, lane 3). These results indicate that the purified 7 kDa fragment obtained after HrcU~207--357~ expression is in fact derived from HrcU~XAC~. N-terminal sequencing by Edman degradation of the 7 kDa fragment was consistent with the N-terminus beginning at position 207 (XXXLFIRDKR), indicating that the initiation Met residue was indeed retained. The mass of the purified 7 kDa fragment determined by MALDI-ToF analysis was very close to the mass expected from the N-terminal fragment (6911 Da with retention of the initiation methionine) produced from cleavage between residues Ans264 and Pro265 of the NPTH sequence within HrcU~XAC\_207--357~. The above results thus allow us to designate the name HrcU~XAC\_207--264~ to the 7 kDa polypeptide that was detected and purified after expression of HrcU~207--357~. Cleavage of HrcU~XAC\_207--357~ between residues Asn264 and Pro265 would be expected to produce two fragments, one N-terminal fragment beginning at residue 207 and ending at residue 264 (6911 Da) and one C-terminal fragment corresponding to residues 265--357 (9931 Da). As mentioned above, only a 7 kDa fragment could be observed to be induced in Coomassie-stained gels (data not shown). While no 10 kDa fragment was observed to be induced in Coomassie-stained gels, a faint band could be observed above the 7 kDa band in the Western blot of *E. coli* lysates after induction of HrcU~207--357~ expression with IPTG ([Fig. 1C](#pone-0017614-g001){ref-type="fig"}, lane 1). Therefore, the evidence so far is consistent with the cleavage at residue 264 possibly followed by a degradation of a significant fraction of the 10 kDa fragment in *E. coli*. Interactions between HrpB2~XAC~ and fragments derived from the cytosolic C-terminal domain of HrcU~XAC~ {#s2c} ------------------------------------------------------------------------------------------------------- We have previously shown that HrpB2~XAC~ interacts with fragments derived from the C-terminal domain of HrcU~XAC~ in yeast two-hybrid assays [@pone.0017614-Alegria1]. In that study, the smallest HrcU~XAC~ fragment observed to interact corresponded to residues 256 to 357 (underlined sequence in [Fig. 1A](#pone-0017614-g001){ref-type="fig"}). It was therefore not clear whether HrcU~XAC~ sequences before or after the conserved NPTH site (or both) were necessary for interaction with HrpB2. We therefore expressed and purified recombinant HrpB2~XAC~ to perform *in vitro* interaction assays with HrcU~XAC\_207--264~ and HrcU~XAC\_207--357(AAAH)~. We also expressed and purified an HrcU~XAC~ fragment corresponding to residues 277--357 with an N-terminal His-tag fusion (HrcU~XAC\_His277--357~) ([Figure 1B](#pone-0017614-g001){ref-type="fig"}, lane 3). This fragment is recognized by polyclonal anti-HrcU~XAC~ antibodies both in *E*. *coli* lysates and after purification ([Figure 1C](#pone-0017614-g001){ref-type="fig"}, lanes 7, 8 and 9) and its estimated mass determined by MALDI-ToF spectrometry corresponds well with the expected mass of a fragment in which the initiation methionine has been retained (data not shown). [Figure 2A](#pone-0017614-g002){ref-type="fig"} shows the results of Far-Western blot analysis of the interaction between HrpB2~XAC~ and HrcU~XAC\_207--357(AAAH)~ using polyclonal antibodies raised against HrpB2~XAC~. HrpB2~XAC~ bound to immobilized HrcU~XAC\_207--357(AAAH)~ (lane 2) but not to an immobilized recombinant C-terminal chicken α-tropomyosin fragment used as a negative control (lane 3). Similar experiments using immobilized HrcU~XAC\_207--264~ failed to detect an interaction (data not shown). [Figure 2B](#pone-0017614-g002){ref-type="fig"} shows that Far-Western assays using immobilized cell lysates obtained before (lane 2) and after induction (lane 1) of HrpB2~XAC~ expression as well as purified HrpB2~XAC~ (lane 3). After incubation of the membranes with HrcU~XAC\_207--357(AAAH)~, bound HrcU~XAC\_207--357(AAAH)~ could be detected with polyclonal anti-HrcU~XAC~ antibodies. Again, no interactions could be detected in similar experiments in which membranes were incubated with HrcU~XAC\_207-264~ (data not shown). ::: {#pone-0017614-g002 .fig} 10.1371/journal.pone.0017614.g002 Figure 2 ::: {.caption} ###### Interaction of HrpB2~XAC~ with HrcU~XAC\_207-357AAAH~ and with HrcU~XAC\_His277-357~. **A**) Far-Western blot assays demonstrating the HrpB2~XAC~ interaction with immobilized HrcU~XAC\_207-357AAAH~. The following purified proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane: Lane 1: HrpB2~XAC~. Lanes 2 and 4: HrcU~XAC\_207-357AAAH.~ Lane 3: chicken muscle tropomyosin fragment Tm~143-284~ [@pone.0017614-Paulucci1]. Nitrocellulose strips corresponding to lanes 2 and 3 were incubated with HrpB2 followed by washing to remove unbound proteins. Nitrocellulose strips corresponding to lanes 1 to 4 were then incubated with polyclonal antiserum raised against HrpB2~XAC~. The strips were rejoined and revealed using anti-mouse IgG conjugated with horseradish peroxidase. **B**) Far-Western Blot assays demonstrating the HrcU~XAC\_207-357AAAH~ interaction with immobilized HrpB2~XAC~. The following samples were separated by SDS-PAGE and transferred to a nitrocellulose membrane: Lysates of *E. coli* cells after (lane 1) and before (lane 2) expression of HrpB2~XAC~ and purified HrpB2~XAC~ (lane 3). The nitrocelulose membrane was incubated with HrcU~XAC\_207-357AAAH~ following by incubation with polyclonal antiserum raised against HrcU~XAC\_207-357AAAH~ and revealed using protein A conjugated with horseradish peroxidase. **C**) Far-Western Blot assays demonstrating the HrpB2~XAC~ interaction with immobilized HrcU~XAC\_His277-357~. The following samples were separated by SDS-PAGE and transferred to a nitrocellulose membrane: Lysates of *E. coli* cells after (lane 1) and before (lane 2) expression of HrcU~XAC\_His277-357~ and purified HrcU~XAC\_His277-357~ (lane 3). The nitrocellulose membrane was incubated with HrpB2 following by incubation with polyclonal antiserum raised against HrpB2~XAC~ and revealed as described in part A. **D**) Pull-down assay demonstrating the interaction of HrpB2~XAC~ with HrcU~XAC\_His277-357~ immobilized on Ni^2+^-chelating resin. HrcU~XAC\_His277-357~ (lane 2), HrpB2~XAC~ (lane 3), a HrpB2~XAC~ plus HrcU~XAC\_His277-357~ mixture (lane 4) and a mixture of HrcU~XAC\_His277-357~ with an *E. coli* BL21(DE3) cell lysate (lane 5) were applied to a Ni^2+^-chelating resin, washed with buffer containing 25 mM imidazole and bound proteins were eluted by washing with 500 mM imidazole. Eluted proteins were separated by SDS-PAGE and visualized by Coomassie brilliant blue staining. Lane 1 shows the contents of the *E. coli* lysate employed in lane 5. Molecular mass markers (M) are shown to the left in kilodaltons. **E**) Since HrpB2~XAC~ and HrcU~XAC\_His277-357~ are not easily separated by SDS-PAGE, the presence of both HrpB2~XAC~ and HrcU~XAC\_His277-357~ in the bound fraction shown in lane 4 of [Figure 2D](#pone-0017614-g002){ref-type="fig"} was demonstrated by Western blot using polyclonal antisera raised against HrcU~XAC\_207-357AAAH~ (lane 1) and against HrpB2~XAC~ (lane 2). The masses of molecular weight markers (in kDa) are indicated to the left of parts A-E. :::  ::: HrcU~XAC\_His277--357~ corresponds to a fragment that begins 10 residues after the conserved NPTH site. Binding of HrcU~XAC\_His277--357~ to HrpB2~XAC~ was demonstrated in Far-Western experiments using *E. coli* lysates obtained after induction of expression of HrpB2~XAC~ as well as purified HrpB2~XAC~. These samples were submitted to SDS-PAGE, transferred to nitrocellulose membranes, overlayed with HrcU~XAC\_His277--357~ and bound HrcU~XAC\_His277--357~ was detected using anti-HrcU~XAC~ antibodies ([Fig. 2C](#pone-0017614-g002){ref-type="fig"}). This interaction was further demonstrated by immobilizing HrcU~XAC\_His277--357~ on a Ni^2+^-chelating resin and testing whether it could retain HrpB2~XAC~ ([Fig. 2D and 2E](#pone-0017614-g002){ref-type="fig"}). While purified HrpB2~XAC~ did not interact with the Ni^2+^-chelating resin on its own ([Fig. 2D](#pone-0017614-g002){ref-type="fig"}, lane 3), it was retained by HrcU~XAC\_His277--357~ bound to the column ([Fig. 2D](#pone-0017614-g002){ref-type="fig"}, lane 4 and [Fig. 2E](#pone-0017614-g002){ref-type="fig"}, lane 2). Since HrpB2~XAC~ (14 kDa) and HrcU~XAC\_His277--357~ (10 kDa) have similar mobility in SDS-PAGE, we detected the individual components of the complex using HrcU~XAC~-specific and HrpB2~XAC~-specific antisera ([Fig. 2E](#pone-0017614-g002){ref-type="fig"}, lanes 1 and 2 respectively). The specific interaction between HrpB2~XAC~ and the region C-terminal to the HrcU~XAC~ NPTH site was further demonstrated in fluorescence perturbation assays. The HrpB2~XAC~ protein does not possess any tryptophan residues. On the other hand, the C-terminal cytosolic domain of HrcU~XAC~ has two tryptophans, one at position 209, before the NPTH site, and the other at position 340, after the NPTH site ([Fig. 1A](#pone-0017614-g001){ref-type="fig"}). We therefore used the intrinsic fluorescence of purified HrcU~XAC\_207--357(AAAH)~, HrcU~XAC\_207--264~ and HrcU~XAC\_His277--357~ as probes to detect interactions with HrpB2. [Figure 3](#pone-0017614-g003){ref-type="fig"} shows that the fluorescence of HrcU~XAC\_207--357(AAAH)~ and HrcU~XAC\_His277--357~ is perturbed by the addition of HrpB2~XAC~ ([Fig. 3B and 3C](#pone-0017614-g003){ref-type="fig"}) while the fluorescence of HrcU~XAC\_207--264~ remains unchanged ([Fig. 3A](#pone-0017614-g003){ref-type="fig"}). The addition of HrpB2 caused slight blue-shifts in the emission spectra of both HrcU~XAC\_207--357(AAAH)~ and HrcU~XAC\_His277--357~ as well as a small increase in intensity. These results confirm that the site of HrpB2~XAC~ interaction on HrcU~XAC~ corresponds to the sequence C-terminal to the NPTH cleavage site. ::: {#pone-0017614-g003 .fig} 10.1371/journal.pone.0017614.g003 Figure 3 ::: {.caption} ###### HrpB2~XAC~ induced changes in HrcU~XAC~ fluorescence. Fluorescence emission spectra of HrcU~XAC\_207-264~ (A), HrcU~XAC\_207-357AAAH~ (B) and HrcU~XAC\_His277-357~ (C) in the absence (dotted lines) and presence (solid lines) of HrpB2. All proteins (2 µM) were dissolved in 5 mM sodium acetate (pH 6.0). Spectra were recorded at 25°C using an excitation wavelength of 280 nm. :::  ::: The NPTH cleavage site is not required for the development of canker symptoms {#s2d} ----------------------------------------------------------------------------- To study the contribution of HrcU~XAC~ and its NPTH site to *Xac* pathogenicity we employed an allelic exchange protocol to the produce the Δ*hrcU Xac* strain containing in-frame deletions of *hrcU* codons 14-347 ([Table 1](#pone-0017614-t001){ref-type="table"}). We also produced plasmids containing the *hrcU* open reading frame plus 1 kb upstream sequences that contain the promoter region (pUFR047\_*hrcU*; [Table 2](#pone-0017614-t002){ref-type="table"} *)*. Furthermore, we introduced mutations in this plasmid that change the NPTH site to AAAH (pUFR047\_*hrcU~AAAH~*). ::: {#pone-0017614-t001 .table-wrap} 10.1371/journal.pone.0017614.t001 Table 1 ::: {.caption} ###### Strains used in this study. ::: {#pone-0017614-t001-1} *Strains* *Relevant characteristics* *Source* ---------------------------------------- --------------------------------------------------------------- ------------------------------ *E. coli* DH10B Recipient for cloning experiments [@pone.0017614-Sambrook1] *E. coli* BL21(DE3) IPTG-inducible T7 RNA polymerase [@pone.0017614-Studier2] *E. coli* BL21(DE3) (RIL) IPTG-inducible T7 RNA polymerase [@pone.0017614-KleberJanke1] *Xac* strain 306 Template for PCR-based cloning [@pone.0017614-daSilva1] *Xac* Δ*hrcU* Xac strain carrying deletion of *hrcU* gene (codons 14--347) This study *Xac* Δ*hrpB2* Xac strain carrying deletion of *hrpB2* gene (codons 10--119) This study *Xac* Δ*hrcU*+pUFR047 \_*hrcU* *Xac* Δ*hrcU* carrying pUFR047\_*hrcU* This study *Xac* Δ*hrcU*+pUFR047\_*hrcU~AAAH~* *Xac* Δ*hrcU* carrying pUFR047\_*hrcU~AAAH~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~1-56~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~1-56~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~1-123~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~1-123~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~LQGPR~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~LQGPR~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~T125A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~T125A~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~L126A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~L126A~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~V127A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~V127A~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~K128A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~K128A~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~N129A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~N129A~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~Q130A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~Q130A~* This study *Strains* *Relevant characteristics* *Source* ---------------------------------------- --------------------------------------------------------------- ------------------------------ *E. coli* DH10B Recipient for cloning experiments [@pone.0017614-Sambrook1] *E. coli* BL21(DE3) IPTG-inducible T7 RNA polymerase [@pone.0017614-Studier2] *E. coli* BL21(DE3) (RIL) IPTG-inducible T7 RNA polymerase [@pone.0017614-KleberJanke1] *Xac* strain 306 Template for PCR-based cloning [@pone.0017614-daSilva1] *Xac* Δ*hrcU* Xac strain carrying deletion of *hrcU* gene (codons 14--347) This study *Xac* Δ*hrpB2* Xac strain carrying deletion of *hrpB2* gene (codons 10--119) This study *Xac* Δ*hrcU*+pUFR047 \_*hrcU* *Xac* Δ*hrcU* carrying pUFR047\_*hrcU* This study *Xac* Δ*hrcU*+pUFR047\_*hrcU~AAAH~* *Xac* Δ*hrcU* carrying pUFR047\_*hrcU~AAAH~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~1-56~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2~1-56~* This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~1-123~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* ~1-123~ This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~LQGPR~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* ~LQGPR~ This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~T125A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* ~T125A~ This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~L126A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* ~L126A~ This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~V127A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* ~V127A~ This study *Xac* Δ*hrpB2*+pUFR047\_*hrpB2~K128A~* *Xac* Δ*hrpB2* carrying pUFR047\_*hrpB2* ~K128A~ This study \*See [Table 2](#pone-0017614-t002){ref-type="table"} for plasmid construction details. ::: [Figure 4A](#pone-0017614-g004){ref-type="fig"} shows the results of inoculation of sweet orange leaf tissue with the *Xac* wild-type, Δ*hrcU, ΔhrcU+*pUFR047*\_hrcU and ΔhrcU+*pUFR047\_*hrcU~AAAH~* strains 15 days after infection. While infection with the wild-type strain showed clear disease symptoms including water-soaking, hyperplasy and necrosis, the Δ*hrcU* strain failed to produce any disease symptoms in the susceptible citrus host. This result is consistent with the absolute requirement for HrcU homologs for the functioning of all T3SS systems [@pone.0017614-Oh1], [@pone.0017614-Allaoui1]. The phenotype of the Δ*hrcU* strain could be reverted by the expression of wild-type HrcU coded by the pUFR047\_*hrcU* plasmid or by expression of the HrcU~XAC\_AAAH~ coded by the pUFR047\_*hrcU~AAAH~* plasmid ([Fig. 4A](#pone-0017614-g004){ref-type="fig"}). In both cases, canker symptoms were less severe than those observed using the wild-type strain. It is not clear why the reversion of disease symptoms was attenuated in these experiments. We note that the native upstream promoter regions contained within these plasmids contain PIP (plant-inducible promoter) boxes [@pone.0017614-daSilva1] that have been shown to be recognized by the HrpX transcription factor that controls *hrp* expression in *Xcv* [@pone.0017614-AstuaMonge1], [@pone.0017614-Fenselau1], [@pone.0017614-Koebnik1]. ::: {#pone-0017614-g004 .fig} 10.1371/journal.pone.0017614.g004 Figure 4 ::: {.caption} ###### HrcU and HrpB2 contribute to *Xac* pathogenicity during infection of *Citrus sinensis*. Macroscopic symptoms 15 days after inoculation on the abaxial surface of leafs with Δ*hrcU* (**A**) and Δ*hrpB2* (**B-D**) mutants. The following strains were used: *Xac* wild-type (WT), Δ*hrcU*, Δ*hrcU*+pUFR047\_*hrcU* (Δ*hrcU*+*hrcU*), Δ*hrcU*+pUFR047\_*hrcU~AAAH~* (Δ*hrcU*+*hrcU~AAAH~*), Δ*hrpB2,* Δ*hrpB2+*pUFR047*\_hrpB2* (Δ*hrpB2+hrpB2)*, Δ*hrpB2+*pUFR047*\_hrpB2~1-56~* (Δ*hrpB2+hrpB2~1-56~*), Δ*hrpB2+*pUFR047*\_hrpB2~1-123~* (Δ*hrpB2+hrpB2~1-123~*), Δ*hrpB2+pUFR047\_hrpB2~LQGPR~* (Δ*hrpB2+hrpB2~LQGPR~)*, Δ*hrpB2+*pUFR047*\_hrpB2~T125A~* (Δ*hrpB2+hrpB2~T125A~*), Δ*hrpB2+*pUFR047*\_hrpB2~L126A~* (Δ*hrpB2+hrpB2~L126A~*), Δ*hrpB2+*pUFR047*\_hrpB2~V127A~* (Δ*hrpB2+hrpB2~V127A~*), Δ*hrpB2+*pUFR047*\_hrpB2~K128A~* (Δ*hrpB2+hrpB2~K128A~*), Δ*hrpB2+*pUFR047*\_hrpB2~N129A~* (Δ*hrpB2+hrpB2~N129A~*) and Δ*hrpB2+*pUFR047*\_hrpB2~L130A~* (Δ*hrpB2+hrpB2~L130A~*). :::  ::: The HrpB2~XAC~ C-terminal region is required to elicit citrus canker symptoms {#s2e} ----------------------------------------------------------------------------- To study the contribution of HrpB2~XAC~ to *Xac* pathogenicity, the allelic exchange protocol was used to produce the Δ*hrpB2* strain with an in-frame deletion of *hrpB2* codons 10-119 ([Table 1](#pone-0017614-t001){ref-type="table"}). We also produced plasmid pUFR047\_*hrpB2* ([Table 2](#pone-0017614-t002){ref-type="table"}) which codes for the wild-type HrpB2~XAC~ protein plus a 1 kb upstream region that includes the *hrpB1* gene between the promoter and *hrpB2*. [Figure 4B](#pone-0017614-g004){ref-type="fig"} shows that the Δ*hrpB2* strain was unable to elicit disease symptoms and that the virulence of the mutant strain was fully restored by transformation with pUFR047\_*hrpB2*. ::: {#pone-0017614-t002 .table-wrap} 10.1371/journal.pone.0017614.t002 Table 2 ::: {.caption} ###### Plasmids used in this study. ::: {#pone-0017614-t002-2} *Plasmids* *Relevant characteristics* *Source* ---------------------------- ------------------------------------------------------------------------------------------ ----------------------------------------------- pET-11d T7 RNA polymerase - based expression vector [@pone.0017614-Studier1] pET-3a T7 RNA polymerase - based expression vector [@pone.0017614-Studier1] pET-28a (+) T7 RNA polymerase - based expression vector Novagen pU1 pET-11d based vector expressing HrcU~XAC\_207-357~ This study pU2 pET-11d based vector expressing HrcU~XAC\_207-357(AAAH)~ This study pU3 pET-28a(+) based vector expressing HrcU~XAC\_His277-357~ This study pB2 pET-3a based vector expressing HrpB2~XAC~ This study pET-Tmy~143-284~ pET-3a based vector expressing chicken alpha tropomyosin [@pone.0017614-Paulucci1] pNPTS138 Suicide vector, Km^r^/*SacB* Dickon Alley[\*](#nt102){ref-type="table-fn"} pUFR047 Wide host range vector, Gm^r^ [@pone.0017614-DeFeyter1] pBBR1MCS-5 Wide host range vector, Gm^r^ [@pone.0017614-Kovach1] pNPTS138\_Δ*hrcU* Suicide vector carrying internal truncation of *Xac hrcU* gene This study pNPTS138\_Δ*hrpB2* Suicide vector carrying internal truncation of *Xac hrpB2* gene This study pBBR\_*hrcU* pBBR1MCS-5 vector carrying *Xac hrcU* gene This study pBBR\_*hrcU~AAAH~* pBBR1MCS-5 vector carrying *Xac hrcU* gene with mutations that change NPTH motif to AAAH This study pUFR047\_*hrcU* pUFR047 based vector for expression of HrcU~XAC~ in *Xac* This study pUFR047\_*hrcU~AAAH~* pUFR047 based vector for expression of HrcU~XAC\_AAAH~ in *Xac* This study pBBR\_*hrpB2* pBBR1MCS-5 vector carrying *Xac hrpB2* gene This study pUFR047\_*hrpB2* pUFR047 based vector for expression of HrpB2~XAC~ in *Xac* This study pUFR047\_*hrpB2~1~* ~-56~ pUFR047 based vector for expression of HrpB2~XAC\_1-56~ in *Xac* This study pUFR047\_*hrpB2~1~* ~-123~ pUFR047 based vector for expression of HrpB2~XAC\_1-123~ in *Xac* This study pUFR047\_*hrpB2~LQGPR~* pUFR047 based vector for expression of HrpB2~XAC\_LQGPR~ in *Xac* This study pUFR047\_*hrpB2~T125A~* pUFR047 based vector for expression of HrpB2~XAC\_T125A~ in *Xac* This study pUFR047\_*hrpB2~L126A~* pUFR047 based vector for expression of HrpB2~XAC\_L126A~ in *Xac* This study pUFR047\_*hrpB2~V127A~* pUFR047 based vector for expression of HrpB2~XAC\_V127A~ in *Xac* This study pUFR047\_*hrpB2~K128A~* pUFR047 based vector for expression of HrpB2~XAC\_K128A~ in *Xac* This study pUFR047\_*hrpB2~N129A~* pUFR047 based vector for expression of HrpB2~XAC\_N129A~ in *Xac* This study pUFR047\_*hrpB2~Q306A~* pUFR047 based vector for expression of HrpB2~XAC\_Q130A~ in *Xac* This study \*unpublished. ::: Multiple sequence alignment analysis of HrpB2 proteins from *Xanthomonas*, *Burkholderia*, *Acidovorax* and *Ralstonia* species ([Figure 5](#pone-0017614-g005){ref-type="fig"}) indicates that there are two regions of sequence conservation in an otherwise variable protein family: i) a five residue motif which we name FQALM that corresponds to positions 35--39 of HrpB2~XAC~ and ii) the last six amino acids of the protein (HrpB2~XAC~ residues 125--130) which we name the TLMKNQ motif (in *Xac* the methionine residue is substituted with a valine). ::: {#pone-0017614-g005 .fig} 10.1371/journal.pone.0017614.g005 Figure 5 ::: {.caption} ###### Graphical Representation of the multiple sequence alignment of the HrpB2 protein family. The Pfam database [@pone.0017614-Finn1] lists 61 sequences in this group (PF09487) from *Xanthomonas* (13 sequences), *Burkholderia* (43 sequences), *Ralstonia* (3 sequences), and *Acidovorax* (2 sequences) species. However, after removal of all sequences with greater that 95% identity, only 16 remain. These 16 sequences were used to generate this representation using the WebLogo server (<http://weblogo.berkeley.edu/>) [@pone.0017614-Crooks1] in which the height of the residue symbol indicates the degree of conservation (the representation obtained using all 61 sequences is highly similar). Numbers refer to residue positions in HrpB2~XAC~. The FQALM and TLMKNQ motifs are underlined. :::  ::: In order to determine whether either or both of these motifs is important for HrpB2~XAC~ function in the elicitation of citrus canker symptoms, we expressed HrpB2~XAC~ fragments or full-length HrpB2~XAC~ variants ([Table 1](#pone-0017614-t001){ref-type="table"}) in the Δ*hrpB2* strain. To test the importance of the FQALM motif we mutated these residues to LQGPR and expressed the mutant protein (HrpB2~XAC\_LQGPR~) in the Δ*hrpB2* strain using the pUFR047*\_hrpB2~LQGPR~* plasmid. The Δ*hrpB2+*pUFR047*\_hrpB2~LQGPR~* strain was able to cause citrus canker symptoms in a manner indistinguishable from the wild-type *Xac* strain ([Fig. 4B](#pone-0017614-g004){ref-type="fig"}). Therefore, the FQALM motif does not seem to be essential for HrpB2 function. When the Δ*hrpB2* strain was transformed with plasmids pUFR047*\_hrpB2~1-56~* and pUFR047*\_hrpB2~1-123~*, leading to the expression of HrpB2~XAC\_1-56~ and HrpB2~XAC\_1-123~ respectively, neither of the resulting strains were able to induce citrus canker symptoms in orange leaves ([Fig. 4C](#pone-0017614-g004){ref-type="fig"}). These results suggested that the C-terminal region of HrpB2~XAC~ which contains the conserved TLMKNQ motif is important for HrpB2 function. To test the importance of each residue in this motif, six *hrpB2* ~XAC~ mutants in which each of these residues were changed to alanine were expressed the Δ*hrpB2* strain. The results showed while the *Xac* strain expressing HrpB2~XAC\_T125A~ was not able produce canker symptoms, the strains expressing HrpB2~XAC\_V127A~, HrpB2~XAC\_K128A~, HrpB2~XAC\_N129A~ and HrpB2~XAC\_Q130A~ produced canker symptoms to the same extent as wild-type *Xac*. Furthermore, Δ*hrpB2* cells expressing H*rpB2~XAC\_L126A~* produced attenuated citrus canker symptoms when compared to the same cells containing the plasmid that expresses wild-type HrpB2~XAC~ ([Fig. 4D](#pone-0017614-g004){ref-type="fig"}). These results point to the importance of the TLMKNQ motif, and especially to the first residue of this motif (T125), in the role of HrpB2~XAC~ in the development disease in citrus plants. HrpB2~XAC~ is secreted by *Xac* in liquid media {#s2f} ----------------------------------------------- Rossier *et al.* [@pone.0017614-Rossier1], showed that in *Xcv* HrpB2 is secreted in a T3SS-dependent manner. In that study, a mutant *Xcv* strain with constitutive expression of the *hrp* locus (due to a constitutively activated HrpG mutation) was used. No such mutant *Xac* strain has yet been isolated or produced. Expression of *hrp* genes in *Xcv* is dependent on unknown plant signals and is controlled by specific promoters with PIP boxes [@pone.0017614-AstuaMonge1], [@pone.0017614-Fenselau1], [@pone.0017614-Koebnik1]. In *Ralstonia solanacearum, hrp* expression is dependent on contact with an unidentified component derived from the host cell wall [@pone.0017614-Aldon1], [@pone.0017614-Brito1], [@pone.0017614-Marenda1] and passion fruit leaf extracts have been shown to modify the proteome of *X. axonopodis* pv. passiflorae [@pone.0017614-Tahara1]. We therefore grew liquid *Xac* cultures in the presence of extracts derived from sweet orange (*C. sinensis*) leaves. Proteins in the secreted fraction were separated by SDS-PAGE and probed for HrpB2~XAC~ by Western blot analysis using anti-HrpB2~XAC~ antiserum. We found that HrpB2~XAC~ could be observed in the secreted fraction of wild-type cells ([Fig. 6A](#pone-0017614-g006){ref-type="fig"}, lane 1). As expected, secretion of HrpB2~XAC~ was abolished in the Δ*hrpB2* mutants and complementation with pUFR047\_*hrpB2* restored HrpB2~XAC~ secretion ([Fig. 6A](#pone-0017614-g006){ref-type="fig"}, lanes 2 and 3, respectively). We did not detect HrpB2~XAC~ in the cellular fractions (data not shown) but note that the secreted fraction was concentrated 60-fold in relation to the cellular fraction (see Experimental Procedures). ::: {#pone-0017614-g006 .fig} 10.1371/journal.pone.0017614.g006 Figure 6 ::: {.caption} ###### HrpB2~XAC~ is secreted by *Xac*. Liquid cultures of *Xac* were grown as described in [Materials and Methods](#s4){ref-type="sec"}. Secreted fractions were concentrated and separated by SDS-PAGE 18% and proteins were transferred to nitrocellulose membranes. HrpB2~XAC~ was detected using anti-HrpB2~XAC~ antiserum and revealed using anti-mouse IgG conjugated with horseradish peroxidase. (**A**) Lane 1: *Xac* wild-type, lane 2: *Xac* Δ*hrpB2,* lane 3: *Xac* Δ*hrpB2+*pUFR047\_*hrpB2*, lane 4: *Xac ΔhrpB2+*pUFR047*\_hrpB2~1-56~*, lane 5: *Xac* Δ*hrpB2+*pUFR047*\_hrpB2~1-123~*, lane 6: *Xac* Δ*hrpB2+*pUFR047*\_hrpB2~LQGPR~*. (**B**) Lane 1: *Xac* wild-type, lane 2: *Xac ΔhrpB2+*pUFR047*\_hrpB2~T125A~*, lane 3: *Xac* Δ*hrpB2+*pUFR047*\_hrpB2~Q130A~,* lane 4: *Xac* Δ*hrpB2*. (**C**) Lane 1: *Xac* wild-type, lane 2: *Xac* Δ*hrcU*, lane 3: *Xac* Δ*hrcU*+pUFR047\_*hrcU* and lane 4: *Xac* Δ*hrcU*+pUFR047\_*hrcU~AAAH~*. :::  ::: We then asked whether the HrpB2~XAC~ mutants described above were secreted when expressed in the Δ*hrpB2* strain. [Figure 6A](#pone-0017614-g006){ref-type="fig"} (lanes 4, 5 and 6) shows that HrpB2~1-56~ (5.7 kDa), HrpB2~1-123~ (13 kDa) and HrpB2~XAC\_LQGPR~ were all observed in *Xac* culture supernatants. Furthermore, all six mutants carrying alanines at each position of the TLMKNQ motif could be detected in *Xac* culture supernatants ([Fig. 6B](#pone-0017614-g006){ref-type="fig"}, lanes 2 and 3 and data not shown). Finally, we observed that HrpB2~XAC~ secretion was abolished in the Δ*hrcU* mutant ([Fig. 6C](#pone-0017614-g006){ref-type="fig"}, lane 2). Complementation of the Δ*hrcU* mutant with pUFR047\_*hrcU* restored HrpB2~XAC~ secretion to wild-type levels ([Fig. 6C](#pone-0017614-g006){ref-type="fig"}, lane 3). Interestingly, complementation of the Δ*hrcU* mutant with pUFR047\_*hrcU~AAAH~* resulted in significantly reduced levels of HrpB2~XAC~ secretion ([Fig. 6](#pone-0017614-g006){ref-type="fig"}, lane 4). This difference in levels of HrpB2~XAC~ secretion may, therefore, be due to the inability of the HrcU~XAC\_AAAH~ protein to undergo the self-cleavage reaction. Apparently, only minimal amounts of HrpB2~XAC~ are necessary to elicit citrus canker symptoms during the infection process. HrcU~XAC~ is not required for *Xac* survival *in planta* {#s2g} -------------------------------------------------------- In order to determine whether HrcU~XAC~ and its NPTH site were necessary for *Xac* survival in inoculated host leafs, we inoculated Citrus leafs with *Xac* bacterial suspensions and accompanied bacterial numbers during a 12 day period after infection. [Figure 7A](#pone-0017614-g007){ref-type="fig"} shows that wild-type *Xac*, Δ*hrcU*, Δ*hrcU*+pUFR047\_*hrcU*, Δ*hrcU*+pUFR047\_*hrcU~AAAH~* strains all presented similar growth curves. This suggests that HrcU~XAC~ is not absolutely required for bacterial survival *in planta*, in spite of the fact that the Δ*hrcU* does not produce canker disease symptoms. In contrast, the Δ*hrpB2* strain presented significantly reduced survival when compared to the wild-type and Δ*hrpB2*+pUFR047\_*hrpB2* strains ([Fig. 7B](#pone-0017614-g007){ref-type="fig"}). ::: {#pone-0017614-g007 .fig} 10.1371/journal.pone.0017614.g007 Figure 7 ::: {.caption} ###### Number of colony-forming units (CFU) of *Xac* strains per cm^2^ of leaf tissue during the first twelve days after inoculation. The abaxial surface of young leaves was pricked by using insect pins whose tips were previously immersed in the bacterial suspension for *Xac hrcU* mutant strains (A) or by infiltration into leaves with needleless syringes for *Xac hrpB2* mutant strains (B). Discs of infected leaves were excised, homogenized and cultured quantitatively by incubation on agar plates. The assays were performed in triplicate and error bars represent the standard deviation of the data. Differences in the initial bacterial populations are due to differences in the inoculation protocols. (A) *Xac* wild-type (diamonds), Δ*hrcU* (squares), *ΔhrcU*+pUFR047\_*hrcU* (triangles), Δ*hrcU*+pUFR047\_*hrcU~AAAH~* (crosses). (B) *Xac* wild-type (diamonds), Δ*hrpB2* (squares), *ΔhrpB2*+pUFR047\_*hrpB2* (triangles). :::  ::: Discussion {#s3} ========== In this study we constructed non-polar knock-out mutants for the *hrcU* and *hrpB2* genes and show that they completely abolish pathogenicity of *Xac* in sweet orange. Complementation of the Δ*hrcU* strain with plasmids pUFR047\_*hrcU* or pUFR047\_*hrcU~AAAH~* recovered the capacity to induce disease symptoms. We also demonstrated that HrpB2~XAC~ is secreted to the extracellular space in a HrcU~XAC~-dependent manner by the *Xac* T3SS. HrpB2~XAC~ secretion was abolished in the Δ*hrcU* knockout and restored in the Δ*hrcU*+pUFR047\_*hrcU* and Δ*hrcU*+pUFR047\_*hrcU~AAAH~* strains, but the amount of HrpB2~XAC~ secreted by the Δ*hrcU*+pUFR047\_*hrcU~AAAH~* strain was reduced with respect to that observed for the wild-type and Δ*hrcU*+pUFR047\_*hrcU* strains ([Figure 6C](#pone-0017614-g006){ref-type="fig"}). This result suggests that while HrcU~XAC~ cleavage may not be absolutely necessary for the proper functioning of the *Xac* T3SS, it may contribute to the efficiency by which it carries out its tasks. In this report we have shown that HrcU~XAC~ expressed in *E. coli* suffers proteolysis at a highly conserved NPTH site in a manner similar to that already described for its paralog FlhB of the flagellar system [@pone.0017614-Minamino1] and its orthologs YscU, EscU and SpaS from the T3SSs of *Yersina* [@pone.0017614-Lavander1], [@pone.0017614-Riordan1], *E. coli* and *Salmonella* [@pone.0017614-Zarivach1] respectively. This, and a similar report for the HrcU protein from *Xanthomonas campestris* pv. *vesicatoria* [@pone.0017614-Lorenz1], are the first observations of NPTH-dependent cleavage of a FlhB homolog from the T3SS of a plant pathogen. Ferris *et al.* [@pone.0017614-Ferris1] have shown that FlhB cleavage at the NPTH site is an autocatalytic process; that is, FlhB catalyzes its own hydrolysis at this site. Furthermore, a series of crystal structures of the C-terminal domains of the FlhB homologs EscU and SpaS from the *E. coli* and *Salmonella* T3SSs [@pone.0017614-Zarivach1], and YscU from *Yersinia enterocolitica* [@pone.0017614-Wiesand1] have recently provided information regarding the mechanism and conformational changes associated with self-cleavage. We also show that the HrcU~XAC~ C-terminal fragment that is released upon HrcU~XAC~ self-cleavage interacts with HrpB2~XAC~, whose only known homologs are found in the phytopathogens *Xanthomonas* spp., *Ralstonia solanacearum*, and *Acidovorax avenae*, as well as *Burkholderia* spp that infect both animals and plants. Our results show that HrpB2~XAC~ does not interact specifically with the site of HrcU~XAC~ cleavage since it could bind to HrcU~XAC\_207-357(AAAH)~ and to HrcU~His277-357~, a fragment that begins 10 residues after the NPTH site. Unfortunately we were not able to detect HrcU~XAC~ in the wild-type or complemented mutant strains using the anti-HrcU~XAC~ polyclonal antibodies in this study (data not shown) and so could not determine relative levels of HrcU~XAC~ in the *Xac* strains nor have we so far been able to determine whether HrcU~XAC~ is in fact cleaved at the NPTH site in *Xac* cells. However, during the preparation of this work, HrcU cleavage was observed in *Xcv* [@pone.0017614-Lorenz1]. In order to understand HrcU function it is useful to recall what we know about the functioning of HrcU homologs. Mutants that inhibit cleavage at the NPTH site of HrcU homologs exhibit defects in the secretion of specific substrates. For example, in FlhB, mutations at this site inhibit the export of "late" flagellar proteins, while normal levels of early substrates, including hook protein FlgE, are secreted [@pone.0017614-Fraser1]. Also, a *Y. enterocolitica* Δ*yscU* strain expressing YscU~N263A~, in which the conserved Asn residue of the NPTH sequence was mutated to Ala, produced longer needles, exported reduced amounts of YscP (a FliK homolog, see below) and did not export the translocator proteins LcrV, YopB and YopD. The first two defects could be compensated by overexpression of YscP (see below) while export of LcrV, YopB and YopD was absolutely dependent on a cleavable NPTH site [@pone.0017614-Sorg1]. The cleaved FlhB C-terminal fragment binds to both early and late flagellar export substrates (FlgD, FliC). Furthermore, the product of the *fliK* gene, FliK or flagellar hook-length control protein, binds to the self-cleavage C-terminal fragment of FlhB [@pone.0017614-Minamino1] and during flagellar assembly FliK is itself secreted subsequent to hook protein subunit secretion [@pone.0017614-Macnab1], [@pone.0017614-Minamino2]. Also, *fliK* mutants do not secrete late substrates but do secrete excessive amounts of hook protein (FlgE), resulting in the production of characteristic polyhooks [@pone.0017614-Williams1]. This phenotype can be reverted by single amino acid substitutions in FlhB, almost all of which map to the C-terminal self-cleavage fragment [@pone.0017614-Fraser1], [@pone.0017614-Williams1], [@pone.0017614-Kutsukake1], [@pone.0017614-Suzuki1]. Thus, in the flagellar system, FlhB and FliK act together to control substrate switching from early to late substrates, though the molecular mechanism by which this is achieved is not fully understood [@pone.0017614-Macnab1], [@pone.0017614-Macnab2]. In the animal pathogens *Yersinia*, *Salmonella* and *Shigella*, the formation of needle complexes and subsequent secretion of virulence factors by T3SSs are controlled by an interplay between FlhB and FliK homologs. In these systems, mutations in the FliK homologs YscP [@pone.0017614-Edqvist1], [@pone.0017614-Journet1], InvJ [@pone.0017614-Kubori1] or Spa32 [@pone.0017614-Tamano1] result in the formation of needles of variable length and compromised virulence factor secretion. In the case of *Yersinia*, the phenotypes can be reverted by mutations in the cytosolic domains of the FlhB homolog YscU [@pone.0017614-Edqvist1]. Futhermore, YscP, InvJ and Spa32 are secreted during T3SS assembly [@pone.0017614-Tamano1], [@pone.0017614-Collazo1], [@pone.0017614-Payne1] in a manner similar to the secretion of FliK in the flagellar system. Finally, in *Yersinia*, YscP secretion appears to be coupled to the secretion of another small protein (YscO) that binds preferentially to the uncleaved form of YscU [@pone.0017614-Riordan2]. Few YscP homologs from non-flagellar T3SS have been identified in plant-associated bacteria: HrpP from *Pseudomonas syringae,* RspP from *P. fluorescens,* HpaP from *R. solanacearum* [@pone.0017614-Agrain1] and the HpaP/HpaC proteins coded by the *hrp* gene clusters of *Xanthomonas spp* (for example HpaC in *Xcv* and HpaP in *Xac*). During the preparation of this work Lorenz *et al.* [@pone.0017614-Lorenz1] published a study on the HpaC and HrpB2 proteins from *Xcv*. They found that: 1) amino acids 10 to 25 of HrpB2 are crucial for its efficient secretion and function and that HrpB2 is necessary for the secretion of effectors and of extracellular components of the secretion apparatus, 2) HrpB2 and HpaC interact with each other and both also interact with the C-terminal domain of HrcU and 3) HrpB2 secretion is suppressed by HpaC. They therefore speculated that HpaC acts to control the switch between the secretion of early to late T3SS substrates (see also reference [@pone.0017614-Rossier1]) and that HpaC binding to HrcU specifically inhibits HrpB2 binding and secretion [@pone.0017614-Lorenz1]. While HpaC from *Xcv* has been shown to be necessary for the secretion of both T3SS effector and translocon proteins, it is not required for the export of the Hrp pilus protein HrpE [@pone.0017614-Buttner3]. In this sense, the *hpaC* mutant phenotype in *Xcv* is similar to that observed for *yscP*, *invJ* and *spa32* mutants (see above). However, HpaC itself is not secreted by *Xcv* and Hrp pilus formation was not affected in *hpaC* mutant strains (different from that observed for *yscP*, *invJ* and *spa32* mutants as described above) [@pone.0017614-Buttner3]. On the other hand, HrpB2 binds to HrcU and, like FliK and YscP, HrpB2 is itself secreted. Since both HrpB2 and HpaC bind to the C-terminal domain of HrcU, the accumulated evidence so far is not clear as to which (if either) HpaC-HrcU or HrpB2-HrcU complexes carry out molecular functions in the *Xanthomonas* T3SS that are orthologous to those of YscU-YscP and FlhB-FliK described above. One interesting observation from our study was that while both Δ*hrcU* and Δ*hrpB2* knockout strains do not induce citrus canker symptoms, only the latter presents a significant reduction in survival in the host tissue. The Δ*hrcU* mutant survives as well as the wild type strain in the host tissue, but does not detectably secrete HrpB2. A similar phenomena has been observed in *X. campestris* pv. *glycines* 8ra where HrcU is required for pathogenicity in its natural soybean host but is not required for multiplication in the host plant nor is it required for the induction of HR in non-hosts [@pone.0017614-Oh1]. The molecular basis for the differences in the Δ*hrcU* and Δ*hrpB2* phenotypes in *Xac* is not yet clear. One possibility is that the ΔhrcU mutant fails to secrete effector(s) that trigger specific host defense mechanisms resulting in the bacterial survival. Another possibility is that intracellular HrpB2 may contribute to *Xac* survival while extracelular HrpB2 contributes to citrus canker symptom development. In *Xcv*, deletion of HrpB2 residues 10-25 impaired protein secretion and disease symptom formation, which led to the conclusion that secretion is required for function [@pone.0017614-Lorenz1]. We demonstrated that while Δ*hrpB2*+pUFR047\_*hrpB2~1-56~*, Δ*hrpB2*+ pUFR047\_*hrpB2~1-123~* and Δ*hrpB2*+pUFR047\_*hrpB2~T125A~* strains are not able to cause citrus canker, the truncated HrpB2~XAC~ polypeptides and HrpB2~XAC~ single amino acid substitution mutants are all however secreted to the extracellular space. Therefore, HrpB2~XAC~ secretion, *per se*, is not sufficient for HrpB2~XAC~ function. Apparently, the conserved C-terminal region of the protein, more specifically residue T125 in the conserved TLMKNQ motif, is especially important for HrpB2~XAC~-dependent pathogenicity. Important unanswered questions remain regarding HrpB2 function at the molecular level. Further studies are needed to determine whether HrpB2 exercises a role in substrate switching or as a minor structural component of the T3SS pilus (as do hook-filament junction and capping proteins in bacterial flagella) or carries out other, as yet not contemplated, functions and also whether these functions are effected within the bacterial cell or in the exterior subsequent to its secretion (or both). Materials and Methods {#s4} ===================== Construction vectors for the expression of HrcU~XAC\_207-357~, HrcU~XAC\_His277-357~, HrcU~XAC\_207-357(AAAH)~ and HrpB2~XAC~ in *E. coli* {#s4a} ------------------------------------------------------------------------------------------------------------------------------------------ *E. coli* strains and plasmids are described in [Table 1](#pone-0017614-t001){ref-type="table"} and [Table 2](#pone-0017614-t002){ref-type="table"}, respectively. *E. coli* cells were cultivated at 37°C in 2xYT media [@pone.0017614-Sambrook1]. When necessary, the appropriate antibiotics were added at the following final concentrations: ampicillin 200 µg/ml, kanamycin 50 µg/ml and chloramphenicol 200 µg/ml. Synthetic oligonucleotide primers ([Table 3](#pone-0017614-t003){ref-type="table"}) for polymerase chain reactions (PCR) were designed containing restriction sites useful for cloning (see below). PCR products were purified from agarose gels using the QIAquick Gel Extraction Kit (Qiagen). To produce a vector for the expression of HrcU~XAC\_207-357~, the DNA sequence coding residues 207-357 of the *hrcU* gene was amplified from genomic *Xac* DNA using the oligonucleotides F-U~207-357~ and R-U~207-357~ ([Table 3](#pone-0017614-t003){ref-type="table"}). The PCR product was digested with endonucleases NcoI and HindIII and inserted into the expression vector pET-11d [@pone.0017614-Studier1], previously digested with the same enzymes to produce plasmid pU1. Primers F-U~AAAH~ and R-U~AAAH~ ([Table 3](#pone-0017614-t003){ref-type="table"}) were used in PCR with pU1 as template in order to change the codons for residues 264-266 to alanine codons using the QuikChange Site-Directed Mutageneis Kit (Stratagene). The resulting recombinant plasmid (pU2) directs the expression of HrcU~XAC\_207-357(AAAH)~. Note that in both recombinant HrcU~XAC\_207-357~ (through which HrcU~XAC\_207-264~ is purified, see below) and HrcU~XAC\_207-357(AAAH)~, residues Gln207 and His208 have been mutated to Met and Asp residues, respectively, due to the introduction of restriction sites used in the cloning protocol. To produce a vector for the expression of HrcU~XAC\_His277-357~, the sequence coding for HrcU residues 277-357 was amplified using primers F-U~His277-357~ and R-U~His277-357~ ([Table 3](#pone-0017614-t003){ref-type="table"}). This product was digested with NdeI and HindIII and ligated into the expression vector pET-28a (Novagen) previously digested with the same enzymes to produce the recombinant plasmid pU3. To produce a vector for the expression of full-length HrpB2~XAC~, the expression vector pET-3a (Studier *et al.,* 1990) was digested with HindIII, filled in with the Klenow fragment of *E. coli* DNA polymerase I and then digested with NdeI. Primers F-B2 and R-B2 were used in a PCR with *Xac* genomic DNA, the product was treated with Klenow fragment and polynucleotide kinase to produce blunt ends, digested with NdeI and then ligated into the pET-3a vector described above to produce the recombinant plasmid pB2. The accession numbers for the complete Xac genome sequence and the HrpB2~XAC~ and HrcU~XAC~ protein sequences are NC\_003919, NP\_640763 and NP\_640761, respectively. ::: {#pone-0017614-t003 .table-wrap} 10.1371/journal.pone.0017614.t003 Table 3 ::: {.caption} ###### Oligonucleotides used in this study. ::: {#pone-0017614-t003-3} *Oligonucleotides* *Sequence* -------------------- ----------------------------------------------------- F-U~207-357~ 5′ CATCCCATGGACTGGCTGTTCATCCGGGAC 3′ R-U~207-357~ 5′ CCCAAGCTTCTCGAGGCTCGCACGCGATCTCCTAG 3′ F-U~AAAH~ 5′ GTGATGGTGGTCGCCGCGGCCCATTACGCGGTGGCAC R-U~AAAH~ 5′ GTGCCACCGCGTAATGGGCCGCGGCGACCACCATCAC 3′ F-U~His277-357~ 5′ TAAATTGCTCATATGGATGACTTCGGCCTA 3′ R-U~His277-357~ 5′ TAAATTGCTCCATGGATGACTTCGGCCTA 3′ F-B2 5′ CGGAATTCCATATGACGCTCATTCCTCCTGTC 3′ R-B2 5′ CCGCTCGAGCTATTGGTTCTTGACCAGTGTCTG 3′ F1-U 5′ TCGGGACTAAAGCTTGCATCAACT TGATCT 3′ R1-U 5′ GGAATTACCATATGCAGTTTCTTCTCGGTCGGCTTCTC 3′ F2-U 5′ GGAATTACCATATGCACAGCGACGGCGATGGAGCT 3′ R2-U 5′ TTTGAACTTGCTAGCTGATCGGTGCCGCTG 3′ R-compU 5′ ATTTTAAGCTTGTCGACCTAGCATGGCAGAGCTCC 3′ F1-B2 5′ CACTACAAGCTTAAGCAACCAGCAAGGGGA 3′ R1-B2 5′ GGAATTACCATATGAATCGCTTGGACAGGAGGAAT 3′ F2-B2 5′ GGAATTACCATATGAAGAACGCCGTGCAGACACTG 3′ R2-B2 5′ AACATTAAATCTAGAGTCGACTGGTTCGCATGCAGGCCGAGC 3′ R-compB2 5′ AATTTAAGCTTGTCGACCTATTGGTTCTTGACCAGTGTC3′ F-M57 5′GCAGCGAGTGGGCAACCCGAGCTAGATGAGCCGCGTGGTCGATGTGC3′ R-M57 5′GCACATCGACCACGCGGCTCATCTAGCTCGGGTTGCCCACTCGCTGC3′ F-Q124 5′GCAATCGGGAAAGAACGCAGTGTAGACACTGGTCAAGAATCAATAG3′ R-Q124 5′CTATTGATTCTTGACCAGTGTCTACACTGCGTTCTTTCCCGATTGC3′ F-FQALM 5′CGCTAGTGAATCGCTTACAAGGGCCGAGGCAGTCCTCTAGC3′ R-FQALM 5′GCTAGAGGACTGCCTCGGCCCTTGTAAGCGATTCACTAGCG3′ F-T125A 5′GGAAAGAACGCAGTGCAGGCACTGGTCAAGAATCAATAG3′ R-T125A 5′CTATTGATTCTTGACCAGTGCCTGCACTGCGTTCTTTCC3′ F-L126A 5′GAAAGAATGCAGTGCAGACAGCGGTCAAGAACCAATAGGT3′ R-L126A 5′ACCTATTGGTTCTTGACCGCTGTCTGCACTGCATTCTTTC3′ F-V127A 5′AGAACGCAGTGCAGACACTGGCCAAGAATCAATAGGTCGAC3′ R-V127A 5′GTCGACCTATTGATTCTTGGCCAGTGTCTGCACTGCGTTCT3′ F-K128A 5′GCCGTGCAGACACTGGTAGCAAACCAATAGGTCGACCTCGA3′ R-K128A 5′TCGAGGTCGACCTATTGGTTTGCTACCAGTGTCTGCACGGC3′ F-N129A 5′CCGTGCAGACACTAGTCAAGCGCCAATAGGTCGACCTCGAGGG3′ R-N129A 5′CCCTCGAGGTCGACCTATTGGCGCTTGACTAGTGTCTGCACGG3′ F-Q130A 5′TGCAGACACTGGTCAAGAACGCATAGGTCGACCTCGAGGGGGG3′ R-Q130A 5′CCCCCCTCGAGGTCGACCTATGCGTTCTTGACCAGTGTCTGCA3′ ::: Expression and purification of recombinant HrpB2~XAC~ and HrcU~XAC~ fragments {#s4b} ----------------------------------------------------------------------------- Plasmid constructs pU1, pU2, and pB2 were used to transform *E. coli* strain BL21(DE3) [@pone.0017614-Studier2] and pU3 was used to transform BL21(DE3)RIL cells [@pone.0017614-KleberJanke1]. The synthesis of recombinant proteins was induced by the addition of 0.4 mM isopropyl-β-D-thiogalactopyranoside when cultures grown at 37°C attained an optical density of 0.8 at 600 nm. After three more hours of growth, cells were collected by centrifugation at 4500 x *g* for 15 min at 4°C and ressuspended in 20 ml/l of culture of 25 mM Tris-HCl (pH 8.0) for HrcU~XAC~ fragments, and 5 mM sodium acetate (pH 6.0) for HrpB2~XAC~. Cells were lysed by passage through a French pressure cell followed by centrifugation at 37000 x *g* for 1 hour at 4°C. Expression of HrcU~XAC\_207-357~ led to the production of a 7 kDa polypeptide, not 17 kDa expected from the size of the protein coded by the gene fragment in the pU1 vector (see [Results](#s2){ref-type="sec"}). This polypeptide was purified from the soluble fraction of the bacterial lysate by Q-Sepharose (Amersham Bioscience) anion-exchange chromatography (25 mM Tris-HCl (pH 8.0), 14 mM β-mercaptoethanol) using a 0-300 mM NaCl gradient, followed by Superdex G-75 (Amersham Bioscience) size exclusion chromatography (25 mM Tris-HCl (pH 8.0), 100 mM NaCl, 14 mM β-mercaptoethanol). HrcU~XAC\_207-357(AAAH)~ and HrpB2~XAC~ recombinant proteins were recovered from the insoluble fraction of the bacterial lysate by solubilizing in 25 mM Tris-HCl (pH 8.0), 14 mM β-mercaptoethanol, 8 M urea for HrcU~XAC\_207-357(AAAH)~ or 5 mM sodium acetate (pH 6.0), 14 mM β-mercaptoethanol, 8 M urea for HrpB2~XAC~. HrcU~XAC\_207-357(AAAH)~ was purified by Q-Sepharose anion-exchange chromatography, using the solubilization buffer (above) and a 0-300 mM NaCl gradient followed by Superdex G-75 size exclusion chromatography using 25 mM Tris-HCl (pH 8.0), 100 mM NaCl, 14 mM β-mercaptoethanol, 8 M urea. HrpB2~XAC~ was purified by passing the protein mixture through a Q-sepharose column equilibrated with 5 mM sodium acetate (pH 6.0), 14 mM β-mercaptoethanol, 8 M urea. HrpB2~XAC~ does not bind to this column under these conditions. The unbound fraction containing HrpB2~XAC~ was concentrated using an 10 kDa Amicon filter (Millipore) and separated by Superdex G-75 size exclusion chromatography using 5 mM sodium acetate (pH 6.0), 100 mM NaCl, 14 mM β-mercaptoethanol, 8 M urea. HrcU~XAC\_207-357(AAAH)~ and HrpB2~XAC~ were refolded by dialyses against 25 mM Tris-HCl (pH 8.0), 14 mM β-mercaptoethanol for HrcU~XAC\_207-357(AAAH)~, or 5 mM sodium acetate (pH 6.0), 14 mM β-mercaptoethanol for HrpB2~XAC~ containing successively reduced amounts of urea: 6 M, 4 M, 2 M, 0 M. HrcU~His277-357~ was purified from the insoluble fraction of the bacterial lysate by solubilizing in 25 mM Tris-HCl (pH 8.0), 10 mM imidazole, 100 mM NaCl, 2 mM β-mercaptoethanol, 8 M urea. The protein mixture was applied to a Ni^2+^-chelating Sepharose column equilibrated with the same buffer and eluted using a 25-500 mM imidazole gradient. HrcU~XAC\_His277-357~ fractions were pooled and the protein was refolded by successive dialyses against 25 mM Tris-HCl (pH 8.0), 14 mM β-mercaptoethanol containing 6 M, 4 M, 2 M and 0 M urea. Production of polyclonal antibodies against HrcU~XAC207-357(AAAH)~ and HrpB2~XAC~ proteins {#s4c} ------------------------------------------------------------------------------------------ Swiss Webster mice were immunized with four injections, separated by one week intervals, of 10 µg soluble HrpB2~XAC~. New Zealand white rabbits were immunized with HrcU~XAC\_207-357(AAAH)~ using four 200 µg injections separated by one week intervals. In both cases, the antigens were diluted with one volume of complete Freund\'s adjuvant (Sigma) for the first immunization and one volume of incomplete Freund\'s adjuvant (Sigma) for the remaining immunizations. Blood was collected and incubated for 1 hr at 37°C and the serum was recovered by centrifugation at 5000 x g for 15 min at room temperature, aliquoted and stored at -20°C. Before use, antiserum aliquots were incubated with an *E. coli* lysate as described [@pone.0017614-Sambrook1]. Edman degradation N-terminal sequencing {#s4d} --------------------------------------- The N-terminus of the 7 kDa polypeptide purified after the expression of HrcU~XAC\_207-357~ was lyophilized and dissolved in ultrapure water. N-terminal sequencing was carried out by Edman degradation using a PPSQ/23 sequencer (Shimadzu Corporation, Tokyo). Mass spectrometry experiments {#s4e} ----------------------------- Purified proteins were analyzed by Matrix Assisted Laser Desorption Ionization (MALDI) Time of Flight (TOF) Mass Spectrometry (MS) using an Ettan MALDI-TOF Pro system (Amersham Biosciences). All MALDI-TOF MS spectra were externally calibrated using a cytochrome C standard (12327 Da). Protein mass was identified in linear mode with positive ionization at 20 kV. The samples were mixed with an equal volume of sinapinic acid matrix dissolved in 50% acetonitrile, 0.5% of trifluoroacetic acid. A 0.5 µl aliquot was loaded onto stainless steel MALDI slides for analysis. Spectra were analyzed using the Ettan Maldi-Tof Pro v2.0 software package. Western blot assays {#s4f} ------------------- Samples were separated by SDS-PAGE (18% acrylamide) and electroblotted onto a nitrocellulose membrane. The membrane was colored with Ponceau red to identify the positions of specific proteins and then blocked for 2 h with 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% Tween 20, 0.1% Triton (TBS-TT) and 5% non-fat dry milk. The membranes were probed for 2 h with the appropriate polyclonal antiserum in 5-10 ml of the above blocking buffer (1∶3000 dilution for anti-HrcU~XAC~ antibody and 1∶20000 dilution for the anti-HrpB2~XAC~ antibody) and then washed four times for 15 min with TBS-TT. The anti-HrpB2~XAC~ antibody was detected using an anti-mouse IgG conjugated with horseradish peroxidase (Sigma) at a dilution of 1∶6000. The anti-HrcU~XAC~ antibody was detected using protein A conjugated with horseradish peroxidase (Sigma) at a dilution of 1∶30000. The membranes were incubated for 2 h with the protein-A or anti-IgG conjugates in 5-10 ml of blocking buffer following by washing with TBS-TT. Reactive bands were detected using the ECL AdvanceTM Western Blotting Detection Kit (GE Heathcare-Amersham) according to the manufacturer\'s instructions. Far-Western assays {#s4g} ------------------ Far-Western blot assays were carried out to detect specific protein-protein interactions. Approximately 15 µg of purified recombinant protein or lysates from *E. coli* cells was separated by SDS-PAGE (18% acrylamide) and electroblotted onto nitrocellulose membranes. The membrane was blocked for 2 h with TBS-TT plus 5% nonfat dry milk followed by 14 h incubation with 50 µg/ml of a second purified recombinant protein (indicated in the figure legends) at 4°C. Unbound proteins were removed by washing the membranes four times for 15 min with TBS-TT. Bound proteins were then detected as described for the Western blot assays (above). In some cases, negative control experiments were performed using a polypeptide derived from residues 143-284 of chicken muscle α-tropomyosin [@pone.0017614-Paulucci1]. His-tag pulldown assays {#s4h} ----------------------- HrcU~XAC\_His277-357~, HrpB2~XAC~ and an *E. coli* lysate were dialyzed at 4°C against 25 mM Tris-HCl, 100 mM NaCl, 2 mM β-mercaptoethanol, 10 mM imidazole (pH 8.0). A mixture of HrcU~His277-357~ (30 µM) and HrpB2 (30 µM) was added to a 0.25-ml aliquot of Ni^2+^-chelating Sepharose resin (Amersham Bioscience) equilibrated in the above buffer at room temperature. In control experiments, the resin was mixed with only HrcU~XAC\_His277-357~ or HrpB2 or with a mixture of HrcU~His277-357~ and a lysate derived from 10 ml of *E. coli* BL21(DE3) culture (OD~600~ = 0.8). The mixtures were washed four times with 1 ml of 25 mM Tris-HCl, 100 mM NaCl, 2 mM 2-β-mercaptoethanol, 25 mM imidazole (pH 8.0). Bound proteins were released by washing with 50 µl of 25 mM Tris-HCl, 100 mM NaCl, 2 mM 2-mercaptoethanol, 500 mM imidazole (pH 8.0). Samples were then analyzed by SDS-PAGE and Western blot assay. Fluorescence experiments {#s4i} ------------------------ HrcU~XAC\_207-357(AAAH)~ and HrpB2~XAC~ (both 2 µM) were dissolved in 5 mM sodium acetate pH 6.0 at 25°C. Fluorescence emission spectra were obtained using an AVIV (Lakewood, NJ) ATF 105 Automated Titrating Differential/Ratio spectrofluorometer and were collected between 320 and 400 nm using an excitation wavelength of 280 nm and excitation and emission bandwidths of 2 nm and 5 nm respectively. Production of *Xac* genes knockouts {#s4j} ----------------------------------- Deletion strains were constructed using the suicide vector pNPTS138 (Alley Dickon, unpublished) by allelic exchange as described [@pone.0017614-Guzzo1]. DNA fragments (1 kb) flanking each side of the *Xac hrpB2* and *hrcU* genes were amplified by PCR using oligonucleotides listed in [Table 3](#pone-0017614-t003){ref-type="table"}. For *hrcU*, primer pairs F1-U + R1-U and F2-U + R2-U were used. For *hrpB2,* primer pairs F1-B2 + R1-B2 and F2-B2 + R2-B2 were used. The products were digested with endonuclease NdeI and specific pairs were joined together with T4 DNA ligase (New England Biolabs). The resulting fragments were cloned into pNPTS138 generating pNPTS138-Δ*hrcU* using HindIII and NheI and pNPTS138-Δ*hrpB2* by using HindIII and SalI. These plasmids were introduced by electroporation into *Xac* strain 306. Kanamycin and ampicillin-resistant colonies were selected and grown on plates containing 5% sucrose and ampicillin. Sucrose-sensitive and kanamycin- and ampicilin-resistant colonies were selected and used to inoculate 10 ml of 2xYT-ampicilin medium, which was incubated overnight with agitation at 28°C. A 100 µl aliquot of this culture was plated without dilution on 2xYT agar plates containing 200mg/L ampicillin. The resulting colonies were transferred in replica on two plates: one containing kanamycin and ampicillin and another containing sucrose and ampicilin. Clones that were simultaneously kanamycin-sensitive and sucrose-resistant were selected, and the deletion was confirmed by PCR. Production of expression vectors for complementation of Δ*hrcU* and Δ*hrpB2 in Xac* {#s4k} ----------------------------------------------------------------------------------- A fragment containing the *hrcU* gene plus 1 kb upstream sequences was amplified by PCR using primers F1-U and R-compU ([Table 3](#pone-0017614-t003){ref-type="table"}). This fragment contains the complete HrcU open reading frame as well as its native promoter. After digestion with HindIII and SalI, this fragment was cloned into the HindIII-SalI sites of pBBR1MCS-5 [@pone.0017614-Kovach1], resulting in pBBR\_*hrcU* ([Table 2](#pone-0017614-t002){ref-type="table"}). To construct pBBR\_*hrcU~AAAH~* ([Table 2](#pone-0017614-t002){ref-type="table"}), primers F-U~AAAH~ and R-U~AAAH~ (see [Table 3](#pone-0017614-t003){ref-type="table"}) were used in a PCR amplification with pBBR\_*hrcU* as template to change the codons for residues 264-266 (NPT) to alanine codons using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). The mutation was confirmed by sequencing. The HindIII/SalI fragments of pBBR\_*hrcU* and pBBR\_*hrcU~AAAH~*, which contain the complete HrcU~XAC~ open reading frame as well as its native promoter, were cloned into the same sites of pUFR047, a broad-host range vector carrying a gentamycin resistance gene [@pone.0017614-DeFeyter1], generating constructs pUFR047\_*hrcU* and pUFR047\_*hrcU~AAAH~* ([Table 2](#pone-0017614-t002){ref-type="table"}). These plasmids were used to transform the Δ*hrcU* mutant strain by electroporation followed by selection on LB plates with 10 µg/ml gentamycin and 200 µg/ml ampicillin. A fragment containing the *hrpB2* gene plus 1 kb upstream sequences was amplified by PCR using primers F1-B2 and R-compB2 ([Table 3](#pone-0017614-t003){ref-type="table"}), digested with HindIII and cloned into the HindIII site of pUFR047. The resulting construct, pUFR047\_*hrpB2* ([Table 2](#pone-0017614-t002){ref-type="table"}) was used to transform the *Xac* Δ*hrpB2* mutant strain by electroporation. Transformed colonies were selected on LB/gentamycin/ampicilin plates to produce strain Δ*hrpB2*+pUFR047*\_hrpB2* ([Table 1](#pone-0017614-t001){ref-type="table"}). To produce *hrpB2* gene mutants for expression in *Xac*, the HindIII/SalI fragment of the PCR product above was cloned between the HindIII and SalI sites of pBBR1MCS-5 generating the construct pBBR\_*hrpB2* ([Table 2](#pone-0017614-t002){ref-type="table"}) which was then used as a template to produce mutants using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). Primers F-M57 and R-M57, F-Q124 and R-Q124 ([Table 3](#pone-0017614-t003){ref-type="table"}) were used to change the codons 171 and 372 to stop codons; primers F-FQALM and R-FQALM ([Table 3](#pone-0017614-t003){ref-type="table"}) were used to change the codons for the FQALM motif (residues 35-39) to LQGPR codons and finally, primers pairs F-T125A and R-T125A, F-L126A and R-L126A, F-V127A and R-V127A, F-K128A and R-K128A, F-N129A and R-N129A and F-Q130A and R-Q130A ([Table 3](#pone-0017614-t003){ref-type="table"}) were used to change the respective codons to alanine codons. The HindIII/SalI fragments from all these pBBR\_*hrpB2* derived constructs were cloned between the same sites of pUFR047 generating the constructions pUFR047*\_hrpB2~1-56~*, pUFR047*\_hrpB2~1-123~,* pUFR047*\_hrpB2~T125A~*, pUFR047*\_hrpB2~L126A~*, pUFR047*\_hrpB2~V127A~,* pUFR047*\_hrpB2~K128A~,* pUFR047*\_hrpB2~N129A,~* and pUFR047*\_hrpB2~Q130A~* ([Table 2](#pone-0017614-t002){ref-type="table"}). All the mutations were confirmed by sequencing. Theses constructions were used to transform the *Xac* Δ*hrpB2* strain by electroporation ([Table 1](#pone-0017614-t001){ref-type="table"}). Plant bioassays {#s4l} --------------- Highly susceptible Navel sweet orange (*Citrus sinensis* (L.) Osbeck) plants were grown under greenhouse conditions and maintained at 28°C with daylight for virulence assays. To visually monitor the development of citrus canker symptoms, *Xac* 306 and mutant strains were grown overnight at 30°C and adjusted to an optical density of 0.3 at 600 nm in 2xYT culture medium. The suspensions were hand-infiltrated with a 1-ml syringe with needle into the abaxial surface of attached leaves. To monitor bacterial growth *in planta*, *Xac* strains were grown overnight at 30°C and adjusted to an optical density at 600 nm (*OD* ~600~) of 0.5 in NB culture medium (8g of nutrient broth liter^-1^, 5 g of NaCl liter^−1^, pH 7). The abaxial surface of young leaves was pricked by using pins whose tips were previously immersed in the bacterial suspension for *Xac hrcU* mutant strains ([Fig. 7A](#pone-0017614-g007){ref-type="fig"}) or by infiltration into leaves with needleless syringes for *Xac hrpB2* mutant strains ([Fig. 7B](#pone-0017614-g007){ref-type="fig"}). In both cases, leaf disks (0.8 cm^2^) from infected plants were removed with a cork borer during a 12 day period post-inoculation, macerated in 0.85% NaCl with a mortar and pestle. Different dilutions were spread on LB plates with the appropriate antibiotics and the bacterial population was determined by counting colonies after a 2-day incubation period at 28-30°C. Experiments were performed in triplicate. Preparation of orange leaf extracts {#s4m} ----------------------------------- Sweet orange leaf extracts were prepared as described previously for passion fruit leaf extracts [@pone.0017614-Tahara1]. Leaves were washed extensively with sterile water. Midribs were excluded and 1 g of tissue was mixed with liquid nitrogen and pulverized to form a fine powder. One-hundred milliliters of MM medium [@pone.0017614-Tahara1] plus carbenicillin 100 µg/ml, pH 7.4 were added and the mixture was macerated followed by centrifugation at 5000 x *g* for 15 min at 4°C. The supernatant was recovered and passed through 0.45 µm and 0.22 µm filters (Millipore) and stored at -80°C. HrpB2 secretion by *Xac* {#s4n} ------------------------ *Xac* 306 cells were cultivated at 30°C in MM medium (pH 5.4) plus 100 µg/ml carbenicillin containing sweet orange leaf extract (extract derived from 1 g of leaf tissue per litre of MM medium). *Xac* cultures (50 mL) were grown for 24 h to an *OD* ~600~ = 0.3 after which cells were collected by centrifugation and resuspended in 3 ml of urea-SB: 8 M urea, 10% glycerol, 52 mM Tris-HCl (pH 6.8), 2% SDS, 0.1% bromphenol blue, 140 mM 2-mercaptoethanol. The extracellular (secreted) fraction from a 50 ml culture was passed through a low protein-binding filter 0.45 µm (Millipore). Proteins in the filtrates were precipitated by adding 10% trichloroacetic acid and freezing at −20°C for 12 h followed by centrifugation at 12000 x *g* (4°C). The precipitate was washed twice with cold acetone and resuspended in 50 µl urea-SB. Note that the above procedure produces a secreted fraction that is derived from 60 times as many bacterial cells per unit volume as the cellular fraction. Equal volumes of cellular and secreted protein fractions were separated by SDS-PAGE (18% acrylamide) and transferred onto the nitrocellulose membrane. HrpB2 was detected by Western blot using anti-HrpB2~XAC~ antibodies (above). We thank Fernando Corrêa for the help in the fluorescence measurements, Izaura Nobuko Toma and Paolo di Mascio for MALDI-TOF MS analyses, Izaura Yoshico Hirata for Edman degradation analyses, Ângela Mika Katsuyama for the clone for expression of HrpB2, and Marcos C. Alegria for helpful discussions. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by research grants from the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP grant \# 2005/59243-3) and the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) to C.S.F. and a graduate student scholarship to P.A.C. from FAPESP. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: PAC AMA CSF. Performed the experiments: PAC RFS AMA RAH TSS. Analyzed the data: PAC AMA MAM CSF. Contributed reagents/materials/analysis tools: MAM CSF. Wrote the paper: PAC CSF.

PubMed Central

2024-06-05T04:04:19.176426

2011-3-9

PMC3052358

Introduction {#s1} ============ Stem cell therapy is a potential promising approach for the treatment of muscular dystrophies such as Duchenne muscular dystrophy (DMD), in which muscle fibres degenerate due to lack of the protein dystrophin [@pone.0017454-Grounds1]--[@pone.0017454-Peault1]. Skeletal muscle regeneration is mediated by muscle-specific stem cells called satellite cells [@pone.0017454-Collins1]; their progeny, myoblasts, can be expanded in culture and retain myogenic differentiative capacity. Despite promising work in mouse models of DMD [@pone.0017454-Partridge1], clinical trials of myoblasts in DMD patients were disappointing [@pone.0017454-Law1]--[@pone.0017454-Mendell1], the main problems being low survival and migration of grafted cells and the low number of donor-derived muscle fibres [@pone.0017454-Mendell2]. Attention has therefore turned to other types of stem cell, with the goal of finding a cell that can be systemically-delivered, give rise to significant numbers of muscle fibres in recipient muscles and functionally reconstitute the muscle stem cell pool, so that dystrophin-negative muscle fibres can be repaired later in life. Amongst the many stem or precursor cells of human origin that make at least some muscle in *in vivo* models of DMD [@pone.0017454-Benchaouir1]--[@pone.0017454-Gang1], blood-vessel associated stem cells - mesoangioblasts from embryonic stages or pericytes from adults - seem to be the most promising [@pone.0017454-Dellavalle1], [@pone.0017454-Morosetti1]--[@pone.0017454-Sampaolesi2]. Human muscle-derived pericytes gave rise to large amounts of muscle after intra-arterial delivery in immunodeficient, dystrophin-deficient (SCID mdx) mice [@pone.0017454-Dellavalle1]. However, despite expressing markers of pericytes and not myoblasts, their precise origin is uncertain, as the method of preparation could lead to contamination with other cell types, e.g. satellite cells, endothelial cells, mesenchymal stem cells and fibroblasts. Here, we have isolated cells (termed muscle-derived cells, or mdcs) from human muscle biopsies following the protocol used previously to prepare human pericytes [@pone.0017454-Dellavalle1] and investigated their phenotype and capacity to undergo myogenic differentiation *in vitro.* Our cell preparations were phenotypically similar to pericytes prepared by Dellavalle et al. in terms of expression of pericyte markers such as ALP and PDGFR-β, except that a proportion of our cells in most of the preparations also expressed the myogenic marker CD56. In addition, our cell preparations contained cells expressing myogenic regulatory factors at the mRNA level prior to their differentiation into myotubes, so we termed them mdcs rather than pericytes. We also found differences in mdcs prepared in the same way from 8 different donors - two preparations showed extensive myogenic differentiation *in vitro*, four were less myogenic and two entered senescence at early stages in culture. *In vivo*, in contrast to human pericytes, which contribute to large numbers of muscle fibres after intra-arterial transplantation, our mdcs, especially the CD56+ subpopulation, contributed to muscle regeneration only after intra-muscular transplantation in our mdx nu/nu mouse model. Materials and Methods {#s2} ===================== 1. Ethics {#s2a} --------- Tissue sampling was approved by the NHS national research ethics service Hammersmith and Queen Charlotte\'s and Chelsea Research Ethics Committee: Setting up of a rare diseases biological samples bank (biobank) for research to facilitate pharmacological, gene and cell therapy trials in neuromuscular disorders (NMD) REC reference number: 06/Q0406/33, in compliance with national guidelines regarding the use of biopsy tissue for research. All patients gave written informed consent. Mice were bred and experimental procedures were carried out in the Biological Services Unit, Institute of Child Health, University College London, in accordance with the Animals (Scientific Procedures) Act 1986. Experiments were performed under Home Office licence numbers 70/6228 or 70/7086. Experiments were approved by the local University College London ethical committee prior to the licence being granted. 2. *In vitro* isolation and maintenance of human muscle derived cells {#s2b} --------------------------------------------------------------------- Human mdcs were isolated as previously described [@pone.0017454-Dellavalle1], [@pone.0017454-Tonlorenzi1]. Muscle biopsies from 3 normal and 5 DMD patients ([Table 1](#pone-0017454-t001){ref-type="table"}) were cut into 1 mm^3^ pieces using a scalpel and placed as explants into 35 cm^2^ culture dishes (Nunc) coated with collagen type I (1 mg/ml from rat tail, Sigma). Explants were kept in M5 medium (Megacell medium (Sigma) + 5% foetal bovine serum (FBS, PAA) + 2 µM glutamine (Sigma) + 1% non essential amino acids (NEAA) + 0.1 mM β- mercaptoethanol (β-ME, Sigma) + 5 ng/ml basic fibroblast growth factor (bFGF, Peprotech) for 10--14 days at 37°C in 5% O~2~ and 5% CO~2~. Small, refractory, non-adherent cells were collected by gentle pipetting, transferred to new collagen I-coated dishes and expanded in M10 medium (Megacell medium (Sigma) + 10% FBS + 2 µM Glutamine + 1% NEAA + 0.1 mM β- ME + 5 ng/ml bFGF), whilst adherent cells and the initial muscle explants were discarded. Once cells reached confluence, the same procedure was repeated to collect the small refractory cells and expand them in new collagen 1-coated dishes. Cells generated from the second dish were counted as passage 1, and mean population doubling times (mpds) were determined from this point. For long-term maintenance, cells were plated at a density of 2.5×10^5^ cells/75 cm^2^ flask, on the substrate and in the medium described above. Cells were trypsinized and passaged every 3--4 days, and mpds were calculated by the following formula: mpd increase  =  Ln \[total cells/cells seeded\]/Ln2 and doubling time were calculated by time from seeding to harvesting/mpd. Total mRNA from a small proportion of cells at each passage was extracted for RT-PCR assay below. Aliquots of cells were frozen at each passage and stored in liquid nitrogen for future studies. ::: {#pone-0017454-t001 .table-wrap} 10.1371/journal.pone.0017454.t001 Table 1 ::: {.caption} ###### Sources of human mdcs. ::: {#pone-0017454-t001-1} No. ID age sex Diagnosis Muscle of origin Myogenicity (fusion index) Comments ----- ----- ---------- ----- --------------------------------------- ------------------ ---------------------------- ------------------------------------ 1. pN1 Adolescent idiopathic scoliosis (AIS) Para-spinal 4% 2. pN2 15 years F AIS Para-spinal 4.5% 3. pN3 14 years F AIS Para-spinal N/A Entered senescence at early stages 4. pD1 4 years M DMD Δ42-43 Quadriceps 0.5% 5. pD2 11years M DMD Δ45-50 EDB 10--40% 6. pD3 11years M DMD Δ45-50 EDB 10--40% 7. pD4 4.5years M DMD (C.5503C-T, P.Gln1835X) Quadriceps Yes (intermediate level) 8. pD5 5 years M DMD with duplicated exon 3-9 Quadriceps N/A Entered senescence at P6 ::: 3. *in vitro* analysis of mdcs {#s2c} ------------------------------ ### 3.1 Semi-quantitative RT-PCR {#s2c1} Total RNA was extracted from mdcs with RNAeasy mini preparation kit (Qiagen, West Sussex, UK) following the manufacturer\'s instructions. RNA was quantified using Nanodrop 1000 3.6.0 (Thermal Scientific). For each RT-PCR reaction, 10 ng of total RNA was used as template. One-step RT-PCR was performed using primers designed to recognize either the myogenic regulatory factors (MRFs) Pax3, Pax7, Myf5, MyoD, desmin, myosin heavy chain (MHC) as previously described [@pone.0017454-Meng1], or other cellular markers: CD34, PDGFR-β, NG2, ALP, GAPDH and CD144. The sequence of the primers is listed in [Table 2](#pone-0017454-t002){ref-type="table"}. PCR products were separated on 1% agarose gel for 20--30 min at 100 V and visualised with SyberSafe DNA gel stain (Invitrogen) and GelDoc imaging software (BioRad). ::: {#pone-0017454-t002 .table-wrap} 10.1371/journal.pone.0017454.t002 Table 2 ::: {.caption} ###### Primers used for identification of mdcs. ::: {#pone-0017454-t002-2} Forward primer Reverse primer Fragment size --------- ------------------------ ----------------------- --------------- CD34 AGAAAGGCTGGGCGAAGACCCT AGTGGGGAAGGGTTGGGCGT 311 bp PDGFR-β CTGCGTCTGCAGCACGTGGA CTGCCCAAAGGCCCCAGAGC 357 bp NG2 GTCCGACGGGCAACACCAGG CACTGGCCCTGCTTCCACGG 340 bp ALP CTGACCACTGCCAGCCCACC GGGCAGCCGTCACTGTGGAG 294 bp CD144 CCGCGGGAAACAGAGCCCAG ACTCGCCCTGCTCGTTGCAC 696 bp GAPDH CCCATCACCATCTTCCAGGA TTGTCATACCAGGAAATGAGC 731 bp ::: ### 3.2 Flow cytometric analysis {#s2c2} Mdcs pN1, pD1 and pD2 at mpds 15--30 were processed for flow cytometry. Cultured cells were expanded for 3 days as described above before being detached from the culture flasks using 0.05% trypsin−0.2% EDTA (Sigma) and centrifuged at 500 g for 5 min. The cell pellet was fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature for flow cytometric analysis. 5×105 cells were processed for each staining and all procedures were performed at room temperature. Cells were blocked with PBS containing 1% bovine serum albumin (BSA) for 30 min before being incubated with primary antibodies for 1 hour, followed by corresponding secondary antibodies for 30 min ([Table S1](#pone.0017454.s006){ref-type="supplementary-material"}). Either secondary antibody alone, or PE/FITC conjugated isotype matched antibodies, were used as controls. After staining, cells were washed with PBS and analysed with a BD LSRII FACS machine. 10,000 events were collected for each sample. Flowjo 7.2.5 software was used to analyse the results. FACS analysis was performed at least 3 times for each marker on each cell preparation. ### 3.3 Immunofluorescent staining {#s2c3} Two cell preparations, pN1 and pD2, at mpds 10--20 were used for immunofluorescent analysis of marker expression. 1×10^5^ cells were plated onto 5 µg/ml poly-D-lysine (PDL) coated coverslips and incubated overnight before being processed for immunofluorescent staining. Staining was performed at room temperature. Cells were fixed with 4% PFA for 15 min and incubated with blocking solution (PBS containing 10% normal goat serum (NGS)/0.3% Triton X100) for 30 min. Cells were then incubated with primary antibodies (listed in [Table S1](#pone.0017454.s006){ref-type="supplementary-material"}) for 1 hour followed by Alexa 488-conjugated goat anti- mouse or rabbit IgG (H+L) (Invitrogen, 1∶500) for 1 hour. Coverslips were then mounted with mounting medium (DAKO) containing 10 µg/ml 4′,6-diamidino-2-phenylindole (DAPI). Cells stained with secondary antibody only were used as control. ### 3.4 In vitro myogenesis {#s2c4} To initiate myogenic differentiation, mdcs were plated on 10 µg/ml laminin (Invitrogen) coated 8-well chamberslides (Nunc) at 5×10^4^ cells/well in M2 medium (Megacell medium containing 2% FBS). Medium was changed every 3--4 days during differentiation. Cells were fixed at different time points after plating and stained as described in [material and methods](#s2){ref-type="sec"} section 3.3 with an antibody to myosin (MF20; DSHB), a marker of muscle differentiation. Fusion index was determined by counting the percentage of nuclei within MF20+ myotubes in 5 randomly-encountered fields per well in 4 replicate wells. The cell preparation (pD2) with the highest fusion index was chosen for all subsequent *in vitro* and *in vivo* studies. 4. Separation of CD56 + and CD56- populations by flow cytometry {#s2d} --------------------------------------------------------------- pD2 mdcs at mpds 10.773, 16.165 and 25.772 were sorted into CD56+ and -- subpopulations. Cells were plated at 2.5×10^5^ cells/75 cm^2^ flask for 3 days before being trypsinized and resuspended in PBS containing 10% BSA and incubated with CD56:PE antibody (Miltenyi biotech, 130-090-755,1∶20) for 30 min at 4°C. Cells incubated with mouse IgG1:PE (1∶20) were taken as a negative control. Cells were washed twice with PBS after antibody incubation, and then sorted on the basis of CD56 expression using a MoFlo XDP cell sorter. A small aliquot (approximately 1×10^4^ cells) of both CD56+ and CD56- cells was collected and processed for RT-PCR analysis (as described in [material and methods](#s2){ref-type="sec"} section 3.1). Sorted CD56+ and CD56- subpopulations and non-sorted pD2 cells were expanded for further *in vitro* analysis and intra-muscular transplantation. To examine the myogenicity of CD56+ and -- cells, pD2 cells were sorted by flow cytometry on the basis of CD56 expression and immediately plated onto laminin-coated chamberslides at 5×10^4^ cells/well in proliferation medium (M10). Differentiation was induced 24 hours later by switching into differentiating medium (M2). Cells were fixed at D1, D3, D5, and D7 after differentiation and immunostained with antibodies to CD56, myosin (MF20), Myf5, desmin, MyoD and myogenin. The percentage of positive cells was counted as described above. 5. BrdU assay {#s2e} ------------- CD56+ and CD56- subpopulations of pD2 cells were labelled with bromodeoxyurindine (BrdU) and stained with an anti-BrdU antibody ([Table S1](#pone.0017454.s006){ref-type="supplementary-material"}) to detect proliferating cells. Cells at mpds 20--27 were plated onto 5 µg/ml Poly-D-Lysine coated 8-well chamberslides at a density of 2×10^4^ cells/well for 24 hours in M10 medium. 10 µM BrdU was then added to the culture medium for 18 hours. Cells were then fixed with 4% PFA for 15 min, followed by treatment with 3N HCl for 10 min. Cells were then stained with a BrdU antibody, as described in [material and methods](#s2){ref-type="sec"} section 3.3. The percentage of BrdU positive nuclei was counted in 5 randomly-encountered fields at 10× using Metamorph software; 3 wells were averaged each group, and the experiment was repeated 3 times. 6. In vivo transplantation of mdcs {#s2f} ---------------------------------- ### 6.1 Intra-muscular transplantation of mdcs {#s2f1} 5×10^5^ cells in 5 µl medium were grafted into cryodamaged tibialis anterior (TA) muscles of 1--2 month old mdx nu/nu mice as previously described [@pone.0017454-Brimah1], [@pone.0017454-Ehrhardt1]. First, we compared the capacity of pN1 (with a low fusion index of 4%) and pD2 (fusion index 40%) to contribute to muscle regeneration *in vivo*. For this experiment, cells injected were at passage 5 (pN1, mpd 12.4) and passage 4 (pD2, mpd 9.8) respectively and injected muscles were analyzed 28 days after transplantation. Next, we grafted non-sorted and CD56+ and CD56− subpopulations of pD2 and analyzed grafted muscles 4 and 8 weeks later. In some experiments, grafted muscles were removed, frozen, cryosectioned and stained with antibodies to human spectrin, human lamin a/c and pan-laminin ([Table S1](#pone.0017454.s006){ref-type="supplementary-material"}) followed by Alexa 488 conjugated goat anti-mouse IgG (H+L) and Alexa 594 conjugated goat anti-rabbit IgG (H+L) secondary antibody as described previously [@pone.0017454-Brimah1], [@pone.0017454-Ehrhardt1]. Images of the transverse sections containing most nuclei of human origin were captured with MetaMorph software. The number of human lamin A/C+ nuclei, human spectrin+ fibres, human spectrin+ fibres containing human lamin A/C+ nuclei and the number of human lamin A/C+ nuclei inside or outside basal lamina were counted using MetaMorph software and compared using either Student\'s t test or one-way ANOVA. ### 6.2 Intra-arterial transplantation of mdcs {#s2f2} Mdx nu/nu mice at 1--2 months of age were anesthetised with isofluorane and an incision was made in the groin to expose the femoral artery of one hindlimb. The femoral artery was carefully dissected away from the flanking femoral vein and nerve. A 30G insulin syringe (Becton Dickinson) was inserted into the artery and 25 µl medium containing 5×10^5^ cells was injected. Pressure was applied to prevent blood loss from the artery. The skin was then sutured and the mouse was given a subcutaneous injection of Vetergesic (buprenorphine hydrochloride, 0.05 g/kg body weight). The contralateral limb remained untreated. In preliminary experiments, we injected 25 µl of 1% Evans blue instead of cells into the femoral arteries of the right legs of host mice. The dye was immediately seen along the downstream blood vessel network and the foot of the injected leg became blue, demonstrating that our intra-arterial injections were successful (data not shown). Details of intra-arterial cell injections are given in [Table 3](#pone-0017454-t003){ref-type="table"}. Two Mdc preparations, pN1 and pD2, were each injected intra-arterially into mdx nu/nu mice: pN1 was injected into 13 host mice, 4 of which were analysed 2 hours after grafting and 9 of which were analysed 4 weeks after grafting. pD2 was injected in to 9 host mice, all of which were analysed 4 weeks after grafting. At each time point after transplantation, TA muscles of the injected leg, the lung and liver were processed for cryosectioning as described for intramuscular transplantation. ::: {#pone-0017454-t003 .table-wrap} 10.1371/journal.pone.0017454.t003 Table 3 ::: {.caption} ###### Summary of intra-arterial transplantation of mdcs. ::: {#pone-0017454-t003-3} Cell preparation Number of mice Time point No. human lamin a/c cells in downstream TA muscles (Mean±SEM) ------------------ ---------------- ------------ --------------------------------------------------------------- pN1 4 2 hrs 3.75±0.48 pN1 9 4 weeks 1.11±0.30 pD2 9 4weeks 8.60±7.41 ::: Results {#s3} ======= 1. Mdcs exhibit long-term proliferative capacity before entering senescence and express myoblast, pericyte and mesenchymal stem cell markers {#s3a} -------------------------------------------------------------------------------------------------------------------------------------------- It has been reported that cells isolated and maintained with the same protocol used in this study were pericytes, not satellite cells [@pone.0017454-Dellavalle1]. Since our cell preparations have phenotypical differences from cells described in this paper, we termed them mdcs. Mdcs proliferated rapidly *in vitro*, but during the first few passages, myotubes were observed within all cell preparations (data not shown), suggesting that satellite cell-derived myoblasts were present. The number of myotubes diminished after the first 2 passages and gradually the morphology of the cells became homogenous and similar to that described for pericytes: a mixture of small, triangular adherent cells and small, round floating cells [@pone.0017454-Dellavalle1], [@pone.0017454-Tonlorenzi1] ([Figure 1a](#pone-0017454-g001){ref-type="fig"}). The average mpd time of 4 separate mdc preparations (pN1, pD1, pD2 and pD3) was 33.79±1.24 hrs ([Figure 1b](#pone-0017454-g001){ref-type="fig"}), similar to that of pericytes (36 hrs) [@pone.0017454-Dellavalle1], but shorter than human synovial stem cells (54 hrs) [@pone.0017454-Meng1] and human myoblasts (49.7 hrs in low O~2~ and 60 hrs in high O~2~ conditions) (data not shown). The proliferation curve of pN1 and pD2 cells shows that our mdcs maintained their proliferative capacity until 30 mpds before entering into senenscence ([figure 1b](#pone-0017454-g001){ref-type="fig"}), slightly longer than reported for pericytes (20--25 mpds) [@pone.0017454-Dellavalle1]. ::: {#pone-0017454-g001 .fig} 10.1371/journal.pone.0017454.g001 Figure 1 ::: {.caption} ###### Characterization and phenotype of human mdcs. a\. Cultured human mdcs consisted of small, triangular adherent cells and small, round floating cells. Scale bar  = 50 µm. b. Proliferation curve of one normal human mdc preparation (pN1) and one DMD mdc preparation (pD1). The average mean population doubling time for these cells is 33.79±1.24 hrs. c, d, e, f. Representative images showing the expression of NG2 (c, green), CD49b (d, green), CD29 (e, green) and CD44 (f, green) by pN1 human mdcs. Nuclei were counterstained with 10 µg/ml DAPI (blue). Scale bar  = 50 µm. g. Representative FACS images showing the expression of PDGFRβ, ALP, CD49b, CD29, CD71, CD90, Stro-1, CD56 and CD34 by human mdcs. Isotype control is shown in red in each image, and positive signal is shown in green. Note that most of these markers, apart from PDGFRβ and CD56, were expressed at similar levels among cell preparations ([Figure S1](#pone.0017454.s001){ref-type="supplementary-material"}). :::  ::: Our mdcs expressed pericyte markers PDGFR-β ([Figure 1g](#pone-0017454-g001){ref-type="fig"}), ALP ([Figure 1g](#pone-0017454-g001){ref-type="fig"}), CD49b ([Figure 1d, 1g](#pone-0017454-g001){ref-type="fig"}), CD146 ([Figure S4a](#pone.0017454.s004){ref-type="supplementary-material"}, a\') and NG2 (detected by immunostaining, but not by flow cytometry) ([Figure 1c](#pone-0017454-g001){ref-type="fig"}), which is in accordance with previous reports [@pone.0017454-Dellavalle1]. But, although pericytes have been previously reported not to express the myogenic marker CD56 [@pone.0017454-Dellavalle1], most cell preparations (pN1, pD1, pD2, pD3, and pD5), maintained under our culture conditions, contained a C56+ subpopulation, ranging from 1.49--47% of the total population, as assessed by flow cytometric analysis ([Figure S1](#pone.0017454.s001){ref-type="supplementary-material"}). Double labeling of CD56:PE and pericyte markers showed that there were CD56+/ALP+, CD56+/CD146+ cells within pD2, suggesting the pericyte origin of the CD56+ cell subpopulation ([Figure S4](#pone.0017454.s004){ref-type="supplementary-material"} c and d). In addition, mdcs contained cells weakly expressing CD34, a marker expressed on satellite cells [@pone.0017454-Collins2], [@pone.0017454-Ieronimakis1], endothelial cells [@pone.0017454-Allegra1], hematopoietic stem cells [@pone.0017454-Nielsen1], [@pone.0017454-Mastrandrea1], myoendothelial cells [@pone.0017454-Zheng1], but not on pericytes [@pone.0017454-Dellavalle1]. To investigate the presence of endothelial cells within our mdcs, we performed FACS analysis of VE-Cadherin (CD144) and immunostaining of Von Willenbrand Factor (vWF) on pD2 cells. pD2 cells did not express VE-Cadherin ([Figure S4](#pone.0017454.s004){ref-type="supplementary-material"} b and b\') or vWF (data not shown). An angiogenesis assay on pD2 cells at 20 mpds (passage 9) clearly showed that these cells made vessel-like structures ([Materials and Methods S1](#pone.0017454.s007){ref-type="supplementary-material"} and [Figure S5](#pone.0017454.s005){ref-type="supplementary-material"}). In addition to pericyte and myogenic markers, the mesenchymal stem cell markers CD29 ([Figure 1e](#pone-0017454-g001){ref-type="fig"}), CD44 ([Figure1f](#pone-0017454-g001){ref-type="fig"}), CD71 ([Figure 1g](#pone-0017454-g001){ref-type="fig"}), CD90 ([Figure 1g](#pone-0017454-g001){ref-type="fig"}) and Stro-1 ([Figure 1g](#pone-0017454-g001){ref-type="fig"}) [@pone.0017454-Pittenger1], [@pone.0017454-Kemp1], were also expressed at similar levels by 3 mdc preparations ([Figure 1](#pone-0017454-g001){ref-type="fig"}). Most of the above cell markers examined were expressed at similar levels in the 3 cell preparations tested, except for PDGFR- β (ranging from 29%--98%) and CD56 (ranging from 4 to 40%) ([Figure S1](#pone.0017454.s001){ref-type="supplementary-material"}). 2. Mdcs express myogenic factors at the RNA level but show variable myogenesis {#s3b} ------------------------------------------------------------------------------ During proliferation *in vitro*, human pericytes do not express the myogenic determination factors Pax7, Myf5, MyoD or MHC, but they do express Pax3 at the RNA level [@pone.0017454-Dellavalle1]. However, RT-PCR on the total mRNA extracted from the different passages (P2-P6, equivalent to mpds 3--17) of 4 different mdc preparations (pN1, pD1, pD2 and pD3) grown under proliferative conditions showed that all four preparations expressed myogenic markers Pax3, Myf5 and MyoD. Two cell preparations, pN1 and pD1, also expressed Pax7 at mpd3-7 ([Figure 2a](#pone-0017454-g002){ref-type="fig"} and data not shown). Expression of Pax7, Myf5 and MyoD on pN1 and pD1 (2 cell preparations exhibiting low myogenicity *in vitro*, [Table 1](#pone-0017454-t001){ref-type="table"}) decreased with time in culture (with the exception of MyoD expression, that did not decrease significantly with time in pN1), whilst Pax3 expression was maintained at the different mpds examined ([Fig. 2a](#pone-0017454-g002){ref-type="fig"}). ::: {#pone-0017454-g002 .fig} 10.1371/journal.pone.0017454.g002 Figure 2 ::: {.caption} ###### Contribution of human mdcs to myogenesis *in vitro* and muscle regeneration *in vivo*. Expression of myogenic regulator factors at the RNA level by 2 representative cell preparations, pN1 (low myogenic) and pD2 (highly myogenic) from passage 2 to passage 6, equivalent to mpd 2 to 17. b, c. *In vitro* myogenesis of pN1(b) and pD2 (c) determined by myosin (MF20, green) immunostaining 7 days after *in vitro* differentiation, nuclei were counterstained with 10 µg/ml DAPI (blue). Scale bar  = 50 µm. d, e, f, g. Contribution of donor mdcs to nuclei and muscle fibres after grafting pN1(d) and pD2 (e) into cryodamaged TA muscles of mdx nude mice. Transverse muscle sections were co-stained with human lamin A/C and human spectrin (both red) to visualize the donor nuclei and donor muscles fibres. Total nuclei were counterstained with 10 µg/ml DAPI (blue). Arrows point to 2 human spectrin+ fibres in figure d. Scale bar  = 50 µm. f and g show the comparison of human lamin A/C+ cells and human spectrin+ fibres between groups of muscles transplanted with different cell preparations. Data was analysed with Student\'s t test. h, i, j, k. Mdcs did not contribute to muscle regeneration after intra-arterial transplantation. Representative images show that, 2 hours after i.a. injection of mdcs, very few human lamin A/C+ nuclei (green) were found either outside (h) blood vessels (vWF immunostaining, red) or (i) inside blood vessels; and few human lamin A/C+ nuclei (green) were present in the lungs of injected mice (j); 4 weeks after i.a. transplantation of mdcs, either no, or very few human lamin A/C + nuclei (green, k) were detected in TA muscles of the injected mice. No human spectrin+ muscle fibres were detected in any grafted mice. Nuclei were counterstained with 10 µg/ml DAPI (blue). Scale bar  = 100 µm for h, i and j, scale bar  = 50 µm for k. :::  ::: Expression of Myf5, MyoD and MHC were maintained in pD2 and pD3, two highly myogenic cell preparations ([Table 1](#pone-0017454-t001){ref-type="table"}). These data suggest that in cell preparations with low myogenicity, myogenic cells were present initially but were lost with increasing time in culture; whilst in highly myogenic preparations, myogenic cells persisted *in vitro*. Mdcs pN1, pN2, pD1, pD2, pD3, pD4 at mpd 5--20 were plated in differentiation conditions ([material and methods](#s2){ref-type="sec"} section 3.4) and the fusion index calculated after 7 days. Cell preparations from both normal and DMD muscles were myogenic to different extents, ranging from 0--40% nuclei in myotubes ([Table 1](#pone-0017454-t001){ref-type="table"}). Although there were inter-experiment differences in fusion index, pD2 mdscs retained their capacity to differentiate into myotubes until at least 22.7 mpds ([figure S3j](#pone.0017454.s003){ref-type="supplementary-material"}). The *in vivo* skeletal muscle regenerative capacity of different mdc preparations correlated with their *in vitro* myogenic capacity. Two mdc preparations with different *in vitro* fusion indices were grafted intra-muscularly into cryodamaged TA muscles of mdx nu/nu mice at passage 5 (pN1, mpd 12.4) and passage 4 (pD2, mpd 9.8) respectively. Although similar numbers of donor nuclei were present in muscles grafted with both preparations ([Figure 2f](#pone-0017454-g002){ref-type="fig"}), pN1, with a fusion index of 4% at mpd7 ([Figure 2b](#pone-0017454-g002){ref-type="fig"}, [Table 1](#pone-0017454-t001){ref-type="table"}) gave rise to only 1.33±0.33 (n = 6) fibres of donor origin (human spectrin+ fibres containing human lamin A/C+ nuclei) *in vivo* ([Figure 2d](#pone-0017454-g002){ref-type="fig"},), whereas pD2, with a 12--25% fusion index at mpd10.8 *in vitro* ([Figure 2c](#pone-0017454-g002){ref-type="fig"} and [Figure S3j](#pone.0017454.s003){ref-type="supplementary-material"}), formed significantly more muscle fibres of donor origin ([Figure 2e](#pone-0017454-g002){ref-type="fig"}) (P = 0.0020) *in vivo* (22.83±5.19, n = 6) than pN1 cells ([Fig. 2f, g](#pone-0017454-g002){ref-type="fig"}). 3. Human mdcs fail to contribute to skeletal muscle regeneration after intra-arterial delivery {#s3c} ---------------------------------------------------------------------------------------------- 2 hours after intra-arterial injection of pN1, human lamin A/C+ cells were detected in downstream TA muscles of the injected leg (3.75±0.48 per representative section), both inside ([Figure 2i](#pone-0017454-g002){ref-type="fig"}) and outside the blood vessels ([Fig. 2h](#pone-0017454-g002){ref-type="fig"}). A few human lamin A/C+ cells were also detected in the lungs ([figure 2j](#pone-0017454-g002){ref-type="fig"}) of all mice examined at 2 hours after grafting, but no human cells were detected in the liver. 4 weeks after intra-arterial injection of pN1 and pD2, there were only 1.11±0.30 (pN1) and 8.60±7.41 (pD2) human nuclei per representative section of TA muscles of the grafted leg ([Table 3](#pone-0017454-t003){ref-type="table"}, [Figure 2k](#pone-0017454-g002){ref-type="fig"}); no muscle fibres of human origin were detected in any TA muscles examined. At 4 weeks after grafting, no human cells were detected in 7 lungs analysed. We also performed experiments in which we did 2 intra-arterial injections of the same cells into one femoral artery of 6 mdx nu/nu mice at 28 days apart, sampling downstream TA and gastrocnemius muscles 42 days after the last injection. We found no muscle of donor origin following these 2 injections. 4. CD56+ and CD56- subpopulations contribute differently to myogenesis *in vitro* and skeletal muscle regeneration *in vivo* {#s3d} ---------------------------------------------------------------------------------------------------------------------------- Our human mdc preparations contained between 0 and 40% CD56+ cells ([Fig. S1](#pone.0017454.s001){ref-type="supplementary-material"}), which may have been either satellite cell-derived myoblasts, or were generated during the culture period from another cell type (e.g. pericytes). To determine whether CD56+ and CD56− cells differed in their myogenic potential, we separated these 2 cell populations from pD2 by flow cytometry and analyzed their *in vitro* characteristics and myogenic properties and *in vivo* contribution to myofibre regeneration. ### 4.1 pD2 cells that do not express CD56 give rise to CD56+ myogenic cells {#s3d1} Flow cytometric analysis of CD56 expression was performed on pD2 at mpds 10.773, 16.165 and 22.273. The percentage of CD56+ cells at each mpd was 6.20%, 32.86% and 8.69% respectively, showing that this molecule is not expressed at consistent levels during cell culture. The CD56+ and CD56− cells separated by FACS at D0 ([Figure 3a](#pone-0017454-g003){ref-type="fig"}) were then cultured as described in [material and methods](#s2){ref-type="sec"} section 2.2 and expression of CD56 was determined by flow cytometry at 17 and 30 days after the initial separation. At 17 days after separation, there were 85.72% CD56+ cells present in the CD56+ subpopulation ([Figure 3b](#pone-0017454-g003){ref-type="fig"}), and 4.40% CD56+ cells present in the CD56− subpopulation ([Figure 3d](#pone-0017454-g003){ref-type="fig"}). However, 30 days after separation, there were only 17.62% CD56+ cells within the subpopulation that was originally 100% CD56+ ([Figure 3c](#pone-0017454-g003){ref-type="fig"}), but in contrast, 39.92% CD56+ cells were present in the originally CD56− subpopulation ([Figure 3e](#pone-0017454-g003){ref-type="fig"}). This shows that the CD56− fraction (likely to contain human skeletal muscle pericytes) generates cells capable of myogenesis, as shown by Dellavalle and colleagues [@pone.0017454-Dellavalle1]. ::: {#pone-0017454-g003 .fig} 10.1371/journal.pone.0017454.g003 Figure 3 ::: {.caption} ###### Dynamic expression of CD56 by pD2 cells during culture. FACS sorting of CD56+ (32.86%) and CD56− (34%) subpopulations from pD2 at D0. b, c. Sorted CD56+ cells were maintained in culture for 17 days (b) and 30 days (c). FACS analysis of CD56 expression show that the percentage of CD56+ cells in the initial CD56+ subpopulation was 85.72% (b) at 17 days and 17.42% (c) at 30 days. d, e. Sorted CD56− cells were maintained in culture for 17 days (d) and 30 days (e). FACS analysis of CD56 expression show that the percentage of CD56+ cells in the initial CD56− subpopulation was 4.40% (d) at 17 days and 39.92% (e) at 30 days. :::  ::: ### 4.2 CD56+ and CD56− display different myogenic capacities both in vitro and in vivo {#s3d2} CD56+ and CD56− cells were left to differentiate ([material and methods](#s2){ref-type="sec"} section 3.4) for up to 21 days. Both CD56+ and CD56− cells differentiated into myotubes *in vitro*, but with different kinetics ([Figure 4](#pone-0017454-g004){ref-type="fig"}). CD56+ cells gave rise to myotubes from the second day after initiating differentiation (data not shown). However, CD56− cells were much slower in giving rise to myotubes. Expression of CD56, MF20, Myf5, desmin, MyoD and myogenin by differentiating cells shows that, although CD56+ cells initiated differentiation more rapidly, CD56− cells differentiated to the same extent after a delay of 3--5 days ([figure 4g](#pone-0017454-g004){ref-type="fig"}). The decrease in fusion index in CD56+ cells may due to loss of the already fully differentiated myotubes during the long term *in vitro* maintenance, while the increase of the fusion index in CD56− cells may due to the continuous maturation of the myotubes during the culture period. The *in vitro* myogenic pattern of CD56− cells was similar to that of human pericytes [@pone.0017454-Dellavalle1]. ::: {#pone-0017454-g004 .fig} 10.1371/journal.pone.0017454.g004 Figure 4 ::: {.caption} ###### CD56+ and CD56− mdscs have different kinetics of myogenesis *in vitro*. a, b, c, d, e, f. *In vitro* myogenesis of CD56+ (a, b, c) and CD56− (d, e, f) cells after 7 days (a, d), 14 days (b, e) and 21 days (c, f) differentiation. Myosin (MF20, green) immunostaining was performed to determine the fusion index at each time point. Total nuclei were counterstained with 10 µg/ml DAPI (blue). Scale bar  = 50 µm. g. Quantification of CD56+, Myosin (MF20)+, Myf5+, Desmin+, MyoD+ and myogenin+ cells within CD56+ and CD56− cell population during *in vitro* differentiation. :::  ::: Next, CD56+, CD56− and non-sorted cells from pD2 was transplanted intra-muscularly to investigate their contribution to muscle regeneration. 4 weeks after grafting, CD56+ cells gave rise to significantly more muscle cells or fibres of donor origin than either CD56− (p\<0.0001) or non-sorted cells (p\<0.0001) ([Figure 5](#pone-0017454-g005){ref-type="fig"} and [Table 4](#pone-0017454-t004){ref-type="table"}). Interestingly, when we compared the percentage of donor nuclei located inside the basal lamina (which must be either myonuclei or satellite cells) between the three groups, the highest percentage was found in the muscles grafted with CD56+ cells (82.15±5.94%, mean ± SEM at 4 weeks and 86.86±3.60%, mean ± SEM at 8 weeks after grafting). This indicates that a higher percentage of cells within the CD56+ population were myogenic, forming either myonuclei or satellite cells, than in the CD56− or non-sorted populations ([Figure 5](#pone-0017454-g005){ref-type="fig"}). On grafting CD56+ cells, there were significantly fewer cells (p = 0.002) and muscle fibres of donor origin (p = 0.002 Student\'s t test) at 8, compared to 4 weeks. However, in muscles grafted with either CD56− or non-sorted cells, there were similar amounts of both donor muscle fibres and donor nuclei at 4 and 8 weeks after grafting ([Table 4](#pone-0017454-t004){ref-type="table"}). ::: {#pone-0017454-g005 .fig} 10.1371/journal.pone.0017454.g005 Figure 5 ::: {.caption} ###### Contribution of CD56+, CD56− and non-sorted pD2 cells to muscle regeneration after intra-muscular transplantation into cryodamaged TA muscles of mdx nu/nu mice. a--c. Representative images showing the contribution of CD56+ (a), CD56− (b) and non-sorted pD2 cells (c) to muscle regeneration determined by human lamin A/C and human spectrin (both green) immunostaining. Total nuclei were counterstained with 10 µg/ml DAPI (blue). Scale bar  = 50 µm. d--f. Quantitative comparison of human lamin A/C+ nuclei (d), human spectrin+ fibres (e) and spectrin+ fibres containing human lamin A/C+ nuclei (f) among 3 groups at 4 weeks and 8 weeks after transplantation. g. Quantitative comparison of the percentage of human lamin A/C+ nuclei either inside or outside the basal lamina in each group at 4 weeks and 8 weeks after transplantation. Numbers within the yellow bar show the number of human lamin A/C+ nuclei inside the basal lamina, numbers within the blue bar show the number of human lamin A/C+ nuclei outside basal lamina, and numbers above each bar show the total number of human lamin A/C+ nuclei in each group. :::  ::: ::: {#pone-0017454-t004 .table-wrap} 10.1371/journal.pone.0017454.t004 Table 4 ::: {.caption} ###### Intra-muscular transplantation of CD56+, CD56− and non-sorted pD2 cells into cryodamaged TA muscles of mdx nu/nu mice: contribution to human nuclei and human muscle fibres. ::: {#pone-0017454-t004-4} A. 4 weeks after grafting ----------------------------------------------------------- ---------------------- --------------------- ---------------------- ----------- hLamin A/C+ nuclei (Mean±SEM) 331.0±10.32 (n = 4) 70.75±13.70 (n = 8) 129.13±20.78 (n = 8) p\<0.0001 \%hLamin A/C+ nuclei inside BL (Mean±SEM) 82.2±5.9 (n = 4) 31.2±3.0 (n = 8) 31.2±4.9 (n = 8) p\<0.0001 \%hLamin A/C+ nuclei outside BL (Mean±SEM) 17.8±5.9 (n = 4) 68.8±3.0 (n = 8) 68.8±4.9 (n = 8) p\<0.0001 hSpectrin+fibres(Mean±SEM) 308.0±40.68 (n = 4) 62.25±10.90 (n = 8) 60.25±11.12 (n = 8) p\<0.0001 hSpectrin+ fibres containing hLaminA/C+ nucleus(Mean±SEM) 213.25±15.46 (n = 4) 32.63±6.0 (n = 8) 34.88±5.63 (n = 8) p\<0.0001 B. 8 weeks after grafting ----------------------------------------------------------- ---------------------- --------------------- --------------------- ------------ hLamin A/C (Mean±SEM) 178.88±24.54 (n = 8) 47.88±15.08 (n = 8) 75.75±18.40 (n = 8) p = 0.0003 \%hLamin A/C inside BL (Mean±SEM) 86.9±3.6 (n = 8) 70.2±7.7 (n = 7) 70.0±3.2 (n = 8) P = 0.0371 \%hLamin A/C outside BL (Mean±SEM) 13.1±3.6 (n = 8) 29.8±7.7 (n = 7) 30.0±3.2 (n = 8) P = 0.0371 hSpectrin+ fibres(Mean±SEM) 150.50±17.68 (n = 8) 33.63±9.87 (n = 8) 54.25±14.60 (n = 8) p\<0.0001 hSpectrin+ fibres containing hLaminA/C+ nucleus(Mean±SEM) 110.25±13.26 (n = 8) 24.25±7.28 (n = 8) 37.00±10.33 (n = 8) p\<0.0001 ::: ### 4.3 Phenotypic differences of CD56+ and CD56− subsets of pD2 {#s3d3} To understand why CD56+ and CD56− cells behaved so differently in terms of myogenesis and to determine if there is an inter-relationship between CD56− and + cells, pD2 cells were sorted on the basis of CD56 expression and analyzed to compare their marker expression and proliferation properties. RT-PCR analysis showed that neither CD56+ nor CD56− cells expressed Pax7, but both expressed similar levels of Myf5 and MyoD. However, the CD56+ cell population expressed more desmin and MHC than the CD56− cell population, suggesting that CD56+ cells were more differentiated than the CD56− population under proliferative conditions ([Figure 6a](#pone-0017454-g006){ref-type="fig"}). ::: {#pone-0017454-g006 .fig} 10.1371/journal.pone.0017454.g006 Figure 6 ::: {.caption} ###### Phenotypic differences between CD56+ and CD56− cells. a\. Expression of MRFs and other relevant markers at the RNA level in 2 batches of total RNAs extracted from 2 separate sorts of CD56+ and CD56− cells. b. Graph showing BrdU incorporation in CD56+ and CD56+ cells. c, d. Expression of desmin (green) in CD56+ (d) and CD56− (e) cells. Nuclei were counterstained with 10 µg/ml DAPI (blue). Scale bar  = 50 µm. e, f. Expression of PDGFRβ (green) in CD56+ (f) and CD56− (g) cells. Nuclei were counterstained with 10 µg/ml DAPI (blue). Scale bar  = 50 µm. g, h. Expression of BrdU (green) in CD56+ (h) and CD56− (i) cells. Nuclei were counterstained with 10 µg/ml DAPI (blue). Quantification of BrdU+ cells in each population was compared with Student\'s t test (b). Scale bar  = 100 µm. :::  ::: Both CD56+ and CD56− cells expressed pericyte markers such as ALP and NG2 at similar levels. In addition, both cell populations expressed similar level of CD34 and CD144 at the RNA level ([Figure 6b](#pone-0017454-g006){ref-type="fig"}). Similar results were obtained by immunofluorescent staining for the myogenic markers Pax7, Myf5, MyoD and desmin. Pax7 protein was not present on either CD56+ or CD56− cells and more CD56+ cells expressed Myf5 ([Figure S2](#pone.0017454.s002){ref-type="supplementary-material"}) and desmin ([Figure 6d](#pone-0017454-g006){ref-type="fig"}) in comparison with CD56− cells ([Figure S2](#pone.0017454.s002){ref-type="supplementary-material"}, 6e). MyoD was expressed in the more confluent areas of CD56+ cell cultures, whereas fewer CD56− cells expressed MyoD ([Figure S2](#pone.0017454.s002){ref-type="supplementary-material"}). There was similar level of PDGFRβ expression in CD56− and CD56+ cells ([Figure 6f, g](#pone-0017454-g006){ref-type="fig"}). BrdU staining showed that 40.34±2.193 of CD56+ cells and 52.44±3.691 of CD56− cells (p = 0.048) were in S phase 3 days after plating in proliferative medium, another indication that the CD56+ sub-population might be undergoing slightly more terminal differentiation than CD56− cells ([Figure 6c, h, i](#pone-0017454-g006){ref-type="fig"}). In summary, the above data suggest that the CD56 expression by pD2 cells was highly dynamic, with CD56+ cells being more terminally differentiated myogenic cells than CD56− cells, as indicated by their expression of desmin and MHC. However, the expression of pericyte markers such as ALP, PDGFRβ and other endothelial markers such as CD34 and CD144 at the RNA level, were similar in CD56+ and CD56− cells, suggesting that both CD56+ and CD56− cells contained multiple cell types, including pericytes, endothelial cells and possibly other cell types. Discussion {#s4} ========== A major goal of studies on the contribution of stem cells to skeletal muscle regeneration is to identify the best human stem cell for treatment of muscular dystrophies [@pone.0017454-Benchaouir1], [@pone.0017454-Dellavalle1], [@pone.0017454-Zheng1]. Such a stem cell should be systemically-deliverable, contribute to widespread muscle regeneration and ideally, functionally reconstitute the muscle stem cell pool. Pericytes - stem cells associated with blood vessels in postnatal tissues - have been derived from human skeletal muscle and shown to contribute to muscle regeneration after delivery via the femoral artery in the SCID mdx mouse, an immunodeficient model of DMD [@pone.0017454-Dellavalle1]. We therefore wished to replicate these promising findings, with the ultimate aim of translating stem cell therapy to the treatment of DMD. Stem cells within skeletal muscle {#s4a} --------------------------------- As skeletal muscle is capable of growth, repair and regeneration that can be prevented by local high doses of radiation [@pone.0017454-Quinlan1], [@pone.0017454-Wakeford1], there must be stem or progenitor cells that mediate these processes within the muscle itself. It is not clear how many stem cells (that can give rise to a more differentiated cell type and self-renew) [@pone.0017454-RamalhoSantos1], as opposed to progenitor cells that cannot self-renew, there are within skeletal muscle. At least some satellite cells are muscle stem cells, able to regenerate muscle fibres and reconstitute the satellite cell niche with functional satellite cells [@pone.0017454-Collins1]. Other stem/progenitor cells within skeletal muscle include blood-vessel associated mesoangioblasts [@pone.0017454-Cossu1] or pericytes [@pone.0017454-Chen1], CD133+ cells [@pone.0017454-Benchaouir1], side population cells [@pone.0017454-Asakura1], [@pone.0017454-Uezumi1], muscle derived stem cells [@pone.0017454-Huard1]--[@pone.0017454-Tamaki1], endothelial cells [@pone.0017454-Tamaki2], myoendothelial cells [@pone.0017454-Zheng1] and other interstitial cells [@pone.0017454-Mitchell1]. However, the relationship of these cell types to each other and the extent to which they contribute to muscle regeneration and/or reconstitute the satellite cell pool are unclear. Difficulties in studying stem cells from skeletal muscle {#s4b} -------------------------------------------------------- A major problem with studying the different precursor or stem cells within skeletal muscle, especially when possible clinical applications are considered, is preparing pure populations of cells for study. The different methodologies for isolation, culture and analysis of muscle-derived cells by different laboratories have made it extremely difficult to make comparisons between the studies. In addition, different muscles, or muscles from individuals of different age [@pone.0017454-Conboy1], [@pone.0017454-Carlson1], sex [@pone.0017454-Deasy1], or affected by different pathological conditions, may contain different types or numbers of stem cells, or they may function in a different way. There is no standard method for releasing all cells from skeletal muscles. In addition, alterations to connective tissue that occur as a consequence of age or pathology make extraction of cells from muscle even more challenging. The methods used, e.g. explant culture, enzymatic disaggregation [@pone.0017454-Tonlorenzi1], [@pone.0017454-Brimah1], [@pone.0017454-Ehrhardt1] and, for enzymatic disaggregation, the type of enzyme used [@pone.0017454-Dellavalle1], [@pone.0017454-Zheng1], [@pone.0017454-Eberli1] and the timing the of disaggregation may give rise to preparations containing either different types or proportions of cell types and may also remove cell surface markers [@pone.0017454-Abuzakouk1]. Pre-plating was originally used to remove the more adherent fibroblasts from a muscle cell preparation [@pone.0017454-Cao1], but has more recently been shown to enrich for a particular type of muscle stem cell [@pone.0017454-QuPetersen1]. Sorting on the basis of cell surface antigens is frequently used to define populations of stem cells [@pone.0017454-Zheng1], [@pone.0017454-Jankowski1], but does not give rise to 100% pure populations of cells and a minority cell within the preparation may give rise to misleading results. Sorting a specific stem cell type from freshly-disaggregated postnatal human skeletal muscle on the basis of cell surface marker expression has been reported for AC133+ cells [@pone.0017454-Benchaouir1] and myoendothelial cells [@pone.0017454-Zheng1]; however, a limitation of this approach is that too few cells are released from a small biopsy for immediate sorting [@pone.0017454-Zheng1], therefore sorting is often done following expansion of cells *in vitro*, which itself leads to changes in muscle-derived cells [@pone.0017454-Jankowski1]. The *in vitro* environment may either affect a cell directly, by altering gene expression and phenotype [@pone.0017454-Lecourt1]--[@pone.0017454-Vauchez1], or indirectly, by allowing one sub-population to predominate over another. It is thought that the reason why culturing mouse satellite cells has a profoundly detrimental effect on their subsequent ability to regenerate skeletal muscle *in vivo* is because the cells begin to differentiate in culture [@pone.0017454-Montarras1]. Different muscles may contain either different numbers, or types, of different muscle stem cell populations, but experiments comparing the same stem cell derived from different human muscles have not been done. In the experiments described here, the same protocol was used by the same person to prepare cells from eight different muscles (3 para-spinal, 3 quadriceps and 2 EDB muscles, [Table 1](#pone-0017454-t001){ref-type="table"}). However, only two of these preparations, pD2 and pD3, derived from the right and left EDB muscles of the same DMD patient, were highly myogenic. One would need to perform many more experiments to determine whether mdcs/pericytes from one muscle are more myogenic than the same cells from another muscle from the same individual, or whether age, sex or pathological status has an effect on number, location or activation status of resident stem or precursor cells. As we did not use the same muscles as Dellavalle et al for our cell preparations, direct comparison of our data with theirs is not possible. However, our data complement and extend their promising findings [@pone.0017454-Dellavalle1]. Our most myogenic mdc preparation, pD2, was derived from a Duchenne muscular dystrophy patient aged 11 that had a deletion of exons 45--50 in the dystophin gene, that could be put back in frame by skipping exon 51 [@pone.0017454-Kinali1]. These cells also provide an invaluable tool for studying the combination of exon skipping strategies with stem cell therapy, which is ongoing in our laboratory. Comparison of muscle stem cells prepared by different laboratories {#s4c} ------------------------------------------------------------------ Pericytes expressed annexin V, alkaline phosphatase (ALP), desmin, smooth muscle actin, vimentin, PDGFRβ, nestin, CD13 and CD44 but did not express M-cadherin, N-CAM (CD56), cytokeratins, CD31, CD34, KDR, CD45, CD62L, CD71, CD106, CD117, CD133. Cells were weakly-positive for CD49b, CD63, CD90, CD105 and CD146. Clonal analysis showed that 20% of cells expressed smooth muscle actin or desmin, 50% expressed NG2 and 90% expressed PDGF receptor β, and these percentages did not change with successive passages in culture. Only on terminal differentiation did they express MyoD and Myf5 [@pone.0017454-Dellavalle1]. Our human mdcs, prepared and cultured as described previously [@pone.0017454-Dellavalle1], were similar phenotypically to pericytes, the main difference being in the expression of myogenic markers -- our cells expressed Myf5, MyoD, MHC and Pax7 at the RNA level at early stages under proliferative conditions, whereas pericytes did not. This difference may be due either to the different sensitivities of the RT-PCR methods used by the two laboratories, or to the fact that our preparations, despite being prepared in the same way, were a mixture of cell types, rather than being derived from pericytes alone. Another difference is that 5 out of 8 of our cell preparations expressed CD56 (NCAM) [@pone.0017454-BellesIsles1]. CD56 is a regulator of cell adhesion, intracellular signaling and cytoskeletal dynamics [@pone.0017454-Ditlevsen1] and is expressed by human satellite cells and myoblasts [@pone.0017454-Fidzianska1], myoendothelial cells [@pone.0017454-Zheng1], NK cells [@pone.0017454-Moretta1], but not pericytes [@pone.0017454-Dellavalle1]. However, non-myogenic cells may express NCAM if they become committed to the myogenic lineage: human synovial-derived mesenchmal stem cells, which do not express CD56, did so after being manipulated to overexpress MyoD and induced to undergo myogenic differentiation ([Materials and Methods S1](#pone.0017454.s007){ref-type="supplementary-material"} and [Figure S3a](#pone.0017454.s003){ref-type="supplementary-material"}--h). Both CD56+ and -- subpopulations of pD2 expressed PDGFR-β, ALP, NG2 and CD146, suggesting the presence of pericytes within the preparation. We found that both the CD56+ and CD56− sub-populations of pD2 mdcs underwent myogenesis *in vitro*, with CD56+ cells differentiating into myotubes more rapidly than CD56− cells. Our data suggest, confirming the findings of Dellavalle et al, that CD56− cells may represent a more primitive stem cell that can give rise to more differentiated CD56+ cells during *in vitro* culture ([Figure 3](#pone-0017454-g003){ref-type="fig"}). In support of this hypothesis, we found that CD56− cells expressed fewer myogenic regulatory factors than CD56+ cells ([Figure 6a](#pone-0017454-g006){ref-type="fig"}), implying that they are less committed to the myogenic lineage than CD56+ cells. In addition, CD56− cells showed delayed myogenesis *in vitro* compared to CD56+ cells, possibly because they need to express CD56 before committing to myogenesis ([Figure 4](#pone-0017454-g004){ref-type="fig"}). Finally, CD56− cells proliferated more rapidly than CD56+ cells ([Figure 6c, h and i](#pone-0017454-g006){ref-type="fig"}), again indicating that they were less committed to differentiation. However, future experiments using clonal analysis would be necessary to determine whether CD56− cells do indeed give rise to CD56+ progeny. Contribution of different muscle-derived cells to muscle regeneration {#s4d} --------------------------------------------------------------------- Pericytes, mesoangioblasts and AC133+ cells contribute to muscle regeneration after intra-arterial delivery [@pone.0017454-Benchaouir1], [@pone.0017454-Dellavalle1], [@pone.0017454-Sampaolesi1], [@pone.0017454-Sampaolesi2], but our mdcs did not. This may be because either mdc preparations did not contain pure pericytes, or that different host mouse strains were used. Both mdx nu/nu mice and mdx/SCID mouse strains have the same mutation in their dystrophin gene, but are on different genetic backgrounds and are immunocompromised in slightly different ways: SCID mice have lower B cell function than nu/nu mice, although both lack T cells and retain macrophage and NK cell activity. Furthermore, SCID mice are "leaky" and may generate mature lymphocytes as they age [@pone.0017454-Bosma1]. However, neither nu/nu nor SCID mice are the optimal host for human cell engraftment. To determine whether this had a significant effect on donor cell engraftment, we grafted the same donor mdscs into more highly immunodeficient C5-/Rag2-/gamma chain- host mice [@pone.0017454-Cooper1], which are deficient in T, B and NK cells and have no innate immunity [@pone.0017454-SilvaBarbosa1], but are not dystrophic. We found similar amounts of muscle of donor origin when cells were grafted into cryodamaged TA muscles of mdx nu/nu and C5-/Rag2-/gamma chain- host mice ([Figure S3i](#pone.0017454.s003){ref-type="supplementary-material"}), indicating that the choice of host strain did not have a negative effect on engraftment efficiency. Nevertheless, it remains possible that non-immunological factors within different mouse strains have an impact on the regenerative capacity of donor stem cells. Recent data have suggested that the genetic background of mdx mice affects endogenous muscle regeneration [@pone.0017454-Fukada1]. The genetic background of mdx SCID and mdx nu/nu mice [@pone.0017454-Partridge1] will be different and this may contribute to the differences between our findings and those of Dellavalle et al. Different muscle injury models used for intra-muscular grafting of putative muscle stem cells may also give rise to discrepancies between groups. We grafted cells into muscles that had been cryo-injured immediately prior to grafting [@pone.0017454-Brimah1], [@pone.0017454-Ehrhardt1], [@pone.0017454-SilvaBarbosa1], but Dellavalle et al. grafted cells into muscles that had been injected 48 hours previously with cardiotoxin, which induces muscle degeneration and regeneration [@pone.0017454-Dellavalle1]. Pisani et al. used the same injury regime as ours, but their hosts were immunodeficient Rag2-/gamma chain- non-dystrophic mice [@pone.0017454-Pisani1] that lack NK cells and may therefore be a better host for xenografts than SCID or nu/nu mice [@pone.0017454-SilvaBarbosa1]. Vauchez et al. grafted into muscles of non-dystrophic SCID mice, injuring the muscles prior to grafting by a combination of irradiation and notexin [@pone.0017454-Vauchez1]; Zheng et al. grafted cells into muscles of SCID mice, that had been injured by cardiotoxin one day previously [@pone.0017454-Zheng1]. How these different injury regimes model the dystrophic muscle environment and to what extent the local environment, genetic background and immunological status of the host mouse affect muscle stem cell behavior are important to determine, for the identification of robust methodologies which could be reliably used for therapeutic trials in muscular dystrophies. In addition to the injury model used, our intra-muscular grafting experiments and methods of analysis differ slightly from those of Dellavalle et al. All muscles were analysed a month after grafting, but whereas Dellavalle et al used a human-specific dystrophin antibody to identify fibres of human origin, we used antibodies to human spectrin and lamin a/c. This was because a human-specific dystrophin antibody would be non-informative for identifying muscle fibres derived from stem cells prepared from dystrophin-deficient DMD patients. Interestingly, although they contributed to much muscle regeneration after intra-arterial injection, pericytes only gave rise to very small numbers of muscle fibres after intra-muscular transplantation. CD56+/ALP- cells (satellite-cell derived myoblasts) gave rise to more muscle than CD56−/ALP+ cells (pericytes), but CD56−/ALP- cells, taken to be fibroblasts, made only the occasional donor muscle fibre [@pone.0017454-Dellavalle1]. We also found that both CD56+ and CD56− pD2 cells contributed to muscle regeneration, CD56+ cells making significantly more muscle than either CD56−, or non-fractionated, cells after intra-muscular transplantation. CD56+ cells contributed predominantly to nuclei inside the basal lamina of muscle fibres, i.e. within either muscle fibres and/or satellite cells. But CD56− or non-sorted cells contributed to significantly more nuclei outside the basal lamina ([Table 5](#pone-0017454-t005){ref-type="table"}), confirming that there were more non-myogenic cells within CD56− cell population. ::: {#pone-0017454-t005 .table-wrap} 10.1371/journal.pone.0017454.t005 Table 5 ::: {.caption} ###### Intra-muscular transplantation of CD56+, CD56− and non-sorted pD2 cells into cryodamaged TA muscles of mdx nu/nu mice: percentage of donor nuclei inside or outside the basal lamina. ::: {#pone-0017454-t005-5} A: 4 weeks after grafting --------------------------- ---------- ---------- --- ---------- ---------- --- CD56+ 82.15143 5.934744 4 17.84858 5.934745 4 CD56− 57.65585 2.987688 8 42.34415 2.987688 8 Non-sorted 31.18838 4.894127 8 68.81161 4.894127 8 B: 8 weeks after grafting --------------------------- ---------- ---------- --- ---------- ---------- --- CD56+ 86.85759 3.598443 8 13.14241 3.598443 8 CD56− 70.23909 7.657588 7 29.7609 7.657588 8 Non-sorted 70.01827 3.152388 8 29.98173 3.152388 8 ::: Zheng et al showed that human skeletal muscle-derived CD56+ cells that also expressed CD34 and CD144 (characteristic of myoendothelial cells) contributed to more muscle regeneration than did CD56+/CD34−/CD144− cells (myoblasts) [@pone.0017454-Zheng1]. This suggests that our CD56+ highly regenerative cell population may contain myoendothelial cells, but our facs analysis ([Figure S4](#pone.0017454.s004){ref-type="supplementary-material"}) showed that myoendothelial cells were rare. In addition, our CD56− sub-population contained very few endothelial cells, suggesting that pericytes, rather than endothelial cells, are the major CD56− contributor to muscle regeneration. Conclusions and future work {#s4e} --------------------------- In conclusion, despite most of our findings being in agreement with the findings of Dellavalle et al, we were not able to replicate the promising work showing that human pericytes gave rise to considerable muscle regeneration following intra-arterial injection in immunodeficient, dystrophin-deficient host mice [@pone.0017454-Dellavalle1]. It is not entirely clear why this might be, but it is likely that our cell preparations, despite being isolated according to the same protocol, consisted of a mixture of myoblasts, myoendothelial cells, endothelial cells and pericytes. In addition, there are some minor technical differences between our experimental protocols and those of Dellavalle et al. In both laboratories, cells were initially plated on collagen; however, Dellavalle et al. then expanded their cells on plastic, whereas we continued to use collagen. Pericytes were re-suspended in PBS for grafting, whereas mdscs were re-suspended in the medium in which they had been grown. Lastly, for intra-muscular injection, we resuspended 5×10^5^ cells in 5 µl, whereas Dellavalle et al. resuspended the same number of cells in 10 µl PBS [@pone.0017454-Dellavalle1]; for intra-arterial transplantation, we resuspended the cells in 25 µl medium, while Dellavalle et al. resuspended the same number of cells in 60--70 µl PBS for systemic injection [@pone.0017454-Sampaolesi2]. Although these are minor variables, they might account for the different results in the two laboratories. Nevertheless, despite inter-preparation variability, our mdcs were myogenic and contributed to muscle regeneration *in vivo*. Furthermore, we showed that mdcs contained both CD56+ and -- cells, with cells expressing CD56+, a marker of both myoblasts and myoendothelial cells, contributing to significantly more muscle regeneration than CD56− cells. However, to confirm that a particular stem cell does indeed contribute to muscle regeneration after systemic or local delivery, the cells would have to be prepared in a way that ensured that no contaminating cell type was present, which is technically challenging. In addition, the conditions under which the cells are expanded *in vitro* must maintain stem cell characteristics. The resolutions of this bottleneck will represent a significant step forward in the development of this approach for the treatment of muscular dystrophies. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Expression of PDGFRβ (green) and CD56 (green) in 3 human mdc preparations (pN1, pD1 and pD2) by flow cytometric analysis.** Control was performed using corresponding isotype control detailed in [Table S1](#pone.0017454.s006){ref-type="supplementary-material"} (red). Note the expression level of these cell markers were highly variable among cell preparations. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Expression of Myf5 and MyoD (both green) by CD56+ and CD56- sub-populations of pD2 cells.** a, a\' and a". Expression of Myf5 by CD56+ cells. b, b\' and b". Expression of Myf5 by CD56- cells. c, c\' and c". Expression of MyoD by CD56+ cells. d, d\' and d". Expression of MyoD by CD56- cells. Nuclei were counterstained with 10μg/ml DAPI (a, b, c and d). Scale bar = 50μm. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### a-h: Expression of CD56 on human synovial stem cells that had been lentivirally-transduced to express hMyoD, 7 days after induction of differentiation *in vitro*. a-d) normal, non-infected human synovial stem cells which were placed in differentiation medium in parallel with e-h) hMyoD synovial stem cells. Cells were stained for myosin (MF20, green) and CD56:PE antibodies (red). Nuclei were counterstained with 10μg/ml DAPI (blue). Scale bar = 50μm. i: Intramuscular transplantation of CD56+, CD56- and non-sorted pD2 cells into cryodamaged TA muscles of C5-/rag2-/γ chain- mice. Quantification of the number of human lamin a/c+ nuclei, human spectrin+ fibres and human spectrin+ fibres containing human lamin a/c+ nuclei showed that the contribution of each population of cells to C5-/rag2-/γ chain- mice were similar to that of transplantation into mdx nu/nu mice (as shown in [Figure 5](#pone-0017454-g005){ref-type="fig"}). j: *In vitro* myogenesis of pD2 cells at different mpds. Fusion index was determined by counting the number of nuclei within MF20+ myotubes. Although there were inter-experiment differences in fusion index, pD2 mdscs retained their capacity to differentiate into myotubes until at least 22.7 mpds. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **FACS analysis of expression of endothelial markers and double staining of CD56:PE and ALP, CD144, CD34 and CD146 on pD2 cells.** a, a\': expression of CD146 on pD2 cells. Approximately 76% of cells were CD146+ (a\'). Mouse IgG1:FITC was used as isotype control (a) of CD146:FITC antibody. b, b\': expression of CD144 on pD2 cells. 0.126% cells were CD144+ (b\'). Rabbit IgG:FITC was used as isotype control (b) of CD144:FITC antibody. c, d, e, f: Double staining of pD2 cells with CD56:PE in combination with ALP (c), CD146 (d), CD144 (e) and CD34 (f). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Angiogenesis of pD2 cells (mpd 20) in culture.** Tube formation can be observed 24 hours after being cultured in 10% FCS containing endothelial basal medium-2 on Matrigel substrate (a). Cells cultured in serum-free medium (b) were taken as negative control. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Antibodies used for FACS analysis or immunostaining.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Materials and Methods S1 ::: {.caption} ###### (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: The support of the MRC Centre for Neuromuscular Diseases Biobank is gratefully acknowledged. We thank Dr. Francesco Conti for his critical reading of the manuscript, and Dr. Ayad Eddaoudi and Mr. Prabhjoat Chana for their help with FACS sorting and analysis. We also thank Dr. Geraldine Edge and Mr. Paolo De Coppi for their help in obtaining the muscle biopsies. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the International Collaborative Effort for DMD (ICE), the Wellcome Trust (<http://www.wellcome.ac.uk/>) grant number 084241/Z/07/Z, and the Medical Research Council (<http://www.mrc.ac.uk/index.htm>) grant number G0900872. J.M. was funded by the International Collaborative Effort for DMD (ICE) and the Medical Research Council, and J.E.M. holds a Wellcome Trust University award. J.E.M. and F.M. are PIs of the MRC Centre for Neuromuscular Diseases. C.A. was funded by the Duchenne Parent Project, Netherlands. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JM JEM CA FM. Performed the experiments: JM JEM CA SX. Analyzed the data: JM JEM CA FM SX. Contributed reagents/materials/analysis tools: JM JEM CA. Wrote the paper: JM JEM CA FM.

PubMed Central

2024-06-05T04:04:19.182027

2011-3-9

PMC3052359

Introduction {#s1} ============ *Candida glabrata* is a haploid member of Ascomycetes normally not found in the environment but which has rather adapted to conditions found in mammals [@pone.0017589-Kaur1]. Among human fungal pathogens, *C. glabrata* is often reported as the second most prevalent species after *Candida albicans* [@pone.0017589-Ruhnke1],[@pone.0017589-Pfaller1]. *C. glabrata* can cause mucosal and bloodstream infection (BSI) mainly in immuno-compromised hosts. Worldwide, *C. glabrata* accounts for an average 11% of infections caused by *Candida* species, however this proportion varies from 7 to 20% depending on geographical locations [@pone.0017589-Pfaller2]. *C. glabrata* infections can be treated with several antifungal agents including amphotericin B, azoles and echinocandins [@pone.0017589-Sanglard1], [@pone.0017589-LassFlorl1]. However, *C. glabrata* can develop antifungal resistance and especially to the class of azole antifungals. Azole resistance surveillance studies have revealed a proportion varying from 10 to 20% of isolates with MIC values reaching clinical breakpoints (e.g. 64 µg/ml for fluconazole, based on CLSI standards). Several countries reported an increase in the proportion of azole-resistant isolates from 2001 to 2007 [@pone.0017589-Pfaller2]. *C. glabrata* is also known for exhibiting intrinsically higher azole MIC values than *C. albicans*. For example, the average of fluconazole MIC values of a *C. glabrata* wild type population is near a value of 4 µg/ml, while it is approximately 32-fold lower for *C. albicans* [@pone.0017589-Borst1], [@pone.0017589-Sanglard2]. We and others showed that azole resistance in *C. glabrata* was mediated almost exclusively by enhanced drug efflux and overexpression of multidrug transporters of the [A]{.underline}TP [B]{.underline}inding [C]{.underline}assette (ABC) transporters. Several genes encoding these transporters were identified including *CgCDR1*, *CgCDR2* (*PDH1*) and *CgSNQ2* [@pone.0017589-Sanglard2], [@pone.0017589-Torelli1], [@pone.0017589-Bennett1], [@pone.0017589-Sanglard3], [@pone.0017589-Izumikawa1]. Azole resistance in clinical isolates can be the result of overexpression of single or several transporters [@pone.0017589-Ferrari1]. The understanding of regulatory circuits controlling the expression of these genes has progressed in the recent years. A major regulator of these genes, *CgPDR1*, was identified [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai1]. This gene belongs to the family of zinc finger transcription factors and functionally resembles *PDR1* and *PDR3* from the baker\'s yeast *Saccharomyces cerevisiae*. Deletion of *CgPDR1* results in a loss of transcriptional control of the major transporters involved in azole resistance and, consequently, decreased resistance to these antifungals [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai1]. *CgPDR1* exhibits mutations, so called gain-of-function (GOF) mutations, which are responsible for intrinsic high expression of ABC transporters and therefore constitute the molecular basis of azole resistance in *C. glabrata* [@pone.0017589-Ferrari1], [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai1]. One striking feature of GOF mutations is their high diversity among *CgPDR1* alleles from azole-resistant isolates. As many as 67 mutations conferring azole resistance are described up to now [@pone.0017589-Ferrari1], [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai1], [@pone.0017589-Tsai2], [@pone.0017589-Berila1]. GOF mutations are found within several domains of the transcription factor corresponding to putative functional elements inferred from comparison to the *S. cerevisiae PDR1* and *PDR3* and including the transcriptional activation domain, a regulatory domain and a so-called middle homology region (MHR) which is found in several zinc finger proteins [@pone.0017589-Ferrari1], [@pone.0017589-Tsai2]. Not only are GOF mutations in *CgPDR1* important for azole resistance in *C. glabrata* but also for fungal-host interactions. We showed that GOF mutations were associated with enhanced virulence and fitness in animal models of systemic infection [@pone.0017589-Ferrari1]. This was unexpected since it is generally accepted that the development of drug resistance in other microbes is usually associated with costs in virulence or fitness. Secondary compensatory mechanisms can however restore the costs of resistance development [@pone.0017589-Andersson1], [@pone.0017589-Anderson1]. In this study we addressed in *C. glabrata* the identification of genes behind the GOF-dependent virulence of *CgPDR1*. Because we rationalized that some genes commonly expressed by GOF mutations could be responsible for this effect, we analysed with transcript profiling analysis *C. glabrata* isolates containing individual GOF mutations but in identical genetic backgrounds. Only two genes (*CgCDR1* and *PUP1*) were identified. We describe here their relevance in the enhanced virulence mediated by *CgPDR1* GOF mutations. Results {#s2} ======= Transcriptional analysis of GOF mutations {#s2a} ----------------------------------------- In a previous study, we reported a high variety of gain-of-function (GOF) mutations in the transcriptional activator *CgPDR1* [@pone.0017589-Ferrari1]. These mutations conferred azole resistance through the differentiated upregulation of several ABC transporters including *CgCDR1*, *CgCDR2* and *CgSNQ2*. It is known that *CgPDR1* controls the expression of many other genes, some of which contain a regulatory domain in their promoter matching the PDRE (Pleiotropic Drug Responsive Element) described in *S. cerevisiae* (TCCRYGSR) [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai2]. We were therefore interested to test whether the differentiated expression pattern observed for a few genes as described earlier [@pone.0017589-Ferrari1] could be generalized to the entire transcriptome of *C. glabrata*. In order to achieve this goal, labeled cRNA from mRNA isolated in triplicates from strains containing seven different *CgPDR1* GOF was applied to oligonucleotides custom arrays. The selection of GOFs was based on their occurrence in putative CgPdr1 functional domains including the regulatory domain (L280F, R376W), the MHR (Y584C, T588A) and the activation domain (D1082G, E1083Q). The GOF P822L was also selected since it was previously associated with a strong upregulation of *CgSNQ2* as compared to other ABC-transporters [@pone.0017589-Torelli1]. The format of one-color hybridization was chosen since it allows direct comparisons between any strains. The strains containing the different GOF were obtained by re-introduction of *CgPDR1* alleles at the genomic locus and were described in our previous study [@pone.0017589-Ferrari1]. As summarized in [Table 1](#pone-0017589-t001){ref-type="table"}, the number of genes differentially regulated (≥2-fold) by individual GOF as compared to the wild type *CgPDR1* ranges from 73 (for the R376W substitution) to 385 (for the T588A substitution) and no GOF regulated a similar number of genes. A total of 626 genes were regulated by at least one GOF (see [File S1](#pone.0017589.s005){ref-type="supplementary-material"}). The degree of similarity between transcription profiles in the 626 genes regulated could also be estimated with linear regression coefficients, which can establish the extent of gene co-regulation by pairs of separate GOF. As summarized in [Table 2](#pone-0017589-t002){ref-type="table"}, approximately half of r^2^ values from pairwise comparisons were above 0.5 (from 0.54 to 0.87) and thus signified a moderate trend towards the co-regulation of the genes by these GOFs. The highest correlation (r^2^ = 0.87) was observed between expression pattern of GOF D1082G (SFY103) with P822L (SFY116) ([Fig. 1A](#pone-0017589-g001){ref-type="fig"}, left side). One GOF (R376W) in SFY101 yielded systematically low r^2^ values with all other GOFs (between 0.0003 and 0.058). Increasing the cut-off for differential regulation to ≥3-fold did not significantly change r^2^ values (data not shown). The expression of genes obtained from GOF P822L (SFY116) and from R376W is shown to illustrate the low level of gene co-regulation between both isolates ([Fig. 1A](#pone-0017589-g001){ref-type="fig"}, right side). Taken together, these data support the concept that individual GOF result each in distinct transcription profiles even though the number of GOF analysed is probably only a portion of the entire mutation spectrum. ::: {#pone-0017589-g001 .fig} 10.1371/journal.pone.0017589.g001 Figure 1 ::: {.caption} ###### Expression profiles of *C. glabrata* genes regulated by GOFs in *CgPDR1*. **Panel A**: Pairwise comparisons of gene expression changes relative to SFY114 carrying the wild type *CgPDR1* allele. Each data point correlate the same gene expressed in strain SFY116 (P822L GOF) versus strain SFY103 (P1082G GOF) (left side) and in strain SFY116 (P822L GOF) versus strain SFY101 (R376W GOF). For each diagram, r^2^ values are given. **Panel B**: K-means clustering of the normalized expression levels of the 626 genes regulated (≥2-fold) by at least one *CgPDR1* GOF. Clustering was performed with Genespring® GX (parameters: Euclidian distance metric, 100 iterations, 4 clusters). For each cluster, enriched biological function and biological component were determined using GO terms of *S. cerevisiae* homologues. Results are given below the cluster analysis. :::  ::: ::: {#pone-0017589-t001 .table-wrap} 10.1371/journal.pone.0017589.t001 Table 1 ::: {.caption} ###### Number of *C. glabrata* genes regulated by ≥twofold in *PDR1* GOF mutants as compared to the wild type. ::: {#pone-0017589-t001-1} Strain *CgPDR1* GOF mutation Genes upregulated Genes downregulated Total -------- ----------------------- ------------------- --------------------- ------- SFY101 R376W 27 46 73 SFY103 D1082G 53 77 130 SFY105 T588A 235 150 385 SFY109 E1083Q 58 103 161 SFY111 Y584C 197 132 329 SFY115 L280F 67 132 199 SFY116 P822L 71 89 160 ::: ::: {#pone-0017589-t002 .table-wrap} 10.1371/journal.pone.0017589.t002 Table 2 ::: {.caption} ###### Correlation coefficients of transcriptional profiles. ::: {#pone-0017589-t002-2} GOF in *CgPDR1* allele L280F R376W Y584C T588A P822L D1082G E1083Q ------------------------ -------- -------- -------- -------- -------- -------- -------- L280F 1 0.016 0.6111 0.3107 0.7761 0.6979 0.8391 R376W 0.016 1 0.0588 0.0316 0.0003 0.0055 0.0003 Y584C 0.6111 0.0588 1 0.5491 0.7798 0.7012 0.7321 T588A 0.3107 0.0316 0.5491 1 0.4591 0.5596 0.4023 P822L 0.7761 0.0003 0.7798 0.4591 1 0.8704 0.7741 D1082G 0.6979 0.0055 0.7012 0.5596 0.8704 1 0.6984 E1083Q 0.8391 0.0003 0.7321 0.4023 0.7741 0.6984 1 ::: Given the diversity of transcriptional profiles provided by each GOF, the generated transcriptional data were clustered in a separate analysis in order to group sets of genes co-regulated by the different GOFs. Four separated groups were thus identified which were enriched in specific biological processes ([Fig. 1B](#pone-0017589-g001){ref-type="fig"}). It is noteworthy that genes from cluster 1 and 4 are enriched in processes related to amino acid metabolism, while others are enriched in signal transduction and protein metabolic processes. We closely inspected the transcription profiles of two isolates, one carrying the GOF mutation D1082G (SFY103) and the other the mutation P822L (SFY116). This choice was based on the fact that these profiles show the highest correlation (r^2^ = 0.87) and similar numbers of up-and downregulated genes, thus facilitating comparisons ([Table 1](#pone-0017589-t001){ref-type="table"} and [2](#pone-0017589-t002){ref-type="table"}). Between the two GOFs, 86 genes were co-regulated (32 upregulated and 54 downregulated) from the total of 626 genes regulated by at least one GOF. The upregulated genes in the SFY103 vs SFY116 comparison showed enrichment for xenobiotic transporter activity (p = 3.7E-3), while the downregulated genes exhibited enrichment in amino acid (arginine, glutamine) biosynthesis processes (p = 5.87E-07 to 2.97E-06). The inspection of conserved motifs in the promoters of upregulated genes yielded the consensus YCCACGGA ([Figure S3](#pone.0017589.s003){ref-type="supplementary-material"}), which closely resembled the PDRE recognition motif of *PDR1* in *S. cerevisiae* ((TCC\[AG\]\[CT\]G\[G/C\]\[A/G\]) [@pone.0017589-Devaux1]. These data are therefore consistent with the role of *CgPDR1* in the regulation of genes by the GOF mutations D1082G and P822L. To determine whether the expression of genes differentially regulated by the GOF mutations was also affected by the absence of *PDR1*, we analysed the expression profile of the *pdr1*Δ strain SFY92. A total of 247 genes were differentially regulated (≥2-fold) in strain SFY92 as compared to SFY114 (containing the *CgPDR1* wild type allele). Analysis of the 99 downregulated genes showed that one third of these genes encode for proteins predicted to be localized in the mitochondria. Moreover, enrichment of specific biological processes (oxidation-reduction, ATP synthesis coupled to electron transport chain, cellular respiration) was observed ([File S2](#pone.0017589.s006){ref-type="supplementary-material"}). Consistent with these observations is that *PDR1* and *PDR3* in *S. cerevisiae* are known to participate into the mitochondria-nucleus signalling pathway [@pone.0017589-Gulshan1], which may also be applied to *CgPDR1*. Finally, 121 genes were differentially regulated not only in absence of *PDR1* but also in the presence of GOF mutations, indicating that these genes might represent the basal set of *PDR1*-dependent genes. Virulence determinants in *C. glabrata* {#s2b} --------------------------------------- We reported that GOF mutations analysed here by transcriptional profiling in *C. glabrata* not only resulted in azole resistance but also in enhanced virulence and fitness in a mice model of infection [@pone.0017589-Ferrari1]. We reasoned that enhanced virulence could be due to specific genes commonly regulated by all *CgPDR1* GOFs, given that this phenotype was shared by all these mutations. Our current analysis revealed that no gene was commonly downregulated and only two genes were commonly upregulated by at least two-fold by all GOFs, i.e. *CgCDR1*, the well-known ABC-transporter involved in azole resistance, and the ORF CAGL0M12947g, which we named *PUP1* (for *[P]{.underline}DR1* [UP]{.underline}regulated gene) in the present study. This gene is highly similar to *YIL077c*, a gene encoding a protein of unknown function thought be located in the mitochondria. We tested this hypothesis in *C. glabrata* by the expression of a GFP-tagged version of *PUP1* in the azole-resistant clinical isolate DSY565. As shown in [Fig. 2](#pone-0017589-g002){ref-type="fig"}, the GFP signal could be detected in DSY565. Moreover, Mitotracker Red staining (Panel C), which specifically reveals mitochondrial punctuate and tubular structures, co-localized with GFP signals of Pup1-GFP. These results therefore confirmed that *PUP1* encodes a mitochondrial protein. ::: {#pone-0017589-g002 .fig} 10.1371/journal.pone.0017589.g002 Figure 2 ::: {.caption} ###### Localization of Pup1p in mitochondria. SFY174 cells expressing the Pup1p-GFP fusion protein were stained with Mitotracker Red and fixed as described in [Materials and Methods](#s4){ref-type="sec"}. **Panel A**: Nomarski images of the cells; **panel B**: Pup1p-GFP; **panel C**: mitochondria stained with Mitotracker Red; **panel D**: merging of B and C. Four individual images are shown. Bar, 5 µm. :::  ::: *CgCDR1* and *PUP1* are overexpressed by all GOFs and therefore they may constitute good candidates to be responsible for the enhanced virulence observed in animal models. *In vitro*, both genes were dependent on the presence of *CgPDR1* ([Fig. 3A](#pone-0017589-g003){ref-type="fig"}). Moreover, *PUP1* contains two PDREs in its promoter (−770 to −763: TCCACGGA; −740 to −733: TCCGTGGA) and *PUP1* expression was inducible by fluconazole ([Fig. 3B](#pone-0017589-g003){ref-type="fig"}) similarly to *CgCDR1*. Because they might be important for the enhanced virulence phenotype, these genes should also be expressed *in vivo*. We tested this hypothesis by injecting mice with strains expressing the GFP under the control of the *CgCDR1* promoter or fused to the *PUP1* ORF. Kidneys homogenates were recovered and analysed by flow cytometry to identify GFP-positive yeast cells. As shown in [Fig. 4](#pone-0017589-g004){ref-type="fig"}, GFP could be easily detected in the azole-resistant background DSY565 (SFY168) that expresses GFP under the control of the *CgCDR1* promoter. This was not the case in the DSY562 background (SFY167), where GFP expression driven by the *CgCDR1* promoter is low. Similarly, GFP signals in yeast cells expressing the GFP-tagged *PUP1* were detectable in the DSY565 background (SFY174), but not in the DSY562 background (SFY173). The results are consistent with the *in vitro* experiments performed with both GFP-tagged genes and thus indicate that *CgCDR1* and *PUP1* are overexpressed by *CgPDR1* GOF both *in vitro* and *in vivo*. ::: {#pone-0017589-g003 .fig} 10.1371/journal.pone.0017589.g003 Figure 3 ::: {.caption} ###### Expression of *CgCDR1* and *PUP1 in vitro*. **Panel A**: Expression of *CgCDR1* and *PUP1* in isolates containing distinct *CgPDR1* alleles. **Panel B**: Expression of *PUP1* after exposure to 256 µg ml^−1^ fluconazole during 150 min. Quantification was performed by qRT-PCR. The values are averages of three separate experiments and represent the increase in gene expression relative to DSY562 (set at 1.00). Strains were constructed from a *pdr1*Δ mutant and were named by the re-introduced GOF mutation or wild type *CgPDR1* allele. The indicated names correspond to the following strains: *pdr1*Δ: SFY92, *PDR1*: SFY114, L280F: SFY115, R376W: SFY101, Y584C: SFY111, T588A: SFY105, P822L: SFY116, D1082G: SFY103, E1083Q: SFY109). :::  ::: ::: {#pone-0017589-g004 .fig} 10.1371/journal.pone.0017589.g004 Figure 4 ::: {.caption} ###### Expression of *CgCDR1* and *PUP1 in vivo*. Flow cytometry analysis of GFP-positive yeast cells was performed from mice kidneys. Groups of 4 mice were injected intravenously with 4×10^7^ CFU of *C. glabrata* strains. Mice were sacrificed at day 7 post-infection. Results are expressed as percents of GFP-positive events in FACS and represent values recorded separately for each mouse. Asterisks indicate statistically significant differences (\*: *P*\<0.05; \*\*: *P*\<0.01, \*\*\*: *P*\<0.001). Strains SFY167 and SFY168 express the *CgCDR1p-3xGFP* construct and are derived from DSY562 and DSY565, respectively. Strains SFY173 and SFY174 express the *PUP1-3xGFP* construct and are derived from DSY562 and DSY565, respectively. As controls, kidneys of uninfected mice (mock) were analyzed alone or mixed with 1×10^7^ cells of SFY168 or SFY174 grown in YEPD. :::  ::: To test whether *CgCDR1* and *PUP1* were involved in *C. glabrata* virulence, mutants were constructed in both the genetic backgrounds of DSY562 and DSY565 resulting in strains SFY148 and SFY149 (*CgCDR1* mutants) and SFY150 and SFY151 (*PUP1* mutants), respectively. The deletion of the genes was verified by Southern analysis (see [Figure S2](#pone.0017589.s002){ref-type="supplementary-material"}). The constructed mutants were next injected intravenously in mice and mice survival was recorded over time. In this model, mice are immuno-compromised by cyclophosphamide treatment. In general, deletion of *CgCDR1* and *PUP1* in DSY562 background had no significant effects as compared to the azole-susceptible isolate DSY562 ([Fig. 5](#pone-0017589-g005){ref-type="fig"}). On the contrary, the deletion of *CgCDR1* or *PUP1* in DSY565 resulted in a significant decrease in virulence as compared to the wild type (SFY149 vs DSY565: p = 0.04; SFY151 vs DSY565: p = 0.02). Deleting both genes from DSY565 (SFY170) had a no significant effect as compared to single mutants. In addition, revertant isolates, SFY160 and SFY162, restored *PUP1* and *CgCDR1* expression, respectively, and the phenotype of the wild type parent. ::: {#pone-0017589-g005 .fig} 10.1371/journal.pone.0017589.g005 Figure 5 ::: {.caption} ###### Virulence of *C. glabrata* is dependent on *CgCDR1* and *PUP1*. Survival curves of mice infected with DSY562 (panel A) and DSY565 (panel B) and derived mutants. Statistical differences were performed using the Log-rank Mantel-Cox test (Prism 5.0) by comparing survival curves of mice infected by the parental strains (DSY562 or DSY565) and by other strains as indicated. Asterisks indicate statistically significant differences (\*: *P*\<0.05; \*\*: *P*\<0.01, \*\*\*: *P*\<0.001). NS indicates no significance (*P*\>0.05). For strains derived from DSY562, the indicated names correspond to the following strains: *pdr1*Δ: SFY92, *cdr1*Δ: SFY148, *CDR1*rev: SFY161, *pup1*Δ: SFY150, *PUP1*rev: SFY159, *cdr1*Δ, *pup1*Δ: SFY152. For strains derived from DSY565, the indicated names correspond to the following strains: *pdr1*Δ: SFY94, *cdr1*Δ: SFY149, *CDR1*rev: SFY162, *pup1*Δ: SFY151, *PUP1*rev: SFY160, *cdr1*Δ, *pup1*Δ: SFY153. :::  ::: Tissue burdens were assessed at day 7 post infection and are shown in [Fig. 6](#pone-0017589-g006){ref-type="fig"}. In this model, mice are immunocompetent and the endpoint measurement is not mice survival but rather tissue colonization by the infection agent. CFU values were compared with each other. In isolates derived from DSY562, it is interesting to observe that the deletion of *PUP1*, even if it did not result in a decrease of mice survival as compared to the wild type, significantly decreased kidney colonization. This decrease was compensated by the reintroduction of *PUP1* in the mutant (SFY160). This decrease was even more pronounced in the absence of both *PUP1* and *CgCDR1* (SFY169). In isolates derived from DSY565, the individual deletion of *CgCDR1* and *PUP1* (SFY150 and SFY151) decreased CFU counts in a significant manner as compared to the parent strain, a change which was restored by revertants of the corresponding genes. The double deletion of *PUP1* and *CgCDR1* decreased CFU counts in comparison to all other conditions, as observed from DSY5652-derived strains, indicating that *CgCDR1* and *PUP1* deletions have an additive effect on tissue colonization. ::: {#pone-0017589-g006 .fig} 10.1371/journal.pone.0017589.g006 Figure 6 ::: {.caption} ###### *C. glabrata* tissue burdens in murine infection models. Fungal tissue burdens in kidneys (panel A) and spleen (panel B) from BALB/c mice infected intravenously with 4×10^7^ viable cells of *C. glabrata* strains. Mice were sacrificed at day 7 post-infection. Results are expressed as CFUs per gram of tissue and represent values recorded separately for each of the ten mice. Geometric means are indicated by horizontal bars. Statistical comparisons are summarized above each panel. Asterisks indicate statistically significant differences (\*: *P*\<0.05; \*\*: *P*\<0.01, \*\*\*: *P*\<0.001). NS indicates no significance (*P*\>0.05). The symbol '-' indicates that the statistical comparison was not performed. Statistical differences were determined using the non-parametric Wilcoxon Rank sum tests (Prism 5.0). The origin of each strain is indicated; strain background (DSY562 and DSY565) is indicated by filled or empty symbols, respectively. See legend of [Fig. 5](#pone-0017589-g005){ref-type="fig"} for strain designations. :::  ::: Taken together, these results strongly suggest that *CgCDR1* and *PUP1*, two genes upregulated by all *CgPDR1* GOF mutations, are important for the enhanced virulence phenotype observed in the azole-resistant isolate DSY565. Decreased virulence from DSY565-derived strains was associated with decreased tissue colonization and mutant phenotypes could be reverted by the corresponding wild type genes. Overexpression of *CgCDR1* and *PUP1* in a *CgPDR1*-independent manner {#s2c} ---------------------------------------------------------------------- The overexpression of *CgCDR1* and *PUP1* is under the control of *CgPDR1* in *C. glabrata*. We showed in the above experiments that both *CgCDR1* and *PUP1* have impact on *C. glabrata* virulence. However, these experiments were carried out in the background of a functional *CgPDR1* and it is possible that other *CgPDR1*-dependent factors contribute to enhanced virulence of azole-resistant isolates. We therefore expressed *CgCDR1* and *PUP1* with a strong constitutive promoter (*TDH3*) in the background of a *CgPDR1* deletion strainto avoid interference with such factors. As observed in [Fig. 7](#pone-0017589-g007){ref-type="fig"}, the engineered strains could overexpress both genes at different levels but still to higher levels than *pdr1*Δ mutants. *CgCDR1* levels were approximately equal to those measured in the azole-resistant isolate DSY565 ([Fig. 7A](#pone-0017589-g007){ref-type="fig"}), while *PUP1* levels were higher (approx. 20-fold) when expressed under the control of the *TDH3* promoter than the native promoter ([Fig. 7B](#pone-0017589-g007){ref-type="fig"}). However, both genes were expressed to similar levels in DSY562 and DSY565 as expected from the constitutive expression from the *TDH3* promoter. Azole MICs strains were 32 µg/ml fluconazole in strains overexpressing *CgCDR1* via the *TDH3* promoter, while the fluconazole MICs were almost identical to the parent strains when *PUP1* was overexpressed (1--2 µg/ml, [Table 3](#pone-0017589-t003){ref-type="table"}), indicating that *CgCDR1* is the major mediator of azole resistance in our strains. ::: {#pone-0017589-g007 .fig} 10.1371/journal.pone.0017589.g007 Figure 7 ::: {.caption} ###### Overexpression of *CgCDR1* and *PUP1* in a *CgPDR1*-independent manner. **Panel A**: *TDH3*-dependent expression of *CgCDR1*. **Panel B**: *TDH3*-dependent expression of *PUP1*. Quantification was performed by qRT-PCR. The values are averages of three separate experiments and represent the increase in gene expression relative to SFY196 (set at 1.00). Strains derived from DSY562 are represented by black bars and the indicated names correspond to the following strains: *PDR1*: SFY196, *pdr1*Δ: SFY198, *pdr1*Δ+*TDH3p-CDR1*: SFY200, *pdr1*Δ+*TDH3p-PUP1*: SFY202. Strains derived from DSY565 are represented by white bars and the indicated names correspond to the following strains: *PDR1^L280F^*: SFY197, *pdr1*Δ: SFY199, *pdr1*Δ+*TDH3p-CDR1*: SFY201, *pdr1*Δ+*TDH3p-PUP1*: SFY203. :::  ::: ::: {#pone-0017589-t003 .table-wrap} 10.1371/journal.pone.0017589.t003 Table 3 ::: {.caption} ###### Fluconazole susceptibilities of *CgCDR1*, *PUP1* and *CgPDR1* mutant strains derived from strains DSY562 and DSY565. ::: {#pone-0017589-t003-3} Fluconazole MIC (µg ml^−1^)[a](#nt101){ref-type="table-fn"} -------- ------------------------------------------------------------- --- --- ---- ----- ----- --- ---- --- DSY562 4 1 1 2 4 4 1 32 1 DSY565 128 1 4 64 128 128 2 32 2 a MICs were determined by the broth microdilution method according to EUCAST document EDef 7.1 [@pone.0017589-EUCASTAFST1]. ::: The strains were next injected intravenously in mice and tissue burden were next assessed from kidneys and spleen from sacrificed animals ([Fig. 8](#pone-0017589-g008){ref-type="fig"}). In general, when *CgCDR1* and *PUP1* were overexpressed in a *pdr1*Δ mutant background, tissue burdens were significantly increased as compared to the parent strains. The colonization was slightly lower when *PUP1* was overexpressed as compared to *CgCDR1*. ::: {#pone-0017589-g008 .fig} 10.1371/journal.pone.0017589.g008 Figure 8 ::: {.caption} ###### Effect of *CgCDR1* and *PUP1* overexpression on tissue colonization. **Panel A**: Fungal tissue burdens in kidneys. **Panel B**: Fungal tissue burdens in spleen. Tissue burden were determined from BALB/c mice infected intravenously with 4×10^7^ viable cells of *C. glabrata* strains. Mice were sacrificed at day 7 post-infection. Results are expressed as CFUs per gram of tissue and represent values recorded separately for each of the ten mice. Geometric means are indicated by horizontal bars and asterisks indicate statistically significant differences (\*: *P*\<0.05; \*\*: *P*\<0.01, \*\*\*: *P*\<0.001). NS indicates no significance (*P*\>0.05). Statistical differences were determined using the non-parametric Wilcoxon Rank sum tests (Prism 5.0). Strain background (DSY562 and DSY565) is indicated by filled or empty symbols, respectively. For strains derived from DSY562, the indicated names correspond to the following strains: DSY562-TDH3p: SFY196; *pdr1*Δ-TDH3p: SFY198; *pdr1*Δ-TDH3p-*CDR1*: SFY200 *pdr1*Δ-TDH3p-*PUP1*: SFY202. For strains derived from DSY565, the indicated names correspond to the following strains: DSY562-TDH3p: SFY197; *pdr1*Δ-TDH3p: SFY199; *pdr1*Δ-TDH3p-*CDR1*: SFY201 *pdr1*Δ-TDH3p-*PUP1*: SFY203. :::  ::: When virulence of the same strains was tested in the immuno-suppressed mice model, the results showed no significant difference between strains overexpressing *CgCDR1* or *PUP1* as compared to the *pdr1*Δ mutants ([Fig. 9](#pone-0017589-g009){ref-type="fig"}). A closer inspection of the obtained data still suggests that strains overexpressing *CgCDR1* or *PUP1* tended to be more virulent than their parents. At day 15 post-infection, 90% of the mice infected with the *pdr1*Δ mutants survived, while approximately 70% survived when infected with the overexpressing strains ([Fig. 9](#pone-0017589-g009){ref-type="fig"}). ::: {#pone-0017589-g009 .fig} 10.1371/journal.pone.0017589.g009 Figure 9 ::: {.caption} ###### Virulence of *C. glabrata* in strains overexpressing *CgCDR1* and *PUP1*. Immuno-suppressed mice were infected as described in [Material and Methods](#s4){ref-type="sec"} with strain derived from DSY562 and DSY565. Statistical differences were performed using the Log-rank Mantel-Cox test (Prism 5.0) by comparing survival curves of mice infected by the strains as indicated. The comparison between DSY565-TDH3p and DSY565-TDH3p was significant (p = 0.04) while comparisons of strains overexpression *CgCDR1* and *PUP1* with parents (*pdr1*Δ-TDH3) was not significant. See legend of [Fig. 8](#pone-0017589-g008){ref-type="fig"} for strain designations. :::  ::: These results support the idea that the individual overexpression of *CgCDR1* and *PUP1* contributed moderately to virulence, however their overexpression was more important for maintaining tissue colonization. Taken together, our results indicate that both *CgCDR1* and *PUP1* are important mediators of *C. glabrata* virulence, but that their individual overexpression *per se* is not sufficient to mimic the increased virulence conferred by *CgPDR1* GOF mutations. Discussion {#s3} ========== In this study we analysed the expression profiles of GOF mutations obtained from azole-resistant isolates in a previous study [@pone.0017589-Ferrari1]. The analysis of transcription profiles gave only two genes commonly upregulated by all GOFs, *CgCDR1* and *PUP1*. Other investigators have analysed transcription profiles of azole-resistant isolates and thus enable comparisons with our study. Recently, Tsai *et al.* [@pone.0017589-Tsai2] obtained the transcription profiles of seven clinical pairs, each containing an azole-susceptible and an azole-resistant isolate. The *CgPDR1* GOF obtained from these strains were different from those investigated here, except for the L280F GOF. Their study highlighted 45 genes regulated (by ≥2-fold change as compared to the susceptible parent) by at least one clinical pair. Our study revealed a larger set of genes regulated by at least one GOF (i.e. 626 genes). *CgCDR1* and *PUP1*, the two genes selected in our study were found commonly upregulated by all GOFs in the Tsai *et al.* [@pone.0017589-Tsai2] study including by decreasing expression levels, *CgCDR1*, CAGL0M12947g (*PUP1*), CAGL0F02717g (*CgCDR2/PDH1*), CAGL0K00715g (*RTA1*), CAGL0C03289g (*YBT1*), CAGL0G00242g (*YOR1*), CAGL0K09702g, CAGL0A00451g (*CgPDR1*) and CAGL0G01122g. In a study published by Vermitsky *et al.* [@pone.0017589-Vermitsky1], one azole-resistant isolate (F15) was compared to an azole-susceptible parent. From the 109 genes regulated by at least two-fold in the resistant isolate, 34 were found regulated (out of 626 genes) in our study, among which *CgCDR1* and *PUP1*, the latter being the most upregulated gene in their study. The differences in transcriptional profiles could be explained by several factors including experimental conditions, type of array technology and intrinsic differenced between isolates used in all three studies. One major difference between our study and others is that we used an isogenic background in the reintroduction of the seven individual *CgPDR1* alleles, which prevents intrinsic strain variations. This is perhaps a reason for the difference between the number of genes regulated in at least one condition in our study (626 genes regulated by at least one GOF) and that of Tsai *et al.* [@pone.0017589-Tsai2] (45 genes regulated in at least one strain pair). This view is supported by separate results obtained with the transcriptional comparison of two related clinical strains, DSY717 and DSY2317, the latter containing the *CgPDR1* GOF L1081F. Between these two isolates, only 39 genes were regulated by at least two-fold ([File S3](#pone.0017589.s007){ref-type="supplementary-material"}), including *CgCDR1* and *PUP1*, thus suggesting that intrinsic strain variations may mask the real effect of GOF on the *C. glabrata* transcriptome. The overlap between our study and others [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai2] falls into 14 regulated genes ([Figure S4](#pone.0017589.s004){ref-type="supplementary-material"}). Besides *CgCDR1* and *PUP1*, which were found consistently upregulated in all three studies, the other genes may constitute a core set of genes regulated by *CgPDR1*. It is interesting to observe that the 14 genes are almost all found upregulated in the data provided by Vermitsky *et al.* [@pone.0017589-Vermitsky1] and Tsai *et al.* [@pone.0017589-Tsai2], while in our case, the regulation of these genes is dependent on the type of reintroduced GOF in the same genetic background. Several hypotheses will be provided below. Given that *CgPDR1* is a major regulator of azole resistance in *C. glabrata* and should act on regulated genes via PDRE binding elements in the promoters of regulated genes, the consensus for *CgPDR1* binding (TCCRYGSR) was proposed and we searched systematically for this motif in the promoter regions of the 626 genes regulated by at least one GOF in our study. Fourty six (46) genes contained the consensus. We asked whether the degree of upregulation obtained by each GOF could be associated by the presence of the consensus. Our data show that the PDRE consensus was present in seven (for SFY101) to 45% (for SFY115) of the upregulated genes in single to several copies (see [File S1](#pone.0017589.s005){ref-type="supplementary-material"}). The presence of the PDRE could be detected in the downregulated genes, however the proportion was low (between 1--4%) and usually the detected PDRE occurred in a single copy. Regulatory elements on genes dependent on individual GOF were also searched with the RSAT tool ([File S4](#pone.0017589.s008){ref-type="supplementary-material"}). The following consensus site (TCCACGGA) could be detected in the promoters of upregulated genes from the GOF L280F (SFY115) and P822L (SFY116) and D1082G (SFY103). It resembles the PDRE consensus proposed by Vermitsky *et al.* [@pone.0017589-Vermitsky1] and fits to the sequence TCCACGGA published by Tsai *et al.* [@pone.0017589-Tsai2]. In complement to these analyses, we also observed that the PDRE consensus was present in 11 out of 14 promoters of regulated genes from three different data sets ([Figure S4](#pone.0017589.s004){ref-type="supplementary-material"}) and thus highlights the relevance of this binding site for the regulation of these genes. Future studies will be needed to address the genome-wide occupancy of CgPdr1 by chromatin immuno-precipitation experiments in *C. glabrata*. One can expect that CgPdr1 will bind to some extent to the genes commonly regulated by the different studies discussed here. We showed here that GOF mutations in *CgPDR1* have differential effect on transcriptional profiles. This result was unexpected since previous results investigating the effect of GOF mutations in regulators of drug resistance in other yeast species (for example *MRR1* or *TAC1* mutations in *C. albicans*) have concluded to a convergence of transcriptional profiles with different mutations on a same regulator [@pone.0017589-Liu1], . As mentioned from data shown in [Fig. 1A](#pone-0017589-g001){ref-type="fig"}, while a pairwise comparison between two GOFs can yield good correlation between expressed genes, another example between R376W and P822L gave striking different results: here, about 55% of the regulated genes showed an inverse expression pattern. Such patterns is not unique to our study: Tsai *et al.* [@pone.0017589-Tsai2] have analysed the expression of a few genes including *CgCDR1*, *CgPDR1*, *CgSNQ2* in a set of isogenic strains into which individual GOF were re-introduced. The authors observed a GOF-dependent gene expression pattern as documented here. Presently, no clear explanations could be provided four our observations. However, taking *S. cerevisiae* homologues Pdr1 and Pdr3 as models, some hypothesis can be formulated. In *S. cerevisiae*, the expression of the ABC-transporters *PDR5*, *SNQ2*, *PDR10*, *PDR15* and *YOR1* is controlled by Pdr1p/Pdr3p. In addition, Yrr1p modulates the expression of both *SNQ2* and *YOR1*. Similarly to *PDR3*, *YRR1* is autoregulated via PDREs in its promoter [@pone.0017589-Martens1], [@pone.0017589-Zhang1]. Pdr1p and Pdr3p can act as homo- or heterodimers and can positively or negatively regulate expression of target genes, indicating that additional factors can modulate their activity [@pone.0017589-Mamnun1], . For instance, the transcriptional regulator Rdr1p, acts as a repressor of *PDR5* in a PDRE-dependent manner and heterodimers of Rdr1p/Pdr1p or Rdr1p/Pdr3p compete with Pdr1p/Pdr3p for binding to PDREs [@pone.0017589-Akache1], [@pone.0017589-Hellauer1]. Similarly, the zinc cluster protein Stb5p also acts through PDREs and forms predominantly heterodimers with Pdr1p (no interaction with Pdr3p or Yrr1p yet described). Yrr1p is only present as a homodimer [@pone.0017589-Akache2]. Pdr1 and Pdr3 can also associate to different subunits of the Mediator complex including Med15 and Med12, which is an important step into the recruitment of RNA polymerase II for target gene transcription. These two subunits are present in the C- and L-Mediator complexes, which may act as positive and negative regulator of transcription, respectively [@pone.0017589-Shahi1]. While both Pdr1 and Pdr3 can bind to Med15, Pdr3 binds in a specific manner to Med12 only in cells with mitochondrial dysfunctions [@pone.0017589-Shahi1]. With respect to CgPdr1, which combined in a single gene properties shared by Pdr1 and Pdr3, these studies suggest that CgPdr1 may interact with other DNA-binding proteins and may also associate with different subunits of the Mediator complex. The different GOF detected in CgPdr1 may alter in a positive or negative manner these interactions and thus could result in differentiated gene expression patterns as observed in our study. Future studies will be needed to verify this hypothesis. Virulence and tissue burden quantitative assays performed in this study support the idea that *CgCDR1* and *PUP1* are important for the pathogenesis of *C. glabrata* at some stage of the infection. Currently our data cannot discriminate whether or not *C. glabrata* can replicate in the tested animal models. At least, the tested strains can persist over the time course of the experimentation, which is consistent with similar experiments performed in mice [@pone.0017589-Jacobsen1]. Interestingly, enhanced virulence has been observed in other *C. glabrata* isolates where azole resistance results from mitochondrial dysfunctions independently of GOF *CgPDR1* mutations. In this case, *CgCDR1* and *PUP1* are strongly upregulated and thus may also contribute to favor *C. glabrata* in host interactions [@pone.0017589-Ferrari2]. The specific role of individual gene in fungal-host interaction remains to be solved however several reports have already identified ABC-transporters as able to contribute to selective advantages under host conditions. For example, the *Cryptococcus neoformans* ABC transporter *AFR1* was shown to interfere with lysosome acidification in macrophages to increase its survival. In particular, azole-resistant isolates showing increased *AFR1* expression were more virulent than their parental azole-susceptible isolates [@pone.0017589-Sanguinetti1], [@pone.0017589-Orsi1], [@pone.0017589-Sionov1], which highlights the relevance of the association between drug resistance and virulence observed here. Interestingly, a recent study reported that *AFR1* upregulation could be obtained by reversible chromosome duplication and thus suggests *C. neoformans* could use this mechanism to modulate its virulence [@pone.0017589-Sionov2]. In another fungal species, *Botrytis cinerea*, which is a fungus causing losses of commercially important fruits, vegetables and vineyards worldwide, ABC-transporter upregulation was associated with drug resistance due to the use of fungicides. *B. cinerea* drug resistance is spreading, thus arguing against a fitness cost due to ABC-transporter upregulation [@pone.0017589-Kretschmer1]. Regarding *PUP1*, no other homologues were found yet involved in microbial pathogenesis and therefore the exact role of the product encoded by this gene in *C. glabrata* pathogenesis remains an open question. We have attempted the overexpression of both genes in a *CgPDR1*-independent manner and animal experiments yielded results in favor of the hypothesis that *CgCDR1* and *PUP1* contribute to virulence. However, while tissue burden of mice were consistently increased when *CgCDR1* and *PUP1* were overexpressed ([Fig. 8](#pone-0017589-g008){ref-type="fig"}), virulence assays failed to discriminate in a statistical manner survival curves obtained with the overexpressing strains ([Fig. 9](#pone-0017589-g009){ref-type="fig"}). Several hypotheses could be provided explaining these results. First, it is possible that enhanced virulence needs the simultaneous overexpression of *CgCDR1* and *PUP1* to result in significant survival differences with parental strains. Second, it is also possible that, because the overexpression was carried out in a *pdr1*Δ mutant, other *CgPDR1*-dependent genes still need to be co-expressed for phenocopying the enhanced virulence of the original strain DSY565. Moreover, it is possible that the animal model used here (mouse intravenous infection) is not best suited to reveal the role of the two investigated genes. Urinary tract infection models might represent an alternative, as demonstrated by Domergue *et al.* [@pone.0017589-Domergue1]. These questions are currently being addressed in the laboratory. In conclusion, our study started from a transcriptional analysis to identify important mediators of azole resistance and virulence in *C. glabrata*. The ABC transporter *CgCDR1* contributes almost solely to azole resistance but but has other activities contributing to the enhanced virulence of azole-resistant isolates. Nevertheless, this protein could be targeted for the design of inhibitors interfering both with resistance and virulence of this yeast species. ABC-transporter inhibitors have been already described and among them some are used in animal health for parasite protection (i.e. mylbemycins) and have low toxicity profiles for mammalian cells [@pone.0017589-Dryden1]. It will be therefore interesting to test these substances in the future to decrease drug resistance and its associated virulence in *C. glabrata*. Materials and Methods {#s4} ===================== Strains and growth media {#s4a} ------------------------ *C. glabrata* strains used in this study are listed in [Table 4](#pone-0017589-t004){ref-type="table"}. Yeasts were grown in complete medium YEPD (1% Bacto peptone, Difco Laboratories, Basel, Switzerland), 0.5% Yeast extract (Difco) and 2% glucose (Fluka, Buchs, Switzerland). To prepare inocula for experimental infections, yeasts were grown in YEPD medium. When grown on solid media, 2% agar (Difco) was added. YPD agar plates containing nourseothricin (clonNAT, Werner BioAgents) at 200 mg ml^−1^ were used as a selective medium for growth of yeast transformant strains. *FLP*-mediated excision of the *SAT1* cassette was induced by growing the cells for 4 h at 30°C in YCB-BSA medium (23.4 g l^−1^ yeast carbon base and 4 g l^−1^ bovine serum albumin; pH 4.0). One hundred to 200 cells were then spread on YPD plates containing nourseothricin (15 µg ml^−1^) and grown for 48 h at 30°C to obtain nourseothricin-sensitive strains. This drug concentration can distinguish between nourseothricin-resistant and nourseothricin-sensitive cells. *Escherichia coli* DH5 was used as a host for plasmid construction and propagation. DH5α was grown in Luria-Bertani broth or on Luria-Bertani agar plates supplemented with ampicillin (0.1 mg ml^−1^) when required. ::: {#pone-0017589-t004 .table-wrap} 10.1371/journal.pone.0017589.t004 Table 4 ::: {.caption} ###### Strains used in this study. ::: {#pone-0017589-t004-4} Strain Parental strain Genotype Reference --------- -------------------- ------------------------------------- --------------------------- DSY562 Related to DSY565 Azole-susceptible clinical strain [@pone.0017589-Sanglard3] DSY565 Azole-resistant clinical strain [@pone.0017589-Sanglard3] DSY717 Related to DSY2317 Azole-susceptible clinical strain [@pone.0017589-Ferrari1] DSY2317 Azole-resistant clinical strain [@pone.0017589-Ferrari1] SFY92 DSY562 *pdr1*Δ::*SAT1-FLIP* [@pone.0017589-Ferrari1] SFY93 SFY92 *pdr1*Δ::*FRT* [@pone.0017589-Ferrari1] SFY94 DSY565 *pdr1*Δ::*SAT1-FLIP* [@pone.0017589-Ferrari1] SFY95 SFY94 *pdr1*Δ::*FRT* [@pone.0017589-Ferrari1] SFY101 SFY93 *pdr1*Δ::*PDR1^R376W^-SAT1* [@pone.0017589-Ferrari1] SFY103 SFY93 *pdr1*Δ::*PDR1^D1082G^-SAT1* [@pone.0017589-Ferrari1] SFY105 SFY93 *pdr1*Δ::*PDR1^T588A^-SAT1* [@pone.0017589-Ferrari1] SFY109 SFY93 *pdr1*Δ::*PDR1^E1083Q^-SAT1* [@pone.0017589-Ferrari1] SFY111 SFY93 *pdr1*Δ::*PDR1^Y584C^-SAT1* [@pone.0017589-Ferrari1] SFY114 SFY93 *pdr1*Δ::*PDR1-SAT1* [@pone.0017589-Ferrari1] SFY115 SFY93 *pdr1*Δ::*PDR1^L280F^-SAT1* [@pone.0017589-Ferrari1] SFY116 SFY93 *pdr1*Δ::*PDR1^P822L^-SAT1* [@pone.0017589-Ferrari1] SFY148 DSY562 *cdr1*Δ::*SAT1-FLIP* This study SFY149 DSY565 *cdr1*Δ::*SAT1-FLIP* This study SFY150 DSY562 *pup1*Δ::*SAT1-FLIP* This study SFY151 DSY565 *pup1*Δ::*SAT1-FLIP* This study SFY152 SFY148 *cdr1*Δ::*FRT* This study SFY153 SFY149 *cdr1*Δ::*FRT* This study SFY154 SFY150 *pup1*Δ::*FRT* This study SFY155 SFY151 *pup1*Δ::*FRT* This study SFY159 SFY154 *pup1*Δ::*PUP1-SAT1* This study SFY160 SFY155 *pup1*Δ::*PUP1-SAT1* This study SFY161 SFY152 *cdr1*Δ::*CDR1-SAT1* This study SFY162 SFY153 *cdr1*Δ::*CDR1-SAT1* This study SFY167 DSY562 *CDR1~p~*::\[pSF109\] This study SFY168 DSY565 *CDR1~p~*::\[pSF109\] This study SFY169 SFY152 *cdr1*Δ::*FRT, pup1*Δ::*SAT1* This study SFY170 SFY153 *cdr1*Δ::*FRT, pup1*Δ::*SAT1* This study SFY173 DSY562 *PUP1*::\[pSF113\] This study SFY174 DSY565 *PUP1*::\[pSF113\] This study SFY196 DSY562 *ScTDH3~p~-SAT1* This study SFY197 DSY565 *ScTDH3~p~-SAT1* This study SFY198 SFY93 *pdr1*Δ::*FRT, ScTDH3~p~-SAT1* This study SFY199 SFY95 *pdr1*Δ::*FRT, ScTDH3~p~-SAT1* This study SFY200 SFY93 *pdr1*Δ::*FRT, ScTDH3~p~-CDR1-SAT1* This study SFY201 SFY95 *pdr1*Δ::*FRT, ScTDH3~p~-CDR1-SAT1* This study SFY202 SFY93 *pdr1*Δ::*FRT, ScTDH3~p~-PUP1-SAT1* This study SFY203 SFY95 *pdr1*Δ::*FRT, ScTDH3~p~-PUP1-SAT1* This study ::: Drug susceptibility assays {#s4b} -------------------------- The *C. glabrata* strains were tested for azole susceptibility with the broth microdilution method described in the EUCAST document EDef 7.1 [@pone.0017589-EUCASTAFST1]. Briefly, aliquots of 1.5×10^5^ cells ml^−1^ were distributed into wells of a microtiter plate in RPMI 1640 containing 2% glucose and incubated at 35°C for 24 h. Endpoint readings were recorded with an automatic plate reader (Multiskan Ascent, Thermo) and the lowest azole concentration that reduced growth to 50% of that of the drug-free control was defined as the MIC. Construction of *C. glabrata* microarrays {#s4c} ----------------------------------------- The nucleotide sequences of the 5283 *C. glabrata* ORFs and the mitochondrial genome were downloaded from the Génolevure Consortium (<http://www.genolevures.org/>). Following the Agilent eArray Design guidelines, two separate probe sets for each ORF were designed, each consisting of 60 base oligonucleotides. The probe selection was performed using the GE Probe Design Tool. Probes were filtered following their base composition and distribution, cross-hybridization potential and melting temperature to yield two probe sets representing each 5210 nuclear and 6 mitochondrial ORFs. These probes cover more than 98% of the nuclear genome and represent 6 out of the 8 mitochondrial protein-encoding genes. For quality control and normalization purposes, 103 probes were selected randomly and spotted 20 times throughout each array in addition to standard Agilent controls including spike controls for intra- and inter-array normalizations. *C. glabrata* custom arrays were manufactured in the 8×15 k format by Agilent Technologies. cRNA synthesis, one-color labelling and *C. glabrata* arrays hybridization {#s4d} -------------------------------------------------------------------------- Sample preparation was performed on three biological triplicates. Total RNA was extracted from log phase cultures in liquid YEPD as previously described [@pone.0017589-Sanglard2]. Briefly, after centrifugation of 5 ml culture (corresponding to 10^8^ cells), the yeast cell pellet was mixed with 0.3 g of glass beads, 300 µl of RNA extraction buffer (0.1 M Tris-HCl at pH 7.5, 0.1 M LiCl, 10 mM EDTA, 0.5% SDS) and 300 µl of phenol-chloroform-isoamyl alcohol (24∶24∶1). After 1 min of vortexing in a bead beater (Fastprep-24 Instrument, MP Biomedicals Switzerland, Zürich), the aqueous phase was re-extracted with phenol-chloroform-isoamyl alcohol, and RNA was precipitated with 600 µl of ethanol at −20°C for 1 h. The RNA pellet was resuspended in 50 µl of diethyl pyrocarbonate-treated H~2~O. The integrity of the input template RNA has been determined prior to labeling/amplification, using Agilent RNA 6000 Nano LabChip kit and 2100 bioanalyzer (Agilent Technologies). Agilent\'s One-Color Quick Amp Labeling Kit (Agilent Technologies) was used to generate fluorescent cRNA according to the manufacturer\'s instructions. Briefly, 1 µg of total RNA from each sample was used to which a spike mix and T7 promoter primers were added, both of which are provided by the manufacturer. cDNA synthesis was promoted by MMLV-RT (Moloney Murine Leukemia Virus Reverse Transcriptase) in the presence of dNTPs and RNaseOUT. Next, cRNA was produced from this first reaction with T7 RNA polymerase, which simultaneously amplifies target material and incorporates cyanine 3-labeled CTP. The labelled cRNAs were purified with RNeasy Mini Kit (Qiagen) and quantified using NanoDrop ND-1000 UV-VIS Spectrophotometer. 600 ng of Cy3-labelled cRNAs were fragmented and hybridized for 17 h at 65°C to each array using the Gene Expression Hybridization Kit (Agilent Technologies) and a gasket slide with a 8 microarrays/slide format for sample hybridization to separate each sample in specific sub-arrays of the 8×15 K format. Microarrays data analysis {#s4e} ------------------------- Slides were washed and processed according to the Agilent 60-mer Oligo Microarray Processing protocol and scanned on a Agilent microarray scanner G2565BA (Agilent Technologies). Data were extracted from the images with Feature Extraction (FE) software (Agilent Technologies). FE software flags outlier features, and detects and removes spatial gradients and local backgrounds. Data were normalized using a combined rank consistency filtering with LOWESS intensity normalization. The gene expression values obtained from FE software were imported into GeneSpring 10.0.2 software (Agilent Technologies) for preprocessing and data analysis. For inter-array comparisons, a linear scaling of the data was performed using the 75th percentile signal value of all of non-control probes on the microarray to normalize one-color signal values. Probe sets with a signal intensity value below the 20th percentile were considered as absent and discarded from subsequent analysis. The expression of each gene was normalized by its median expression across all samples. Genes were included in the final data set if their expression changed by at least 2-fold between each strain expressing a *CgPDR1* GOF allele and the strain SFY114 expressing the *CgPDR1* wild type allele in at least two independent experiments. Corrected p-value (\<0.05) was chosen as the cut-off for significance. Validation of genes found regulated by microarray analysis was performed by qRT-PCR analysis (see below for technical details) on a set of nine different genes. In general, the correlation found between qRT-PCR and microarray data was excellent (see [Figure S1](#pone.0017589.s001){ref-type="supplementary-material"}). Microarray data have been uploaded to the NCBI GEO microarray repository. The GEO accession number for the *C. glabrata* Agilent array is GPL10713 and the accession numbers for the data are GSE23827, GSE23828 and GSE23829. Use of bioinformatic tools {#s4f} -------------------------- The analysis of consensus pattern on *C. glabrata* promoters (−800 to −1) was performed using the Regulatory Sequence Analysis Tools (RSAT: <http://rsat.ulb.ac.be/rsat/index.html>) and implemented to the pattern discovery tool (oligo-analysis). The settings were those supplied by default by the tool provider. The position-specific scoring matrices (PSSM) consensus matrices were converted using statistical parameters to consensus patterns and viewed via Weblogo [@pone.0017589-ThomasChollier1]. GO term enrichment analysis in the investigated genes was carried out using the Generic Gene Ontology (GO) Term Finder online tool available at <http://quantbio.princeton.edu/toolsResources.html>. Quantitative real-time RT-PCR (qRT-PCR) {#s4g} --------------------------------------- Total RNA was extracted from log phase cultures with an RNeasy Protect Mini kit (Qiagen) by a process involving mechanical disruption of the cells with glass beads and an RNase-free DNase treatment step as previously described [@pone.0017589-Sanguinetti2]. Expression of the *CgCDR1*, *CgCDR2* and *CgSNQ2* genes was quantitatively assessed with real-time RT-PCR in an i-Cycler iQ system (Bio-Rad). All primers and probes [@pone.0017589-Sanguinetti2] were designed with Beacon Designer 2 (version 2.06) software (Premier Biosoft International) and synthesized by MWG Biotech (Ebersberg, Germany). qRT-PCRwere carried out as previously described [@pone.0017589-Sanguinetti2]. Each reaction was run in triplicate on three separate occasions. For relative quantification of the target genes, each set of primer pairs and the Taqman probes were used in combination with the primers and probe specific for the *CgACT1* reference gene in separate reactions [@pone.0017589-Torelli1]. *CgPDR1* and *PUP1* expression levels were determined by real-time qRT-PCR in a StepOne Real-time PCR System (Applied Biosystems) [@pone.0017589-Ferrari1] using the Mesa Blue qPCR Mastermix Plus for Sybr assay kit (Eurogentec). Each reaction was run in triplicate on three separate occasions. *CgPDR1* and *PUP1* expression levels were normalized by *CgACT1* expression. Changes (*n*-fold) in gene expression relative to that of control isolate SFY114 were determined from *CgACT1-*normalized expression levels. The primers used for *PUP1* quantification are PUPa (5′-cactggtgcgctgaaaggtg-3′) and PUPb (5′-tgtcccaggctatctttgcc-3′). The primers used for *CgPDR1* and *CgACT1* quantification were previously described [@pone.0017589-Ferrari1]. A two-fold increase in the expression level of each gene was arbitrarily considered as significant [@pone.0017589-Torelli1]. Disruption and replacement of *CgCDR1* {#s4h} -------------------------------------- For the disruption of *CgCDR1*, the *SAT1* flipping method was employed (Reuss *et al.*, 2004). The complete *CgCDR1* ORF flanked by 500 bp was amplified by PCR from genomic DNA of DSY562 using the primers CgCDR1-ApaI (5′-gcgcaaaGGGCCCtacatgttggcaaacccagg-3′) and CgCDR1-SacII (5′-gcgcaaaCCGCGGttggacaattgaatcagccg-3′) containing *Apa*I and *Sac*II restriction sites, respectively, and inserted into pBluescript II SK(+) to yield pSF87. *CgCDR1* deletion was created by PCR using the primers CgCDR1-XhoI (5′-gcgcaaaCTCGAGtgttacttttctttactttg-3′) and CgCDR1-NotI (5′-gcgcaaaGCGGCCGCtaatttatttagcctgcgct-3′) and pSF87 as a template. The resulting PCR product was digested with *Xho*I and *Not*I and ligated to a 4.7 kb *Xho*I-*Not*I fragment containing the *SAT1* flipper cassette from pSFS1 (referred as to *FLIP*) [@pone.0017589-Reuss1] to yield pSF91. This plasmid was linearised by digestion with *Apa*I and *Sac*II and transformed into DSY562 and DSY565. After selection of transformants on nourseothricin-containing YEPD plates (200 µg/ml), the *CgCDR1* deletion strains SFY148 and SFY149, respectively, were obtained. For *CgCDR1* replacement, the *SAT1* cassette was excised in SFY148 and SFY149 to obtain the nourseothricin-sensitive strains SFY152 and 153 respectively. The 600-bp of the 3′UTR of *CgCDR1* ORF was amplified by PCR from DSY562 genomic DNA using the primers CgCDR1-NotIb (5′-gcgcaaaGCGGCCGCaaattttagacagcgctcgg-3′) and CgPDR1-SacIIb (5′-gcgcaaaCCGCGGtttgcgacaaattgggcagc-3′) and inserted into pSFS1 to yield pSF97. The complete *CgCDR1* ORF flanked by 500-bp upstream and 250-bp downstream was amplified using the primers CgCDR1-ApaI (see above) and CgCDR1-XhoIb (5′-gcgcaaaCTCGAGtatacctatgagcagatttc-3′) and inserted into pSF97 to yield pSF103. This plasmid was linearised by *Apa*I and *Sac*II and transformed into SFY152 and SFY153. After selection of transformants on, the *CgCDR1* revertant strains SFY161 and SFY162 were obtained. Disruption and replacement of *PUP1* {#s4i} ------------------------------------ For the disruption of *PUP1* (CAGL0M12947g), the complete *PUP1* ORF flanked by 500-bp was amplified using the primers PUP-KpnI (5′-gcgcaaaGGTACCcattcatacccattccgtgg-3′) and PUP-SacI (5′-gcgcaaaGAGCTCtaggattcctgaaatgctgg-3′) containing *Kpn*I and *Sac*I restriction sites, and inserted into pBluescript II SK(+) to yield pSF90. *PUP1* deletion was created by PCR using the primers PUP-ApaI (5′-gcgcaaaGGGCCCattgtaacttatgttgtctg-3′) and PUP-SacII (5′-gcgcaaaCCGCGGagtgaccatactacacatta-3′) and pSF90 as a template. The resulting PCR product was digested with *Apa*I and *Sac*II and ligated to a 4.7 kb *Apa*I-*Sac*II fragment containing the *SAT1* flipper cassette from pSFS1 [@pone.0017589-Reuss1] to yield pSF94. This plasmid was linearised by digestion with *Kpn*I and *Sac*I and transformed into DSY562 and DSY565 to obtain the *PUP1* deletion strains SFY150 and SFY151, respectively. Another *PUP1* deletion cassette was constructed to obtain strains with deletion in both *CgCDR1* and *PUP1*. As described above, pSF90 was amplified using the primers PUP-ApaI and PUP-SacII. The *SAT1* marker without the flipper system was amplified using the primers SAT1-ApaI (5′-gcaaaGGGCCCggaccacctttgattgtaaatagt-3′) and SAT1-SacII 5′-(ataagaatCCGCGGgtcaaaactagagaataataaag-3′) and pSFS1 as template. The resulting PCR products were digested with *Apa*I and *Sac*II and ligated to yield pSF101. This plasmid was transformed into the *CgCDR1* deletion strains SFY148 and SFY149 to obtain the *CgCDR1* and *PUP1* double deletion strains SFY169 and SFY170, respectively. For *PUP1* replacement, the *SAT1* cassette was excised in SFY150 and SFY151 to obtain the nourseothricin-sensitive strains SFY154 and SFY155 respectively. *PUP1* replacement cassette was created by PCR using the primers PUP-ApaIb (5′-gcgcaaaGGGCCCcgaatctattggtcgcaagg-3′) and PUP-SacIIb (5′- gcgcaaaCCGCGGgtaagtcatggagcttatgc-3′) and pSF90 as a template. The resulting PCR product was digested with *Apa*I and *Sac*II and ligated to a 4.7 kb *Apa*I-*Sac*II fragment containing the *SAT1* flipper cassette from pSFS1 [@pone.0017589-Reuss1] to yield pSF98. This plasmid was linearised by *Kpn*I and *Sac*I and transformed into SFY154 and SFY155 to obtain the *PUP1* revertant strains SFY159 and SFY160. All above-constructed strains were verified by Southern blot analysis (see [Figure S2](#pone.0017589.s002){ref-type="supplementary-material"}). Transformants were selected onto nourseothricin-containing YEPD plates. Overexpression of *CgCDR1* and *PUP1* {#s4j} ------------------------------------- For *CgCDR1* and *PUP1* overexpression, the *SAT1* marker was amplified using the primers SAT1-NotI (5′-ataagaatGCGGCCGCgtcaaaactagagaataataaag-3′) and SAT1-BamHI (5′-gcaaaGGATCCggaccacctttgattgtaaatagt-3′) and inserted into the *Not*I-*Bam*HI restriction sites of pBluescript II SK(+) to yield pSF30. This plasmid was then digested with *Xho*I and *Eco*RI and ligated to a 1.3 kb *Xho*I-*Eco*RI fragment containing the *C. glabrata CEN-ARS* from pCgACU-5 (Kitada *et al.*, 1996) to yield pSF126. The 0.7 kb *Eco*RI-*Bam*HI fragment from yEpGAP-Cherry-MCS [@pone.0017589-KepplerRoss1] containing the constitutive *S. cerevisiae TDH3* promoter, was ligated into pSF126 to yield pSF127. The complete *CgCDR1* and *PUP1* ORFs were amplified by PCR from genomic DNA of DSY562 using the primers CgCDR1-EcoRIfor (5′-actGAATTCatgtctcttgcaagtgacaag-3′) and CgCDR1-EcoRIrev (5′-ataGAATTCtatacctatgagcagatttc-3′), and PUP-EcoRIfor (5′-actGAATTCatgtcagacagcagggaaat-3′) and PUP-EcoRIrev (5′-ataGAATTCcgaatctattggtcgcaagg-3′), respectively. The resulting PCR products were digested by *Eco*RI and inserted downstream of the *TDH3* promoter of pSF127 to yield the *CgCDR1* and *PUP1* overexpressing vectors, pSF129 and pSF130, respectively. The plasmids pSF129 and pSF130 were transformed into the *PDR1* deletion strains SFY93 and SFY95 to obtain strains overexpressing *CgCDR1* (SFY200 and SFY201) or *PUP1*, (SFY202 and SFY203). As controls, plasmid pSF127 was introduced in strains DSY562, DSY565 and derivatives *pdr1*Δ mutants SFY93 and SFY95 to yield strains SFY196, SFY197, SFY198 and SFY 199, respectively. Transformants were selected onto nourseothricin-containing YEPD plates. Construction of the fusions *CgCDR1p-*3x*GFP* and *PUP1*-3x*GFP* {#s4k} ---------------------------------------------------------------- To express *GFP* under the control of the *CgCDR1* promoter, the *SAT1* marker was amplified using the primers SAT1-StuI (5′-ataagaatAGGCCTgtcaaaactagagaataataaag-3′) and SAT1-BamHI (see above) and inserted into the *Stu*I-*Bgl*II restriction sites of pBS-3xGFP--TRP1 [@pone.0017589-Lee1] containing three tandemly fused *GFP* genes (3x*GFP*) to yield pSF104. Five hundred bp of the *CgCDR1* promoter were amplified from genomic DNA of using the primers CgCDR1p-BamHI (5′-gcgcaaaGGATCCtacatgttggcaaacccagg-3′) and CgCDR1p-BclI (5′-gcgcaaaTGATCAtgttacttttctttactttg-3) containing *Bam*HI and *Bcl*I restriction sites, respectively, and inserted into the *Bam*HI site of pSF104 to yield pSF109. This plasmid was linearised by digestion with *Sph*I and transformed into DSY562 and DSY565 to obtain strains SFY167 and SFY168, respectively. To fuse the 3x*GFP* gene and the *PUP1* ORF, the complete *PUP1* ORF was amplified from DSY562 genomic DNA using the primers PUP-BglIIf (5′-gcgcaaaAGATCTatgtcagacagcagggaaat-3′) and PUP-BglIIr (5′-gcgcaaaAGATCTtgtatgatcattatcctt-3′) and inserted into the *Bam*HI site of pSF104 to yield pSF113. This plasmid was linearised by digestion with *Nco*I and transformed into DSY562 and DSY565 to obtain strains SFY173 and SFY174, respectively. Transformants were selected onto nourseothricin-containing YEPD plates. Confocal microscopy {#s4l} ------------------- To label mitochondria, log phase cultures of strain SFY174 were treated with 0.25 µM Mitotracker® Red CMXRos (Molecular Probes) for 30 min and washed with PBS. *C. glabrata* cells were fixed in 3.5% para-formaldehyde at 4°C for 5 min followed by 10 min at room temperature. Cells were then washed 3--5 min with phosphate-buffered saline (10 mM Na~2~HPO~4~, 2 mM KH~2~PO~4~, 140 mM NaCl, 3 mM KCl, pH 7.4). The remaining fixative was quenched with 100 mM Tris-HCl, pH 8.0. Fluorescence was analyzed with a confocal fluorescence microscope (Zeiss LSM 510 Meta, Jena, Germany). Flow cytometry {#s4m} -------------- Groups of four female BALB/c mice (20 to 25 g; Charles-River) were injected into their lateral vein with saline suspensions containing 4×10^7^ colony-forming units (CFU) of the *C. glabrata* strains (each in a volume of 250 µl). After seven days, mice were sacrificed by CO~2~ inhalation, and kidneys were excised aseptically and homogenized in 10 ml sterile water. Kidneys homogenates were washed twice with FACS buffer (1×PBS pH 7.4, 5% FCS, 2 mM EDTA pH 8.0) and resuspended in 2 ml FACS buffer. Remaining tissue aggregates and cell clumps were eliminated by filtration through 50-µm cell strainers. A FACSCalibur® system (BD Bioscience) and the CellQuest™ software were used for analysis. Animal studies {#s4n} -------------- Female BALB/c mice (20 to 25 g) were purchased from Harlan Italy S.r.l (San Pietro al Natisone, Udine, Italy) and inbred in-house. The mice were housed in filter-top cages with free access to food and water. To establish *C. glabrata* infection, mice were injected into their lateral vein with saline suspensions of the *C. glabrata* strains (each in a volume of 200 µl). In virulence studies, a group of ten immuno-suppressed mice was established for each yeast strain. Mice were rendered neutropenic by intraperitoneal administration of cyclophosphamide (200 mg kg^−^1 of body weight per day) three days before challenge and on the day of infection. Mice were injected with 7×10^7^ colony-forming units (CFU) of each of the investigated strains. For tissue burden experiments, immuno-competent mice were inoculated with 4×10^7^ CFU. After seven days, mice were sacrificed by use of CO~2~ inhalation, and target organs (spleen and kidney) were excised aseptically, weighted individually, and homogenized in sterile saline by using a Stomacher 80 device (Pbi International, Milan, Italy) for 120 s at high speed. Organ homogenates were diluted and plated onto YPD. Colonies were counted after two days of incubation at 30°C, and the numbers of CFU g^−1^ of organ were calculated. For survival experiments, mice were made neutropenic as previously described [@pone.0017589-Kaur2] and then injected with 7×10^7^ CFUs of each of the strains studied. Mice were monitored with twice-daily inspections and those that appeared moribund or in pain were sacrificed by use of CO~2~ inhalation. CFU counts were analyzed with non-parametric Wilcoxon Rank sum tests, while mean survival times were compared among groups by using the long-rank test. A *P*-value of less than 0.05 was considered to be significant. Ethics Statement {#s4o} ---------------- The animal experiments were performed under a protocol approved by the Institutional Animal Use and Care Committee at Università Cattolica del S. Cuore, Rome, Italy (Permit number: L21, 10/02/2008) and authorized by the Italian Ministry of Health, according to Legislative Decree 116/92, which implemented the European Directive 86/609/EEC on laboratory animal protection in Italy. Animal welfare was routinely checked by veterinarians of the Service for Animal Welfare. Animal experiments carried out for *in vivo* detection of GFP-tagged proteins (see above) were performed at the University of Lausanne and University Hospital Center under the surveillance of the local governmental veterinarian offices. These experiments were approved by the local governmental veterinarian offices and are registered under number 1734.2. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Validation of microarrays results by qRT-PCR.** **Panel A**: Gene expression relative to the strain SFY114 (containing the wild type *CgPDR1* allele) obtained by microarray analysis for each of the investigated GOF mutation in *CgPDR1*. Color code for up- and downregulated genes is given. **Panel B**: Gene expression relative to the strain SFY114 obtained by qRT-PCR. The values are averages of three separate experiments and represent increase in gene expression relative to SFY114 (set at 1.00). Primers used for *CgPDR1*, *PUP1*, *CgCDR1* and the normalization control *CgACT1* are described in the [Material and Methods](#s4){ref-type="sec"} section. Other primers used for qRT-PCR are listed below. The comparison between qRT-PCR results and microarrays was estimated by linear regression between relative expression changes. R^2^ values ranged from 0.4 and 0.89 between comparisons. Two comparisons including values obtained for CAGL0A00473g and CAGL0A00451g (*PDR1*) gave low correlation coefficients. This is explained by the fact that microarrays values of regulated genes were 10--100 fold different than observed for qRT-PCR. However, these discrepancies do not change the categorization of these genes being up- and downregulated by a given GOF mutation and taking a 2-fold change as a cut-off value. Forward and reverse primers are the following for CAGL0K00715g: 5′-TGCATCATCGAAGTCGTTGG-3′ and 5′-CCCACGAGTAACAGCACCACT-3′; for CAGL0E03894g: 5′-AAGCCGCAGACAAAGAGCAG-3′and 5′-CATCACCATTCTCGCCGTG-3′; for CAGL0A00473g: 5′-CACTGGTGCGCTGAAAGGTG-3′ and 5′-TGTCCCAGGCTATCTTTGCC-3′; for CAGL0F01111g: 5′-GTTTGGCTACTTGAGCACCGA-3′ and 5′-CGATCTCCCCTAGGCCATC-3′; for CAGL0I09724g: 5′-GCCTGAGAGCTTGGACCACT-3′ and 5′-TTGTTGGACGTGGTCTTCGA-3′; for CAGL0D02662g: 5′-CGCTGATGTTTCTGCGATGT-3′ and 5′-CACCGAATGCGATCATCAAA-3′. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Southern blot analysis and diagram illustrating strategies for disruption and replacement of** ***CgCDR1*** **and** ***PUP1*** **in** ***C. glabrata*** **isolates.** DNA was purified from isolated colonies, digested with the restriction enzyme *Pvu*II, analyzed by gel electrophoresis and hybridized to specific probes. **Panel A**: Analysis of *CgCDR1* loci. The expected sizes for *CgCDR1* analysis are: 1.7 kb for DSY562 and DSY565 (wild type *CgCDR1* locus); 6.1 kb for SFY148 and SFY149 (*cdr1*Δ::*SAT1-*FLIP); 1.3 kb for SFY152, SFY153, SFY169 and SFY170 (*cdr1*Δ::*FRT*); 1.7 kb for SFY161 and SFY162 (*cdr1Δ*::*CgCDR1-SAT1*). **Panel B**: Analysis of *PUP1* loci. The expected sizes for *PUP1* analysis are: 1.2 kb for DSY562 and DSY565 (wild type *PUP1* locus); 12.6 kb for SFY150 and SFY151 (*pup1*Δ::*SAT1-*FLIP); 7.8 kb for SFY154 and SFY155 (*pup1*Δ::*FRT*); 1.2 kb for SFY159 and SFY160 (*pup1*Δ::*PUP1*-*SAT1*); 9.7 kb for SFY169 and SFY170 (*pup1*Δ::*SAT1*). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Promoter consensus analysis of genes upregulated in SFY103 (GOF mutation D1082G) and SFY116 (GOF mutation P822L).** The data was obtained using RSAT (<http://rsat.ulb.ac.be/rsat/index.html>) and the oligo-analysis tool with default settings. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Comparisons of transcript profiling experiments of azole resistance in** ***C. glabrata*** **.** **Panel A**: Venn diagram was obtained by comparisons of published studies [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai2] with the present study and included all genes regulated by ≥2-fold. **Panel B**: List of the 14 genes commonly regulated as reported by published studies [@pone.0017589-Vermitsky1], [@pone.0017589-Tsai2] and by the present study. Color codes and abbreviations are detailed in [File S1](#pone.0017589.s005){ref-type="supplementary-material"}. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: File S1 ::: {.caption} ###### **List of genes regulated by the GOF mutations in** ***CgPDR1*** **.** (XLSX) ::: ::: {.caption} ###### Click here for additional data file. ::: File S2 ::: {.caption} ###### **List of genes regulated by** ***CgPDR1*** **in** ***C. glabrata*** **.** (XLSX) ::: ::: {.caption} ###### Click here for additional data file. ::: File S3 ::: {.caption} ###### **List of genes regulated in a pair of isolate containing an azole-susceptible (DSY717) and an azole-resistant isolate (DSY2317).** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: File S4 ::: {.caption} ###### **Putative regulatory sequences in genes regulated by GOF mutations in** ***CgPDR1*** **.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors thank Françoise Ischer for excellent technical assistance, Marilena La Sorda for performing animal experiments, Anthony Croxatto for confocal microscopy and Albert Spicher for flow cytometry analysis. The authors thank A. Coste and P. Hauser for critical reading of the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by a grant of the Swiss Research National Foundation 31003A\_127378 to DS. M.S. was supported by a grant from the Istituto di Ricovero e Cura a Carattere Scientifico (IRCCS) "Lazzaro Spallanzani" (Strategic Research Program 2006, Italy) and from Università Cattolica del S. Cuore (Linea D1, 2010). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SF DS MS. Performed the experiments: SF RT MS. Analyzed the data: DS SF BP MS. Contributed reagents/materials/analysis tools: DS SF BP MS. Wrote the paper: DS SF MS.

PubMed Central

2024-06-05T04:04:19.187140

2011-3-9

PMC3052360

Introduction {#s1} ============ *Legionella pneumophila* is an opportunistic bacterial pathogen, which is ubiquitous in the environment as a parasite of fresh water amoebae. Inhalation of *L. pneumophila* contaminated aerosols by immuno-compromised individuals can lead to an atypical acute pneumonia known as Legionnaires\' disease [@pone.0017638-Fields1]. The cell biological features associated with infections of amoeba and mammalian cells are highly similar, suggesting that amoeba serves as the "training site" for its ability to colonize higher organisms [@pone.0017638-Swanson1]. Similarly, most genetic determinants important for multiplying within amoebae cells also are essential for its growth in mammalian cells [@pone.0017638-Solomon1]. The single most important virulence factor of *L. pneumophila* is the Dot/Icm type IV secretion system [@pone.0017638-Segal1], [@pone.0017638-Vogel1]. Built with about 26 proteins, this apparatus connects the bacterial cytoplasm to the extracellular environment and functions as the conduit through which effector proteins are delivered into host cells [@pone.0017638-Nagai1]. Protein substrates of the Dot/Icm are directly involved in the construction of the niche called Legionella containing vacuole (LCV) that supports intracellular bacterial growth [@pone.0017638-Isberg1]. Elucidating the functions of these substrates will reveal not only the mechanisms used by *L. pneumophila* to subvert host cellular processes but also could potentially reveal novel host pathways undetectable or difficult to study under normal physiological conditions. Thus, tremendous efforts have been invested to identify such effector proteins, characterize their functions, and to understand their roles in *L. pneumophila* pathogenesis [@pone.0017638-Newton1]. Several methods have been developed to measure Dot/Icm-mediated protein translocation: i) immunostaining of infected cells or/and isolated LCVs with antibodies specific for candidate effector proteins [@pone.0017638-Nagai2], [@pone.0017638-Luo1]; ii) interbacterial protein transfer by detecting in recipient cells the deletion of genes by chimeras of candidate proteins and the Cre recombinase [@pone.0017638-Luo1]; iii) the restoration of a transfer-deficient mutant of the effector SidC [@pone.0017638-VanRheenen1], [@pone.0017638-Huang1]; iv) the use of the calmodulin-dependent adenylate cyclase from *Bordetella pertussis* as a reporter [@pone.0017638-Chen1], [@pone.0017638-Bardill1], [@pone.0017638-Nagai3], [@pone.0017638-Rasis1] and v), a FRET assay based on β-lactamase activity and the reporter reagent CCF4-AM [@pone.0017638-Charpentier1]. Candidate genes used in these translocation assays were obtained by a number of strategies, including bioinformatics analyses to retrieve proteins harboring structural features or functional domains typically found in proteins of eukaryotes origins [@pone.0017638-deFelipe1]; second, proteins that physically interact with components of the Dot/Icm complex or chaperones [@pone.0017638-Luo1], [@pone.0017638-Bardill1]; third, proteins capable of disrupting cellular processes of *Saccharomyces cerevisiae* [@pone.0017638-Campodonico1], [@pone.0017638-Xu1], [@pone.0017638-Shen1]; fourth, proteins whose expression appears to be regulated similarly to known substrates [@pone.0017638-Rasis1], [@pone.0017638-Zusman1], [@pone.0017638-Altman1]; fifth, computational tools to search for proteins which have one or more of the above features [@pone.0017638-Burstein1]. The combination of these gene search methods and the use of one or more translocation reporter systems have led to the identification of 204 proteins transferred by Dot/Icm. A hydrophobic residue at the -3^rd^ position [@pone.0017638-Nagai3] and the E-block [@pone.0017638-Huang1] are the two known features important for Dot/Icm-dependent translocation of subsets of substrates. Other characteristics, such as frequent occurrence of small side-chain residues at -11th to -5th residues, and a polar residue at the -16^th^ position had been found in many substrates [@pone.0017638-Kubori1]. Whether these features are important for protein translocation is unknown. Dot/Icm substrates identified from candidates that have not been prescreened will likely lead to reveal novel features recognizable by the transporter. Consistent with early observations that the LCVs are closely associated with the rough endoplasmic reticulum (ER) of host cell and that disruption of the vesicle budding from the ER repressed intracellular bacterial replication [@pone.0017638-Swanson2], [@pone.0017638-Kagan1], several bacterial proteins have been shown to target molecules that regulate protein trafficking between the ER and the Golgi apparatus. For example, RalF activates the small GTPase Arf1 on the surface of LCVs [@pone.0017638-Nagai2] whereas SidM/DrrA and LepB are guanine nucleotide exchange factor (GEF) and GTPase activation protein (GAP) for the Ras-family small GTPase Rab1, respectively [@pone.0017638-Machner1], [@pone.0017638-Ingmundson1]. In addition to vesicle trafficking, several other cellular processes are modulated by substrates of the Dot/Icm system. Mammalian cells harboring replicating *L. pneumophila* exhibits a strong resistance to exogenous cell death stimuli [@pone.0017638-Losick1], [@pone.0017638-AbuZant1], probably by synergized effects resulted from NF-κB activation and the activities of some effectors such as SdhA and SidF, which have been shown to contribute to such resistance by different mechanisms [@pone.0017638-Laguna1], [@pone.0017638-Banga1]. Two effectors, LegK1 and LnaB are able to activate NF-κB when ectopically expressed in mammalian cells, but whether such activation plays any role in modulating host cell death is unknown [@pone.0017638-Ge1], [@pone.0017638-Losick2]. Like other intravacuolar pathogens, *L. pneumophila* is able to maintain a neutral luminal pH in LCVs [@pone.0017638-Horwitz1]. Recently, the effector SidK has been shown to contribute to such regulation by antagonizing the activity of v-ATPase, the proton transfer machinery that controls organellar pH [@pone.0017638-Xu1]. The cytoplasmic face of LCVs is decorated by ubiquitinated proteins and the host proteasome function is important for intracellular bacterial replication [@pone.0017638-Dorer1]. Not surprisingly, a number of Dot/Icm substrates are involved in protein ubiquitination [@pone.0017638-Kubori1], [@pone.0017638-Price1], [@pone.0017638-Lomma1]. Finally, at least four proteins are capable of inhibiting host protein synthesis by inactivating the translation elongation factor eEF1A, which may contribute to the induction of host stress response and other unknown cellular processes [@pone.0017638-Shen1], [@pone.0017638-Belyi1], [@pone.0017638-Belyi2]. Given the large number of substrates, the diverse host functions modulated by these proteins and the possibility that translocation signals may be one of the parameters that control temporal translocation of these proteins, it is likely that previous screens have missed some substrates with undefined features. In this study we have greatly expanded the repertoire of Dot/Icm substrates by performing a comprehensive screen to test for translocation all the open reading frames larger than 300 base pairs annotated as hypothetical proteins in the genome *L. pneumophila* strain Philadelphia 1. Our efforts led to the identification of 70 novel Dot/Icm substrates. Furthermore, a protein translocated at a high efficiency does not share most of the features found in one or more groups of established substrates, indicating that the Dot/Icm system has a great flexibility in recognizing substrates. Results {#s2} ======= Construction of a library expressing β-lactamase and Legionella proteins fusions {#s2a} -------------------------------------------------------------------------------- In order to obtain a more complete list of protein substrates transferred by the *L. pneumophila* Dot/Icm system, we initiated a comprehensive screen to test Dot/Icm-dependent translocation of hypothetical proteins in strain Philadelphia 1. We chose the β-lactamase as the reporter because this system is applicable for large-scale screens while holding comparable sensitivity to other systems such as the adenylate cyclase (Cya) assay [@pone.0017638-Charpentier1]. In this method, each candidate gene is fused to TEM1 (β-lactamase) and the bacterial strain expressing the fusion protein is used to infect host cells. Host cells are then loaded with CCF4-AM which, when excited at 409 nm, emits green fluorescence (520 nm) due to fluorescence resonance energy transfer (FRET) between the coumarin and fluorescein fluorophores. Delivery of the β-lactamase fusion protein into host cells leads to cleavage of the β-lactam ring of CCF4-AM, releasing the two fluorophores and changing the fluorescence emission from green to blue (447 nm) when excited at the same wavelength. Translocation detected by this reporter can be easily quantitated by the percentage of infected cells emitting blue fluorescence signals [@pone.0017638-Charpentier1]. To construct the fusion library, we first retrieved all the open reading frames larger than 300 base pairs annotated as hypothetical genes from the *L. pneumophila* Philadelphia 1 genome ([Table S1](#pone.0017638.s002){ref-type="supplementary-material"}). After eliminating genes that have been reported as substrates of the Dot/Icm system at the time the project was initiated, a total of 833 candidate genes were obtained ([Table S1](#pone.0017638.s002){ref-type="supplementary-material"}). To clone the genes, we designed primer pairs to amplify each open reading frame by PCR and inserted them individually into pXDC61M to generate translational fusions with the upstream β-lactamase gene ([Fig. 1A](#pone-0017638-g001){ref-type="fig"}). A total of 798 plasmids expressing the β-lactamase fusion were constructed. To examine the quality of the library, we randomly chose 30 *L. pneumophila* strains harboring the plasmid to detect the expression of the fusion proteins. Although the level of expression varies, a protein corresponding to the expected sizes of the chimeras was detected in most of the samples ([Fig. 1B](#pone-0017638-g001){ref-type="fig"}), indicating that we have successfully constructed a library expressing β-lactamase fusions in *L. pneumophila*. ::: {#pone-0017638-g001 .fig} 10.1371/journal.pone.0017638.g001 Figure 1 ::: {.caption} ###### Construction of a library expressing fusions of β-lactamase and *L. pneumophila* hypothetical proteins. **A.** A schematic structure of the fusion proteins and the screening strategy. In most cases, the gene was fused with the β-lactamase by inserting into the vector as a *Bam*HI/*Sal*I fragment. After infection, samples were loaded with the CCF4-AM dye and were inspected under a fluorescence microscope. **B.** Evaluation of the library for expression of the fusion proteins. Plasmids directing expression of β-lactamase fusions were introduced into a wild type *L. pneumophila* strain. Total cell lysates of bacteria grown in the presence of IPTG were used to examine the steady state of the fusion proteins by immunoblot. In each image, the detection of a protein nonspecifically recognized by the antibody (arrow) was used as a loading control. :::  ::: Identification of proteins transferred by the Dot/Icm system {#s2b} ------------------------------------------------------------ After verifying the expression of the fusion proteins in *L. pneumophila*, we infected U937 macrophages with *L. pneumophila* strains expressing β-lactamase fusions grown to post exponential phase. One hour after infection, cells were loaded with CCF4-AM dye and incubated for an additional 2 hours at room temperature. Translocation of the β-lactamase chimera was assessed by the presence of cells emitting blue fluorescence signals. A group of 24 bacterial strains expressing β-lactamase were used to infected host cells for each screen. In each experiment, we used the TEM-RalF and TEM1-FabI hybrid proteins as the positive and negative controls, respectively [@pone.0017638-Charpentier1]. Samples were visually inspected under a fluorescence microscope and strains that gave more than 5% of blue cells were retained for further study. Under this experimental condition with a multiplicity of infection (MOI) of 20, 95% of the cells infected by wild type *L. pneumophila* strain expressing the TEM-RalF fusion emits blue fluorescence, whereas similar infections with the strain expressing TEM- FabI results in no blue cells (data not shown), which are consistent with results from an earlier study [@pone.0017638-Charpentier1]. Infections were repeated at least twice for strains that gave positive translocation results. Constructs harboring genes exhibiting detectable transfer were introduced into the *dot/icm*-deficient strain Lp03 and the resultant strains were similarly tested for delivery the β-lactamase fusions into host cells. None of these fusions caused detectable translocation in this *dot/icm*-deficient strain (data not shown). A total of 164 proteins that consistently promote TEM translational fusions in a Dot/Icm-dependent manner were obtained ([Tables S2](#pone.0017638.s003){ref-type="supplementary-material"} and [S3](#pone.0017638.s004){ref-type="supplementary-material"}). Among these, 94 proteins had been reported as Dot/Icm substrates ([Table S3](#pone.0017638.s004){ref-type="supplementary-material"}), further validating the reliability of our screen strategy. Thus, our efforts have added 70 proteins to the inventory of the substrate pool of the *L. pneumophila* Dot/Icm transporter ([Table S2](#pone.0017638.s003){ref-type="supplementary-material"}). The transfer efficiencies of these proteins vary greatly, ranging from 5% to 95% ([Fig. 2](#pone-0017638-g002){ref-type="fig"}). Among these, 5 proteins exhibited transfer efficiencies comparable to that of RalF, causing more that 90% of infected cells to emit blue fluorescence signals ([Fig. 2](#pone-0017638-g002){ref-type="fig"}). These 5 proteins do not share any detectable common features. Instead, the primary sequence of the C-terminal portion of Lpg2844 is quite different from all known substrates (see below). Twenty-five proteins converted 50%--80% of the green cells into blue cells, 13 proteins exhibited translocation efficiencies between 20 to 45% and 27 proteins showed low transfer efficiencies with less than 20% blue cells in the samples ([Fig. 2](#pone-0017638-g002){ref-type="fig"} and [Table S2](#pone.0017638.s003){ref-type="supplementary-material"}). ::: {#pone-0017638-g002 .fig} 10.1371/journal.pone.0017638.g002 Figure 2 ::: {.caption} ###### Dot/Icm-dependent translocation of substrates. Effectors identified in this study were divided into four groups according to their transfer efficiencies, 5 genes from each group were shown as representatives. U937 cells seeded in 96-well plates were infected with wild type or *dot/icm*-deficient *L. pneumophila* strains expressing a gene fusion and images were acquired 2 hrs after CCF4-AM loading with a DP72 color camera (Olympus). Group I, genes with translocation efficiency \>90%; Group II, genes with translocation efficiency between 50% and 80%; Group III, genes with translocation efficiency between 20% and 45% and Group IV, genes with translocation efficiency less than 15%. :::  ::: Although all candidates were annotated as hypothetical proteins in the genome database of the Philadelphia 1 strain, careful bioinformatic analysis revealed that a small fraction of them harbor motifs of known or putative functions ([Table S2](#pone.0017638.s003){ref-type="supplementary-material"}). Further, most of these proteins have a homolog in the genome of the Paris, Lens or the Corby strain; the number of proteins without a detectable homolog are 7, 12 and 13 for these three strains, respectively ([Table S2](#pone.0017638.s003){ref-type="supplementary-material"}). Two proteins, Lpg1083 and Lpg1684 are specific for the Philadelphia 1 strain ([Table S2](#pone.0017638.s003){ref-type="supplementary-material"}). In some cases, two or more genes were clustered in a specific region in the chromosome ([Table S2](#pone.0017638.s003){ref-type="supplementary-material"}), a common phenomenon in gene organization of *L. pneumophila* type IV substrates [@pone.0017638-Luo1] and often accounts for the remarkable plasticity of genomes of this organism [@pone.0017638-Cazalet1], [@pone.0017638-Ninio1]. To examine whether the difference in transfer efficiency among proteins was due to the stability of the fusion proteins, we examined the levels of the fusion proteins in several strains from each group. In general, there was no clear correlation between the steady state levels of the hybrids and translocation efficiencies. For example, the steady state levels of the TEM-Lpg0021 and TEM-Lpg0181 were among the highest in these strains, but their transfer efficiencies were not the highest ([Fig. 3](#pone-0017638-g003){ref-type="fig"}). On the other hand, the poorly expressed TEM-Lpg2555 and TEM-Lpg2874 were translocated at high efficiency ([Fig. 3](#pone-0017638-g003){ref-type="fig"}). To further determine to what extent the lack of detectable translocation was a result of a failure to express the protein fusions, we examined the β-lactamase fusions in 27 transfer deficient strains. Although the protein levels vary, all of these strains produced readily detectable fusion proteins ([Fig. S1](#pone.0017638.s001){ref-type="supplementary-material"}). Therefore, the level of fusion protein expressed in *L. pneumophila* is not the sole factor determining the translocation competency of a particular substrate. ::: {#pone-0017638-g003 .fig} 10.1371/journal.pone.0017638.g003 Figure 3 ::: {.caption} ###### Translocation efficiency does not correlate with levels of β-lactamase fusion expressed in *L. pneumophila*. **A.** The translocation efficiency of 9 substrates in the β-lactamase assay. After CCF4-AM loading, macrophages infected with bacterial strains expression fusions between β-lactamase and individual genes were inspected under a fluorescence microscope, translocation efficiencies were obtained by enumerating cells emitting blue and green fluorescence signals, respectively. Experiments were performed in triplicates and at least 300 cells were counted each sample. Similar results were obtained in at least two independent experiments. **B.** The levels of the fusion proteins in *L. pneumophila* strains used for infections shown in A. Bacterial cells equivalent to one OD~600~ unit were lysed in 200 µl of SDS loading buffer, 15 µl of boiled supernatant were resolved by SDS-PAGE. After transferring to nitrocellulose membranes, the fusion protein was detected with a β-lactamase specific antibody by immunoblot. The isocitrate dehydrogenase (ICDH) was probed as a loading control. Samples: 1. Lpg1776; 2, Lpg0021; 3, Lpg2425; 4, Lpg1147; 5, Lpg0181; 6, Lpg2555; 7, Lpg2874; 8, Lpg0405; 9, Lpg0195. :::  ::: Recognition of diverse translocation signals by the Dot/Icm transporter {#s2c} ----------------------------------------------------------------------- In our efforts to identify common features in the last 100 amino acids of the substrates that may be important for Dot/Icm-dependent protein translocation, we found that a few proteins that have amino acid composition greatly different from the known features. One such example is Lpg2844, a 361 aa protein in which more than 1/3 of the residues are serine. Interestingly, the last 100-aa region of this protein contains few of the features known to be important for Dot/Icm-dependent translocation. A hydrophobic residue (methionine) at the -3^rd^ position [@pone.0017638-Nagai3] is the only recognizable characteristics associated with translocation found on this region of this protein ([Fig. 4A](#pone-0017638-g004){ref-type="fig"}). Full-length Lpg2844 promoted the translocation of lactamase with 85% efficiency ([Table S2](#pone.0017638.s003){ref-type="supplementary-material"} and [Fig. 4 B&D](#pone-0017638-g004){ref-type="fig"}). Importantly, a chimera containing β-lactamase fused to the last 100 amino acids of Lpg2844 promoted translocation at efficiencies only marginally lower than those of full-length proteins, indicating that like other Dot/Icm substrates, signals important for translocation localized to the C-terminal portion of this protein ([Fig. 4B--D](#pone-0017638-g004){ref-type="fig"}). As expected, a fusion that contained Lpg2844 lacking the last 100 amino acids failed to promote translocation at a detectable level ([Fig. 4 B&D](#pone-0017638-g004){ref-type="fig"}). These data indicate that the Dot/Icm transporter is capable of recognizing diverse features in the C-terminal portions of its substrates and that in some cases such differences do not affect translocation efficiencies of the proteins. ::: {#pone-0017638-g004 .fig} 10.1371/journal.pone.0017638.g004 Figure 4 ::: {.caption} ###### Diverse features presented in the C-terminal end of Dot/Icm substrates. **A.** Alignment of the last 50 amino acids of three well-established effectors and the new substrate Lpg2844 to highlight the features important for translocation found in Dot/Icm substrates, including: i) The hydrophobic residue at the -3^rd^ position (▽, red) [@pone.0017638-Nagai3] and ii) the E-block [@pone.0017638-Huang1] (the three blue residues in SidF). Note the different amino acids composition in Lpg2844. **B--D.** A region containing the last 100 amino acids of Lpg2844 is important and sufficient for promoting translocation. Bacterial strains expressing fusions of β-lactamase to full-length Lpg2844 (I), it\'s last 100 aa (II) or a fragment lacking the last 100 aa (III) were used to infect macrophages and infected cells were loaded with the CCF4-AM dye. Translocation efficiency (B) was obtained as described in [Fig. 3](#pone-0017638-g003){ref-type="fig"}, data shown are the average of three independent experiments done in triplicates; stable expression of the fusions by *L. pneumophila*, equal amount of protein samples resolved by SDS-PAGE was probed for the fusions with a β-lactamase specific antibody. The ca. 60-kDa non-specific band detected by the antibody was used as a loading control (arrow in panel C). Representative images of infected cells loaded with CCF4-AM (D). Similar results were obtained in at least two independent experiments. :::  ::: Discussion {#s3} ========== Analyses of the functions of proteins delivered into host cells by specialized transporters have provided great insights into the biology of both the pathogens and their hosts. Earlier studies that combined appropriate reporter systems and genetic screens and/or bioinformatics analyses have identified a large of number of protein substrates translocated by the *L. pneumophila* Dot/Icm system. Although effective, these strategies do not allow truly unbiased screens. In this study, we used a comprehensive method to test Dot/Icm-dependent translocation of hypothetical proteins larger than 100 amino acids in the genome of *L. pneumophila* strain Philadelphia 1. Our efforts had added 70 novel proteins to the Dot/Icm substrate inventory, which with previously proteins, have expanded experimentally confirmed Dot/Icm substrates to 275 ([Table S4](#pone.0017638.s005){ref-type="supplementary-material"}). Thus, judging by the number of proteins transferred, the *L. pneumophila* Dot/Icm system is arguably the most prolific bacterial translocator, whose substrates are more than five times of the effectors secreted by the Hrp type III secretion system of *Pseudomonas syringae*, a plant pathogen known to have a large repertoire of effectors [@pone.0017638-Lindeberg1]. Many Dot/Icm substrates identified by an earlier genetic method are larger than 100 kDa [@pone.0017638-Luo1]. With a more complete list of its substrates, we analyzed the size distribution of these proteins. Surprisingly, only 39 genes are larger than 2 kbp of which 12 are more than 3 kbp ([Fig. 5](#pone-0017638-g005){ref-type="fig"}). The majority of the genes are shorter than 2 kbp, with 113 sizing between 1 to 2 kbp and 91 shorter than 1 kbp ([Fig. 5](#pone-0017638-g005){ref-type="fig"}). Thus, the length of most of the substrate genes is about 1 kbp, typical for proteins of bacterial origins [@pone.0017638-Smith1]. That many substrates genes identified in an earlier genetic screen are longer than 2 kbps [@pone.0017638-Luo1] may be a result of high probability of generating in-frame fusions of longer genes in the random library used for bacterial two-hybrid screenings. ::: {#pone-0017638-g005 .fig} 10.1371/journal.pone.0017638.g005 Figure 5 ::: {.caption} ###### The distribution of all verified Dot/Icm substrates according to their length. Proteins experimentally shown as substrates of the Dot/Icm transporter were collected and sorted according to the length of the gene, which were then divided into four groups: I, genes larger than 3 kbp; II, genes ranging between 2 to 3 kbp; III, genes ranging between 1 to 2 kbp and IV, genes smaller than 1 kbp. :::  ::: Type IV transporter-mediated protein translocation is determined by at least three factors: Components of the transporter involved in recognizing the signals, signals encoded by amino acids embedded in the carboxyl portion of the substrates and chaperones responsible for proper folding of the substrate. In *L. pneumophila*, DotF, an important component of the Dot/Icm system interacts with a set of substrates [@pone.0017638-Luo1], but whether this protein is involved in recognizing the substrates by interacting with patches of amino acids important for translocation is unknown. The roles played by the chaperones IcmS/IcmW in substrate translocation probably is by inducing conformational changes [@pone.0017638-Cambronne1], which presumably will allow the translocation signals more properly exposed to the transporter. Accumulating evidence suggests that signals residing in the carboxyl end of Dot/Icm substrates are quite diverse. For examples, a hydrophobic residue located at -3 or -4 position is present in many Dot/Icm substrates and has been shown to be important for the translocation of RalF [@pone.0017638-Nagai3]. More recently, Huang et al showed that an E-block motif is important for the translocation of a subset of substrates [@pone.0017638-Huang1]. Further, our results have revealed Lpg2844, a protein with a carboxyl amino acid composition highly different from most substrates can be translocated at high efficiencies ([Fig. 4](#pone-0017638-g004){ref-type="fig"}). Thus, it appears that in *L. pneumophila,* successful translocation does not need the presence of all known signals. Residues important for protein translocation can contribute to substrate recognition by forming structural entities that directly bind to the transporter or by indirectly involved in this process by supporting the formation of such structures. Further experiments aiming at the interactions between transporter components and the substrates are required to assign specific roles to these amino acids in translocation. Clearly, the highly diverse translocation signals would allow a transporter to accommodate structurally different substrates, which is in great agreement with the large substrate pool possessed by the Dot/Icm system. Moreover, because the Dot/Icm system is functioning in actively replicating intracellular bacteria [@pone.0017638-Liu1], translocation signals can determine the amount of a protein delivered into host cells, thus contributing to temporal control of effector activity during infection. Although comprehensive, our direct screen method was limited by a number of factors and clearly was not exhaustive: First, we only examined proteins larger than 100 amino acids, yet at least one substrate, Lpg0045 (70 aa) is smaller than this threshold (Kubori et al., 2008). Second, our library may have excluded candidates annotated as proteins with detectable homology to proteins of known activity. Third, the β-lactamase fusion reporter may not work for some substrates. Fourth, some substrates may be translocated at efficiencies beyond the detection sensitivity of the β-lactamase assay. Finally, some of the β-lactamase fusions in our library may not be expressed at levels sufficient for detectable translocation. However, the number of false negative genes resulting from this reason, if any, should be low because high transfer efficiencies do not necessarily correlate with high protein levels ([Fig. 3](#pone-0017638-g003){ref-type="fig"}). Further, the β-lactamase fusions of 27 examined transfer-deficient genes all expressed at readily detectable levels ([Fig. S1](#pone.0017638.s001){ref-type="supplementary-material"}). In support of these potential limitations, 8 proteins capable of promoting the translocation of SidCΔC100 are not detectably positive in our assay ([Table S5](#pone.0017638.s006){ref-type="supplementary-material"} and Ref. [@pone.0017638-Huang1]). With the exception of lpg0926, the β-lactamase fusions of the other seven genes are stably expressed in *L. pneumophila* (data not shown). We further examined the translocation of this protein using the Cya fusion method; no translocation was detected despite the fusion was stably made in *L. pneumophila* (data not shown). The reasons for these discrepancies can be multiple. It is possible that some translocation signals still present in SidCΔC100, which would aid the transfer of proteins harboring low efficient transferring signals. Clearly, these potential limitations could be applied to each of the available reporter systems used to measure protein translocation. Nevertheless, that we have re-identified 94 of the substrates reported in the last several years indicates the reliability of the β-lactamase reporter system. Taken together, these observations suggest that the repertoire of the Dot/Icm substrate will likely continue to expand as more saturated screens are performed and/or more sensitive reporter systems are developed to detect translocation. This extremely large substrate repertoire may explain the observation that deletion mutants lacking one or more of these genes rarely exhibit significant intracellular growth defect, probably due to functional redundancy exerted by subsets of these proteins [@pone.0017638-Luo1]. Or, the accumulation of these effectors may be a result of the challenges faced by the bacterium to colonize phylogenetically diverse amoebae host in its natural niches. Alternatively, these proteins may not necessarily contribute to intracellular bacterial growth, the most common phenotype examined in the study of *L. pneumophila* pathogenesis. Instead, they may help the host adapt to challenges such as those brought by detrimental environmental changes and the fluctuations in nutrient supplies. Regardless of the reason, future studies directed to the elucidation of the biochemical and cell biological activities of these substrates will undoubtedly contribute greatly to our understanding the biology of both the pathogen and its hosts. Materials and Methods {#s4} ===================== Bacterial strains, cell culture and media {#s4a} ----------------------------------------- The *L. pneumophila* strains used are derivatives of the strain Lp02 (*thyA Δ(hsdR-lvh) rpsL*) [@pone.0017638-Berger1], and were grown on charcoal-yeast extract (CYE) solid medium or ACES-yeast extract (AYE) [@pone.0017638-Berger1]; Lp03 contains the *dotA3* mutation [@pone.0017638-Berger1]. The growth media were supplemented with thymidine at 100 µg/ml when appropriate. For infection experiments, the *L. pneumophila* strains used in all assays were grown to post-exponential phase (OD~600~≈3.4--3.7) unless stated otherwise. Antibiotics were used at the following concentrations with *E. coli* strains: ampicillin 100 µg/ml and chloramphenicol, 30 µg/ml. For *L. pneumophila* strains, chloramphenicol was used at 5 µg/ml. U937 cells were cultured in RPMI medium supplemented with 10% FBS prior to being induced by phorbol myristate acetate (PMA) (0.1 µg/ml). For assays in 96-well microtiter plates, differentiated U937 cells were plated in 96-well having optically clear bottoms in a density of 1×10^5^/well. Antibody and Western blotting {#s4b} ----------------------------- The antibody against β-lactamase was purchased from Abcam (Cambridge, MA). The antibody specific for the isocitrate dehydrogenase (ICDH) of *L. pneumophila* was described elsewhere [@pone.0017638-Xu1]. To detect the expression of the fusion proteins in *L. pneumophila*, strains harboring the plasmids were grown in CYE medium supplemented with thymidine (200 mg/ml), chloramphenicol 5 mg/ml and IPTG (0.5 mM) to post exponential phase (OD~600~ = 3.4--3.7). Cells equal to 1 OD unit were withdrawn and dissovled in 200 ml of SDS loading buffer. After boiling for 5 min, cleared supernatant was resolved by SDS-PAGE. Separated proteins were transferred onto nitrocellulose membranes and proteins were detected by Western blot using an appropriate IRDye infrared secondary antibody (Li-Cor\'s Biosciences Lincoln, Nebraska, USA) and the signals were detected with an Odyssey infrared imaging system as described [@pone.0017638-Xu1]. Construction of β-lactamase fusion library {#s4c} ------------------------------------------ To accommodate the fusion of genes to the β-lactamase gene as *Bam*HI/*Sal*I DNA fragments in the first open reading frame, we inserted a DNA fragment obtained from annealing oligos 5′-CGGATCCCTGCAGGCGGCCGCGTCGACT-3′ and 5′-CATGGCCTAGGGACGTCCGCCGGCGCAGCTGAGATC-3′ into *Kpn*I and *Xba*I digested pDXC61[@pone.0017638-Charpentier1] to give pDXC61M. To make the fusion library, open reading frames larger than 300 base pairs that code for hypothetical proteins were retrieved from the genome of *L. pneumophila* strain Philadelphia 1. 19-base primers were designed to amplify the entire gene by PCR with the Pfu UltraII high fidelity DNA polymerase (Agilent, Santa Clara, CA). In each case, DNA sequences recognized by the restriction enzymes *Bam*HI and *Sal*I were added to the 5′ and 3′ primers, respectively ([Table S1](#pone.0017638.s002){ref-type="supplementary-material"}). For genes whose sequences contain the *Bam*HI or *Sal*I recognition site, sequences for *Bgl*II or *Xho*I were added. For a number of genes that harbor one or more of these sites, other restriction enzymes were used ([Table S1](#pone.0017638.s002){ref-type="supplementary-material"}). After digestion with the appropriate restriction enzymes the PCR products were inserted into similarly digested pDXC61M. Plasmids containing correct inserts were introduced into *L. pneumophila* strains by electroporation. To make β-lactamase fusions with specific regions of genes, the target regions were amplified by PCR with the appropriate primers ([Table S6](#pone.0017638.s007){ref-type="supplementary-material"}) and were inserted into pDXC61M as described above. Screen for fusions that transfer the β-lactamase into mammalian cells {#s4d} --------------------------------------------------------------------- To test Dot/Icm-dependent transfer of the fusion proteins into host cells, *L. pneumophila* strains expressed the fusions grown to post exponential phase in the presence of 0.5 mM IPTG were used to infect monolayers of U937 cells seeded in 96-well plates at an MOI of 20. One hour after infection, the CCF4-AM substrates (Invitrogen, Carlsbad, CA) were mixed with medium in the wells. After further incubation for 2 hours at 25°C, infected cells were visually inspected under a Nikon IX-80 fluorescence microscope equipped with a β-lactamase FL-Cube (U-N41031, Chroma Technology Corp, Bellows Falls, VT). Images of infected cells were obtained by a DP-72 color fluorescence camera (Olympus). The percentage of infected cells was determined by counting the number of cells emiting blue fluorescence in specified areas of the wells. Experiments were performed in triplicate and in each sample and at least 300 cells were counted. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Expression of β-lactamase fusions in representative strains exhibiting undetectable protein translocation. *L. pneumophila* strains harboring fusions of β-lactamase to the indicated genes were grown in liquid medium containing 0.5 mM of IPTG to post-exponential phase (OD~600~ = 3.5--4.2). Cells corresponding to one OD unit were withdrawn and solubilized with 200 µl of SDS sample buffer. Ten µl of cleared supernatant were resolved in SDS-PAGE, proteins transferred onto nitrocellulose membranes were probed with indicated antibodies and detected with an Odyssey image system. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Primers used for the construction of the β-lactamase fusion library. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Characteristics of Dot/Icm substrates identified in this study. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### Experimentally confirmed substrates re-identified in this study. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### Experimentally confirmed protein substrates of the Dot/Icm transporter. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### Expression of proteins fusions for candidates positive for translocation in the SidCΔC100 assay but negative in the β-lactamase reporter assay. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S6 ::: {.caption} ###### Primers for construction of deletion mutants of Lpg2844. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Dr. Howard Shuman (University of Chicago, Chicago, IL, USA) for plasmids and members of our laboratory for helpful discussion. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by NIH-NIAID grants R01AI069344, K02AI085403 and R21AI092043 (Z.-Q.L). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: ZQL WZ. Performed the experiments: WZ SB YT CZ RS JG. Analyzed the data: ZQL WZ CZ. Contributed reagents/materials/analysis tools: CZ. Wrote the paper: WZ ZQL.

PubMed Central

2024-06-05T04:04:19.193758

2011-3-9

PMC3052361

Introduction {#s1} ============ 'Cognitive control' refers to the coordination and direction of lower-level cognitive processes critical to complex, goal-directed behaviour. These processes, including attentional selection, conflict resolution, and the online maintenance of goal-relevant information (and inhibition of goal-irrelevant information), may underlie higher many cognitive functions, permitting the flexibility and sophistication of human thought and behaviour across a wide variety of task situations. While cognition has traditionally been conceptualized as separate from affect, it has been increasingly recognized that affective significance is a major factor in goal-directed behaviour, both in establishing goals and in shaping how information is processed during goal pursuit. Emotionally salient stimuli in the environment may be prioritized for processing over non-emotional stimuli [@pone.0017635-Eastwood1], [@pone.0017635-Ohman1], but it remains unclear whether qualitatively distinct neural circuitry is engaged for the processing of affectively-valenced stimulus dimensions. The present study examines the neural systems engaged in the detection and management of conflict, a canonical control function, when the information being processed (i.e., the source of conflict) is emotional versus non-emotional in nature. *Conflict* can be defined mechanistically in terms of cross-talk caused by the simultaneous concurrent processing of goal-relevant and goal-irrelevant information competing for common resources [@pone.0017635-Botvinick1]. The Stroop task [@pone.0017635-Stroop1] is a classic conflict task: participants must name the colour of presented words while ignoring the word\'s meaning. In some trials, goal-irrelevant information is congruent with goal-relevant information (e.g., the word 'RED' printed in red ink); in other trials, the goal-relevant and irrelevant information are incongruent (e.g., 'RED' printed in green ink), leading to conflict. Using tasks such as the Stroop (as well as related incompatibility paradigms such as Simon, flanker and others, e.g., [@pone.0017635-Kornblum1]), conflict has been extensively studied in the cognitive realm. Functional neuroimaging methods have been used to identify a number of frontal and parietal brain regions canonically associated with cognitive control in this and other tasks [@pone.0017635-Braver1], [@pone.0017635-Cole1], [@pone.0017635-Owen1], [@pone.0017635-Wager1], with the ACC in particular being associated with conflict processing functions [@pone.0017635-Botvinick2], [@pone.0017635-Kerns1], [@pone.0017635-Carter1], [@pone.0017635-Carter2]. Evidence from early neuroimaging studies examining conflict elicited by emotional versus non-emotional distracters resulted in an influential hypothesis postulating that emotional and cognitive conflict detection are mediated by distinct rostral and dorsal subdivisions of the ACC, respectively (Bush et al., 2000). However, subsequent investigations of emotional and cognitive conflict processing have yielded mixed evidence regarding the domain-specificity of their underlying neural systems. Most studies focusing on emotion conflict using Stroop-like variants tend to find activation in dorsal rather than ventral ACC, as well as other areas associated with cognitive control, such as the lateral PFC [@pone.0017635-Buhle1], [@pone.0017635-Haas1]. In a recent study comparing activity in closely matched emotion and non-emotional variants of a face-word Stroop paradigm, Egner and colleagues [@pone.0017635-Egner1] again reported that conflict detection was associated with dorsal ACC in both conditions; activation was observed in the rostral ACC (and amygdala) only during conditions examining emotional conflict resolution (i.e., modulation based on previous trial conflict). Ochsner and colleagues [@pone.0017635-Ochsner1] compared an emotional versus non-emotional flanker task, and also found a number of areas commonly engaged by conflict in both tasks, including the dorsal ACC. However, consistent with the cognitive/emotion division hypothesis, they also observed that affective conflict selectively engaged the rostral medial PFC, with brain-behavior correlations observed in rostral ACC. Likewise, another recent study [@pone.0017635-Lee1] reported distinct patterns of conflict-related neural activity in conditions involving emotional stimulus-response (S-R) incompatibility (emotion expression interference; elicited via making facial expressions incongruent with those of presented faces) with cognitive S-R incompatibility (elicited via the Simon task). A major challenge in this research area has been to utilize appropriate paradigms that enable valid and closely matched comparisons of emotional and cognitive forms of conflict. The hypothesis that emotional versus cognitive conflict may depend on distinct subdivisions of the ACC was based on evidence from emotional adaptations of the Stroop task, which examine interference from emotional distracters (e.g., performance of the colour-naming task for emotional relative to non-emotional words; [@pone.0017635-Mathews1]). However, it has been asserted that interference in the colour-naming emotional Stroop task may occur because of lower-level lexical effects [@pone.0017635-Larsen1] or general attention capture [@pone.0017635-Algom1] rather than the direct conflict effects present in the traditional Stroop. To improve upon this design, face-word Stroop variants have been utilized, in which positive and negatively valenced words (e.g., 'HAPPY' or 'FEAR') are superimposed on compatible or compatible facial expressions [@pone.0017635-Haas1], [@pone.0017635-Egner1], [@pone.0017635-Etkin1], [@pone.0017635-Krug1]. This design improves on the colour-naming emotional Stroop in that the responses require affective classification and the task-relevant and irrelevant information are semantically related, leading to affective incompatibility effects more closely related to the direct conflict present in the traditional cognitive Stroop. However, all of these tasks involve an incompatibility between a task-relevant stimulus and a task-irrelevant stimulus (thus, stimulus-stimulus \[S-S\] incompatibility). In contrast, studies of cognitive conflict have explored both S-S and S-R incompatibilities [@pone.0017635-Kornblum1]. The emotion expression interference paradigm developed by Lee and colleagues [@pone.0017635-Lee1] is a first step in exploring S-R incompatibility in the context of emotional conflict: this paradigm examines interference when participants make emotional facial expressions as a behavioral response, capitalizing on their role as an index of emotional experience and expression [@pone.0017635-Dimberg1]. However, the Lee et al paradigm requires participants to make an expression in response to a presented face. As such, it leaves open the possibility that interference effects in the task may be caused by overriding imitation tendencies instead of being due to conflicting emotional influences, per se. In view of these considerations, our goal was to examine emotional conflict with a paradigm that similarly capitalized on emotional facial expressions to index stimulus-response incompatibility, but that improved upon this paradigm by avoiding possible imitative influences. Accordingly, we developed a new paradigm to examine emotional conflict via S-R incompatibility using emotional facial expressions to emotional, but non-face stimuli [@pone.0017635-Chiew1]. This task was adapted from the *AX Continuous Performance Task* (AX-CPT), which has been repeatedly established as a robust probe of context processing, cognitive conflict, and cognitive control [@pone.0017635-Carter2], [@pone.0017635-Cohen1], [@pone.0017635-Braver2], [@pone.0017635-Braver3]. The emotional AX-CPT requires participants to respond to emotionally evocative cue-probe combinations with emotionally congruent or incongruent facial expressions. This task was developed on the rationale that interference elicited by a mismatch between evoked emotion and required facial response may more closely approximate situations of emotional conflict that people experience in 'real-life' (e.g., acting pleasant to a rude customer; smiling graciously after a defeat), thus achieving a higher level of ecological validity. In prior work using facial electromyography (EMG) to index expression responses in this task, we demonstrated that behavioural interference can be robustly elicited, and furthermore, that such interference was greater when emotional influences were present relative to when they were absent [@pone.0017635-Chiew1]. In the AX-CPT, conflict and cognitive control are varied on a trial-by-trial basis through the use of contextual pre-cues. Certain cue-probe combinations require a target response (e.g., 'A' followed by 'X'), whereas all other cue-probe combinations require a non-target response. The target ('AX') combination occurs with high frequency, which leads to high levels of interference in two low-frequency cue-probe combinations: AY (target cue, non-target probe) and BX (non-target cue, target probe). In AY trials, interference arises from expectancy established by the target cue, while in BX trials interference arises via a dominant target response bias to the probe. In both trial combinations, target-related response biases produce stimulus-response interference because a non-target response is required. In the emotional AX-CPT we developed, text instructions ('SMILE' and 'FROWN') were used as cues and emotionally evocative pictures (from the International Affective Picture System \[IAPS\];[@pone.0017635-Lang1] served as probes; participants were required to smile or frown in response. The target cue-probe-response combination was always emotionally congruent (i.e., smiling to 'SMILE'+pleasant picture, or frowning to 'FROWN'+unpleasant picture). BX trials (non-target cue, target probe) involved incompatibility between the probe presented and the required facial response (e.g., smiling to an unpleasant picture); in contrast, interference in AY trials (target cue, non-target probe) was due to incompatibility between the instructions of the cue and the required facial response (e.g., frowning after 'SMILE' cue). When contrasting performance in the emotion AX-CPT relative to a tightly matched non-emotional condition (in which probes were emotionally neutral), utilizing EMG measures to quantify the facial expression response, we observed that interference effects were present under both emotional and non-emotional conditions, but were strongest in the emotional AX-CPT, when both emotional and non-emotional sources of incompatibility were present [@pone.0017635-Chiew1]. In this condition, interference was due not only to standard sources of S-R incompatibility, but also because of the automatic, but inappropriate affective response to the target (e.g. being cued to smile to a negative IAPS picture). This paradigm is unique among present tasks probing emotional conflict, in that it requires integrated processing of *both* cue and probe in order to perform successfully, as opposed to requiring inhibition of the emotional information. Additionally, a major strength of the paradigm is the ability to create a closely matched analog task that permits a direct comparison of emotional vs. non-emotional conflict. Specifically, by changing probe stimuli to be affectively neutral (i.e., arbitrary symbol categories instead of emotionally evocative pictures), but retaining the other aspects of the task structure (including using facial expressions as response modality), sources of S-R incompatibility in the affective dimension are eliminated, while the standard non-affective S-R association effects driving AX-CPT effects remain (i.e., probe-driven biases and cue-driven expectancies). By comparing effects in the two conditions, it is possible to isolate the additive conflict effects specifically associated with S-R incompatibility in the affective dimension. The present study builds on our previous behavioural work by using event-related fMRI to examine whether brain activity associated with processing emotional vs. non-emotional conflict involves the same general control-related regions or qualitatively different neural circuits. Such a comparison may help to clarify further some of the outstanding contradictions present in previous emotion conflict research. On the basis of previous neuroimaging evidence, we hypothesized that both emotional and non-emotional versions of this task would engage common control-related regions including the dorsal ACC and lateral PFC. Further, based on our previous behavioural evidence, we predicted that conflict-related interference would be greater in the emotional task than in the non-emotional task, and that this would be reflected in increased levels of elicited activity within these control-related brain regions. Finally, we tested whether the emotional task was associated with the activation of potentially affectively-specialized regions, such as the rostral ACC/ventromedial PFC and amygdala, that might be selectively recruited to detect emotional conflict. Methods {#s2} ======= Ethics Statement {#s2a} ---------------- Ethics approval to conduct this study was granted by the Institutional Review Board of Washington University. Each participant provided written, informed consent prior to participation, in accordance with the human subjects guidelines established by Washington University. Participants {#s2b} ------------ Twenty-four healthy young adults (8 males, 16 females; mean age = 25.5 years, SD = 5.63) were scanned using fMRI while participating in the task. All fMRI participants were right-handed, native English speakers, and screened to ensure no neurological or psychiatric disorders, psychotropic medications, or other factors were present that contraindicated fMRI. Task Procedure {#s2c} -------------- Participants performed an emotional (Emotion condition) and non-emotional (Neutral condition) variant of the AX-CPT. The AX-CPT paradigm follows a cue-probe trial structure, in which cue stimuli set a context that is needed for appropriate response selection to the subsequent probe. The Emotion and Neutral conditions were identical in all respects except for the category of stimuli used as probes. Cue stimuli in the task were the words 'SMILE' and 'FROWN'. For probes, the Emotion condition used IAPS pictures as probes and the Neutral condition used alphanumeric symbols (i.e., letters served as target probes, and digits served as nontarget probes). New pictures/symbols were used as probes on each trial, except for a pre-specified neutral picture/punctuation mark on no-go trials (described below). Across participants the particular cue-probe combination that comprised the "AX" target trial type was counter-balanced. Thus, for approximately half of the participants (11/24) the AX target was 'SMILE'/positive picture ("SMILE"/letter in Neutral) requiring a smile response (facial expression) and the other half (13/24) the AX target was 'FROWN/negative picture ("FROWN"/letter in Neutral) requiring a frown response. However, on nontarget trials (AY, BX, BY), the opposite facial expression was required. All other details of the task paradigms described below were identical for the Emotion and Neutral conditions, and for both participant groups. Trials were presented in pseudorandom sequence, with target (AX) trials occurring at a 7∶1 frequency compared to all non-target task trials, leading to a total of 84 AX trials, 12 AY trials (target cue, non-target probe), 12 BX trials (non-target cue, target probe), 12 BY trials (non-target cue, non-target probe). Although the absolute numbers of high conflict (BX and AY) trials is somewhat low, our prior results suggest that this number was sufficient to robustly detect significant interference effects. In addition to primary task trials, no-go trials were also included to ensure that participants responded on the basis of the cue-probe combination and not solely and prematurely to the cue. No-go trials were indicated by a pre-specified neutral picture in Emotion (punctuation mark in Neutral), to which no response was to be made (24 no-go trials total; occurring both after target and non-target cues). Participants performed four scanning runs each of the Emotion and Neutral conditions of the AX-CPT (eight runs in total). Within each run, task blocks (three per run; 135 seconds each) alternated with short fixation blocks (four per run; 30 seconds each). Each scanning run began with 10 seconds of rest (later discarded) to allow the scanner to reach steady state; total run duration was ∼9 minutes. Each of the three task blocks within a scanning run consisted of 12 trials; thus participants performed eight runs of 36 AX-CPT trials each for 288 trials in total (144 Emotion, 144 Neutral). AX-CPT trials consisted of cue-probe pairs shown in sequence. Trial structure ([Figure 1](#pone-0017635-g001){ref-type="fig"}) was as follows: cue (750 ms), inter-stimulus-interval (ISI; 3250 ms), probe (2500 ms), and minimum inter-trial-interval (ITI) of 1000 ms (for a minimum total trial length of 7.5 seconds). ITIs included additional jittering to facilitate event-related response estimation, in increments of 2500 ms (no jitter, 2500 ms, 5000 ms, or 7500 ms). 72 trials were presented at each of the four ITI lengths. ::: {#pone-0017635-g001 .fig} 10.1371/journal.pone.0017635.g001 Figure 1 ::: {.caption} ###### Trial structure with timing. \(A) Example of the target (AX) cue-probe-response for the smile-AX condition of the Emotion AX-CPT; (B) Example of BX (non-target cue, target probe) and (C) AY (target cue, non-target probe) conflict trials for smile-AX condition of the task. :::  ::: fMRI Data Collection {#s2d} -------------------- Structural and functional imaging data was collected on a 3T Siemens TIM Trio whole-body scanner at Mallinckrodt Institute of Radiology at Washington University School of Medicine. High-resolution anatomical images were acquired for each participant using a sagittal T1-weighted MP-RAGE sequence (TE = 3.16 ms, TR = 2400 ms, flip angle = 8° 176 slices, 1×1×1 mm voxels). Anatomical images were aligned with each individual\'s functional images. To facilitate registration of the T1 and functional scans, a T2-weighted image was also acquired in the same space as the functional scans \[TE = 96 ms, TR = 5000 ms, 189×256 acquisition matrix, 48 slices, 1.02×1×3 mm voxels\]. The functional images were collected in eight 210TR (∼9 minutes) runs using an asymmetric spin-echo echo-planar sequence sensitive to blood oxygenation level-dependent (BOLD) contrast (T2\*) \[TE = 27 ms, TR = 2500 ms, flip angle = 90°, FOV = 256 mm, skip = 0 mm, 36 slices, 4×4×4 mm voxels\]. Stimuli were presented using E-Prime (Psychology Software Tools, Pittsburgh, PA) on a Dell PC. As described in the Task Procedure section, participants responded to each trial using emotional facial expressions. A custom-built mirror apparatus positioned over the head coil served both to reflect the projected image of the task screen towards the participant and to reflect the view of the participant\'s face such that it could be recorded using a videocamera positioned at the head end of the bore. Video recording served to ensure participant compliance in the task and was visually inspected to verify that such compliance was occurring. However, due to technical difficulties and poor video quality, this video was not quantitatively evaluated for measures of behavioural performance. A fiber-optic button box interfaced with E-Prime facilitated communication with the participant. fMRI Data Analysis {#s2e} ------------------ The fMRI data were analyzed with in-house software. Data analysis was conducted with a general linear model (GLM), including nuisance regressors for linear trends within runs and baseline shifts between runs. Additionally, the GLM contained task-related regressors for block and event-related activity. Block-related activity related to each task condition (Emotion and Neutral) were modeled as boxcar functions, but because examining sustained activity did not permit the examination of conflict effects in the data, these functions were also treated as regressors of no interest. Our experimental design follows the specifications of Visscher et al. [@pone.0017635-Visscher1] in permitting the dissociation of block and event-related fMRI activity (using alternating blocks of task and rest, as well as jittered trials within each task block); [using event-related regressors that are estimated (using delta or FIR functions) rather than assumed via a model of the hemodynamic response function. With this estimation approach, multicollinearity between the sustained and event-related regressors has been shown not to be a major concern.]{.underline} The primary task-related analysis focused on event-related activity as a function of trial type and task condition. Event-related estimates were created for each trial type within task conditions (AX, AY, BX, BY, no-go within Emotion and Neutral task versions). Given the complex trial structure, event-related effects were analyzed without reference to a fixed hemodynamic response function, using a delta-function estimation approach. Thus, within a 25-second response epoch following trial onset, independent values were estimated for each of 10 timepoints (corresponding to the 10 TR frames). The estimates from the individual subject GLMs were analyzed using appropriately designed analyses of variance (ANOVAs) that treated participants as a random factor. ### Regions of interest identification {#s2e1} We examined event-related brain activity in analyses within *a priori* defined regions of interest (ROIs). Analyses were conducted within two 'networks' of interest (selected not on the basis of functional connectivity but as coherent sets of regions observed in prior literature to be functionally related to cognitive control and reward processing). The first analysis examined activity within regions associated with cognitive control and working memory (established using meta-analyses; primarily including dorsal medial and lateral prefrontal and parietal regions [@pone.0017635-Owen1], [@pone.0017635-Wager1]. The ROI mask for the cognitive control network (CCN) was created by using anatomical coordinates identified by the aforementioned meta-analyses as seed points with 10 mm radius spheres drawn around them. The second analysis examined activity within anatomical regions associated with emotion and reward processing (hereafter EMO network) including the amygdala, portions of the basal ganglia (putamen, caudate, substantia nigra and nucleus accumbens), anterior insula, medial orbitofrontal cortex, and ventromedial prefrontal cortex, with regions drawn according to anatomical criteria identified using the Talaraich atlas [@pone.0017635-Talairaich1], and previous studies [@pone.0017635-Ahsan1], [@pone.0017635-Jensen1], [@pone.0017635-Kable1], [@pone.0017635-Knutson1], [@pone.0017635-Knutson2], [@pone.0017635-Kringelbach1], [@pone.0017635-Nitschke1], [@pone.0017635-ODoherty1], [@pone.0017635-ODoherty2]. A separate region of interest included the rostral ACC, defined anatomically, which was near to, but not overlapping the ventromedial PFC ROI [@pone.0017635-Whalen1]. For coordinates for ROIs in both networks, please refer to [Table S1](#pone.0017635.s002){ref-type="supplementary-material"} and [Table S2](#pone.0017635.s003){ref-type="supplementary-material"}. The exact masks for both networks are available from the authors by request. Significant activity within each network mask was corrected for multiple comparisons using a cluster size criterion based on Monte Carlo simulations [@pone.0017635-McAvoy1], [@pone.0017635-Forman1], via the AlphaSim software within AFNI [@pone.0017635-Ward1]. To assure a multiple comparisons corrected *p*\<.05 criteria, significant regions were identified based on a per-voxel minimum *z*\>2.32 and minimum cluster size of 37 voxels within the CCN mask (or 30 voxels within the EMO mask). Within each mask, we were interested in identifying regions demonstrating general sensitivity to conflict (e.g., across both the emotional and non-emotional tasks) and then examining whether brain activity within these conflict-associated regions differed as a function of emotional task content. Thus, the first stage analysis consisted of the following voxelwise contrast: high conflict trials (AY + BX collapsed, averaged across timepoints 4--7) \> low conflict trials (AX + BY collapsed, averaged across timepoints 4--7). This analysis further collapsed across the Emotion and Neutral conditions, in order to enable unbiased identification of regions. Timepoints 4--7 were selected to capture probe-related activity, which is necessary for the elicitation of conflict in the AX-CPT paradigm. In the second stage of analysis, we conducted ROI-based ANOVAs on significant regions identified as sensitive to conflict in the first-stage analysis. Two different kinds of region-wise analyses were carried out. In the first ANOVA, we examined which, if any, of these conflict-defined regions showed independent differences in brain activity as a function of task condition, i.e., Emotion vs. Neutral, and time (the ANOVA included all 10 timepoints). This analysis enabled a direct test of whether conflict-related regions showed increased responsivity under emotion conditions. In the second ANOVA, we only included the high-conflict trials AY and BX, to examine whether task condition effects were still exhibited selectively during conflict. Additionally, by including trial-type as a factor, we tested whether condition effects differed by the type of conflict elicited (AY = cue-based; BX = probe-based; again, timepoint was also included as a factor in the ANOVA). In addition to analyses within these networks of interest, we conducted more focused analyses within the rostral ACC and amygdala ROIs, as these regions have been specifically implicated in emotional conflict processing [@pone.0017635-Egner1], [@pone.0017635-Bush1]. The amygdala was part of the general EMO mask, but the additional analyses focused exclusively on amygdala and rACC regions, and as such utilized a more liberal corrected threshold specific to the size of each ROI (i.e., small-volume correction). Thus, for these analyses a reduced cluster-size criterion of 12 voxels for rostral ACC and 9 voxels for amygdala was employed (again with voxelwise minimum *z*\>2.32). In addition to the analyses described above, we also conducted a focused test with the rACC and amygdala ROIs to examine whether these regions show a selective response to conflict only under Emotion conditions. As such, a voxelwise contrast of high conflict (AY + BX, timepoints 4--7) \> low conflict (AX + BY, timepoints 4--7) was conducted, but only using trials from the Emotion condition. Results {#s3} ======= Behavioural Performance {#s3a} ----------------------- As described in [Methods](#s2){ref-type="sec"}, participants performed the emotional AX-CPT with voluntary emotional facial expressions as the response modality. Facial expressions were monitored in the scanner using video recording and video footage was inspected following each participant to ensure compliance with the task, but poor video quality and technical difficulties rendered this video unusable for the purposes of evaluating behavioural performance. Previously published data from our laboratory [@pone.0017635-Chiew1] investigating the Emotion and Neutral versions of the AX-CPT used here found no significant main effects of task condition on performance (indexed by error rates and response onsets), suggesting that the overall difficulty of emotional and non-emotional versions of the task may be comparable. Additionally, that study indexed performance using facial electromyography (EMG), which enables a much more fine-grained behavioural analysis than video coding would have permitted in the present study. We discuss issues with the present study\'s behavioural data and present behaviour from our previous EMG study in [Text S1](#pone.0017635.s005){ref-type="supplementary-material"} and [Figure S1](#pone.0017635.s001){ref-type="supplementary-material"}. We compared areas defined by task conflict (high \> low conflict) within the CCN and EMO networks with and without discernable errors, and found relatively few differences. These results are shown in [Table S3](#pone.0017635.s004){ref-type="supplementary-material"}. Imaging Results: ROI Analyses {#s3b} ----------------------------- As described in the [Methods](#s2){ref-type="sec"} section, event-related brain activity was examined within two 'networks' of interest: the cognitive control network (CCN) and the emotion/reward processing network (EMO). We also analyzed brain activity within more focused ROIs of the rostral ACC and bilateral amygdala. ### Conflict-defined regions within CCN and EMO ROIs {#s3b1} Within each ROI, we identified regions showing conflict-related increases in activity through the high-conflict \> low-conflict contrast, collapsing across Emotion and Neutral conditions to provide an unbiased test. Fourteen regions within the CCN, as well as five regions within the EMO network, were identified as showing conflict responses. These conflict-defined regions are summarized in [Table 1](#pone-0017635-t001){ref-type="table"}, with cortical regions shown in [Figure 2](#pone-0017635-g002){ref-type="fig"}. As expected, conflict-related regions within the CCN included the dorsal ACC and bilateral PFC, along with additional activation in the inferior parietal lobule, precuneus, thalamus, and cerebellum. The EMO regions showing sensitivity to conflict included bilateral dopaminergic midbrain, bilateral anterior insula, and left putamen. However, in this contrast, conflict-related activation was not observed in ventromedial PFC or amygdala. ::: {#pone-0017635-g002 .fig} 10.1371/journal.pone.0017635.g002 Figure 2 ::: {.caption} ###### Cortical areas sensitive to the High \> Low conflict contrast. These areas fall within the CCN and REW masks and were identified as showing significant (AY + BX) \> (AX + BY) activation, collapsed across Emotion and Neutral conditions. :::  ::: ::: {#pone-0017635-t001 .table-wrap} 10.1371/journal.pone.0017635.t001 Table 1 ::: {.caption} ###### Activity in areas defined by task conflict (high\>low conflict) within anatomically defined ROIs. ::: {#pone-0017635-t001-1} Coordinates Cluster Size (mm^3^) ROI Z BA Area Sig. Condition\*Time effect in High vs. Low Conflict Contrast Sig. Condition\*Time effect in AY vs. BX Trial Contrast Sig. Condition\*Time effect in AY vs. BX Trial Contrast --------------- ---------------------- ----- ------ --------- -------------------------- --------------------------------------------------------------- --------------------------------------------------------- --------------------------------------------------------- 0, 11, 48 7722 CCN 3.63 32 Dorsal ACC \* **\*** **\*** 41, 28, 35 1512 CCN 2.88 9 R DLPFC \* \* \* −44, 8, 33 5589 CCN 3.48 9 L IFJ \* **\*** **\*** 45, 5, 32 2673 CCN 2.93 9 R IFJ \* \* \* −49, 13, 3 2457 CCN 3.20 47 L IFG **\*** 28, 0, 54 7155 CCN 4.15 8 R superior frontal (FEF) \* **\*** −28, −1, 55 7938 CCN 3.83 8 L superior frontal (FEF) 18, −60, 43 23031 CCN 4.83 7 R precuneus **\*** −37, −52, 40 6534 CCN 4.24 40 L IPL \* **\*** 10, −12, 4 3915 CCN 3.88 \-\-\-- R thalamus \* −9, −11, 6 1215 CCN 2.89 \-\-\-- L thalamus **\*** −31, −67, −45 2052 CCN 3.08 \-\-\-- L cerebellum 32, −60, −44 1215 CCN 3.07 \-\-\-- R cerebellum 33, −62, −26 999 CCN 4.27 \-\-\-- R cerebellum \* **\*** \* 39, 20, 0 3996 EMO 4.83 47/13 R anterior insula \* −36, 17, 0 3051 EMO 4.02 47/13 L anterior insula \* \* −16, 5, −2 1161 EMO 3.36 \-\-\-- L putamen \* \* 8, −17, −10 1026 EMO 3.12 \-\-\-- R DA midbrain −6, −18, −10 1242 EMO 3.35 \-\-\-- L DA midbrain −16, −1, −11 324 EMO 2.84 \-\-\-- L amygdala \* \* 1 Significant effects of interest within these areas (condition\*time interactions within high vs. low conflict contrast and AY vs. BX trials contrast; trial\*time interactions within AY vs. BX trials contrast) are marked by asterisks in their respective columns. 2 Abbreviations: ROI = region of interest; CCN = cognitive control network; EMO = emotion/reward network; BA = Brodmann area; IFG = inferior frontal gyrus; IFJ = inferior frontal junction; DLPFC = dorsolateral prefrontal cortex; ACC = anterior cingulate cortex; IPL = inferior parietal lobule; DA = dopaminergic.; FEF = frontal eye fields. ::: ### Condition-related effects within the high versus low conflict contrast {#s3b2} In the next stage of analysis, each of these conflict-defined ROIs was subjected to an ANOVA that tested for effects of condition type, using timepoint as an additional factor to define event-related effects (i.e., in terms of a condition × time interaction). Nine ROIs showed such condition × time effects -- these areas are marked in a column in [Table 1](#pone-0017635-t001){ref-type="table"}. The areas showing sensitivity to both conflict and emotional task content included, most prominently, the dorsal ACC, right dorsolateral PFC, and bilateral posterior PFC, near the inferior frontal junction. The examination of timecourses in these nine regions revealed that, in all of them, the condition × time interaction was due to Emotion \> Neutral activation, especially in the middle timepoints where activity peaked (approximately timepoints 4--7). The timecourse of the effect within the dorsal ACC is shown in [Figure 3](#pone-0017635-g003){ref-type="fig"}, as a representative illustration of this pattern. In this and the other regions, the effects of condition did not interact with conflict, but instead were present as an additive increase in activation. In only one region, the right dopaminergic (DA) midbrain, was there evidence of a condition\*conflict interaction (at trend-level, *p* = .057). However, this interaction was due to increased activity in both the high and low conflict trials of the Emotion condition (i.e., with a reduced conflict-related increase), compared to the Neutral condition. ::: {#pone-0017635-g003 .fig} 10.1371/journal.pone.0017635.g003 Figure 3 ::: {.caption} ###### Timecourses illustrating High \> Low Conflict and Emotion \> Neutral effects. Representative regions demonstrating both a high \> low conflict and Emotion \> Neutral pattern (due to a condition\*time interaction), but no conflict\* condition interaction: (A) dorsal ACC; (B) left amygdala. :::  ::: ### Emotion and trial-type effects under high conflict {#s3b3} Because the ANOVA described above showed Emotion effects that did not interact with conflict, we conducted a follow-up ANOVA to address two additional questions: 1) Was the Emotion-related increase in activation present even when only considering high-conflict trials (i.e., AY and BX)? 2) Were there any differential effects of Emotion related to the type of conflict experienced, i.e., cue-based (AY trials) versus probe-based (BX trials)? To address these questions, the second ANOVA included only the high-conflict trial types (AY, BX) and excluded the low-conflict trials (AX,BY), to examine potential effects of condition (Emotion, Neutral) and high-conflict trial-type (AY,BX) as primary factors of interest (additional factors again included timepoint, and target expression). The primary pattern observed in the first ANOVA, a condition × time interaction, was replicated in the second ANOVA. Eight regions showed this effect, denoted in [Table 1](#pone-0017635-t001){ref-type="table"}; again, these included dorsal ACC, right dorsolateral PFC, and bilateral PFC regions. Importantly, the same Emotion \> Neutral pattern was observed in these regions, confirming that high emotion-conflict trials increased activation of the cognitive control system relative to non-emotion conflict conditions. A second pattern that was observed in the ANOVA was a trial-type × time interaction, which was significant in 11 ROIs. In all of these regions, the pattern was due to increased activation on BX trials relative to AY, during the early part of the trial (timepoints 2--5), but then comparable activation later in the trial (timepoints 6--10). [Figure 4](#pone-0017635-g004){ref-type="fig"} demonstrates this timecourse pattern in an example region, the right lateral PFC. Although a BX \> AY pattern is consistent with conflict being increased under probe-based conditions, the early, rather than late timecourse of the effect suggests that the trial-type effect might be anticipatory or expectancy-related. Note that the expectancy for high conflict is significantly greater following a B-cue (probability BX \| B-cue∼0.4) than following an A-cue (probability AY \| A-cue∼0.1). Thus, differential conflict anticipation or expectancy may account for the trial-type effects, rather than a differential response to experienced conflict during probe processing. Similar conflict expectancy effects have been observed in prior studies of the AX-CPT [@pone.0017635-MacDonald1], [@pone.0017635-Paxton1] and other conflict paradigms [@pone.0017635-Sohn1]. ::: {#pone-0017635-g004 .fig} 10.1371/journal.pone.0017635.g004 Figure 4 ::: {.caption} ###### Timecourse illustrating BX \> AY trial-type effect. Representative region in right lateral PFC exhibiting BX \> AY activity early, and comparable levels of activity in both trial types later in the timecourse. :::  ::: Although effects of condition and trial-type were present, the two factors did not appear to interact, as no regions showed evidence of condition × trial-type or condition × trial-type × time interactions. Thus, the BX \> AY pattern did not differ significantly between Emotion & Neutral conditions. ### Focused analysis of rostral ACC and amygdala activity: conflict and condition effects {#s3b4} As done previously within the CCN and EMO masks, we computed contrasts (high \> low conflict) within the rostral ACC and amygdala ROIs. However, to test whether these regions were particularly sensitive to emotion conflict per se, we conducted a follow-up ANOVA using a high \> low conflict contrast, but restricting to the Emotion condition only. No voxels within the rostral ACC or amygdala survived this contrast, contrary to evidence from previous studies suggesting their sensitivity to emotional conflict. As a final test to ensure that we did not produce any false negatives, we tested the high \> low conflict contrast, using all the data (Emotion and Neutral), but with lowered statistical thresholds, utilizing a small-volume correction for each region individually. Even with these more liberal thresholds, no rostral ACC clusters were observed; however, a small voxel cluster within the left amygdala was identified (see [Table 1](#pone-0017635-t001){ref-type="table"}). Within this conflict-sensitive left amygdala region, there was a significant effect of task condition in the full ANOVA (i.e., involving conflict, condition and timepoint as factors; see [Figure 3](#pone-0017635-g003){ref-type="fig"}). This interaction was due to a similar pattern of activity to that observed in several other regions within the CCN and EMO networks (i.e., Emotion \> Neutral activity). Similarly, as with these other regions, no condition\*conflict interaction was observed. Indeed, if anything, the high \> low conflict effect was weaker in the Emotion condition relative to Neutral ([Figure 3](#pone-0017635-g003){ref-type="fig"}), consistent with the absence of a significant conflict effect in this region when only the Emotion condition was examined. Additionally, in this left amygdala region, no effects of trial type were observed in the ANOVA contrasting AY and BX trials. Together, these results confirm that the rostral ACC and amygdala did not show any selective emotion conflict effects, and the small left amygdala region that was identified showed a pattern of activation that was very similar to other regions within cognitive control network, i.e., sensitivity to both to the presence of task conflict and to emotional processing, but no preferential response to emotional conflict (e.g., these factors did not interact with one another). Discussion {#s4} ========== With the present study, we examined neural activity associated with emotional versus non-emotional conflict using a novel paradigm: the emotional AX-CPT. This paradigm capitalized on the use of controlled facial expressions as a response modality to generate S-R incompatibility that was either emotional or non-emotional in nature. The examination of brain activity associated with the processing of these two forms of S-R incompatibility helps clarify the extent to which emotional conflict relies on neural circuitry common to that associated with more traditionally studied forms of cognitive conflict. Specifically, the current findings suggest that both emotional and non-emotional conflict commonly engage a number of brain regions associated with cognitive control, including the dorsal ACC and lateral PFC, as well as certain areas implicated in both emotional processing and cognitive control, such as bilateral anterior insula. Additionally, most of these common regions demonstrated higher activity when processing emotional (versus non-emotional) conflict; in contrast, we observed no conflict-sensitive regions where the non-emotional task elicited greater activity than the emotional task. Our findings are in line with several other studies examining the neural basis of emotional versus non-emotional conflict. Processing of both kinds of conflict may rely on cognitive control-related brain areas [@pone.0017635-Ochsner1], [@pone.0017635-Compton1]. In particular, mechanisms underlying both emotional and non-emotional conflict detection have been localized to the dorsal ACC [@pone.0017635-Haas1], [@pone.0017635-Egner1], [@pone.0017635-Wittfoth1]. However, the present results are inconsistent with the older hypothesis that rostral and dorsal subdivisions of the ACC are devoted to processing emotional and cognitive conflict, respectively [@pone.0017635-Bush1]. In particular, although we observed robust conflict-related activation in the dorsal ACC in both Emotion and Neutral conditions, no such patterns were observed in the rostral ACC, even when focusing exclusively on Emotion conflict. The absence of emotion-specific conflict regions in the ACC during task processing may be surprising from the perspective of classic theoretical distinctions, but is actually relatively consistent with the prior literature. As discussed previously, original variants of the emotion Stroop actually target emotional distraction or even non-affective variables, and as such may not be appropriate for the study of emotional conflict, as suggested by recent conceptual analyses [@pone.0017635-Buhle1]. More recent studies that utilize conflict-based variants of the emotional Stroop and related tasks have been equivocal as to whether rostral ACC is either engaged, or associated with the detection (rather than resolution) of emotional conflict [@pone.0017635-Haas1], [@pone.0017635-Egner1]. Additionally, in one recent study rostral ACC activity during the emotion-conflict Stroop was dependent on the trait anxiety level of participants [@pone.0017635-Krug1]. Thus, the current study adds to prior literature in suggesting that caution is warranted regarding whether the rostral ACC should in fact be associated with emotion conflict processing per se. Instead, further investigation of this region is needed, that focus on examining potential alternative accounts such as emotional distraction, conflict resolution, and individual trait anxiety. In contrast to the pattern in the rostral ACC, there were significant effects of emotion on activation in a number of regions associated with cognitive control functions, including the dorsal ACC and lateral PFC. Interestingly, these effects were observed as significant condition (Emotion \> Neutral) and conflict (Conflict \> No Conflict) effects, without a significant condition × conflict interaction. In other words, the emotion effects were additive to conflict, rather than interactive, which suggests two independent mechanisms. At first glance, this pattern seems somewhat counter-intuitive, since the presence of affectively-valenced content did not selectively modulate the magnitude of the conflict effect, but instead increased activation equivalently on both high and low-conflict trials. Nevertheless, the pattern may actually be fairly consistent with interpretations regarding the nature of emotional conflict and control. In particular, a key feature of the Emotional AX-CPT is that in the Emotion condition, there should be a relatively automatic, but task-irrelevant, subjective emotional reaction to the affective content present in the probe. This emotional reaction is task-irrelevant because correct response selection requires consideration only of the cognitive classification of the probe as having positive or negative content (and in integrating this information with cue classification). Indeed, the subjective emotional response to the probe, which may automatically trigger a tendency to activate the associated facial expression, can lead to an additional source of response uncertainty. For example, if viewing a negatively valenced probe stimulus triggers a tendency to make a frown expression (or likewise, if viewing a positively valenced probe stimulus triggers a tendency to smile), confusion can be generated regarding whether this "expression tendency" is appropriate for the current trial (i.e., correct on AX but incorrect on BX trials). Under such circumstances, from a cognitive control perspective, the optimal task strategy would be to suppress any subjective emotional responses that might be experienced in order to reduce response uncertainty. Because such task-irrelevant emotional response tendencies can occur on all trials in the Emotion condition, there would be generally higher cognitive control demands in this condition relative to Neutral. In addition to the additive effects of emotion and conflict observed in regions associated with cognitive control, this same pattern was also present in the left amygdala, at least under an adjusted statistical threshold. The amygdala has typically been thought of as an emotion processing region whose activity, in conflict, distraction, and regulation paradigms, will reflect the emotional valence of stimuli, rather than tracking cognitive control demands [@pone.0017635-Egner1], [@pone.0017635-Dolcos1], [@pone.0017635-Ochsner2]. However, prior findings of amygdala activity associated with increased cognitive control have also been repeatedly observed, although they typically receive less attention in the literature. For example, in one study increased amygdala activation was associated with improved behavioral performance during working memory, selectively under high-load conditions [@pone.0017635-Schaefer1]. This finding, and others [@pone.0017635-Holland1], [@pone.0017635-Holland2], [@pone.0017635-Ousdal1], supports alternative theoretical views of amygdala function, in which this regions is postulated to mediate general vigilance/goal-relevance-detection processes that contribute to enhanced cognitive performance as well as processing of emotional demands [@pone.0017635-Davis1], [@pone.0017635-Sander1]. The pattern of left amygdala activation in the present task -- associated with both emotion and conflict-processing -- might be better characterized by such an explanation, especially considering that emotional information must be evaluated for valence, while at the same time suppressing subjective emotional responses, in order to optimally perform the task. Beyond the main effects of condition and conflict, a number of regions also exhibited distinct patterns of activity as a function of the type of interference present. As in the original AX-CPT, the emotional AX-CPT involves non-target trials eliciting conflict via two different forms of interference: AY trials, where interference is cue-based and relatively top-down in nature, and BX trials, where interference is probe-based and relatively bottom-up in nature. A number of frontal and parietal regions associated with cognitive control demonstrated significant trial effects in the present study, primarily because of BX \> AY activity early in the trial (with comparable activity levels late in the trial). Previous studies of the AX-CPT have observed similar patterns of activation within the lateral PFC and other regions, demonstrating the robustness of the effect [@pone.0017635-MacDonald1], [@pone.0017635-Paxton1]. The pattern of activity is typically interpreted as reflecting the higher degree of interference expectancy associated with B-cues (i.e., associated with non-target responses) relative to A-cues (i.e., associated with target responses), and thus increased demands for proactive cognitive control [@pone.0017635-Braver3]. The current study extends this finding by demonstrating that this interference expectancy effects can be exhibited during emotional as well as non-emotional AX-CPT conditions. As such, the current results support the general notion that participants utilize the same types of proactive control strategies even when experiencing high demands for such control as a result of emotional conflict. The emotional AX-CPT paradigm presented in the present study, and the use of emotional facial expressions as a response modality more generally, have the potential to provide a more naturalistic technique from which to probe emotional conflict, relative to the previous laboratory paradigms that have been used. Facial expressions have direct, automatic associations with different emotional experiences [@pone.0017635-Cacioppo1]; thus, they potentially provide a performance measure that is a more sensitive index of both trial-by-trial fluctuations and individual differences in emotional processing. In the present study we were not able to obtain behavioural performance measures due to technical difficulties, but future studies capitalizing on this technique should explore this possibility (e.g., via simultaneous EMG and fMRI recordings). Additionally, using facial expressions as responses permits elicitation of conflict via S-R interference, which is a robust form of interference that has nevertheless been understudied (relative to S-S interference) in the domain of emotion. The utilization of facial expressions as a response modality provides a potential means to probe emotional conflict via S-R interference in other paradigms as well, such as the Stroop adaptations utilized by Egner and colleagues [@pone.0017635-Egner1], [@pone.0017635-Etkin1]. For example, in Stroop conditions that require participants to make facial expressions to semantically associated words (e.g., "smile", "frown") while ignoring irrelevant but superimposed affectively-valenced pictures, it would be possible to manipulate congruency in an analogous manner to that examined here. One of the advantages of developing adaptations of the Stroop and related paradigms (e.g., Flankers, Simon) that include facial expressions as a response modality is that it would permit exploration of experimental manipulations not easily implemented in the AX-CPT. In particular, conflict-related shifts in control state (e.g., conflict adaptation or resolution effects) have been profitably examined through manipulation and examination of trial-by-trial sequential effects [@pone.0017635-Gratton1], changes in relative trial frequencies [@pone.0017635-Carter3], and other similar effects. As a means of eliciting emotional conflict in a naturalistic, ecologically valid manner, the S-R incompatibility elicited through facial expression-based responding has the potential to be exploited in a similar variety of experimental manipulations, contributing to our knowledge of the behavioural and neural mechanisms underlying emotional conflict processing. It is our hope that this technique may provide one direction by which investigations of emotional conflict may approach the rigor and sophistication of similar research within the more traditional realm of cognitive control. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### \(a) Error rates and (b) response onset times measured via EMG in the Emotion AX-CPT, from Chiew & Braver (2010), as a function of Condition (Emotion vs. Neutral) and Conflict (high vs. low). (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Centres of mass for cognitive control network (CCN) regions of interest (ROIs) used to mask the neuroimaging data. (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Coordinates for hand-drawn emotion/reward-related (EMO) regions of interest (ROIs) used to mask the neuroimaging data. (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### Activity in areas defined by task conflict (high \> low conflict) within anatomically defined ROIs with discernable errors eliminated, and comparable areas with all trials included (from [Table 1](#pone-0017635-t001){ref-type="table"}). (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S1 ::: {.caption} ###### (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by grant RO1 MH66078 from the National Institutes of Health to T.S.B. and a Natural Sciences and Engineering Research Council of Canada Postgraduate Scholarship to K.S.C. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: KSC TSB. Performed the experiments: KSC. Analyzed the data: KSC TSB. Wrote the paper: KSC TSB.

PubMed Central

2024-06-05T04:04:19.197502

2011-3-9

PMC3052362

Introduction {#s1} ============ Mice are highly vocal animals, with both males and females vocalizing in same-sex and cross-sex social encounters [@pone.0017460-Panksepp1], [@pone.0017460-Gourbal1], [@pone.0017460-Maggio1], [@pone.0017460-Stowers1], [@pone.0017460-Liu1]. Mouse pups are also highly vocal, producing isolation vocalizations when they are cold or removed from the nest, despite the fact that they cannot hear until postnatal day 10 (p10) [@pone.0017460-Ehret1], [@pone.0017460-Romand1]. Adult mice can discriminate between vocalizations of pups and adults. For example, virgin female mice are attracted to playbacks of male song, but not pup vocalizations [@pone.0017460-Hammerschmidt1]. In contrast, playback of pup vocalizations to mothers, but not to pup-naïve virgins, elicits search and retrieval behavior [@pone.0017460-Haack1], [@pone.0017460-Ehret2]. These behavioral differences have a correlate in the auditory cortex, where physiological responses to pup syllables differ for maternal and virgin animals [@pone.0017460-Liu2], [@pone.0017460-GalindoLeon1]. The complexity of acoustic communication behaviors and the presence of neural correlates that underlie some of these behaviors provide a strong rationale to explore the details of this vocal communication system and how it changes developmentally. Previous work has shown that the frequency and duration of pup vocalizations change during development. Liu and colleagues [@pone.0017460-Liu1] reported that adult syllables are shorter than pup syllables and that the peak frequency of the syllables also differs between adults and pups, with adult syllables having frequencies that fall between the two frequency ranges occupied by pup syllables. However, it is not clear that these changes occur for all of the pup syllable types characterized by Scattoni and colleagues [@pone.0017460-Scattoni1]. If so, the changes might reflect maturation of the vocal tract or vocal control. If all of the syllable types do not change in the same way, it may suggest that vocal learning occurs as the acoustic features of syllables are refined. In adults, syllables are produced in bouts that have a song-like structure [@pone.0017460-Holy1]; different syllable types are produced in temporally organized patterns. How these adult patterns develop is not known, nor is the relationship between them and the patterns of pup isolation vocalizations understood. Changes in syllable sequencing that occur during development may provide additional cues for adult animals to differentiate the calls of mouse pups from those of adults. To address these issues, we examined the structure and sequencing of syllables across development and compared these features to those in adults. We characterized the vocalizations of CBA/CaJ mice because they are the standard *normal control* strain used for auditory research, due to their sensitive hearing thresholds that are maintained up to at least 39 weeks [@pone.0017460-Zheng1]. We recorded isolation vocalizations from pups between ages p5 and p13, and from adults in a variety of social settings. The results show that both the probability of different syllable types being produced and their sequencing within bouts changed with the age of pups, and that these features differed from adults. We used these data to generate a MATLAB program we have called a 'virtual mouse vocal organ'. This probabilistic model of mouse vocalizations generates bouts of mouse syllables with acoustic features appropriate to mice of different ages. This program allows for highly controlled generation of vocal stimuli that correspond to those of mice at different developmental stages for use in behavioral and neurophysiological experiments. Materials and Methods {#s2} ===================== Ethics Statement {#s2a} ---------------- All procedures were approved by the Institutional Animal Care and Use Committee at the Northeastern Ohio Universities Colleges of Medicine and Pharmacy (Approval ID number 10-001). A total of 42 CBA/CaJ mice were used in this study, comprising 18 adult males, 12 adult females (aged between p91 and p140) and 15 pups from 3 litters with different parents. Acoustic recordings {#s2b} ------------------- Recordings of mouse vocalizations were carried out in a single-walled acoustic chamber (Industrial Acoustics, New York, NY) lined with anechoic foam. Mice were situated within an open topped cylindrical arena (D 165 mm, H 110 mm) placed over a heating pad. Acoustic signals were recorded by ultrasonic condenser microphone (CM16/CMPA, Avisoft Bioacoustics, Berlin, Germany), located 5 cm (for pups) or 7 cm (for adults) above the arena floor. The acoustic signals were amplified and then digitized at 500 kHz with 16-bit resolution (UltraSoundGate, Avisoft Bioacoustics). Recorded signals were displayed in real-time on a computer with commercial software (Recorder\_USGH, Avisoft Bioacoustics). Gain was adjusted online to prevent signal saturation. The recording system was flat (±3 dB) from 20 kHz to 140 kHz, with a low frequency roll-off of 12 dB per octave. The system provided a strong signal-to-noise ratio that did not require offline filtering (see unfiltered recording sequence in [Figure 1](#pone-0017460-g001){ref-type="fig"}). ::: {#pone-0017460-g001 .fig} 10.1371/journal.pone.0017460.g001 Figure 1 ::: {.caption} ###### Spectrograms and waveforms of typical syllables recorded from CBA/CaJ pups. All of the syllables shown from a p9 pup (top) have frequency steps and one harmonic, the sound levels of consecutive syllables varied substantially. The final two syllables shown from a p13 pup (bottom) have nonlinearities within their 2nd elements. :::  ::: Recordings of pup isolation vocalizations, lasting 5 minutes, began 5--10 seconds prior to placing the pup in the center of the arena. Pup vocalizations were recorded from 15 pups at p5, and from 14 pups at the subsequent ages p7, p9, p11 and p13. Pups were from 3 litters with different parents. Mouse pups did not produce isolation vocalizations after p13, in agreement with previous work [@pone.0017460-Haack1]. For recording adult vocalizations, several behavioral paradigms were used: male-male interactions (4 pairs), male-female interactions (6 pairs), female-female interactions (3 pairs) and male vocalizations in the presence of female bedding (4 animals). In adult encounters, one animal was introduced to the recording chamber a few seconds prior to the second animal. Recording was started when the first animal was introduced and continued for 30--60 minutes. Data Analysis {#s2c} ------------- ### Basic analyses of spectro-temporal features {#s2c1} Syllables were detected offline using SASLab Pro 5.1 (Avisoft Bioacoustics). Since syllable amplitude was variable ([Fig. 1](#pone-0017460-g001){ref-type="fig"}) and sometimes close to the background noise floor, automated thresholding was not used to detect sounds. Instead, the start and end times of the syllables were manually tagged onto the sound file, providing a measure of syllable duration and marking syllables for further analysis. The dominant frequency, the frequency that was produced at the maximum amplitude, was also measured at several points within a syllable: the start (+2 ms), center and end (−2 ms). All syllables recorded were included in the analysis so long as they were not saturated or associated with movement noise. Syllables were separated into distinct categories based on their spectro-temporal characteristics. The classification scheme used by Scattoni *et al.*, (2008) provided an initial guide, and in most cases we used their syllable categories. Several frequency metrics were calculated depending on the spectro-temporal characteristics of the syllable type. The dominant frequency was computed for constant frequency syllables (the flat syllable and the short syllable) and syllables that had both upward and downward frequency modulations (the complex syllable and the chevron syllable). The bandwidth of frequency modulated syllables was calculated by subtracting the dominant frequency at the lowest frequency of the syllable from the dominant frequency at the highest frequency of the syllable. For syllables with frequency steps, the dominant frequency of each 'component' was computed along with the frequency difference between components, i.e., the step size. These characteristics were compared among pups of different ages and between each pup age group and adults. ### Zipf relation {#s2c2} Zipf\'s law was developed to describe the proportion of word usage within human language; it holds that within a natural language, the frequency of occurrence of a word is inversely proportional to its rank in the frequency table [@pone.0017460-Zipf1], [@pone.0017460-Zipf2]. We compared the Zipf\'s relation value for the vocal repertoire in order to identify whether the use of each syllable type was non-random and to calculate the capacity for the mouse vocal repertoire to carry information. The Zipf\'s relation was compared among pup groups and adults. Slopes of the Zipf relation were computed as the regression coefficient when the logarithm of frequency of occurrence of each syllable is plotted against the logarithm of the syllable rank (the most common syllable, the second most common syllable and so on) [@pone.0017460-Zipf1], [@pone.0017460-Zipf2]. The resulting slope was used as a measure of the potential of the repertoire to carry information from the caller to the listener. A slope of −1 is considered to represent the optimal balance between the number of syllable types in an animal\'s vocal repertoire and the level of repetition of each syllable type [@pone.0017460-McCowan1]. A shallower, slope (closer to 0) represents a repertoire that is more diverse and random, whereas a steeper, more negative, slope represents a repertoire that is less diverse and more repetitious. Syllable sequence analysis {#s2d} -------------------------- The Zipf\'s statistic provides a measure of the potential for the vocal repertoire as a whole to carry information, but it does not reveal higher-order structure that may be present in sequences of syllables within bouts. We used two statistical methods to test whether successive syllables within a syllable bout were independent of one another, or conversely, whether there was an effect caused by one or more of the preceding syllables in the bout. We used both a chi-squared goodness-of-fit test and a higher-order entropy model based on information theory. Chatfield and Lemon [@pone.0017460-Chatfield1] outlined the advantages of using both methods in unison. The chi-squared test determines whether the repertoire is random or has some higher-order structure. However, it does not clearly determine the sequential level at which the repertoire has higher-order structure. Information theory entropy analysis provides a graphical representation of the data that allows for comparisons between the levels of organization, information entropy, of the repertoire at higher-orders. The information theory entropy measure also allows for comparison between syllable repertoires from different groups of animals. We used the chi-squared goodness-of-fit test at the one-syllable and the two-syllable levels to test whether the repertoire had internal structure. At the one-syllable level, the proportions of each syllable type were compared against a random model using a chi-squared analysis, testing whether the syllables were produced in a nonrandom pattern. At the two-syllable level, we examined syllable combinations for all bouts having more than 3 syllables, at all ages. The frequency of occurrence of each possible two-syllable type pairing was compared, using chi-squared, to a random model. A significant difference reflects an ordered structure within the repertoire where the probability of a particular syllable being produced is affected by the previous syllable; however it does not provide specific information as to what the probable patterns are. We further used the two-syllable pairings to assess the complexity of the repertoire across development. We considered a repertoire to be more complex if there was a higher probability of switching between syllable types than repeating the same syllable type. This was achieved by summing all repetitions, regardless of syllable type, and dividing the sum by the total two-syllable pairings. We used information theory to measure the entropy of the repertoire at several different levels; zeroth order, first order, second order and third order. Zeroth order measures the diversity of the vocal repertoire, in this case how many syllable types there are at each age. First-order entropy measures the simple organizational structure of the repertoire, how often each syllable type is used; this is akin to the Zipf\'s statistic. Second-order entropy measures the level of organization at the two-syllable sequence level. Third-order entropy describes the extent to which the bouts are organized at the three-syllable level. The highest level we compared is fourth-order entropy; this investigates to what extent the bouts are organized at the four-syllable level. If a repertoire has organization at higher levels, then the entropy should reduce significantly with each increasing level. We used equations for the formation of informational entropies described by Doyle *et al.* [@pone.0017460-Doyle1]:where H~0~ is the maximum entropy of the vocal repertoire in bits, the zeroth-order entropy, and *N* is the total number of syllable types. First-order entropy is calculated using:where *p(i)* is the probability of a syllable type being produced within the mouse vocal repertoire at a given age. Second-order entropy is reduced by any higher-order structure present within the repertoire:where *p(i,j)* is the joint probability of two syllable types being produced and *P~i~(j)* is the conditional probability that syllable *j* will be produced given that syllable *i* has just occurred within a bout. Third-order entropy goes on to measure the entropy, in bits, at the three-syllable level:where *p(i, j, k)* is the combined probability of three syllable types being produced and *P~ij~(k)* is the conditional probability that *k* will be produced given that both *i* and *j* have just been produced sequentially. When the entropy value of the repertoire, in bits, is plotted against the entropic level, zeroth to fourth order, the degree of change between two entropic orders reflects the level to which the repertoire is reliant on higher-order sequential patterns. In a truly random system the entropic value would be equal at all levels. A significant reduction in entropy between orders is indicative of a pattern in the sequencing of syllables within bouts, at least up to that level. It has been proposed that vocal learning animals learn the higher-order structure of their communication system, so that the adult repertoire would have a greater level of higher-order structure than that of pups [@pone.0017460-McCowan1]. We compared the differences between the entropy at different levels using a chi-squared analysis. For example, comparison of the change in entropy between and was made using the following [@pone.0017460-Chatfield1]:where N~1~ is the total number of syllables sampled and is -and is -. This can be used to compare differences between values at each entropic level, where would be the difference between and , where *a* and *b* are the repertoires from animals of different ages. This measure gives the diversity of the proportion of each syllable used in the repertoire, which is akin to comparing the Zipf\'s statistic ratios across ages. For this analysis, the vocal repertoire should comprise similar numbers of syllable types, as the T values would be expected to be greater when there are more syllable types. Large samples are needed for these analyses because the degrees of freedom become very large at the higher entropic levels. The degrees of freedom are calculated as described by Chatfiel and Lemon [@pone.0017460-Chatfield1] (see [Table 1](#pone-0017460-t001){ref-type="table"}). Where c is the number of syllable types produced. When testing for a significant difference in the level of entropy between two non-adjacent entropy levels, for example from and , the degrees of freedom are the sum of those at and : (DF = (*c*-1)^2^+*c*(*c*-1)^2^). If there is a significant decrease in entropy between and , but not between and , it shows that the communication system has internal structure up to the two-syllable level, but no additional structure at the three-syllable level. We compared the magnitude of the change, across age, between and . This measured the difference in the level of higher-order structure across ages, providing a measure of the age-related changes in the organization of the bouts at the two-syllable level (the change in entropy between and) up to the four-syllable level (the change in entropy between and ). If the change is increasingly negative as pup development proceeds, there is an increasing level of higher-order structure in the sequencing of syllables. ::: {#pone-0017460-t001 .table-wrap} 10.1371/journal.pone.0017460.t001 Table 1 ::: {.caption} ###### Equations for calculating the degrees of freedom at each entopic level. ::: {#pone-0017460-t001-1} Entropic level Degrees of freedom ---------------- -------------------- Zeroth *c*-1 First (*c*-1)^2^ Second *c*(*c*-1)^2^ Third *c^2^*(*c*-1)^2^ ::: The virtual mouse vocal organ {#s2e} ----------------------------- We generated a 'virtual mouse vocal organ' within MATLAB. This program generates bouts of mouse vocalizations with acoustic features that are appropriate to each of the mouse ages studied. First, a version of each syllable type which was typical in its spectro-temporal characteristics was selected for each age. Sequences of syllables are generated using a probabilistic Markov model. We examined syllable combinations at all ages for all bouts with more than 3 syllables. The probability of occurrence of each possible two-syllable type pairing was measured, as was the probability of occurrence of each three-syllable sequence. The probability of each syllable occurring first within a bout was also computed at each age. The transitional probabilities of the model were calculated individually for each age group from the proportion of syllable pairs and triplets. For each age group either a first, second, or third-order model can be selected. Initially, the first-order Markov model pseudo-randomly selects from the syllable types on the basis of their probability of occurrence. Thus the most probable first syllable will not always be selected. The second syllable is selected in the same way, using the transitional probabilities of each syllable being vocalized following the syllable type that was selected for syllable one. The third syllable is selected based on the probability of following the second syllable; this is independent of the first syllable. In this way an infinite number of syllables can be produced, and the selection of each syllable added to the syllable sequence is affected by the probability with which it follows the previous syllable. Within the second-order Markov model, each new syllable is selected pseudo-randomly based on the associated probability that it follows the preceding two syllables. The program outputs wav files made up of the typical syllables arranged in a probable sequential pattern that is appropriate for each age from p5 to p13 and for adults. The bouts are generated with the associated average inter-syllable intervals. Results {#s3} ======= Although syllables were occasionally produced in discrete utterances, both pups and adults generally produced syllables in bouts that comprised several syllables. A large number of syllables were collected and analyzed from animals at each age; p5, n = 3145; p7, n = 4329; p9, n = 6306; p11, n = 4560; p13, n = 3082; adults, n = 6963 (male-male, n = 1382, male-female, n = 4121, female-female, n = 188, male with female bedding, n = 1272). We first discuss the prevalence of different syllable types across pup age and then compare the spectro-temporal characteristics of the syllables. Finally, we characterize the age-related changes in the sequencing of the syllable bouts. Vocal repertoire {#s3a} ---------------- ### Syllable types {#s3a1} We identified 11 syllable types, 9 of which were similar to those described by Scattoni *et al.*, [@pone.0017460-Scattoni1]. In contrast to their study, however, we did not classify harmonic (termed composite by Scattoni *et al.*, [@pone.0017460-Scattoni1]) or nonlinear (termed harmonic by Scattoni *et al.*, [@pone.0017460-Scattoni1]) sounds as separate syllable types. Instead, within each syllable type, we noted whether a syllable was tonal, harmonic or nonlinear; these features are described in a later section. Holy and Guo [@pone.0017460-Holy1] characterized syllables as having either frequency jumps or a sinusoidal structure; they did not break down the sinusoidal syllable type into subtypes based on the direction of frequency change, for example an upward frequency modulation or a downward frequency modulation. The syllable types we found are described below and are shown in [Figure 2](#pone-0017460-g002){ref-type="fig"}. 1. *Complex syllables* were monosyllabic with two or more directional changes in frequency \>6 kHz. 2. *1 Frequency step syllables* (1 freq. step) had two elements, in which the second element was ≥10 kHz different from the preceding element and there was no separation in time between steps (these are similar to the *two syllable* calls described by Scattoni *et al.*, [@pone.0017460-Scattoni1] and the single frequency jumped syllables described by Holy and Guo [@pone.0017460-Holy1]). 3. *2 Frequency step syllables* (2 freq. step) have three elements, in which the second element was ≥10 kHz different from the first and the third element was ≥10 kHz different from the second. There was no separation in time between elements (similar to the *frequency steps* syllable described by Scattoni *et al.*, [@pone.0017460-Scattoni1] and the multiple frequency jumped syllables described by Holy and Guo [@pone.0017460-Holy1]). 4. *Up-FM syllables* were upwardly frequency modulated with a frequency change ≥6 kHz. 5. *Down-FM syllables* were downwardly frequency modulated with a frequency change ≥6 kHz. 6. *Flat syllables* were constant frequency syllables with modulation \<6 kHz. 7. *Short syllables* lasted ≤5 ms. 8. *Chevron syllables* were shaped like an inverted U. The highest frequency was at least 6 kHz greater than the starting and ending frequencies. 9. *Reverse Chevron syllables* were shaped like a U. The lowest frequency was at least 6 kHz less than the starting and ending frequencies. 10. *Low Frequency Harmonic syllables* (LFH) were harmonic stacks with fundamental frequencies below 5 KHz. These syllables often had harmonics extending into the ultrasonic range in adults (\>20 kHz). These have previously been classified in pups as wriggling calls [@pone.0017460-Ehret3] and in adults as pain sounds [@pone.0017460-Williams1] or low frequency harmonics [@pone.0017460-Portfors1]. These syllables were not analyzed for developmental differences because they occur at frequencies within the sharp low-frequency cutoff of the recording microphone. 11. *Noisy syllables* were warbled, noisy, harmonic syllables in the 10--120 kHz range (these are different from the low frequency harmonic syllables given by mice). ::: {#pone-0017460-g002 .fig} 10.1371/journal.pone.0017460.g002 Figure 2 ::: {.caption} ###### Spectrograms of the different syllable types produced by CBA/CaJ mice. The majority of syllables have energy solely in the ultrasonic range (20 kHz) with the exception of the Noisy syllable and the low frequency harmonic (LFH) syllable. Note substantial energy at frequencies above 100 kHz for many syllables. Tonal syllables are marked (A); these syllables have no harmonics or nonlinearities. Harmonic sounds are marked with a (B). Nonlinear sounds are marked with a (C); these syllables have subharmonics or deterministic chaotic elements. :::  ::: In pups of any age or in adults, some syllable types were consistently produced more commonly than other types ([Fig. 3](#pone-0017460-g003){ref-type="fig"}) (chi-square range 1880--8434, p\<0.001). Among pups, the most commonly produced syllables were the 1 freq. step syllable (p5), the flat syllable (p7), and the 2 freq. step syllable (p9, p11, and p13). Adult mice most commonly produced the up-FM syllable. Only one syllable type was unique to adult animals, the noisy syllable. ::: {#pone-0017460-g003 .fig} 10.1371/journal.pone.0017460.g003 Figure 3 ::: {.caption} ###### Changes in the proportions of syllable types across postnatal development. Each column shows proportions of most common syllables at the specified age. The proportions of the noisy syllable are not shown as it was only produced very rarely by adult mice, representing only 0.6% (44/6936) of syllables. :::  ::: Across age, the numbers and proportions of different syllable types changed ([Fig. 3](#pone-0017460-g003){ref-type="fig"}). To analyze this, we performed chi-squared analyses on scaled proportional data for the 6 most common syllable types; flat, 1 freq. step, 2 freq. step, chevron, down-FM and up-FM. There was a highly significant effect of age on the proportions of each of these major syllable types. For example, the proportion of the 2 freq. step syllable increased steadily as pups aged ([Fig 3](#pone-0017460-g003){ref-type="fig"}); this syllable was relatively rare among p5 syllables but was the most common syllable type by p13. The proportion of 1 freq. step syllables changed with age (Χ^2^(5, *N* = 5436) = 2475, *p*\<0.001); this syllable was the most common at p5 but was rarely produced by adult animals. The most common adult syllable was the up-FM (see [Fig. 2](#pone-0017460-g002){ref-type="fig"}). This syllable occurred with differing proportions in different pup ages (Χ^2^(4, *N* = 562) = 383, *p*\<0.001)\], becoming increasingly common with age. There were also significant age-related changes in the probabilities of the flat syllable (Χ^2^(5, *N* = 7670) = 1087, *p*\<0.001), the chevron syllable (Χ^2^(5, *N* = 4077) = 345, *p*\<0.001) and the down-FM syllable (Χ^2^(5, *N* = 1603) = 342, *p*\<0.001). ### Complexity of vocal repertoire {#s3a2} A Zipf\'s statistic was used to compare the structural complexity of the mouse repertoire across ages. As pups increased in age, the Zipf\'s statistic slope became closer to −1, meaning that the second most common syllable is used half as frequently as the most common syllable, and so on (see [Fig. 4A](#pone-0017460-g004){ref-type="fig"}). At p5, the slope was near −2, indicating that the vocal repertoire was highly repetitious (Zipf slopes by age: p5, −1.97; p7, −1.98; p9, −1.66; p11, −1.44; p13, 1.03 and adult-1.48). By age p13, the slope was very close to the optimal −1. This value was even greater than that found from adult mouse communication syllables. ::: {#pone-0017460-g004 .fig} 10.1371/journal.pone.0017460.g004 Figure 4 ::: {.caption} ###### Change in the sequencing of syllables within bouts, across age. \(A) The Zipf slopes increase with age, showing that the repertoire became less repetitious with increasing age. (B) The probability of a switch between syllables, in consecutive calls, increased with age. (C) For mice of different age groups, entropy declines as function of structural order. The negative slope indicates a higher-order structure in the syllable sequences at each age. (D) Slopes of the first to third order entropies, normalized to the value for adults for graphical purposes only. Note that the negative slope is steeper for p13 pups than for the younger pups, indicating greater higher-order structure within song bouts. P13 animals had significantly more sequential structure than younger animals. Adults had significantly more sequential structure than pups at any age. :::  ::: The change in entropy between H~0~ and H~1~ increased steadily from p5 to p13, indicating that the repertoire became increasingly diverse (See [Fig. 4C](#pone-0017460-g004){ref-type="fig"}). The entropic change between H~0~ (zero order) and H~1~ (first order) was 1.24 bits at p5 and 0.48 bits at p13 (Χ^2^ (8) = 6568, p\<0.001). This change reflects the changes in Zipf\'s statistic that suggest that the repertoire became more diverse and less repetitious over pup development. Acoustic Features of Syllables {#s3b} ------------------------------ The acoustic features of most pup syllables changed during development, and almost always differed from the syllables of adults. The nature of these changes was complex. We illustrate developmental changes for two syllables that were common both in pups and adults, the flat syllable ([Fig. 5](#pone-0017460-g005){ref-type="fig"}) and the chevron syllable ([Fig. 6](#pone-0017460-g006){ref-type="fig"}), then follow with quantitative and statistical comparisons across age for the remaining syllables. In these analyses, all adult syllables of the same type are grouped together, since we found no significant differences in acoustic features of these syllables as a function of the category of social interaction. ::: {#pone-0017460-g005 .fig} 10.1371/journal.pone.0017460.g005 Figure 5 ::: {.caption} ###### Analysis of the duration and frequency of flat syllables across ages. \(A) The most typical flat syllable at each age. (B) Distributions of durations of the flat syllable at each age. Older pups are progressively less likely to produce longer flat syllables. (C) Distributions of the dominant frequency of the flat syllables at each age. At p5, p7 and p9 there are two clear peaks in the distribution, but by p11 there are fewer low frequency syllables. The frequency of the higher peak gradually reduces with age. (D) The kurtosis of the duration distribution gets more positive with age, indicating a more peaked distribution of duration. Bars represent the standard error of the mean. (E) Scatter plots of duration against frequency for flat syllables from animals aged p7 and adults. Although there is some overlap, the frequencies of adult syllables fall between those of the pup syllables. :::  ::: ::: {#pone-0017460-g006 .fig} 10.1371/journal.pone.0017460.g006 Figure 6 ::: {.caption} ###### Analysis of the duration and frequency of chevron syllables across ages. \(A) The most typical chevron syllable at each age. Note the similarity of pup syllables. (B) Distributions of durations of the chevron syllable at each age. The mean duration changes only slightly. (D) The Kurtosis of the distribution increases with age. Bars represent the standard error of the mean. (C) Distributions of the dominant frequency of the chevron syllable at each age. At p5 and p7 there are two clear peaks in the distribution, but by p9 there are fewer low frequency syllables being produced. The frequency of the higher peak gradually reduces with age. (E) Scatter plots of duration against frequency from animals aged p7 and adults. Although there is some overlap, the spectro-temporal features of adult syllables are distinct from pup syllables. :::  ::: In adults, the typical flat syllable had a peak frequency near 76 kHz and lasted for 19 ms ([Fig. 5A](#pone-0017460-g005){ref-type="fig"}). Pup syllables were either higher or lower in peak frequency, and usually longer in duration. The developmental change in duration was straightforward; younger pups produced syllables with greater variation in duration ([Fig. 5A and 5D](#pone-0017460-g005){ref-type="fig"}), but the proportions of longer syllables diminished as pups aged. The kurtosis of the distribution became more positive with age ([Fig. 5D](#pone-0017460-g005){ref-type="fig"}), showing that the mice are honing in on a more typical duration. The peak frequency of pups\' flat syllables was bimodally distributed, with the lower peak just below the adult peak and the higher peak substantially higher than in adults ([Fig. 5C](#pone-0017460-g005){ref-type="fig"}). Over pup development, the average peak frequency increased as more syllables were produced in the higher frequency band. Although there is substantial overlap in both duration and peak frequency between pup syllables and those of adults, duration vs frequency distributions reveal much less overlap. Thus, the flat syllables of p7 pups have a distinctive distribution compared to adults ([Fig. 5E](#pone-0017460-g005){ref-type="fig"}). Chevron syllables in adult mice typically had peak frequencies near 83 kHz and lasted about 40 ms. Most pup syllables were higher in frequency and longer in duration ([Fig. 6](#pone-0017460-g006){ref-type="fig"}). The predominant change in syllable duration over pup development was a decrease in the variance across syllables that is reflected in the change in the kurtosis of the distribution ([Fig. 6D](#pone-0017460-g006){ref-type="fig"}), and a slight increase in the mean ([Fig. 6B](#pone-0017460-g006){ref-type="fig"}). Adult versions of the chevron syllable were dramatically lower frequency than pup versions. Although the peak frequency of pup syllables was usually higher than in adults, chevron syllables in young pups, like flat syllables in pups, were bimodally distributed. As pups increased in age, most syllables had frequencies within the higher band and the average value of this band reduced ([Fig. 6C](#pone-0017460-g006){ref-type="fig"}). Adult chevron syllables had a unimodal distribution with less variation. Adult and pup syllables are distinct from one another in duration and dominant frequency ([Fig. 6E](#pone-0017460-g006){ref-type="fig"}). Both pup and adult chevron syllables had similar bandwidth. ### Syllable duration {#s3b1} Overall, pup syllables were significantly longer than adult syllables (pups, 52 ms \[SD 23\]; adults, 29 ms \[SD 20\]; F (1, 28384) = 5188, p\<0.001). In [Figure 7](#pone-0017460-g007){ref-type="fig"}, we compare the duration of each syllable type across ages. There were age-related differences in the durations of all syllable types apart from the short syllable (not shown). Most syllable types were generally shorter with age: the flat syllable ([Fig. 5](#pone-0017460-g005){ref-type="fig"}) (F (5, 7664) = 367, p\<0.001), the 1 freq. step syllable (F (5, 5430) = 166, p\<0.001), the 2 freq. step syllable (F (5, 4810) = 102, p\<0.001) the down-FM syllable (F (5, 1597) = 38, p\<0.001) and the complex syllable (F (5, 570) = 19, p\<0.001). A common pattern was an initial increase in syllable duration between p5 and p7 or p9, followed by a decrease as pups aged further ([Fig. 7A](#pone-0017460-g007){ref-type="fig"}). Flat, 1- and 2-freq. step, up-FM, and complex syllables followed this pattern. The durations of the low frequency harmonic and reverse chevron syllables were not compared because too few examples were recorded from pups. ::: {#pone-0017460-g007 .fig} 10.1371/journal.pone.0017460.g007 Figure 7 ::: {.caption} ###### Developmental changes in syllable duration. The means and standard errors of the durations are shown. (A) These syllables showed a progressive decrease in duration over development. The average durations across all syllables (*dashed black line*) reflect the pattern of these individual syllable types. (B) These syllables had more complex patterns of duration change across age. Age-dependent durations of the short syllable are not shown, as the criterion for classifying it was duration dependent. :::  ::: It is noteworthy that some syllables did not follow the trend described above ([Fig. 7B](#pone-0017460-g007){ref-type="fig"}). For example, the duration of the chevron syllable increased significantly at each age ([Fig. 6B](#pone-0017460-g006){ref-type="fig"}) (p\<0.001), with the exception of syllables from p11 animals. Chevron syllables from p13 animals were significantly longer than those from all other pup ages (mean = 61.9 ms, 95% CI \[60.5, 63.3\]) p\<0.001). Further, while the variability in duration often decreased with pup age, that was not always the case (e.g., complex syllables). ### Syllable frequency {#s3b2} Several spectral features of syllables changed between pups and adults. On average, pup syllables were significantly higher in frequency than adult syllables (F (5, 28385) = 261, p\<0.001). Moreover, when all syllables were pooled, there was an overall reduction in syllable frequency with age with the exception of p13, where there was a jump back to a frequency similar to those of p5 animals ([Fig. 8A](#pone-0017460-g008){ref-type="fig"}). ::: {#pone-0017460-g008 .fig} 10.1371/journal.pone.0017460.g008 Figure 8 ::: {.caption} ###### Changes in frequencies for each of several syllables, and for all syllables (*dotted black line*). \(A) Dominant frequencies declined somewhat with pup age. (B) Syllable types that showed an increase in dominant frequency with pup age. In all cases, dominant frequency declined between p13 pups and adults. (C) The dominant frequencies of the three components of the 2 freq. step syllable; the stepped pattern was reversed for adult animals. :::  ::: Although each syllable type exhibited age-related differences in spectral features, these changes were often not consistent with the general pattern ([Fig. 8B](#pone-0017460-g008){ref-type="fig"}). For instance, while the stepped syllables generally decreased in frequency with pup age, the peak frequency in flat and complex syllables increased. The peak frequency of the chevron syllable increased and then decreased by age p13. These observations suggest that no single mechanism is responsible for changes in syllable frequency, and they emphasize the importance of analyzing each syllable type separately. Further inspection revealed the limitations of focusing on a single population measure (e.g., the mean peak frequency) to characterize age-related changes. Thus, [Figure 5C](#pone-0017460-g005){ref-type="fig"} shows that the peak frequency of flat syllables had a bimodal distribution, with a lower frequency peak at 75 kHz and a higher peak at 95 kHz. Developmental changes in the mean value represented a shift in the population from one to the other of the modes. Thus, the p5 distribution is well balanced, whereas at p7 the lower frequency syllables are more common. Conversely, at ages p9, p11, and p13, the higher frequency syllables are more common. These age-related differences are reflected in the *median* dominant frequency at each age (p5 = 73.2; p7 = 70.2; p9 = 98.6; p11 = 93.7; p13 = 91.7 and adult = 77.6). Like the flat syllable, the chevron syllable showed a bimodal distribution of peak frequency, particularly at ages p5 and p7 ([Fig. 6](#pone-0017460-g006){ref-type="fig"}). At p5 there were peaks near 70 kHz and 100 kHz, with a larger proportion of higher frequency syllables (mean = 90.1 kHz; median = 108.3 kHz). By p7 there were very few syllables in the lower frequencies (mean = 92.4 kHz; median = 103.0 kHz). From ages p9 onward, the decrease in peak frequency was the result of a decrease in the mean value of the higher frequency distribution (F (5, 4071) = 106; p\<0.001). Bonferoni post-hoc tests showed that the mean dominant frequencies of chevron syllables from adult animals were significantly lower in frequency than pup syllables by at least 10 kHz (mean = 83.2 kHz, 95% CI \[82.7, 83.6\], p\<0.001). These differing patterns suggest a refinement of vocal frequency rather than modifications of the vocal tract. The frequency step size between the two components of the 1 freq. step syllable did not change as a function of age (mean = 30.4 kHz, 95% CI \[30.0, 30.8\]). Within pups the dominant frequency of the lower frequency component reduced significantly as a function of age ([Fig. 8A](#pone-0017460-g008){ref-type="fig"}) (F (5, 5259) = 47; p\<0.001). In adult mice this component had similar frequencies to younger pups. A similar pattern of frequency change was observed in the high frequency component ([Fig. 8A](#pone-0017460-g008){ref-type="fig"}). The most striking difference in the 2 freq. step syllable is that the direction of the frequency step differs between pups and adults ([Fig. 8C](#pone-0017460-g008){ref-type="fig"}). In pups the syllable starts at high frequency, steps to a lower frequency and then steps back up to higher frequency; the adult version starts at a low frequency, steps to a higher frequency then steps back to a lower frequency. Within pups, each component of the 2 freq. step syllable reduced steadily in dominant frequency with age ([fig. 8C](#pone-0017460-g008){ref-type="fig"}). The first frequency step, between component 1 and component 2, was of similar magnitude amongst pups (roughly +31 kHz). However, within the adult version the step was downward and significantly smaller (mean = −15.1 kHz, 95% CI \[−19.7, −10.4\], p\<0.001). The second frequency step was of similar magnitude amongst pups and adults. Spectral analysis was not undertaken on the up-FM syllable at p5 as it was rarely produced at that age. Both the start end the end frequency of the up-FM syllable changed with age, with adult syllables starting at significantly lower frequencies than pup syllables (F (5, 1895) = 397, p\<0.001). The upper frequencies of the up-FM syllables were significantly lower in adult animals than any pup age (M = 84.4 kHz, 95% CI \[84.1, 84.7\], p\<0.001). The bandwidth of these syllables did not change as a function of age. The same pattern was observed for the down-FM syllable; adult syllables started and ended at lower frequencies than pup syllables and there was no overall change in the bandwidth ([Fig. 8A](#pone-0017460-g008){ref-type="fig"}). Surprisingly, the dominant frequencies of many pup syllables were above 100 kHz, the proposed upper limit of adult mouse hearing [@pone.0017460-Muller1]. These frequencies are also above the range of mouse pup hearing; pups are deaf until p10 [@pone.0017460-Ehret1] and cannot hear frequencies above 50 kHz until after p14 [@pone.0017460-Ehret1], [@pone.0017460-Romand1]. At p5, 30.1% (946/3145) of syllables had dominant frequencies above 100 kHz. This was still high at p11, where 13.8% (629/4560) of syllables had dominant frequencies above 100 kHz ([Fig. 9](#pone-0017460-g009){ref-type="fig"}). A smaller number of pup syllables, between 6% and 11% depending on the age group, had all of their energy above 100 kHz. These completely inaudible syllables would be perceived by adults as silent intervals within the bout, changing its temporal properties. Syllables with dominant frequencies above 100 kHz were rare in adult animals, 0.6% (43/6963), and no adult syllables had energy exclusively above 100 kHz ([Fig. 9](#pone-0017460-g009){ref-type="fig"}). ::: {#pone-0017460-g009 .fig} 10.1371/journal.pone.0017460.g009 Figure 9 ::: {.caption} ###### Syllables with dominant frequencies above 100 kHz are common in pups but not adults. :::  ::: ### Harmonic structure {#s3b3} Most pup syllables and nearly all adult syllables were tonal, characterized by a single harmonic element ([Fig. 2](#pone-0017460-g002){ref-type="fig"}, all syllables marked "A"). In contrast, some pup syllables had either additional harmonics or nonlinear components. [Figures 1](#pone-0017460-g001){ref-type="fig"} and [2](#pone-0017460-g002){ref-type="fig"} display several examples of harmonic syllables, with a second harmonic in the 100--200 kHz range (e.g., [Fig. 1](#pone-0017460-g001){ref-type="fig"}, all p9 syllables; [Fig. 2](#pone-0017460-g002){ref-type="fig"}, all syllables marked "B"). The two main types of nonlinearities were subharmonics, composed of frequency components that are integer fractions of the fundamental frequency, and deterministic chaos, composed of broadband noisy components (see [@pone.0017460-Fitch1]) ([Fig. 2](#pone-0017460-g002){ref-type="fig"}, all syllables marked "C"). Across pup age, there were significant age-related changes in the proportions of these syllables ([Fig. 10](#pone-0017460-g010){ref-type="fig"}). In particular, nonlinear syllables became progressively more common, increasing from 0.3% in p5 animals to 23.5% in p13 animals. Very few adult syllables were harmonic (1.7%, 121/6815) or had nonlinearities (1.6%, 106/6815). ::: {#pone-0017460-g010 .fig} 10.1371/journal.pone.0017460.g010 Figure 10 ::: {.caption} ###### Harmonic and non-linear syllables are common in pups but not adults. With increasing pup age the proportion that had nonlinearities changed significantly (X^2^ (5, *N* = 2416) = 1417, *p*\<0.001); increasing steadily developmentally. There were also significant changes in the proportions of harmonic syllables (X^2^ (5, *N* = 7672) = 1091, *p*\<0.001). Adult mouse syllables were almost always tonal, the proportions of tonal syllables changed significantly over development (X^2^ (5, N = 17915) = 1737, p\<0.001). :::  ::: Features of vocal bouts {#s3c} ----------------------- Mouse syllables were produced in bouts that could be composed of either the same syllable or a combination of different syllables. A bout was defined to include at least 3 syllables produced successively with silent intervals less than 1569.8 ms. This criterion, based on our analysis of all intervals less than 5000 ms, for all ages, is 2 standard deviations above the mean (mean = 337.8, SD = 616.0). Using this criterion, we analyzed 23608 syllables from 1337 bouts. We examine temporal features of the bouts, then the sequential ordering of syllables within a bout. ### Temporal features of bouts {#s3c1} We compared inter-call interval (the silent interval between syllables), bout duration, and number of syllables in a bout across ages. The number of syllables in a bout decreased as a function of age (F (5, 1337) = 19, p\<0.001). At p5, syllable bouts were comprised of an average of 24 syllables (95% CI 19, 26), decreasing to 15 syllables per bout at age p13. Adult bouts comprised of fewer syllables per bout than those of any pup age group (mean = 11, \[95% CI 9, 12\], p\<0.001). The mean inter-syllable interval decreased significantly and substantially as a function of age (F (5, 23608) = 232, p\<0.001). Because the distributions of syllable intervals were skewed towards shorter intervals, we describe this change using median values. The median inter-syllable interval was longest at p5 (190.2 ms), decreasing at each successive pup age except p11 (median values: p7, 149.1 ms; p9, 93.4 ms; p11, 96.5 ms; p13; 84.8 ms). The intervals between adult syllables were substantially shorter than those of pups of every age (median value, 69.4 ms). ### Bout structure: probability analysis {#s3c2} Bouts of syllables typically comprised 3--4 syllable types in both pups and adults (mean = 3.8, \[95% CI 3.7, 3.9\]). We examined whether the occurrence of syllables within bouts followed non-random patterns, and whether patterns changed over development. Note that some developmental change in patterns is expected even if syllables are randomly distributed within bouts, since different syllable types have greater probabilities of being produced by animals of different ages. To analyze whether non-random patterns occurred, we examined pairings of syllable types using a two-syllable model (e.g. Flat-to-Chevron, Chevron-to-Chevron etc.). This approach tested whether the number of occurrences of syllable pairings differed from the number predicted by a random model. Syllable combinations were examined for all bouts with more than 3 syllables (n = 1337 bouts). The two-syllable model showed that the sequence of syllables within a bout was not random (Χ^2^ (9, *N* = 2039) = 10511, *p*\<0.001). We next compared the probabilities of switching between syllable types within a bout of syllables, across ages. With increasing age, the probability that animals switched between syllable types increased steadily from 0.43 at p5 to 0.63 by p13 (see [Fig. 4B](#pone-0017460-g004){ref-type="fig"}). The oldest pups, at p13, were similar to adults in their high probabilities of switching between syllable types within a bout. The increased probability of switching back and forth between those syllable types resulted in a more complex sequence of syllables being produced by older animals. The transition with the highest associated probability changed developmentally. At both p5 and p7 the most probable transition was from a 1 freq. step syllable to a flat syllable; at p9 and p11, this changed to transitions from a 2 freq. step syllable to a 1 freq. step syllable. At p13 the most common transition was from a 2 freq. step syllable to a chevron syllable. In adult animals the most probable transition was from a 1 freq. step syllable to an up-FM syllable. Bouts were most likely at every age to begin with a flat syllable, even though this was not the most common syllable type produced at each age. Flat syllables made up only 26% of adult syllables, yet the probability of a flat syllable being produced at the start of a bout was higher, at 0.46. The same pattern was observed for pups of all ages; there was a probability of between 0.32 and 0.64 of a flat syllable occurring first in a bout. The most probable 10-syllable bouts are shown in [Table 2](#pone-0017460-t002){ref-type="table"}. The first syllable was computed based on the probability of that syllable occurring first in a bout. The second syllable is the syllable type that had the greatest transitional probability of following the first syllable type. Each successive syllable was the syllable type that had the highest transitional probability of following the previous two syllables, based on third-order transitional probabilities. ::: {#pone-0017460-t002 .table-wrap} 10.1371/journal.pone.0017460.t002 Table 2 ::: {.caption} ###### Most probable sequences of syllables in 10-syllable bouts at each age, based on third-order transitional probabilities. ::: {#pone-0017460-t002-2} Syllable type at each sequential position ----------- ------------------------------------------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- **p5** Flat Flat Flat Flat Flat Flat Flat Flat Flat Flat **p7** Flat 1F. step Flat Flat Flat Flat Flat Flat Flat Flat **p9** Flat Chevron Chevron Chevron Chevron Chevron Chevron Chevron Chevron Chevron **p11** Flat 2F. step 2F. step 2F. step 2F. step 2F. step 2F. step 2F. step 2F. step 2F. step **p13** Flat Flat Flat Flat Flat Flat Flat Flat Flat Flat **Adult** Flat Flat Up Flat Up Flat Up Flat Up Flat ::: ### Bout structure: information theory analysis {#s3c3} The two-syllable model described in the preceding section showed that the sequence of syllables within a bout was not random. This shows that at every age tested there was some sequential structure, in which the preceding syllable had an effect on what syllable type was likely to follow. We used information theory to compare the level of sequential organization across ages at the two-, three- and four-syllable levels. The overall entropy reduced with increasing entropy level ([Fig. 4C](#pone-0017460-g004){ref-type="fig"}); the negative entropic slope is indicative of higher-order structure within the sequencing of syllables. The entropy reduced significantly between H~1~ (first order) and H~2~ (second order) at all pup ages, and in adults (Χ^2^ (7--10, *N*\>3143)\>2059, *p*\<0.001). This supports the findings of the two-syllable model. The entropy also reduced significantly between H~2~ and H~3~ (second to third order) at all pup ages and in adults (Χ^2^ (49--100, *N*\>3143)\>1204.19, *p*\<0.001), showing that at all ages the probability of a particular syllable type being produced is affected by the two preceding syllable types. There was a further reduction of entropy between H~3~ and H~4~, this difference was not significant. Thus, at each age, bouts appear to be organized up to at least the three-syllable level. A comparison of the degree of entropic change between H~1~ and H~3~ across development revealed an increase in the level of sequential structure. There was no difference in entropic change between p5 and p11 pups, but there was a significant difference between p11 and p13 pups. We also compared the overall difference in entropic change between pups younger than p13 and adults, and between p13 pups and adults; adults showed a greater degree of higher-order structure than pups of any age. There was an increasing degree of higher-order structure as pups developed and from pups to adults. This was evident at both the two-syllable sequence level, H~1~ to H~2~, and the three-syllable sequence level, H~2~ to H~3~. There was a significantly larger reduction in entropy between H~1~ and H~2~ in p13 pups than in p5 pups (Χ^2^ (64) = 499, p\<0.001), and between p13 pups and adults (Χ^2^ (100) = 6256, p\<0.001). This difference was not evident between p5 and p11 pups. There was also a significant reduction in entropy developmentally between H~2~ and H~3~, reflecting a greater degree of organization at the three-syllable level. The entropy change between H~2~ and H~3~ increased between p5 and p13 (Χ^2^ (576) = 1347, p\<0.001), and from pups younger than p13 to adults (Χ^2^ (1100) = 1295, p\<0.001). Adults had 0.40 bits more organization at the two-syllable level than p13 pups (Χ^2^ (64) = 5627, p\<0.001), however adults had no greater organization at the three-syllable level than p13 pups. Virtual Mouse Vocal Organ {#s3d} ------------------------- The preceding sections described many developmental changes in the vocalizations of mice. We used these developmental changes to implement a probabilistic Markov model in order to generate bouts of syllables that have features appropriate for animals of different ages (instructions, program and associated sound files in [Program S1](#pone.0017460.s001){ref-type="supplementary-material"}). [Figure 11](#pone-0017460-g011){ref-type="fig"} shows spectrograms of bouts of 20 syllables generated using the third order Markov model from p5, p9, p13, and adult mice. The bouts become dramatically shorter with development, as a result of overall decreases in both the duration of the individual syllables and the inter-syllable interval. The decrease in variability between the average frequencies of the individual syllable types is also evident. Note that some syllables, especially of younger pups, have significant or all energy above 100 kHz. These syllables would likely change the perceived temporal structure of bouts, as adult mice only hear frequencies up to approximately 100 kHz. ::: {#pone-0017460-g011 .fig} 10.1371/journal.pone.0017460.g011 Figure 11 ::: {.caption} ###### Spectrograms of syllable bouts generated by the Virtual Mouse Vocal Organ. Syllable bouts correspond to mice aged p5, p9, p13 and adult. :::  ::: Discussion {#s4} ========== Our investigation into the developmental changes in mouse vocalizations combined several experimental features: examination of developmental changes for each of several syllable types, recording of spectral components in syllables above 100 kHz, and use of quantitative methods to examine the organization of mouse syllable sequences. This approach produced several key findings. 1) The CBA/CaJ mouse has 11 syllable types, 10 of which are produced by both pups and adults. The proportions of these syllable types vary developmentally. 2) Developmental changes in spectral and temporal features differ among syllable types. 3) The temporal characteristics of bouts changes with age, with older pups and adults having shorter intervals between syllables and fewer syllables within a bout. 4) The vocal repertoire becomes more diverse developmentally, with the overall complexity and higher-order structure of bouts increasing. Overall, this work characterizes the differences within mouse vocal communication that could be used by adult mice to determine the age of the vocalizing animal. The combination of age-dependent changes in the structure of vocalizations and the temporal sequencing of syllables was used to create age-appropriate bouts using the Virtual Mouse Vocal Organ. Developmental changes in syllable types {#s4a} --------------------------------------- We observed all syllable types described in previous work in pups and adults [@pone.0017460-Liu1], [@pone.0017460-Scattoni1], [@pone.0017460-Portfors1], [@pone.0017460-Bell1], [@pone.0017460-Branchi1]. In addition to the flat calls observed previously in CBA/CaJ pups [@pone.0017460-Liu1], we recorded several syllable types with frequency transitions (see also Portfors [@pone.0017460-Portfors1]). Our study shows that each of the types observed in adults also occurs in pups, with the exception of the noisy syllable. Scattoni and colleagues [@pone.0017460-Scattoni1] reported that at p8, the 2 freq. step was the most common syllable type produced by three commonly used mouse strains (30%, C57BL/6J; 41%, 129X1; and 43% FVB/NJ); this is in agreement with our finding that the 2 freq. step syllable was the most common at p9 in CBA/CaJ mice, representing 31% of syllables. These results suggest that even young mouse pups have well developed vocal production mechanisms which may be common across mouse strains. The proportions of syllable types change over pup development. In p5 animals, most syllables were flat, 1 freq. step, or down-FM syllables. As pups age, they produced a greater variety of syllables, with generally more complex features. The changing proportions of different syllable types could be used as cues for mothers to distinguish pups on the basis of age. This change continues in juvenile mice, since syllable type distribution is quite different between p13 and adult animals. Developmental changes in spectro-temporal features of syllables {#s4b} --------------------------------------------------------------- Our study corroborates results from previous studies showing decreases in the duration and frequency of adult syllables compared to those of pups [@pone.0017460-Liu1], [@pone.0017460-Scattoni1], [@pone.0017460-Hahn1], [@pone.0017460-Motomura1]. When we analyzed each syllable type separately, however, we found different patterns of developmental change. For some syllables, these patterns suggested a gradual change in the direction of the adult feature, but for other syllables or features, there was no such change. ### Syllable frequency {#s4b1} Across all types of syllables, the frequency of syllables did not change in a systematic way with pup age. The average dominant frequency reduced steadily from p5 to p11, however after the onset of hearing, the dominant frequency of pup syllables returned to a value similar to p5 animals. This pattern was not reflective of all syllable types. Viewed individually, syllable types show different patterns of change in frequency. The complex syllable types had dominant frequencies that increased steadily with pup development, whereas the mean dominant frequency of the chevron syllable changed very little across development. The flat syllable showed a more complex pattern of change, as frequency formed a bimodal distribution. As development progressed, pups syllables more often fell within the higher frequency peak of the distribution, resulting in dominant frequencies that were above those of adult syllables. Many calls from younger animals had the majority, if not all, of their energy at frequencies above 100 kHz, the proposed upper end of the mouse hearing range [@pone.0017460-Muller1]. Pups of all ages produced many high frequency calls over 100 kHz, whereas adult animals did not produce any calls entirely above 100 kHz. It is possible that these calls reflect a lack of vocal control, though this seems unlikely because they were able to produce versions of each syllable type that are well within the mouse audiogram. Puzzling in these patterns is the fact that many features change over the pup ages prior to hearing onset. Mouse pups are deaf until p10 [@pone.0017460-Ehret1] and cannot hear frequencies above 50 kHz until p14 [@pone.0017460-Ehret1], [@pone.0017460-Romand1], thus they lack the acoustic feedback to recognize that these calls are inaudible. The changes we see could reflect increased motor control, although such explanations would need to account for changes in different directions for different types of syllables. Another factor that may underlie changing spectral features is selection by mothers. For these features, a more complete understanding of changes across development, including what occurs in juvenile animals, will help to assess the underlying mechanisms and functional significance. ### Harmonic structure {#s4b2} Within pups, there was a steady increase in the proportion of syllables containing nonlinearities, from 1% at p5 to 23% at p13. However, in adults, there were virtually no syllables with non-linear spectral features. The increase in nonlinearities with pup age could relate to their functional properties in communication with mothers, or may be part of a developmental trajectory of the vocal organs. Our view below is that these nonlinearities are functionally important for older pups. However, any resolution of this issue requires an understanding of the developmental profile between p13 animals and adults. Nonlinear components such as subharmonics, frequency jumps and deterministic chaos have been described in the vocalizations of several species, both vocal learners (zebra finch: [@pone.0017460-Fee1]) and non-vocal learners (rhesus macaque:[@pone.0017460-Fitch1], pig: [@pone.0017460-Tokuda1]; frog *Amolops torotus*: [@pone.0017460-Suthers1]). Several evolutionary benefits have been proposed for the use of nonlinearities within vocalizations. The acoustic variability they generate within calls is thought to enhance caller recognition in African wild dogs [@pone.0017460-Tooze1]. The formant frequency of vocalizations is used to ascertain the size of the caller in primates [@pone.0017460-Fitch2], [@pone.0017460-Rendall1], [@pone.0017460-Fitch3]. Primates potentially incorporate nonlinear phenomena so that their vocalizations seem to have a lower formant frequency, thus making the caller seem larger and audible at a greater distance [@pone.0017460-Fitch1]. We propose four plausible evolutionary reasons why producing non-linear pup vocalizations may be advantageous for mice. First, nonlinear components within syllables often make a sound lower frequency without reducing its overall sound level. In mice, smaller animals are often killed or neglected by the mother in favor of the larger ones [@pone.0017460-Gandelman1], making it beneficial for the pup to appear larger. Second, broad band sounds are easier to locate than tonal sounds [@pone.0017460-Marler1], thus the production of nonlinear calls may have developed to enhance the probability of the mother finding a pup when it has been removed from the nest. Third, nonlinear components may convey salience or emotional state that may be appropriate to isolation states in pups. Adult mice still produce nonlinear calls, though they are rare. It may be that nonlinear calls are a characteristic that adults associate with juvenile animals. Fourth, non linear components would generate mechanical distortion products in the cochlea at lower frequencies that would be audible by pups, thus pups may make many of these calls at p11 and p13 because they are self-audible. For example the non-linear 2 Freq. step syllable in [Figure 2](#pone-0017460-g002){ref-type="fig"} would generate strong distortion product near 50 kHz and other, less strong, distortion products at lower frequencies (24--36 kHz). These are likely to be audible by pups above p10. ### Temporal properties {#s4b3} A major difference between pup syllables and adult syllables is their duration [@pone.0017460-Liu1], [@pone.0017460-Scattoni1], [@pone.0017460-Hahn1], [@pone.0017460-Motomura1]. Pup syllables are on average 52 ms, while adult syllables are on average 29 ms. For some syllables, including the complex and the down-FM syllables, most of the decrease in duration occurs over pup development, with less or no change between p13 animals and adults. For other vocalizations (e.g., chevron and the 2 freq. step syllables), the major change occurs between p13 animals and adults. Still other syllables do not change substantially in duration over the times studied here (e.g., up-FM and short syllables). A major change across age, however, seems to be that older pups and adults produce an increasingly uniform distribution of syllable durations. For duration more than any other feature of syllable types, pups seem to be honing in on a particular duration across development, although how and when this occurs clearly varies for different syllable types. ### Overall view {#s4b4} Viewed across the vocal repertoire of pups or adults, there are a variety of acoustic features of syllables, from frequency to duration to frequency-time properties, that permit adult animals to distinguish pups of different ages and pups from adults [@pone.0017460-Liu1], [@pone.0017460-Scattoni1]. We also recognize that pup syllables may be optimized to activate retrieval behavior by mothers, and so may undergo selection pressures and developmental trajectories that may be different than for adult syllables. To clarify these issues, it will be important to further characterize development of syllables in juveniles, and to test whether syllables develop similarly under different pup or adult hearing capabilities. Developmental changes in bout structure {#s4c} --------------------------------------- There are age-related differences in the temporal patterns of bouts, with younger animals producing longer syllables with longer inter-syllable intervals. A greater repetition rate may convey information related to size of the caller. Syllables may be produced in bouts because sounds with sudden temporal breaks are easier to locate; arrival times at the two ears are easier to detect for sounds in a repetitious temporal pattern [@pone.0017460-Marler1]. We found bouts were most likely, at every age, to start with a flat call. This is similar to the introductory notes characterized in zebra finch song, in which some notes are most likely to be produced first in a song [@pone.0017460-Zann1]. ### Complexity of the vocal repertoire {#s4c1} If an animal\'s vocal repertoire is too unified (repetitious), then there is little communicative complexity, as a highly repetitious sequence of syllable types conveys only a little information. Conversely, if a repertoire is too diverse, or randomly distributed, the same message could be represented in multiple ways, resulting in less information being communicated by any one syllable type [@pone.0017460-Zipf1], [@pone.0017460-McCowan1]. The principle of Zipf\'s law [@pone.0017460-Zipf1] is that communication has an optimal balance between unification and diversification. The majority of human languages have an optimal balance within the proportions of word usage between unity and diversity with slopes of around −1[@pone.0017460-Zipf1]. Dolphin whistles become closer to this balance developmentally, achieving a value in adult dolphins that is close to human language, −0.95. Their developmental change in the slope is thought to reflect vocal learning. However age dependant differences in the Zipf slope have also been described in the squirrel monkey, an non-vocal learning species [@pone.0017460-Zipf1]. In both humans and bottlenose dolphin the repertoire begins as more diverse, becoming more repetitious over development. However, our data show changes that are more like the squirrel monkey, where the repertoire starts out overly repetitious and becomes more diverse across development. The same proportion of syllable types can be composed within a bout of syllables with different levels of complexity. For example, the two sequences {A, A, A, A, B, B, B, B, C} and {A, B, C, B, B, A, A, B, A} have the same proportions of letters, however the overall complexity of the first sequence is less than the complexity of the second sequence. We rated the complexity of bouts of syllables by calculating the probability of transitions between syllable types occurring. As the animals aged, the probability of a switch between syllable types increased from 0.5 at p5 to 0.7 in p13 pups and adults, resulting in increasingly complex sequences. Finally, adult mouse vocalizations had significantly more higher-order sequential organization than p13 pup vocalizations, implying that the rules of the vocal system are still not fully developed by p13 and that more sequential learning may take place throughout adolescence. ### Functional significance {#s4c2} Duration and frequency of sounds are critical for evoking pup retrieval behavior from the mother [@pone.0017460-Ehret4], [@pone.0017460-Ehret5]. Maternal mice respond with retrieval behavior to strings of pure tones at frequencies similar to pup calls, almost as well as they do to strings of complex pup calls [@pone.0017460-Ehret4]. If a pup need only produce flat tonal signals in order to evoke retrieval by its mother, why does it produce such complex strings of syllables? Pups may increase the complexity of their bouts to make them distinct from those of littermates. In younger animals where the vocal repertoire is more repetitious, it is likely that the calls from one pup will be masked by the calls of other pups in the litter that are making similar sounds, making it harder for the mother to localize a specific pup. Pups may learn to produce more complex sequences to counter this masking and thus increase the likelihood of being localized, and retrieved, by their mother. A similar mechanism for the release of background masking has been reported in king penguin communication. Playback studies conducted in a quiet environment demonstrate that frequency of a syllable is the critical feature for retrieval. However, in an environment with background noise similar to the natural environment, the syllabic organization facilitates recognition [@pone.0017460-Aubin1]. Our results are consistent with vocal learning in mice, but do not provide sufficient supportive evidence. The changes we describe could result from vocal learning, developmental changes in motor control, or changes in the vocal tract. Vocal learning issues could be addressed by employing the analytic methods used here, comparing the vocalizations of normally hearing mice with those deafened prior to hearing onset and those raised by deafened dams. The virtual mouse vocal organ {#s4d} ----------------------------- We found several developmental changes within mouse syllables and sequences that could be used as acoustic cues for adults to determine the age of the caller. However, it is still unclear which of these cues are used by adult mice. We developed a virtual mouse vocal organ that creates appropriate syllable sequences based on age-related changes in several features that we measured: frequency, bandwidth, duration, inter-syllable interval, syllable probability and the sequential pattern (transitional probability). This virtual mouse vocal organ consists of a third order Markovian probabilistic model of CBA/CaJ bouts of syllables. The model draws on the most typical version of each syllable type corresponding to each age tested and produces bouts of syllables in wav format. This can be used to generate well-controlled experimental stimuli for behavioral or neurophysiologic research (see attached MATLAB program). Supporting Information {#s5} ====================== Program S1 ::: {.caption} ###### This compressed folder holds the \'Virtual Mouse Vocal Organ\' program. The instructions for running the program are included along with the MATLAB code and a library of the most typical version of each syllable type produced at each age. (RAR) ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to thank Dr Shobhana Sivaramakrishnan for kindly allowing us to record from her mouse pups, Rachel Anderson and Matthew Wenstrup for their help with syllable analysis, and Dr. Merri Rosen and Marie Gadziola for their helpful comments on the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The work was supported by NIH R01-DC000937-20 and DC000937-18S1. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JG JM JW. Performed the experiments: JG. Analyzed the data: JG JM JW. Contributed reagents/materials/analysis tools: JG JM JW. Wrote the paper: JG JM JW.

PubMed Central

2024-06-05T04:04:19.202776

2011-3-9

PMC3052363

Introduction {#s1} ============ For centuries, free-will and self-consciousness have been striking subjects of interest in various sectors of society. Within this framework, distinguishing oneself from others appears to be crucial for social interactions. However, this ability is far from absolute, since it may be impaired in certain pathological conditions, such as in schizophrenia, in which patients regularly fail to identify their own actions or thoughts, by over-attributing them to external sources [@pone.0017500-Mellor1], [@pone.0017500-Spence1]. In fact, a growing body of evidence suggests that the capacity to experience oneself as the agent of one\'s own actions (agency) might concur, along with the sense of body-ownership, with self-identification in general [@pone.0017500-vandenBos1]. Importantly, in daily-life activities, we do not constantly think about the source of our perceptions, and it could be postulated that this moment-to-moment selfhood experience may be mainly linked to non-conscious cognitive processes, which only manifests itself when a situation requires it. One of the most influential models for understanding the nature of brain processes involved in distinguishing oneself from other is the central monitoring theory (CMT). In the CMT, the agent is determined by online monitoring of the degree of congruence between internally-generated predictions and peripheral signals, generated by an action [@pone.0017500-Frith1], [@pone.0017500-Feinberg1], [@pone.0017500-Frith2]. Some neuroimaging studies in the field of visuo-motor agency suggest that the temporo-parietal junction (TPJ), and especially the inferior parietal lobule (IPL), plays a key role in disentangling the origin of sensory events [@pone.0017500-Farrer1], [@pone.0017500-Ruby1], [@pone.0017500-Blanke1]. However, the CMT is unable to account for agency without execution of an action, as in mental imagery, in which an action is represented by the motor system, without being executed [@pone.0017500-Decety1], [@pone.0017500-Buccino1]. An alternative hypothesis particularly useful for resolving this problem is proposed by the shared-representation model (SRM). According to the SRM, the motor system simulates an action, by whomever the agent is (oneself or somebody else), and these two modalities of action representations share the same neural structures [@pone.0017500-Decety2], [@pone.0017500-Georgieff1]. This theory is supported by the existence of mirror neurons which were first shown in the Macaque monkey brain. Mirror neurons discharge during both execution and observation of an action [@pone.0017500-Gallese1], [@pone.0017500-Rizzolatti1]. A Mirror Neuron System (MNS), similar to that of the monkey, has been found in humans and may be involved in many higher motor functions, from coding intended actions [@pone.0017500-Iacoboni1] to language processing [@pone.0017500-Jardri1]. Interestingly, the previously mentioned right IPL, has also been shown to be involved in the MNS [@pone.0017500-Chong1], nevertheless separate areas for mirror responses and agency were identified within the parietal region [@pone.0017500-Agnew1]. This result is consistent with the idea that the MNS does not directly provide by itself, a representation of self. Consequently, the SRM assumes that the overlap between neural activations, resulting from both of these forms of action representation, is incomplete. Thus, the most suitable explanation to account for the differentiation between oneself and another agent, is that it occurs on the basis of non-overlapping areas [@pone.0017500-Georgieff1], such as in the IPL. Our team recently developed and validated an fMRI paradigm in healthy subjects, which allows shared representations to be examined during situations with speech exchange [@pone.0017500-Jardri1]. In order to suppress potential contaminations of the task by means of body-ownership sensations commonly observed in visuo-motor agency tasks [@pone.0017500-Tsakiris1], we assessed the neural substratum of self-other distinction in the language domain. We used two types of sensorimotor verbal conditions: covert speech production and passive listening to somebody else\'s voice. This procedure showed in healthy subjects that the self-other distinction relied on the activity modulation of a medial fronto-parietal network and the right IPL. While another sub-part of the IPL is recruited more during the encoding mirror response, we also checked that these areas were not activated simply because the subject recognized the voice of the person speaking [@pone.0017500-Jardri1]. As some authors proposed to directly relate language to nuclear symptoms of schizophrenia [@pone.0017500-Crow1], it seems particularly important in this disorder to test the self/non-self distinction in the verbal domain. In the present study, we used the previously described procedure to investigate the neural substrates of the self-other distinction in schizophrenia. Schizophrenia is a particularly devastating disorder, in which thought insertions, delusions of alien control and auditory-verbal hallucinations constitute "first-rank symptoms": FRS [@pone.0017500-Schneider1]. We have already mentioned that these symptoms may be related to an incapacity to recognize their own thoughts and actions as being internally generated, but instead attribute them to alien entities [@pone.0017500-Waters1]. Interestingly, the temporo-parietal junction has been shown to be involved in this type of process in schizophrenia. Using a PET-scan, Spence et al. observed that schizophrenic patients experiencing delusions of control during testing, showed abnormally high activity in the parietal region, when making voluntary movements [@pone.0017500-Spence1]. Using fMRI in a task that revealed any discordance between hand movements being performed by the subjects, and sensory feedback displayed on a computer screen [@pone.0017500-Farrer1], other authors have noted an abnormal relationship between the degree of control of the movements, and activity in the right IPL in patients with schizophrenia [@pone.0017500-Farrer2]. However the vast majority of experiments conducted in schizophrenia, only explored explicit self-other distinction, based on attribution judgment paradigms [@pone.0017500-Farrer2], [@pone.0017500-Franck1], [@pone.0017500-Allen1], [@pone.0017500-Johns1], [@pone.0017500-Lindner1]. These studies were insufficient to clearly determine if external misattributions by patients result more from a perceptual experience disorder, or from an impairment of judgment itself, which would be more linked to a thought disorder [@pone.0017500-Lafargue1]. In such a basic context, it seems important to test self-other distinction at a lower level of processing, that is to say when no explicit distinction between self and non-self is required. Our main hypothesis was that distinguishing between internally and externally-generated stimuli in schizophrenia would rely on a low-level impairment that might be related to an increased overlap between self/other neural representations. Our study would also make it possible to determine the exact role of the IPL in patients suffering from schizophrenia, during the self-other distinction process. Methods {#s2} ======= This research was approved by the local ethical committee (Comité Consultatif de Protection des Personnes se prêtant à une Recherche Biomédicale de Lille, CCPPRB Lille, n°CP 06/52). Participants {#s2a} ------------ 30 right-handed participants, according to the Edinburgh laterality test [@pone.0017500-Oldfield1], who spoke French as their first language, and who gave their written consent, were included in our study. They were split in two groups: 15 patients with a diagnosis of paranoid schizophrenia according to the DSM-IV-TR classification [@pone.0017500-APA1] and 15 matched controls. Exclusion criteria for the whole population were the presence of an axis II diagnosis or another axis I diagnosis, a neurological or an Ear, Nose and Throat disorder, or a history of substance or alcohol misuse. Patients with an IQ below normal range were also excluded (WAIS-R). The main medical, social and demographic characteristics of these subjects are summarized in [***Table 1***](#pone-0017500-t001){ref-type="table"}. All of the patients had received regular doses of antipsychotic medication for at least one month prior to testing. A quantitative assessment of the symptoms was performed using the PANSS [@pone.0017500-Kay1] either one day before, or on the day of the fMRI scan. All of the patients presented with at least two FRS. Finally, even though some of them had experienced refractory hallucinations during the weeks before they were examined, none of them reported auditory or visual hallucinations during the fMRI scanning procedure. ::: {#pone-0017500-t001 .table-wrap} 10.1371/journal.pone.0017500.t001 Table 1 ::: {.caption} ###### Demographic information of the participants included in the 'implicit self-other distinction' study (n = 30). ::: {#pone-0017500-t001-1} Patients with schizophrenian = 15; mean +/− sd Healthy controlsn = 15; mean +/− sd Group comparison Significance Comments -------------------------------------- ------------------------------------------------ ------------------------------------- ------------------ -------------- ----------------------------------------------------------- Age (yr) *30.8+/−7.6* *30.1+/−6.6* *Z = 0.23* *p = 0.82* *-* Handedness-ratio (R/L) *15/0* *15/0* *NA* *NS* *-* Sex-ratio (M/F) *13/2* *10/5* *Z = 1.60* *p = 0.11* *-* Education (yrs from High school) *2.4+/−1.8* *5.9+/−1.6* *Z = 3.41* *p\<0.01* *introduced as a covariate in later statistical analyses* Illness duration (yrs) *11.2+/−5.9* *NA* *NA* *NA* *-* Antipsychotic dosage (mg/day CPZ-Eq) *683+/−462* *NA* *NA* *NA* *-* PANSS total score *82.8+/−14.6* *NA* *NA* *NA* *-* PANSS positive subscale *22.5+/−5.5* *NA* *NA* *NA* *-* *The z-scores shown in this table come from the Wilcoxon non-parametric test; sd: standard deviation; ; Yrs: years; NA: not applicable; NS: not significant; M: male, F: female subjects; R: right; L: left; CPZ-Eq: daily therapeutic equivalent dose for the antipsychotic medications, using Chlorpromazine as a reference*. ::: Stimuli {#s2b} ------- Three conditions with different stimuli were used: Self Generated Voice \<SGV\>, unfamiliar Externally-Generated Voices \<EGV\> (2 males, 2 females) and Reverse-Taped Voice \<RTV\>. During the sound recording, the speakers read Paul Eluard\'s poem "La ville de Paris renversée". Such connected-speech-stimuli were chosen because they including both semantic and prosodic information, and not just words or phonemes [@pone.0017500-Hesling1]. Their voices were recorded using a digital recorder (*Tascam TEAC DA-P1*) with a sampling rate of 44.1 kHz, and a resolution of 16 bits. One of the unfamiliar voices was altered and played backwards \<RTV\>, which disrupted the language making it incomprehensible, but kept the same frequency and prosodical properties. This transformation allowed a stimulus that the subject was unable to reproduce himself, either by means of an object, such as a musical instrument, or by his voice to be used, since it is biomechanically impossible to join together the breathed or blown syllables of this artificially reversed voice. All of the stimuli lasted for a total of 21 seconds. The amplitude of the sound files was normalized to 96%, in order to obtain a comparable amplitude for the different recordings (75 dB SPL). Experimental procedure {#s2c} ---------------------- The paradigm used in our study was designed to find the neural correlates of implicit self-other distinction in language. The participants had to perform the experiment inside the scanner. The subjects lay down with their eyes closed, wearing MR-compatible headphones, which transmitted sound stimuli, and attenuated the ambient noise of the scanner (a reduction of about 30 dB SPL) (*MRI Devices Corporation, USA*). The experiment used a block-design paradigm, consisting of an initial silent period lasting 1 minute, to allow the subjects to get used to the noise of the scanner, followed by an alternation of stimuli lasting 21 seconds, and a rest lasting 12 seconds. The alternating stimulus/rest cycle was repeated 8 times per condition. The different conditions were presented to the different subjects in a random order. The subjects were asked to repeat the poem to self and not move their mouth during the \<SGV\> condition, while listening to their own voice through the headphones at the same time. Artifacts caused by face and head movements were avoided by this covert speech procedure. During the \<EGV\> and \<RTV\> conditions, the subjects had to listen to the voices passively. The subjects were never informed about the origin of the voice heard (self or other), and a sound signal (500 msec) of a different frequency preceded the different stimuli. Moreover, giving the subjects the task to mentally repeat or pasively listen to, diverged them from making explicit self-other judgements. The implicit testing condition for self and other was achieved, since during the task, the subject had to covert speak after hearing a certain high-pitched sound signal, and was not required to use their ability to distinguish between their own voice and the somebody else\'s voice (judgment of attribution). Testing self-other distinction at an implicit level as in this study required that no behavioral measurements of the subjects\' responses were made during scanning. Two control-situations were implemented outside the scanner in order to check that participants did not perform explicit judgments about the voices heard during scanning. First, the experiment included a preliminary training step for each subject, allowing us to ensure that the explicit task of covert speech was well understood and correctly performed. We did not use the subject\'s pre-recorded voice for the \<SGV\> condition during this training step. Second, a post-session interview was systematically realized to ensure that any of the participant performed attribution judgments during scanning. The full procedure is shown in [***Figure 1***](#pone-0017500-g001){ref-type="fig"} for the condition with a high-pitched initial signal (\<SGV\>). ::: {#pone-0017500-g001 .fig} 10.1371/journal.pone.0017500.g001 Figure 1 ::: {.caption} ###### Diagram of the fMRI block-design with the expected results. In the upper left panel, a model of each experimental condition is shown: RTV (reverse-tape voice: yellow), EGV (externally-generated voice: green) and SGV (self-generated voice: red). Each block lasted 21 sec. and was repeated 8 times in a random order, each time being followed by a 12 sec. silent period. Part of the recorded voice spectrogram is shown in detail, as an example of the SGV condition, preceded by a 500 msec. high-pitched cue, which indicated that the subject had to mentally repeat what he was hearing (covert speech). The other conditions \[EGV, SGV\] were preceded by low-pitched cues, indicating that the subject should listen passively. Two contrast analyses were performed, and are shown in the right panel of the figure. C1 contrasted EGV and RTV, and revealing brain areas from the Mirror-Neuron-System. C2 contrasted EGV with SGV and revealed the cortical network involved in self-other distinction. The middle column shows known results from healthy subjects for C1 and C2 [@pone.0017500-Jardri1]. The right column lists the expected results in the patients with schizophrenia for the C1 and C2 contrasts. In our study, we hypothesized that in schizophrenics, there would be more overlap between the neural networks activated for representation of other or self (C2). This result cannot be explained by dysfunction in the mirror neuron system (C1). IFG: Inferior Frontal Gyrus/IPL: Inferior Parietal Lobule/ACC: Anterior Cingulate Cortex/PCC: Posterior Cingulate Cortex/STS: Superior Temporal Sulcus/SMA: Supplementary Motor Area/R-L: Right or left side of the brain, respectively. :::  ::: Data acquisition {#s2d} ---------------- Imaging was performed using a 1.5 Tesla MRI scanner (*Intera Achieva*, *Philips*, *The Netherlands*) with a SENSE-Head coil containing 8 elements. The T1-weighted anatomical sequence was a 3D multi-shot TFE with the following properties: 140 slices, thickness = 1.1 mm, FOV = 240 mm^2^, matrix = 256×256, TR = 8.2 msec, TE = 4 msec, flip angle = 8°, TFE factor = 192. The T2\*-weighted functional sequence was a single-shot sensitivity-encoded echo planar imaging sequence (SENSE). A run of 280 volumes was obtained with a 30-slice Fast Fourrier Echo, thickness = 4 mm, FOV = 240 mm^2^, matrix = 64×64, TR = 3000 ms, TE = 70 ms, flip angle = 90°, SENSE factor g = 1.4. Data analysis {#s2e} ------------- The functional data were pre-processed and analysed using *BrainVoyager QX v1.9* software (*Brain Innovation*, *The Netherlands*, *2008*). Images were pre-processed using slice scan time correction, 3D head motion correction, temporal high-pass filtering with 3 cycles/point, linear trend removal and 3D spatial smoothing, with a Gaussian filter of 5.0 mm. The anatomical data were submitted to an intensity unhomogeneity correction algorithm, resampled to 0.5 mm resolution and normalized in the stereotactic Talairach\'s space [@pone.0017500-Talairach1]. Head-tissue, subcortical structures and the cerebellum were then removed, to allow for advanced cortical segmentation processing. This segmentation was performed at the grey/white-matter and the grey-matter/cerebrospinal fluid boundaries, and each resulting hemisphere was submitted to a "bridge-removal" algorithm. For visualization of the statistical maps, the slice-based functional data were aligned on a high-quality 3D anatomical image. Finally, the cortical surface was reconstructed and inflated. *Cortex Based Alignment* was performed using curvature information, to improve anatomical inter-subject correspondence mapping beyond Talairach\'s transformation [@pone.0017500-Fischl1]. At a first level, *multisubject random effects analysis* was performed for each group, according to the *general linear model* \[RFX-GLM\] [@pone.0017500-Kiebel1]. We defined a factorial design, in which the predictors of interest (\<SGV\>, \<EGV\> and \<RTV\>) and of no-interest (x,y,z motion parameters in translation and rotation) were derived by convolution of an appropriate box-car waveform, with a double-gamma haemodynamic response function. During second level analysis, we compared the two groups using a *repeated measure ANOVA model*, with the participants\' group being defined as an inter-subject factor. Educational level was introduced as a covariate of no interest in this 2^nd^ level analysis. Since all subjects received the same stimuli, the conditions applied to the subjects in each group were defined as intra-subject factors (the \[EGV -- SGV\] contrast for example). This bi-factorial ANOVA model was run over all of the voxels to obtain RFX maps. The statistical maps resulting from these first- and second-level analyses, were thresholded using the *false discovery rate* approach for multiple comparison correction [@pone.0017500-Genovese1]. To investigate with greater precision the spatial differences in the activated clusters between the groups for specific contrasts, we calculated *probabilistic functional maps* using the aligned surface maps previously produced. In order to achieve the macro-anatomical alignment of the gyri and sulci, the specified activation clusters of interest for each subject were mapped onto a common group space, prior to calculation. Finally, in the patient group, we performed *correlation analysis* between the quantitative functional activities in the regions of interest identified in the second-level analysis, and symptom severity, based on the PANSS total and subscale scores using the *SigmaPlot v10.0 software* (*Systat*, *USA 2007*). In this correlation analysis, the dosage of the antipsychotic drugs was introduced as a covariate of no interest. Results {#s3} ======= The t-test, F-test peaks and the corresponding corrected p-values are shown in [***Tables 2***](#pone-0017500-t002){ref-type="table"} ***and*** [***3***](#pone-0017500-t003){ref-type="table"}. ::: {#pone-0017500-t002 .table-wrap} 10.1371/journal.pone.0017500.t002 Table 2 ::: {.caption} ###### Cortical areas involved in the identification of intelligible speech in healthy controls and patients with schizophrenia. ::: {#pone-0017500-t002-2} BA Side t-values p (corrected values) Talairach coordinates ---------------------------------------- -------- ------ ---------- ---------------------- ----------------------- ***Healthy controls*** Pre-central gyrus 4 R+L 4.5 0.001 (-)44/-18/39 Medial frontal gyrus (SMA) 6 R+L 4.2 0.001 (-)3/-4/59 Insula 13 L 6.1 0.001 -43/8/5 Middle temporal gyrus 21 R+L 3.7 0.002 (-)51/-28/-10 Anterior cingulate gyrus 32 L 4.5 0.002 -3/32/29 Inferior parietal gyrus 40 R 5.9 0.001 47/-33/40 Inferior frontal gyrus (Broca\'s area) 44, 45 L 6.8 0.001 -46/18/16 ***Patients with schizophrenia*** Pre-central gyrus 4 R+L 4.1 0.001 (-)45/-17/37 Medial frontal gyrus (SMA) 6 R+L 4.2 0.001 (-)2/-4/56 Insula 13 L 5.7 0.001 -45/8/4 Middle temporal gyrus 21 R\>L 4.1 0.001 (-)53/-27/-9 Anterior cingulate gyrus 32 L 4.6 0.002 -1/29/31 Inferior parietal gyrus 40 R 5.9 0.001 42/-34/43 Inferior frontal gyrus (Broca\'s area) 44, 45 R+L 6.5 0.001 -46/19/17 BA: Broadmann\'s areas; R/L: right or left side of the brain; SMA: supplementary motor area. ::: ::: {#pone-0017500-t003 .table-wrap} 10.1371/journal.pone.0017500.t003 Table 3 ::: {.caption} ###### Cortical areas involved in implicit self-other distinction in healthy controls and patients with schizophrenia. ::: {#pone-0017500-t003-3} BA Side t-values p (corrected values) Talairach coordinates ----------------------------------- ---- ------ ---------- ---------------------- ----------------------- ***Healthy controls*** *Post-central gyrus* 3 R −8.5 0.002 (-)58/-16/26 *Pre-central gyrus* 4 L −11.8 0.001 -57/-8/22 *Medial frontal gyrus* 6 R −10.9 0.001 3/-5/56 *Medial frontal gyrus* 8 R+L −7.9 0.003 (-)3/46/39 *Inferior frontal gyrus* 44 L −7.8 0.003 -57/7/11 *Dentate nucleus (cerebellum)* NA R+L −10.3 0.001 (-)19/-53/-33 *Caudate nucleus* NA L\>R −7.9 0.003 (-)17/-1/22 *Thalamus* NA R+L −7.6 0.003 (-)10/-18/10 *Posterior cingulate gyrus* 23 R+L 8.0 0.002 (-)3/-54/20 *Posterior cingulate gyrus* 31 R+L 8.3 0.002 2/-54/26 *Anterior cingulate gyrus* 24 R+L 7.6 0.003 (-)3/32/12 *Anterior cingulate gyrus* 32 R+L 9.2 0.002 (-)3/34/18 *Superior temporal gyrus* 39 R+L 5.9 0.004 (-)47/-54/12 *Inferior parietal lobule* 40 R\>L 7.8 0.003 44/-38/40 ***Patients with schizophrenia*** *Post-central gyrus* 3 R −8.2 0.002 (-)59/-15/26 *Pre-central gyrus* 4 L −10.5 0.001 -56/-7/23 *Medial frontal gyrus* 6 R −10.7 0.001 2/-4/54 *Medial frontal gyrus* 8 R+L −8.0 0.002 (-)4/41/43 *Inferior frontal gyrus* 44 L −7.6 0.003 -58/6/13 *Dentate nucleus (cerebellum)* NA R+L −9.9 0.002 (-)9/-53/-30 *Caudate nucleus* NA L\>R −8.0 0.002 (-)13/4/14 *Thalamus* NA R+L −7.6 0.003 (-)9/-17/7 *Posterior cingulate gyrus* 23 R+L 9.8 0.002 (-)1/-56/21 *Posterior cingulate gyrus* 31 R+L 10.1 0.001 1/-51/32 *Anterior cingulate gyrus* 24 R+L 9.3 0.002 (-)2/33/12 *Anterior cingulate gyrus* 32 R+L 11.2 0.001 (-)1/32/20 *Superior temporal gyrus* 39 R+L 5.5 0.004 (-)47/-52/11 *Middle temporal gyrus* 39 R\>L 5.9 0.004 (-)45/-59/22 *Inferior parietal lobule* 40 R\>L 9.7 0.002 43/-39/38 BA: Brodmann\'s areas; R/L: right or left side of the brain; SMA: supplementary motor area. ::: Behavioral data {#s3a} --------------- All participants without exception performed successfully the training phase before the MRI-scanning for the explicit task (mentally repeat during \<SGV\>, listen during the other conditions). Moreover, post-session interviews revealed that no subjects from the patient and the control groups identified their own voices in the SGV condition, avoiding the possibility of explicit attribution judgments during the task. Activation network for intelligible speech {#s3b} ------------------------------------------ Intelligible language, tested by \[EGV -- RTV\] contrast RFX analysis, showed bilateral activations in both groups, in brain regions which are part of the MNS (Cf. [***Table 2***](#pone-0017500-t002){ref-type="table"}). More precisely, activity was measured in Broca\'s area, the left insula, in the middle part of the precentral gyrus (bilaterally), the supplementary motor area (SMA), the middle temporal gyrus (MTG) and the right inferior parietal lobule (IPL). The second level of this particular contrast analysis, corrected for "education" disparities, found increased activity in the right part of the MNS network (including the middle temporal gyrus), for patients with schizophrenia compared to the matched controls. Note that no activity was found in the mirror-like areas with \[RTV -- EGV\] reverse contrast analysis. Activation network for self-verbal agency {#s3c} ----------------------------------------- ### 1st level analysis {#s3c1} Implicit other-self differentiation was tested by \[EGV -- SGV\] contrast RFX analysis in each group (Cf. [***Table 3***](#pone-0017500-t003){ref-type="table"}). As already shown for healthy subjects, this analysis showed three different haemodynamic patterns [@pone.0017500-Jardri1]. First of all areas, that are classically more activated in language production \<SGV\> than in the listening condition \<EGV\>, were identified in the cortical and sub-cortical regions. As expected, higher activity at the level of premotor and motor cortices with the production condition was found in each group. Then, brain areas that were more activated in the listening conditions than in speech production were grouped into two different clusters. The first cluster corresponded to the areas activated during the \<EGV\> condition, which were simply activated less, or stayed at the baseline level during the \<SGV\> condition: the middle temporal gyri (Brodmann\'s area BA 39), and the right IPL at the level of the supramarginalis gyrus (BA 40). The second cluster included the areas that presented no reactivity for the listening conditions, but were deactivated during speech production, with respect to the resting condition: the ventro-medial prefrontal cortex (BA 32, 33) and the posterior cingulate cortex (BA 23, 31). ### 2nd level analysis {#s3c2} Finally, repeated ANOVA measurements showed an interaction effect between the groups and the conditions within the parieto-temporo-frontal network when differentiating others from self. This second-level analysis, corrected for "education" variations, showed increased activity for this contrast for the healthy subjects, compared with the patients with schizophrenia in the right IPL (**F~28~ = 20.22**; **p\<10^−4^**), the right MTG (**F~28~ = 21.80**; **p\<10^−5^**) and in the medial frontal (**F~28~ = 22.93**; **p\<10^−6^**) and medial parietal cortices (**F~28~ = 23.46**; **p\<10^−4^**), in the anterior and posterior cingulate cortices respectively. As an illustration, the lowest difference of amplitude of the haemodynamic responses observed between externally generated voices and the self (EGV and SGV respectively) in the right IPL in the patients with schizophrenia compared with the healthy controls, results from an increased activity during the self condition (Cf. [***Figure 2***](#pone-0017500-g002){ref-type="fig"}). In each group, probabilistic functional maps were then computed. These probabilistic maps confirmed up to 80% spatial consistency for the activity patterns in this neural network in each group, amongst all of the subjects tested, and good homogeneity for each experimental group. Within this network, we compared the surface extent of the clusters in the other-self contrast between the two groups using the following formula: (\[voxels~EGV~−voxels~SGV~\]/voxels~EGV~), and found significantly larger cluster sizes in the healthy subjects compared with the patients, in the right IPL (**t~28~ = −2.96**; **p\<10^−3^**), the right MTG (**t~28~ = −2.46**; **p\<10^−2^**), the medial frontal (**t~28~ = −3.67**; **p\<10^−4^**) and medial parietal cortices (**t~28~ = −3.29**; **p\<10^−3^**). These results all support a higher degree of overlap between self/non-self activations in schizophrenics compared with controls. ::: {#pone-0017500-g002 .fig} 10.1371/journal.pone.0017500.g002 Figure 2 ::: {.caption} ###### The self-other distinction network in patients with schizophrenia, compared with healthy controls. In the middle column, the red-to-yellow colour-code represents the relative percentage of subjects with greater activity in the \[EGV -- SGV\] contrast, after we performed cortex-based normalization between the healthy matched controls (CTL) and the patients suffering from schizophrenia (SCZ). %PM: percentage of activity in the subjects included at a specific spatial location, in a range from 20 to 90 (spatial consistency). In the patients, the difference in the \[EGV -- SGV\] contrast was significantly less marked than for the controls, which was also confirmed by the mean haemodynamic response of the fMRI signal in the right inferior parietal lobule (IPL). This is shown for each group in the left panel. The green bars represent the EGV condition, and the red bars show the SGV condition. In the right column, correlation analysis (adjusted for the dosage of the antipsychotic drugs) between the intensity of the fMRI signal in the self-other distinction network and the severity of the positive symptoms in schizophrenia (PANSS positive subscale) is shown. :::  ::: Correlation with severity of the positive symptoms {#s3d} -------------------------------------------------- Correlation analysis was performed in the parieto-temporo-frontal network, which has been previously identified. The activity level in the right IPL was initially found to correlate positively with the PANSS positive scores (**r^2^ = 0.68**; **p\<10^−5^**), although no such correlation was found in the temporal cortex or the medial brain areas of this network (r^2^\<10^−3^). Furthermore, no correlation was identified between the BOLD level in these areas and the other scores (negative, general or composite subscales, total PANSS score). After adjustment for the "dosage" factor, the analysis was repeated, but the response profile remained the same. The signal increase in the right IPL remained strongly correlated with the PANSS positive scores (**r^2^ = 0.65**; **p\<10^−4^**), but no such link was found between the severity scores and the temporal or medial brain areas. Discussion {#s4} ========== Using fMRI, we investigated the neural correlates of implicit self-other distinction in the verbal domain, in normal adults and patients suffering from schizophrenia. For both groups, other/self differentiation was accompanied with a high spatial consistency, by activation on the medial brain surface, of a neural network composed of the anterior and posterior cingulate cortices, coupled with right-sided structures, including the temporal cortex and the inferior parietal lobule (IPL). This result supports the idea that the schizophrenic patients use a comparable neural network to controls during social interaction, such as speech exchange situations. The involvement of these brain areas in implicit self-other distinction agrees with a meta-analysis showing IPL involvement in both explicit and low-level social cognitive processes [@pone.0017500-Decety3]. Another key finding in our study is that the difference between non-self and self cortical maps was smaller in the patients suffering from schizophrenia than in the controls. Indeed, the patients recruited more overlapping brain maps for self and non-self, as demonstrated by the cluster-size calculation between the two groups in addition to the weakest BOLD amplitude differences between the self and non-self conditions in the patients (cf. [*Figure 2*](#pone-0017500-g002){ref-type="fig"}). Finally since antipsychotic drugs can potentially disturb BOLD measurements, the strength of our results was reinforced by the finding that the effect remained, even after adjustment for the dosage factor. Altogether, our data are compatible with the idea of selectively-impaired implicit self-other processing in schizophrenia, which may account for passivity phenomena, independently of any MNS disturbance [@pone.0017500-Buccino2], since no hypofunctioning of this mirror network was observed during the intelligible speech listening contrast in the patients, compared with the healthy controls. Such a small difference between self and other related activations in patients with schizophrenia could be interpreted by referring to the SRM [@pone.0017500-Decety2], [@pone.0017500-Georgieff1]. As already mentioned in the introduction, this model is supported by neuroimaging studies showing a partial overlap in cortical activations during three types of action: those produced by oneself, those entirely mentally simulated, or those observed in other people. In this way, discrimination between self and non-self may occur, on the basis of non-overlapping areas, by means of a neural 'Who' system [@pone.0017500-Georgieff1]. In this setting, the pathological phenomenon of FRS in patients with schizophrenia, may result in self/non-self ambiguities, as suggested by Jeannerod [@pone.0017500-Jeannerod1]. However due to the results of the present experiment, different response profiles have to be distinguished within this network. On the one hand, no correlation with the severity of positive symptoms in the patients with schizophrenia was demonstrated for the medial parieto-frontal structures. This leads us to propose that dysfunctions of this medial network would more traduce a state-independent pattern, i.e. whatever the severity of actual symptoms is. A finest examination of the developmental trajectory of these brain areas also supports a 'trait-marker' hypothesis. First, the perigenual portion of the ACC, part of the medial parieto-frontal network recruited during self-other distinction, has a more progressive growth than caudal or dorsal portions [@pone.0017500-Kelly1]. In the same line of thoughts, long-range white-matter tracts connecting together the parieto-frontal areas of this network in the antero-poterior direction mature slowly during childhood, compared with transcallosal tracts that support functional connectivity between homotopic regions [@pone.0017500-Dubois1]. Such a grey and white-matter heterochronous maturation make these structures more fragile during brain development and could account for the expression of a medial parieto-frontal trait-pattern in patients vulnerable to schizophrenia, independently of the potential future emergence of acute psychotic episodes. On the other hand, anatomo-functional changes in the TPJ, frequently reported in schizophrenia [@pone.0017500-Torrey1], [@pone.0017500-Wible1], have been shown linked to the presence of particular symptoms. First, different sulcal displacements of this structure were evidenced in patients with inner and outer-space hallucinations [@pone.0017500-Plaze1]. Second, TPJ volume-reductions in schizophrenia were correlated with the severity of symptoms [@pone.0017500-Wilke1] or increased in patients who reported delusions of passivity [@pone.0017500-Maruff1]. Third, functional disturbances of this region were also correlated with the presence of psychotic symptoms [@pone.0017500-Spence1], [@pone.0017500-Farrer2]. In our study, the correlation analysis results between the BOLD signal amplitude and the intensity of positive symptoms ([Figure 2](#pone-0017500-g002){ref-type="fig"}) fully support the idea that hyperactivity in the right IPL could be considered a 'state-marker' for acute psychotic episodes. In such a perspective, the IPL may play a critical role in deciphering the origin of a stimulus, since the more the information is ambiguous, the more the IPL is functionally recruited. Increased activity measured within this structure in patients with schizophrenia compared with controls during the 'self' condition, might result from specific difficulties on the part of the patients in eliminating the ambiguity of the signals when the representations are self-generated. Therefore, it seems reasonable to assume that schizophrenic patients may have a tendency to experience self abnormally, even without having to explicitly judge whether the stimulus is from self, or not. It could be hypothesized that severe positive symptomatology would result in complete overlap between the self and non-self cortical maps at the level of the IPL. This is exactly what we found in a recent case-study of a child suffering from early-onset schizophrenia [@pone.0017500-Jardri2] using the same self-other fMRI contrast-analysis as in our current study. Using repetitive Transcranial Magnetic Stimulation (rTMS) over the child\'s right IPL allowed increasing source-monitoring and agency performances, as well as normalizing the activity pattern in the cortical site. Interestingly, studies of virtual lesions in healthy volunteers examining the exact opposite to this case-report, confirmed that the participants made more self false-alarms after TMS was applied over the right IPL, than when it was applied to a control contra-lateral site [@pone.0017500-Uddin1]. In order to test whether externally-generated stimuli would rely on a low-level impairment in schizophrenia, we used an implicit procedure. As a consequence, this procedure did not allow for task-performance monitoring within the scanner. It seems therefore essential to exclude alternative explanations of the results which could be due for instance to a poorer performance by the patients or to the fact they might be more likely than controls to repeat someone else\'s voice within their head. Our results allow us to think that such interpretations are unlikely. First due to its simplicity the task itself did not result in significant differences in the behavioral performances between the two groups. Indeed, control tasks implemented outside the scanner confirmed that patients with schizophrenia performed the task similarly to healthy subjects. Secondly, the activation patterns measured in the other vs. self contrast (\<EGV\> - \<SGV\>) confirmed that participants from the both group performed the task adequately within the scanner (i.e. covert speak during \<EGV\> and passively listen during \<EGV\>). Moreover assuming a greater tendency to action (i.e. repeat voices) in the patients seems improbable since no group difference was evidenced within premotor and motor areas in this contrast analysis. For all these reasons, we believe that our fMRI procedure provides strong evidences for an increased overlap between the neural representations for self and others in schizophrenia. Even though self-awareness impairment has already been suggested in schizophrenia, our study is the first evidence of neural impairments during implicit self-other distinction, in accordance with recent behavioural data [@pone.0017500-Bulot1]. Within this network, the anterior and posterior cingulate cortices were not shown to have a proportional relationship with the severity of positive symptoms. This alteration could therefore constitute a potential bio-marker for the disease. The recent research program for editing the DSM-V classification clearly emphasized the need to apply this sort of neuroscience findings, which reflect pathophysiological processes, to help with screening and earlier diagnosis of psychiatric disorders [@pone.0017500-Charney1]. However, the activity measured in the right IPL correlated strongly with the current status of the patients, and so should perhaps be considered more as a state-marker for FRS. The right IPL could in fact prove to be a possible target for dimensional treatment, such as rTMS. Finally, in order to distinguish between the different specific FRS, which are linked with the feeling of agency, further studies which replicate our paradigm and include different sub-groups, such as hallucinators vs. non-hallucinators, and healthy subjects with a high genetic risk of developing schizophrenia are required. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The authors have no support or funding to report. [^1]: Conceived and designed the experiments: RJ DP PT. Performed the experiments: RJ DP EV AA CD. Analyzed the data: RJ DP GL. Contributed reagents/materials/analysis tools: RJ DP CD. Wrote the paper: RJ DP GL EV AA CD PT.

PubMed Central

2024-06-05T04:04:19.209578

2011-3-9

PMC3052364

Introduction {#s1} ============ MicroRNAs (miRNAs) are an abundant class of small non-coding RNAs of ∼22 nucleotides in length that function as negative gene regulators [@pone.0017648-Ambros1], [@pone.0017648-Bartel1]. In animals, miRNAs are involved in processes such as tissue development and cell differentiation [@pone.0017648-Stefani1], apoptosis [@pone.0017648-Spizzo1], in which fine regulation of gene expression in time and space is required for the correct execution of these processes; and in diseases such as cancer [@pone.0017648-Garzon1]. These known functions may represent just a small part of a much bigger scenario; the main known function of miRNAs is the regulation of gene expression at the post-transcriptional level either by protein translation inhibition or mRNA decay [@pone.0017648-Brodersen1], [@pone.0017648-Guo1]. One third of the genes in the human genome are predicted to be miRNAs targets [@pone.0017648-Lewis1] and the continuing discovery of new miRNAs functions suggest that these molecules are implicated in the regulation of almost every physiological processes. Profile studies have already shown that many miRNAs are specifically expressed in certain organs, cell types and developmental stages [@pone.0017648-Stefani1]. To date only two recent studies have demonstrated that key haemostatic proteins, PAI-1 and fibrinogen, may be regulated by miRNA [@pone.0017648-Fort1], [@pone.0017648-Muth1], probably just reflecting the tip of the iceberg concerning regulation of haemostasis by miRNA. In this study, we tried to show the potential relevance that miRNAs regulation might have in the whole haemostatic system, by evaluating the differential expression of miRNAs in the liver of mice associated with the extraordinary change in the expression of hepatic haemostatic proteins after birth. Developmental haemostasis refers to the age-related changes in the coagulation system that are most marked during neonatal life and childhood [@pone.0017648-Monagle1]. Actually, the haemostatic system is incompletely developed at birth and matures throughout infancy. Neonates have low levels of the most procoagulant and anticoagulant proteins, although the levels of the factors V, VIII, XIII, and fibrinogen in neonates are similar to adult [@pone.0017648-Andrew1]. The most intriguing aspect of developmental haemostasis is to understand the mechanisms and rationale for such marked age-related changes. Previous studies have confirmed that post-translational modifications in coagulation proteins do occur with age. Moreover, significant differences at the transcriptional levels can also contribute to these differences. However, due to the role of miRNAs in development, these molecules may also contribute directly or indirectly to the dramatic changes in the haemostatic system observed in neonates [@pone.0017648-Bartel1], [@pone.0017648-Andrew1]. Results {#s2} ======= We studied the differential expression pattern of 558 mature miRNAs in liver from an adult and a neonate mouse. As expected, we found that miRNAs expression profiles significantly differed in these two stages. The expression level of 81 miRNAs significantly changed between livers from adult and neonate ([Figure 1](#pone-0017648-g001){ref-type="fig"}). We filtered these data by the fold change, and found that 68 out of 81 miRNAs showed at least a two-fold expression change between neonate and adult liver ([Table 1](#pone-0017648-t001){ref-type="table"}). Among these miRNAs, 41 were overexpressed in neonate in comparison with adult and 27 miRNAs were overexpressed in adult when compared to neonate ([Table 1](#pone-0017648-t001){ref-type="table"}). If certain miRNAs were directly involved in the control of the expression of haemostatic proteins, we could find such candidates among the miRNA overexpressed in neonates. Thus, we performed an exhaustive *in silico* search of haemostatic proteins that might be potential targets of the 41 miRNAs overexpressed in neonate mice. Interestingly, 21 out of 41 miRNAs overexpressed in neonate mice, have haemostatic mRNA as potential targets ([Table 2](#pone-0017648-t002){ref-type="table"}). When we focused on those targets predicted by both algorithms, we found 6 miRNA that potentially can bind the 3′UTR regions of mRNA from *F7* (encoding the coagulation factor VII), *F9* (encoding the coagulation factor IX), *F12* (encoding the coagulation factor XII), *FXIIIB* (encoding the coagulation factor XIII, B polypeptide), *PLG* (encoding plasminogen), and *SERPINC1* (encoding the anticoagulant antithrombin) genes. All these proteins are significantly lower in neonates compared with adults in humans although no data are available in mice. Then we validated the expression of these miRNA observed in the microarray in 14 neonate mice (+1 day after bird) and 6 adult mice by qRT-PCR. The results obtained were fully comparable, and the relative fold changes of miRNA expression were similar to those detected by microarrays (r = 0.78, p\<0.05) ([Figure 2](#pone-0017648-g002){ref-type="fig"}). We were particularly interested to deepen in two miRNAs: miR-18a and miR-19b which are overexpressed 5.4 and 8.2-fold and both have antithrombin as a potential target. Antithrombin is a key anticoagulant that also plays other relevant roles outside the haemostatic system such as strong anti-angiogenic [@pone.0017648-SchedinWeiss1] and anti-inflammatory [@pone.0017648-Afshari1] roles. Moreover, the levels of this serpin are naturally reduced in newborns to less than 50% of the levels observed in adults and then increase to approach adult levels by approximately six months of age in humans [@pone.0017648-Monagle2]. First, we confirmed the lower antigenic levels of antithrombin in plasma of neonate mice compared with adults (34±4 *vs.* 86±7, respectively), reaching adult levels at +13 days after birth. Interestingly, when quantifying miR-18a, miR-19b, and antithrombin mRNA during the 19 days after birth, we found an inverse and significant correlation (miR-19b: R = 0.81; p = 0.03; miR-18a: R = 0.91; p\<0.001) ([Figure 3](#pone-0017648-g003){ref-type="fig"}). We point out that miR-18a and miR-19b are also expressed in human liver where their expression pattern during development is similar to that observed in our study [@pone.0017648-Tzur1]. ::: {#pone-0017648-g001 .fig} 10.1371/journal.pone.0017648.g001 Figure 1 ::: {.caption} ###### miRNA expression profile. Differentially expressed miRNAs were clustered by Cluster3.0. Log2 scale was employed. In red miRNAs overexpressed in neonate liver; in green miRNAs overexpressed in adult liver. :::  ::: ::: {#pone-0017648-g002 .fig} 10.1371/journal.pone.0017648.g002 Figure 2 ::: {.caption} ###### miRNA expression validation. \(A) Selected miRNAs were validated by RT-PCR using specific assays. (B) Correlation between data from microarray and RT-PCR for selected miRNAs was performed using lineal regression (SPSS15.0 software). P\<0.05 was considered as statistically significant. :::  ::: ::: {#pone-0017648-g003 .fig} 10.1371/journal.pone.0017648.g003 Figure 3 ::: {.caption} ###### Expression of antithrombin and miRNAs miR-18a and miR-19b during post-natal development in mouse. \(A) miRNAs and antithrombin mRNA were measured in liver from mice during postnatal development using cDNA specific assays. RT-PCR was performed in three mice from each age. Data are shown as relative level (% ± SD) taking adults as reference for antithrombin mRNA and +1 day neonates for miRNAs. (B) Expression correlation between miR-19b and miR-20a levels with AT mRNA levels, during post-natal development. :::  ::: ::: {#pone-0017648-t001 .table-wrap} 10.1371/journal.pone.0017648.t001 Table 1 ::: {.caption} ###### Potential haemostatic factor mRNA targeted by miRNAs overexpressed in neonate liver. ::: {#pone-0017648-t001-1} miRNA Fold change mirGen Targetscan --------------------- ------------- ---------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------- **mmu-miR-376b** 85,6 \- Protein C, Factor V **mmu-miR-154** 26,5 \- Factor IX **mmu-miR-379** 26,9 Factor VIII Factor V **mmu-miR-223** 15,1 **Factor IX** [\#](#nt101){ref-type="table-fn"}, **plasminogen** [\#](#nt101){ref-type="table-fn"} Factor VIII, **Factor IX** [\#](#nt101){ref-type="table-fn"}, **Plasminogen** [\#](#nt101){ref-type="table-fn"} **mmu-miR-495** 14,3 Factor II \- **mmu-miR-382** 17,9 Factor XIII (beta subunit), Plasminogen \- **mmu-miR-200a** 8,4 Factor XIII (beta subunit) \- **mmu-miR-130b** 7,5 Antithrombin, Factor XIII (beta subunit) Protein C **mmu-miR-19b** 8,2 **Antithrombin** [\#](#nt101){ref-type="table-fn"}, Factor XIII (beta subunit) **Antithrombin** [\#](#nt101){ref-type="table-fn"} **mmu-miR-300** 9,2 **Factor XIII (beta subunit)** [\#](#nt101){ref-type="table-fn"} **Factor XIII (beta subunit)** [\#](#nt101){ref-type="table-fn"} **mmu-miR-134** 8,2 **Factor VII** [\#](#nt101){ref-type="table-fn"} **Factor VII** [\#](#nt101){ref-type="table-fn"} **mmu-miR-299\*** 9,0 Factor II, Factor XI \- **mmu-miR-199a-5p** 4,3 \- Factor XI, Factor XII **mmu-miR-18a** 5,4 **Antithrombin** [\#](#nt101){ref-type="table-fn"} **Antithrombin** [\#](#nt101){ref-type="table-fn"} **mmu-miR-667** 5,2 Protein S (alpha), Factor VII \- **mmu-miR-99b** 4,0 Plasminogen, Factor II \- **mmu-miR-30e** 2,3 \- Factor XI **mmu-miR-145** 2,8 Plasminogen \- **mmu-miR-30a** 2,5 \- Factor XI **mmu-miR-130a** 2,0 Antithrombin, Factor XIII (beta subunit) Protein C **mmu-miR-181a** 2,4 Factor VII, Factor X, **Factor XII** [\#](#nt101){ref-type="table-fn"}, Protein S (alpha) **Factor XII** [\#](#nt101){ref-type="table-fn"}, FactorXI \# mRNA targeted in both algorithms. ::: ::: {#pone-0017648-t002 .table-wrap} 10.1371/journal.pone.0017648.t002 Table 2 ::: {.caption} ###### Potential haemostatic factor mRNA targeted by miRNAs over-expressed in neonate liver. ::: {#pone-0017648-t002-2} *miRNA* *mirGen* *Targetscan* ---------------------------------------------------- ---------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------- [¶](#nt102){ref-type="table-fn"} **mmu-miR-223** **Factor IX** [\*](#nt104){ref-type="table-fn"}, **Plasminogen** [\*](#nt104){ref-type="table-fn"} Factor VIII, **Factor IX** [\*](#nt104){ref-type="table-fn"}, **Plasminogen** [\*](#nt104){ref-type="table-fn"} [¶](#nt102){ref-type="table-fn"} **mmu-miR-19b** **Antithrombin** [\*](#nt104){ref-type="table-fn"}, Factor XIII (beta subunit) **Antithrombin** [\*](#nt104){ref-type="table-fn"} [¶](#nt102){ref-type="table-fn"} **mmu-miR-300** **Factor XIII (beta subunit)** [\*](#nt104){ref-type="table-fn"} **Factor XIII (beta subunit)** [\*](#nt104){ref-type="table-fn"} [¶](#nt102){ref-type="table-fn"} **mmu-miR-134** **Factor VII** [\*](#nt104){ref-type="table-fn"} **Factor VII** [\*](#nt104){ref-type="table-fn"} [¶](#nt102){ref-type="table-fn"} **mmu-miR-18a** **Antithrombin** [\*](#nt104){ref-type="table-fn"} **Antithrombin** [\*](#nt104){ref-type="table-fn"} [\#](#nt103){ref-type="table-fn"} **mmu-miR-181a** Factor VII, Factor X, **Factor XII** [\*](#nt104){ref-type="table-fn"}, Protein S (alpha) Factor XI, **Factor XII** [\*](#nt104){ref-type="table-fn"} ¶ Overexpressed 4--32-fold in comparison with adult liver, \# overexpressed 2--4-fold in comparison with adult liver. \*mRNA targeted in both algorithms. ::: Discussion {#s3} ========== Our study has identified 41 miRNAs overexpressed in livers of neonate mice, some of them with potential direct effect on hepatic haemostatic proteins that dramatically change their levels after birth. The inverse correlation observed between miR-18a and miR-19b levels with antithrombin mRNA, one potential target of these miRNAs, suggests that certain miRNAs may be involved in the regulation of selected hepatic haemostatic proteins during development by targeting mRNA coding for these proteins and be in part responsible of the observed decay in neonates [@pone.0017648-Andrew1]. Indeed, Tzur et al. suggested that in humans several biological processes and pathways, in particular blood coagulation, may be regulated during the embryonic period by differentially expressed miRNAs [@pone.0017648-Tzur1]. Beside the potential role of miRNAs in the regulation of haemostatic proteins in developmental haemostasis in liver, these molecules may also be of relevance in adults where the wide inter-individual variability for haemostatic proteins may be potentially responsible of bleeding or thrombotic disorders. In addition, miRNAs may also play a role through indirect mechanisms, by regulating transcriptional factors or post-translational modifications affecting haemostatic elements. Several studies have shown that certain miRNAs may regulate the levels of transcription factors expressed in the liver, potentially involved in the regulation of the transcription of a large panel of hepatic haemostatic mRNAs [@pone.0017648-Guttilla1], [@pone.0017648-Lei1], [@pone.0017648-Wan1]. Additionally, our results are based on a microarray performed on the expression of 558 miRNAs, the use of novel miRNA databases would certainly add additional miRNA candidates as potential regulators of other hepatic haemostatic proteins not described in our study. Obviously, the main limitation of our study is that it did not demonstrate through *in vitro* experiments the validity of our hypothesis. To note, we looked for a cellular model that endogenously express antithrombin in order to inhibit its expression with oligonucleotide precursors. First, we examined AML12, a murine hepatocyte cell line that unexpectedly, did not express antithrombin. We next move to NIH3T3 cells that do express antithrombin and we performed transfection assays with oligonucleotide precursors and inhibitors (pre-miRs and antagomiRs from Applied Biosystems, Madrid, Spain) of miR-18a and miR-19b to evaluate their effect in the expression of antithrombin. Our results suggest that these miRNAs might not have a direct effect on antithrombin levels in this cell line (data not shown). However, this cellular model had several pitfalls. Particularly, NIH3T3 cells are not a hepatic cell line, and accordingly, cytosolic environment may be completely different from that of hepatocytes. Moreover, the level of expression of antithrombin is very low, and fibroblasts do not express antithrombin in the adulthood. Further studies are required to sustain the potential role of miRNA in the dramatic change of the haemostatic system after birth. Thus, it is necessary to clarify the mechanisms of action of candidate miRNA overexpressed in neonates on the levels of haemostatic proteins. Any potential direct inhibitory effect of miRNAs must be demonstrated under appropriate conditions: i.e. embryonic mouse hepatocytes transfected with miRNA inhibitors or adult mouse hepatocytes transfected with miRNA oligonucleotide precursors. On the other hand, the indirect effect of miRNA on haemostatic protein regulation by targeting transcription factors may also be extremely relevant to evaluate. In conclusion, these results open new fascinating perspectives in thrombosis and haemostasis by introducing novel elements, miRNAs, as potential regulators of the haemostatic system. Methods {#s4} ======= Mouse samples {#s4a} ------------- For microarray assay, we sacrificed one mice of different stage of development: one day after birth mouse (neonate) and fifty days after birth mouse (adult). On the other hand, we used 6 adult mice and 14 neonate mice from at 4 different litters, which did not include the samples used for the microarray assays, to perform miRNA expression validation assay. Finally, we sacrificed 3 mice from different litters, from neonate stage (day +1) to adult stage (day +50) each two days, to perform the quantification of antithrombin mRNA levels and miRNAs miR-18a and miR-19b levels, during post-natal development. In all case we used swiss (ICR CD-1™) mice. Livers were finely dissected and immediately frozen in liquid nitrogen and kept at −80°C until their use. Blood was anticoagulated with citrate and plasma stored at −80°C. All experimental procedures were performed in accordance with the approved Institutional Animal Care Guidelines from the University of Murcia. The study was approved by the Ethics Committee from Fundación para la Formación e Investigación Sanitarias de la Región de Murcia (04--23/2008). RNA Isolation {#s4b} ------------- Total RNA was isolated from frozen liver using Trizol® Reagent (Invitrogen, Carlsbad, CA) following manufacturer\'s instructions. The RNA concentration and 260/280 ratio were determined by using NanoDrop spectrophotometer (Thermo Scientific, Wilmington, DE) and RNA integrity was verified by lab-on-chip technology using the Experion automated electrophoresis system (Bio-Rad Laboratories, Madrid, Spain). MicroRNA microarray {#s4c} ------------------- MicroRNAs microarray profiling was performed using total RNA extracted from the liver from one adult mouse (day +50) and one neonate mouse (day +1) using the LC Sciences technology (LC Sciences, Houston, TX). The arrays were designed to detect and quantify miRNA transcripts corresponding to 558 mature miRNAs contained in the Sanger mirBase Release 10.0 (miRMouse 10.0) <http://www.sanger.ac.uk/Software/Rfam/mirna/>. We used two chips in which RNAs from each sample were labeled either with cy3 or with cy5. The signal values were derived by background subtraction and normalization. A transcript to be listed as detectable must meet at least two conditions: signal intensity higher than 3×(background SD) and spot CV\<0.5. CV is calculated by (SD)/(signal intensity). When repeating probes were present on an array, a transcript was listed as detectable only if the signals from at least 50% of the repeating probes were above detection level. Signals were listed in median signal values of repeating probes of p-value\<0.01. Median values were used to minimize the effect of occasional "non-uniform spots" that may have signal values deviate from average signal values but have p-values\<0.01. All differentially expressed transcripts with p-value\<0.01, were presented in log2 scale, positive log2 value indicates an upper regulation and a negative log2 value indicates a down regulation. Validation by qRT-PCR {#s4d} --------------------- The RT reaction was performed using 200 ng of total RNA for each sample according to the manufacturer instructions (SuperScript First Strand, Invitrogen, Madrid, Spain). The Real Time reactions were performed using Taqman Gene Expression Assay on a LC480 Real Time PCR (Roche, Madrid, Spain). For mature miRNA quantification one set of primers and a probe were chosen from Applied Biosystems. Expression analysis was performed in triplicates for each sample. Expression of U6 snRNA was used as endogenous reference control. The fold difference for each sample was obtained using the 2^−ΔΔCt^ method. Target analysis {#s4e} --------------- *In silico* search of potential miRNA targets was performed in the following miRNA-target prediction algorithms : TargetScan (release 5.1, <http://www.targetscan.org>) [@pone.0017648-Lewis1], mirGen (<http://www.diana.pcbi.upenn.edu/miRGen>) [@pone.0017648-Megraw1]. Statistical analysis {#s4f} -------------------- Lineal regression was performed by Statistical Package for Social Science (version 15.0; SPSS, Chicago, IL, USA). Data are presented as mean ± standard deviation. Statistical significance was taken as P\<0.05. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**RT holds an FPI fellowship from the Spanish Ministry of Science and Innovation (BES-2007-16973). VPA holds a Research Fellowship Rio Hortega from ISCIII. CM and IMM are investigators from Fundación para la Formación e Investigación Sanitarias de la Región de Murcia (FFIS). This work was supported by research grants from ISCIII ([www.isciii.es](http://www.isciii.es)) and FEDER (PI08/1506, EMER07/035 and RECAVA RD06/0014/0039), MCYT ([www.mityc.es](http://www.mityc.es)) & FEDER (SAF2009-08993), Fundación Séneca ([www.f-seneca.org](http://www.f-seneca.org)) 07703/GERM/07 and Fundación Mutua Madrileña ([www.fundacionmutua.es](http://www.fundacionmutua.es)). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: RT JC CM. Performed the experiments: RT VPA IMM. Analyzed the data: RT JC CM. Wrote the paper: RT JC CM. Critically read the manuscript: VV.

PubMed Central

2024-06-05T04:04:19.213633

2011-3-9

PMC3052365

Introduction {#s1} ============ Highly pathogenic avian influenza (HPAI) subtype H5N1 (hereafter H5N1) first emerged in domestic geese of south-eastern China in 1996 [@pone.0017622-Guo1], [@pone.0017622-Xu1] and has since become endemic in poultry across much of Eurasia [@pone.0017622-OIE1], [@pone.0017622-Sims1]. Despite extensive eradication and vaccination campaigns, the virus continues to persist, re-emerging across much of its range. Waterfowl belonging to the family Anatidae (ducks, geese, and swans) are natural reservoirs for low pathogenic forms of avian influenza [@pone.0017622-Alexander1] and are rarely observed with HPAI infection [@pone.0017622-Alexander2]. H5N1 is unique, however, being the first HPAI virus to repeatedly cross the poultry species barrier back to wild birds. The first reports of spillover to wild birds occurred in low numbers of captive waterbirds at a Hong Kong waterfowl park in 2002 [@pone.0017622-Ellis1]. However, it was not until April 2005 that the first large epizootic (more than 6000 birds) occurred in wild species on the remote breeding grounds of Qinghai Lake (QHL), north-western China [@pone.0017622-Chen1], [@pone.0017622-Liu1]. Infections began in bar-headed geese *(Anser indicus)* soon after their migratory return to QHL, and were followed by infections in great black-headed gulls *(Larus ichthyaetus)*, brown-headed gulls *(Larus brunnicephalus)*, and great cormorants *(Phalacrocorax carbo)* 10 days later, and ruddy shelducks *(Tadorna ferruginea)* within 3 weeks [@pone.0017622-Chen2]. Over half of the reported cases were in bar-headed geese. The QHL outbreak was significant not only because it was the first major H5N1 epizootic in wild populations, but also because it occurred in a region generally lacking poultry [@pone.0017622-FAO1], which raised questions about the source of the virus and the ability of wild birds to transport virus over long distances. Following the QHL epizootic, H5N1 expanded beyond Asia, spreading northward and westward into Europe and Africa. All viruses isolated from these regions were subsequently traced back to QHL (clade 2.2) [@pone.0017622-Sims1], [@pone.0017622-Chen3] which fueled debate regarding the role wild birds play in H5N1 spread [@pone.0017622-GauthierClerc1], [@pone.0017622-Fouchier1], [@pone.0017622-Normile1], [@pone.0017622-Weber1]. QHL is a critical breeding ground and staging area for migratory waterbirds, supporting 150,000 migrants each year [@pone.0017622-Li1], including 15% of the global breeding population of bar-headed geese [@pone.0017622-Miyabayashi1]. It is situated at the intersection of two major flyways: the Central Asian Flyway which extends from western Siberia through central Asia and south to India, and the East Asian Flyway which ranges from Russia through eastern China and south to Australia [@pone.0017622-Boere1]. QHL holds multiple international designations of ecological importance: Important Bird Area [@pone.0017622-BirdLife1], Key Staging Site for migrating Anatidae (waterfowl) [@pone.0017622-Miyabayashi1], Wetland of International Importance [@pone.0017622-Ramsar1], and national nature reserve of China. Despite its importance for migratory birds, little is known about wintering and breeding locations for species using the lake during different seasons [@pone.0017622-Miyabayashi1]. If wild birds had played an integral role in transporting H5N1 to QHL, we hypothesized that the following conditions would be observed: (1) exposure to H5N1 virus on their wintering grounds or migratory stopovers during the northern hemispheric spring migration, (2) significant migration distances and viral shedding before becoming physiologically compromised, and (3) high sequence similarity between virus isolates from areas of exposure and those from QHL. To examine these questions, we sought to develop an improved understanding of the migratory movements, geographic range, habitat use, and overlap of bar-headed geese with poultry in zones of infection [@pone.0017622-Miyabayashi1], [@pone.0017622-Muzaffar1], [@pone.0017622-Yasue1]. In 2007 and 2008, we tracked the migratory movements of 29 bar-headed geese from QHL using GPS satellite telemetry. We compared the spatial ecology of the host with the trajectory of H5N1 across the Qinghai-Tibet Plateau using phylogenetic analysis. This interdisciplinary approach used traditionally disparate tools -- satellite telemetry and virological analysis - to assess the role of bar-headed geese in the transmission dynamics of H5N1. Materials and Methods {#s2} ===================== Study Area {#s2a} ---------- We sampled birds at the Qinghai Lake National Nature Reserve (QLNNR), Qinghai Province, in north-western China (36.82 N, 99.81 E). QHL is China\'s largest saltwater lake (4500 km^2^) and is located 3200 m above sea level on the eastern edge of the Qinghai-Tibet Plateau. The lake remains frozen November--March, and has a short rainy season from June--August (0.35 m annual average rainfall) [@pone.0017622-Xu2]. The lake is a closed basin fed by 25 freshwater streams, the majority of which have intermittent flow [@pone.0017622-Rhode1]. Marking and Virology Sampling {#s2b} ----------------------------- Bar-headed geese return to QHL in late March when temperatures are below freezing and the saline lake is still frozen. During this time, geese use freshwater springs and wetlands surrounding the lake before moving to one of three breeding colony sites in early April. We captured geese during the pre-breeding season at four non-colony sites to reduce disturbance to breeding colonies. Capture and marking occurred in late March 2007--2008 at QLNNR using monofilament noose sets and a net launcher (Coda Enterprises, Mesa, Arizona, USA). We recorded standard morphometrics (mass, flat wing chord, and diagonal tarsus [@pone.0017622-Dzubin1]), age, and sex. Individuals were sampled for avian influenza (cloacal and tracheal swabs, blood serology) following standard procedures [@pone.0017622-FAO2]. Virology samples were analyzed by the Chinese Academy of Sciences, Wuhan Institute of Virology with the following methods: (1) type A influenza with an ELISA test (Optical Density 630 above 0.23 as positive), (2) H5 subtype with RT-PCR [@pone.0017622-Fouchier2], and (3) H5, H7, H9, and H10 antibodies with hemagglutinin inhibition [@pone.0017622-OIE2]. Each bird was equipped with a 45-g, solar-powered GPS platform transmitter terminal (PTT; Microwave Telemetry, Columbia, Maryland, USA). PTTs measured 57×30×20 mm and were attached dorsally with a double-threaded backpack harness made of Teflon ribbon (Bally Ribbon Mills, Bally, Pennsylvania, USA). Transmitter packages weighed on average \<2.1% of the bird\'s body mass. Capture, handling, and marking procedures were approved by the United States Geological Survey Patuxent Wildlife Research Center Animal Care and Use Committee and University of Maryland Baltimore County Institutional Animal Care and Use Committee (Protocol EE070200710). The PTTs were programmed to take GPS locations every 2 hours, and data were uploaded to satellites every 2 days (CLS America Inc., Maryland, USA). We used ArcGIS 9.3 (Environmental Systems Research Institute, Inc., Redlands, California, USA) and Google Earth 5.0 (Google, Mountain View, California, USA) to plot and analyze telemetry locations. Telemetry, Movements, and H5N1 Risk Factors {#s2c} ------------------------------------------- We evaluated chronology, movement rates, seasonal habitat use, wintering locations, and migration stopover sites of the geese. Four seasons of the life cycle of the geese were defined: breeding (including post-breeding molt), fall migration, wintering, and spring migration. Migratory stopovers were defined as areas used during migration in which a goose moved no more than 20 km in a 24 h period. Stopover boundaries were drawn with minimum convex polygons (MCP) since we lacked enough locations to apply kernel home range methods. We calculated cumulative distance and time to complete each migration leg (the path between two consecutive stopover locations) for all geese that completed a minimum of one fall migration. The longest migration leg (km) for each goose was used to estimate the greatest distance travelled between stationary periods. Total migration distance for each individual was defined by the Euclidian distance between its northernmost and southernmost location. We used ArcGIS 9.3 and Hawth\'sTools [@pone.0017622-Beyer1] to conduct these analyses. Habitat use was evaluated by extracting land cover variables from telemetry locations according to season, pooled across individuals. Habitat use (expressed as a percentage) was based on the number of telemetry locations recorded at each land cover type divided by the total number of locations recorded in a season. Habitat features were derived from 1-km land-cover data produced by the Chinese Academy of Sciences [@pone.0017622-Liu2], [@pone.0017622-Liu3]. We reduced 25 land-cover classes into six for analysis: urban, cropland, grassland, wetland, woodland and other. Unlike conventional land cover data which represent the landscape as a single class per pixel, continuous fields in this dataset included the percent cover of all classes present within a pixel. Spatial overlap between goose movements and poultry farming was examined based on 1-km poultry densities from UNFAO Geonetwork [@pone.0017622-FAO1]. ### Brownian bridge utilization distributions {#s2c1} To examine relationships between goose movements, landscape features, and potential risk factors associated with H5N1 transmission, we created Brownian Bridge utilization distributions (BBUD) [@pone.0017622-Horne1] to describe goose migration patterns. Advantages of using the BBUD model over kernel density estimates for migration analyses is that the BBUD method avoids issues of serial correlation between points, assumes locations are temporally dependent, and explicitly includes length of time between locations in the model. This approach removes subjectivity in estimating temporal weights between intervals of unequal length and uses observed movements and measurable location error to model probability of occurrence. This effect is critical for migration when animals are moving long distances in short periods of time [@pone.0017622-Horne1]. We created fall and spring migration BBUDs for each individual with Animal Space Use software [@pone.0017622-Horne2]. An estimate of spring and fall migration routes were calculated as the mean probability of occurrence across individuals [@pone.0017622-Horne1]. This required defining consistent seasonal dates across individuals from earliest departure and latest arrival dates of marked birds. Fixed kernel home ranges were created for breeding and winter seasons. Least squares cross-validation was applied to obtain kernel smoothing parameters [@pone.0017622-Horne3], [@pone.0017622-Ackerman1] and utilization distribution models were developed for breeding and wintering with Animal Space Use [@pone.0017622-Horne2]. Geese were weighted equally by including the same number of locations selected randomly for each individual. Probabilistic utilization distribution (UD) models for the group were created for all birds as a group with a Brownian bridge movement model for the spring and fall migration and fixed kernel home range models for breeding and non-breeding seasons. ### HPAI H5N1 outbreaks {#s2c2} HPAI H5N1 outbreak data for 2003--2009 were obtained from two databases: the People\'s Republic of China Ministry of Agriculture Prevention and Control of Avian Influenza database [@pone.0017622-MOA1] and the UNFAO\'s Emergency Prevention System for Transboundary Animal and Plant Pests and Diseases (EMPRES-i) database. Outbreak events were cross-checked between the two databases ([Table S1](#pone.0017622.s001){ref-type="supplementary-material"}) and imported into a GIS framework for analysis. Statistical Analyses {#s2d} -------------------- We conducted a geospatial analysis of bar-headed goose movements in relation to H5N1 risk factors and outbreak locations. This is the first effort we are aware of to compare separate analyses of risk factors for poultry and wild bird outbreaks. Following an information theoretic approach (AIC) to compare outbreak and non-outbreak (random) locations under two sets of *a priori* defined logistic regression models [@pone.0017622-Burnham1], we hypothesized that poultry outbreaks would be explained by anthropogenic factors (poultry density, etc.) and wild bird outbreaks by habitat and goose utilization distributions. Covariate predictors in the poultry models included latitude (Lat), bar-headed goose utilization distribution (BHGO UD), poultry density (PD), and cropland (Crop). Wild bird model covariates included Lat, BHGO UD, PD, grassland (Grass), and wetland (Wetl). Non-outbreak locations were represented by 10 random locations for each outbreak [@pone.0017622-Manly1] and were drawn proportionally from the spatial extent of outbreaks in each season. A separate *a priori* univariate logistic regression (BHGO UD) was conducted to examine goose exposure to H5N1 outbreaks. Analyses were performed with the R statistical package [@pone.0017622-R1] 'glm' function (family = binomial, link = −logit). A classification and regression tree (CART) identified significant model covariates with the Rpart package in R [@pone.0017622-R1], [@pone.0017622-Therneau1]. Binary trees were built by recursively partitioning explanatory variables into high and low categories that significantly contributed to the prediction of outbreak versus random locations [@pone.0017622-Breiman1]. Temporal aspects of outbreaks were examined for wild birds and domestic poultry by (a) comparing seasonal differences in numbers of outbreaks of each type (poultry or wild bird) and (b) comparing the observed and expected number of outbreaks based on the length of the season for each outbreak type using Fisher\'s Exact Tests [@pone.0017622-Agresti1], [@pone.0017622-Fisher1]. Expected numbers were calculated under the assumption that outbreaks were proportional to the number of days within a seasonal period. Phylogenetics {#s2e} ------------- We conducted a phylogenetic analysis to determine genetic relatedness of H5N1 viruses isolated from wild birds and poultry with a focus on the Qinghai-Tibet Plateau. Sequences of HPAI H5N1 were obtained from the Genbank database hosted by the National Center for Biotechnology Information (NCBI) as of 30 September 2010. All phylogenetic analyses were performed with MEGA version 4.0 [@pone.0017622-Tamura1]. A total of 38 sequences of the HA (hemagglutinin) gene were trimmed to a length of 1550 bp and aligned with ClustalW following default settings. A phylogenetic tree was created by applying the neighbor-joining method [@pone.0017622-Saitou1] and evolutionary distances were computed with the Kimura 2-parameter method [@pone.0017622-Kimura1]. Bootstrapping (×1000) was used to assess the reliability of the tree topology. The tree was rooted to A/goose/Guangdong/1/96 and structured according to the World Health Organization system of avian influenza cladistics. Results {#s3} ======= Marking and Virology {#s3a} -------------------- We marked 29 bar-headed geese at QHL; 14 in 2007 and 15 in 2008 ([Table S2](#pone.0017622.s002){ref-type="supplementary-material"}) including 25 adults (14 male, 11 female) and 4 juveniles (2 male, 2 female). We obtained 20,150 Argos Doppler and 45,021 GPS locations ([Table S2](#pone.0017622.s002){ref-type="supplementary-material"}). Average transmitter lifespan was 10 months, but some PTTs performed for more than 2 years. GPS data was used for our analyses because they provided a large number of locations and lower spatial error compared with Argos data (15 m versus \>100 m error). Twenty-two of the 29 marked geese were sampled for avian influenza virus (sampling materials were unavailable for the first seven birds marked). However, none of the sampled birds tested positive for avian influenza virus. Migration from Qinghai Lake {#s3b} --------------------------- We defined the seasonal periods as: breeding (23 May--26 September; 126 days); fall migration (27 September--9 December; 74 days); wintering (10 December--5 March; 87 days); and spring migration (6 March--22 May; 78 days). We mapped migration routes, stopover sites, breeding locations, and wintering locations ([Fig. 1](#pone-0017622-g001){ref-type="fig"}). ::: {#pone-0017622-g001 .fig} 10.1371/journal.pone.0017622.g001 Figure 1 ::: {.caption} ###### Migration routes, stopover locations, breeding, and wintering areas for bar-headed geese marked at Qinghai Lake. Migration routes are shown for the 15 individuals (each in unique color) that completed at least one fall migration. White polygons represent stopover areas. Individual 82081 (red path, full path in inset) wintered in India. Individuals 74898 (orange) and 82084 (yellow) are highlighted in [Figure 3](#pone-0017622-g003){ref-type="fig"}. :::  ::: ### Breeding {#s3b1} Twenty-seven geese remained at QHL through the breeding season while 2 moved farther north ([Table 1](#pone-0017622-t001){ref-type="table"}): \#82084 moved to Hala Lake and \#67693 moved to Terkhiin Tsagaan Lake, Mongolia (within 200 km of H5N1 outbreaks in 2005, 2006, and 2009). Short movements (200--250 km) from the breeding grounds to molting locations occurred from 13 June to 19 September with most geese departing QHL in late June. The Zhaling-Eling Lake Region, a high-elevation (4290 m) wetland area 250 km southwest of Qinghai Lake was the most important post-breeding site ([Table 1](#pone-0017622-t001){ref-type="table"}, [Fig. 1](#pone-0017622-g001){ref-type="fig"}). ::: {#pone-0017622-t001 .table-wrap} 10.1371/journal.pone.0017622.t001 Table 1 ::: {.caption} ###### Breeding, post-breeding, spring and fall stopover, and wintering areas used by bar-headed geese captured at Qinghai Lake, 2007--2008. ::: {#pone-0017622-t001-1} Site Name Distance (km) and Direction from Qinghai Lake Latitude, Longitude (Degrees) Time Period[a](#nt101){ref-type="table-fn"} N[b](#nt102){ref-type="table-fn"} Date Range[c](#nt103){ref-type="table-fn"} Mean Length of Stay (range in days) Number of Telemetry Fixes ------------------------------------------------------------ ----------------------------------------------- ------------------------------- --------------------------------------------- ------------------------------------ -------------------------------------------- ------------------------------------- --------------------------- Terkhiin Tsagaan Lake Region, Arkhangai Province, Mongolia 1240 N 48.09 N, 99.92 E B07 1 5/7--6/5 29 49 Longbaotan Nature Reserve 570 SW 33.19 N, 96.47 E B/PB 08 1[d](#nt104){ref-type="table-fn"} 5/22--10/23 154 222 Hala Lake Region 260 NW 38.13 N, 97.95 E B08-09 2[d](#nt104){ref-type="table-fn"} 4/27--10/22 139 (103--178) 1,611 Qinghai Lake and Zhaling-Eling Lake Regions 0--350 SW 36.72 N, 99.02 E B/PB 07--09 27[d](#nt104){ref-type="table-fn"} 3/25--11/4 154 (7--223) 19,338 Hala Lake Region 260 NW 38.13 N, 97.95 E S08 1 5/13--5/13 1 3 Zhaling-Eling Lake Region 350 SW 34.93 N, 97.33 E F07-08 7[d](#nt104){ref-type="table-fn"} 9/28--11/4 11 (2--23) 783 S08-09 5 4/11--5/16 8 (2--29) 231 East of Wuli, Qumalai County 750 SW 34.63 N, 93.78 E S09 1 4/8--5/8 31 131 Zhamucuo Wetland/Sanjiangyuan Nature Reserve 850 SW 33.09 N, 93.66 E F07-08 10[d](#nt104){ref-type="table-fn"} 10/9--11/2 6 (1--11) 560 S08 1 4/30--5/5 6 46 North of Xagquka, Biru County 970 SW 32.06 N, 92.86 E S08 1 5/14--5/19 6 52 Selincuo Black-necked Crane Nature Reserve 1120 SW 31.65 N, 91.55 E F07-08 9[d](#nt104){ref-type="table-fn"} 10/17--11/11 7 (5--14) 839 S08-09 6 3/15--5/13 13 (3--38) 517 Namucuo Lake Region 1220 SW 30.50 N, 91.13 E F07-08 7 10/22--11/17 13 (4--22) 808 S09 2 3/14--4/18 9 (1--17) 178 Yamdrok Lake Region 1420 SW 28.84 N, 90.79 E F08 2 11/2--12/7 18 (4--32) 309 S09 1 3/11--4/7 28 229 Lhasa and Yarlung River Basins 1350 SW 29.40 N, 90.34 E W07-09 15[d](#nt104){ref-type="table-fn"} 10/29--4/24 123 (17--175) 16,632 Bhitarkonika National Park, Orissa, India 2330 SW 20.65 N, 86.91 E W08-09 1 12/9--3/6 87 1,001 a B = Breeding, PB = Post-breeding, S = Spring stopover, F = Fall stopover, W = Wintering area. b N is the number of unique individuals that used a site in a given time period. c Date ranges include arrival of first individual to departure of last individual. d Includes individuals that returned during a second cycle, second visit included in tally. ::: ### Fall migration {#s3b2} Mean fall departure date was 13 October (range 27 September--30 October). Thirteen geese initiated their migration from Zhaling-Eling Lake region while 2 left from QHL. All geese followed similar routes: in general, they flew along a southwesterly path with stopover locations at Zhamucuo wetland, Selincuo Black-necked Crane Reserve, and Namucuo Lake ([Table 1](#pone-0017622-t001){ref-type="table"}, [Fig. 1](#pone-0017622-g001){ref-type="fig"}). ### Wintering {#s3b3} Geese arrived at wintering areas from 29 October--9 December (mean = 9 November). All birds wintered within 100 km of Lhasa except for \#82081 which wintered south of the Himalaya at Bhitarkonika National Park in Orissa, eastern India. Lhasa wintering areas included the Pengbo River Basin (25 km northeast of Lhasa), the Lhasa River Basin (south of the cityflowing southwest for 50 km before meeting the Yarlung River), and the Yarlung River Basin (flowing eastward through Gongga then south becoming the Brahmaputra River) ([Fig. 1](#pone-0017622-g001){ref-type="fig"}). ### Spring migration {#s3b4} The mean spring departure date for the geese was 1 April (range 6 March--April). Spring stopover locations included the Namucuo Lake region and Zhaling-Eling Lakes ([Table 1](#pone-0017622-t001){ref-type="table"}, [Fig. 1](#pone-0017622-g001){ref-type="fig"}). The mean breeding arrival date was 29 April (range 28 March--22 May). Seven geese returned to breeding locations used the previous year (six to Qinghai Lake and one to Hala Lake). Two geese used different breeding locations in 2008: \#74901 used Longbaotan wetland 570 km southwest of QHL; and \#74902 used the Hala Lake region, 260 km northwest of QHL. The spring migration pathway (n = 9, orange line) was broader than the fall route (n = 15, yellow line) despite the fact that it was represented by fewer individuals. ### Migration rates and stopover duration {#s3b5} Ninety-four migration legs were recorded (fall: 65 legs flown by 15 geese; spring: 29 legs flown by 9 geese). Migration was rapid -- the longest leg (1158 km) was completed within 5.1 days ([Table 2](#pone-0017622-t002){ref-type="table"}). Individual migrations included an average of two stopovers and were completed in \<1 month (fall migration = 26 d; spring migration = 29 d). Migration rates (including time at stopovers) for the fall and the spring did not differ significantly (17 and 14 km/h, respectively; p = 0.23). Distance between QHL and the wintering grounds was 1300 km for 14 geese wintering near Lhasa and 2300 km for goose \#82081 that wintered in India. ::: {#pone-0017622-t002 .table-wrap} 10.1371/journal.pone.0017622.t002 Table 2 ::: {.caption} ###### Migration chronology and movement rates of bar-headed geese marked at Qinghai Lake during fall and spring migration, 2007--2008. ::: {#pone-0017622-t002-2} Fall Migration Spring Migration Longest Leg[a](#nt105){ref-type="table-fn"} ------- --------- ---------------- ------------------ --------------------------------------------- --- --------- --------- ---- --- -------- ----- ------- ------ 67582 3/25/07 10/10/07 11/13/07 35 3 4/7/08 4/22/08 16 2 534.1 1.2 457.8 1250 67690 3/25/07 10/10/07 11/3/07 25 2 293.6 1.3 234.9 1264 67695 3/25/07 10/12/07 11/1/07 21 2 4/24/08 5/10/08 17 2 815.5 5.8 141.8 1385 10/9/08 10/29/08 21 2 67698 3/31/07 10/24/07 10/31/07 8 1 716.7 2.6 277.4 1382 74898 3/30/07 10/28/07 11/3/07 7 1 500.2 1.1 461.7 1316 74900 3/31/07 10/30/07 11/13/07 15 1 571.5 0.8 685.8 1454 74901 3/31/07 10/1/07 11/11/07 42 2 4/6/08 5/22/08 47 2 584.9 0.7 877.3 1233 10/23/08 10/29/08 7 1 74902 3/30/07 10/24/07 11/17/07 25 2 4/8/08 5/13/08 36 2 981.6 1.6 620.0 1239 82079 4/2/08 10/20/08 11/12/08 24 1 3/25/09 3/28/09 4 0 1158.4 5.1 227.9 1235 82080 4/2/08 9/30/08 11/5/08 37 4 3/24/09 1 708.0 2.3 314.7 1329 82081 4/1/08 10/14/08 12/9/08 57 3 3/6/09 4/16/09 42 3 1036.0 4.3 239.1 2336 82082 3/30/08 9/30/08 11/15/08 47 4 4/1/09 4/27/09 27 2 780.1 2.3 346.7 1285 82084 3/30/08 10/22/08 11/9/08 19 2 4/5/09 5/21/09 47 2 642.8 2.1 308.5 1270 82085 3/30/08 9/27/08 11/6/08 41 4 342.6 1.1 316.2 1362 82086 3/31/08 10/15/08 10/29/08 15 1 3/14/09 4/9/09 27 2 950.7 4.9 193.4 1252 a Longest flight between two consecutive stationary areas. b from northern-most to southern-most points. ::: HPAI H5N1 outbreaks on the Qinghai-Tibet Plateau {#s3c} ------------------------------------------------ Sixteen outbreaks were reported on the Qinghai-Tibet Plateau from 2003--2009; nine in wild birds and seven in poultry ([Table S1](#pone.0017622.s001){ref-type="supplementary-material"}). Fourteen outbreaks (87.5%) were located within the BHGO UD ([Fig. 2a](#pone-0017622-g002){ref-type="fig"}). All poultry cases occurred near Lhasa, whereas wild bird outbreaks were clustered around QHL and between Lhasa and QHL ([Fig. 2a](#pone-0017622-g002){ref-type="fig"}). Poultry outbreaks occurred in chickens and wild bird outbreaks occurred in several waterbird species including bar-headed geese, brown-headed gulls, great black-headed gulls, great cormorants, ruddy shelducks and great-crested grebes (*Podiceps cristatus*). However, the bar-headed goose was the primary species infected during the outbreaks, both in the total numbers killed and the frequency of outbreaks. ::: {#pone-0017622-g002 .fig} 10.1371/journal.pone.0017622.g002 Figure 2 ::: {.caption} ###### Brownian bridge utilization distributions in relation to poultry density and HPAI H5N1 outbreaks. Brownian bridge utilization distributions (A) describe fall (yellow; 27 Sep--9 Dec) and spring (orange; 6 Mar--22 May) goose migrations. Fixed kernel home ranges depict (B) population level breeding and post-breeding areas (C) and wintering areas, with only locations near outbreaks shown. Brown shading indicates poultry densities. H5N1 outbreak events in wild birds (white) and poultry (black) are indicated for 2003--2009. Two shading levels indicate isopleths containing 95% (red) and 99% (yellow-orange) of total locations. :::  ::: ### Habitat use and overlap with poultry and captive-reared geese {#s3c1} Habitat use varied seasonally ([Table 3](#pone-0017622-t003){ref-type="table"}). Breeding season habitats included natural wetlands (54% of locations) and grasslands (35%). Grasslands were primarily used during migration (78% fall, 77% spring). In the winter, a combination of natural wetlands (34%) and agricultural fields (39%) were used by the geese ([Table 3](#pone-0017622-t003){ref-type="table"}, [Fig. 3](#pone-0017622-g003){ref-type="fig"}). Poultry densities at goose locations were near zero during all seasons except the winter which averaged 35.8 poultry head per km^2^ ([Table 3](#pone-0017622-t003){ref-type="table"}). Potential exposure of geese to H5N1 occurred on the wintering grounds as was evidenced by their spatial and temporal overlap with a poultry outbreak in January 2008 ([Fig. 3a](#pone-0017622-g003){ref-type="fig"}) and with a bar-headed goose captive-rearing facility in 2009 ([Fig. 3b](#pone-0017622-g003){ref-type="fig"} and [4](#pone-0017622-g004){ref-type="fig"}). ::: {#pone-0017622-g003 .fig} 10.1371/journal.pone.0017622.g003 Figure 3 ::: {.caption} ###### Concurrent use of natural wetlands and agricultural fields by wintering bar-headed geese near Lhasa. \(A) Winter movements for goose 74898 (3 November 2007--2 April 2008; 1205 locations) in relation to a confirmed HPAI H5N1 outbreak in chickens on 21 January, 2008 (black circle). (B) Winter movements (9 November 2008--5 April 2009; 961 locations) for goose 82084 in relation to a captive bar-headed goose farm (red circle). :::  ::: ::: {#pone-0017622-g004 .fig} 10.1371/journal.pone.0017622.g004 Figure 4 ::: {.caption} ###### Bar-headed goose farming in Tibet. \(A) Captive bar-headed geese in Gonggar County, Tibet as shown in a December 2007 China Tibet Information Center article (Wu 2007). (B) Wild bar-headed geese (foreground) shown in close proximity to a captive bar-headed goose farm (blue building in background) in a January 2007 photo from an anonymous source. Approximately 250 bar-headed geese were counted in outdoor net pens attached to the building (out of view in photo B). Approximate location of this farm is shown in a red circle ([Fig. 3b](#pone-0017622-g003){ref-type="fig"}). :::  ::: ::: {#pone-0017622-t003 .table-wrap} 10.1371/journal.pone.0017622.t003 Table 3 ::: {.caption} ###### Percent habitat type, poultry density, and human population densities at 43,841 satellite telemetry locations in China for 29 bar-headed geese marked at Qinghai Lake, 2007--2008. ::: {#pone-0017622-t003-3} Percent Habitat Type ------------------------ ---------------------- ------ ----- ------ ----- ------ ------ ------- Breeding/Post-breeding 0.2 34.9 0.0 53.6 0.1 11.1 0.6 0.7 Fall Migration 0.3 78.4 0.0 15.0 0.0 6.3 0.6 1.1 Spring Migration 0.0 76.6 0.0 12.3 0.0 11.1 0.5 1.5 Winter 38.5 25.0 0.2 33.6 0.8 1.9 35.8 134.9 ::: ### Poultry outbreak risk factors {#s3c2} The AIC analysis for poultry outbreaks indicated two competing top models ([Table 4](#pone-0017622-t004){ref-type="table"}): Model 1 included poultry density and cropland (AIC weight *w~i~* = 0.72) while Model 2 included poultry density, cropland, latitude, and BHGO UD (*w~i~* = 0.26). The top model indicated that domestic outbreak locations were explained by measures of poultry density and cropland area. The CART analysis separated outbreaks into those above 66 poultry per km^2^ and above 0.3 ha of cropland per km^2^ ([Fig. 5a](#pone-0017622-g005){ref-type="fig"}). Although the BHGO UD was not a strong indicator in predicting locations of poultry outbreaks, exposure of BHGO to H5N1 virus via outbreaks in poultry on the wintering grounds was confirmed (*a priori* univariate logistic regression, BHGO UD, p = 0.032). ::: {#pone-0017622-g005 .fig} 10.1371/journal.pone.0017622.g005 Figure 5 ::: {.caption} ###### Classification and regression tree describing poultry (A) and wild bird (B) outbreaks on the Qinghai-Tibet Plateau. Numbers of random points (numerator), outbreak locations (denominator), and percentage of total sample are reported at each terminal node. :::  ::: ::: {#pone-0017622-t004 .table-wrap} 10.1371/journal.pone.0017622.t004 Table 4 ::: {.caption} ###### Akaike Information Criterion (AIC) best-fit model results for ten a priori models of domestic poultry and wild bird H5N1 outbreaks during 2003--2009 (n = 7 domestic, n = 9 wild) compared with random points drawn from minimum convex polygons per season (n = 70 domestic, n = 90 wild) on the Qinghai-Tibet Plateau. ::: {#pone-0017622-t004-4} Model K ΔAIC Likelihood *w~i~* ---------------------------------------------------------- --- ------- ------------ -------- **Poultry Outbreaks** [a](#nt107){ref-type="table-fn"} PD + Crop 3 0.00 1.00 0.72 PD + Crop + Lat + BHGO UD 5 2.05 0.36 0.26 PD + Lat 3 9.34 0.01 0.01 PD + Lat + BHGO UD 4 9.80 0.01 0.01 PD 2 11.37 0.00 0.00 PD + BHGO UD 3 11.78 0.00 0.00 Crop 2 14.35 0.00 0.00 Lat 2 15.22 0.00 0.00 BHGO UD 2 21.92 0.00 0.00 null 0 25.07 0.00 0.00 **Wild Bird Outbreaks** [b](#nt108){ref-type="table-fn"} Wetl + Lat 3 0.00 1.00 0.70 Lat 2 2.29 0.32 0.22 Wetl 2 6.41 0.04 0.03 Wetl + PD 3 8.17 0.02 0.01 Wetl + Gras 3 8.31 0.02 0.01 Wetl + BHGO UD 3 8.36 0.02 0.01 null 0 9.47 0.01 0.01 Gras 2 10.83 0.00 0.00 PD 2 11.07 0.00 0.00 BHGO UD 2 11.44 0.00 0.00 a Covariates for poultry model were bar-headed goose utilization distributions (BHGO UD), poultry density (PD), latitude (Lat), and cropland (Crop). b Covariates for wild bird models were BHGO UD, PD, Lat, grassland (Gras), and wetlands (Wetl). K = number of model parameters, ΔAIC = AIC differences, *w~i~* = Akaike weights. ::: ### Wild bird outbreak risk factors {#s3c3} Two models best fit the data to explain locations of H5N1 outbreaks in wild birds on the Qinghai-Tibet Plateau. The top model (*w~i~* = 0.70) included latitude and wetlands while the second-ranked model (*w~i~* = 0.22) included latitude only ([Table 4](#pone-0017622-t004){ref-type="table"}). CART results indicated differences at 36.14°N with wetland area larger than 0.3 ha ([Fig. 5b](#pone-0017622-g005){ref-type="fig"}). Contrary to our hypothesis, BHGO UD was not a significant predictor for wild bird outbreaks which tended to occur in remote regions. However, outbreaks were concentrated north of 36°N (near QHL), with fewer outbreaks located between QHL and Lhasa ([Fig. 2a](#pone-0017622-g002){ref-type="fig"}) where the BHGO UD is geographically broad and with lower use per pixel. In this region, the BHGO UD does not provide probabilities sensitive to predicting outbreak versus non-outbreak locations given the low number of total outbreaks. Thus, we conclude that while bar-headed geese are an important species in H5N1 events [@pone.0017622-Chen1], [@pone.0017622-Liu1], [@pone.0017622-Chen2], [@pone.0017622-Lei1], the BHGO UD is not a strong predictor across the Qinghai-Tibet Plateau, except near QHL where there are high concentrations of this species on the breeding grounds and greater number of outbreaks. ### Timing of outbreaks {#s3c4} We found a temporal lag between peak seasons for poultry and wild bird outbreaks on the Qinghai-Tibet Plateau (Fisher\'s Exact Test, p = 0.008, [Fig. 6a](#pone-0017622-g006){ref-type="fig"}). Poultry outbreaks occurred during the winter, spring and breeding seasons, whereas wild bird outbreaks were not reported until the spring and breeding months ([Fig. 6b](#pone-0017622-g006){ref-type="fig"}, [Table S1](#pone.0017622.s001){ref-type="supplementary-material"}). More poultry outbreaks were found during the winter and the spring than expected (p = 0.079) and subsequently during the spring and the breeding seasons for wild birds (p = 0.017). In contrast, fewer outbreaks than expected occurred in the breeding and the fall for poultry, and the fall and the winter for wild birds ([Fig. 6b](#pone-0017622-g006){ref-type="fig"}). The first H5N1 outbreak reported on the Qinghai-Tibet Plateau occurred in poultry near Lhasa in February 2004, followed in the spring (Apr--Jun) of 2005 with the epizootic at QHL. Both results from [Fig. 6](#pone-0017622-g006){ref-type="fig"} and timing of the first two outbreaks reported on the Qinghai-Tibet Plateau suggest initial outbreaks in poultry, followed by outbreaks in wild birds. ::: {#pone-0017622-g006 .fig} 10.1371/journal.pone.0017622.g006 Figure 6 ::: {.caption} ###### Comparison by season of wild and domestic outbreaks on the Qinghai-Tibet Plateau. \(A) Observed versus expected values differed for both poultry and wild bird outbreaks when tested separately (Fisher Exact Test, poultry P = 0.079, wild P = 0.017). (B) Temporal distribution of wild bird outbreaks differed from domestic bird outbreaks (Fishers Exact Test, P = 0.008). Expected numbers were calculated under the assumption that outbreaks are proportional to the number of days within the seasonal period (Winter = 87 d, Spring Migration = 78 d, Breeding = 126 d, and Fall Migration = 74 d). :::  ::: Phylogenetics {#s3d} ------------- Phylogenetic analyses indicated that HPAI H5N1 isolates from clades 2.2 and 2.3 infected migratory waterfowl including the bar-headed goose and poultry on the Qinghai-Tibet Plateau. The directionality of virus transmission from domestic to wild birds or vice-versa was however, difficult to ascertain based on the limited number of sequences available within this geographic region. Viruses isolated from bar-headed geese formed distinct sub-clades based on year of outbreak. For instance, isolates from 2005 grouped within sub-clade 2.2.1, and similarly isolates from 2006 and 2008 grouped within sub-clade 2.2.2 and 2.3.2, respectively ([Fig. 7](#pone-0017622-g007){ref-type="fig"}). This pattern suggested reintroduction of the virus into bar-headed goose populations, rather than continuous circulation or persistence. Isolates from A/bar-headed goose/Tibet/8/06 and the A bar-headed goose/Qinghai/F/06 formed a monophyletic grouping within sub-clade 2.2.2 and showed high bootstrap support (80%). The evolutionary distance between these two isolates was 0.008 base substitutions per site, lower than the overall average (0.046) for the 38 isolates representative of HPAI H5N1 divergence since 1996. ::: {#pone-0017622-g007 .fig} 10.1371/journal.pone.0017622.g007 Figure 7 ::: {.caption} ###### Phylogenetic relationships of HPAI H5N1inferred by neighbor-joining analysis based on 1550 bp fragment of the HA gene. Viruses isolated from the bar-headed goose are highlighted (blue) and the monophyletic grouping of isolates from Tibet and Qinghai are indicated by a symbol (♦). :::  ::: Discussion {#s4} ========== We hypothesized that if bar-headed geese were a significant contributor in the spread of H5N1 to QHL, certain conditions would have had to occur: (1) exposure of geese to virus sources prior to arriving at QHL, (2) onset of migration and viral shedding before individuals became physiologically compromised, and (3) evidence of genetic relatedness between isolates from virus source regions frequented by geese and QHL. Our findings indicate that these conditions were met; however, we discuss caveats in our findings based on the most current H5N1 disease ecology. Both the timing of migratory movements and distribution of wintering habitat used by geese around Lhasa exposed them to H5N1 sources (Condition 1) as was evidenced in four ways: (a) location of outbreaks within the BHGO UD ([Fig. 2a](#pone-0017622-g002){ref-type="fig"}); (b) detailed movements showing proximity in timing and spatial distribution between marked geese, H5N1 outbreak in chickens, and a captive bar-headed goose farm ([Fig. 3](#pone-0017622-g003){ref-type="fig"} and [4](#pone-0017622-g004){ref-type="fig"}); (c) temporal analysis showing peaks in poultry outbreaks followed by peaks in wild bird outbreaks ([Fig. 6](#pone-0017622-g006){ref-type="fig"}); (d) and a significant *a priori* logistic regression between BHGO UD and poultry outbreaks ([Table 5^a^](#pone-0017622-t005){ref-type="table"}). The evidence for goose exposure to H5N1 sources on the Lhasa wintering grounds is consistent. What was unexpected is the discovery of Lhasa as a major wintering area for geese from QHL and the potential of the region as an important H5N1 transmission route -- both of which have conservation implications for this species. Prior to this work, little was known about migration of breeding and staging waterbirds at QHL [@pone.0017622-Miyabayashi1], [@pone.0017622-Zhang1]. Of 800 bar-headed goose banding records from the lake in 1987 and 1988, only three returns have been reported (one each in northeast and southwest India [@pone.0017622-Uttangi1] and one in Chittagong, southern Bangladesh [@pone.0017622-National1]). Bishop et al. [@pone.0017622-Bishop1] reported that 25% of the global bar-headed goose population winters on the southern Qinghai-Tibet Plateau, however, breeding connectivity for these individuals was previously unknown. Hypotheses regarding the epidemiology of the event at QHL suggested by previous authors [@pone.0017622-Chen2], [@pone.0017622-Chen3], [@pone.0017622-Liang1], [@pone.0017622-Kilpatrick1] were based on inadequate migration data available at the time, and associated models therefore did not examine Tibet as a potential H5N1 transmission link to QHL. For example, Chen et al. [@pone.0017622-Chen3] and Liang et al. [@pone.0017622-Liang1] hypothesized that wild birds transported the virus to QHL from Poyang Lake, southeastern China, where 6 apparently healthy wild ducks (species not reported) tested positive for HPAI H5N1. Results from our study and related work at Poyang Lake [@pone.0017622-Takekawa1], which examined migration patterns of 9 waterfowl species (n = 62), showed no evidence of migratory connectivity between the 2 lakes located in separate flyways. These findings affirmed the importance for including knowledge of host ecology to inform the debate on wild birds and HPAI transmission. ::: {#pone-0017622-t005 .table-wrap} 10.1371/journal.pone.0017622.t005 Table 5 ::: {.caption} ###### Top-ranked Akaike Information Criterion (AIC) models with parameter values for domestic poultry and wild bird H5N1 outbreaks on the Qinghai-Tibet Plateau. ::: {#pone-0017622-t005-5} Model[a](#nt110){ref-type="table-fn"} Parameter[b](#nt111){ref-type="table-fn"} *w~i~* β SE z-value Pr(\>\|z\|) Pr(\>\|×^2^\|) --------------------------------------- ------------------------------------------- -------- --------- -------- --------- ------------- ---------------- Poultry Outbreaks 1 Intercept 0.72 −6.451 2.714 −2.377 0.018 **--** PD 0.088 0.042 2.122 0.034 \<0.001 Crop 0.001 0.000 2.167 0.030 \<0.001 Poultry Outbreaks 2 Intercept 0.26 11.660 17.780 0.656 0.512 -- PD 0.078 0.043 1.826 0.068 \<0.001 Crop 0.001 0.000 1.968 0.049 \<0.001 Lat −0.249 0.494 −0.503 0.615 0.480 BHGO UD −10.280 9.393 −1.094 0.274 0.228 Wild Outbreaks 1 Intercept 0.70 −17.330 5.633 −3.076 0.002 **--** Lat 0.045 0.166 2.696 0.007 0.004 Wetl 0.000 0.000 2.222 0.026 0.024 Wild Outbreaks 2 Intercept 0.22 −17.476 5.463 −3.199 0.001 -- Lat 0.462 0.161 2.860 0.004 0.003 a *A priori* analysis of poultry outbreaks and BHGO UD indicated BHGO exposure to H5N1 sources on the wintering grounds; Single variable *a priori* poultry model: BHGO UD (p = 0.032). b Parameters for poultry models included bar-headed goose utilization distribution (BHGO UD), poultry density (PD), latitude (Lat), and cropland (Crop). Parameters for wild bird models include BHGO UD, PD, Lat, grassland, and wetlands (Wetl). *w~i~* = Akaike weights, β = model coefficients, SE = standard error (SE), Pr(\>\|z\|) = significance value, Pr(\>\|×^2^\|) = Chi-square goodness of fit test. ::: A clear understanding of poultry farming in this remote region is also critical to assessing H5N1 transmission potential. There has been confusion over whether captive bar-headed goose farms exist at QHL, and whether such farms provided introduction of H5N1 virus to this region generally lacking poultry [@pone.0017622-Butler1]. Through a combination of local investigation, communication with experts, and detailed review of the original blog postings (in Chinese) regarding the farms, we have learned that the farms are not located at QHL, but instead exist 1200 km south of QHL near Lhasa. These conclusions are independently supported by [@pone.0017622-Feare1]. The largest of the captive bar-headed goose breeding facilities discussed above is included in our study ([Fig. 3b](#pone-0017622-g003){ref-type="fig"} and [4](#pone-0017622-g004){ref-type="fig"}), where one of our marked geese spent the winter foraging in the local fields and wetlands. We found evidence suggesting that migratory movement could occur before the virus impaired a bird\'s ability to migrate (Condition 2). Recent challenge studies identified the average duration of asymptomatic virus shedding to be 6.5. days in experimentally infected bar-headed geese (A/whooper swan/Mongolia/244/2005H5N1) [@pone.0017622-Brown1]. Geese from QHL averaged 380 km per day (range 294 km in 1.3 days to 1158 km in 5.1 days; [Table 2](#pone-0017622-t002){ref-type="table"}), indicating a capacity to move the distance between QHL and Lhasa within the asymptomatic period of virus shedding. Migration for most individuals, however, did not occur without time spent at stopover locations, which increased total migration time to an average of one month. Thus, while movement of virus the entire distance (within the asymptomatic period) is possible, as was observed for one bar-headed goose, a more likely mode of transmission is through a relay effect [@pone.0017622-Gaidet1]. This would occur if a virus is transported to stopover locations where it can spread among individuals within concentrated areas of feeding and roosting [@pone.0017622-Alexander3], [@pone.0017622-Brown2] or through environmental persistence [@pone.0017622-Stallknecht1], [@pone.0017622-Stallknecht2], and then be forwarded to newly infected individuals along the migratory pathway. We have evidence that stopover locations used by marked geese in this study ([Table 1](#pone-0017622-t001){ref-type="table"}, [Fig. 1](#pone-0017622-g001){ref-type="fig"}) are important both for bar-headed geese as well as a number of other important waterbird species (Y.S. Hou, unpubl. data). In addition, recent studies [@pone.0017622-Costa1], [@pone.0017622-Fereidouni1] have indicated that mallards *(Anas platyrhynchos)* with previous exposure to homologous LPAI viruses may remain healthy enough to migrate, and this might apply to other waterfowl such as bar-headed geese. The first H5N1 outbreak on the Plateau occurred in chickens in Lhasa in February, 2004 ([Table S1](#pone.0017622.s001){ref-type="supplementary-material"}), and was followed by the QHL wild bird epizootic in spring of 2005. Unfortunately, none of the seven reported poultry outbreaks have sequences available in open sources such as GenBank. Release of these data would improve our analyses by allowing us to test relationships between isolates from poultry and wild birds from the 2 regions. However, Li et al. [@pone.0017622-Li2], in an updated analysis of H5N1 virus evolution in China, includes one sequence from a 2008 outbreak in chickens from a live-bird market (CH/TB/6/08). Here they discovered high sequence similarity between the 2005 QHL epizootic (bar-headed goose isolates, BHG/QH/3/05) and chicken isolates from Tibet in one of the eight H5N1 genes (PB2), supporting a connection between the Lhasa and QHL outbreaks (Condition 3). In addition, our phylogenetic analysis of publicly available sequences included 2 isolates from wild birds in Tibet: a bar-headed goose (A/BHG/TB/8/06) and great cormorant (A/GC/TB/12/06), both from 2006. The analysis indicated that geese were first infected by HPAI H5N1 belonging to clade 2.2 that emerged at QHL during 2005. Since this time, the virus has continued to infect bar-headed geese periodically within the Central Asian Flyway, most recently during the 2009 outbreak in Mongolia. Goose isolates from the QHL and Tibet outbreaks in 2006 formed a monophyletic group within sub-clade 2.2.2 and were closely related, providing additional evidence that migratory birds were agents of transmission between these two outbreak sites in China. Further analysis of isolates in combination with remaining unreleased sequences would help elucidate the route by which viruses moved from southeastern China to QHL including the directionality of virus transmission between poultry and wild birds [@pone.0017622-Chen1], [@pone.0017622-Chen2], [@pone.0017622-Chen3], [@pone.0017622-Lei1]. Lhasa provides a unique situation for intermixing and potential transfer of disease between wild and domestic birds. Each winter, the Lhasa region experiences increased concentrations of humans, poultry, and wild birds when nomadic herders return to populated centers, chicken production peaks preceding the Chinese New Year festivities in late January [@pone.0017622-Pfeiffer1], and up to 50% of the global population of bar-headed geese winter in sheltered river valleys surrounding Lhasa [@pone.0017622-Bishop1], [@pone.0017622-Bishop2], [@pone.0017622-Lang1]. H5N1 outbreaks in domestic birds spiked in frequency under such winter conditions [@pone.0017622-OIE1] followed by an increase in wild bird outbreaks during the spring and breeding seasons ([Fig. 6](#pone-0017622-g006){ref-type="fig"}). The H5N1 situation involving wild birds in the Central Asian Flyway is unique relative to results from studies in the East Asian Flyway (EAF) along the Pacific coast [@pone.0017622-Takekawa1], [@pone.0017622-Newman1]. The EAF boasts some of the world\'s most productive poultry systems including rice-paddy duck farming in China and parts of Southeast Asia [@pone.0017622-Muzaffar2]. Whooper swans marked in eastern Mongolia within the EAF demonstrated spatial proximity to poultry outbreaks in Korea and north-eastern China; however, a lack of correspondence in timing and micro-habitat use precluded the likelihood of transmission between the two groups [@pone.0017622-Newman1]. Eight duck species marked at Poyang Lake in south-eastern China also showed a temporal mismatch between H5N1 outbreaks and arrival of wild ducks to the wintering grounds [@pone.0017622-Takekawa1]. Poultry production in the Central Asian Flyway is extensive in the south, particularly in India and Bangladesh, and limited to absent in the central and northern sections (Tibet and northward). H5N1 outbreaks along the flyway mirror human and poultry densities whereby more domestic bird outbreaks occur in the south and wild bird outbreaks in the sparsely populated north [@pone.0017622-OIE1]. The Qinghai-Tibet Plateau lies within the transition zone along this gradient, and it appears that bar-headed geese may be an important vector in H5N1 spread as evidenced by the size of the infected population at Qinghai Lake [@pone.0017622-Liu1], [@pone.0017622-Chen2]. Our study identifies QHL and Lhasa as important linkages between wild and domestic transmission of H5N1 and provides new supporting information regarding the role of wild birds in long distance spread of this virus. Further investigation of wild birds and H5N1 transmission within the Central Asian Flyway will increase our understanding of how wild birds may contribute to virus circulation and the unique pattern of outbreaks in this remote region. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **Confirmed HPAI H5N1 outbreaks in wild birds and domestic poultry on the Qinghai--Tibet Plateau, 2003--2009.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **PTT performance (as of 1 Sep 2009) for 29 bar-headed geese marked in 2007--2008 at Qinghai Lake, China.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors thank Sue Haseltine, Patti Bright, Steve Schwarzbach, and Judd Howell of USGS and Joseph Domenech and Juan Lubroth of FAO for support of this project. We are grateful to the QLNNR staff, Qinghai Forestry Bureau (Sandan Li), Chinese Academy of Sciences (Xudong Hu, Lianling Hu, Ning Kong, Ze Luo), U. S. Embassy (Bill Chang, Dan Jassem, Qihong Qian), and USGS (Shane Heath, Kyle Spragens) for field and logistical support. We thank Chris Hamilton and Julio Pinto for providing outbreak data from the EMPRES-i database; and Julie Yee (USGS WERC) and Shay Howlin (West-Inc) for statistical advice. Mike Erwin and anonymous reviewers provided helpful comments to strengthen earlier versions of this manuscript. We thank Mary Anne Bishop (International Crane Foundation) for providing local information on the Lhasa region. The use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded by the United States Geological Survey (Patuxent Wildlife Research Center, Western Ecological Research Center, Alaska Science Center, and Avian Influenza Program); the United Nations FAO, Animal Production and Health Division, EMPRES Wildlife Unit; National Science Foundation Small Grants for Exploratory Research (No. 0713027); and the Chinese Academy of Sciences (No. 2007FY210700, INFO-115-D02, KSCX2-YW-N-063 and 2005CB523007). The founders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: DJP JYT SHN. Performed the experiments: DJP PC JYT MT YH BMC BY TL YL FL SG ZX YH YZ DCD WMP. Analyzed the data: DJP BMC NJH DCD WMP. Contributed reagents/materials/analysis tools: DJP PC JYT MT YH BMC BY TL YL FL SG ZX YH YZ DCD WMP SHN. Wrote the paper: DJP NJH JYT SHN.

PubMed Central

2024-06-05T04:04:19.216093

2011-3-9

PMC3052366

Introduction {#s1} ============ 3-Deoxy-*D*-manno-octulosonate 8-phosphate (KDO8P) synthase (KDO8PS) is a bacterial enzyme that synthesizes KDO8P from phosphoenolpyruvate (PEP) and arabinose 5-phosphate (A5P). This reaction is of significant biological relevance, as KDO8P is the phosphorylated precursor of KDO, which is an essential component of the endotoxin of Gram negative bacteria [@pone.0017459-Raetz1]. The enzyme exists in two forms, differing in the requirement (or lack thereof) of a divalent metal for activity [@pone.0017459-Duewel1]. We and others have determined high-resolution structures of both metallo- and non-metallo forms of KDO8PS [@pone.0017459-Duewel1], [@pone.0017459-Wagner1], [@pone.0017459-Radaev1], [@pone.0017459-Duewel2], [@pone.0017459-Cochrane1], which revealed the features of the active site: in metal dependent KDO8PSs the divalent metal ion (Zn^2+^, or Fe^2+^, [@pone.0017459-Kona1]) is coordinated by the side chains of a cysteine, a histidine, a glutamic acid, and an aspartic acid. In non-metallo KDO8PSs an asparagine replaces the metal binding cysteine and there is no metal ion. Several mutagenesis studies [@pone.0017459-Cochrane1], [@pone.0017459-Oliynyk1], [@pone.0017459-Shulami1], [@pone.0017459-Li1], [@pone.0017459-Kona2], [@pone.0017459-Allison1] and more recently, quantum mechanical simulations [@pone.0017459-Kona1], [@pone.0017459-Tao1], [@pone.0017459-Tao2], have established that substitution of the metal binding cysteine with asparagine is absolutely required to convert metal-dependent to metal-independent KDO8PSs, although additional changes in the highly conserved CysAspGlyPro motif of the loop that contains the metal binding aspartic acid are necessary to achieve appreciable levels of activity [@pone.0017459-Cochrane1], [@pone.0017459-Allison1]. These observations are intriguing and raise the possibility that during the evolution of KDO8PS, after the initial choice between cysteine + metal and asparagine, additional sequence divergence occurred primarily to maximize the stability of the protein in the environment of hundreds of different bacteria. In this study we aimed to establish if the climb to stability fitness was a key factor in determining the amino acid sequence of KDO8PSs, and whether it affected only conserved positions or also positions with a high propensity for coevolution. A new combination of tools from information theory and structural modeling provided an avenue to quantify the contribution of coevolving residues to the stability of KDO8P synthase: this methodology may be of general value in the study of all protein families. Results {#s2} ======= The stability landscape of KDO8PS {#s2a} --------------------------------- From the point of view of protein stability the organization of enzyme active sites is inherently unstable because these sites are optimized for catalysis, which means they are pre-organized to stabilize the transition state(s), rather than the protein [@pone.0017459-Warshel1], [@pone.0017459-Roca1]. Thus, the substitution of a catalytic side chain (most often to alanine) will typically increase the overall protein stability, while sacrificing function [@pone.0017459-Nagatani1], [@pone.0017459-Beadle1]. Conversely, most mutations that introduce a new function are destabilizing [@pone.0017459-Wang1], [@pone.0017459-Bloom1]. The generality of this stability-function tradeoff must be viewed within the context of the fact that regardless of their effect on functions most mutations are destabilizing [@pone.0017459-Matthews1], [@pone.0017459-Guerois1], [@pone.0017459-Schymkowitz1], [@pone.0017459-Tokuriki1]. With respect to KDO8PS, several attempts were made (e.g., [@pone.0017459-Cochrane1], [@pone.0017459-Kona2]) to map the evolutionary paths from metallo- to non-metallo forms (and vice versa) without giving sufficient consideration to the fact that sequence differences between and within these forms in different bacterial backgrounds, may be related not only to activity but also to stability. One way we looked at the relevance of individual residues in the context of the three-dimensional structure of KDO8PSs was by computing a Hidden Markov Model (HMM) for the entire KDO8PS protein family (multiple sequence alignment [MSA S1](#pone.0017459.s006){ref-type="supplementary-material"}, Supporting Information). This family contains the 'C23' metallo sub-family (175 sequences with Cys at position 23 of *Neisseria meningitidis* (*Nm.*) KDO8PS (Uniref Q9JZ55, PDB 2QKF), used here as reference [@pone.0017459-Cochrane1]), and the 'N23' non-metallo sub-family (173 sequences with Asn at position 23). Only the positions of the MSA corresponding to the sequence of *Nm.* KDO8PS ([MSA S2](#pone.0017459.s007){ref-type="supplementary-material"}, Supporting Information) were used in the calculation of the HMM profile. A HMM profile specifies a probability distribution over the alphabet of the 20 common amino acids, taking into consideration the background frequency for each amino acid, computed by counting amino acid occurrences in all known proteins, or only in the proteins of the family under consideration [@pone.0017459-Eddy1],[@pone.0017459-Sonnhammer1]. If the background frequency of amino acid *j* is π*~j~*, then the important positions are those whose distribution differs from π. Therefore, the relative entropy between the observed distribution *P* ~i~ at the *i*-th position of the profile and π, *H*(*P* ~i~ \|\| π) = *I*(*P* ~i~), defines the information content of *P* ~i~. This information can be visualized as a HMM histogram or vector ([Figure 1A](#pone-0017459-g001){ref-type="fig"}). In practice, the height (relative entropy  =  information content) of the histogram at each position indicates how much the observed frequency of amino acids at that position deviates from the background frequency. In the HMM vector of [Figure 1A](#pone-0017459-g001){ref-type="fig"}, position numbers refer to the reference sequence of *Nm.* KDO8PS [@pone.0017459-Cochrane1]. There is a "wavy" pattern in the histogram derived from the MSA of all KDO8PSs (1^st^ inset of [Figure 1A](#pone-0017459-g001){ref-type="fig"}), which is clearly evident in the superimposed Loess [@pone.0017459-Cleveland1], [@pone.0017459-Cleveland2] fit. Regions of the histogram higher than the mean value are marked with an orange bar below the sequence; these regions invariably include the strands of the β-barrel in the structure ([Figure 1B](#pone-0017459-g001){ref-type="fig"}), a clear indication that the sequence of the interior of the protein is less determined by chance. The HMM vectors of metallo (C23) or non-metallo (N23) KDO8PSs, calculated independently, are shown in the 2^nd^ inset of [Figure 1A](#pone-0017459-g001){ref-type="fig"}. Their difference is shown in the 3^rd^ inset of [Figure 1A](#pone-0017459-g001){ref-type="fig"}. The difference vector identifies positions that best fulfill the specific demands of the two subfamilies of KDO8PS: while partly still in regions of the HMM profile with high relative entropy, these positions tend to cluster outside the β-barrel ([Figure 1B](#pone-0017459-g001){ref-type="fig"}). ::: {#pone-0017459-g001 .fig} 10.1371/journal.pone.0017459.g001 Figure 1 ::: {.caption} ###### HMM histograms of metallo and non-metallo KDO8PSs. A. 1^st^ inset, HMM vector for all KDO8PSs with superimposed Loess fit (calculated with a span of 20 residues). Regions of the histogram higher than the mean are highlighted with an orange bar below the sequence. 2^nd^ inset, HMM vectors for the C23 and N23 sub-families of KDO8PS. 3^rd^ inset, difference between the HMM vectors of the C23 and N23 sub-families. B. Stereo view of a monomer of *Nm.* KDO8PS shown here as a ribbon drawing. The regions of the structure corresponding to the orange bars in the 1^st^ inset of Panel A are colored in yellow. Residues corresponding to positions with values \>2σ in the difference histogram of panel A are shown as sticks with green carbon atoms. Residues corresponding to positions with values \>2σ for the difference between C23 and N23 KDO8PSs in the correlation between stability and amino acid usage (last inset in [Figure 2B](#pone-0017459-g002){ref-type="fig"}) are shown as sticks with magenta carbon atoms. Helices are labeled (H1 to H8) according to their position in the amino acid sequence. :::  ::: Unfortunately the HMM vectors do not reveal whether the driving force for these effects is only function or whether stability is also a factor. This question was addressed by using the experimentally validated FoldX algorithm [@pone.0017459-Guerois1], [@pone.0017459-Schymkowitz1], [@pone.0017459-Schymkowitz2] to calculate the ΔΔG changes associated with introducing any one of the 20 possible amino acids at each position of the structure of *Nm.* KDO8PS (again taken here as the reference structure for this class of proteins). This type of calculation was initially introduced by Tokuriki *et al.* [@pone.0017459-Tokuriki2] to study the overall distribution of stability effects for all possible mutations in a large set of different single domain globular proteins. In our case we used an entire tetramer of *Nm.* KDO8PS, which is known to be the biological unit of the enzyme [@pone.0017459-Cochrane1], and the individual mutations were introduced simultaneously in all four subunits. Thus, the calculated ΔΔG changes account also for the effects of mutations at the interface between subunits. Furthermore, as ΔΔG changes can be dependent on a particular conformation of the enzyme, a three-dimensional model of tetrameric *Nm.* KDO8PS derived from the X-ray structure [@pone.0017459-Cochrane1] (see [Methods](#s4){ref-type="sec"} section) was subjected to 12 nanoseconds (ns) of molecular dynamics (MD) simulation at 300 K under solvated conditions. This very long simulation time progressively eliminated possible errors in the original model and assured that the equilibrium structure of *Nm.* KDO8PS that is used in the FoldX calculations is as close as possible to the native structure in solution. Progressive convergence of the structure toward equilibrium was evaluated from the changes in the Cα root mean square deviation, Cα-RMSD, from the structure at time *t* = 0 ([Figure S1A](#pone.0017459.s001){ref-type="supplementary-material"}, Supporting Information). The final part of the simulation (6--12 ns) was selected to gather statistics about the conformational properties of the enzyme. The fluctuations around the average structure (Cα root mean square fluctuation, Cα-RMSF) that occur in this part of the simulation reflect the degree of mobility in the solution structure ([Figure S1A](#pone.0017459.s001){ref-type="supplementary-material"}). Frames from the final 6 ns of the simulation were investigated further with a clustering procedure (incorporated in the program X-Cluster, Schrodinger, LLC) that, based on the cross-RMSDs between frames, identified three structures ([PDB S1](#pone.0017459.s011){ref-type="supplementary-material"}, [PDB S2](#pone.0017459.s012){ref-type="supplementary-material"}, [PDB S3](#pone.0017459.s013){ref-type="supplementary-material"}, Supporting Information) as representative of all the states sampled during this part of the MD run. Thus, the final 6 ns of the MD simulation can be considered as fluctuations around these three main conformations, which appear in a 58∶24∶18 relative ratio. ΔΔG changes associated with mutating every amino acid of all four subunits to all 20 possible amino acids were calculated in duplicate for each of the three conformers. Values derived from each configuration were then merged by weighing each configuration according to its contribution to the population of states in solution as determined from the MD run ([Figure S1B](#pone.0017459.s001){ref-type="supplementary-material"}). The final ΔΔG values reflect not only the distribution of energies in the solution ensemble of tetrameric *Nm.* KDO8PS, but also the differences originating from the slightly different environment that each residue senses in the four subunits of the tetramer. The outcome of this calculation is a "stability landscape" of KDO8PS ([Figure 2A](#pone-0017459-g002){ref-type="fig"}; [Energy Matrix S1](#pone.0017459.s014){ref-type="supplementary-material"}, Supporting Information): the peaks in the landscape represent positions in the protein where introduction of a certain amino acid would significantly increase the ΔG, and therefore decrease the overall stability. It is worth noting that while the energies derived from FoldX are clearly not on an absolute scale [@pone.0017459-Tokuriki2], the relative trends are expected to be correct [@pone.0017459-Potapov1], [@pone.0017459-Khan1]. In general, it can be seen how bulky aromatic residues (W,Y,F,H) tend to decrease stability ( = increase energy) at every position, and in four positions (21,68,231,232) any residue besides glycine or alanine decreases stability dramatically. These effects appear to be due to very large energy terms derived from van der Waals clashes of these residues with their surroundings, which are not sufficiently relieved by the relaxation of the structure. The observed amino acid frequencies at each position in the C23 and N23 sub-families of KDO8PSs are shown as lolly-pops (yellow for C23, green for N23, height proportional to the corresponding frequency) superimposed to the stability landscape of KDO8PS ([Figure 2A](#pone-0017459-g002){ref-type="fig"}). In most positions, the amino acids most often used are those that do not decrease stability: in other words, only the planes of the stability landscape are significantly populated. ::: {#pone-0017459-g002 .fig} 10.1371/journal.pone.0017459.g002 Figure 2 ::: {.caption} ###### Stability landscape of KDO8PS. A. ΔΔG changes associated with introducing any one of the 20 possible amino acids at each position of all four subunits of the structure of *Nm.* KDO8PS were calculated using the FoldX algorithm to produce a "stability landscape" of KDO8PS: the peaks in the landscape represent positions in the protein where introduction of a certain amino acid would significantly decrease the overall stability. The observed frequency of different amino acids at each position in the two subfamilies of C23 and N23 KDO8PSs is superimposed to the stability landscape as lolly-pops (yellow for C23, green for N23), whose height is proportional to the frequency of a certain amino acid at a certain position. B. Stability/(amino acid usage) correlation in the two subfamilies of KDO8PSs. Correlation coefficients are based on the stability landscape and the relative frequency of each amino acid at each position as shown in panel A. From top to bottom: 1^st^ inset, correlation between stability and amino acid usage at each position in the entire family of KDO8PSs; a Loess fit (calculated with a span of 20 residues) is superimposed. The regions of the structure with the highest relative entropy in the HMM vector of [Figure 1A](#pone-0017459-g001){ref-type="fig"} are shown with an orange bar below the stem plot. 2^nd^ inset, *p*-values for the correlations shown in the 1^st^ inset. 3^rd^ and 4^th^ insets, correlation between stability and amino acid usage at each position of C23 (yellow circles) and N23 (green circles) KDO8PSs. 5^th^ inset, difference between the correlations in C23 and N23 KDO8PSs. Residues corresponding to positions with values \>2σ for this difference are shown as sticks with magenta carbon atoms in [Figure 1B](#pone-0017459-g001){ref-type="fig"}. :::  ::: The data presented in [Figure 2A](#pone-0017459-g002){ref-type="fig"} offer an interesting opportunity to understand the relationship between sequence and stability in KDO8PS: for each position in the reference structure of *Nm.* KDO8PS we consider two vectors (each with 20 elements): the first is the vector of the ΔΔG changes associated with mutating the original sequence to each of the 20 amino acids; the other contains the frequencies of each of the 20 amino acids for that position in each of the two subfamilies of KDO8PS (or in the entire family). We recall that if a MSA is composed of independent sequences all producing stable folds with approximately the same structure, and if individual residues contribute additively to stability (no epistasis) [@pone.0017459-Horovitz1], [@pone.0017459-Tracewell1], then the stability contribution ΔΔG*~a,i~* of a particular amino acid *a* at a given position *i* should be a roughly logarithmic function of its frequency *f~a,i~* in the MSA [@pone.0017459-Ohage1]: Approaches based on this idea have been generally successful in engineering more stable proteins [@pone.0017459-Polizzi1], [@pone.0017459-ChaparroRiggers1], [@pone.0017459-Lehmann1], [@pone.0017459-Lehmann2]. It follows that for each of the 280 positions in the reference structure of *Nm.* KDO8PS we can calculate the linear correlation coefficient (*corr*) between the stability vector \[exp(-ΔΔG*~a,i~*)\] derived with FoldX and the amino acid frequency vector derived from the MSA. The result of this calculation is shown in the insets of [Figure 2B](#pone-0017459-g002){ref-type="fig"}; in the family of all KDO8PSs and in both subfamilies there is a clear distinction between positions in which the amino acid usage is influenced by an evolutionary drive to increase stability, and other positions in which there is no such trend. There is a wavy pattern in the correlation between stability and observed amino acid frequencies (see the Loess fit in the first inset of [Figure 2B](#pone-0017459-g002){ref-type="fig"}), which is similar to that already noticed in the HMM vector ([Figure 1A](#pone-0017459-g001){ref-type="fig"}). The 2^nd^ inset of [Figure 2B](#pone-0017459-g002){ref-type="fig"} shows the *p*-values for the correlations in the 1^st^ inset, calculated by testing the *null* hypothesis of zero correlation against the *alternative* hypothesis of non-zero correlation. The large *p*-values for all the sequence positions with correlation near zero, and the small *p*-values (\<5E-4) for all the sequence positions with large correlation, lend credibility to the apparent correlation between stability gains and choice of particular amino acids in certain regions of the protein. The correlation coefficients between the HMM vectors ([Figure 1A](#pone-0017459-g001){ref-type="fig"}) and the stability/sequence correlation vectors are 0.34 (*p* = 7.1E-9), 0.31 (*p* = 1.5E-7), and 0.39 (*p* = 1.9E-11), for all KDO8PSs and for the C23 and N23 subfamilies, respectively, suggesting that the choice of sequence in the regions of high relative entropy is at least in part aimed at increasing the overall stability of the structure. If the correlation vectors between sequence and stability for each sub-family (3^rd^ and 4^th^ insets of [Figure 2B](#pone-0017459-g002){ref-type="fig"}) are subtracted from each other (5th inset of [Figure 2B](#pone-0017459-g002){ref-type="fig"}), a new set of positions is identified in the structure of *Nm.* KDO8PS, which represent places in which the correlation with stability changes significantly between the two sub-families. These are positions in which one sub-family selectively affects stability with respect to the other sub-family by preferentially adopting or discarding certain amino acids. By and large, these positions ([Figure 1B](#pone-0017459-g001){ref-type="fig"}, residues colored in magenta) are also found outside the β-barrel, as already noted for the positions derived from the difference between the HMM vectors ([Figure 1B](#pone-0017459-g001){ref-type="fig"}, residues colored in green). The contribution of coevolving residues to the stability of KDO8PS {#s2b} ------------------------------------------------------------------ In a MSA some positions are highly conserved, while others vary. The conserved positions are clearly important, but the non-conserved positions are not irrelevant because the net stabilization of the folded state, relative to the unfolded state is usually so small, that all positions may contribute significantly to the protein stability. This is clearly evident in the stability landscape of [Figure 2A](#pone-0017459-g002){ref-type="fig"}. Thus, the destabilizing effects of a given amino acid at one position can be compensated by the stabilizing effect of a certain amino acid at another position: in other words, two positions could be coevolving. A wide variety of algorithms have been developed to detect coevolving positions from a MSA (reviewed in [@pone.0017459-Horner1], [@pone.0017459-Caporaso1], [@pone.0017459-Codoner1]). Some of these methods use -tests [@pone.0017459-Kass1], [@pone.0017459-Fodor1], some are perturbative [@pone.0017459-Dekker1], [@pone.0017459-Lockless1], [@pone.0017459-Halabi1], others employ amino acid substitution matrices [@pone.0017459-Gobel1], and many work within the frame of information theory [@pone.0017459-Reza1]. Information entropy, *H*(X), is a measure of the uncertainty associated with a discrete random variable X that assumes values {x~1~,\..., x~n~}:where *b* is the base of the logarithm used and *p* is the probability mass function of the variable X [@pone.0017459-Shannon1], [@pone.0017459-Shannon2]. Related to *H*(X), mutual information, *MI*(X;Y), measures the mutual dependence of two discrete random variables X and Y:where *p*(x,y) is the joint probability mass function of X and Y, and *p*(x) and *p*(y) are the marginal probability mass functions of X and Y, respectively. Intuitively, MI measures how much knowing one of the two variables reduces the uncertainty about the other. In a MSA, the amino acids in a given column can be considered as a set of observations (x~i~) of a random variable X. An estimate of the entropy *H*(X) is obtained by using the observed amino acid frequencies, *f*(x~i~), in place of the underlying probabilities, *p*(x~i~); likewise, *MI*(X;Y) for a pair of columns can be derived using the frequencies, *f*(x~i~,y~j~), of all ordered pairs occurring in the two columns. In practice, MI between positions (columns in a MSA) reflects the extent to which knowledge of the amino acid at one position allows us to predict the identity of the amino acid at the other position [@pone.0017459-Atchley1], [@pone.0017459-Gloor1], [@pone.0017459-Martin1]. If amino acids occur independently at the two sites, the theoretical value for MI is zero; conversely, MI is high if the two positions are correlated. However, significant background MI can originate from random pairings of residues when the number of sequences in the multiple sequence alignments is small (in practice, less than 125 sequences [@pone.0017459-Martin1]). In addition, positions with high entropy (non-conserved positions) have more background MI than positions with low entropy. MI is also affected by various sources of bias, because the sequences in a MSA do not exactly meet the assumption of independent evolution. For example, the appearance of a mutation in an ancestral protein, which is clearly a single evolutionary event, would be considered in a MI analysis as representing an independent event that occurred in each of the proteins in the MSA that descended from that ancestor. This treatment of a single event as multiple independent events acts as a phylogenetic bias that increases the mutual information among residues. Normalization of MI values reduces the effect of positional entropy and phylogenetic bias [@pone.0017459-Martin1], and several normalized variants of MI have been proposed. A useful and symmetric type of normalized MI is symmetric uncertainty (*SU*) [@pone.0017459-Witten1], [@pone.0017459-Press1]: *MIr*, a form of MI normalized by the joint entropy of the variables, *H*(X,Y), instead of the sum, *H*(X)+*H*(Y), is conceptually similar to SU and is also widely used [@pone.0017459-Gloor1], [@pone.0017459-Martin1]. We have calculated the matrix of symmetric uncertainties *SU*(X,Y) for all columns in the MSAs of the entire family of metallo + non-metallo KDO8PSs. As recommended by some authors [@pone.0017459-Merlo1], [@pone.0017459-Dunn1], [@pone.0017459-Fernandes1] gaps were not included in the amino acid alphabet as this can lead to artificially high MI values, and if a gap appeared in a row in at least one of the two columns, that row did not contribute to the SU value for just those two columns. On the other hand, also this exclusion may lead to an artificial increase of the SU value. In fact, while the theoretical value for MI is zero if amino acids occur independently at the two sites, MI can be zero only if the observed pair frequencies reflect all possible pairs for the observed amino acid frequencies. If all 20 amino acids are present in each column and are equally probable, MI vanishes only if the frequency of each pair of amino acids is 1/20^2^. This condition is not met in a MSA with less than 400 sequences. Thus, in order to weigh the significance of pairs of columns containing different numbers of ungapped rows, the SU value was scaled linearly by the fraction of ungapped rows in the two columns with respect to the total number of rows in the alignment. Other forms of (non-linear) correction for the number of rows included in the MI calculation have also been proposed [@pone.0017459-Buslje1]. Many have noticed that MI between positions *i* and *j* in a MSA is highly correlated to the product of the average value of MI at each of these positions [@pone.0017459-Dunn1], [@pone.0017459-Little1]. While the origin of this correlation is uncertain, and can perhaps be attributed to positional entropy effects and a combination of both phylogenetic and stochastic bias [@pone.0017459-Little1], it is nonetheless clear that it produces a high background MI that obscures coevolution patterns. Several corrections have been proposed to eliminate this correlation, giving rise to new formulations or scoring of MI defined respectively as positional MI, *MIp* [@pone.0017459-Dunn1], Z-scored residual MI, *ZRes* [@pone.0017459-Little1], Z-scored-product normalized MI, *ZNMI* [@pone.0017459-Brown1]. The *Zpx* score introduced by Gloor *et al.* [@pone.0017459-Gloor2] supercedes the *MIp* score. More recently it has been pointed out that attempts to infer coevolution only from pair-count data are heavily affected by assumptions on the consistency between joint and marginal frequencies [@pone.0017459-Fernandes1], and that additional biological knowledge may be necessary for a meaningful derivation of coevolution patterns. In this line of thinking, Codoner *et al.* [@pone.0017459-Codoner2] have proposed to consider the correlation in the hydrophobicity and/or molecular weight of coevolving amino acid sites *a priori* to determine statistically their biological significance, and have shown that the application of these statistical filters to the number of pairs detected as coevolving reduces significantly the number of false positives. However, while this is clearly an emerging strategy in coevolution studies, here we are primarily interested in determining whether there is a statistical correlation between coevolution and stability, and therefore we do not include in the identification of coevolving pairs any statistical filters based on structural properties that directly affect stability (like hydrophobicity or molecular weight). Finally, the sensitivity of MI analyses to the size and quality of the MSA is also a matter of concern [@pone.0017459-Buslje1], [@pone.0017459-Brown1], [@pone.0017459-Codoner2], [@pone.0017459-Weil1]. We have studied the effect of different levels of sequence redundancy in the MSA, by calculating MI not only with the original data set of 348 sequences but also with a series of progressively smaller MSAs in which the highest level of identity between any two sequences was 98, 96, 94, 92, 90, 88, 86, 84, 82, 80%. These MSAs consisted respectively of 308, 266, 241, 221, 203, 179, 165, 154, 146, 130 sequences. A threshold point at which the trend in the statistics for the total number and distribution of coevolving pairs appeared to change could be recognized in the MSA with 86% maximal sequence identity, which consisted of only 165 sequences ([Figure S2](#pone.0017459.s002){ref-type="supplementary-material"}, [MSA S3](#pone.0017459.s008){ref-type="supplementary-material"}, Supporting Information). While throughout the manuscript we refer to the complete data set of 348 sequences, the corresponding results obtained with the data set of 165 sequences are also provided ([Table S1](#pone.0017459.s003){ref-type="supplementary-material"}, [Table S2](#pone.0017459.s004){ref-type="supplementary-material"}, [Table S3](#pone.0017459.s005){ref-type="supplementary-material"}, Supporting Information). Overall these analyses were fairly insensitive to the size/redundancy of the MSAs, suggesting that in the specific case of KDO8PS MI studies are not particularly affected by stochastic and/or phylogenetic bias. Despite normalization and scaling, the SU matrix of the MSA for the KDO8PS family is contaminated by a significant level of background MI (*corr* = 0.78 to the column product matrix). The *Zpx*, *ZRes* and *ZNMI* matrices all work well with the KDO8PS data set in reducing this background MI; as expected they are highly correlated to each other (*corr*(*Zpx,ZRes*) = 0.90, *corr*(*Zpx,ZNMI*) = 0.81, *corr*(*ZNMI,ZRes*) = 0.73). Since it is not clear yet which of these formulations of MI is more accurate or appropriate for a particular study [@pone.0017459-Brown1], we report the results obtained with each one in our analysis of coevolution in KDO8PS ([Tables 1](#pone-0017459-t001){ref-type="table"}, [2](#pone-0017459-t002){ref-type="table"} and [3](#pone-0017459-t003){ref-type="table"}). In this study we were not interested in determining whether there are patterns of spatial relationship among the coevolving positions, but whether there is a relationship between coevolving pairs and the stability landscape of KDO8PS described in the previous section. As an example of our approach, the number of coevolving pairs with score higher than 1 to 5 σ over the mean of all scores in the *ZRes* matrix, are shown for every position in the MSA of all KDO8PSs as histograms in the insets of [Figure 3](#pone-0017459-g003){ref-type="fig"}. The same regions of sequence highlighted for the HMM vector of [Figure 1A](#pone-0017459-g001){ref-type="fig"} are shown below the baseline of the lowest inset, as horizontal orange bars. The correlation coefficients between the histogram vectors for each type of MI matrix and the HMM vector of [Figure 1A](#pone-0017459-g001){ref-type="fig"} are shown in [Table 1](#pone-0017459-t001){ref-type="table"}. These coefficients were calculated including only positions of the MSA that do not correspond to fully conserved residues, as there are no coevolving pairs between these positions. Some modest level of correlation (at best *corr* = 0.30) with reasonable statistical significance (*p*\<0.01) can be recognized only for the strongest signals in the *ZRes*, and *ZNMI* matrices. ::: {#pone-0017459-g003 .fig} 10.1371/journal.pone.0017459.g003 Figure 3 ::: {.caption} ###### Mutual information in KDO8PS. The *ZRes* matrix for the MSA of all KDO8PSs is shown in the upper panel. For better contrast, matrix values are displayed in the z-score range 0 to 20 with the color ramp shown in the side bar; the full range of the matrix is from −3.9 to 245.7. Fully conserved positions appear as uniform light blue rows and columns. For each position of the MSA of all KDO8PSs, the insets below the matrix show the number of coevolving pairs with score values larger 1, 2, 3, 4, or 5 σ over the mean of all the scores in the matrix. The regions of the structure with the highest relative entropy in the HMM vector of [Figure 1A](#pone-0017459-g001){ref-type="fig"} are shown as orange bars below the histograms; these are the same regions also shown as orange bars in [Figures 1](#pone-0017459-g001){ref-type="fig"} and [2](#pone-0017459-g002){ref-type="fig"}. :::  ::: ::: {#pone-0017459-t001 .table-wrap} 10.1371/journal.pone.0017459.t001 Table 1 ::: {.caption} ###### Correlation coefficients between the histogram vectors derived from the MI matrices and the HMM vector of [Figure 1A](#pone-0017459-g001){ref-type="fig"}. ::: {#pone-0017459-t001-1} Threshold for coevolving pairs 1 σ 2 σ 3 σ 4 σ 5 σ -------------------------------- -------------------- ------------------ ------------------ ------------------ ------------------ ***Zpx*** **matrix** −0.393 \[2.0e-10\] −0.094 \[0.144\] 0.061 \[0.346\] 0.104 \[0.012\] 0.118 \[8.0e-4\] ***ZRes*** **matrix** −0.013 \[0.838\] 0.118 \[0.066\] 0.171 \[0.007\] 0.222 \[4.9e-4\] 0.201 \[1.6e-3\] ***ZNMI*** **matrix** −0.270 \[1.9e-5\] 0.104 \[0.105\] 0.227 \[3.4e-4\] 0.297 \[2.3e-6\] 0.300 \[1.7e-6\] ::: ::: {#pone-0017459-t002 .table-wrap} 10.1371/journal.pone.0017459.t002 Table 2 ::: {.caption} ###### Contribution to KDO8PS stability from coevolving pairs with score \>5 σ in the *ZRes* matrix of [Figure 3](#pone-0017459-g003){ref-type="fig"}. ::: {#pone-0017459-t002-2} *i,j pair* score ΔΔG(i+j) \|ΔΔG(i-j)\| *i,j pair* score ΔΔG(i+j) \|ΔΔG(i-j)\| ------------- ------------ ------------ -------------- ------------- ------------ ----------- -------------- 23,179 13.464 1.473 0.866 145,169 18.448 0.481 0.299 23,201 16.307 1.264 0.674 **145,193** **13.953** **0.380** **0.399** 24,65 20.772 0.494 0.302 156,157 23.782 1.169 0.428 24,66 16.083 0.600 0.203 169,193 26.494 0.217 0.209 24,74 19.011 0.594 0.217 169,218 14.159 0.686 0.553 **26,244** **245.65** **−0.021** **0.024** 172,204 20.79 0.294 0.205 45,215 13.477 0.602 0.505 172,205 16.916 0.516 0.424 45,268 13.995 1.112 0.651 172,206 15.603 0.413 0.333 **62,107** **188.1** **−0.009** **0.030** 195,221 71.347 0.285 0.252 **73,76** **18.029** **0.174** **0.387** 204,205 24.25 0.709 0.235 74,173 18.367 0.378 0.285 204,206 19.581 0.606 0.356 **91,106** **16.741** **0.272** **0.323** 205,206 16.747 0.828 0.579 116,137 14.649 2.573 2.497 208,211 15.733 0.326 0.274 119,120 41.6 0.025 0.025 215,249 14.89 0.176 0.081 119,175 13.509 0.094 0.083 215,268 14.422 0.662 0.518 120,124 13.63 0.073 0.058 **222,248** **14.622** **0.241** **0.256** 133,234 114.09 0.111 0.092 227,229 14.026 1.276 0.971 **139,213** **184.35** **−0.067** **0.083** 231,268 13.469 2.913 1.800 141,172 17.902 0.194 0.136 240,243 15.497 0.872 0.805 141,204 17.453 0.388 0.120 **248,249** **45.125** **0.162** **0.186** 144,205 13.579 0.782 0.160 268,272 28.615 0.820 0.440 ΔΔG\'s are in kcal/mol. Rows in which \|ΔΔG(i-j)\|\>ΔΔG(i+j) are shown in bold. ::: ::: {#pone-0017459-t003 .table-wrap} 10.1371/journal.pone.0017459.t003 Table 3 ::: {.caption} ###### Correlation coefficients between the vectors of MI scores for *i,j* pairs above a threshold σ value and the vectors representing the average effect of those pairs on the stability of KDO8PS. ::: {#pone-0017459-t003-3} Threshold for coevolving pairs. 1 σ 2 σ 3 σ 4 σ 5 σ ------------------------------------------------------------------------------------- ----------------------- -------------------- ------------------- ------------------- ------------------- ***Zpx*** **matrix** **No. of unique coevolving pairs** [a](#nt102){ref-type="table-fn"} 4319 1402 467 162 75 ***corr*** **(MI~ij~,ΔΔG~i~+ΔΔG~j~)** \[*p*-value\][b](#nt103){ref-type="table-fn"} −0.019 \[0.103\] −0.110 \[1.9e-5\] −0.205 \[3.8e-6\] −0.272 \[2.3e-4\] −0.258 \[0.013\] ***corr*** **(MI~ij~,\|ΔΔG~i~−ΔΔG~j~\|)** \[*p*-value\] −0.082 \[3.0e-8\] −0.156 \[2.0e-9\] −0.223 \[5.5e-7\] −0.283 \[1.3e-4\] −0.260 \[0.013\] **% of pairs with opposite effects** 0.343 0.301 0.244 0.222 0.223 ***ZRes*** **matrix** **No. of unique coevolving pairs** 861 267 105 66 42 ***corr*** **(MI~ij~,ΔΔG~i~+ΔΔG~j~)** \[*p*-value\] −0.116 \[3.3e-4\] −0.183 \[1.3e-3\] −0.236 \[7.7e-3\] −0.260 \[0.017\] −0.353 \[0.011\] ***corr*** **(MI~ij~,\|ΔΔG~i~--ΔΔG~j~\|)** \[*p*-value\] −0.129 \[7.1e-5\] −0.194 \[7.3e-4\] −0.231 \[9.0e-3\] −0.240 \[0.026\] −0.288 \[0.032\] **% of pairs with opposite effects** 0.271 0.232 0.238 0.227 0.190 ***ZNMI*** **matrix** **No. of unique coevolving pairs** 4157 892 225 78 32 ***corr*** **(MI~ij~,ΔΔG~i~+ΔΔG~j~)** \[*p*-value\] −0.051 \[4.5e-4\] −0.166 \[3.3e-7\] −0.242 \[1.2e-4\] −0.235 \[0.019\] −0.451 \[4.8e-3\] ***corr*** **(MI~ij~,\|ΔΔG~i~−ΔΔG~j~\|)** \[*p*-value\] −0.105 \[5.5e-12\] −0.201 \[7.1e-10\] −0.262 \[3.5e-5\] −0.238 \[0.018\] −0.445 \[5.3e-3\] **% of pairs with opposite effects** 0.362 0.314 0.280 0.269 0.250 a Since the MI matrix is symmetric the total number of pairs (as represented for example in the histograms of [Figure 3](#pone-0017459-g003){ref-type="fig"}) is twice that of the unique part of the matrix. b The *null* hypothesis of zero correlation was tested against the *alternative* hypothesis of negative correlation. ::: Since the HMM vector represents deviations from the expected background probability distribution to satisfy the demands of structure, stability and function, it is reasonable to suggest that strongly coevolving positions of KDO8PS may be associated with one or more of these demands. In order to determine the specific contribution of coevolution to stability we compared the level of MI of each pair with the average contribution to stability by that pair in the MSA of KDO8PS, as derived by applying to each sequence in the MSA the information contained in the stability matrix shown in [Figure 2A](#pone-0017459-g002){ref-type="fig"} ([Energy Matrix S1](#pone.0017459.s014){ref-type="supplementary-material"}, Supporting Information). For example, based on that matrix a proline would contribute 3.27 kcal/mol at position 31 of a sequence in the MSA, but −1.05 kcal/mol at position 56. We were interested not only in the sum (ΔΔG*~i~*+ΔΔG*~j~*) of the stability contributions of each member of the *i,j* pair (as these contributions can be considered approximately additive under the assumption of no epistasis), but also in the absolute value of the difference \|ΔΔG*~i~*−ΔΔG*~j~*\|, which depends on whether the two contributions have similar or opposite effects. Since MI is already the log of a probability, consistent with equation (1) the two magnitudes of interest are: where the average in equation (6) is over all the sequences in the MSA. As an example, these magnitudes are shown in [Table 2](#pone-0017459-t002){ref-type="table"} for all coevolving pairs with a score \>5 σ in the *ZRes* matrix of [Figure 3](#pone-0017459-g003){ref-type="fig"}. Less than 20% of these pairs (rows in bold in [Table 2](#pone-0017459-t002){ref-type="table"}) have values of \|ΔΔG*~i~*−ΔΔG*~j~*\|\>(ΔΔG*~i~*+ΔΔG*~j~*), and thus identify positions that, on average, exert opposite effects on the stability of KDO8PS. For each type of MI matrix, the correlations between the score vector (for example, the score columns in [Table 2](#pone-0017459-t002){ref-type="table"}) and the (ΔΔG*~i~*+ΔΔG*~j~*) and \|ΔΔG*~i~*−ΔΔG*~j~*\| vectors are shown in [Table 3](#pone-0017459-t003){ref-type="table"}. In general, it can be seen that, as progressively higher σ thresholds are chosen, the score vectors become more clearly anti-correlated to the (ΔΔG*~i~*+ΔΔG*~j~*) vector: this trend is consistent with equation (5), because we expect that pairs whose components raise the ΔΔG are less likely to occur. The score vectors are also anti-correlated to the \|ΔΔG*~i~*−ΔΔG*~j~*\| vector, suggesting that on average coevolving pairs have a preference for positions whose individual effects on stability are similar. The trends are similar in the various matrices and *p*-values are acceptable, with *ZRes* and *ZNMI* yielding the largest values of anti-correlation ([Table 3](#pone-0017459-t003){ref-type="table"}). These results can be interpreted in the light of current theories of coevolution. According to one hypothesis, a reason for coevolving pairs is to suppress a decrease in stability or function produced by a mutation at one site with an increase in stability or function provided by a mutation in a residue near the site of the first mutation [@pone.0017459-Yanofsky1], [@pone.0017459-Poon1], [@pone.0017459-Poon2]. This tenet is challenged by the observation that most suppressor mutations are not close in space to the initial mutations [@pone.0017459-Klig1], while coevolving sites are most often spatially clustered [@pone.0017459-Gloor1], [@pone.0017459-Martin1], [@pone.0017459-Poon2], [@pone.0017459-Tillier1]. Directed evolution studies also suggest that epistatic paths can be bypassed (and they certainly are in laboratory evolution), because there are multiple sequences that satisfy a given fitness goal, and there are many different paths to these sequences [@pone.0017459-Tracewell1], [@pone.0017459-Romero1], [@pone.0017459-Bloom2]. According to an alternative model for coevolution, "covarions" arise when both the original residue and the mutated residue are compatible with function, but the spectrum of residues possible at other positions in the protein is altered by the mutation [@pone.0017459-Fitch1]. In this context one might expect that most single mutations that finally become stabilized in coevolving pairs are neutral in the context of the protein in which they occur, but become beneficial (even at a much later time) in the presence of the partner in the pair [@pone.0017459-Gloor2]. The data in [Table 3](#pone-0017459-t003){ref-type="table"} suggest that both mechanisms were operational in the evolution of KDO8PS. Approximately 1/4 of all strongly coevolving pairs (for example those shown in bold in [Table 2](#pone-0017459-t002){ref-type="table"}; see also [Table 3](#pone-0017459-t003){ref-type="table"}) may have originated from cycles of mutation and suppression that affected stability. Other pairs, for which both values of (ΔΔG*~i~*+ΔΔG*~j~*) and \|ΔΔG*~i~*−ΔΔG*~j~*\| are small, are best explained by a succession of neutral or nearly neutral covarions. Discussion {#s3} ========== In the metal dependent forms of KDO8PS the metal is not directly involved in an activation process, but together with its ligands stabilizes the position of the reactants that favors the condensation reaction [@pone.0017459-Kona1], [@pone.0017459-Tao1], [@pone.0017459-Tao2]. In the non-metal forms of the enzyme, this role is performed by an asparagine side chain that replaces the combination cysteine + metal [@pone.0017459-Tao1], [@pone.0017459-Tao2]. In both forms of the enzyme additional stabilization of the reactants is provided by variations in the CysAspGlyPro motif of the loop that contains the metal binding aspartic acid [@pone.0017459-Cochrane1], [@pone.0017459-Allison1]. Outside these demands placed on the structure by the need to catalyze a specific reaction, the factors (like protein stability) that contributed to the evolution of hundreds of different forms of KDO8PS are not well understood. The starting point of our study was the derivation of the stability landscape of KDO8PS ([Figure 2A](#pone-0017459-g002){ref-type="fig"}) by calculating the ΔΔG changes associated with introducing any one of the 20 possible amino acids at each sequence position of the structure of *Nm.* KDO8PS (taken here as the reference structure for this class of proteins). Superposition on this landscape, of the actual amino acid frequencies at each position in the two sub-families of KDO8PSs confirms the intuitive expectation that in most positions of high relative entropy ([Figure 1A](#pone-0017459-g001){ref-type="fig"}, 1^st^ inset), the amino acids most often used are those that do not decrease stability ([Figure 2B](#pone-0017459-g002){ref-type="fig"}, 1^st^ inset). Furthermore, the plane regions of the landscape tend to be large, such that random drifts of sequence have a high chance of producing only small gains or losses of stability, which can be easily offset by changes in other regions. This feature of the landscape may be part of the physical basis for the threshold robustness (tolerance) to mutations that was observed in other proteins [@pone.0017459-Bershtein1], [@pone.0017459-Bershtein2]. Another important factor in constraining sequence variation is the need to retain a common core structure, while possibly adapting to specific bacterial environments. Ultimately, the preservation of structure is related to the preservation of function because small changes in atom distances in the active site (produced by overall changes of structure) can have effects on the activation energies as dramatic as those produced by very localized mutations in the active sites [@pone.0017459-Sigala1]. Both sequence conservation and coevolution were involved in the preservation of the structure of KDO8PS, but coevolution had a marginal role ([Table 1](#pone-0017459-t001){ref-type="table"}). This result can be rationalized by observing that if most mutations shifted a position in the sequence away from a clearly defined valley in the stability landscape, then it would be more likely for a given amino acid at that position to become fully conserved. Conversely, if most mutations moved a position around between regions of similar height, a large number of different amino acids at other positions might easily compensate the small ΔΔG changes of the initial mutation, and a clear pattern of coevolution would be less likely to emerge. In our study we have used various formulations of MI to identify correlated positions, but other methods may provide different information on the potential cross-talks between conserved and correlated positions, and be better suited for different tasks [@pone.0017459-Fodor1]. For example, using the Observed Minus Expected Squared (OMES) covariance method [@pone.0017459-Kass1] Kowarsch *et al.* found a statistically significant association between correlated positions and disease-causing mutations [@pone.0017459-Kowarsch1]. The association was more pronounced for the exposed accessible sites of the proteins studied, which are expected to be more involved with function, rather than for the structural cores, which are more likely to be involved in stability. Thus, altogether, despite the significantly different algorithm used, the conclusions of Kowarsch\'s study are compatible and complementary to ours in assessing the contributions of correlated positions to the stability of proteins. Derivation of the stability landscape is still at an early stage of refinement and its utility as a tool to understand the evolution of amino acid sequences in protein families needs additional confirmations. Current calculations were carried out with the FoldX algorithm, which uses a full atomic description of the structure of proteins and whose different energy terms have been weighted using empirical data obtained from protein engineering experiments [@pone.0017459-Guerois1], [@pone.0017459-Schymkowitz2], [@pone.0017459-Mendes1]. While this or similar algorithms involving empirical functions provide a relatively fast estimation of the importance of the interactions contributing to the stability of proteins and protein complexes, the calculated ΔΔG changes are unlikely to be on a correct absolute scale [@pone.0017459-Tokuriki2] and are certainly not as accurate as those derived from more sophisticated approaches like, for example, Free Energy Perturbation (FEP [@pone.0017459-Zwanzig1]); unfortunately, FEP is still computationally too expensive and time consuming, to be compatible with full sequence scans. Two recent comparisons of various methods (including CC/PBSA [@pone.0017459-Benedix1], Rosetta [@pone.0017459-Kortemme1], EGAD [@pone.0017459-Pokala1], I-Mutant2.0 [@pone.0017459-Capriotti1], I-Mutant3.0 [@pone.0017459-Capriotti2], CUPSAT [@pone.0017459-Parthiban1], MultiMutate [@pone.0017459-Deutsch1], Dmutant [@pone.0017459-Zhou1], Hunter [@pone.0017459-Potapov1], and FoldX) designed to predict the ΔΔG changes associated with mutations, ranked FoldX among the overall best performing ones [@pone.0017459-Potapov1], [@pone.0017459-Khan1]. However, while all the methods showed a correct trend in their predictions, they failed to provide precise values, with the best predictors being only moderately (60%) accurate. All these considerations point to the conclusion that significantly better tools are needed for a quantitative analysis of mutations by computational means. However, regardless of whether further improvements are easily achievable in calculating the stability landscape of proteins, an aim of this study is to stress how such landscape can be an important tool to integrate structural data with information theory to understand the evolution of proteins. For example, it should be noted that HMM profiles would be quite different if the information content of the distribution *P* ~i~ at the *i*-th position of the profile was calculated with respect to the expected frequency of different amino acids as derived from the values of the stability matrix at that position, rather than from the background frequencies of different amino acids in all known proteins, or just in the family under consideration. In that case we might expect that the positions with high information content would be those in which the observed sequence deviates significantly (perhaps for functional reasons) from the stability constraints associated with a certain structure. Similar considerations might apply also to the calculation of mutual information, where the expected frequencies derived from the stability profile could be used as the probability mass function of each variable. This approach is complementary to that adopted by Codoner *et al.* [@pone.0017459-Codoner2] to filter out (based on the hydrophobicity and molecular weight of residues at specific positions) potential false positives in the ensemble of coevolving pairs inferred from MI analyses. For example, if a model based on the size of side chains is adopted for the distribution of aa\'s at any given position, we can see how the *largest* contribution to MI between any two positions will originate from deviations (like the match of one large and one small side chain in a defined region of space) of the observed joint distribution with respect to the model. The interpretation of MI as the Kullback-Leibler divergence [@pone.0017459-Kullback1], [@pone.0017459-Kullback2], [@pone.0017459-Kullback3] of the observed *versus* the model joint distribution of two variables becomes particularly meaningful in this context. While our study was primarily aimed at studying the effect of coevolving residues on the overall stability of KDO8PS, indirectly it provided important information on the functional role of specific pairs. For example, residues 23--26 of *Nm.* KDO8PS listed at the top of [Table 2](#pone-0017459-t002){ref-type="table"} represent the sequence AsnValLeuGlu, with Asn being the key residue the replaces the metal binding Cys in all non-metal KDO8PS\'s. Position 26 appears to coevolve with position 244, which immediately precedes the conserved motif CysAspGlyPro (residues 246--249 in *Nm.* KDO8PS), whose functional role in KDO8PS was already noted [@pone.0017459-Allison1]. Coevolution of position 24 with positions 65,66, and 74 is explained by the fact that residue 24 is in van der Waals contact with 66 and within 10 Å of 65 and 74. Other pairs, like position 23 (Cys or Asn in all KDO8PS) with positions 179 and 201, are of interest but at the moment still unexplained. It is also worth noting that, besides its relevance for the coevolution analysis, the stability matrix of *Nm.* KDO8PS ([Energy Matrix S1](#pone.0017459.s014){ref-type="supplementary-material"}, Supporting Information) provides the platform for designing new mutations of KDO8PS that may be useful to stabilize or destabilize specific regions of the enzyme. Finally, it is legitimate to ask whether the conclusions about the role of coevolving residues in the thermodynamic stability of KDO8PS can be extended to other proteins. First, the analysis appears to be fairly insensitive to the choice of the reference sequence in the MSA. In fact, we obtained very similar results using the X-ray structure of KDO8PS from the hyperthermophile *Aquifex aeolicus* (PDB entry 1FWW) as the reference structure for the calculation of the stability matrix (not shown). Second, preliminary application of the method to two other structurally unrelated protein families (the Atp12p chaperones involved in F~1~-ATPase assembly [@pone.0017459-Ludlam1], and B1 type metallo β-lactamases [@pone.0017459-Bebrone1]) gave results that are consistent with those obtained with KDO8PS ([Table S4](#pone.0017459.s009){ref-type="supplementary-material"}, [Table S5](#pone.0017459.s010){ref-type="supplementary-material"}, Supporting Information). While these observations are very encouraging, clearly more testing with a larger set of different protein families will be necessary before any general principles about the role of coevolution in protein stability can be derived from the application of the tools described in this study. Methods {#s4} ======= Multiple Sequence Alignments {#s4a} ---------------------------- Multiple sequence alignments (MSAs) of 348 sequences (175 metallo and 173 non-metallo KDO8PSs) were calculated independently with T-Coffee [@pone.0017459-Notredame1], Muscle [@pone.0017459-Edgar1], and Mafft [@pone.0017459-Katoh1] and then merged together with T-Coffee ([MSA S1](#pone.0017459.s006){ref-type="supplementary-material"}, Supporting Information). For clarity and ease of comparison with the X-ray structure, the sequence numbering of *Nm.* KDO8PS was used as reference for the entire family, and only positions in the MSA with a corresponding residue in *Nm.* KDO8PS were finally retained ([MSA S2](#pone.0017459.s007){ref-type="supplementary-material"}, Supporting Information). Hidden Markov Models (HMMs) [@pone.0017459-Eddy1] were calculated with the HMMER 3.0 package [@pone.0017459-Eddy2]. Molecular Dynamics Simulations {#s4b} ------------------------------ A complete three-dimensional model of *Nm.* KDO8PS was built with Prime 2.1 (Schrodinger, LLC) using primarily the X-ray structure of *Nm.* KDO8PS (PDB 2QKF) as template, and that of *Aa.* KDO8PS (PDB 1FWW) only to build the residues missing in the *Nm.* structure. The ensemble for the MD simulation was constructed with Desmond (D.E. Shaw Research) [@pone.0017459-Bowers1] by solvating the enzyme with SPC water [@pone.0017459-Berendsen1] inside an orthorhombic box of 87.7, 110.0, 105.4 Å: a minimum distance of 12 Å was left between any protein atom and the edge of the box. The final ensemble contained a tetramer of *Nm.* KDO8PS, approximately 26,000 solvent molecules, 72 Na and 72 Cl ions in the solvent regions to neutralize charges on the proteins and to achieve a final salt concentration of 150 mM. Prior to additional steps the ensemble was subjected to energy minimization under periodic boundaries condition with a totals of 2,000 iterations first with the steepest descent (SD) method until a gradient threshold of 25 kcal/mol/Å was achieved, and then with the LBFGS method [@pone.0017459-Byrd1] until a convergence threshold of 1 kcal/mol/Å was met. The 2005 release of the OPLS-AA force-field [@pone.0017459-Jorgensen1] was used in this and all subsequent calculations. Short range Coulombic interactions were calculated with a cutoff radius of 9.0 Å, while long range interactions were calculated with the smooth particle mesh Ewald method [@pone.0017459-Essmann1] using an Ewald tolerance of 1e^−9^. A 12 nanoseconds (ns) MD simulation of the solvated tetrameric *Nm.* KDO8PS was carried out with Desmond in the NPT ensemble at 300 K. For this purpose the Nose-Hoover thermostat method [@pone.0017459-Evans1] with a relaxation time of 1.0 picoseconds (ps), and the Martyna-Tobias-Klein barostat method [@pone.0017459-Martyna1] with isotropic coupling of the cell along all three axes to a reference pressure of 1.01325 atm and a relaxation time of 2 ps were used. Integration was carried out with the RESPA integrator [@pone.0017459-Tuckerman1] using time steps of 2.0 fentomseconds (fs), 2.0 fs, and 6.0 fs for the bonded, van der Waals and short-range, and long-range electrostatic interactions. SHAKE constraints [@pone.0017459-Ryckaert1] were imposed on all the heavy-atom-hydrogen covalent bonds. Coulombic interactions were calculated as for the minimization protocol. Coordinates were saved every 4.8 ps. Before the 12 ns productive run of the simulation, the ensemble was relaxed using the following protocol: 1) 12 ps in the NVT ensemble at 10 K with a fast relaxation constant and non-hydrogen solute atoms restrained; 2) 12 ps in the NPT ensemble at 10 K and 1 atm, with a fast temperature relaxation constant, a slow pressure relaxation constant, and non-hydrogen solute atoms restrained; 3) 24 ps at 300 K and 1 atm with other conditions as in step 2; 4) 24 ps at 300 K and 1 atm with a fast temperature relaxation constant and a fast pressure relaxation constant. Conformational clustering based on the cross-RMSDs between the frames contained in the 6--12 ns interval was carried out with the program X-Cluster (Schrodinger, LLC). FoldX calculations of protein stability {#s4c} --------------------------------------- ΔΔG changes associated with introducing any one of the 20 possible amino acids at each position in all four monomers of a tetramer (the biological unit) of *Nm.* KDO8PS were calculated with FoldX v3.0b4 [@pone.0017459-Guerois1], [@pone.0017459-Schymkowitz2] following the procedure described in [@pone.0017459-Kiel1]. First, each of the representative structures of *Nm.* KDO8PS derived from the MD simulation was optimized using the repair function of FoldX. Then, structures corresponding to each of the point mutants were generated and their energies calculated using the FoldX energy function. Finally, the energy of the optimized wild-type structure was subtracted from that of the mutant. Each calculation was carried out in duplicate to ensure convergence: in this case the FoldX algorithm repeated the same mutations twice changing the rotamer set used and the order of moves such that alternative solutions could be explored. Mutual Information Analysis {#s4d} --------------------------- Correlation coefficients were calculated with the Statistics Toolbox of Matlab 7.10 (The MathWorks™). Mutual Information (MI) and Symmetric Uncertainty (SU) were calculated with the Information Theory Toolbox v.1.0 for Matlab developed by Joaquin Goni (Dept. of Physics and Applied Mathematics, University of Navarra, Pamplona, Spain), and available for download at Matlabcentral ([www.mathworks.com/matlabcentral](http://www.mathworks.com/matlabcentral)). Z-scored cross-product positional MI, *Zpx* [@pone.0017459-Gloor2], product based Z-scored residual MI, *ZRes* [@pone.0017459-Little1], Z-scored-product normalized MI, *ZNMI* [@pone.0017459-Brown1] were implemented according to the published algorithms as Matlab programs. All the data sets and Matlab programs required to reproduce the results of the study are available from the authors upon request. Figures were generated with Matlab 7.10 and Pymol 1.2r3 (Schrodinger, LLC). Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Derivation of the stability landscape of** ***Nm.*** **KDO8PS.** A. MD simulation of tetrameric *Nm.* KDO8PS at 300K. The Cα root mean square deviation, Cα-RMSD, from the structure at time *t* = 0 is shown in the upper trace colored in blue (0--6 ns) and red (6--12 ns). The fluctuations around the average structure (Cα root mean square fluctuation, Cα-RMSF) that occur during the 6--12 ns part of the simulation are shown in the lower green trace, and reflect the degree of mobility in the solution structure. B. ΔΔG changes associated with mutating every amino acid of all four subunits of *Nm.* KDO8PS to all 20 possible amino acids (a total of 5600 mutations in each subunit) were calculated in duplicate for each of the three main conformers ([PDB S1](#pone.0017459.s011){ref-type="supplementary-material"}, [PDB S2](#pone.0017459.s012){ref-type="supplementary-material"}, [PDB S3](#pone.0017459.s013){ref-type="supplementary-material"}) observed in the 6--12 ns part of the MD simulation. Values derived from these three representative configurations were then merged (magenta trace; see also [Energy Matrix S1](#pone.0017459.s014){ref-type="supplementary-material"}) according to their contribution (relative ratio of 58∶24∶18 for [PDB S1](#pone.0017459.s011){ref-type="supplementary-material"}∶[PDB S2](#pone.0017459.s012){ref-type="supplementary-material"}∶[PDB S3](#pone.0017459.s013){ref-type="supplementary-material"}) to the solution ensemble. Standard deviations of the ΔΔG values are shown as a black trace. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Effect of redundancy in the MSA of KDO8PSs on the identification of coevolving pairs.** A. MI with *ZRes* scoring was calculated for the original data set of 348 sequences (maximum 99% identity between any two sequences) and for a series of smaller MSAs in which the highest level of identity between any two sequences was 98, 96, 94, 92, 90, 88, 86, 84, 82, 80%. The total number of coevolving positions (left panel) or unique coevolving pairs (right panel) is shown for different σ levels in the MI matrix. B. The level of contrast in the MI matrix is expressed as the ratio between the number of coevolving positions (left panel) or unique coevolving pairs (right panel) at 5 and 1 σ, respectively. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Correlation coefficients between the histogram vectors derived from the MI matrices and the HMM vector of a MSA of 165 KDO8PS sequences ([MSA S3](#pone.0017459.s008){ref-type="supplementary-material"}) in which the highest identity allowed between any two sequences is 86%. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Contribution to KDO8PS stability from coevolving pairs with score \>5 σ in the *ZRes* matrix of a MSA of 165 KDO8PS sequences ([MSA S3](#pone.0017459.s008){ref-type="supplementary-material"}) in which the highest identity allowed between any two sequences is 86%. ΔΔG\'s are in kcal/mol. Rows in which \|ΔΔG(i−j)\|\>ΔΔG(i+j) are shown in bold. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### Correlation coefficients between the vectors of MI scores for *i,j* pairs above a threshold σ value and the vectors representing the average effect of those pairs on the stability of KDO8PS, based on a MSA of 165 KDO8PS sequences ([MSA S3](#pone.0017459.s008){ref-type="supplementary-material"}) in which the highest identity allowed between any two sequences is 86%. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: MSA S1 ::: {.caption} ###### MSA in fasta format of all KDO8P synthases. (FASTA) ::: ::: {.caption} ###### Click here for additional data file. ::: MSA S2 ::: {.caption} ###### MSA in fasta format of all KDO8P synthases with only the positions corresponding to the sequence of *Nm.* KDO8PS (Uniref entry 9JZ55) retained. (FASTA) ::: ::: {.caption} ###### Click here for additional data file. ::: MSA S3 ::: {.caption} ###### MSA in fasta format of KDO8P synthases with no more than 86% identity between any two sequences. Only the positions corresponding to the sequence of *Nm.* KDO8PS (Uniref entry 9JZ55) were retained. (FASTA) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### Correlation coefficients between the vectors of MI scores (based on a MSA of 249 sequences) for *i,j* pairs above a threshold σ value and the vectors representing the average effect of those pairs on the stability of Atp12p. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### Correlation coefficients between the vectors of MI scores (based on a MSA of 145 sequences) for *i,j* pairs above a threshold σ value and the vectors representing the average effect of those pairs on the stability of B1 type metallo β-lactamases. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: PDB S1 ::: {.caption} ###### Atomic coordinates in pdb format for the structure of *Nm.* KDO8PS which, based on a clustering procedure, is the most representative conformer of all the states sampled during the 6--12 ns part of the MD run shown in [Figure S1](#pone.0017459.s001){ref-type="supplementary-material"}. (PDB) ::: ::: {.caption} ###### Click here for additional data file. ::: PDB S2 ::: {.caption} ###### Atomic coordinates in pdb format for the structure of *Nm.* KDO8PS which, based on a clustering procedure, is the 2^nd^ most representative conformer of all the states sampled during the 6--12 ns part of the MD run shown in [Figure S1](#pone.0017459.s001){ref-type="supplementary-material"}. (PDB) ::: ::: {.caption} ###### Click here for additional data file. ::: PDB S3 ::: {.caption} ###### Atomic coordinates in pdb format for the structure of *Nm.* KDO8PS which, based on a clustering procedure, is the 3^rd^ most representative conformer of all the states sampled during the 6--12 ns part of the MD run shown in [Figure S1](#pone.0017459.s001){ref-type="supplementary-material"}. (PDB) ::: ::: {.caption} ###### Click here for additional data file. ::: Energy Matrix S1 ::: {.caption} ###### FoldX energy matrix. The rows (1--280, top to bottom) of the matrix correspond to the amino acid sequence of *Nm.* KDO8PS. The columns (1--20, left to right) of the matrix correspond to the 20 common amino acids in the following order: G A L V I P R T S C M K E Q D N W Y F H. ΔΔG values are in kcal/mol. (TXT) ::: ::: {.caption} ###### Click here for additional data file. ::: **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by U.S. Public Health Service Grants GM41857 to SHA, GM69840 to DLG, and by a Wayne State University Research Enhancement Program in Computational Biology Grant to DLG. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SHA DLG. Analyzed the data: SHA DLG. Wrote the paper: SHA DLG.

PubMed Central

2024-06-05T04:04:19.222691

2011-3-9

PMC3052367

Introduction {#s1} ============ *Mycobacterium tuberculosis* has an array of proteins to ensure its existence during the course of infection. In order to thrive and maintain its homeostasis, the pathogen continuously influences its surroundings mainly through surface-located sensor proteins. Extracellular signals are communicated through the sensors to the cytosol leading to the appropriate cell responses. Apparently, a large number of pathogens employ reversible phosphorylation of proteins by kinases and phosphatases as a way of transmitting the signals from extracellular milieu which helps in their survival and pathogenicity [@pone.0017871-Echenique1]--[@pone.0017871-Wang1]. Kinases carry out the phosphorylation by transferring the phosphate moiety on target proteins and phosphatases convert them back to the unphosphorylated state, either by dephosphorylating the substrate or by regulating the activity of kinases. Apart from the well recognized two component systems targeting His/Asp residues in bacteria, Ser, Thr and Tyr residues are also the major targets for phosphorylation. *M. tuberculosis* is known to have 11 Ser/Thr protein kinases (STPKs PknA-L, except C), one tyrosine kinase (PtkA), one Ser/Thr phosphatase (PstP) and two tyrosine phosphatases (PtpA and PtpB) [@pone.0017871-Bach1], [@pone.0017871-Cole1]. Till date a large number of mycobacterial proteins are shown to be regulated through phosphorylation by STPKs [@pone.0017871-Chao1]--[@pone.0017871-Arora1]. Some of these substrates are also known to be dephosphorylated by PstP [@pone.0017871-Gupta1], [@pone.0017871-Arora1]--[@pone.0017871-Sureka1]. PstP is a PP2C phosphatase (PPM family) that strictly requires Mn^2+^-ion for its activity [@pone.0017871-Chopra1]. It is a membrane localized enzyme with intracellular catalytic domain of 237 amino acids joined by a juxtamembrane region to the extracellular domain of 191 residues with a single transmembrane helix [@pone.0017871-Pullen1]. Using multi-wavelength anomalous diffraction studies, Pullen *et* al. determined the structure of the catalytic phosphatase domain of PstP [@pone.0017871-Pullen1]. PstP contains three metal-binding centers in its structure in contrast to two metal centers found in most of the PP2C phosphatases. Using atomic absorption spectroscopy and X-ray analysis, it has been shown that all the bound metal-ions are Mn^2+^. Similarities between Human Ser/Thr phosphatase PP2Cα and the mycobacterial enzyme have been explained on the basis of structural folds, metal binding and conserved residues [@pone.0017871-Pullen1]. Mutational analyses of PP2Cα have depicted the significance of certain conserved amino acid residues [@pone.0017871-Jackson1]. The corresponding residues in PstP are involved in binding to metal-ions and catalysis in addition to managing the binding and release of phosphate moiety. These residues in PP2Cα are critical for its activity [@pone.0017871-Jackson1] and thus, they are hypothesized to be important for PstP also. The interesting feature of *M. tuberculosis* Ser/Thr signaling molecules is that both the essential STPKs, PknB (Rv0014c) and PknA (Rv0015c) and the only Ser/Thr phosphatase PstP (Rv0018c) are located in the same genomic cluster which is conserved in several mycobacterial species [@pone.0017871-Cole1], [@pone.0017871-Gupta1], [@pone.0017871-Kang1]. Transcriptional analysis in earlier studies revealed that PknA, PknB and PstP show similar expression profiles [@pone.0017871-Kang1] and thus, implicate that strong regulation is required for their own functions as both the classes of enzymes functionally counteract each other. In this study, we show that the activity of PstP is modulated by phosphorylation. This is the first report on the regulation of any bacterial Ser/Thr phosphatase by post-translation modification. PstP was found to be phosphorylated differentially by PknA and PknB, both *in vitro* and in the surrogate host *Escherichia coli*. Additionally, we found that zinc ions (Zn^2+^) and inorganic phosphate (Pi) can inhibit the activity of PstP which in turn affects the phosphorylation status of both the kinases and phosphatase. Materials and Methods {#s2} ===================== Bacterial strains and growth conditions {#s2a} --------------------------------------- *E. coli* DH5α strain (Novagen) was used for cloning and BL21 (DE3) (Stratagene) was used for the expression of recombinant proteins. *E. coli* cells were grown and maintained with constant shaking (220 rpm) at 37°C in LB medium supplemented with 100 µg/ml ampicillin. Gene manipulation {#s2b} ----------------- The genes coding for PknA~c~ (*rv0015c*, representing the **[c]{.underline}**ytosolic region of 1-337aa) and PstP (*rv0018c*, PstP: 1-514aa) were PCR amplified using *M. tuberculosis* H37Rv genomic DNA. Resulting PCR products were digested with corresponding restriction enzymes and ligated into the vectors pProEx-HTc (Invitrogen) and/or pGEX-5X-3 (GE Healthcare Bio-Sciences) previously digested with the same enzymes. *Htc-PknB~c~* and *Htc-PstP~c~* were obtained as described earlier [@pone.0017871-Gupta1]. *pGEX-PknB~c~* was sub-cloned from *Htc-PknB~c~* using standard protocols under the same restriction sites. For cloning in dual-expression vector pETDuet-1 (Novagen), genes coding for PstP~c~ or PstP~c~ ^D38G^ were inserted in MCS1 having N-terminal His~6~-tag while kinases PknA and PknB (full length) were cloned in MCS2 with N-terminal MBP-tag (Maltose-binding protein tag upstream of the kinase). MBP-alone (without kinase) was taken as control vector having PstP~c~ or PstP~c~ ^D38G^ in MCS1. The protocols used for cloning in pETDuet-1 have been discussed earlier [@pone.0017871-Khan1]. Mutagenesis of specific residues was carried out using the QuikChange XL site-directed mutagenesis kit (Stratagene) as per manufacturer\'s instructions. Mutants of PstP and PstP~c~ were created as R20G, D38G and D229G using *Htc-PstP* and *Htc-PstP~c~* as templates. *Htc-PstP~c~* was utilized for the generation of *Htc-PstP~c~^t5a^* and *Htc-PstP~c~^t141e^*. *Htc-pknB~c~* was employed as template for generation of double mutant *Htc-pknB~c~^t171/173d^*. The details of all the primers and clones are provided in [tables 1](#pone-0017871-t001){ref-type="table"} and [2](#pone-0017871-t002){ref-type="table"}, respectively. The integrity of all clones was confirmed by DNA sequencing (TCGA, New Delhi). ::: {#pone-0017871-t001 .table-wrap} 10.1371/journal.pone.0017871.t001 Table 1 ::: {.caption} ###### Primers used in the study. ::: {#pone-0017871-t001-1} Primer Name Sequence Details (5′→3′) [\*\*](#nt101){ref-type="table-fn"} ------------------------ -------------------------------------------------------------------------- PknB~c~ ^T171/173D^ FP CGGCAACAGCGTG[GAC]{.underline}CAG[GAC]{.underline}GCAGCAGTGATCG PknB~c~ ^T171/173D^ RP CGATCACTGCTGC[GTC]{.underline}CTG[GTC]{.underline}CACGCTGTTGCCG PknA FP TGATCGAAGCCG[GAATTC]{.underline}AGGGGGAACCATGA EcoR1 PknA~c~ RP AGCACCCCC[GCGGCCGC]{.underline}GAGCAGCGC[TCA]{.underline}CTGACCGGAC Not1 PstP~c~ ^D38G^ FP CTATTGGCCCTGGCCG[G]{.underline}CGGCATGGGTGGGCAT PstP~c~ ^D38G^ RP ATGCCCACCCATGCCG[C]{.underline}CGGCCAGGGCCAATAG PstP~c~ ^R20G^ FP GATCGCGGCTTGGTA[G]{.underline}GCGCCAACAACGAAGACTCGGTC PstP~c~ ^R20G^ RP GACCGAGTCTTCGTTGTTGGCGC[C]{.underline}TACCAAGCCGCGATC PstP~c~ ^D229G^ FP GGCGGCGGCCCCG[G]{.underline}CAACGTCACTGTCGTCGTC PstP~c~ ^D229G^ RP GACGACGACAGTGACGTTG[C]{.underline}CGGGGCCGCCGCC PstP~c~ ^T5A^ FP GGAGAGTGGCGCGCGTG[G]{.underline}CCCTGGTCCTGCGATAC PstP~c~ ^T5A^ RP GTATCGCAGGACCAGGG[C]{.underline}CACGCGCGCCACTCTCC PstP~c~ ^T141E^ FP GACGACACGTTTGTCCAA[GC]{.underline}GCTGGTCGACGAAGGCCG PstP~c~ ^T141E^ RP CGGCCTTCGTCGACCAGC[GC]{.underline}TTGGACAAACGTGTCGTC pETDuet-PstP FP CACC [GCGGCCGC]{.underline}TCATATG GCGCGCGTGACCCTGG Not1 pETDuet-PstP~c~ RP CGGTCACCAGT[GCGGCCGC]{.underline}GAATGCTCACCGTCGGCC Not1 \*\*Restriction sites/stop codon/mutated sequences have been underlined. ::: ::: {#pone-0017871-t002 .table-wrap} 10.1371/journal.pone.0017871.t002 Table 2 ::: {.caption} ###### Description of the plasmids used in this study. ::: {#pone-0017871-t002-2} Plasmid construct Description Reference ---------------------------------- ------------------------------------------------------------------------------------------------------------------ ----------------------------------- pProEx-HTc *E. coli* expression vector containing N-terminal His~6~-tag Invitrogen pProEx-HTc-PknB~c~ Expression of His~6~PknB~1-331~ (cytosolic domain) [@pone.0017871-Gupta1] pProEx-HTc-PknB~c~ ^T171/173D^ pProEx-HTc-PknB~c~ with activation loop residues Thr^171^ and Thr^173^ mutated to Asp, phosphomimetic amino acid This study pProEx-HTc-PknA~c~ Expression of His~6~PknA~1-337~ (cytosolic domain) This study pProEx-HTc-PstP~c~ Expression of His~6~PstP~1-300~ (cytosolic domain) [@pone.0017871-Gupta1] pProEx-HTc-PstP~c~ ^R20G^ pProEx-HTc-PstP~c~ with Arg^20^ mutated to Gly This study pProEx-HTc-PstP~c~ ^D38G^ pProEx-HTc-PstP~c~ with Asp^38^ mutated to Gly This study pProEx-HTc-PstP~c~ ^D229G^ pProEx-HTc-PstP~c~ with Asp^229^ mutated to Gly This study pProEx-HTc-PstP~c~ ^T5A^ pProEx-HTc-PstP~c~ with Thr^5^ mutated to Ala This study pProEx-HTc-PstP~c~ ^T141E^ pProEx-HTc-PstP~c~ with Thr^141^ mutated to Glu This study pGEX-5X-3 *E. coli* expression vector containing N-terminal Glutathione S-Transferase tag GE Healthcare pGEX-5X-3-PknA~c~ Expression of GST-PknA~1-337~ (cytosolic domain) This study pGEX-5X-3-PknB~c~ Expression of GST-PknB~1-331~ (cytosolic domain) This study pETDuet1 *E. coli* dual expression vector containing N-terminal His~6~-tag in MCS1 and C-terminal S-tag in MCS2 Novagen pETDuet1-PstP~c~ ^D38G^/MBP Expression of His~6~-PstP~c~ ^D38G^ in MCS1 with Myelin basic protein (MBP) in MCS2 This study, [@pone.0017871-Khan1] pETDuet1-PstP~c~ ^D38G^/MBP-PknA Expression of His~6~-PstP~c~ ^D38G^ in MCS1 with MBP-tagged PknA in MCS2 This study, [@pone.0017871-Khan1] pETDuet1-PstP~c~ ^D38G^/MBP-PknB Expression of His~6~-PstP~c~ ^D38G^ in MCS1 with MBP-tagged PknB in MCS2 This study, [@pone.0017871-Khan1] ::: Protein expression and purification {#s2c} ----------------------------------- Proteins were expressed and purified from *E. coli* as described previously [@pone.0017871-Gupta1]. The purified proteins were assessed by SDS-PAGE and concentrations were estimated by Bradford assay (Bio-Rad). *In vitro* kinase assays and phosphoamino acid analysis {#s2d} ------------------------------------------------------- *In vitro* phosphorylation of PstP~c~ or its mutants (0.5--3 µg) by PknA~c~ (0.5--1 µg) or PknB~c~ (1--3 µg) was carried out in kinase buffer (20 mM PIPES \[pH 7.2\], 5 mM MnCl~2~, 5 mM MgCl~2~) containing 2 µCi \[γ-^32^P\]ATP (BRIT, Hyderabad, India) followed by incubation at 25°C for 20 min. Reactions were terminated by 5X SDS sample buffer followed by boiling at 100°C for 5 min. Proteins were separated by 12% SDS-PAGE and analyzed by PhosphorImager (FLA 2000, Fuji). Zn^2+^ and Pi were added to the kinase assay reactions as per requirement of the assay. For the visualization of phosphorylation signal on cleaved proteins, removal of recombinant tags was achieved by addition of TEV protease (for His~6~-tagged PstP/PstP~c~ and their mutants) in TEV buffer (Tris-Cl \[pH 8.5\], 5 mM EDTA, 300 mM NaCl and 1 mM DTT) after the kinase reaction followed by an additional incubation for 2 hr at 20°C. For phosphoamino acid analysis, PstP~c~ ^D38G^ was phosphorylated by PknB~c~ and PknA~c~ and cleaved with TEV protease as mentioned above, separated by SDS-PAGE and electroblotted onto Immobilon PVDF membrane (Millipore). Phosphoamino acid analysis by two-dimensional thin layer electrophoresis was performed as described earlier [@pone.0017871-Gupta1], [@pone.0017871-Boyle1]. *In vitro* dephosphorylation and *p*-nitrophenol phosphate (*p*NPP) hydrolysis assays {#s2e} ------------------------------------------------------------------------------------- PknB~c~ and PknA~c~ were autophosphorylated by *in vitro* kinase assays using \[γ-^32^P\]ATP. 1 µg of purified PstP~c~/PstP~c~ ^D38G^/PstP~c~ ^R20G^/PstP~c~ ^D229G^ were added in four sets of reactions and incubated at 25°C for increasing time points up to 30 min to measure the dephosphorylation potential of PstP~c~ and its mutants. For auto-dephosphorylation assays, PknB~c~ and PknB~c~ ^T171/173D^ (2 µg each) were autophosphorylated by *in vitro* kinase assays and exposed to dephosphorylation by PstP~c~ and PstP~c~ ^D38G^ (1 µg). Reactions were stopped by adding 5X SDS sample buffer and boiled for 5 min at 100°C. The samples were separated by 12% SDS-PAGE and phosphorylated bands were observed and analysed by PhosphorImager. *p*NPP hydrolysis assay was performed as a measure of phosphatase activity. PstP~c~ was added to a reaction mixture containing phosphatase assay buffer (50 mM Tris pH 8.0, 5 mM DTT, 4 mM MnCl~2~) and 10 mM *p*NPP in a 96-well plate and incubated at 37°C for indicated time points and absorbance was read at 405 nm (Microplate reader, Bio-Rad). To assay the relative activity of PstP~c~ and its phosphatase-deficient variants, increasing concentrations of enzymes were added to the reaction mix and processed as above. Alkaline phosphatase (Roche) and PknB~c~ were taken as positive and negative controls, respectively, for the *p*NPP hydrolysis assays. Variations of PstP~c~ activity by addition of Zn^2+^ and Pi was assessed by adding ZnCl~2~ or sodium phosphate \[pH 7.2\] to the reaction mixture as above, to achieve the indicated final concentrations. pETDuet-1 purified PstP~c~ and PstP~c~ ^D38G^, co-expressed with or without kinases, were employed for *p*NPP-assays to measure the effect of phosphorylation on their activities. Metabolic labeling in *E. coli* {#s2f} ------------------------------- The procedure described by Kumar *et* al. was followed for metabolic labeling [@pone.0017871-Kumar1]. *E. coli* (BL21-DE3) transformants harbouring either *pETDuet-PstP~c~/PstP~c~^D38G^-mbp* or *pETDuet-PstP~c~/PstP~c~^D38G^-mbpPknA* or *pETDuet-PstP~c~/PstP~c~^D38G^-mbpPknB* were grown in 5 ml LB medium containing 100 µg/ml ampicillin to an O.D~600~ of ∼0.6. The cells were induced with 1 mM IPTG and further grown for 4 hr at 16°C. Cultures were harvested, washed with 5 ml of M9 medium \[pH 7.0\] without phosphate salts (for 1 L: NH~4~Cl-1 g, NaCl-0.5 g, 20% Glucose-10 ml, MgSO~4~.7H~2~O-1 ml, Thiamine-HCl-1 ml, CaCl~2~-1 ml). The cells were resuspended in 2 ml of M9 media supplemented with 1 mCi of \[^32^P\]orthophosphoric acid (BRIT, Hyderabad, India), 100 µg/ml ampicillin and 1 mM IPTG and further grown at 16°C for 4 hr. Under specific conditions, Sodium phosphate \[pH 7.2\] (2 mM) or ZnCl~2~ (4 mM) were added to M9 media and subsequent processing steps of metabolic labeling. The cells were harvested and lysed by sonication in the lysis buffer containing phosphate-buffered saline, 5% glycerol and protease inhibitor cocktail. The cell lysate was clarified and the lysates containing His~6~-fusion protein were incubated with lysis buffer equilibrated Ni^2+^-NTA affinity beads for 2 hr at 4°C. The beads were then thoroughly washed with lysis buffer containing 20 mM imidazole and resuspended in 5X SDS sample buffer followed by boiling for 15 min. The samples were resolved on SDS-PAGE followed by autoradiography. Identification of phosphorylation sites in PstP~c~ ^D38G^ {#s2g} --------------------------------------------------------- PknB~c~ and PknA~c~ were employed for *in vitro* kinase assay using 50 µM cold ATP and PstP~c~ ^D38G^. The samples were run on 12% SDS-PAGE, stained with Coomassie Brilliant Blue and de-stained. Bands corresponding to PstP~c~ ^D38G^ were excised from the gel and washed with MilliQ water. The samples were processed for identification of phosphorylation sites by using Thermo-Finnagen LTQ electrospray instrument (Proteomics Core Facility, Children\'s Hospital, Boston). The detailed protocol of sample processing for identification of phosphorylation sites has been provided in [File S1](#pone.0017871.s006){ref-type="supplementary-material"}. Generation of polyclonal antibodies for PstP~c~ in rabbit and immunoblotting {#s2h} ---------------------------------------------------------------------------- Polyclonal antibodies against PstP~c~ were generated in rabbit. To confirm the presence of PstP~c~/PstP~c~ ^D38G^ in Ni^2+^-NTA pulled-out proteins after metabolic labeling by western blot analysis, the samples were resolved by SDS-PAGE along with positive (purified PstP~c~) and negative controls (GST-PknB~c~) and transferred onto nitrocellulose membrane (Bio-Rad). Standard procedure for immunoblotting was followed [@pone.0017871-Gupta1], [@pone.0017871-Arora1]. The blots were developed using SuperSignal^R^ West Pico Chemiluminescent Substrate kit (Pierce Protein Research Products) according to manufacturer\'s instructions. Results {#s3} ======= Identification of the residues critical for the activity of PstP {#s3a} ---------------------------------------------------------------- On the basis of structural data available for PstP and alignment with the residues important for Human PP2Cα activity [@pone.0017871-Pullen1], PstP~c~ mutants were generated using site-directed mutagenesis. These residues include the Mn^2+^-ion binding sites-Asp^38^ and Asp^229^ and phosphate (Pi) binding residue-Arg^20^ ([Figure 1A](#pone-0017871-g001){ref-type="fig"}). In the resulting mutants, these sites were converted to Glycine (PstP~c~ ^D38G^, PstP~c~ ^D229G^ and PstP~c~ ^R20G^). The activity of these mutants was compared using chromogenic substrate *p*NPP. To confirm the authenticity of the assay, increasing concentrations of alkaline phosphatase were utilized as a positive control while PknB~c~ was used as negative control ([Figure S1](#pone.0017871.s001){ref-type="supplementary-material"}). The *p*NPP assay with increasing amounts of PstP~c~-mutants showed that the mutation of Asp^38^ and Asp^229^ to Gly resulted in \>90% loss of the dephosphorylation activity of PstP~c~, while the PstP~c~ ^R20G^ mutant lost about 60% of its activity ([Figure 1B and 1C](#pone-0017871-g001){ref-type="fig"}). Thus, Arg^20^, Asp^38^ and Asp^229^ were identified as the residues required for optimum activity of PstP. To confirm that the loss in activity was specifically due to mutagenesis of Asp^38^, Asp^229^ and Arg^20^, irrelevant residues (Thr^5^ and Thr^141^) in PstP~c~ were mutagenized to generate PstP~c~ ^T5A^ and PstP~c~ ^T141E^. The relative activities of these mutants were compared with the native enzyme through *p*NPP-assay ([Figure S2](#pone.0017871.s002){ref-type="supplementary-material"}). There were no significant changes observed in the mutants in comparison to PstP~c~, thus reinforcing the importance of Arg^20^, Asp^38^ and Asp^229^ residues. ::: {#pone-0017871-g001 .fig} 10.1371/journal.pone.0017871.g001 Figure 1 ::: {.caption} ###### Critical residues of PstP. (**A**) Schematic representation of PstP with critical residues (Arg^20^, Asp^38^ and Asp^229^) being highlighted with upward arrows. (**B**) Activity profiles of PstP~c~ and its mutants: Activity assays were performed by *p*NPP-hydrolysis mediated by PstP~c~, PstP~c~ ^R20G^, PstP~c~ ^D38G^ and PstP~c~ ^D229G^. Increasing concentrations of proteins were taken with constant substrate concentration (10 mM *p*NPP) and incubated at 37°C for 30 mins. As shown in the graph, the mutants had lost phosphatase activity to different extents. Activity is calculated as a measure of µmoles of *p*NPP hydrolyzed per min. at a given enzyme concentration. (**C**) The relative activity of all the phosphatase variants (5 µg each, 30 min.) showed that PstP~c~ ^D38G^ and PstP~c~ ^D229G^ had lost \>90% of activity while PstP~c~ ^R20G^ lost ∼60% of the activity as compared to PstP~c~. The error bars indicate the SD of three individual experiments. :::  ::: Phosphatase activity of PstP~c~ and its mutants {#s3b} ----------------------------------------------- The dephosphorylation potential of PstP~c~ and its mutants was also assessed by their ability to dephosphorylate PknB~c~ in a time-dependent dephosphorylation ([Figure 2A](#pone-0017871-g002){ref-type="fig"}) and *p*NPP hydrolysis assays ([Figure S3](#pone.0017871.s003){ref-type="supplementary-material"}). PstP~c~ ^R20G^ dephosphorylated the autophosphorylated PknB~c~ to some extent, whereas substantial loss of phosphatase activity was observed with PstP~c~ ^D38G^ and PstP~c~ ^D229G^ ([Figure 2A](#pone-0017871-g002){ref-type="fig"}). The activity of PstP~c~ ^D229G^ was relatively higher than that of PstP~c~ ^D38G^ as opposed to the observation in *p*NPP-assays ([Figures 1C](#pone-0017871-g001){ref-type="fig"} and [S3](#pone.0017871.s003){ref-type="supplementary-material"}). Similar observations have been reported earlier where the activity of an enzyme, specifically Ser/Thr phosphatases, is shown to be dependent on the nature of substrate [@pone.0017871-Dahche1]--[@pone.0017871-Sugiura1]. *p*NPP is an artificial substrate while PknB is a natural substrate of PstP, which may be recognized and subsequently dephosphorylated more optimally. Additionally, in this case, the activity of the phosphatase also depends on the activity of PknB, as discussed in later sections. The assays were also performed using autophosphorylated PknA~c~ which showed similar results (data not shown). Surprisingly, in this assay, additional phosphorylated bands corresponding to the size of PstP~c~ ^D38G^ were observed when incubated with kinase for longer time. No such bands were observed with PstP~c~, PstP~c~ ^R20G^ and PstP~c~ ^D229G^ at the given concentrations. ::: {#pone-0017871-g002 .fig} 10.1371/journal.pone.0017871.g002 Figure 2 ::: {.caption} ###### Dephosphorylation by PstP~c~ and its mutants. (**A**) Autoradiogram showing autophosphorylated PknB~c~, exposed to dephosphorylation by PstP~c~, PstP~c~ ^R20G^, PstP~c~ ^D38G^ and PstP~c~ ^D229G^. Time-dependent dephosphorylation was performed with 1 µg of phosphatase after carrying out autophosphorylation of PknB~c~ (2 µg) in an *in vitro* kinase assay. Noticeably, PstP~c~ ^D38G^ was observed to be phosphorylated with increasing time points (3^rd^ panel from the top). (**B**) Autoradiogram showing phosphorylation of PstP~c~ by PknA~c~ (1 µg). Increasing concentrations of PstP~c~ were used to measure the extent of dephosphorylation. Unexpectedly, the phosphatase itself got phosphorylated at higher kinase to phosphatase ratio, though kinase was completely dephosphorylated. No phosphorylation was observed at higher PstP~c~ concentrations. :::  ::: To further assess this observation, PknA~c~ or PknB~c~ were incubated with increasing concentrations of PstP~c~. Interestingly, PstP~c~ was phosphorylated by PknA~c~ at higher kinase to phosphatase ratio ([Figure 2B](#pone-0017871-g002){ref-type="fig"}). An increase in the concentration of PstP~c~ resulted in complete dephosphorylation of both the proteins. This serendipitous observation intrigued us to explore whether PstP is a target of Ser/Thr protein kinases. Due to strong dephosphorylation activity of PstP, it was difficult to achieve the phosphotransfer on native phosphatase. Therefore, further studies were carried out with the mutants of PstP that were deficient in phosphatase activity. Phosphorylation of PstP~c~ ^D38G^, PstP~c~ ^D229G^ and PstP~c~ ^R20G^ {#s3c} --------------------------------------------------------------------- After identification of the residues critical for PstP~c~ activity and measuring the activity of corresponding mutants, the phosphorylation status of PstP~c~ mutants was studied. PknA and PknB were employed for the phosphorylation assays. PstP~c~ ^D38G^ and PstP~c~ ^D229G^ were efficiently phosphorylated by both PknA~c~ and PknB~c~ ([Figure 3A](#pone-0017871-g003){ref-type="fig"}), whereas faint signal on PstP~c~ ^R20G^ was observed owing to its partial phosphatase activity. Phosphorylation of PstP~c~ (at 3 µg concentration) was not observed by *in vitro* kinase assay as it completely dephosphorylated PknA~c~ and PknB~c~, making them inactive (heat-inactive PstP~c~ was found to be phosphorylated-data not shown). To confirm that the observed phosphorylation is on PstP~c~-mutants and not on the N-terminally attached His~6~-tag, TEV-protease cleavage of the tag was performed after the kinase assays. Phosphorylation was confirmed to be specifically localized on the cleaved substrate protein (data not shown). Additionally, the R20G, D38G and D229G mutants were also created in full length PstP construct and *p*NPP-hydrolysis assays and phosphorylation reactions were also confirmed using full length PstP and its mutants (data not shown). ::: {#pone-0017871-g003 .fig} 10.1371/journal.pone.0017871.g003 Figure 3 ::: {.caption} ###### Phosphorylation of PstP~c~ and its mutants by PknA and PknB. (**A**) Phosphorylation of PstP~c~ and its mutants (3 µg each) by 2 µg PknB~c~ (upper panel) and 0.5 µg PknA~c~ (middle panel). PstP~c~ ^D38G^ and PstP~c~ ^D229G^ were efficiently phosphorylated by both the kinases due to loss of phosphatase activity. Phosphorylation on PstP~c~ ^R20G^ mutant was low due to its partial phosphatase activity. The corresponding SDS-PAGE is shown (lowest panel) as a loading control. (**B**) Phosphoamino acid analysis by 2D-TLE illustrates that both PknA~c~ (upper panel) and PknB~c~ (lower panel) phosphorylates PstP~c~ ^D38G^ on Thr residues. (**C**) Sites of phosphorylation of PknB~c~ (blue) and PknA~c~ (green) in PstP~c~ ^D38G^ were identified by mass spectrometric analysis. PknB~c~ phosphorylates PstP~c~ ^D38G^ majorly on four Thr residues-Thr^137^, Thr^141^, Thr^174^ and Thr^290^ while two Thr residues were phosphorylated by PknA~c~-Thr^137^ and Thr^174^. :::  ::: Phosphoamino acid analysis and identification of phosphorylation site(s) of PknA and PknB in PstP~c~ ^D38G^ {#s3d} ----------------------------------------------------------------------------------------------------------- Phosphoamino acid analysis by two-dimensional thin layer electrophoresis showed that both PknA~c~ ([Figure 3B](#pone-0017871-g003){ref-type="fig"}, upper panel) and PknB~c~ ([Figure 3B](#pone-0017871-g003){ref-type="fig"}, lower panel) phosphorylated PstP~c~ ^D38G^ on Thr residues while no signal was observed on the spots corresponding to pSer and pTyr. For further experiments, PstP~c~ ^D38G^ was utilized. The sites of PknA and PknB phosphorylation on PstP~c~ ^D38G^ were identified through mass-spectrometric analysis by Thermo-Finnagen LTQ electrospray Mass-Spectrometer, using *in vitro* phosphorylated protein. The results showed that four Thr residues were phosphorylated by PknB (Thr^137^, Thr^141^, Thr^174^ and Thr^290^) while PknA phosphorylated PstP~c~ ^D38G^ on two residues (Thr^137^ and Thr^174^) ([Figure 3C](#pone-0017871-g003){ref-type="fig"}, supplementary file 2). Thus, PstP~c~ ^D38G^ is differentially phosphorylated by PknA and PknB which may have important implications on the activity of PstP. Validation of PstP phosphorylation in *E. coli* {#s3e} ----------------------------------------------- To further substantiate our results, the phosphorylation status of PstP~c~ and PstP~c~ ^D38G^ was examined specifically by PknA and PknB in *E. coli* using a dual expression system. PstP~c~ and PstP~c~ ^D38G^ were cloned in pETDuet1 expression vector along with either MBP alone or MBP-tagged PknA or PknB. *E. coli* BL21 (DE3) cells transformed with *pETDuet1-PstP~c~/PstP~c~^D38G^-MBP* or *pETDuet1-PstP~c~/PstP~c~^D38G^-MBP-kinase* (kinase, PknA or PknB) were metabolically labelled with \[^32^P\]orthophosphoric acid. Phosphorylation of PstP~c~ and PstP~c~ ^D38G^ could only be detected when PknA or PknB were co-expressed ([Figures 4A and 4B](#pone-0017871-g004){ref-type="fig"}), suggesting the phosphorylation of phosphatase by both the kinases in native conditions in *E. coli*. Western blot analysis of Ni^2+^-NTA purified samples using rabbit anti-PstP~c~ antibodies confirmed the metabolically labelled protein to be PstP~c~ (data not shown). ::: {#pone-0017871-g004 .fig} 10.1371/journal.pone.0017871.g004 Figure 4 ::: {.caption} ###### Co-expression analysis of STPKs and PstP~c~/PstP~c~ ^D38G^. (**A**) Metabolic labeling of PstP~c~: PstP~c~ co-expressed with MBP-PknA (lane 2) or MBP-PknB (lane 3) gets phosphorylated in *E. coli* under native conditions while PstP~c~ co-expressed with MBP alone (lane 1) was not phosphorylated. (**B**) Metabolic labeling of PstP~c~ ^D38G^: PstP~c~ ^D38G^ co-expressed with MBP-PknA (lane 2) or MBP-PknB (lane 3) gets phosphorylated in *E. coli* while PstP~c~ ^D38G^ co-expressed with MBP alone (lane 1) was not phosphorylated. As expected, the intensity of phosphorylation on PstP~c~ ^D38G^ was comparatively higher than that of PstP~c~. (**C**) Relative activity profile of pETDuet1 purified PstP~c~ and (**D**) PstP~c~ ^D38G^: *p*NPP assays were performed with PstP~c~ and PstP~c~ ^D38G^ (1 µg each) purified from pETDuet1 co-expressing MBP or MBP-PknA/PknB. The dephosphorylation potential of phosphorylated PstP~c~ and PstP~c~ ^D38G^ (co-expressed with either kinase) is higher than that of unphosphorylated protein. For PstP~c~ ^D38G^, activity was evaluated over long time points due to its low dephosphorylation activity. Activity is calculated as a measure of µmoles of *p*NPP hydrolyzed per µg of protein at a given time. The error bars indicate the SD of three individual experiments. (**E**) Relative dephosphorylation of PknA~c~ by pETDuet-1 purified PstP~c~ ^D38G^: Autophosphorylated PknA~c~ was incubated for 30 mins with unphosphorylated and phosphorylated PstP~c~ ^D38G^ and the extent of dephosphorylation was assessed by *in vitro* dephosphorylation assays. The image obtained after autoradiography was analyzed by ImageGauge software (Fuji) and relative intensity of phosphorylation was measured: (1) PknA~c~ alone, (2) PknA~c~+MBP-PstP~c~ ^D38G^, (3) PknA~c~+PstP~c~ ^D38G^ phosphorylated by PknA and (4) PknA~c~+PstP~c~ ^D38G^ phosphorylated by PknB. As shown, the PknA-phosphorylated PstP~c~ ^D38G^ dephosphorylated the kinase to a greater extent in comparison to the unphosphorylated PstP~c~ ^D38G^. The error bars represent the SD of the three individual experiments. The corresponding autoradiogram is shown in [Figure S4](#pone.0017871.s004){ref-type="supplementary-material"}. :::  ::: Activity assays of pETDuet1-purified PstP~c~ and PstP~c~ ^D38G^ {#s3f} --------------------------------------------------------------- The activity profiles of PstP~c~ and PstP~c~ ^D38G^ co-expressed with and without PknA/PknB, were evaluated. According to the *p*NPP assays, the activity of phosphorylated PstP~c~ (co-expressed with PknA or PknB) was higher than that of unphosphorylated phosphatase (co-expressed with MBP alone) ([Figure 4C](#pone-0017871-g004){ref-type="fig"}). The phenomenon was also confirmed by measuring the activity of PstP~c~ ^D38G^. As already discussed, PstP~c~ ^D38G^ had retained about 10% of the dephosphorylation activity as a result of which, it was phosphorylated efficiently by kinases. The relative activity of phosphorylated PstP~c~ ^D38G^ with PknA/PknB and unphosphorylated protein was measured for 420 min. Interestingly, the activity of phosphorylated PstP~c~ ^D38G^ was remarkably higher than that of unphosphorylated protein, thus the similar profile as that of PstP~c~ was observed ([Figure 4D](#pone-0017871-g004){ref-type="fig"}). Also, the activity of PknA phosphorylated phosphatase was even more than the protein phosphorylated by PknB. Noticeably, the increase in phosphatase activity after phosphorylation may also account for the observed increase in the activity of PstP~c~ ^D229G^ in the time-dependent dephosphorylation assays ([Figure 2A](#pone-0017871-g002){ref-type="fig"}). The dephosphorylation of *in vitro* autophosphorylated PknA~c~ was assessed by PstP~c~ ^D38G^+MBP, PstP~c~ ^D38G^+MBP-PknA and PstP~c~ ^D38G^+MBP-PknB. As expected, due to higher activity of phosphorylated PstP~c~ ^D38G^, intensity of phosphorylation on PknA~c~ was low as compared to the reaction containing unphosphorylated PstP~c~ ^D38G^+MBP ([Figures 4E](#pone-0017871-g004){ref-type="fig"} and [S4](#pone.0017871.s004){ref-type="supplementary-material"}). Also, since PknA-phosphorylated PstP~c~ ^D38G^ was more active than PknB-phosphorylated PstP~c~ ^D38G^ ([Figure 4D](#pone-0017871-g004){ref-type="fig"}), the extent of dephosphorylation was more in lane 3 as compared to lane 4. Auto-dephosphorylation of PstP~c~ {#s3g} --------------------------------- Next, we tried to understand whether the inability of PstP~c~ to be effectively phosphorylated was due to its dephosphorylation activity on the kinases resulting in their inactivation or it was due to auto-dephosphorylation. Consequently, phosphomimetic mutants of PknB~c~ were generated for the Thr residues of activation loop in catalytic domain [@pone.0017871-Boitel1], forming PknB~c~ ^T171/173D^ which cannot be dephosphorylated by PstP~c~ on Thr^171^ and Thr^173^. As reported by Boitel *et* al., PknB does not lose phosphorylation signals after mutagenesis of Thr^171^ and Thr^173^. Through a series of careful analysis of single and double mutants of PknB, it has been shown that PknB can be additionally phosphorylated on Ser^166^ and/or Ser^169^ residues [@pone.0017871-Boitel1]. Thus, we utilized PknB~c~ and PknB~c~ ^T171/173D^, that were autophosphorylated in an *in vitro* kinase assay using \[γ-^32^P\]ATP, before incubation with PstP~c~. Phosphorylation of PstP~c~ was still not observed with constitutively active PknB~c~ ^T171/173D^, as confirmed by phosphotransfer observed on PstP~c~ ^D38G^ ([Figure 5A](#pone-0017871-g005){ref-type="fig"}). This suggests that PstP~c~ can dephosphorylate itself. Additionally, PknB~c~ ^T171/173D^ was completely dephosphorylated in presence of PstP~c~, suggesting that PstP could also dephosphorylate the surplus sites Ser^166^/Ser^169^. ::: {#pone-0017871-g005 .fig} 10.1371/journal.pone.0017871.g005 Figure 5 ::: {.caption} ###### Factors affecting PstP activity. (**A**) Auto-dephosphorylation of PstP~c~: Autoradiogram showing phosphorylation by PknB~c~. PstP~c~ and PstP~c~ ^D38G^ (3 µg each) were used for *in vitro* phosphorylation assay by PknB~c~ and PknB~c~ ^T171/173D^ (2 µg each). Since PknB~c~ ^T171/173D^ cannot be dephosphorylated by PstP~c~, lack of signal signifies auto-dephosphorylation of phosphatase. PstP~c~ ^D38G^ was used as positive control to show that PknB~c~ ^T171/173D^ is active. Regulation of PstP~c~ activity: *p*NPP assay showing the effect on activity of PstP~c~ (1 µg) by (**B**) Zn^2+^ and (**C**) Pi. *p*NPP assay was carried out for 30 mins and activity was calculated as a measure of µmoles of *p*NPP hydrolyzed per min per µg of protein. The error bars show SD of three independent experiments. (**D**) Phosphorylation of PstP~c~: Autoradiogram showing the phosphorylation of PstP~c~ (1 µg) by GST-PknA~c~ (left panel) and GST-PknB~c~ (right panel) in presence of 0.2 mM Zn^2+^ and 0.5 mM Pi. Since His~6~-tagged STPKs were not resolved properly from PstP~c~ on SDS-PAGE ([Figure S5](#pone.0017871.s005){ref-type="supplementary-material"}), the assay was also performed with GST-tagged kinases having higher molecular weights. (**E**) Metabolic labeling of PstP~c~ by PknA and PknB in *E. coli* in presence of Zn^2+^ and Pi: Phosphorylation level of PstP~c~ was observed to be increased when Zn^2+^ (4 mM) and Pi (2 mM) were added during the culture conditions and subsequent processing steps. The autoradiograms obtained after SDS-PAGE were analyzed by ImageGauge software and intensity of the band corresponding to PstP~c~ phosphorylation without any added factor was taken as 100%. Relative phosphorylation is depicted in the bar graph. :::  ::: Identification of the factors affecting the activity of PstP {#s3h} ------------------------------------------------------------ The phosphorylation of PstP suggested that additional factors may be involved in the cellular milieu that can regulate and control the phosphatase activity, preceding its phosphorylation. In general, phosphatases are known to be affected by a number of factors like metal-cations, Pi, creatine phosphate (CP) and ATP/ADP ratio. PstP~c~ activity assay was carried out in the presence of selected factors. Interestingly, activity of PstP~c~ was reduced in the presence of Zn^2+^ and Pi, as assessed by *p*NPP assay. Reduction of almost 50% activity was observed at 0.2 mM Zn^2+^ ([Figure 5B](#pone-0017871-g005){ref-type="fig"}) and 0.5 mM Pi ([Figure 5C](#pone-0017871-g005){ref-type="fig"}). Maximum inhibition of PstP~c~ was observed at 1 mM Zn^2+^ and 4 mM Pi. Inhibition by Zn^2+^ at \>1 mM was not calculable due to protein precipitation in the reaction mixture. Phosphorylation of PstP~c~ in the presence of Zn^2+^ and Pi {#s3i} ----------------------------------------------------------- The inhibition of PstP~c~ in the presence of Zn^2+^ and Pi provided a condition that could favour the phosphorylation of PstP~c~ by STPKs. PstP~c~ was indeed phosphorylated by PknA~c~ and PknB~c~ in presence of 0.2 mM Zn^2+^ or 0.5 mM Pi ([Figure 5D](#pone-0017871-g005){ref-type="fig"}), under *in vitro* conditions. Since the phosphorylated bands of His~6~-tagged PknA~c~/PknB~c~ and PstP~c~ were not able to resolve on SDS-PAGE ([Figure S5](#pone.0017871.s005){ref-type="supplementary-material"}), the assay was performed with GST-tagged kinases and similar results were obtained. To further assess the effects of Zn^2+^ and Pi, metabolic labeling of PstP~c~ by co-expressed kinases PknA and PknB was performed in *E. coli* in the presence of Zn^2+^ (4 mM) and Pi (2 mM) ([Figure 5E](#pone-0017871-g005){ref-type="fig"}). Phosphorylation of PstP~c~ was indeed enhanced in the presence of Zn^2+^ by ∼40%-50%. The enhancement in phosphorylation in the presence of Pi was not as prominent (∼10%--20%), possibly due to competition of phosphate ions with \[^32^P\]orthophosphoric acid. Nevertheless, as a proof of principle, Zn^2+^ and Pi were identified as the novel regulators which can inhibit the activity of PstP~c~ and facilitate its phosphorylation. Discussion {#s4} ========== The coordinated regulation of Ser/Thr protein kinases and phosphatases is essential for maintaining the appropriate equilibrium of protein phosphorylation. Membrane associated kinases and phosphatases are known or hypothesized to be regulated by external stimulus. It is of great relevance to decipher the regulatory mechanisms especially in the systems like *M. tuberculosis* where one Ser/Thr phosphatase PstP is accountable for the effects caused by 11 STPKs. In general, the processes involved in regulating the phosphatases include some external signals, variation in pH [@pone.0017871-Fjeld1], cellular concentrations of ATP, ADP, Pi (or their ratios) [@pone.0017871-Chen1], [@pone.0017871-Zhao1], cytosolic cations like Mn^2+^, Zn^2+^, Mg^2+^, Ca^2+^ [@pone.0017871-Chopra1], [@pone.0017871-Fjeld1], [@pone.0017871-Zhao1]--[@pone.0017871-Taylor1] and post-translation modifications (phosphorylation, methylation) [@pone.0017871-Chen1], [@pone.0017871-Shi1], [@pone.0017871-Ahn1]--[@pone.0017871-Kobayashi1]. Present study demonstrates an example of PknA and PknB mediated regulation of PstP through inter-dependent phosphorylation-dephosphorylation reactions. Regulation of phosphatases by phosphorylation is a critical step for cell signaling pathways. It is also associated with feedback phenomena in case where phosphatases are phosphorylated by the kinases that are in turn dephosphorylated by the same phosphatase. Certain examples illustrate the phosphorylation of PP2C phosphatases such as rat Mg^2+^-dependent protein phosphatase α (MPPα) by casein kinase II [@pone.0017871-Kobayashi1], Soybean kinase associated protein phosphatase (Soybean KAPP) [@pone.0017871-Miyahara1], *Oryza sativa* KAPP [@pone.0017871-vanderKnaap1], but these have not been detailed in terms of feedback regulation. PstP has conserved domain architecture of PP2C-phosphatases (PPM family). PPM family phosphatases play an imperative role in a number of systems described earlier [@pone.0017871-Adler1]--[@pone.0017871-Umezawa1]. Except a few PP2C-phosphatases like Human PP2Cα [@pone.0017871-Das1] and Arabidopsis KAPP [@pone.0017871-Stone1], not much is known about other members of this family. For PstP, we have previously shown that PknA and PknB are the targets for dephosphorylation by PstP and detailed the basic biochemical requirements of this enzyme along with its membrane localization [@pone.0017871-Chopra1]. In a later study, Pullen *et* al. resolved the crystal structure of PstP catalytic domain and described the most important features of this molecule having characteristic PP2C-fold along with three-metal binding centers that associate with Mn^2+^ [@pone.0017871-Pullen1]. The discovery of third-metal centre was a unique feature of PstP as other PP2C phosphatases were found to have two metal-binding centres. In the recent studies, the PP2C-phosphatases of *Streptococcus agalactiae* and *Thermosynechococcus elongatus* have been shown to have a similar third-metal binding centre [@pone.0017871-Rantanen1], [@pone.0017871-Schlicker1]. The third metal ion center in PstP is proposed to be involved in structural perturbations leading to altered phosphoprotein recognition profiles. In this study, three conserved residues were selected for generation of site-directed mutants in PstP~c~, on the basis of similarity with Human phosphatase PP2Cα [@pone.0017871-Pullen1]. Arg^20^ (PP2Cα Arg^33^) is responsible for hydrolysis of phosphate moiety from pSer/pThr residues in target proteins. Asp^38^ (PP2Cα Asp^60^) and Asp^229^ (PP2Cα Asp^282^) constitute a part of Mn^2+^-metal centers and coordinate with the two critical Mn^2+^. Mutations of Asp^38^ and Asp^229^ affected the activity of PstP rendering it active to minimal level, though R20G mutant retained about 40% activity. Thus, the residues that are involved in Mn^2+^-ion binding and hydrolysis of phosphate are deciphered to be critical for its activity. Accordingly, the extent of phosphorylation of each mutant was dependent on the remaining dephosphorylation activity, so that PstP~c~ ^D38G^ and PstP~c~ ^D229G^ were efficiently phosphorylated by PknA and PknB. Association with metals is crucial for PP2C phosphatases and any perturbation with inherently associated metals may lead to altered functional profile. The minimum requirement for PstP~c~ activity is the presence of Mn^2+^ [@pone.0017871-Chopra1]. For PP2C-class of phosphatases, divalent ions other than Mn^2+^/Mg^2+^ can inhibit their activity by competitively replacing the Mn^2+^ in the core enzyme structure [@pone.0017871-Fjeld1] and Zn^2+^ are the most potent regulators, having comparable ionic radii with that of Mn^2+^. PstP~c~ was partially inactive in the presence of 0.2 mM ZnCl~2~ and displayed lower activity on increasing the Zn^2+^-ion concentration upto 2 mM, as observed by *p*NPP assays. *In vitro* kinase assays with PknA~c~ and PknB~c~ in presence of Zn^2+^ resulted in phosphorylation of PstP~c~. Also, there was increase in phosphorylation of PstP~c~ during metabolic labeling by PknA and PknB in the presence of Zn^2+^ added in the *E. coli* culture. These results indicate that in mycobacterial cell, if cytosolic Zn^2+^ concentration increases, it may inhibit PstP perhaps leading to its phosphorylation. In an elaborative elemental analysis, Wagner *et* al. have reported that during infection, intravacuolar Zn^2+^-ion concentration increases from 0.037 mM to 0.46 mM in macrophages infected with *M. tuberculosis* [@pone.0017871-Wagner1]. Although there is no report of concomitant increase in mycobacterial Zn^2+^-ion concentration, it can only be speculated that if these changes in vacuolar ionic concentrations alter the mycobacterial ionic profile, a condition may develop where the enzymes that respond to Zn^2+^ (like PstP) can be activated or deactivated. End-product inhibition of enzymes is a well established phenomenon to prevent the accumulation of a particular metabolite. In case of reversible reactions, end-product accumulation can change the direction of the reaction. Similarly, Pi is known to inhibit a number of phosphatases [@pone.0017871-Fjeld1], [@pone.0017871-Halbedel1], [@pone.0017871-Das1] and in present study, PstP~c~ mediated *p*NPP hydrolysis is inhibited by Pi. To confirm that this effect is not limited to *p*NPP, *in vitro* kinase assays and metabolic labeling in *E. coli* showed PstP~c~ to be phosphorylated by PknA and PknB in presence of Pi because of its inhibition. Pi content is indicative of nutrient availability and energy status of the cell. In general, high Pi is associated with energy-starved conditions, when all the ATP is depleted and metabolite homeostasis is in unbalanced state. Such conditions usually arise during late-log and stationary phases in culture conditions. Metabolic labeling by \[^32^P\]orthophosphoric acid in the presence of co-expressed STPK (PknA or PknB) in *E. coli* lead to the specific phosphorylation of PstP~c~ and PstP~c~ ^D38G^. Co-expression in pETDuet-1 has previously been utilized extensively to assess the interaction of mycobacterial STPKs with their cognate substrates in the surrogate host *E. coli* [@pone.0017871-Khan1], [@pone.0017871-Kumar1]. Such dual-expression systems are increasingly becoming useful for analysis of protein-protein interactions specifically for challenging systems like mycobacteria [@pone.0017871-Molle1]. Activity assays of the pETDuet-1 purified PstP~c~ and PstP~c~ ^D38G^ revealed the higher activity of PknA-phosphorylated phosphatase as compared to the PknB-phosphorylated protein. Prominent variations in the activity of phosphorylated and unphosphorylated PstP~c~ ^D38G^ were observed with phosphorylated protein being proficient to hydrolyze *p*NPP to a greater extent (∼15-fold higher) in contrast to the unphosphorylated protein. The difference in the activities of phosphorylated and unphosphorylated PstP~c~ was not as prominent as that of PstP~c~ ^D38G^ (∼2--3 fold higher). These differences may be attributed to the fact that PstP~c~ may get auto-dephosphorylated to a greater extent than PstP~c~ ^D38G^ during expression and purification procedures. Higher activity of phosphorylated phosphatase is suggestive of reverse regulation of signaling cascade emanating from the kinases. In the constitutively active state, STPKs perform their regular functions and phosphorylate the target substrates following the stimulus. This may ultimately lead to the phosphorylation of PstP. The resulting increase in the activity of phosphatase may itself act as a control mechanism for kinases, eventually impeding the continued effect of that particular stimulus. The overall process has to be dynamic due to auto-dephosphorylation of PstP, eventually ceasing the effect of signaling cascade. In the conditions of high Zn^2+^ or high Pi content of the cell, PstP may not be active and will allow the kinase to work at its maximal activity. The proposed phosphorylation of PstP in such conditions may act as a mechanism to overcome the inhibition of PstP, hence balancing the cellular signaling pathways. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### ***p*** **NPP-assay.** To confirm the authenticity of pNPP assay, increasing amounts of alkaline phosphatase (0-100 ng) was used a positive control and PknB~c~ (0--5 µg) was used as a negative control. The assay was performed for 30 mins at 37°C and the activity is calculated as µmoles of *p*NPP hydrolyzed per min at a given amount of enzyme used. As clearly evident, alkaline phosphatase showed very high activity while no such activity was detected in PknB~c~. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Effect of mutations on the activity of PstP~c~.** To show that the loss in activity of PstP~c~ was specifically due to mutations of Arg^20^, Asp^38^ and Asp^229^, PstP~c~ was mutagenized on irrelevant residues Thr^5^ and Thr^141^ to Ala and Glu, respectively and *p*NPP hydrolysis was performed for 30 mins at 37°C. Activity of PstP~c~ was taken as 100% and relative activity was calculated. As evident from the bar graph, there were no significant changes in the activity of the mutants PstP~c~ ^T5A^ and PstP~c~ ^T141E^ as compared to PstP~c~. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Time-dependent** ***p*** **NPP-assay.** *p*NPP-hydrolysis was performed in a time-dependent manner for 30 mins using PstP~c~, PstP~c~ ^R20G^, PstP~c~ ^D38G^ and PstP~c~ ^D229G^ variants (2 µg each) at 37°C. Alkaline phosphatase (2 ng) was used a positive control and PknB~c~ (5 µg) was used as a negative control. Activity was calculated as nmoles of *p*NPP hydrolyzed per µg of enzyme used at a given time and depicted in logarithmic scale. Nevertheless, the results are essentially similar as that of time-dependent dephosphorylation of PknB~c~ ([Figure 2A](#pone-0017871-g002){ref-type="fig"}). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### ***In vitro*** **dephosphorylation activity of pETDuet-1 purified PstP~c~^D38G^.** Autophosphorylated PknA~c~ was incubated with unphosphorylated and phosphorylated PstP~c~ ^D38G^. As shown in the autoradiogram, the PknA-phosphorylated PstP~c~ ^D38G^ dephosphorylated the kinase to a greater extent in comparison to the unphosphorylated PstP~c~ ^D38G^. The image was also analyzed by ImageGauge software and corresponding values are depicted by bar-graph ([Figure 4E](#pone-0017871-g004){ref-type="fig"}). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Phosphorylation of PstP~c~.** Autoradiogram showing the phosphorylation of PstP~c~ (1 µg) by His~6~-tagged STPKs PknA~c~ (upper panel) and PknB~c~ (lower panel) in presence of 0.2 mM Zn^2+^ and 0.5 mM Pi. Due to overlapping molecular weights of PknA~c~ and PknB~c~ with PstP~c~, the bands were not separated properly. Still, the phosphotransfer on PstP~c~ was evident in presence of Zn^2+^ and Pi by both the kinases. The reaction was also performed with GST-tagged STPKs to clearly depict the reaction ([Figure 5D](#pone-0017871-g005){ref-type="fig"}). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: File S1 ::: {.caption} ###### **Detailed protocol of sample processing for identification of phosphorylation sites.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Zachary Weldon (Proteomics Core Facility, Children\'s hospital, Boston) for providing assistance in using Mass-spectrometric analysis for identification of phosphorylation sites. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Financial support to the work was provided by Council of Scientific and Industrial Research (NWP-0038). AS is a Senior Research Fellow of University Grants Commission, India. GA and MG are Senior Research Fellows of Council of Scientific and Industrial Research, India. SU is a Junior Research Fellow of Council of Scientific and Industrial Research, India. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: AS GA. Performed the experiments: AS GA MG SU. Analyzed the data: AS GA VKN YS. Contributed reagents/materials/analysis tools: VKN YS. Wrote the paper: AS GA MG.

PubMed Central

2024-06-05T04:04:19.228199

2011-3-9

PMC3052368

Introduction {#s1} ============ New neurons born in the adult mammal Dentate Gyrus (DG) functionally integrate into the hippocampal network [@pone.0017689-vanPraag1], [@pone.0017689-RamirezAmaya1], [@pone.0017689-Toni1]. The integration process resembles the one described during early development [@pone.0017689-Rihn1], [@pone.0017689-Jones1] but is slower [@pone.0017689-Rao1], [@pone.0017689-OverstreetWadiche1]. During the first 4 weeks ∼80% of the new neurons die, and the remaining 20% survive for at least 11 months [@pone.0017689-Kempermann1]. Thus, the first 4 weeks are critical for the survival of these new neurons and also for their maturation [@pone.0017689-Abrous1], [@pone.0017689-Piatti1]. For example, in the first week after birth new neurons are partially differentiated [@pone.0017689-Dayer1] and express doublecortin (DCX), which is important for neuronal migration [@pone.0017689-Brown1], and by the third week ∼90% of the new neurons express NeuN, a marker of mature neurons [@pone.0017689-Brown1], [@pone.0017689-Kempermann1]. New neurons [do not]{.underline} show electrophysiological features of maturity until the third week [@pone.0017689-Ambrogini1], [@pone.0017689-Esposito1]: their GABAergic response, which is initially depolarizing, becomes hyperpolarizing around this time [@pone.0017689-Ge1], [@pone.0017689-Karten1], and the glutamatergic input matures during weeks 3 and 4 [@pone.0017689-Ambrogini1], [@pone.0017689-Esposito1], [@pone.0017689-Ge1]. Anatomically, the axon, dendrites, and their spines reach maturity around weeks 3 to 4 [@pone.0017689-Hastings1], [@pone.0017689-Esposito1], [@pone.0017689-Zhao1]. By detecting the expression of the immediate early gene (IEG) Arc induced by spatial exploration, we previously examined the functional integration of 5-month-old, adult-born granular neurons into the behaviorally relevant hippocampal network [@pone.0017689-RamirezAmaya1]. We found that more adult-born neurons expressed Arc in animals allowed to explore, than in cage control animals, indicating that adult-born neurons have a specific Arc-expression response to spatial exploration [@pone.0017689-RamirezAmaya1]. Nevertheless, it is not clear when this specific response to exploration appears. By detecting the expression of the IEGs cFos and Arc evoked during a water-maze task in previously trained animals, Kee and colleagues [@pone.0017689-Kee1] observed that 6-week-old new neurons, but not younger ones, are recruited into circuits that can be re-activated at 10 weeks. They also showed that the expression of cFos after a single water-maze session occurs only in 6-week- but not in 1-week-old new neurons, similar to previous findings showing cFos expression after seizures only in 3-week-old new neurons [@pone.0017689-Jessberger1]; this earlier work was the first to show responsiveness of adult-born granular neurons to spatial behavior [@pone.0017689-Jessberger1]. Additionally, Tashiro and colleagues found that when they are 2 weeks old, new neurons are preferentially recruited into circuits that process information from experience in an enriched environment, which enhances new neuron survival [@pone.0017689-Tashiro1]. However, the question remains as to when, during this critical period for survival and maturation, do adult-born neurons integrate into the behaviorally relevant hippocampal network, developing a specific response to spatial exploration? The expression of the IEG Arc and its protein product, stimulated by behaviorally induced neural activity [@pone.0017689-RamrezAmaya1] and important for synaptic plasticity [@pone.0017689-Guzowski1], [@pone.0017689-Plath1], is detected as early as 24 hours after the birth of adult-born granular neurons [@pone.0017689-Kuipers1]. Therefore, this question can be answered by comparing Arc protein expression in new granular neurons 30 minutes after spatial exploration or cage control conditions during the critical time for survival and maturation, without affecting these processes by behavioral stimulation. Animals were administered BrdU on day 0 and sacrificed on day 1, 7, 15, 30, or 45, in each case thirty minutes after a 5-minute exploration session or cage control conditions ([Fig. 1](#pone-0017689-g001){ref-type="fig"}). Their brains were processed for triple immunohistochemistry for NeuN, Arc, and BrdU (also for DCX&BrdU and Arc&DCX), and semi-confocal and confocal microscopy images were taken for later analysis (see [Fig. 2A&B](#pone-0017689-g002){ref-type="fig"}, [Fig 3](#pone-0017689-g003){ref-type="fig"}, [Fig. 4A](#pone-0017689-g004){ref-type="fig"}, and [Fig. 5D--H](#pone-0017689-g005){ref-type="fig"} ). ::: {#pone-0017689-g001 .fig} 10.1371/journal.pone.0017689.g001 Figure 1 ::: {.caption} ###### Schematic representation of the experimental procedure. BrdU was administered on day 0, in 4 separate ip injections (50 mg/Kg each) every 4 h from 9:00 A.M to 9:00 P.M. Animals were sacrificed 1, 7, 15, 30, or 45 days post BrdU injections (PBI), either from their home cages (cage control, CC) or 30 min after a 5-min spatial exploration (SE). :::  ::: ::: {#pone-0017689-g002 .fig} 10.1371/journal.pone.0017689.g002 Figure 2 ::: {.caption} ###### Survival and migration of adult-born granular cells. Confocal images, taken with a 25× objective, of the DG stained with DCX in red, BrdU in green, and counterstained with DAPI in blue. A) shows BrdU-positive cells co-localizing (yellow) with DCX from an animal sacrificed 7 days PBI; B) shows BrdU-positive cells from an animal sacrificed 45 days PBI that do not co-localize with DCX. C) Average proportion of BrdU-positive cells from the total granular cell population found in each (1, 7, 15, 30, and 45 days) PBI group. \*P\<0.01 as compared to day 1 PBI group, † P\<0.01 between indicated groups. Note the significant decrease in the proportion of BrdU-positive cells found with time after new neurons were born. D) Absolute BrdU-positive cell counts found in each PBI group (1, 7, 15, 30 and 45 days); note that the results are very similar to the proportions shown in C. E) Percentage of BrdU-positive cells from the total BrdU-positive cell population found in each DG layer: the subgranular zone (SGZ), DG layer 1 (L1), DG layer 2 (L2), and DG layer 3 (L3), for the different PBI times. The different proportions of BrdU-positive cells found in the various layers with increasing time suggest that new cells migrate through the DG layers. :::  ::: ::: {#pone-0017689-g003 .fig} 10.1371/journal.pone.0017689.g003 Figure 3 ::: {.caption} ###### Arc expression after spatial exploration. A\) Percentage of the whole granular cell population that expresses Arc after spatial exploration. Note that Arc expression is observed in a higher proportion of cells in the exploration group (solid bar) than in cage controls (pattern bar). \*P\<0.001. B) Percentage of BrdU-positive cells that expressed Arc in each PBI group, i.e., sacrificed either directly from its home cage (CC, pattern bar) or after spatial exploration (SE, solid bar). Arc expression in BrdU-positive cells occurs with no behavioral stimulation from day 1 until day 15 PBI, but on days 30 and 45 PBI, Arc expression was observed in significantly more BrdU-positive neurons from exploration animals than from cage controls. \*P\<0.01. C) Percentage of BrdU-positive cells expressing Arc in each DG layer, either after cage control conditions (patterned bar) or spatial exploration (solid bar). The yellow horizontal bars represent the proportion of the total BrdU-positive cells found in each layer on days 30 and 45. Note that no Arc expression was found in BrdU-positive neurons in the SGZ. In L1, the IEG Arc was expressed in a larger proportion of new neurons from exploration animals than from the cage controls, \*P\<0.01, indicating that BrdU-positive neurons located in L1 are more likely to respond to exploration. The opposite was observed in L2, i.e. the proportion of BrdU-positive cells in L2 that expressed Arc was much lower in SE than in CC animals. D to H) 40× confocal images taken from the middle plane of a confocal microscope image stack were used to verify the co-localization of Arc and BrdU. D) NeuN in blue, E) Arc protein in red, F) BrdU in green, G) NeuN, Arc, and BrdU merge; note that BrdU-positive neurons expressing Arc appear yellow. H) Flat image obtained from a 3D boxels reconstruction. The rotation of the 3D projection was used to confirm BrdU and Arc co-localization. The scale bar in G is 100 μm. :::  ::: ::: {#pone-0017689-g004 .fig} 10.1371/journal.pone.0017689.g004 Figure 4 ::: {.caption} ###### Arc expression in DCX-positive cells. The images show double immunohistochemistry for Arc and DCX, and were taken with the Apotome microscope system (Zeiss) with a 40×/1.3 NA objective using Z section resolution. A) DAPI is shown in blue, B) Arc protein is shown in green, C) DCX is shown in red, and D) the 3 channels were merged. Scale bar = 100 μm. :::  ::: ::: {#pone-0017689-g005 .fig} 10.1371/journal.pone.0017689.g005 Figure 5 ::: {.caption} ###### Contribution of adult-born neurons to spatial information processing: a hypothesis suggested by integrating our current data with that obtained previously. A\) Linear regression plot calculated using the percentage of BrdU-positive cells that specifically respond to spatial exploration by expressing Arc from animals sacrificed on days 30 and 45 PBI (red dots), and those sacrificed 150 days PBI (yellow dot) [@pone.0017689-RamirezAmaya1]. Using this model we calculated the maximum length of time throughout the life span of the animal during which adult-born neurons respond to spatial exploration (time in days means age of animals in X axis for all graphs). The result suggests that after 301 (271+30) days, the neurons born in the adult brain may no longer respond to spatial exploration. B) The data obtained from Kuhn [@pone.0017689-Kuhn1] was scaled to fit our BrdU-positive cell proportions, and a power model was used to calculate the number of neurons that survived at the different times through the animal\'s life. C) Cumulative number of new neurons that survived after 30 days including all the neurons born since 70 days post natal, when the animal is already mature. D) Total number of neurons born after DG development that contributes to the DG spatial exploration response over the course of the animal\'s life span. The red line shows the time (day 217) when the maximum number of newly incorporated neurons participates in the DG network response to spatial information processing. At this time, 2057 new neurons respond to spatial exploration, which represents 60% of the total DG granular cell population that responds to spatial exploration. The super-imposed graph (shown in light colors) represents the same calculation including the number of neurons born since postnatal day 1 that survive for 30 days. Notice in particular that the contribution to spatial information processing of neurons born after the DG development is complete increases through the early age of the animals, and after reaching its maximum, it rapidly decreases to a plateau reached in late adulthood. :::  ::: We found that the expression of Arc occurs without stimulation in 1- to 15-day-old new granular neurons. A specific response to exploration was observed in ∼5% of the 30- to 45-day-old new-neurons. This temporal pattern of Arc expression in response to spatial exploration reflects the course of selective survival and integration in a network that processes spatial information. Ninety percent of the BrdU-positive cells expressing Arc in response to spatial exploration were located in DG layer 1 (L1), and no Arc expression was detected in new cells located in the subgranular zone (SGZ). Based on calculations using the current data and that obtained previously [@pone.0017689-RamirezAmaya1], we hypothesize that new neurons may no longer respond to spatial exploration after they are 301 days old, and that in a 7-month-old animal the majority (60%) of the neurons that respond to exploration were born during adulthood. Results {#s2} ======= Survival and migration of new neurons {#s2a} ------------------------------------- The proportion of BrdU-positive cells in the DG upper blade varied significantly across the different maturation times (*F* ~4,30~ = 29.771, *p*\<0.001) ([Fig. 2C](#pone-0017689-g002){ref-type="fig"}). One day post BrdU injection (PBI), 0.26% of the entire population of granular DG cells included in the study had incorporated BrdU, and seven days PBI, the percentage of BrdU-positive cells (0.35%) was significantly higher (the Bonferroni test yields a *p*\<0.01). At 15 days PBI, the proportion of new granular cells was 0.23%, which was significantly lower than on day 7 (*p*\<0.01) but not statistically different from day 1. Thirty days PBI, the percentage of BrdU-positive cells was only 0.17% and was significantly lower than that found on days 1, 7, and 15 PBI (*p*\'s\<0.01). Finally, 45 days PBI, the percentage of BrdU-positive cells was only 0.15%, which was significantly lower than on days 1, 7, and 15 (*p*\'s\<0.01) but was not different from day 30. The ANOVA of the absolute BrdU-positive cell counts revealed a similar pattern through time (*F* ~4,30~ = 89.772, *p*\<0.001, Bonferroni test yields p\'s\<0.01). These counts represent the average number of BrdU-positive cells found in each PBI group ([Fig. 2D](#pone-0017689-g002){ref-type="fig"}), without adjusting for the total population of granular cells included in the analysis. In the animals sacrificed on day 1 PBI ([Fig. 2E](#pone-0017689-g002){ref-type="fig"}), 60% of the BrdU-positive cells were located in the sub granular zone (SGZ), 32% in DG layer 1 (L1), 7% in layer 2 (L2), and only 1% were found in layer 3 (L3). On day 7 PBI, 32% of the BrdU-positive cells were located in the SGZ, 54% in L1, 11% in L2, and 3% in L3. On day 15 PBI, 16% of the BrdU-positive cells were located in the SGZ, 58% in L1, 19% in L2, and 6% in L3. At 30 days PBI, 13% of the BrdU-positive cells were located in the SGZ, 54% in L1, 24% in L2, and 10% in L3. Finally, at 45 days PBI, 12% of the BrdU-positive cells were located in the SGZ, 51% in L1, 23% in L2, and 13% in L3. The proportion of cells found in each DG layer was compared between the different PBI groups using a one-way ANOVA ([Fig. 2E](#pone-0017689-g002){ref-type="fig"}). The proportion of cells in the SGZ differed among groups (*F* ~4,30~ = 176.086, *p*\<0.001), where the proportion of BrdU-positive cells found in animals sacrificed on day 1 PBI was significantly different (using the Bonferroni post hoc correction) from all other groups (*p*\<0.01). For L1, the proportion of BrdU-positive neurons differed significantly among groups (*F* ~4,30~ = 47.687, *p*\<0.001), and the post hoc analysis showed differences between animals sacrificed at day 1 PBI and all other groups (*p*\<0.01). Differences were also found between animals sacrificed at day 15 and day 45 PBI (*p*\<0.01). In L2, significant differences were found in the proportion of BrdU-positive cells among groups (*F* ~4,30~ = 40.070, *p*\<0.001), and the post hoc analysis revealed differences between animals sacrificed at day 1 PBI and the animals sacrificed 15, 30, and 45 days PBI (*p*\<0.01); differences were also found between animals sacrificed at day 7 PBI and those sacrificed at days 15, 30, and 45 PBI (*p*\<0.01); in L2, significant differences were found between animals sacrificed on day 15 and those sacrificed on day 30 PBI (*p*\<0.01). In L3, significant differences were found among groups (*F* ~4,30~ = 35.403, *p*\<0.001), and the post hoc analysis revealed that most of the PBI groups differed from each other (*p*\<0.01), except those sacrificed on day 1 vs. day 7 PBI and animals sacrificed on day 7 vs. 15 PBI. The differences in the proportion of BrdU-positive cells located in the different regions between the animals sacrificed at different PBI times indicate that the newly incorporated cells migrate through the different DG layers. Sparse Arc expression in DG granular cells after spatial exploration {#s2b} -------------------------------------------------------------------- After spatial exploration (SE), the expression of the immediate early gene Arc was observed in ∼1.5% of the granular cell population, while in the cage control (CC) animals only ∼0.3% of the granular cells were classified as Arc positive. The percentage of cells expressing Arc was significantly different (*F* ~1,33~ = 12.042, *p*\<0.001) between CC and SE animals, demonstrating that a small population of granular neurons in the DG expressed Arc in response to a novel spatial exploration in an open box ([Fig. 3A](#pone-0017689-g003){ref-type="fig"}), as previously reported [@pone.0017689-RamrezAmaya1], [@pone.0017689-Rosi1], [@pone.0017689-Chawla1], and suggesting a sparse DG code for spatial information. Specific Arc expression response to spatial exploration develops in new neurons over a 30-day period {#s2c} ---------------------------------------------------------------------------------------------------- By using double staining, we observed that in animals sacrificed at early PBI times (1--15), BrdU is found mostly in DCX-positive cells and in late PBI time (30--45) primarily in NeuN-positive neurons ([Fig. 3D--H](#pone-0017689-g003){ref-type="fig"}). We also observed that Arc can be expressed in DCX positive cells. At early times ([Fig. 4](#pone-0017689-g004){ref-type="fig"}.) For these reasons we included all BrdU-positive cells found in the different DG layers in further analysis. Arc expression in the BrdU-positive population differed across the different treatment groups ([Fig. 3B](#pone-0017689-g003){ref-type="fig"}). A 2-way ANOVA showed significant differences between days PBI (*F* ~4,25~ = 3.79, *p*\<0.02); between exploration and cage control conditions (*F* ~4,25~ = 13.603, *p*\<0.01), and among days and the behavioral treatment (*F* ~4,25~ = 5.437, *p*\<0.0027). One day PBI, Arc expression was observed in 1.3% of the BrdU-positive cells found in cage control animals (CC) and in 1.01% of the BrdU-positive cells found in animals exposed to spatial exploration (SE); these values were not significantly different. Arc was expressed in 2.1% of the 7-day-old new neurons from CC animals and in a similar percentage (1.56%) of the corresponding new neurons from SE animals. Moreover, the proportion of 7-day-old, BrdU-positive cells expressing Arc (BrdU+/Arc+ cells) did not differ from the proportion of new cells expressing Arc on day 1 PBI from either the SE or CC group. Fifteen days PBI, Arc was expressed in a similar proportion of new cells from CC animals (∼2.3%) and from the SE group (2.9%). Although Arc expression in BrdU-positive cells tended to increase from day 1 to day 15 PBI, this increase was not statistically significant. Importantly, these results suggest that Arc expression at this early time after these new neurons were born may not be driven by spatial behavior stimulation. In contrast, when new neurons were 30 days old, the proportion of BrdU+/Arc+ cells was significantly greater (*p*\<0.01) in SE (∼5%) than in CC animals (∼1.5%). The proportion of BrdU+/Arc+ cells at 45 days PBI also differed significantly (p\<0.001) between SE (4.8%) and CC animals (0.7%). A repeated measures ANOVA was done on the percentage of cells expressing Arc at the different PBI times and revealed significant changes with time (*F* ~1,5~ = 26.839, *p*\<0.01), differences between SE and CC groups (*F* ~4,4~ = 3.385, *p*\<0.05), and also a significant interaction (*F* ~4,6~ = 5.887, *p*\<0.01), indicating that a specific Arc expression response to spatial exploration developed between 30 and 45 days after new neurons were born. It is important to note that the proportion of BrdU+/Arc+ cells in response to exploration at 45 days PBI was significantly greater than the percentage of total granular cells expressing Arc in response to exploration (T ~6~ = 7.995 *p*\<0.001). Exploration-induced Arc expression in BrdU-positive cells is observed mainly in DG L1 {#s2d} ------------------------------------------------------------------------------------- The position of the BrdU-positive cells expressing Arc within the different granular layers of the dentate gyrus [@pone.0017689-Kempermann1] was identified, and for each layer we calculated the proportion of BrdU-positive cells expressing Arc relative to the total population of BrdU-positive cells that express Arc. These proportions were compared between CC and SE groups ([Fig. 3C](#pone-0017689-g003){ref-type="fig"}) at all 5 time points when the animals were sacrificed. Significant differences were found between CC and SE animals in L1 (*F* ~1,33~ = 8.754, *p*\<0.01), where 89.3% of the BrdU-positive cells expressing Arc from the SE group but only 63.78% from the CC group were located. In L2, the proportion of Arc-expressing cells also differed significantly between CC and SE animals (*F* ~1,33~ = 14.509, *p*\<0.001) 29.56% of the BrdU-positive cells expressing Arc from the CC group were located in L2 as compared to 1.67% from the SE animals. In L3, no significant differences were found between groups. The MANOVA analysis revealed significant differences between the various layers in the proportion of BrdU-positive cells expressing Arc (Wilks lambda *F* ~2,31~ = 7.183, *p*\<0.01), and it is clear that the L1 region had the highest Arc expression in BrdU-positive cells. Importantly, BrdU-positive cells expressing Arc were not found in the SGZ.Using a Student\'s t-test we compared the proportion of BrdU+/Arc+ cells with the proportion of BrdU-positive cells among the different DG layers. Since the specific Arc expression response to exploration develops after 30 days, in this analysis we included only the animals sacrificed 30 and 45 days after the BrdU injection. The proportion of BrdU+/Arc+ cells were significantly higher than the proportion of BrdU-positive cells in L1 only for SE animals and not for CC animals (p\<0.01). In contrast, in layer 2 of SE but not CC animals, the proportion of BrdU+/Arc+ cells was significantly lower (p\<0.01) than the proportion of all BrdU-positive cells. No differences were found in layer 3. A hypothesis suggesting that adult-born neurons modify their contribution to spatial information processing throughout their life and the animal\'s life span {#s2e} ------------------------------------------------------------------------------------------------------------------------------------------------------------- Here, we found that the proportion of new granular cells that responded to exploration 30 days after these neurons were born was 5%, and at day 45 PBI it was 4.8%. Previously, we reported that ∼2.8% of the 5-month-old, newly incorporated neurons responded to exploration [@pone.0017689-RamirezAmaya1]. This suggests that the likelihood of a new neuron to respond to behavioral exploration decreases with time [@pone.0017689-Bischofberger1]. In this study, we included ∼80,000 upper blade DG granular cells per animal, obtained from 450 µm in the antero-posterior axis. The whole sample region measured 1300 µm, and we calculated that it contained 231,000 DG granular neurons. From this population of neurons, 1.5% (3,465) responded to spatial exploration, a result similar to what has been reported previously [@pone.0017689-RamrezAmaya1], [@pone.0017689-Rosi1], [@pone.0017689-Chawla1]. We hypothesize that the probability of the adult-born granular neurons to respond to spatial exploration changes linearly with time:where P~0~ indicates the percentage of neurons that respond when they are 30 days old. We used a linear regression to estimate the parameters of our model (P~0~ = .054, r = −0.0002, *R^2^* = 0.9989, *p* = 0.0208), and the result is shown in [Figure 5A](#pone-0017689-g005){ref-type="fig"}, were the red circles represent our current data, and the yellow circle represents our earlier result [@pone.0017689-RamirezAmaya1]. The line represents the model, with which we calculate the probability that granular cells will respond to spatial exploration after they are more than 150 days old. Obtaining the intercept of the line to reach a probability of 0 by the following formula: This suggests that 301 days after the new neurons were born, they no longer respond to spatial exploration. We also used the individual data from each animal to calculate this linear regression and obtained a significant regression (*P* ~0~ = 0.0505, r = −0.0002, *R^2^* = 0.5101, *p* = 0.0135), validating the conclusion obtained with the average proportions. We acknowledge that a linear model with only 3 data points presents clear limitations, and further research is needed to add more data points between 45 and 360 days. This will allow us to determine if a linear model is adequate and will test the prediction that cells no longer respond to exploration after they are ∼1 year old. The number of new granular neurons found 45 days after birth in the sampled regions was 345. Adult-born neurons that survived for 4 weeks remained stable for at least 11 months [@pone.0017689-Kempermann1]; therefore, we can assume that the number of newly incorporated neurons detected at days 30--45 PBI represents the number of cells born on day 1 that will survive for the rest of the animal\'s life. However, the rate of cell proliferation varies across the animal\'s life span [@pone.0017689-Kuhn1], and this modifies the number of stable new neurons throughout the life time.By using these data [@pone.0017689-Kuhn1] we propose a power model: Since the DG is fully assembled after the first post-natal weeks [@pone.0017689-Schlessinger1], sexual maturity is reach after 6 weeks [@pone.0017689-Adams1], and the neurons that survive for 4 weeks remain, we perform further calculations using the number of granular cells that were born and survived since post-natal day 70. N~0~ is then the number of new granular neurons that survive for 30 days. In order to estimate the parameters of our model, we calculate a linear regression. We choose 's' such that the equation predicts the number of BrdU-positive cells that we detect after 45 days in the current work ([Figure 2C&D](#pone-0017689-g002){ref-type="fig"}). (*N(150) = 345*). The result, *N* ~0~ * = 2218.2*, is shown in [Figure 5B](#pone-0017689-g005){ref-type="fig"}, where the red dots are the data and the solid line is the model. From these data we calculate the cumulative number of neurons that were born after the animal was sexually mature and that survived throughout the animal\'s life span, assuming that there is no mortality after 30 days of maturation [@pone.0017689-Kempermann1]. The cumulative number of adult-born granular neurons over the course of the animal\'s life span is shown in [Figure 5C](#pone-0017689-g005){ref-type="fig"}. In order to estimate the number of adult-born granular neurons that contribute to the DG response to spatial exploration throughout the animal\'s life span, we considered the number of neurons that are added through time (N(t)), and the probability that they will respond at different time intervals (P(t)) after they were added. In order to combine these measurements we used a convolution of the variables N and P. This is the overlap level of N and the function of P transferred and inverted, which implies the number of new neurons at time t (throughout the animal\'s life span) under the proportion *P* (*t*−τ) that contribute to spatial information precesing (see [Fig. 6A and 6B](#pone-0017689-g006){ref-type="fig"} for a graphic explanation). The resulting calculation showed that the time t, when the maximum number of neurons, that were born after the DG network was fully formed, contribute to spatial information processing, is at day 217, where 2057 of the adult-born neurons in the sample region are predicted to respond to spatial exploration ([Fig. 5D](#pone-0017689-g005){ref-type="fig"}). This represents 60% of the total granular cell population that responds to this behavior. If we include in our calculation all the neurons born after the animal\'s birth that survive for 30 days, the result suggest that at day 129 of the animal\'s life, 99% of the neurons that respond to exploration are neurons that were born post-natally ([Figure 5D](#pone-0017689-g005){ref-type="fig"}, shaded line). ::: {#pone-0017689-g006 .fig} 10.1371/journal.pone.0017689.g006 Figure 6 ::: {.caption} ###### Illustration of the calculation procedure. A\) As an example of our calculation that lead to the results found in [Figure 6D](#pone-0017689-g006){ref-type="fig"}, we can estimate the number of cells that contribute to the spatial exploration response at day 400 in the life of (red arrow), from cells born on day 250. We found ∼11,000 new neurons that were born on day 250 and survive for at least 30 days; when we multiplied this by the proportion of neurons that responds to exploration at day 400 (2%), we calculated that ∼220 cells respond to exploration (purple arrow in B); B) When all the new neurons are considered (area under the line in yellow) we have a total number of neurons that respond at day 400, from the cumulative number of new neurons at that time. C) To determine the background due to Arc expression during early development of new neurons, we calculated the probability of neurons to express Arc from 1 to 30 days, using 2 linear models. D) A similar convolution as the one used for our previous calculation was done to estimate the number of adult-born neurons found from 1 to 30 days PBI ([Figure 5 C](#pone-0017689-g005){ref-type="fig"}). Note that the maximum proportion of neurons that express Arc with no stimulation is ∼0.16%; for comparison, cage control animals present 0.31% of cells expressing Arc with no previous stimulation ([Figure 5 A](#pone-0017689-g005){ref-type="fig"}), and this represents ∼10% of the granular cells expressing Arc after spatial exploration. :::  ::: We emphasize that this model is presented only as a new hypothesis suggesting that the neurons born in the adult mammal hippocampus change their contribution throughout both, their own and the animal\'s life span. These results, though inconclusive, should stimulate further studies that evaluate the relevance and contribution of newly incorporated granular neurons through time. Discussion {#s3} ========== During the first 6 weeks, the average proportion of BrdU-positive cells found in the DG from all animals fell from ∼0.35% on day 7 to ∼0.15% on day 45, suggesting that ∼60% of the neurons detected on day 7 died over the course of the next 5 weeks. It has been reported that in mouse, new neurons that survive more than 4 weeks represent 20% to 45% of the proliferating cells [@pone.0017689-RamirezAmaya1], [@pone.0017689-Snyder1] and that in rats, they represent 42 to 60% of the newly born neurons [@pone.0017689-Snyder1], [@pone.0017689-Dayer1]. In the present work, the number of BrdU-positive cells found on day 1 PBI was lower than the number of cells found on day 7 PBI. This may be explained by the fact that on day 1, only 12 hrs had passed since the last BrdU administration, and more new cells may have incorporated BrdU in the following hours [@pone.0017689-Cameron1]. This is consistent with an earlier report that used a single exposure to \[^3^H\]-methyl-thymidine and found more labeled cells on day 7 than on day 1 [@pone.0017689-Cameron2]. Here, we found that the proportion of granular cells born on one day that survive for 45 days is 0.15%, and that these cells may survive for 3 to 11 months [@pone.0017689-Kempermann1], [@pone.0017689-Snyder1]; the result was similar using the absolute BrdU-positive cells counts. We calculated that the whole DG contains ∼1,200,000 granular cells, which is consistent with previous reports [@pone.0017689-West1], [@pone.0017689-Rapp1] and suggests that the number of cells born in the DG during one day that survive is ∼1,800. A similar number can be calculated by using the estimated number of cells born each day in the adult rat DG [@pone.0017689-Cameron1] and subtracting the dying neurons [@pone.0017689-Kempermann1], [@pone.0017689-Snyder1]. The location of the BrdU-positive cells among the different DG layers throughout time indicated that new cells migrate from the SGZ through the rest of the layers, as previously reported [@pone.0017689-Kempermann1], [@pone.0017689-Esposito1]. The highest migration from the SGZ to DG L1 and L2 was reported to occur between day 7 and 14 [@pone.0017689-Esposito1]. We observed that from day 1 to 15 PBI, when expression of DCX is highest [@pone.0017689-Brown1], the location of new cells changed significantly, primarily from the SGZ to DG L1 and L2, indicating that this new neurons migrate through the DG layers. Our results also show that a small proportion of BrdU-positive cells (∼10%) stay in the SGZ. Most (51%) of the BrdU-positive cells that survive for 45 days stay in DG L1, consistent with previous observations that ∼60% of the newly incorporated neurons were located in this layer [@pone.0017689-Kempermann1]; at day 45 PBI the percentage of new cells that we found in L2 was 22.6%, and in L3 it was 13.5%, also consistent with the earlier report. The proportion of DG neurons expressing Arc after exploration was 1.5%, significantly higher than that observed in the cage control animals and in agreement with the notion of a sparse code for spatial information processing in the DG [@pone.0017689-RamrezAmaya1], [@pone.0017689-Rosi1], [@pone.0017689-Chawla1]. The likelihood to observe Arc expression in BrdU-positive cells increases throughout new neuron maturation, as recently reported for other IEGs such as zif268 and cFos [@pone.0017689-Snyder1]. The highest IEG (zif268 and cFos) expression, after kainite-induced seizures, was observed at 4 weeks and remained stable at 10 weeks. The expression of zif268 induced by water maze training was maximal at 3 weeks; however, it is not clear if such expression was specifically induced by the behavioral treatment, since no behavioral controls were shown [@pone.0017689-Snyder1]. In our case, from day 1 until day 15 PBI the proportion of BrdU-positive cells expressing Arc was similar in cage control and exploration animals. Likewise, in animals that received LTP-inducing stimulation, Arc expression was reported in both the stimulated and the non-stimulated hemispheres in 1-day-old new neurons [@pone.0017689-Kuipers1]. These data suggest that Arc expression in new neurons at an early stage of their development occurs independently of sensory or behavioral stimulation. We cannot suggest that this seemingly spontaneous or constitutive Arc expression [@pone.0017689-Kuipers1] is independent of neuronal activity, but it may occur without synaptic stimulation since immature granular neurons do not respond to synaptic input or are "silent" at this early stage [@pone.0017689-Esposito1]. It is possible then, that these new neurons may be activated in response to paracrine BDNF release or other signals [@pone.0017689-Babu1], which is an interesting idea since this time is critical for the maturation and synaptic integration of adult-born neurons. Arc expression at this early stage may be indicative of an ongoing biological process involved in the synaptic integration of adult-born neurons [@pone.0017689-Kuipers1]. This hypothesis is supported by recent evidence showing that Arc increases the density of immature spines and regulates spine morphology [@pone.0017689-Peebles1]. Studying the role of Arc expression in young, adult-born granular neurons may help to understand the cellular mechanisms underlying the synaptic integration of these neurons. On the other hand, Arc expression at these early stages may represent background noise in the system. For this reason, we calculated the number of new neurons that contribute to Arc expression in the general population. The result indicated that in cage control animals ∼50% of all the granular neurons expressing Arc may be new granular neurons born between 1 and 30 days before sacrifice (See [Fig. 6C & D](#pone-0017689-g006){ref-type="fig"}). However, this represents only ∼10% of the neurons expressing Arc in response to spatial exploration, suggesting that the background noise produced by this Arc expression in young new neurons contributes relatively little to the whole DG network response to exploration. At days 30 and 45 PBI, Arc expression in the BrdU-positive cells was significantly higher in animals allowed to explore than in cage controls, indicating that a specific response to exploration develops with maturation. Previously, it was reported that the unspecific expression of Arc in response to LTP stimulation lasted for about 28 days, but the number of BrdU-positive cells expressing Arc under these conditions increased on day 28 in the granular cell layer, and from day 14 onwards in the SGZ [@pone.0017689-Kuipers1]. In contrast, we found no Arc expression in the BrdU-positive neurons located in the SGZ ([Fig. 3C](#pone-0017689-g003){ref-type="fig"}), indicating that new granular cells found in this DG region are not responsive to behavioral exploration and may not be functionally integrated into the behaviorally relevant network. This agrees with our own analysis of Arc expression in the different DG layers ([Fig. 7A](#pone-0017689-g007){ref-type="fig"}), where we did not find Arc-expressing cells after spatial exploration in the SGZ. This may suggest that migration from the SGZ to the DG granular layer is mandatory for new neurons to be able to respond to behavioral exploration and express Arc. The discrepancies with Kuiper\'s work may also be explained by the different stimulation methods and the animals that were used. It is possible, even in the non-stimulated hemisphere and in the SGZ [@pone.0017689-Plath1], that a strong electrophysiological stimulus can activate the CA3c commissural projections [@pone.0017689-Scharfman1], thereby inducing Arc expression in new neurons due to their enhanced plasticity [@pone.0017689-Ge2]. Another possibility is that the functional integration of adult-born neurons into the hippocampal network may take longer in Sprague Dawley than in Wistar rats. ::: {#pone-0017689-g007 .fig} 10.1371/journal.pone.0017689.g007 Figure 7 ::: {.caption} ###### Dentate gyrus layer segmentation and Arc expression in the different layers. A\) Shows a representative DG layer image segmented into 4 equidistant layers. NeuN-positive granular cells are shown in blue, Arc-expressing cells in pink, and BrdU-positive cells in green. The grid was used as a guide for the segmentation procedure, and the cells that were mainly (\>50%) located in grids from a particular layer were considered to belong to that layer. The segmentation is represented by an overlay of translucent colors. In brown is the SGZ, in purple L1, in green L2, and in gray L3. B) Percentage of granular cells (from the general population) expressing Arc in the different DG layers after spatial exploration. The red bar is the proportion of cells expressing Arc in all layers. :::  ::: The enhanced plasticity of newly incorporated neurons (4 to 6 weeks of age) was confirmed in the present study with the observation that ∼5% of the new neurons found at 30--45 days PBI responded to exploration by expressing Arc, whereas only 1.5% of the total DG neurons responded. This difference is similar to a previous report [@pone.0017689-Kee1] and agrees with our earlier findings [@pone.0017689-RamirezAmaya1], where a significantly higher proportion of 5-month-old new neurons than of total DG neurons expressed Arc in response to spatial exploration. The timing for the development of this Arc expression in response to spatial exploration coincides with the dynamic cellular process that takes place in these cells [@pone.0017689-Abrous1], [@pone.0017689-Esposito1], [@pone.0017689-Piatti1] and reflects the course of selective survival and integration into a network that processes spatial information. Around the time new neurons become responsive [@pone.0017689-Ambrogini1], [@pone.0017689-Esposito1] and their glutamatergic and GABAergic input matures [@pone.0017689-Ambrogini1], [@pone.0017689-Esposito1], [@pone.0017689-Ge1], [@pone.0017689-Karten1], the expression of Arc rises and becomes specific. This suggests that the enhanced survival effect of behavioral stimulation soon after new neurons are born [@pone.0017689-Kempermann2], [@pone.0017689-Ambrogini2] depends on a non-specific activation of DG neurons, probably related to BDNF expression [@pone.0017689-Rossi1], [@pone.0017689-Lee1]. In animals exposed to spatial exploration, ∼90% of their BrdU-positive cells that express Arc were located in DG L1 compared to 60% in the cage control group. The distribution pattern of BrdU-positive neurons among the DG layers ([Figure 2D](#pone-0017689-g002){ref-type="fig"}) is similar to that observed for BrdU-positive neurons expressing Arc in cage control animals but notably different from the distribution of new neurons that respond to spatial exploration by expressing Arc. The observations that mature (28-day-old), new neurons found in DG L1 exhibit a well-defined dendritic structure, a conspicuous axonal projection extending towards the hilus, and spiny dendrites reaching the outer molecular layer [@pone.0017689-Esposito1] suggest that neurons located in L1 mature earlier, developing a functional response to the input, while neurons located in L2 and L3 develop these features later. This may explain why most of the adult-born neurons that are functionally integrated into the behaviorally relevant network are located in this inner layer, and suggests that the time required for migration may delay the functional integration of new neurons; thus, it is expected that neurons located in L2 and L3 will acquire the ability to respond specifically to spatial exploration at a time point later than 45 days, since Arc-expressing cells are found in all 3 granular layers with a bias towards those located in the inner DG (see [Fig. 4B](#pone-0017689-g004){ref-type="fig"}). When we estimated the proportion of BrdU-positive cells that respond to spatial exploration by expressing Arc over the course of the animal\'s life span, we found a highly significant linear regression; thus, our hypothesis is that neurons born in the adult DG after day 301 do not respond to exploration anymore; this does not suggest that new neurons lose their integration within the network but rather that they can no longer be recruited into circuits that process spatial information ([Fig. 5A](#pone-0017689-g005){ref-type="fig"}), which may be related to a network mechanism that ensures that the DG code for spatial information can change. It is important to acknowledge that a non-linear model may also fit the data, and that more data points are required to demonstrate that our hypothesis is correct; however, the hypothesis is supported by the highly significant linear regression analysis and by the fact that the number of new neural units under the curve throughout the animal\'s life ([Fig. 6D](#pone-0017689-g006){ref-type="fig"}) agrees with the total number of Arc-expressing cells after exploration. The hypothesis that old granular cells no longer process spatial information after a certain time agrees with the granular cell retirement hypothesis [@pone.0017689-Alme1], which suggests that granular neurons process spatial information for a limited period of time. This satisfies an important requirement of the associative memory model from Treves and Rolls [@pone.0017689-Treves1], which proposes that: "...during retrieval..., the mossy fiber input should be absent or strongly reduced..., in order not to blur the signal relayed by the perforant path". Our data suggest that, instead of shutting down or reducing the DG input to the CA3 during retrieval, a few granular neurons retire every day, gradually changing the DG code of a particular experience. This hypothesis is of great interest, but its validity must be tested by further research with more time points, especially at late maturation times for adult-born neurons. The model also suggests that in a 7-month-old animal, 60% (\[5-month-old and 99%\] if we include all neurons added postnatally) of the total granular neurons that responded to exploration should be neurons born after the DG was fully formed. This suggests that adult-born neurons may contribute more to spatial information processing than pre-natally born granular neurons. This idea was recently tested in mice that received the administration of the thymidine analog CidU either on embryonic day 18 or postnatal day 7 and another thymidine analog IdU on postnatal day 60 [@pone.0017689-Stone1]. The results showed no differences in the proportion of granular cells responding to spatial behavior by expressing cFos. However, if we compare the time of analysis (Postnatal day 60) with our model ([Fig. 5D](#pone-0017689-g005){ref-type="fig"}) and consider that rats and mice present slight differences in the speed of maturation and life span [@pone.0017689-Adams1], [@pone.0017689-Drickamer1], it is possible that differences may be observed at later time points, congruently, other groups had recently found a higher proportion of adult born granular neurons responding to behavior compare to prenatally born granular neurons (Nora Abrous, Personal Communication). In any case, the model suggests that adult neurogenesis plays a pivotal role in spatial information processing, and that this contribution changes with time, which may explain cognitive changes related to hippocampal function over the course of the animal\'s life. Methods {#s4} ======= Subjects {#s4a} -------- Thirty-five adult male Wistar rats (4 months of age) were provided by the bioterium of our Institute. Rats were individually housed, had access to water and food *ad libitum*, and were maintained on an inverted 12 h∶12 h light-dark cycle, with lights on at 9:00 am. The animals were allowed to habituate to the room conditions and handled daily for at least 10 days before experiments began. The "bioethics committee" from our institute headed by Dr. Magdalena Giordano Noyola approved all the protocols and experimental procedures performed with the animals in the present study, which were done in accordance with international ethical guidelines for animal care and handling (ID:INEU/SA/CB/034). BrdU administration {#s4b} ------------------- Given that our goal was to accurately establish the date on which new neurons were able to respond to behavioral stimulation, we compared different 5-Bromo-2′-deoxyuridine (BrdU) (Sigma, St Louis MO ) administration procedures to maximize the number of newborn neurons detected on a single day. A dose of 200 mg/Kg, divided into four, 50-mg/Kg injections (diluted in 0.15 M NaCl solution) and administered every 4 hrs gave the highest number of BrdU-positive cells in the hippocampal dentate gyrus, as compared to other procedures tested. The first BrdU injection was at 9:30 am and the last was at 9:30 pm on the same day. After the BrdU administration, the animals remained undisturbed in their home cages until they were sacrificed after or without a 5-min spatial exploration session (see [Fig. 1](#pone-0017689-g001){ref-type="fig"}). Spatial exploration {#s4c} ------------------- In order to examine the response of new neurons to spatial exploration at different maturation times, 1, 7, 15, 30, or 45 days post BrdU injections (PBI), the animals were exposed to a 5-min exploration session (n = 5, 4, 4, 4, and 3 respectively) and were sacrificed 30 min later ([Fig. 1](#pone-0017689-g001){ref-type="fig"}). The exploration environment was an open square box, 70×70 cm, with 20-cm-high walls made of translucent acrylic. All the walls were covered with orange foamy paper, and the floor was partitioned into nine grids using black foamy paper strips. Each rat was fully covered with a white towel, then individually transported to the behavioral room, placed in the center of one of the grids in the apparatus, and moved to the center of a different grid every 15 s. This ensures that each of the grids was visited two or three times during the 5-min exploration session [@pone.0017689-RamrezAmaya1]. Immediately after exploration, the animal was placed back in its home cage and kept undisturbed. Cage control animals for each PBI group (n = 3) remained undisturbed in their home cages during the behavioral session and were sacrificed the same day and time as their respective exploration group. Brain extraction {#s4d} ---------------- Thirty min after the exploration session, each animal was killed by quick decapitation. The rat\'s brain was quickly and carefully extracted and frozen in 2-methylbutane (Sigma) by immersing it in a dry ice/ethanol slurry. The rat brains were stored at −70°C. Blocking and sectioning {#s4e} ----------------------- Using a stainless steel matrix (Electron Microscopy Sciences, Hatfield, PA), brain hemisections including the whole hippocampi were obtained. From 8 to 10 brain sections were molded into a block with Tissue-Tec OCT compound® (Sakura Finetek, Torrance, CA), such that each block contained brains from all groups and the position of each group differed in each block. The blocks (4 total) were cryosectioned into 20-ìm-thick coronal sections at −18°C in a CM1850 Leica cryostat (Nussloch, Germany), captured on slides (Lauka, MEX) pre-treated with diethyl-polycarbonate (Sigma) solution, and kept in a sealed box at −70°C before the immunostaining procedure. Immunostaining {#s4f} -------------- In order to maximize the detection of BrdU-positive cells expressing Arc, 30 to 40 serial sections from the dorsal hippocampus (Range between −2.60 to −4.3 from bregma) from each block were selected for the staining procedure. We used a triple immnunostaining protocol similar to that described previously to detect NeuN, Arc, and BrdU [@pone.0017689-RamirezAmaya1]. The tissue was fixed in 2% paraformaldehyde, pH 7.4, for 8 min at 4°C, washed in Tris-buffered saline (TBS), pH 7.0, and quenched in TBS with 2% H~2~O~2~ for 20 min. The sections were blocked for 40 min in tyramide signal amplification (TSA) kit blocking buffer (Perkin Elmer Life Sciences, Emeryville, CA). The tissue was then incubated sequentially with biotinylated mouse anti-NeuN antibody (1∶2000; Chemicon, Bedford, MA), with polyclonal rabbit anti-Arc antibody for the second detection (1∶500; a kind gift from Paul F. Worley\'s laboratory), and for the third detection with mouse anti-BrdU monoclonal antibody (1∶500 BD Biosciences México DF, México). After detecting NeuN and before Arc antibody incubation, the tissue was permeabilized with acetone/methanol (50∶50, v/v; Sigma) at 4°C for 15 min. For the detection of BrdU, the tissue was taken through a DNA denaturing procedure, consisting of an incubation with 50% formamide in 2×SSC buffer (Sigma) at 65°C for 2 h, washed in 2×SSC for 10 min, incubated in 2 N HCl at 37°C for 30 min, and washed in 0.1 M boric acid, pH 8.5, for 10 min. Biotinylated anti-NeuN was detected with the avidin-biotin A+B Vectastain amplification kit (Vector laboratories, Burlingame, CA) and the cyanine-5 (Cy5) TSA fluorescence system (PerkinElmer); before Arc detection, the A/B blocking kit (Vector laboratories) was used to block the remaining A+B, and the rabbit anti-Arc was detected with biotinylated anti-rabbit antibody (Vector laboratories), amplified with the A+B Vectastain amplification kit, and finally visualized with the Cy3 TSA fluorescence system (PerkinElmer). Mouse IgG from the first detection was blocked using the mouse-on-mouse blocking kit (Vector Laboratories) before detection of BrdU. The mouse anti-BrdU antibody was detected with a biotinylated anti-mouse antibody in which the signal was amplified using an A+B Vectastain amplification kit, and finally observed using the FITC TSA fluorescence system (PerkinElmer). In order to determine the neuronal lineage of BrdU-positive cells detected early in their maturation, a double staining for doublecortin (DCX) and BrdU was done. In this case, BrdU detection as described above was carried out first, and DCX was detected afterwards using a goat anti-DCX antibody (1∶200, Santa Cruz®), then amplified with A+B Vectastain amplification kit, and revealed with Cy3. We also performed double immunohistochemistry for Arc and DCX combining the methods described above. No staining was observed in the absence of the primary or secondary antibodies, for all antigens. Imaging and Analysis {#s4g} -------------------- The MosaiX module for the APOTOME system (Carl Zeiss, México, DF. México) with the 25×/0.80NA LCI Plan-Apochromat oil immersion objective was used to obtain whole dentate gyrus (DG) mosaic image stacks (with 1.5-ìm optical Z sections). About 8 to 12 individual image stacks were collected and assembled by the MosaiX system (Carl Zeiss) for each DG. About twenty-three whole DG mosaics, taken from the serial stained sections, were imaged for each animal, corresponding to a dorsoventral length of ∼460 ìm from the dorsal hippocampus. Note that only those sections that were optimally stained were included in the analysis. The most anterior section was ∼6.0 mm and the most posterior was ∼4.7 mm from the interaural plane [@pone.0017689-Paxinos1]. Using the Metamorph imaging software, a 2D image was constructed using the middle plane image from each DG MosaiX stack. This was used as the reference image, in which the DG granular layer was partitioned into 4 layers [@pone.0017689-Kempermann1] (see [Fig. 7B](#pone-0017689-g007){ref-type="fig"} for a visual description of the segmentation); one represented the subgranular zone (SGZ), and the other 3 represented the inner (L1), middle (L2), and outer (L3) part of the DG granular layer. It is important to note that the DG granular layer thickness varies across the length of the blade; for this reason, the proportion was adjusted throughout the whole length of the DG granular layer, assuring that the inner, medial, and outer DG granular layers always represented 33% of the whole layer thickness in order to analyze the position of the new granular neurons within the DG granular layer. Meanwhile, the MosaiX stack was used to identify the NeuN-positive cells (image in blue), the cells that had incorporated BrdU (image in green), and those expressing Arc (image in red). The BrdU-positive cells co-localized with NeuN, particularly in the animals sacrificed 15 or more days after BrdU administration; in animals sacrificed earlier, BrdU co-localized primarily with DCX ([Fig. 2A and B](#pone-0017689-g002){ref-type="fig"}). Nearly all of the BrdU-positive cellsco-localized with DCX on days 1 to 15, and most of them co-localized with NeuN from day 30 onwards. These expression time points are similar to those reported earlier [@pone.0017689-Brown1], in which the DCX or NeuN cells represented ∼90% of the total BrdU-positive population. Moreover, we found that in animals sacrificed at early time points (1--15 days PBI) during the maturation of these new neurons, Arc expression was found in DCX cells ([Fig. 7A](#pone-0017689-g007){ref-type="fig"}). For this reason, we included all BrdU-positive cells in the study, and the NeuN staining was used to delineate the DG granular layer. The Arc-positive cells were considered to be the activated neurons. Each cell was classified as BrdU-positive, Arc-positive or positive for both BrdU and Arc, and its classification was marked in the reference image, according to its position in the granular layer. It is important to clarify that the image stack was used to properly classify each cell as Arc-positive or BrdU-positive; additionally, when a cell was classified as both BrdU-positive and Arc-positive, a 40× confocal image stack was projected as a 3D image to confirm this classification (see [Fig. 3D to H](#pone-0017689-g003){ref-type="fig"}). The absolute BrdU-positive cell counts are the summation of all BrdU-positive cells found in all DG images from each animal (∼23 images per animal), and an average per group was obtained. After classification, the DG granular layer volume was calculated in the reference image, using a 40× confocal image in which the useful planes were obtained that, in combination with the area of the DG granular layer, were used to calculate the depth of the tissue (see [Figure 2C and D](#pone-0017689-g002){ref-type="fig"}). The total number of granular cells within the volume of the DG granular layer was estimated for each animal, as described before [@pone.0017689-Kuipers1], [@pone.0017689-Chawla1]. We estimated that about ∼80,000 granular neurons per animal were included in the analysis. Using the total number of granular cells per animal, the proportions of BrdU-positive cells, Arc-expressing cells, and BrdU-positive Arc-expressing cells were calculated. Statistics {#s4h} ---------- One-way ANOVA with Bonferroni as a Post-hoc test, MANOVA, or a student\'s t-test was used where appropriate to compare the proportion of BrdU-positive, Arc-expressing, and BrdU-positive Arc-expressing cells in the different conditions and throughout the DG granular layer. We thank Nydia Hernández-Rios from the Confocal Microscopy Unit (INB UNAM) and Cutberto Dorado Mendieta from lab A-13 (INB UNAM) for technical assistance. This work is part of Claudia Jimena Sandoval Arroyo\'s Ph.D. thesis for the "Programa de Doctorado en Ciencias Biomédicas" of the "Universidad Nacional Autónoma de México". **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded by CONACyT 51028 and PAPIIT IN213907 and IN216510-21. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: VRA CJS. Performed the experiments: CJSA MMC PCBM. Analyzed the data: VRA CJSA OP. Contributed reagents/materials/analysis tools: NHR CDM. Wrote the paper: VRA CJSA.

PubMed Central

2024-06-05T04:04:19.232032

2011-3-9

PMC3052369

Introduction {#s1} ============ The rapid improvements in sequencing technologies are adding new sequences to the databases faster than the pace at which insights into their function could be gained. As a consequence, the vast majority of known genes and proteins have not been characterized experimentally, and their function is yet unknown [@pone.0017679-Punta1]. Moreover, biological functions are not, in general, realized by individual proteins, but, rather, by networks of intricate interactions between numerous genes. The understanding of biological processes requires, therefore, a better knowledge of the functional organization of such networks. Indeed, the study of biological processes increasingly relies on the analysis of biological networks (BN), which has been used to tackle different levels of the functional organization of the cell. On the level of individual proteins, BN are often used to help to elucidate the molecular function of specific proteins [@pone.0017679-Sharan1], [@pone.0017679-Hu1]. On the systems level, they are studied to reveal modules and functional sub-networks [@pone.0017679-Alon1], [@pone.0017679-Stuart1]. An issue that has hardly been faced is that of the meta-organization of different functions in a single, integrated, network. Yook *et al.* [@pone.0017679-Yook1] have concluded that most functional classes appear as segregated sub-networks of the full protein interaction network (PIN). Like most of the studies of BN, the results of Ref. [@pone.0017679-Yook1] are based on parsing the static network, and do not allow the exploration of the meta-organization and the interactions of the sub-networks. We here, instead, give evidence that a dynamical approach to the analysis of BN based on their meta-organization not only enhances the prediction of the function of individual proteins, but also can reveal information on the network macro-scale of interactions between different biological functions. As for predicting the function of individual proteins, two main strategies have been followed so far. The first relies on the analysis of the protein itself: e.g. its similarity to already annotated proteins, its structure, or its biophysical features [@pone.0017679-Punta1], [@pone.0017679-Lee1], [@pone.0017679-Rost1]. The second one is, instead, based on high-throughput technologies providing data that may highlight the context in which the protein acts such as its sub-cellular localization, interactions with other proteins, and the conditions under which it is expressed (or the genes that are co-expressed with it) [@pone.0017679-Sharan1], [@pone.0017679-Hu1]. High-throughput protein-protein interactions detection experiments allow nowadays a representation of the global cell functioning in terms of a network, with nodes representing proteins and edges representing the detected mutual interactions, with the goal of exploiting the properties of these networks for prediction purposes on the function of specific proteins. Notwithstanding the accomplishments of these analyses it is important to highlight that most high-throughput methods can suffer from high false positive and false negative rates [@pone.0017679-vonMering1] and, therefore, functional assignments that are based on these tools may lead to misclassifications. Several past studies attempted already to determine to what extent the function of a protein depends on the way it is interacting with the others in the PIN. However, the use of such network representation for prediction requires the determination of the specific scale of the PIN that one has to consider for unveiling the individual protein\'s function. And, in this latter framework, the current state of the art includes, again, two types of approaches. From one side, several direct annotation schemes have been devised [@pone.0017679-Schwikowski1]--[@pone.0017679-Karaoz1], with the common inspiration of analyzing the local scale features of the PIN, i.e. either basing the function prediction on the information that can be directly extracted from the protein neighborhood, or statistically assessing a probability for a protein to be assigned to a given function, depending on the actual number of its neighbors that are (or are not) pertinent to the same function. From the other side, more recent module assisted techniques [@pone.0017679-Yeang1]--[@pone.0017679-Kelley1] have attempted to make use of the extra knowledge arising from the meso-scale of clustered structures of the PIN, with first identifying dense agglomerates in the network that are loosely connected to other areas of the graph, and then to use this topological information for predictions on the protein specific function. The approach we lay out constitutes a third, novel, strategy. We provide evidence that an alternative source of information is, in fact, the one arising from the analysis of how the modular PIN structure actually organizes the synchronization dynamics of an ensemble of oscillators. In particular, we show how the combination of synchronization features emerging in the PIN structure with a rudimentary classification of proteins based on expert manual assignment, allows, indeed, to gather information on misclassification problems, as well as to offer a more accurate function assignment that is consistent with more recent (and better refined) manual annotation of these proteins\' function. Not less important is the ability of the approach we introduce to assess the coupling of different functional categories, to determine how closely associated they are, and which proteins participate in both of them. Materials and Methods {#s2} ===================== Data {#s2a} ---- For our research we have used a typical and important network with rudimentary functional assignments derived from a *Saccharomyces cerevisiae* PIN, as reported in [@pone.0017679-Bu1]. The data set is based on the work by von Mering *et al.* [@pone.0017679-vonMering1] who scored the reliability of 80,000 reported protein-protein interactions in the yeast. These were based on high-throughput interaction detection methods, such as *i)* yeast two-hybrid systems [@pone.0017679-Uetz1], [@pone.0017679-Ito1], *ii)* protein complex purification techniques using mass spectrometry [@pone.0017679-Gavin1], [@pone.0017679-Ho1], *iii)* correlated messenger RNA expression profiles [@pone.0017679-Cho1], [@pone.0017679-Hughes1], *iv)* genetic interaction data [@pone.0017679-Tong1], [@pone.0017679-Mewes1], and *v)* "in silico" interaction predictions derived from gene context analysis. From this set, Bu *et al.* [@pone.0017679-Bu1] focused on 11,855 interactions (those featuring high and medium confidence levels) among 2,617 proteins. We here focus on the giant connected component of the PIN given in Bu *et al.* [@pone.0017679-Bu1], consisting of proteins and interactions. As for the modular structure of the PIN, we initially refer to the partition in 13 functional categories given by the yeast protein catalog at the Munich Information Center for Protein Sequences (MIPS) [@pone.0017679-Mewes1]. Particularly, we use the data set in which each given protein is assigned to one of the functional categories (with proteins in multiple categories manually assigned by Bu *et al.* [@pone.0017679-Bu1] to only one). In order to test the validity of our findings, we will use the classification provided by the Gene Ontology consortium (GO) [@pone.0017679-Ashburner1]. While MIPS attempts to provide a simple hierarchy with intuitive category structure that allows for manual browsing, GO aims at representing a fine granular description of proteins that provides annotation with a wealth of detailed information. Thus, MIPS gives a very rough division into a couple of dozens of categories and several hundreds of subcategories, whereas GO includes 29,983 different functional terms (as of March 2010). GO also provides a reduced version of its ontology (GOslim) that allows one to trace the detailed terms into more coarse-grained categories. In our analysis, we start with the single MIPS classification for each protein, and use the dynamical overlap method for identifying those proteins that are likely to be involved in more than one of the functional categories in our data (those ones forming the overlapping structures). As a validation, we refer to the classification of these proteins in GOslim, Namely, by manually mapping each GOslim term to one of the 13 MIPS categories, one is able to verify whether or not the assignment of the second function (provided by our method for each one of the proteins in the overlapping sets) is consistent with the functional annotation in GO. Dynamical Overlap Formalism {#s2b} --------------------------- The method is based on the inspection of how oscillators organize in a modular network of dynamical interactions [@pone.0017679-Boccaletti1], by forming synchronization interfaces and overlapping communities [@pone.0017679-Li1], [@pone.0017679-Almendral1]. Here, we will consider a network of phase oscillators on top of the PIN. Thus, the transfer of function between neighboring proteins is performed through the synchronization of coupled oscillators. In order to explain how the method works, let us assume the PIN of the yeast is topologically divided into two main modules, and , each one of them associated to a specific protein function. Every node (protein) in the network is an oscillator whose frequency is set to () whenever the node belongs to (), with . The phase dynamics of this network of coupled oscillators can be described bywhere dot denotes temporal derivative, is the phase of the -th oscillator, is the number of interactions that the -th protein has with the rest of proteins, is some coupling strength, and are the elements of the adjacency matrix representing the PIN [@pone.0017679-Boccaletti1], with if there is an interaction between proteins and , and otherwise. In the extreme case of fully separated modules the network dynamics would eventually (at large coupling strength ) result in the clusters and oscillating synchronously at a constant, different, frequency. If, however, there are just a few interactions between proteins of the two modules, the onset of a *synchronization interface* overlapping the two modules occurs, made of all those nodes displaying an instantaneous frequency that are actually oscillating in time around the mean value of the two frequencies characterizing the clusters [@pone.0017679-Li1]. The rest of nodes, out of the synchronization interface, oscillate at the frequency of the module they belong to. To quantify this behavior, we monitor the instantaneous frequency of each oscillator and we calculate the indicator ,which accounts for how close in time the frequency associated to protein is to the average frequency of the two clusters, and . By fixing a confidence threshold , those proteins belonging to module () have () as they were assigned initially the frequency (), while is the signature of a protein whose module membership is not clear, belonging to the synchronization interface between and . This behavior is graphically sketched in [Fig. 1](#pone-0017679-g001){ref-type="fig"}. There, a small graph composed of 8 nodes ([Fig. 1A](#pone-0017679-g001){ref-type="fig"}) clearly has two densely connected modules that do not coincide with the given functional classification denoted by the color of the nodes. Actually, node 8 does not have any link within its functional module, the yellow one, while node 4 is classified within the blue functional module but shares the same number of links with the other functional module. After solving Eq. (1) by assigning to nodes 1--3 and 8 (functional module ), and to nodes 4--7 (functional module ), the corresponding values extracted from Eq. (2) indicate that nodes 1--3 really belong to module (as ), nodes 5--7 belong to module (as ), while nodes 4 and 8, whose , are the ones candidates to be overlapping between and . To solve this uncertainty, nodes 4 and 8 are reassigned to (blue) and (yellow) respectively ([Fig. 1B](#pone-0017679-g001){ref-type="fig"}) and we observe that whereas falls now within the area of module , increasing the cohesion of the functional module, node 4 still lies within the synchronization interface () overlapping between both modules. ::: {#pone-0017679-g001 .fig} 10.1371/journal.pone.0017679.g001 Figure 1 ::: {.caption} ###### Graphical description of the dynamical overlap method. (**A**) A two module small graph composed of 8 nodes colored according to their membership to the functional module (yellow) or (blue), and corresponding values after solving Eq. (1) with for and for . Nodes 4 and 8 have with this functional classification. (**B**) Same as in (**A**) but nodes 4 and 8 has been reassigned to modules and respectively. Now, node 8, behaves as a node truly from while node 4 behaves as an overlapping node between and as is again close to zero. All the network representations in this manuscript were produced with Cytoscape. :::  ::: For the real situation of a PIN with different functional modules (), this can be done by integrating times the network dynamics described by Eq.(1). In each trial, the -th module () is assigned to the cluster frequency , whereas the rest of the PIN is given the second cluster frequency , resulting in a series of values. This time all those proteins initially assigned to whose , actually belong to module , while if belong to another module different from . All those nodes whose are labeled as belonging to the synchronization interface between module and the rest of the network. Then, a node is identified as an overlapping node between modules and if, being a node from either or , is in both and , that isFinally, the set of nodes of module overlapping with module , with , iswhich has two implications: *i)* while is symmetric in the indexes, is not, and *ii)* and, since , . Eventually, the degree of overlapping between two modules is then given by:which, therefore, provides a measure of how many nodes out of the clusters and are forming the corresponding overlapping structure. The main result of our method is, therefore, an index accounted by Eq.(2), that, for each protein , measures its degree of membership to module (i.e. a protein function). A value indicates that the protein exhibits a dynamical behavior different from that of the majority of proteins in , thus clearly belonging to other module. On the other hand, occurs when the protein performs as the rest of proteins assigned to the same module , and this confirms that it is certainly member of . Finally, a value of close to zero is the signature of a protein whose module membership requires further analysis as it could be the case of a protein belonging to two or more functional modules. Therefore, we are introducing an index that allows to check the accuracy of the initial functional assignment as well as predicting a second (or more) function of a protein. Results and Discussion {#s3} ====================== The application of the method given by Eq. (1) to the PIN and modular classification with , , , and , as described in the [Materials and Methods](#s2){ref-type="sec"} section leads to 13 different series for (being the functional module index and the protein index). In order to proceed with the full analysis of this data, we have to consider all possible combinations of these series to check whether a protein belongs to the functional module initially assigned or whether it is involved in more than one functional module. This can be done efficiently, as shown in the [Figure S1](#pone.0017679.s001){ref-type="supplementary-material"}, but, to illustrate the principles underlying the method, we will just focus on a single pair of functions. [Figure 2A](#pone-0017679-g002){ref-type="fig"} shows the values of the indexes and , being and the *Cellular fate/organization* and *Genome maintenance* functional modules. We plot proteins initially assigned to () in blue (red), while the rest of proteins are plotted in black. Notice that most of the black points are concentrated around , as the corresponding proteins neither belong to nor . The majority of proteins in (blue) and (red) are located close to and , respectively. The blue points inside the ellipse correspond to proteins initially classified as that are not belonging to (as ), but for the very same to is under question (). When examining the indexes for the rest of modules, one finds out . Therefore, we infer that these proteins do, indeed, belong to although weakly. The same arguments apply for the red points lying within the other ellipse: they are proteins weakly ascribed to . A completely different situation is that of those points distributed around (inside the circle, mostly of the points superimposed). They correspond to 15 proteins whose unique membership to and cannot be asserted. When checking the rest of values, one finds that none of these proteins can be assigned to modules other than and , thus again they are weakly associated to both functions and (one of them being the initially assigned function, and the other the predicted one). The novelty here is that there is a twofold assignation, which could be considered as the trace of multi-functional proteins. ::: {#pone-0017679-g002 .fig} 10.1371/journal.pone.0017679.g002 Figure 2 ::: {.caption} ###### Identification of misclassified proteins. (**A**) and values for all proteins in the PIN of the yeast. The color indicates the functional module initially assigned to each protein (blue for , red for and black for the rest). The method identifies 15 proteins (within the circle) with a twofold assignation (the initial and the predicted one). After re-assignation to the predicted function, the new values of the 15 proteins are depicted as circles bordered with the color of that function, and lie together with those other proteins of the same function, indicating an original misclassification. (**B**) Visualization of the network backbone, made of the 15 misclassified proteins and their neighbors. Same color code as for (**A**). :::  ::: Before claiming for multi-functionality, it is mandatory to check if such a multi-assignment holds when the initial modular structure changes. This is tantamount to reassign each one of these proteins to the predicted function and check whether the corresponding protein is still located around , otherwise the multi-functionality is simply an artifact. The new values for the 15 proteins (after reclassification) are shown in [Fig. 2A](#pone-0017679-g002){ref-type="fig"} as circles bordered with the color of the predicted function. The remarkable result is that the emerging dynamics behavior agrees with the new classification, as the 15 proteins are no longer overlapping and move now to the areas corresponding to the predicted function. If we take into account the number of connections a given protein is forming with elements belonging to any one of the other modules in the graph, , the emerging dynamics is reflecting the fact that the original and predicted assignments correspond, respectively to and , that is, the predicted classification makes the functional module more cohesive (see [Fig. S2B](#pone.0017679.s002){ref-type="supplementary-material"}). For the sake of visualization, [Fig. 2B](#pone-0017679-g002){ref-type="fig"} shows the backbone of the original PIN made of the 15 proteins and all their neighboring proteins. While the original function assignment classified the proteins in modules in which they do not have physical interactions, the reclassification is able to unveil the participation of the proteins to the correct module. For example, according to GO, YHR172W is not involved in Cellular fate/organization but in Genome maintenance (see [Table S1](#pone.0017679.s004){ref-type="supplementary-material"}), which is in agreement with the classification pointed by our method. Notice that, in the full analysis, the number of proteins featuring an overlapping behavior is 418 (see the full list in [Table S1](#pone.0017679.s004){ref-type="supplementary-material"} and [Fig. S2](#pone.0017679.s002){ref-type="supplementary-material"}) out of which 103 proteins have no functional annotation in GO and 200 had two or more different function annotations in GOslim. For these latter ones, a comparison with the functions assigned by GO reveals that in 87 cases the predicted function is in agreement with one of the GO assignments. The expected average number of matching of the proteins in for a random function assignment is 25. The p-value for the significance of this result is 0.0001, and it can be established by performing 1,000 random reshuffles of function assignment, and verifying the average number of matches (which in this case was 25). The highest number of random matches was 50 (in 1/1,000 cases), well below the observed 87. As a result, one can claim an original misclassification and, consequently, the method can be used to cure errors in a given protein function classification. With the guidance of the information obtained so far, we have reclassified all proteins of to the corresponding predicted functions, and extracted the subgraph of the original PIN for which each functional module corresponds to a connected component (i.e. we pruned out all those other proteins that were assigned a given function in the MIPS classification, but did not have any interaction with other elements of the same function). The result is a new interaction network made of 2,049 nodes and 9,941 links, that we take for a new set of numerical trials, resulting in a second list of 211 potentially multi-functional proteins (reported in [Table S2](#pone.0017679.s005){ref-type="supplementary-material"}). The situation, is now radically different: at variance with the results of [Fig. 2](#pone-0017679-g002){ref-type="fig"}, [Fig. 3](#pone-0017679-g003){ref-type="fig"} shows that the multi-functional nature of the 30 proteins inside the circle (the subset of obtained when comparing (*Transcription*) and (*Translation*), is indeed genuine, as the final outcome does not depend on whether the proteins are classified according to the assigned or predicted functional modules (see [Fig. S3A](#pone.0017679.s003){ref-type="supplementary-material"}). This is further confirmed by the simultaneous reclassification of each one of the proteins of into the predicted function, and by monitoring the change in the out-degree, , calculated with the predicted and the original classification (shown in [Fig. S3B](#pone.0017679.s003){ref-type="supplementary-material"}). ::: {#pone-0017679-g003 .fig} 10.1371/journal.pone.0017679.g003 Figure 3 ::: {.caption} ###### Identification of multi-functional proteins. (**A**) and values for the 2,049 proteins in the PIN of the yeast after curation. The color indicates the functional module initially assigned to each protein (orange for , green for and black for the rest). The 30 proteins located inside the circle remain there after re-assignation to the predicted function, and are depicted as circles bordered with the color of that function. (**B**) Visualization of the network backbone made of 6 (out of 30) of the multi-functional proteins in (**A**). :::  ::: An independent test of the validity of that assignment is to assess the multi-functionality character of the proteins in by comparison with the more accurate GO classification scheme. One can count the number of different GO annotations for each of the proteins in , and the corresponding distribution of multiple assignments in the rest of the data. The difference between the two distributions (see [Figure 4](#pone-0017679-g004){ref-type="fig"}) is significant (p-value, as for conventional t-test). Namely, the average number of different function assignments in is 6.7, with mode 4, while in the other proteins one finds 4.9 and 3 respectively. Moreover, the standard deviation of the distribution of functions in is significantly greater than that of the other proteins. This confirms that the proteins in come from a population with higher multi-functionality with respect to the population of other proteins. ::: {#pone-0017679-g004 .fig} 10.1371/journal.pone.0017679.g004 Figure 4 ::: {.caption} ###### Statistical assessment of protein multi-functionality. Probability density function of the number of different GO annotations (see [Materials and Methods](#s2){ref-type="sec"} section) of the 211 overlapping proteins in (blue diamonds), as compared to the probability of other proteins in the rest of the data (red squares). Continuous lines are shape-preserving interpolations. :::  ::: Finally, the method allows also to assess a coarse-grain representation of the PIN, showing the way each biological function is interacting with the others. In [Figure 5](#pone-0017679-g005){ref-type="fig"}, each specific cell function is represented by a node whose size is proportional to the total number of proteins participating in that function. The width of each link is proportional to the number of multi-functional proteins provided by our method (Equation (4)). The resulting network representation of the full cell functioning suggests numerous insights about the organization and control of biological functions. As one might expect, there is a strong link between Transcription, Translation and Transcriptional control. But these functions have almost no common proteins to functions like Genome maintenance, Cellular organization or Metabolism. Interestingly, the results show that there are no shared proteins between Amino-acid metabolism and Protein fate, suggesting that even though these two processes may seem related there are no known common mechanisms that control both functions. ::: {#pone-0017679-g005 .fig} 10.1371/journal.pone.0017679.g005 Figure 5 ::: {.caption} ###### Coarse grained representation of the PIN in terms of cell functioning and coordination. The size of nodes is proportional to the total number of proteins participating to the corresponding function, the width of the links is proportional to the size of the corresponding overlapping interface. The full picture of the structure of these overlaps is reported in the [Table S3](#pone.0017679.s006){ref-type="supplementary-material"}. :::  ::: We have then given evidence that a proper inspection on the meso-scale interactions of a generated network of dynamical systems can provide useful information on the micro- and macro-scale processes through which biological processes are organized in a cell. The method is not only able to predict and reassign the function of a given protein, but also to describe qualitatively the main functional interactions that lead to the global functioning of the organism. It is worth highlighting that the present application only focused on unveiling proteins with double functionality, while the method can be easily applied to gather information also on proteins bridging among more than two different biological functions (such an evidence will be reported elsewhere). The core of the presented results gives insights on how molecular functions are networking at different scales, as well as on how to design (or engineer) proper drugs, or mechanisms to control (or regulate) the biological interactions responsible for the functioning, or malfunctioning, of a cell. Supporting Information {#s4} ====================== Figure S1 ::: {.caption} ###### **Identification of misclassified proteins.** The proposed tool is providing the behavior of each protein in the PIN through the indicator , that crucially depends on its original functional classification. Each panel corresponds to the competition trial between module at frequency (in black symbols) and the rest of modules at frequency (in different symbols and colors). The size of each module is written between brackets. Nodes belonging to the corresponding synchronization interface (, gray band) are marked in full face. Those nodes corresponding to overlapping proteins (those appearing in two synchronization interfaces, and ) are encircled with the color of the corresponding overlapping function. Parameters used in Equation (1): , , and (**A**), (**B**). (Transcription), (Other metabolism), (Uncharacterized), (Cellular fate/organization), (Protein fate), (Translation), (Amino-acid metabolism), (Genome maintenance), (Cellular organization), (Energy production), (Stress and defence), (Transcriptional control), (Transport and sensing), and (Transport and sensing). (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Identification of misclassified proteins.** (**A**) Dynamical behavior of the 418 overlapping nodes. In blue when the modules are defined according to the original classification (MIPS). Given that the overlapping node is simultaneously in and , we represent with a circle its value in and with a square its value in . In red we represent the same values as before but when the modules are modified to take into account the function predicted by our method for the overlapping nodes. Same parameters as in [Fig. S1B](#pone.0017679.s001){ref-type="supplementary-material"}. (**B**) Topological behavior. , change in the ratio between out-degree (, number of connections a given protein is forming with elements belonging to any one of the other modules in the graph, and the underscores predicted/original stay for the calculation of in the corresponding annotation) and total degree (, degree of the protein, independent on the specific classification of the protein) of the proteins in (green dots) and the rest of the proteins (black dots) when reassigning the function given by MIPS to the predicted one. The results show that, while all non overlapping proteins (black points) are grouped around (i.e. they do not substantially change their *in-out* connections due to the change in the classification of the overlapping proteins), the members of (green points) appear grouped around , thus reflecting the fact that the original and predicted assignments correspond, respectively to and . This indicates that in the original classification of the proteins in they did not have interactions with other elements of the original functional module, whereas the predicted classification assigns them to the proper functional class. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Identification of multi-functional proteins.** (**A**) Dynamical behavior of the new set of overlapping proteins. In blue, values of the set of overlapping proteins between modules and with the new cured classification (same as in [Fig. 3](#pone-0017679-g003){ref-type="fig"}). As in [Fig. S2](#pone.0017679.s002){ref-type="supplementary-material"}, we plot the value of the overlapping node with circles when is in and with squares when in . In red we represent the same values as before but when the modules are modified to take into account the function predicted by our method for the overlapping nodes. (**B**) Topological properties of the cured PIN. Change in the ratio between out-degree () and total degree () of the proteins in (green dots) and the rest of the proteins (black dots) when reassigning the function given by MIPS to the predicted one. Parameters used in Eq. (1): , , and . (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **List** **of proteins.** Full list with the 418 overlapping proteins resulting from the first iteration of the dynamical overlap method for the PIN of the yeast (see [Materials and Methods](#s2){ref-type="sec"} and [Fig. 2](#pone-0017679-g002){ref-type="fig"}). For each protein, we provide the OLN (Ordered Locus Names), the MIPS classification, whether or not this function is annotated in GOslim, the predicted function and whether or not this predicted function is also provided by GOslim. The first 87 proteins correspond to cases in which the predicted function is in agreement with one of the GO assignments. (PS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **List** **of proteins.** Full list with the 211 overlapping proteins resulting from the second iteration of the dynamical overlap method for the curated PIN of the yeast (see Text and [Fig. 3](#pone-0017679-g003){ref-type="fig"}). The curation of the PIN consists in exchanging the annotated function by MIPS of the 418 proteins from with the function predicted by the overlap and removing those proteins that become isolated within the functional module. Again, for each protein, we provide the OLN (Ordered Locus Names), the MIPS classification and the predicted function. (PS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Multifunctional distribution of proteins in** **.** Module index. Number of proteins within the -module. Overlapping nodes belonging to . Number of proteins belonging to the -module overlapping with module . (PS) ::: ::: {.caption} ###### Click here for additional data file. ::: We wish to thank the EU project DAPHNet, ONR, DTRA, the Israel Science Foundation. The authors acknowledge CRESCO (Computational Research Center for Complex Systems, co-founded by ENEA, Italian National Agency for New Technologies, Energy and Sustainable Economic Development) for providing access to their computing facilities. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Work partly supported by EU contract 043309 GABA, the Spanish Ministry of S&T under Project n. FIS2009-07072, and the Community of Madrid under the R&D Program of activities MODELICO-CM/S2009ESP-1691. No additional external funding received for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SB SH. Performed the experiments: ISN. Analyzed the data: ISN YO JA. Contributed reagents/materials/analysis tools: JA JB IL DL. Wrote the paper: SB SH.

PubMed Central

2024-06-05T04:04:19.236994

2011-3-9

PMC3052370

Introduction {#s1} ============ Collective foraging by social insects is a complex process that involves assessment of resource quality by workers, decision making, food carriage and information transfer (e.g., bees [@pone.0017667-Nez1], [@pone.0017667-Nez2]; bumblebees [@pone.0017667-Bertsch1]; wasps [@pone.0017667-Pflumm1]; ants [@pone.0017667-Roces1], [@pone.0017667-Gordon1]). A social insect colony does not "decide" about, for instance, the selection of a given food source, yet the colony\'s foraging responses arise from the decisions made by each individual worker, and from the interactions among colony members. Central for the understanding of social foraging is the fact that most of the resources collected by foragers are not for their own consumption, but fed to the brood and to nestmates, or stored. Whatever the final end of the collected resources, individual foragers are expected to behave in a way that maximizes food intake at the colony level, because one can assume that the foraging performance of a colony, measured as the delivery rate of food, correlates with colony fitness. The analysis of individual foraging performance, and its extrapolation to account for overall colony responses, is nonetheless not straightforward because returning foragers deliver two commodities to the colony: food and information [@pone.0017667-Nez3]--[@pone.0017667-Roces2]. Because food gathering consumes both time and energy, one could a priori expect a trade-off between time spent collecting on one side, and time used for information transfer on the other side, if colony foraging performance is enhanced through communication, even at the expense of a reduced individual foraging performance. In this scenario, workers could potentially favor the delivery of one of these commodities at the expenses of the other, i.e., workers may spend less time collecting at a food source, thus reducing their own food-delivery rate, but allocate more effort to pass information about the discovery, thus leading to a higher food intake at the colony level. An early return to the nest shortens the foragers\' time to transmit information both via recruitment signals and through direct interactions with nestmates, and may allow potential recruits to get quickly informed about a food discovery [@pone.0017667-Roces2]--[@pone.0017667-Detrain1]. When this dual aspect is kept in mind, it becomes clear that the optimal policies for delivering each of these two commodities, food and information, might be different, and that the maximization of food provisioning at the individual level, with the associated collecting and carrying costs, is incompatible with a rapid delivery of information [@pone.0017667-Roces2], [@pone.0017667-Ydenberg1], [@pone.0017667-Dornhaus1]. The extreme strategy of time saving at the food source for information transfer would be to return completely unladen to the nest, but displaying recruitment behavior, as described for leaf-cutting ants [@pone.0017667-Jaff1]. Leaf-cutting ants (Attini, genera *Atta* and *Acromyrmex*) have repeatedly been used as a model system to explore how communication demands influence foraging decisions, because workers can make, through flexibility in cutting behavior, their own decisions about the fragment size to be cut [@pone.0017667-Roces4]. Leaf-cutting ants are conspicuous herbivores of the Neotropics that cut vegetation into small fragments, which are then transported to the nest as substrate for a symbiotic fungus. It has repeatedly been observed that polymorphic leaf-cutting ants, because of their geometric mode of cutting roughly semicircular fragments, frequently cut leaf pieces that correspond in size to that of the ants\' body [@pone.0017667-Lutz1]. However, not all workers cut fragments of maximal size, indicating the involvement of more flexible mechanisms of load-size determination. For instance, foragers cut smaller fragments from sources located close to the nest [@pone.0017667-Roces5], and also from hard leaves [@pone.0017667-Roces6]--[@pone.0017667-Cherrett1]. In addition, leaf-cutting ant workers were shown to use flexible cutting rules at a newly-discovered, highly-attractive food source: workers shorten their foraging time by cutting smaller fragments and return to the nest with partial loads, intensively recruiting additional nestmates, as discussed in the framework of the so-called information-transfer hypothesis [@pone.0017667-Nez2], [@pone.0017667-Roces6]. This hypothesis states that at newly-discovered food sources, foragers select loads that do not maximize their individual intake rates, but allow them to return earlier to the colony for information transfer. Foragers\' performance as food carriers is therefore reduced, but the colony as a whole increases its harvesting rate due to the workers that gained information and participate in the resource-gathering activity [@pone.0017667-Nez2], [@pone.0017667-Roces2], [@pone.0017667-Roces4], [@pone.0017667-Roces6], [@pone.0017667-Roces7]. Thus, saving cutting time/energy because of the selection of small leaf fragments contributes to speed-up information transfer. So far, the extent to which the decision to transfer information about a food discovery influences individual cutting behavior was only analyzed under controlled laboratory conditions, and only for leaf-cutting ants (reviewed in [@pone.0017667-Roces7]). Upon discovery of a highly-attractive food source, they are known to cut smaller leaf fragments and to engage in intense recruitment behavior, which leads to a faster build-up of workers at the source [@pone.0017667-Roces2], [@pone.0017667-Roces6], [@pone.0017667-Roces7]. While variation in leaf fragment size is necessarily associated with different cutting effort in leaf-cutting ants that cut semicircular fragments, the situation is rather different for grass-cutting ants. Workers of grass-cutting ants climb on a grass blade and cut across its width, which results in the selection of a longish, more or less rectangular grass fragment. Therefore, cutting length is represented by the grass width, which usually does not differ very much along the blade except at its tip. Hence, cutting a larger (longer) fragment does neither imply a higher cutting effort nor a longer cutting time, if grass toughness remains unchanged along the blade. Grass-cutting ant workers may therefore harvest more material per unit cutting effort by simply cutting longer fragments. Grass-cutting ants offer a unique opportunity to investigate whether workers use flexible cutting rules and compromise their load delivery for a faster return to the nest, because changes in load-size determination (fragment length) are not associated with additional costs or energy savings, as outlined above [@pone.0017667-Rschard1]. In addition, since workers do not anchor their hind legs at the grass end while cutting, body size poses no upper limit to the fragment size to be cut, as it is the case for leaf-cutting ants. Differences in fragment length are known to affect walking speed and individual delivery rates [@pone.0017667-Rschard2], [@pone.0017667-Moll1], so that the selected fragment size is expected to largely influence the duration of a worker\'s foraging cycle. When should information be worth transferring? We predict that the needs for information about a food discovery should be particularly high at the beginning of foraging, in order to allow a collective foraging process to get established. This study explores the extent to which grass-cutting ant workers use different cutting rules depending on the foraging phase, and save time as a response to the high demands for information transfer at the beginning of each daily foraging process. For that, we first quantified the foraging dynamics of several field colonies of the grass-cutting ant *Acromyrmex (Moellerius) heyeri*, counting the number of outgoing, returning laden and unladen workers, as well as the rates of their head-on encounters, at two distinct phases: initial and established foraging. These two phases are expected to be associated with different demands for information transfer: high at the beginning of a foraging process, and low on well-established trails [@pone.0017667-Roces2], [@pone.0017667-Detrain1], [@pone.0017667-Jaff1], [@pone.0017667-Detrain2], [@pone.0017667-Traniello1]. In addition, the sizes of the grass fragments cut by workers at the two different foraging phases were recorded, as well as the workers\' walking speed, allowing the later calculation of the individual gross transport rates of vegetable tissue. To control for the marked differences in environmental temperature during the field measurements, which are expected to largely influence walking speed, the relationship between walking speed and temperature was established over a wide range of temperatures in a laboratory colony, thus allowing the calculation of Q~10~-values for walking speeds, and the standardization of carrying performance at a given temperature for meaningful comparisons. Finally, both handling and cutting times by workers offered similar grasses as those harvested in the field were measured in a laboratory colony, to have a more comprehensive picture of the total time investment of single workers at both the initial and established foraging phases. Methods {#s2} ======= The field study to quantify foraging dynamics and fragment-size determination was conducted between March and November 2009 in Tabaré, Florida, Uruguay (33°21′33.74″S, 55°35′33.38″W). A total of 15 *Acromyrmex heyeri* colonies were investigated. Counts of outgoing, returning (laden), returning (unladen) workers were performed, as well as records of walking speed, head-on encounters, fragment size and body size of laden workers. Two different phases along a natural daily foraging process were first determined. The "initial foraging phase" was considered to begin as soon as the first laden worker was observed to walk to the nest coming from the foraging area. The "established foraging phase" was considered to begin 2.5 hours later, based on the counts as presented in the results. Three to four hours later, foraging activity ceased due to the high environmental temperature, and not because of exhaustion of the available grasses. Data were collected on each colony\'s main foraging trail, 2 m away from the nest. Traffic flow and head-on encounters at the two foraging phases {#s2a} -------------------------------------------------------------- Traffic flows of outgoing and returning (both unladen and laden) workers were determined along the colony\'s main foraging trail at regular time intervals, as well as the number of head-on encounters between focal workers and nestmates coming from the opposite direction, for both the initial and established foraging phases. Counts were performed on 12 *A. heyeri* field colonies during November 2009. At the measurement point, workers passing by were counted for 5 minutes, beginning at the time in which the first laden worker was observed to walk to the nest (initial phase). A second count was performed in the same colony at the established foraging phase. A similar procedure was used for the remaining 11 colonies. Immediately after each of the two counts, 12 randomly-selected workers (4 outgoing, 4 laden and 4 unladen walking to the nest) were carefully observed as they walked along a 30 cm trail section, and both the number of contacts (head-on encounters) with workers coming from the opposite direction, and the time elapsed to walk along the trail section, were recorded, so as to calculate the contact rates. Since the probability of contacting a worker directly depends on the number of workers coming from the opposite direction, and this number is expected to depend on the foraging phase considered, comparative values were obtained by dividing the contact rates (contacts\*s^−1^) by the traffic flow (number of workers\*s^−1^), which was recorded as indicated above. In order to quantify more accurately the traffic dynamics as a function of day time, traffic flows were measured in an additional colony every 20 min throughout both phases, over six consecutive days. Fragment-size determination, walking speed and carrying performance at different temperatures {#s2b} --------------------------------------------------------------------------------------------- Walking speed of 10 laden workers was measured over 30 cm on a smooth trail section, once for each of the 15 colonies investigated during March 2009, for both the initial and established foraging phase. Immediately thereafter, laden ants were collected, placed singly in 10 ml vials, and frozen. The scored ants were randomly chosen by collecting every fifth forager that passed by the collecting point. Considering that the collection of 10 ants took approximately 15 min in each foraging phase, no more than two colonies could be investigated each day, so as to keep the measurements at both the initial and established foraging phase within a time window of 30 min. Only workers carrying fresh grass fragments, i.e., those cut immediately before, were considered for further analysis. Workers carrying falling flowers or dry grasses were ignored because these items were directly collected without the involvement of cutting. Workers and their loads were later weighed to the nearest 0.1 mg. Both fragment length and width were measured with a field binocular to the nearest 0.01 mm. For each collected ant, day time and ambient temperature (0.01°C resolution) at the ground level beside the collecting point were precisely determined via the use of temperature Dataloggers (Tinytag, Gemini Data Loggers). The obtained data allowed the calculation of the individual transport rates (mass of vegetable tissue carried per mm and per second) at the actual foraging temperatures. Comparing carrying performance at the two foraging phases: standardization of walking speeds at 15°C {#s2c} ---------------------------------------------------------------------------------------------------- Marked differences in environmental temperature during the field measurements, which are expected to largely influence walking speed, were taken into account by analyzing the relationship between temperature and walking speed over a wide range of temperatures, in workers from a laboratory colony. These data allowed the calculation of Q~10~-values for walking speed, for temperatures ranging from 15 to 35°C, at which foraging activity actually occurred in the field. Field values of speed could therefore be transformed at a given temperature, allowing the comparison of transport rates between the foraging phases irrespective of temperature. During May--June 2009, the relationship between speed and temperature was established by recording walking speeds of laden workers, in independent assays, at 15°C, 20°C, 25°C, 30°C and 35°C. For that, an *A. heyeri* colony was collected at the same location where the field experiments were performed, transported to the Department of Behavioral Physiology and Sociobiology at the University of Würzburg, Germany, and maintained in a climatic room under 25°C and 12∶12 Light∶Dark cycle. In the laboratory, the colony was allowed to collect grasses, and after a foraging column was well established, the time spent by 50 laden workers as they walked across a 50 cm section of the 2 m long wooden bridge connecting the colony with the foraging arena was recorded, for each of the temperatures assayed. To control for the effects of the load size on speed, standardized, previously-cut fragments of the grass *Festuca rubra* (Poaceae) in the size range harvested by field colonies (9.7±1.7 mm long; 0.61±0.30 mm wide; mean±SD) were offered on the foraging arena. The dependence of walking speed on temperature was expressed by calculating the Q~10~-values for a given temperature interval, with the equation: Q~10~ = 10\*exp\[10\*(logV~t2~−logV~t1~)/(T~2~−T~1~)\], where V~t1~ and V~t2~ are the speeds at the temperatures T~1~ and T~2~, respectively. The Q~10~-values were then used to convert the walking speeds measured in the field at different temperatures during both foraging phases, to comparable values at 15°C, as follows: V~15~ = 10\*exp\[logQ~10~\*(15−T~1~)/10+logV~t1~\], where V~t1~ is the walking speed at temperature T~1~ and the Q~10~-value is that including the temperature interval 15−T~1~. Time investment before load carriage: Cutting and lift-up times {#s2d} --------------------------------------------------------------- Time investment during foraging does not only include the travel time, as indicated above, but also cutting and handling (lift-up) times. Although no direct measurements of the time spent by workers during cutting could be done in the field, grass toughness, measured as tissue area density (fragment mass divided by fragment area, in mg\*mm^−2^), is expected to largely determine the average cutting time by workers [@pone.0017667-Roces6]. Based on the tissue area density measured for the fragments collected in the field, grasses with a similar area density were offered to workers from the laboratory colony, and the time needed to cut one millimeter of tissue was recorded. Such measurements allowed the cutting time workers spent under field conditions to be indirectly, but properly estimated. In order to account for the effect of the different ambient temperatures recorded at the two foraging phases, cutting times were recorded at 15°C and 25°C in independent laboratory series. Handling (lift-up) times, i.e., how long a worker needs to bring the harvested fragment into an upright position, is expected to largely depend on fragment size, because of the torque that the worker must overcome during the collection. To investigate whether fragment size affects lift-up times, paper fragments with an average length, width and mass similar to the fragments harvested in the field were offered to workers from the laboratory colony during a foraging cycle. The time needed by foragers to lift-up the fragments from the floor to the upright position was recorded for fragment sizes that matched those of both the initial and established foraging phases. In order to stimulate workers to readily pick-up the paper fragments, they were previously impregnated with diluted orange juice (30% in water) and dried. As for cutting times, lift-up times were recorded at 15°C and 25°C in independent laboratory series. Results {#s3} ======= Traffic flow and head-on encounters at the two foraging phases {#s3a} -------------------------------------------------------------- At the beginning of the counts at sunrise, several outgoing *A. heyeri* workers were already observed on the trail. The first returning laden workers arrived at the counting point usually around 6:30AM. The number of returning workers (both laden and unladen) increased, and the number of outgoing workers decreased with day time, and the morning foraging activity ended approximately 5 hours later, around 11:30 AM. The regular counts of traffic flows on a single colony allowed the clear-cut differentiation of two foraging phases: the initial one, before 8:30 AM, in which the number of outgoing workers still surpassed that of returning laden or unladen workers, and the established foraging phase, later than 8:30 AM, in which outgoing workers decreased in number and were surpassed by the sum of laden plus unladen workers ([Fig. 1A](#pone-0017667-g001){ref-type="fig"}, lines). ::: {#pone-0017667-g001 .fig} 10.1371/journal.pone.0017667.g001 Figure 1 ::: {.caption} ###### Traffic flows of outgoing, returning laden and unladen workers in field colonies. **A**) *Lines*: average traffic flow recorded every 20 minutes in a single colony during six consecutive days. *Large symbols*: average number of workers (mean±SD) from single measurements on 12 colonies at both the initial (07:10h) and established (10:10h) foraging phases. **B**) Ratio between unladen and laden returning workers. *Line*: average ratio obtained from the counts every 20 minutes in a single colony, during six consecutive days. *Large black circles*: average ratio from single measurements on 12 colonies at both the initial (07:10h) and established (10:10h) foraging phases. Asterisks indicate a statistical difference at p\<0.001. :::  ::: During the initial foraging phase, workers were likely to return to the nest without transporting a load ([Fig. 1A](#pone-0017667-g001){ref-type="fig"}). The average number of returning unladen workers was higher than the number of laden workers during the initial phase, yet the opposite occurred at the established phase. Such a pattern was also observed in the 12 additional field colonies in which single counts at the two foraging phases were performed ([Fig. 1A](#pone-0017667-g001){ref-type="fig"}, triangles, unladen vs. laden; initial: paired t-test, t~11~ = 9.53, p\<0.001; established: paired t- test, t~11~ = 8.25, p\<0.001). The ratio between unladen and laden workers was higher at the initial phase, in which for each worker returning laden to the nest, 3 to 4 workers returned unladen ([Fig. 1B](#pone-0017667-g001){ref-type="fig"}). The ratio significantly decreased over time, with the lowest ratios at the end of the established phase, at which most workers returned laden to the nest (t-paired test, t~11~ = 17.98, p\<0.001, Log~10~ transformed data). [Figure 2](#pone-0017667-g002){ref-type="fig"} summarizes the head-on encounter rates for outgoing, returning unladen and laden workers, at both foraging phases, with an average ambient temperature of 15.55°C (±SD = 0.93, N = 12) and 26.81°C (±SD = 1.28, N = 12), respectively. At the initial one, the recorded contact rates for returning workers, laden and unladen, were higher than for outgoing workers (in average, each returning worker experienced one contact every two seconds). The inverse situation was observed at the established phase ([Fig. 2](#pone-0017667-g002){ref-type="fig"}, top; statistics at the figure caption). Since the probability of contacting a worker directly depends on the number of workers coming from the opposite direction, and this is dependent of the foraging phase (see [Fig. 1](#pone-0017667-g001){ref-type="fig"}), comparative values were obtained by dividing the contact rates (contacts\*s^−1^, [Fig. 2](#pone-0017667-g002){ref-type="fig"}, top) by the opposite traffic flow (number of workers\*s^−1^, [Fig. 1a](#pone-0017667-g001){ref-type="fig"}). After this standardization, outgoing workers at the initial phase were significantly more likely to contact a worker coming from the opposite direction than returning workers, both laden and unladen. The observed ratio, with a value of approximately 0.7, means that 70% of the returning workers experienced a contact with an outgoing worker. This difference between outgoing and returning workers vanished at the established phase ([Fig. 2](#pone-0017667-g002){ref-type="fig"}, bottom; statistics at the figure caption). ::: {#pone-0017667-g002 .fig} 10.1371/journal.pone.0017667.g002 Figure 2 ::: {.caption} ###### Interactions between outgoing and returning workers on the foraging trail. **Top**) Rate of head-on-contacts (mean±SD) with workers coming from the opposite direction for single outgoing, returning laden and unladen workers during both the initial and established foraging phases (Initial: F~2,33~ = 28.03, p\<0.001; Established: F~2,33~ = 8.23, p\<0.01; asterisks indicate a difference at p\<0.01 after a Tukey post-hoc test, Log~10~ transformed data). **Bottom**) Number of head-on-contacts divided by the number of workers in the opposite flow, for both foraging phases (Initial: F~2,33~ = 9.82, p\<0.001; Established: F~2,33~ = 0.009, p = 0.91, NS; asterisks indicate a difference at p\<0.01 after a Tukey post-hoc test, Log~10~ transformed data). :::  ::: Regarding their walking speed, outgoing workers moved at the initial phase at 19.73 mm\*s^−1^ (±SD = 3.76, N = 36), and at the established phase at 33.02 mm\*s^−1^ (±SD = 5.92, N = 36). Walking speed of returning workers was much lower at the initial phase, and similar for unladen (15.03 mm\*s^−1^ ±SD = 4.32, N = 36) and laden workers (14.81 mm\*s^−1^ ±SD = 4.51, N = 36; t-test, t~70~ = 0.21, p = 0.83, NS). At the established phase, walking speed was higher than at the initial phase, averaging 32.21 mm\*s^−1^ (±SD = 6.93, N = 36) for unladen and 31.52 mm\*s^−1^ (±SD = 7.30, N = 36) for laden ants (t-test, t~70~ = 0.41, p = 0.68, NS), being both similar to that of outgoing workers (One-way ANOVA, F~2,105~ = 0.45, p = 0.64, NS). Fragment-size determination, walking speed and carrying performance at different temperatures {#s3b} --------------------------------------------------------------------------------------------- *A. heyeri* workers cut fragments of different size depending on the foraging phase ([Fig. 3](#pone-0017667-g003){ref-type="fig"}). The length of the fragment cut was independent of the ant size, but fragments were shorter at the initial foraging phase ([Fig. 3A](#pone-0017667-g003){ref-type="fig"}, ANCOVA, F~1,268~ = 232.96, p\<0.001). In addition, the harvested fragments were narrower ([Fig. 3B](#pone-0017667-g003){ref-type="fig"}, ANCOVA, F~1,268~ = 47.81, p\<0.001) and lighter at the initial foraging phase ([Fig. 3C](#pone-0017667-g003){ref-type="fig"}, ANCOVA, F~1,296~ = 161.38, p\<0.001, Log~10~ transformed data). ::: {#pone-0017667-g003 .fig} 10.1371/journal.pone.0017667.g003 Figure 3 ::: {.caption} ###### Load-size determination by foraging workers in the field. Length (**A**), width (**B**) and mass (**C**) of the grass fragments cut by different-sized foragers in the field, at both the initial and established foraging phases. :::  ::: Tissue area density of the fragments cut during the initial phase (median: 0.085 mg\*mm^−2^, 25--75%: 0.05--0.17 mg\*mm^−2^) was significantly lower than that of the fragments cut at the established phase (median: 0.13 mg\*mm^−2^, 25--75%: 0.09--0.19 mg\*mm^−2^, ANCOVA, F~1,268~ = 8.63, p\<0.01, Log~10~ transformed data), indicating that workers selected tender grasses at the onset of foraging. To rule out the possibility that the observed differences in fragment-size determination between the two foraging phases were simply the result of workers harvesting the tips of the grass blades at the beginning of their daily foraging activity, which are likely less dense than the central and basal parts of the grass blades, the proportion of grass tips occurring in the sample of all laden workers collected was evaluated. Grass tips represented 53.2% of the loads collected at the initial phase, and 50.2% of those collected at the established phase, being these figures statistically similar (Chi-square = 0.093, p = 0.76). Environmental temperature markedly varied during the measurements at the two different foraging phases. Usually, the initial foraging phase started at values above 15°C (mean temperature 17.08°C ±SD = 1.98, N = 150), whereas the phase of established foraging took place at a mean temperature of 27.51°C (±SD = 2.6, N = 150). While at the initial phase the walking speed averaged 20.31 mm\*s^−1^ (±SD = 5.85), at the established phase it averaged 31.71 mm\*s^−1^ (±SD = 6.09). [Figure 4](#pone-0017667-g004){ref-type="fig"} shows the logarithmic relationship between walking speed of laden workers and temperature over a wide range, from 15° to 35°C, as obtained under controlled laboratory conditions ([Fig. 4](#pone-0017667-g004){ref-type="fig"}, dotted line, black triangles). When the walking speeds measured in the field at the two foraging phases are compared to the corresponding values as measured in the laboratory, laden workers were observed to walk in average 52% faster than expected for this temperature at the initial phase, yet not at the established phase ([Fig. 4](#pone-0017667-g004){ref-type="fig"}, white and black circles; expected walking speed based on logarithmic fit at 17.08°C = 13.82 mm\*s^−1^; expected at 27.51°C = 32.19 mm\*s^−1^). ::: {#pone-0017667-g004 .fig} 10.1371/journal.pone.0017667.g004 Figure 4 ::: {.caption} ###### Average walking speed (±SD) of laden workers as a function of temperature, measured in the laboratory (*dotted line*: walking speed = −95.56+88.75\*(log~10~T), R^2^ = 0.89). The black and white circles represent the average walking speeds (±SD) measured in the field at the indicated mean (±SD) environmental temperatures, for both the initial and the established foraging phases. :::  ::: To allow comparisons of transport rates at different temperatures, field data on walking speed were transformed at 15°C using the relationship between walking speed and temperature obtained for laden workers in the laboratory ([Fig. 4](#pone-0017667-g004){ref-type="fig"}). Due to the observed logarithmic relationship, the fractional increase in walking speed for a temperature increase of 10°C, i.e., the Q~10~-value, was dependent of the temperature range considered. The fractional increase in walking speed for a temperature increase of 10°C is described by the equation: Q~10~ = 10\*exp\[10\*(logV~t2~−log V~t1~)/(T~2~−T~1~)\]. The Q~10~-values decreased with increasing temperatures, being Q~10~ = 4.1 for the range 15°--20°C, 2.3 for the range 20°--25°C, 1.5 for the range 25°--30°C, and 1.4 for the range 30°--35°C. In addition to temperature, the size of the load carried is known to markedly influence walking speed of laden workers. To rule out the possibility that the observed higher walking speeds at the initial phase ([Fig. 4](#pone-0017667-g004){ref-type="fig"}) resulted from the smaller loads carried as compared to the established phase ([Fig. 3](#pone-0017667-g003){ref-type="fig"}), load mass was standardized regarding the size of the ant that carry it, as follows: Loading Ratio = \[Load mass+Ant mass\]/Ant mass [@pone.0017667-Rissing1]. [Figure 5](#pone-0017667-g005){ref-type="fig"} shows the relationship between walking speed and loading ratio for workers at the two foraging phases. Data collected at 15°C during the initial phase are presented as obtained; data collected at other temperatures were transformed to 15°C using the obtained Q~10~-values (see above), to allow comparisons at the same temperature. The average loading ratio was significantly lower at the initial phase (F-test, F~1,297~ = 167.94, p\<0.001, Log~10~ transformed data), and walking speed (WS) negatively depended on the loading ratio (LR): WS = 13.39--1.64\*LR (R^2^ = 0.14, p\<0.001). As a consequence, harvesting small grass fragments at the initial phase allowed workers to return faster to the nest, as compared to the established phase. Furthermore, if workers carrying similar loads relative to their body masses are compared, i.e., workers showing similar loading ratios ([Fig. 5](#pone-0017667-g005){ref-type="fig"}, inset), results indicated that they returned to the nest at the initial phase faster than at the established phase, irrespective of their loads (statistics at the figure caption). ::: {#pone-0017667-g005 .fig} 10.1371/journal.pone.0017667.g005 Figure 5 ::: {.caption} ###### Walking speed of laden workers as a function of the Loading Ratio (LR = \[Load mass+Ant mass\]/Ant mass), for both the initial and established foraging phases. Speeds measured at temperatures other than 15°C, during both foraging phases, were transformed to comparable values at 15°C using the Q~10~-values obtained from the speed measurements in the laboratory, as shown in [Figure 4](#pone-0017667-g004){ref-type="fig"} (further details in text). Walking speed of workers walking at 15°C (±0.1°C) during the initial foraging phase remained untransformed (gray triangles). **Inset**: Detailed view of the mean walking speeds for different ranges of loading ratios: 1.2--1.3; \>1.3--1.4; \>1.4--1.5. Asterisks denote statistical differences between the initial and established phase at p\<0.001 (for the loading ratios values 1.2--1.3: t~32~ = 5.41; \>1.3--1.4: t~31~ = 3.70; \>1.4--1.5: t~31~ = 4.17). :::  ::: Data of temperature-corrected walking speed (at 15°C) were used to calculate the gross transport rates of vegetable tissue, a measure of individual performance, for laden workers at both the initial and the established foraging phases. For the two foraging phases, there was a positive relationship between gross transport rate and ant body mass ([Fig. 6](#pone-0017667-g006){ref-type="fig"}). The foraging phase had a very significant effect on the relationship between ant body mass and gross transport rate (ANCOVA, F~1,295~ = 133.68, p\<0.001). Workers showed an average gross transport rate of 8.26 mg\*mm\*s^−1^ (±SD = 1.42, N = 148) at the initial foraging phase, lower than later at the established foraging phase, with an average of 22.61 mg\*mm\*s^−1^ (±SD = 26.16, N = 150). ::: {#pone-0017667-g006 .fig} 10.1371/journal.pone.0017667.g006 Figure 6 ::: {.caption} ###### Gross transport rate of plant tissue (expressed as mass carried per mm and per second) as a function of ant body mass, for both the initial and established foraging phases. Data were Q~10~-transformed at 15°C (further details in text). :::  ::: Cutting and lift-up times depending on the foraging phase {#s3c} --------------------------------------------------------- In the laboratory, the time needed to cut a millimeter of *Festuca* grass tissue, with a density of 0.11 mg\*mm^−2^, averaged 77.12 s (±SD = 35.69, N = 19) at a temperature similar to the average value of the initial phase (15°C), and 25.27 s (±SD = 8.82, N = 19) at a temperature of 25°C, as in the established phase, being these values significantly different (t-test, t~36~ = 6.21, p\<0.001, Log~10~ transformed data). The time needed to lift-up a paper fragment similar in size to those cut in the initial phase ([Fig. 3](#pone-0017667-g003){ref-type="fig"}, 7 mm length, 1 mm width, 0.9 mg) averaged 41.91 s (±SD = 18.60, N = 30) at 15°C, the mean temperature at the initial foraging phase in the field. For a paper fragment as harvested in the established phase ([Fig. 3](#pone-0017667-g003){ref-type="fig"}, 14 mm length, 1 mm width, 2.2 mg), lift-up times were in average much longer, 83.62 s (±SD = 57.58, N = 30; t-test, t~58~ = 4.49, p\<0.001, Log~10~ transformed data). At 25°C, the time needed to lift-up a paper fragment similar in size to that cut at the established phase averaged 37.02 s (±SD = 23.37, N = 30), while for a fragment similar in size to that cut at the initial phase, the lift-up time was significantly shorter, averaging 15.21 s (±SD = 6.57, N = 30; t-test, t~58~ = 5.82, p\<0.001, Log~10~ transformed data). Cutting, loading and carriage: overall returning times by laden and unladen workers {#s3d} ----------------------------------------------------------------------------------- Taken into account the cutting times, handling times and walking speeds, the overall returning time at the *initial foraging phase* can be calculated for *i)* unladen workers, which do not spend any cutting or handling times, *ii)* workers carrying loads as they select at the initial foraging phase, and *iii)* workers that would not decide to select smaller loads as expected, but harvest larger loads as it occurs in the *established foraging phase*. Calculations are based on a returning trip along 20 meters, which is the average *A. heyeri*\'s trail length (personal observations). Since the average walking speed of unladen and laden workers did not differ, the returning time of *unladen* workers, based on the speed of laden workers ([Fig. 5](#pone-0017667-g005){ref-type="fig"}, black circles), would average 26 min 40 s. For *laden* workers carrying small loads as observed in the initial phase at 15°C, both the time needed to cut a grass of 0.68 mm width ([Fig. 3B](#pone-0017667-g003){ref-type="fig"}), and to lift-up the resulting fragment of 7 mm length ([Fig. 3A](#pone-0017667-g003){ref-type="fig"}) should be added, resulting in a returning time of 28 min 35 s. For *laden* workers that would not decide to cut a small fragment at 15°C, but a larger one as in the established phase, both the time needed to cut a grass of 0.98 mm width ([Fig. 3B](#pone-0017667-g003){ref-type="fig"}), and to lift-up the resulting fragment of 14 mm length ([Fig. 3A](#pone-0017667-g003){ref-type="fig"}) have to be added, as well as the additional travel time resulting from the carriage of a larger load at lower speed ([Fig. 5](#pone-0017667-g005){ref-type="fig"}, white circles). The returning time for those workers would average 38 min 22 s. Taken together, *A. heyeri* workers saved between 30% and 26% of their time by returning unladen or with smaller loads as observed at the initial foraging phase, respectively. Discussion {#s4} ========== Foraging phases, contact rates, and information transfer {#s4a} -------------------------------------------------------- Leaf-cutting ants initiate their daily foraging activity by leaving the nest in large numbers without the need of recruitment [@pone.0017667-Lutz1], as also observed in the investigated field colonies ([Fig. 1](#pone-0017667-g001){ref-type="fig"}). Foragers search for suitable resources and, upon discovery, decide whether a given resource is worth communicating to others. The acceptance of a plant is mainly based on chemical and physical features of its leaves [@pone.0017667-Howard1], and also on the workers\' foraging experience [@pone.0017667-Saverschek1]. If the source is suitable, workers return to the nest (laden or unladen) laying a chemical trail, and interact with nestmates both on the trail and inside the nest [@pone.0017667-Roces1], [@pone.0017667-Roces3], [@pone.0017667-Jaff1]. Our field measurements showed that at the beginning of their daily foraging activity, *A. heyeri* workers had a greater tendency to return to the nest unladen. The proportion of unladen workers diminished over time, and attained the lowest values at the time when the traffic flows to and from the nest reached roughly similar values. Why a large proportion of workers, at the initial foraging phase, returned to the nest unladen? It is known that over an entire foraging process, this proportion may be high, varying between 13% and 75% [@pone.0017667-Cherrett1], [@pone.0017667-Hodgson1]--[@pone.0017667-Lewis1]. While a number of these unladen workers can be involved in trail-clearing [@pone.0017667-Daguerre1] or transport of plant sap [@pone.0017667-Stradling1], others are engaged in the reinforcement of the chemical trail or in a combination of food transport and recruitment communication [@pone.0017667-Jaff1], [@pone.0017667-Jaff2]. The observed interactions between outgoing and returning workers, and published results [@pone.0017667-Roces2], [@pone.0017667-Roces6], provide indirect support to the idea that the decision to return unladen to the nest, with the concomitant time savings, represents a response to the high needs for information at the beginning of a daily foraging process. Under the field conditions of our study, it was unfeasible to quantify the intensity of recruitment communication, so that direct evidence for information transfer is lacking. Returning workers were not necessary to stimulate workers to leave the nest, which already occurred in large numbers on the trail at the time the first workers returned to the nest ([Fig. 1](#pone-0017667-g001){ref-type="fig"}) [@pone.0017667-Lutz1]. We suggest that returning workers, through their interactions with outgoing workers on the trail and the expected trail-marking, are responsible for the initial establishment of a foraging process. Outgoing workers, after contacting returning workers and following the chemical trail, are expected to arrive shortly thereafter at the discovered source, and make foraging decisions (flexible fragment-size determination, trail-laying, further contacts) that amplify and help maintain the foraging process already initiated. Head-on encounters between workers moving in opposite directions were very frequent, dependent on the foraging phase, and different for outgoing and returning workers, even after standardization to the traffic flow. This clearly indicates that such contacts did not simply represent collisions in a probabilistic sense. Interestingly, outgoing workers at the initial phase were significantly more likely to contact a worker coming from the opposite direction than returning workers, both laden and unladen. Counts indicated that each outgoing worker contacted in average 70% of the returning workers at the initial phase, which suggest an active search for contacts and/or information, and only 20% at the established phase. Such an active search was described for *Atta cephalotes* as early as 1929 by Lutz, who observed that "frequently, a returning laden forager is stopped momentarily by an outgoing nestmate which is apparently interested in what is being carried" [@pone.0017667-Lutz1]. Returning foragers, on the contrary, showed similar contact rates at the two foraging phases. It appears that they do not actively search for contacts, but just return as fast as possible to the nest, likely laying a chemical trail [@pone.0017667-Jaff1]. Outgoing *Atta cephalotes* workers are also known to experience higher encounter rates than returning ants [@pone.0017667-Burd2], and it has been speculated that they may actively seek encounters with returning ants for information acquisition. The present results go beyond those findings by showing that the contact rates experienced by outgoing workers depended on the time since the onset of foraging, being initially very high, and dropping drastically when a foraging column was already established. It can be argued that the active search for interactions in outgoing workers represents an indirect measure of the workers\' information needs, which is strongly phase-dependent. Even though we were unable to establish whether such a contact indeed involved information transfer or not, it is known that foraging decisions of outgoing workers, and their probability to find a recently-discovered food source, are influenced by the interactions with returning nestmates and the odour of the fragment they carry [@pone.0017667-Roces3], [@pone.0017667-Howard2]--[@pone.0017667-FarjiBrenner1]. In fact, there is a large body of evidence, particularly for harvester ants, emphasizing the role of ant encounters for the regulation of foraging activity [@pone.0017667-Detrain1], [@pone.0017667-Gordon2]--[@pone.0017667-Schafer1]. Even the mere fact of an encounter may provide information, such as the magnitude of the colony\'s foraging activity, and therefore influence the probability of food collection in ants [@pone.0017667-TorresContreras1]--[@pone.0017667-Gordon5]. Fragment-size determination: individual vs. social demands {#s4b} ---------------------------------------------------------- *A. heyeri* foragers cut grass fragments of different size depending on the time since the onset of their foraging activity. Fragments were shorter, narrower and lighter at the initial foraging phase, and workers selected in addition tender grass blades for harvesting, i.e., grasses with lower area density. Considering that load size has a large influence on maneuverability and speed of transport and, therefore, on the rate of plant tissue delivery [@pone.0017667-Rudolph1], [@pone.0017667-Rschard2], [@pone.0017667-Moll1], [@pone.0017667-Burd3], foragers should *a priori* be expected to carry loads that maximize their individual delivery rate, i.e., the amount of plant tissue carried to the nest per unit of foraging time. Do *A. heyeri* workers perform different as individual carriers, depending on the foraging phase? The observed differences in environmental temperature during both foraging phases, which markedly influence locomotion speed of ectothermic animals like ants [@pone.0017667-Shapley1], preclude a direct comparison of individual transport rates between phases. By converting the absolute velocities to standardized values at 15°C (using Q~10~-values for walking speed measured in the laboratory), individual transport rates at both the initial and established foraging phases could be properly compared. Our results showed that even though workers at the initial foraging phase walked at a significant faster pace than expected for this temperature ([Fig. 4](#pone-0017667-g004){ref-type="fig"} and [Fig. 5](#pone-0017667-g005){ref-type="fig"}), their individual transport rate was markedly low, averaging 37% of the rate observed at the established foraging phase ([Fig. 6](#pone-0017667-g006){ref-type="fig"}), as known for the leaf-cutting ant *Atta cephalotes* under controlled laboratory conditions [@pone.0017667-Roces6]. Interestingly, laden workers walked at a fast pace despite their rate of encounters with outgoing workers, which moved in large numbers at the beginning, suggesting that returning workers experienced no significant delays on their way caused by these encounters. A worker\'s foraging cycle: when is time worth saving? {#s4c} ------------------------------------------------------ The observed differences in fragment-size determination at the two different foraging phases have a number of consequences on both individual performance and colony-wide patterns. At the individual level, cutting length and therefore cutting time was shorter at the initial foraging phase, because of the selection of narrow grasses. The time needed to lift-up the selected fragments was also shorter than the time needed for a larger fragment, as selected at the established phase. And, in addition, workers were able to walk faster because of the transport of lighter (shorter) grass fragments. Plasticity in workers\' cutting rules at the initial phase, compared to the alternative of cutting larger fragments as in the established phase, led therefore to a reduction of time costs of around 30%. At the colony level, such time savings may allow a higher frequency of individual roundtrips during a daily foraging process, with the resulting increase in the probability of interacting with outgoing workers for information transfer. As mentioned above, the foraging phase did not only influence cutting behaviour, but walking speed of laden foragers ([Fig. 4](#pone-0017667-g004){ref-type="fig"}). By comparing the walking speeds at the two foraging phases with the speeds measured in the laboratory at the same temperatures, workers were observed to walk faster than expected at the initial phase, yet not at the established phase. Moreover, when controlling for differences in the fragment-size carried, workers still walked faster at the initial phase than at the established phase ([Fig. 5](#pone-0017667-g005){ref-type="fig"}, inset). It follows that beyond the expected physiological effects of temperature on ant locomotion [@pone.0017667-Shapley1], [@pone.0017667-Barnes1], an additional drive motivates workers at the initial foraging phase to increase their walking speed, as described for a related species [@pone.0017667-Roces1], [@pone.0017667-Roces2]. It is an open question how outgoing workers identify the actual foraging phase, so as to adjust their behaviour accordingly. Workers may acquire information via the rate of contacts they experience, the probability of encountering laden workers on their way to the food source, or respond to potential differences in the intensity of pheromone marking on the trail. Why do laden workers walk faster at the initial foraging phase, but carry less material than later at the established phase? The adaptive value of this response, with the associated time savings, might be related to the high information needs at the time of a food discovery, thus allowing both a faster establishment and the maintenance of a foraging process. We argue that at the initial phase of trail development, saving time would be of great importance to monopolize a food source as soon as possible. In fact, the extreme strategy of time-saving at the source for information transfer is that described for the leaf-cutting ant *Atta cephalotes*: scout workers, upon discovery of a newly-food source, cut no fragments at all but return quickly to the nest laying chemical trails, and only start cutting leaves when a foraging column has been established [@pone.0017667-Jaff1]. Whether information needs also influences load-size determination under routine foraging conditions on well-established trails, is unclear [@pone.0017667-Burd4]. In conclusion, our results support the hypothesis that at the initial foraging phase, workers compromise their individual transport rates of material in order to return early to the colony, as predicted by the information-transfer hypothesis [@pone.0017667-Nez2], [@pone.0017667-Roces2], [@pone.0017667-Roces7]. The selection of partial loads at the beginning of foraging, as well as the decision to return to the nest unladen, appear to be driven by the need of a rapid information transfer at the initial phase of the monopolization of resources [@pone.0017667-Roces6], beneficial in an environment where the foraging time windows are constrained by environmental variables [@pone.0017667-Lewis1], and neighbouring colonies may compete for resources [@pone.0017667-Whitehouse1]. MB dedicates this work to Silvia Cardozo, María B. Sosa and Roberto Bollazzi for their unconditional support, help and patience over the years, and is indebted to Alba B. Pereira and Carlos S. Carbonell for their help during the early study times. The authors deeply thank Prof. Dr. Josué A. Núñez, University of Buenos Aires, Argentina, for his deep insights into social insect foraging behavior, and for lively discussions over the years. They also thank Nicole Saverschek for helpful suggestions, the Academic Editor Ryan Earley, and three anonymous reviewers for comments that improved the manuscript. Thanks are also due to Andreas Leimbach, Dirk Ländle and Oliver Berberich for help during data collection in the laboratory, and to Annette Laudahn and Adrienne Gerber-Kurz for the care of the *A. heyeri* colony. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by the German Academic Exchange Service (DAAD, fellowship granted to MB) and the German Research Foundation (DFG - SFB 567 "Mechanisms of Interspecific Interactions of Organisms"). The publications costs were covered by the German Research Foundation (DFG) in the funding programme Open Access Publishing. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MB FR. Performed the experiments: MB. Analyzed the data: MB FR. Contributed reagents/materials/analysis tools: MB FR. Wrote the paper: FR MB.

PubMed Central

2024-06-05T04:04:19.240312

2011-3-9

PMC3052371

Introduction {#s1} ============ In temperate regions, wheat (*Triticum spp.*) and barley (*Hordeum vulgare*) can be sown in autumn to grow vegetatively through winter before flowering in spring. Autumn sowing can enhance yield relative to later sowing times, but can also expose plants to freezing winter conditions [@pone.0017900-Entz1]. Consequently, the capacity to survive winter frosts is an important trait for autumn-sown wheat and barley varieties grown in regions that experience cold winters [@pone.0017900-King1]--[@pone.0017900-Koemel1]. Tolerance to winter frosts is established through cold acclimation, the process where freezing tolerance increases as temperatures decrease during autumn [@pone.0017900-Thomashow1]. Molecular analyses have identified low-temperature responsive genes that are induced during cold acclimation, such as ice crystallisation inhibitors and dehydrins, which protect against freezing damage [@pone.0017900-Thomashow1]. *C-REPEAT BINDING FACTOR* (*CBF*) genes encode transcription factors that play a critical role in the cold acclimation process [@pone.0017900-Thomashow1]. *CBF* genes are rapidly induced by low temperatures to activate genes that contribute to increased freezing tolerance [@pone.0017900-Thomashow1]. The *FROST RESISTANCE 2* (*FR2*) locus on chromosome 5A of wheat and barley (5H), which is associated with variation in frost tolerance [@pone.0017900-Roberts1]--[@pone.0017900-Fowler1], has been mapped to a cluster of *CBF* genes [@pone.0017900-Francia1]--[@pone.0017900-Knox1], reviewed in [@pone.0017900-Galiba1]. An increased number of *CBF* genes at the *FR2* locus might enhance cold acclimation in frost tolerant varieties [@pone.0017900-Knox2]. Another feature of many autumn-sown wheats and barleys is the vernalization requirement; the requirement for prolonged exposure to cold to make plants competent to flower. The requirement for vernalization delays reproductive growth and stem elongation before winter, minimising the risk of frost damage to cold-sensitive reproductive organs. Furthermore, since the capacity for cold acclimation decreases during reproductive growth, the requirement for vernalization allows more time for cold acclimation by lengthening the vegetative growth phase [@pone.0017900-Limin2]. Consistent with this hypothesis, varieties sown in autumn in regions that experience cold winters typically have a strong requirement for vernalization and can acclimate to cold over an extended period. In comparison, varieties that flower without vernalization have a shorter vegetative growth phase and less time for cold acclimation, and are consequently less frost tolerant [@pone.0017900-Limin1], [@pone.0017900-Koemel1], [@pone.0017900-Limin2]. During winter, low temperatures overcome the vernalization requirement and trigger a quantitative flowering response, so that longer exposure to cold causes more rapid flowering until the vernalization response is saturated after several weeks at low-temperatures [@pone.0017900-Gassner1]--[@pone.0017900-Chouard1]. The initial response to prolonged cold can be separated from the flowering response; for example, when sprouting seeds are vernalized, inflorescence development (flowering) does not begin until plants are shifted to normal growth temperatures [@pone.0017900-Purvis1], [@pone.0017900-Sasani1]. This implies a memory of cold that mediates a quantitative change in the rate of development after vernalization. In cereals, vernalization-induced flowering is mediated by the activation of *VERNALIZATION1* (*VRN1*), a gene that promotes flowering (reviewed in [@pone.0017900-Trevaskis1], [@pone.0017900-Distelfeld1]). *VRN1* transcript levels show a quantitative response to cold, with longer durations of cold activating *VRN1* expression to greater extents, and expression of *VRN1* is maintained when vernalized plants are shifted to normal growth temperatures [@pone.0017900-Trevaskis2]--[@pone.0017900-Yan1]. This long term activation of *VRN1* might be mediated through vernalization-induced changes in the state of chromatin at the *VRN1* locus [@pone.0017900-Oliver1]. Following vernalization, *VRN1* accelerates flowering by promoting the transition to reproductive development at the shoot apex and by making the leaves competent to respond to increasing daylength during spring, which accelerates inflorescence development and stem elongation [@pone.0017900-Purvis1], [@pone.0017900-Sasani1], [@pone.0017900-Hemming1]. Many varieties of wheat and barley carry alleles of *VRN1* that are expressed at elevated levels without prior cold treatment [@pone.0017900-Trevaskis2]--[@pone.0017900-Yan1], [@pone.0017900-Takahashi1]--[@pone.0017900-Hemming2]. These varieties flower without vernalization and can be grown in warm climates, where winter temperatures are not cold enough for vernalization, or sown in spring. By accelerating the transition to reproductive growth, when cold acclimation is inhibited, active alleles of *VRN1* are likely to limit the potential for cold acclimation and reduce frost tolerance [@pone.0017900-Limin2]. Consistent with this hypothesis, *VRN1* is linked to the *FR1* locus, which influences the activity of cold acclimation pathways and frost tolerance [@pone.0017900-Roberts1], [@pone.0017900-Hayes1], [@pone.0017900-Francia1], [@pone.0017900-Sutka1]--[@pone.0017900-Stockinger1]. Furthermore, deletion of the *VRN1* region increases the activity of cold acclimation pathways, and frost tolerance, in the *maintained vegetative phase mutant* (*mvp*) of *Triticum monococcum* [@pone.0017900-Dhillon1], which grows vegetatively indefinitely [@pone.0017900-Shitsukawa1]. These observations suggest that *VRN1* and *FR1* are likely the same gene [@pone.0017900-Dhillon1]. The impact of the *VRN1* deletion or active *VRN1* alleles on frost tolerance is greatest in long-days, where floral development occurs rapidly [@pone.0017900-Limin2], [@pone.0017900-Dhillon1]. Cold acclimation and vernalization occur concomitantly when autumn-sown plants experience low-temperatures during winter, but vernalization continues to influence development when plants return to warm temperatures. In this study we use microarrays to compare and contrast gene expression in seedlings of a vernalization-responsive barley (cv Sonja) exposed to short or prolonged cold treatments. Additionally, we compare gene expression in the leaves of vernalized versus non-vernalized plants. By comparing the effects of short and prolonged cold on the transcriptome we identify genes that show distinct low temperature responses. We were also able to identify genes that show lasting responses to prolonged cold treatment, an expression pattern that defines vernalization-responsive genes. Possible roles for these genes in the vernalization response of cereals are discussed. Results {#s2} ======= The Affymetrix 22K Barley1 chip [@pone.0017900-Close1] was used to examine the effects of short or prolonged cold on the transcriptome of barley seedlings (see [materials and methods](#s4){ref-type="sec"}, [Figure 1A](#pone-0017900-g001){ref-type="fig"}). Pair wise comparisons of the different treatments were used to identify contigs that have altered expression in the short term, prolonged or post-cold treatments ([Tables S1](#pone.0017900.s002){ref-type="supplementary-material"}, [S2](#pone.0017900.s003){ref-type="supplementary-material"}, [S3](#pone.0017900.s004){ref-type="supplementary-material"}, [S4](#pone.0017900.s005){ref-type="supplementary-material"}, [S5](#pone.0017900.s006){ref-type="supplementary-material"}, [S6](#pone.0017900.s007){ref-type="supplementary-material"}). Large numbers of contigs show altered expression (\> two fold change relative to the non-treated control, p\<0.01) in short (613 contigs, [Table S1](#pone.0017900.s002){ref-type="supplementary-material"}) or prolonged (786 contigs, [Table S2](#pone.0017900.s003){ref-type="supplementary-material"}) cold treatment ([Figure 2](#pone-0017900-g002){ref-type="fig"}). The contigs that showed altered expression after short term cold were predominantly up-regulated, whereas many contigs were down-regulated in the prolonged cold treatment ([Figure 2](#pone-0017900-g002){ref-type="fig"}). Fewer contigs showed altered expression in the post cold treatment relative to the non-treated control (150 contigs, [Table S3](#pone.0017900.s004){ref-type="supplementary-material"}). A limited number of contigs showed a lasting response to prolonged cold; i.e. significantly changed in both the prolonged cold and 1 day after prolonged cold samples, but not the short term cold sample (60 contigs) ([Figure 3](#pone-0017900-g003){ref-type="fig"}). Expression levels were verified for a subset of contigs with contrasting expression patterns, using quantitative RT-PCR ([Figure S1](#pone.0017900.s001){ref-type="supplementary-material"}). ::: {#pone-0017900-g001 .fig} 10.1371/journal.pone.0017900.g001 Figure 1 ::: {.caption} ###### Overview of sampling methods for microarray analysis. A\) Barley seeds (cv. Sonja) were germinated and grown in darkness at either 20°C over 5 days (control) or 4°C over 49 days (prolonged cold). Seedlings were then shifted from the control treatment to 4°C for 24 hours (short term cold) or shifted from the prolonged cold treatment to 20°C for 24 hours (post cold). In all treatments the shoot apex remained at an early stage of vegetative development, but plants grown from seedlings that experienced prolonged cold flower rapidly when shifted to normal growth conditions, unlike control seedlings germinated at 20°C [@pone.0017900-Sasani1]. B) To identify contigs that show a sustained response to prolonged cold, barley seeds were germinated in the dark at 4°C for 49 days and then transferred to growth in glasshouse conditions until they reached the three leaf stage (10 days after the end of cold treatment). Non-vernalized control plants were grown simultaneously under the same conditions and were sampled at the equivalent developmental leaf stage (14 days). Grey shading indicates low-temperature. :::  ::: ::: {#pone-0017900-g002 .fig} 10.1371/journal.pone.0017900.g002 Figure 2 ::: {.caption} ###### Changes in gene expression in seedlings after different low-temperature treatments. A\) The total numbers of contigs showing differential expression (\>2 fold change, p\<0.01) in the short term cold versus control (ST-C), the prolonged cold versus control (PLC-C) or the post cold versus control (POC-C) samples. B) The total numbers of genes up-regulated in the same comparisons. C) Genes down-regulated in the same comparisons. D) The total number of contigs up or down-regulated in the prolonged versus short term cold treatments. E) The number of contigs up or down-regulated in the post versus prolonged cold treatments. Dark shading indicates 8 fold change of greater, intermediate shading indicates 4--8 fold, light shading indicates 2--4 fold change. :::  ::: ::: {#pone-0017900-g003 .fig} 10.1371/journal.pone.0017900.g003 Figure 3 ::: {.caption} ###### Venn diagram showing the differentially expressed contigs across the different treatments. A summary of the contigs that showed a two fold or greater change in transcript levels across the different treatments when compared to the control treatment (p\<0.1). Shaded area indicates contigs that were significantly changed in the samples treated with prolonged cold and one day after prolonged cold treatment, relative to the control. :::  ::: Principal component analysis (PCA) [@pone.0017900-Wit1] was used to visualise the overall changes in gene expression in the different treatments ([Figure 4](#pone-0017900-g004){ref-type="fig"}). The first two principal components, which explained 97% of the total variance (95% 1^st^ component, 1.8% second component), show that the short term and prolonged cold treatments were dissimilar to each other and to the control treatment ([Figure 4](#pone-0017900-g004){ref-type="fig"}). In comparison the post cold treatment was more similar to the control treatment than to either the short or prolonged cold treatments. Replicates showed a high degree of similarity for all four treatments ([Figure 4](#pone-0017900-g004){ref-type="fig"}). The results of PCA analysis are consistent with the number of contigs that show altered expression in each treatment, relative to the control ([Figures 2](#pone-0017900-g002){ref-type="fig"}, [3](#pone-0017900-g003){ref-type="fig"}, [Table S1](#pone.0017900.s002){ref-type="supplementary-material"}, [S2](#pone.0017900.s003){ref-type="supplementary-material"}, [S3](#pone.0017900.s004){ref-type="supplementary-material"}, [S4](#pone.0017900.s005){ref-type="supplementary-material"}, [S5](#pone.0017900.s006){ref-type="supplementary-material"}, [S6](#pone.0017900.s007){ref-type="supplementary-material"}). Taken together, these observations indicate that prolonged cold treatment has a strong influence on the transcriptome of barley, distinct to that of short term cold treatment, but the majority of cold-induced alterations to the transcriptome are not maintained when plants are shifted to warm conditions after prolonged cold treatment. ::: {#pone-0017900-g004 .fig} 10.1371/journal.pone.0017900.g004 Figure 4 ::: {.caption} ###### Principle component analysis of microarray data. Principal component analysis was applied to differentially expressed contigs identified in the different seedling treatments (see [methods](#s4){ref-type="sec"}). Closed triangles (▴) indicate samples from the no cold control treatment. Open triangles (**Δ**) represent the 1 day after prolonged cold treatment. Closed squares (▪) represent the short term cold treatment. Open squares (□) represent the prolonged cold treatment. :::  ::: K-means cluster analysis {#s2a} ------------------------ Cluster analysis was used to analyse the expression patterns of contigs that showed significant changes in expression (p\<0.01) in any of the two way comparisons between seedling treatments. Ten primary clusters were identified; each showing distinctive expression patterns ([Figure 5](#pone-0017900-g005){ref-type="fig"}, ). ::: {#pone-0017900-g005 .fig} 10.1371/journal.pone.0017900.g005 Figure 5 ::: {.caption} ###### K-means cluster analysis of differentially expressed contigs. The mean for each cluster is shown as black dots and lines and the gray lines represent the expression pattern of individual contigs. The three replicates for each treatment data are shown. :::  ::: A cluster of 114 contigs showed elevated expression at low-temperature, irrespective of the duration ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 1). Contigs belonging to this cluster correspond to cold acclimation genes, such as *DEHYDRIN5* (*DHN5*) (contig1717\_s\_at), *DELTA1-PYRROLINE-5-CARBOXYLATE SYNTHASE 1* (*P5CS1*) (contig3814\_at) and *GALACTINOL SYNTHASE* (contig3810\_at and contig3811\_at) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}; contig sequences are available at [www.plexdb.org](http://www.plexdb.org), see Barley Annotation [@pone.0017900-Wise1]). A second cluster of 98 contigs (cluster 2) showed an inverse expression pattern, and is made up of contigs that are down-regulated by low-temperatures ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 2). This cluster includes contigs corresponding to heat-shock genes (contig2006\_s\_at and contig5597\_s\_at) and a *PROLINE OXIDASE* (contig68\_at) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). A group of 85 contigs showed increased expression only in the short term cold treatment ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 3). This cluster includes several contigs annotated as transcription factors, examples include: basic leucine zipper domain, zinc finger domain, WRKY and NAC domain transcription factors. There were also several contigs annotated as *GLUTATHIONE-S-TRANSFERASE* genes ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Another cluster showed the inverse pattern and was made up of 122 contigs ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 4). This cluster includes contigs annotated as core histone domain containing proteins (contig175\_at and contig175\_x\_at) and basic helix-loop-helix transcription factors (contig4559\_s\_at, contig4560\_at, contig4560\_x\_at and contig26382\_at) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Clusters 3 and 4 define contigs that respond to short term cold exposure or "cold shock". A group of 47 contigs showed elevated expression in the short term cold treatment, the prolonged cold treatment and the post cold sample ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 5). Contigs belonging to this cluster include several contigs corresponding to cold acclimation genes ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Twenty three contigs showed elevated expression only in the post cold samples ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 6). This cluster includes heat-shock genes (contig2004\_s\_at and contig2007\_s\_at) and auxin or jasmonate responsive genes (contig17690\_at and contig2906\_at, HVSMEg0005M23r2\_at respectively) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Contigs that showed elevated expression only in the prolonged cold treatment were grouped in cluster 7 (138 contigs) ([Figure 5](#pone-0017900-g005){ref-type="fig"}). This cluster includes contigs corresponding to *FLOWERING LOCUS T-like2* (HVSMEl0003G02r2\_at) and an *APETALA2-like* gene (contig18652\_at) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Conversely, contigs that had decreased expression in prolonged cold treatment were grouped into cluster 8 (217 contigs) ([Figure 5](#pone-0017900-g005){ref-type="fig"}). Examples from this cluster include contigs described as zinc finger transcription factors (contig4486\_at and contig8233\_s\_at) and cysteine proteases (HB26O11r\_at, HVSMEl0003G02r2\_at and contig11505\_at) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Of most interest to the aims of this study were contigs that showed a lasting response to prolonged cold. A cluster of 46 contigs showed increased expression in the prolonged cold and post cold samples ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 9). This cluster includes *VRN1* (rbaal14f06\_s\_at), contigs corresponding to *23kd JASMONATE INDUCED* genes (rbags15p13\_s\_at and contig1679\_s\_at), a putative glucan synthase (contig19065\_at) and a calcium binding EF-hand protein (AJ250283\_at) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Another cluster of 73 contigs showed the inverse pattern, with decreased expression in the prolonged cold and post cold samples ([Figure 5](#pone-0017900-g005){ref-type="fig"}, cluster 10). This group includes contigs corresponding to *ODDSOC2* (*HvOS2*) (contig12031\_at), rubisco activase (contig1019\_at) and RNase S-like proteins (contig5059\_s\_at and contig5058\_x\_at) ([Table S7](#pone.0017900.s008){ref-type="supplementary-material"}). Vernalization-responsive genes {#s2b} ------------------------------ Gene expression was assayed in the fully expanded second leaf of vernalized or non-vernalized plants using the Affymetrix 22K Barley1 chip [@pone.0017900-Close1] (see [materials and methods](#s4){ref-type="sec"}, [Figure 1B](#pone-0017900-g001){ref-type="fig"}). This allowed comparisons between vernalized and non-vernalized plants to be made using developmentally equivalent tissues, which cannot be made if apex tissue is included (vegetative in non-vernalized versus reproductive in vernalized plants at this growth stage, [@pone.0017900-Sasani1]). A total of 60 contigs showed greater than two fold change in expression level (p\<0.01) in vernalized versus non-vernalized leaves. A less stringent criteria identified 244 contigs that showed greater than 1.5 fold changes in expression level (p\<0.05) ([Table S8](#pone.0017900.s009){ref-type="supplementary-material"}). Of these, 120 were up-regulated after vernalization, including contigs corresponding to *VRN1* (rbaal14f06\_s\_at), *COR14b* (HVSMEa0015E13r2\_s\_at) and a *JUMONJI* transcription factor (contig24321\_at). A total of 128 contigs showed lower expression levels after vernalization, including contigs corresponding to *XYLOGLUCAN ENDOTRANSGLYCOSYLASE* (*XET*) (HVSMEb0004L16r2\_at, contig2673\_at and contig2670\_x\_at), *HvOS2* (contig12031\_at) and *CBF9* (HVSMEn0019L21f\_at). The prolonged cold treatment sample (see above) corresponds to the end of the vernalization treatment (49 days at 4°C). Of the contigs that showed altered expression in the leaves of plants after vernalization, 14 showed altered expression in both the prolonged cold and 1 day after prolonged cold seedling treatments (clusters 9 and 10) (end of vernalization treatment): six were up-regulated including *VRN1* (rbaal14f06\_s\_at) and a calcium binding protein (AJ250283\_at) ([Figure 6A, B](#pone-0017900-g006){ref-type="fig"}, [Table 1](#pone-0017900-t001){ref-type="table"}). In addition, eight were down-regulated including contigs for *HvOS2* (contig12031\_at), RNAse-S-like protein (contig5059\_s\_at) and a PR17d secretory protein (HW03O22u\_s\_at) ([Figure 6C--E](#pone-0017900-g006){ref-type="fig"}, [Table 1](#pone-0017900-t001){ref-type="table"}). ::: {#pone-0017900-g006 .fig} 10.1371/journal.pone.0017900.g006 Figure 6 ::: {.caption} ###### Examples of temperature responsive contigs. The normalized expression levels for individual contigs (A--L), as assayed by Affymetrix Barley1 Chip (RMA normalization) for control seedlings, short term or prolonged cold treatment, and one day after prolonged cold (a). Also shown are expression levels for contigs in the fully expanded second leaf of plants after vernalization (10 days post-vernalization), versus the equivalent leaves from non-vernalized control plants (b). :::  ::: ::: {#pone-0017900-t001 .table-wrap} 10.1371/journal.pone.0017900.t001 Table 1 ::: {.caption} ###### Contigs that show a lasting response to cold in the leaves of vernalized plants. ::: {#pone-0017900-t001-1} Probe Set Best Match Fold Change *P* Value -------------------- ---------------------------------------------- ------------- ----------- **Cluster 9** AJ250283\_at Calcium binding EF-hand protein 11.39 5.90E-04 rbaal14f06\_s\_at *VRN1* (MADS box gene) 6.12 2.63E-09 rbags15p13\_s\_at Jasmonate induced protein 5.43 1.47E-03 Contig1679\_s\_at Jasmonate-induced protein 1.81 1.33E-04 Contig372\_s\_at Unknown 1.70 1.69E-04 Contig19065\_at *Glucan synthase-like 3* (*Hordeum vulgare*) 1.60 7.59E-04 **Cluster 10** Contig12031\_at *ODDSOC2* (MADS box gene) −1.52 3.19E-03 Contig23272\_at Serine acetyltransferase protein, putative −1.55 1.07E-04 HS07I12u\_s\_at FAD binding domain, monooxygenase −1.58 3.72E-03 HE01O15u\_at extracellular dermal glycoprotein-like −1.67 3.60E-04 HVSMEf0015C12f\_at unknown −1.68 1.13E-03 Contig5059\_s\_at RNase S-like protein −1.73 5.94E-04 HW03O22u\_s\_at PR17d/secretory protein −1.75 3.07E-03 Contig9743\_at Unknown −1.79 1.61E-03 ::: Some contigs that respond to short term cold also had altered expression in the leaves of vernalized plants. These include *COR14b* (HVSMEa0015E13r2\_s\_at) ([Figure 6F](#pone-0017900-g006){ref-type="fig"}, [Table S9](#pone.0017900.s010){ref-type="supplementary-material"}), a cold acclimation protein WCS19 (baakp18\_s\_at) ([Figure 6G](#pone-0017900-g006){ref-type="fig"}, [Table S9](#pone.0017900.s010){ref-type="supplementary-material"}) and a LEA-like protein (contig10150\_at) ([Figure 6H](#pone-0017900-g006){ref-type="fig"}, [Table S9](#pone.0017900.s010){ref-type="supplementary-material"}). Others did not show consistent expression patterns when the seedling treatments were compared to the leaf samples. *CBF9* (HVSMEn0019L21f\_at) showed elevated expression after short term and prolonged cold but was down-regulated in the leaves of plants after vernalization ([Figure 6I](#pone-0017900-g006){ref-type="fig"}, [Table S9](#pone.0017900.s010){ref-type="supplementary-material"}). The expression pattern of these contigs is distinct to other contigs that are cold regulated. For example, contigs corresponding to *DHN5* (contig1717\_s\_at and HVSMEa0006I22r2\_s\_at) are also induced by cold but show no change in expression in the leaves of vernalized plants ([Figure 6J](#pone-0017900-g006){ref-type="fig"}). A full list of the contigs with altered expression in seedlings and leaves is provided in [Table S9](#pone.0017900.s010){ref-type="supplementary-material"}. Discussion {#s3} ========== Previous studies investigated the effects of low-temperature on the transcriptome of wheat or barley by examining short to medium term cold responses (1 day--2 weeks) [@pone.0017900-Svensson1]--[@pone.0017900-Winfield1]. In this study, transcriptional responses to short or prolonged cold were assayed and compared ([Figure 1](#pone-0017900-g001){ref-type="fig"}). The data presented show that transcriptional responses to short or prolonged cold differ markedly. This is evident from the lists of contigs with significantly changed expression for each treatment relative to the control ([Figure 2](#pone-0017900-g002){ref-type="fig"}, [Tables S1](#pone.0017900.s002){ref-type="supplementary-material"}, [S2](#pone.0017900.s003){ref-type="supplementary-material"}, [S3](#pone.0017900.s004){ref-type="supplementary-material"}, [S4](#pone.0017900.s005){ref-type="supplementary-material"}, [S5](#pone.0017900.s006){ref-type="supplementary-material"}, [S6](#pone.0017900.s007){ref-type="supplementary-material"}) and is further highlighted by PCA ([Figure 4](#pone-0017900-g004){ref-type="fig"}). Comparing and contrasting the effects of different lengths of cold treatment identified contigs potentially involved in different low-temperature responses. Contigs that respond to short term cold treatment are likely to function during cold "shock", to adjust homeostasis to rapid decreases in temperature, whereas contigs that show altered expression after prolonged cold are likely to be important for long term growth at low temperatures. Comparison of this dataset with previous microarray analyses of low-temperature responses in barley identified 55 contigs that showed a significant response to cold in all experiments (8 down and 47 up) ([Table S10](#pone.0017900.s011){ref-type="supplementary-material"}) (Plexdb accession no. BB65 and BB81; [@pone.0017900-Svensson1]). These contigs define a core set of low-temperature responsive genes from barley, including genes previously identified as cold responsive, such *DHN5* [@pone.0017900-Tommasini1], that are likely to play critical roles in cold acclimation (HVSMEa0006I22r2\_s\_at and contig1717\_s\_at) ([Table S10](#pone.0017900.s011){ref-type="supplementary-material"}). A limited number of contigs show a sustained response to prolonged cold after plants were shifted to warm conditions ([Figures 2](#pone-0017900-g002){ref-type="fig"}, [3](#pone-0017900-g003){ref-type="fig"}, [5](#pone-0017900-g005){ref-type="fig"}). Moreover, few of the contigs that showed altered expression in the prolonged cold treatment showed altered expression in the leaves of vernalized plants at the third leaf stage (10 days after the end of a prolonged cold treatment) ([Table 1](#pone-0017900-t001){ref-type="table"}). This observation is important with regards to the phenomenon of vernalization-induced flowering, since contigs that show a lasting response to prolonged cold potentially contribute to the acceleration of reproductive development that occurs in vernalized plants (see [@pone.0017900-Sasani1]). Indeed, the list of contigs identified as showing a sustained response to prolonged cold includes *VRN1*, a central regulator of the vernalization response in cereals [@pone.0017900-Trevaskis2]. *HvOS2* was identified amongst contigs that are down-regulated by vernalization ([Figure 6C](#pone-0017900-g006){ref-type="fig"}, [Table 1](#pone-0017900-t001){ref-type="table"}). This is consistent with previous studies, which showed that *HvOS2*, and two closely related wheat genes, *Triticum aestivum AGAMOUS-like 33* and *42* (*TaAGL33*, *TaAGL42*), show reduced transcript levels during and after vernalization [@pone.0017900-Trevaskis3], [@pone.0017900-Winfield2], [@pone.0017900-Greenup1]. Down-regulation of *HvOS2* in vernalized plants is likely to contribute to accelerated flowering [@pone.0017900-Greenup1], and although *HvOS2* is down-regulated by cold independently of *VRN1*, maintained repression of *HvOS2* after vernalization requires *VRN1* [@pone.0017900-Greenup1]. Thus, down-regulation of *HvOS2* in the leaves of vernalized plants can be considered a consequence of *VRN1* expression. An *RNase S-like* gene (contig5059\_s\_at), which is up-regulated by *HvOS2* [@pone.0017900-Greenup1], was down-regulated in the leaves of vernalized plants ([Figure 6D](#pone-0017900-g006){ref-type="fig"}, [Table 1](#pone-0017900-t001){ref-type="table"}). Although the function of this gene is not known, this expression pattern is consistent with the hypothesis that transcription of this gene is activated by *HvOS2*. A number of other vernalization-responsive contigs were identified. These include a contig predicted to encode a calcium binding EF-hand protein (AJ250283\_at), which showed increased expression after vernalization ([Figure 6B](#pone-0017900-g006){ref-type="fig"}, [Table 1](#pone-0017900-t001){ref-type="table"}). EF hand proteins act as calcium sensors and likely contribute to diverse biological processes, including hormone metabolism, cell signalling and gene expression (reviewed in [@pone.0017900-DeFalco1]). Calcium signalling might play a role during short term cold responses and cold acclimation [@pone.0017900-Winfield1], [@pone.0017900-Knight1], [@pone.0017900-Doherty1]. The identification of a vernalization-responsive gene encoding an EF-hand protein (contig AJ250283\_at) suggests that altered calcium signalling might also play a role during post-vernalization development in cereals. Contigs corresponding to two 23kd jasmonate induced proteins (rbags15p13\_s\_at and contig1679\_s\_at), which might regulate cell wall polysaccharide synthesis [@pone.0017900-Oikawa1], were induced by vernalization, as was a contig corresponding to a glucan synthase (contig19065\_at) ([Table 1](#pone-0017900-t001){ref-type="table"}). Conversely, a contig corresponding to a FAD binding domain containing protein (HS07I12u\_s\_at) was down-regulated ([Table 1](#pone-0017900-t001){ref-type="table"}). Altered transcript levels for these contigs might reflect adjustment of metabolism in vernalized plants to facilitate the transition to reproductive growth. Alternatively, metabolism might adjust to compensate for changes in metabolite pools that occur during prolonged growth at low-temperatures, which would not occur in control seedlings germinated at 20°C. A previous microarray study investigating seasonal flowering responses in wheat showed that transcript levels for key enzymes in the gibberellin biosynthesis pathway, ent-kaurene synthase and ent-kaurene oxidase, increase during step wise decreases in both temperature and photoperiod [@pone.0017900-Winfield2]. We found no evidence that these enzymes play a role in the vernalization response of barley seedlings; expression levels of *ENT-KAURENE SYNTHASE* (contig11470\_at) and *ENT-KAURENE OXIDASE* (contig15315\_at) remained at similar levels during and after vernalization ([Figure 6K, L](#pone-0017900-g006){ref-type="fig"}). The different findings of this study versus that of Winfield et al [@pone.0017900-Winfield2] might be due to the conditions used in each study; decreasing daylength versus darkness, or the age of plants examined (plants versus seedlings). Regardless, changes in transcript levels for these gibberellin biosynthetic enzymes are probably not required for the early stages of the vernalization response in barley seedlings. This does not rule out important roles for these genes in regulating responses to different temperature and daylength combinations, as suggested by Winfield et al [@pone.0017900-Winfield2], but highlights the advantage of using seedling vernalization as an experimental system; the effects of low-temperature can be separated from the effects of development or daylength. This is important because *sensu stricto* vernalization is a response to cold [@pone.0017900-Gassner1]--[@pone.0017900-Chouard1]. Furthermore, separating the effects of different seasonal flowering cues allows better prediction of physiological responses in complex environments. Some contigs that show a rapid response to cold also show altered expression following prolonged cold treatment. For example, *COR14b* (HVSMEa0015E13r2\_s\_at) and *WCS19* (baak1p18\_s\_at) were induced by short term cold, similar to previous studies [@pone.0017900-Vgjfalvi1], [@pone.0017900-Stockinger1], [@pone.0017900-Tommasini1]. These genes were also activated by prolonged cold treatment and expression remained high in seedlings a day after prolonged cold treatment and in the leaves of vernalized plants ([Figure 6G](#pone-0017900-g006){ref-type="fig"}). This contrasts with the behaviour of most other cold responsive genes; *DHN5* for example, returns to expression levels similar to the control treatment when plants are shifted to warm temperatures after prolonged cold ([Figure 6J](#pone-0017900-g006){ref-type="fig"}). The COR14b and WCS19 proteins localise to the chloroplast, possibly to reduce photo-oxidative stress [@pone.0017900-Gray1], [@pone.0017900-Crosatti1], but it is unclear why expression of these contigs is maintained after cold treatment. The lasting change in expression of these contigs following prolonged exposure to cold could be mediated by changes in chromatin state. Cold induced histone modifications are maintained at the promoters of some cold responsive genes in *Arabidopsis thaliana*, although expression of these genes is not maintained [@pone.0017900-Kwon1]. Two contigs corresponding to *CBF* genes were expressed at lower levels in the leaves of vernalized plants; *CBF2* (AF442489\_at) and *CBF9* (HVSMEn0019L21f\_at) ([Figure 6I](#pone-0017900-g006){ref-type="fig"}, [Table S9](#pone.0017900.s010){ref-type="supplementary-material"}). Reduced expression of these contigs in the leaves of vernalized plants, which are beginning reproductive growth at the time point sampled [@pone.0017900-Sasani1], might contribute to the reduced capacity for cold acclimation that is associated with reproductive growth in cereals [@pone.0017900-Limin1]. The equivalent *CBF* genes show elevated expression during cold acclimation in the *T. monococcum VRN1* deletion mutant (*mvp*), which is unable to initiate reproductive growth, consistent with this hypothesis [@pone.0017900-Dhillon1]. It might be possible to increase frost tolerance during flowering if expression of these genes could be maintained during reproductive growth, through constitutive expression in transgenic plants for example. Increased frost tolerance during flowering would be a valuable trait in areas where sudden spring frosts occur at the time of flowering, causing reduced yield. In conclusion, we have identified barley genes that respond to prolonged cold and show lasting changes in transcriptional activity when plants are shifted to normal growth conditions. The observation that only a limited number of contigs show lasting responses to prolonged cold, at least within the detection limits of microarray analysis, highlights the importance of *VRN1* in the vernalization response of temperate cereals. A key question for further research is how does prolonged cold lead to increased *VRN1* expression? By identifying genes that are differentially expressed during short and prolonged cold we have begun to address this question. Materials and Methods {#s4} ===================== Plant Growth {#s4a} ------------ To compare the effects of short and prolonged cold on transcription in barley seeds (cv. Sonja; a well characterised vernalization-responsive barley [@pone.0017900-Sasani1] with the genotype: *HvVRN1*, *HvVRN2*, *PPD-H1*, *ppd-H2*) were germinated and grown in darkness to an average coleoptile length of 4 cm at either 20°C over 5 days (control) or 4°C over 49 days (prolonged cold). Seedlings were then shifted from the control treatment to 4°C for 24 hours (short term cold) or shifted from the prolonged cold treatment to 20°C for 24 hours (post cold) ([Figure 1A](#pone-0017900-g001){ref-type="fig"}). In all treatments the shoot apex remained at an early stage of vegetative development, but plants grown from seedlings that experienced prolonged cold flower rapidly when shifted to normal growth temperatures, unlike control seedlings germinated at 20°C [@pone.0017900-Sasani1]. Seedlings were harvested from each treatment for RNA extraction. To examine gene expression after prolonged cold, barley plants (*Hordeum vulgare*) (cv. Sonja) were grown in pots covered in foil at 4°C for 49 days. The foil was then removed and plants were grown in a glasshouse (18±2°C) in long days (16-h light/8-h dark), with supplementary light when natural levels dropped below 200 µE until they reached the three leaf stage (10 days) ([Figure 1B](#pone-0017900-g001){ref-type="fig"}). Non-treatment control plants were grown at the same time under the same conditions and were sampled at the equivalent stage of development ([Figure 1B](#pone-0017900-g001){ref-type="fig"}). Microarray analysis {#s4b} ------------------- RNA was extracted using the method of Chang et al. [@pone.0017900-Chang1] and then further purified using RNeasy columns (Qiagen). Sample labelling and hybridisation to the Barley1 Gene chip [@pone.0017900-Close1] were conducted at the Australian Genome Research Facilities (AGRF; Melbourne, VIC, Australia), following the manufacturer\'s recommendations (Affymetrix, Santa Clara, CA). Microarray analyses were performed on 3 biological replicates of each sample. The resulting dataset was analysed in R v2.7.1 and analysed using packages from Bioconductor [@pone.0017900-Gentleman1] (<http://www.bioconductor.org/>), with default settings. Normalisation was carried out by Robust Multichip Analysis (RMA) and differentially expressed contigs were identified across the normalised microarray datasets for biological replicates using linear modelling in limma in R v2.7.1 (Linear Models for Microarray Data) [@pone.0017900-Smyth1]. Each experimental sample was compared with the control sample (e.g. Short Cold vs Control), or to each other, and multiple testing was corrected for by controlling the FDR [@pone.0017900-Benjamini1]. For PCA clustering, genes that were differentially expressed in any of the three seedling treatments relative to the control (empirical Bayes test, no minimum fold change cut off applied; 11057 contigs) were grouped based on condition using the \'cluster.samples\' function in smida (R v.2.7.1). The method chosen was 'pca' and 'euclidean' was selected as the distance measured. The clusters were plotted using the first two principal components from the PCA analysis. Comparisons between lists of contigs with significantly changed expression in the different treatments and the generation of preliminary Venn diagrams was performed using the FiRe macro in Excel® [@pone.0017900-Garcion1]. Raw microarray data has been deposited in the Plant Expression Database ([www.plexdb.org](http://www.plexdb.org)), a MIAME/Plant Compliant Gene Expression Resources for Plants and Plant Pathogens (Experiments BB94 and BB95). K-means Clustering {#s4c} ------------------ K-means cluster analysis was performed contigs that showed a two fold or greater change in transcript levels (p\<0.01,) in any of the two-way comparisons between the different seedling treatments (1800 in total, [Table S1](#pone.0017900.s002){ref-type="supplementary-material"}, [S2](#pone.0017900.s003){ref-type="supplementary-material"}, [S3](#pone.0017900.s004){ref-type="supplementary-material"}, [S4](#pone.0017900.s005){ref-type="supplementary-material"}, [S5](#pone.0017900.s006){ref-type="supplementary-material"}, [S6](#pone.0017900.s007){ref-type="supplementary-material"}). Cluster analysis was performed using the MeV software from the TM4 microarray software suite using the default settings (Euclidean distance, and a maximum of 50 iterations), using the RMA normalized expression values from all 12 samples (3 replicates for each of the four treatments) [@pone.0017900-Saeed1]. Contigs that did not group with the primary clusters or that showed high variability between replicates were manually omitted from the final clusters ([Table S11](#pone.0017900.s012){ref-type="supplementary-material"}). Gene Expression Analysis {#s4d} ------------------------ Total RNA was extracted using the method of Chang et al., [@pone.0017900-Chang1] or the Qiagen RNeasy Plant Miniprep kit (Qiagen). cDNA was synthesised using an oligo(T) primer (T18\[G/C/A\]) to prime first-strand complementary DNA (cDNA) synthesis from 1-5 µg of total RNA with SuperScript III reverse transcriptase enzyme (Invitrogen). qRT-PCR was performed on a Rotor-Gene 3000 real-time cycler (Corbett Research). The primers used for *HvACTIN*, *HvVRN1 and HvOS2* have been described previously [@pone.0017900-Greenup1], [@pone.0017900-Trevaskis4]. The sequence of primers used for contig6358\_at were as follows; forward 5′ TCCTCGTGTGATTTTGCAG 3′ and reverse 5′ TTGAGTTCAGCGATGCTACG 3′ and for HU14M19u\_at; forward 5′ TCAAAAAGGATGCCCAAAAG 3′ and reverse 5′ AACAAGCTTGGCAAAACACA 3′. qRT-PCR was performed using Platinum Taq DNA polymerase (Invitrogen) with SYBR green. Cycling conditions were 4 minutes at 94°C, 40 cycles of 10 s at 95°C, 15 s at 60°C, and 20 s at 72°C, followed by a melting-curve program (72°C--95°C with a 5-s hold at each temperature). Fluorescence data were acquired at the 72°C step and during the melting-curve program. Expression levels of genes of interest were calculated relative to *ACTIN* using the comparative quantification analysis method (Rotogene-5; Corbett Research), which takes into account the amplification efficiency of each primer set. Data presented are the average of a minimum of three biological replicates (unless stated otherwise) and the error bars show standard error. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Quantitative RT-PCR assay versus array analysis of gene expression for selected contigs. A. Expression values as assayed by Affymetrix Barley1 chip, RMA normalisation, for four contigs with contrasting expression patterns and different activity levels. *HvVRN1* (rbaal14f06\_s\_at), a gene encoding a ribosomal protein (HU14M19u\_at), a zinc finger transcription factor gene (contig6538\_at), and *HvOS2* (contig12031\_at). B. Quantitative RT-PCR assay of expression levels for the same contigs. Expression is shown relative to *ACTIN*. Error bars show standard error. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Contigs with altered expression in the short cold treatment relative to no treatment control. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Contigs with altered expression in the prolonged cold treatment relative to the no treatment control. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### Contigs with altered expression 1 day after prolonged cold treatment relative to the no treatment control. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### Contigs with altered expression in the prolonged cold treatment relative to the short cold treatment (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### Contigs with altered expression after one day after prolonged cold relative to short cold treatment. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S6 ::: {.caption} ###### Contigs with altered expression in the prolonged cold treatment relative to one day after prolonged cold treatment. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S7 ::: {.caption} ###### Numbers and descriptions of contigs clustered together according to expression behaviour. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S8 ::: {.caption} ###### Contigs with altered expression in the leaves of barley plants after vernalization. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S9 ::: {.caption} ###### Contigs represented in primary cluster analysis with altered expression in the leaves of barley plants after vernalization. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S10 ::: {.caption} ###### A core set of low-temperature responsive contigs in barley. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S11 ::: {.caption} ###### Contigs omitted from top the ten main clusters. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Jen Taylor for assistance with cluster analysis. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by the Commonwealth Scientific and Industrial Research Organization (CSIRO) Division of Plant Industry and the CSIRO Office of the Chief Executive (OCE). Aaron Greenup was supported by a scholarship from the Australian National University College of Science. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: AGG SS SNO BT. Performed the experiments: AGG SS SNO BT. Analyzed the data: AGG SNO SAW BT. Contributed reagents/materials/analysis tools: AGG SAW BT. Wrote the paper: AGG SAW AAM BT.

PubMed Central

2024-06-05T04:04:19.244750

2011-3-9

PMC3052372

Introduction {#s1} ============ The circadian timing system in mammals exerts control over a wide range of physiology and behavior, including daily environmental changes, the circadian system can be reset by external time cues (Zeitgeber) such as light and metabolic correlates of feeding [@pone.0017655-Lamont1]--[@pone.0017655-Dibner1]. Two separate, but coupled oscillators, LEO and FEO, are involved in circadian system(s) in mammals. Whereas the master LEO is located in the SCN of the hypothalamus [@pone.0017655-Schibler1], [@pone.0017655-Escobar1], localization of the FEO(s) still remains to be determined [@pone.0017655-Krieger1]--[@pone.0017655-Stephan1]. An output of a putative FEO is FAA, which can be assessed by restricted feeding (RF), daily limiting food availability to a restricted time window. SCN lesion studies demonstrated that FEO-driven FAA persists without the LEO, suggesting that FEO resides outside of the SCN [@pone.0017655-Krieger1], [@pone.0017655-Stephan2], [@pone.0017655-Stephan3]. It is also reported that FEO dominates LEO under RF in the entrainment of activity phase [@pone.0017655-Schibler1]. Moreover, disruption of the known circadian clock genes resulted in mostly partial or little alterations of FAA [@pone.0017655-Challet1]. These data suggest that the FEO, which functions under limited nutrient availability, is a dominant circadian oscillator independent of LEO. To elucidate molecular machinery driving circadian clock(s) in mammals, a list of candidate circadian clock/clock-controlled genes have been identified by microarray [@pone.0017655-Stephan1], [@pone.0017655-Mistlberger1], [@pone.0017655-Mistlberger2]. Among these was *Rgs16*, a member of RGS family. RGS proteins regulate G protein-coupled receptor (GPCR)-mediated signaling by negatively or positively interacting with downstream effectors [@pone.0017655-Willars1]. In the brain, *Rgs16* is predominantly expressed in the SCN and thalamus [@pone.0017655-GrafsteinDunn1], [@pone.0017655-Ueda1]. RGS16 exhibits robust circadian rhythms both in the SCN and in liver with its peak at 4--6 (zeitgeber time, ZT4-6) and 8--10 (ZT8-10) hours after light-on, respectively, suggesting that RGS16 is involved in the central and peripheral circadian clocks and/or their outputs [@pone.0017655-Ueda1]. Interestingly, a previous study demonstrated that *Rgs16* expression in liver was up-regulated during fasting and rapidly down-regulated by re-feeding [@pone.0017655-Huang1]. In addition, expression of *Rgs16* in liver was restricted to periportal hepatocytes, the predominant locations of gluconeogenesis and lipolysis [@pone.0017655-Huang1]. Considering that a number of GPCR ligands including vasoactive intestinal peptide (VIP; [@pone.0017655-Maywood1]), glutamate [@pone.0017655-Hannibal1], [@pone.0017655-Gillette1], melatonin [@pone.0017655-Pevet1], [@pone.0017655-Dubocovich1], orexin [@pone.0017655-Akiyama1], [@pone.0017655-Mieda1] and ghrelin [@pone.0017655-LeSauter1], [@pone.0017655-Blum1] have been implicated in the central or peripheral circadian clocks, these data raise the possibility that RGS proteins function in the circadian systems by modulating signaling through as yet unknown GPCR/GPCR ligand(s). Here, to elucidate the possible role of RGS16 in the circadian clock *in vivo*, we generated two independent knockdown (KD) mouse lines using lentiviral vectors expressing two separate short hairpin RNAs (shRNAs) targeting the *Rge16* mRNA. Results {#s2} ======= Generation of the *Rgs16* knockdown mice {#s2a} ---------------------------------------- We designed different shRNAs against different regions of *Rgs16*. Each shRNA expression vectors was transfected into NIH3T3 cell line and levels of the *Rgs16* mRNA were quantified by quantitative RT-PCR (qPCR). The two shRNAs silencing endogenous *Rgs16* mRNA with the greatest efficiency (\#41 and \#53) were selected for producing two independent transgenic mouse lines ([Fig. S2](#pone.0017655.s002){ref-type="supplementary-material"}). We then produced separate high-titer lentiviral vector lots encoding the two shRNA expression cassettes as well as the GFP protein, and introduced them into fertilized one-cell stage mouse embryos by microinjection into the perivitelline space [@pone.0017655-Lois1], [@pone.0017655-Shaughnessy1]. The transgenic mice were selected by detecting GFP expression and the presence of the transgene confirmed by PCR-based genotyping. Expression of the *Rgs16* mRNA in the KD and wild-type brain and liver {#s2b} ---------------------------------------------------------------------- We first examined the spacio-temporal expression patterns of the *Rgs16* mRNA in brain and liver of the KD and wild-type (WT) mice by *in situ* hybridization (ISH) and quantitative PCR (qPCR). In the WT brain, predominant expression of *Rgs16* was observed in the SCN, and at a lower level in thalamus ([Fig. 1A](#pone-0017655-g001){ref-type="fig"}). We also confirmed that *Rgs16* transcription exhibits robust circadian rhythms in the SCN ([Fig. 1B](#pone-0017655-g001){ref-type="fig"}). As shown in [Fig. 1C](#pone-0017655-g001){ref-type="fig"}, average *Rgs16* expression level was reduced the KD SCN ([Fig. 1C](#pone-0017655-g001){ref-type="fig"}). In the WT liver, robust circadian rhythms were observed peaking at ZT11 ([Fig. 1D, F](#pone-0017655-g001){ref-type="fig"} = 21.7, *P\<*0.001). In the KD liver, average *Rgs16* expression level was reduced, and circadian changes were weakened ([Fig. 1D, F](#pone-0017655-g001){ref-type="fig"} = 6.06, *P\<*0.05) relative to the controls (F = 21.7, *P\<*0.001). In thalamus, no daily expression rhythms were observed in both WT (F = 0.65, *P\>*0.05) and KD (F = 0.36, *P\>*0.05) mice ([Fig. 1E](#pone-0017655-g001){ref-type="fig"}). The average expression level of *Rgs16* in the KD thalamus was lower than that of WT control, however, the knockdown efficiency was lower than that in liver ([Fig. 1D, E](#pone-0017655-g001){ref-type="fig"}). ::: {#pone-0017655-g001 .fig} 10.1371/journal.pone.0017655.g001 Figure 1 ::: {.caption} ###### Spatial and temporal expression patterns of the *Rgs16* mRNA in brain and its reduction in the KD mice. **A, B,** *in situ* hybridization performed on brain sections. **A,** Specific and intense signals were observed in the SCN and thalamus (Th). **B,** Diurnal rhythms of the *Rgs16* transcript were observed in the SCN. **C,** ISH quantification of the *Rgs16* mRNA in the KD and WT SCN (n = 8 each) at ZT5. \*\**P*\<0.01 vs. WT (Student\'s *t*-test). **D,** Expression of *Rgs16* in the KD and WT liver observed by qPCR. \*\**P*\<0.01 vs. WT (Student\'s *t*-test). \#\#, *P*\<0.01 circadian gene expression profile by 2-way ANOVA (WT vs. KD mice; n = 3--4). **E,** Expression of *Rgs16* in thalamus of the KD and WT mice (n = 3--4). \**P\<*0.05 vs. WT. :::  ::: Free-running period of locomotor activity rhythm was shorter in the KD mice {#s2c} --------------------------------------------------------------------------- To study the possible involvement of RGS16 in the central circadian clock, we examined locomotor activity rhythms of the KD and control mice. Wheel-running activities of individual mice were monitored under 12 hr light and 12 hr dark (LD) and constant dark (DD) conditions ([Fig. 2A, B](#pone-0017655-g002){ref-type="fig"}). In DD, The KD mice showed an average free-running period significantly shorter than that of the controls (23.84±0.05 hr in WT vs. 23.45±0.07 hr in KD, *P\<*0.01, [Fig. 2C](#pone-0017655-g002){ref-type="fig"}). After locomotor activity rhythms were assessed, brains were sampled at ZT5 from individual KD and WT mice and quantitative ISH was performed on the brain sections. Average *Rgs16* mRNA level in the SCN was significantly decreased in the KD mice relative to controls ([Fig. 1C](#pone-0017655-g001){ref-type="fig"}). ::: {#pone-0017655-g002 .fig} 10.1371/journal.pone.0017655.g002 Figure 2 ::: {.caption} ###### Short free-running period of locomotor activity rhythm in the KD mice. **A, B,** Locomotor activity rhythms of the WT (A) and KD (B) mice were entrained to 12∶12 LD. But in DD, the KD mice exhibited circadian period of locomotor activity shorter than that of WTs. The vertical axis in each graph indicates the day path. Horizontal open and closed bars indicate the light and dark periods, respectively. **C,** The difference in the circadian period between the KD and WT mice (n = 15 each) was statistically significant \*\**P*\<0.01 vs. WT (Student\'s *t*-test). :::  ::: Total amount of locomotor activity was reduced in the *Rgs16* KD mice {#s2d} --------------------------------------------------------------------- We next compared total amount of locomotor activity of the *Rgs16* KD mice with that of controls. Total activity counts were assessed by an infrared sensor of KD and WT mice averaged every 30 minutes. The average activity of the KD mice was significantly lower than that of the controls regardless of day or night ([Fig. 3A, B](#pone-0017655-g003){ref-type="fig"}), whereas averaged day/night ratio of locomotor activity was comparable between the two genotypes ([Fig. 3C](#pone-0017655-g003){ref-type="fig"}). ::: {#pone-0017655-g003 .fig} 10.1371/journal.pone.0017655.g003 Figure 3 ::: {.caption} ###### Reduced amount of locomotor activity in the *Rgs16* KD mice. **A,** Averaged daily activity plot of the KD and WT mice under LD condition. **B,** Average daytime or nighttime locomotor activity was significantly decreased in the KD mice as compared to that of WT controls (n = 21 each). \*\**P\<*0.01 vs. WT (Student\'s *t*-test). **C** Averaged day/night ratio of locomotor activity was comparable between the KD and WT mice (n = 21 each). :::  ::: Attenuated FAA in the *Rgs16* KD mice {#s2e} ------------------------------------- The *Rgs16* mRNA in liver is not only circadianly regulated, but is also up-regulated by fasting and down-regulated by re-feeding [@pone.0017655-Huang1]. These data raise the possibility that RGS16 is involved in metabolism in liver and in the food-driven behavior (FAA) regulated by a certain brain region(s) and/or liver. To test this, we next examined the effects of RF, daytime scheduled feeing for 4 hours (ZT6-ZT10), which is usually a resting period for nocturnal rodents. When mice are food-restricted during the day, FAA is observed about 3--4 hour prior to the feeding time [@pone.0017655-Mistlberger1]. In WT mice, a remarkable increase in the locomotor activities was observed for about 4 hours before the scheduled feeding time ([Fig. 4A, C](#pone-0017655-g004){ref-type="fig"}). On the other hand, the FAA of the KD mice was significantly attenuated in comparison with the controls ([Fig. 4B, C](#pone-0017655-g004){ref-type="fig"}). The reduction in the average percent FAA in the KD animals compared to that in the controls was statistically significant ([Fig. 4D, *P*](#pone-0017655-g004){ref-type="fig"} *\<*0.01). [Fig. 4E](#pone-0017655-g004){ref-type="fig"} demonstrates FAA counts of individual KD and WT mice under RF as compared with baseline activity counts under free feeding (FF, as measured activity during ZT2-6). Bouts of locomotor activity during daytime was strongly reduced in KD group by RF, however, those in nighttime were comparable between WT and KD mice under FF or RF (Compare [Fig. 3A](#pone-0017655-g003){ref-type="fig"} and [4C](#pone-0017655-g004){ref-type="fig"}). It is of note that, under the RF schedule, both WT and KD mice obtained almost the same amount of food during ZT6-ZT10 (0.88±0.08/10 g body weight/day for the WT; 0.80±0.05/10 g body weight/day for KD). Although body weight was slightly reduced by RF in both WT (33.1±2.8 g prior to and 30.0±1.6 g after RF) and KD mice (30.7±1.6 g prior to and 27.6±0.8 g after RF), there was no significant difference in the amount of food intake or the body weight between the two genotypes (*P*\>0.05, Student\'s *t*-test). ::: {#pone-0017655-g004 .fig} 10.1371/journal.pone.0017655.g004 Figure 4 ::: {.caption} ###### Attenuated FAA in the KD mice under daytime-RF. **A, B,** Examples of double-plotted actograms of WT (A) or KD (B) mice under RF. The RF time was set at ZT6--ZT10 (open box). Whereas WT mouse showed a strong FAA (A), those of KD were very weak (B). **C,** Averaged daily activity plot of the KD and WT mice under RF. **D,** Average percentages of FAA \[100× activity counts (ZT2-6)/24-hour activity counts\] during the last three days of RF between WT and KD mice were compared (n = 11 each). \*\**P*\<0.01 vs. WT (Student\'s *t*-test). **E,** FAA counts of individual mice during 4 hours preceding RF (ZT2-6) averaged over the first 7 days of FF and the following 12 days of RF were compared in each genotype (n = 11 each). \**P\<*0.05, vs. WT in FF; \#\#*P\<*0.01 vs. WT in RF (Student\'s *t*-test). :::  ::: Daytime-RF phase-advanced the *Per2* mRNA rhythm in WT, but not in KD {#s2f} --------------------------------------------------------------------- Reduced FAA has been reported in several knockout mice, but the mechanism underlying such behavioral change remains to be elucidated. To examine whether RF elicits FAA by affecting expression profiles of circadian clock gene and/or *Rgs16* in brain and liver, we sampled KD and WT mouse tissues at 4 time points (ZT5, 11, 17, and 23) and performed qPCR. [Fig. 5](#pone-0017655-g005){ref-type="fig"} demonstrates the circadian rhythms in the *Per2* clock gene expression in liver and thalamus in both the WT and the KD mice, where *Rgs16* transcripts are abundantly expressed. Under FF, the *Per2* mRNA exhibited circadian rhythms both in liver (F = 21.7 *P\<*0.01 for WT, F = 16.1 *P\<*0.05 for KD, one-way ANOVA) and thalamus (F = 5.1 *P\<*0.05 for WT) with a peak at ZT17 in both genotypes. Under RF in WTs, the *Per2* rhythms were significantly phase-advanced, peaking at ZT11 and ZT5 in liver and thalamus, respectively ([Fig. 5A, C](#pone-0017655-g005){ref-type="fig"}; F = 25.2 *P\<*0.01 for liver, F = 4.1 *P\<*0.05 for thalamus). In contrast, the phase of the *Per2* rhythms in the KD mice was unaffected by RF either in liver or thalamus ([Fig. 5B, D](#pone-0017655-g005){ref-type="fig"}). ::: {#pone-0017655-g005 .fig} 10.1371/journal.pone.0017655.g005 Figure 5 ::: {.caption} ###### Daytime-RF phase-advanced circadian rhythms of *Per2* in liver and thalamus in WT, but not in the KD mice. Under RF (ZT6-10), the WT mice liver (**A**) and thalamus (**C**) phase-advanced *Per2* rhythms by 6 hours and 12 hours, respectively, as compared to the control mice under FF. In contrast, the circadian phase of *Per2* in the KD mice was not shifted in either liver (**B**) or thalamus (**D**) by RF. \#\#*P\<*0.01 circadian gene expression profile by 2-way ANOVA (FF vs. RF mice, A, C, D). \*\**P\<*0.01, \**P\<*0.05 vs. WT (Student\'s *t*-test). n = 3--4 for each ZT point. FF, free feeding; RF, restricted feeding. :::  ::: Discussion {#s3} ========== In the present study, we generated transgenic *Rgs16* KD mice using lentiviral vectors carrying a shRNA expression cassette and a modified microinjection method, i.e., perivitelline injection [@pone.0017655-Lois1], [@pone.0017655-Shaughnessy1]. This gene silencing technique enabled us to efficiently generate sufficient numbers of transgenic mice within a few months, because of the higher rate of transgenesis by lentiviral vector injection compared with the conventional pronuclear DNA microinjection [@pone.0017655-Park1]--[@pone.0017655-GamaSosa1]. Moreover, this *in vivo* KD strategy could be a method of choice when disruption of the gene of interest is likely to cause lethality *in vivo*, or leads to a lack of phenotype because redundant functionalities from within the gene family or from related gene(s) result in compensation for the gene loss. We anticipated that *Rgs16* knockout could potentially be compensated by other *Rgs* genes (*Rgs2* and/or *Rgs4*), because these genes, which belong to the same R4 subfamily along with *Rgs16*, are also circadianly expressed in the SCN [@pone.0017655-Ueda1]. Consistent with this expectation, another group reported no significant phenotype in the *Rgs16* knockout mice [@pone.0017655-Bansal1]. Our study provided evidence that G protein signaling modulated by RGS16 plays a role not only in the master LEO, but also in the FEO(s). Previous studies have indicated potential circadian roles of several GPCRs in the SCN. For instance, knockout and other studies strongly suggest that VIP and its receptor VPAC2, a member of GPCR family, are important for synchronization of the circadian clock and for sustaining circadian rhythmicity [@pone.0017655-Maywood1]. Likewise, GPCR ligands/GPCRs such as orexin/receptor and ghrelin/receptor are also involved in the FEO [@pone.0017655-Akiyama1]--[@pone.0017655-Blum1]. Further studies to identify upstream GPCR(s) will shed light on a significant role(s) of RGS16 in the SCN, thalamus and liver. In the present study, expression level of *Per2* was up-regulated by RF, consistent with the previous report demonstrating an RF-induced change in basal level of gene expression [@pone.0017655-Oishi1]. Food seeking/satiety and energy metabolism may be one of the factors for FAA formation [@pone.0017655-Akiyama1], [@pone.0017655-Mieda1], [@pone.0017655-Stephan4]. In addition, previous studies have shown that the volume of food can produce large phase-shifts in the FAA rhythm in rats, and also large phase-advance in the *Per2* gene expression in mouse liver in a food volume-dependent manner [@pone.0017655-Hirao1]. In our data, food availability and body weight change during RF conditions were not significantly different between the WT and KD mice, suggesting that deficit of FAA formation and phase-shift of the *Per2* expression in liver did not lead to impairment of food seeking ability and/or the digestive system. RGS16 is the first molecule that has been reported to control not only FAA formation but also phase-shift of liver clock by RF. Thus, our data could provide a new insight into where and how these signaling molecules including RGS proteins affect the circadian timing systems. Our present study also demonstrated that, whereas the circadian rhythms of the *Per2* mRNA were significantly phase-advanced by daytime-RF, both in liver and thalamus compared to free-feeding WT controls, the *Per2* rhythms in the KD mice did not phase-shift in either organ. These results suggest that RGS16 is potentially an upstream regulator of the *Per2* rhythms in liver and thalamus, whereas expression of *Rgs16* itself is regulated by the circadian clock. The data also suggests that RGS16 is involved in the phase-shift of both behavioral and molecular rhythms evoked by RF. It is noteworthy that a phase-advance in the *Per2* rhythms by daytime-RF was observed in thalamus. These data, together with the specific expression of *Rgs16* in this brain region, imply that thalamus plays a role in the FEO. It has been reported that complete paraventricular thalamic nucleus (PVT) lesion affects LEO, but not FEO [@pone.0017655-Landry1], although involvement of thalamic regions other than PVT in the circadian systems needs to be determined. Interestingly, the report also indicated that PVT ablation increased total daily activity, suggesting that this thalamic region is involved in regulating behavioral output. Altogether, our data along with previous studies raise a possibility that RGS16 in thalamus modulates light-entrained and/or food-entrained behavior by regulating activity levels and/or circadian behavioral rhythms under different environmental conditions. Our study provided for the first time evidence suggesting that two distinct regions of the body, thalamus and liver, where *Rgs16* mRNA is abundantly expressed, are involved in the regulation of the FAA under restricted food availability. The data raise the question of whether liver or thalamus plays a lead role in the FEO-operated FAA, or both of these regions regulate the behavior, either independently or cooperatively. Further studies in search of upstream GPCRs/ligands and downstream signaling molecules will elucidate the mechanism of how RGS16 specifically regulates food-entrained behavior in different loci. Materials and Methods {#s4} ===================== Animals {#s4a} ------- C57BL/6J inbred mice were purchased from CREA Japan and used for generating transgenic mice and all other experiments. Animals were provided with food and water *ad libitum*, and entrained to 12 hr light: 12 hr dark condition (12∶12 LD) for at least 2 weeks at 23±2°C, All animals were cared for in accordance with the Law (No. 105) and Notification (No. 6) of the Japanese Government, and all experiments were conducted under permission of the Experimental Animal Welfare Committees of Kinki University (Permission \#KDMS-16-002) and Waseda University (Permission \#08A36). shRNA {#s4b} ----- Six shRNAs were designed from the *Rgs16* cDNA sequence, and annealed double-stranded oligomers were subcloned into the pSilencer 1.0-U6 siRNA Expression Vector (Ambion). Knockdown efficiencies of the expression vectors carrying individual shRNAs were evaluated by transient transfection of the vectors into mouse NIH3T3 (RIKEN Cell Bank) fibroblast cell line by Lipofectamine 2000 (Life Technologies) according to manufacturer\'s instruction. After 48 hours of incubation, cells were lysed using Sepazol-RNA I (Nacalai Tesque), RNA was extracted, and quantitative PCR (qPCR) was performed using *Rgs16* primers. Two shRNAs (\#41, 53), which significantly suppressed the endogenous *Rgs16* expression levels, were used for transgenesis. The sequences of annealed oligonucleotides for expressing \#41 or \#53 shRNA were as follows: \#41: 5′-GCGAGGAGTTCAAGAAGATTTCAAGAGAATCTTCTTGAACTCCTCGCTTTTTT-3′ and 5′-AATTAAAAAAGCGAGGAGTTCAAGAAGATTCTCTTGAAATCTTCTTGAACTCCTCGCGGCC-3′ \#53: 5′-GAGAACTGACCAAGACAAATTCAAGAGATTTGTCTTGGTCAGTTCTCTTTTTT-3′ and 5′-AATTAAAAAAGAGAACTGACCAAGACAAATCTCTTGAATTTGTCTTGGTCAGTTCTCGGCC-3′ Lentiviral vector construction and production of high-titer lentiviral vector particles {#s4c} --------------------------------------------------------------------------------------- The two shRNA sequences (\#41 and \#53) targeting *Rgs16* along with the mouse U6 promoter were PCR-amplified from the pSilencer vectors and cloned into the pTZV TranzVector™ (Tranzyme) as depicted in [Fig. S1](#pone.0017655.s001){ref-type="supplementary-material"}. An eGFP fluorescent marker driven from the human immediate early cytomegalovirus virus (CMV) promoter is co-expressed with the shRNA hairpin sequence. High-titer lentiviral vector particles were generated using the Trans-Lentiviral™ Vector Packaging system [@pone.0017655-Wu1]. Viral particles were concentrated by ultracentrifugation. Functional titers were determined by transducing HEK293T (ATCC) cells with limiting dilutions of virus and counting GFP-positive colonies. Generation of *Rgs16* KD mice {#s4d} ----------------------------- Fertilized oocytes were harvested from super-ovulated C57BL/6J female mice. Viral vectors at a concentration of 1×10^9^/ml were microinjected into the perivitelline space of the oocytes using a FemtoJet injector (Eppendorf), and then reimplanted into the oviduct of pseudo-pregnant recipient females after 2--4 hours of microinjection. Injections were performed under 400× magnification (DM IRE2, Leica), and injection volume was approximately 100 nl. Genotyping and evaluation of the KD mice {#s4e} ---------------------------------------- For genotyping, genomic DNA was extracted from a tail tip of each mouse by a standard protocol using proteinase K and following phenol/chloroform extraction. Using the genomic DNA as a template, PCR was performed using GFP primers. To confirm expression of the GFP protein, we observed tail tips under fluorescent microscopy. qPCR {#s4f} ---- Total RNA was extracted from tail tip, liver, and thalamus of each KD or WT control mouse using RNeasy Mini Kit (Qiagen). The extracted RNA was reverse-transcribed into cDNA using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). For qPCR reaction, SYBR GREEN Premix ExTaq (Takara) was used following manufacturer\'s instruction. The sequences of primers used in this study are as follows. GFP primers: 5′-AGCAAAGACCCCAACGAGA-3′ and 5′-GGCGGCGGTCACGAA-3′; *Rgs16* primers: 5′-TGCTTGTGAACAGGGCTAACTG-3′ and 5′-CTCCCTCCTTAGACCCCATCTT-3′; *Per2* primers: 5′-TGTGTCTTACACGGGTGTCCTA-3′ and 5′-ACGTTTGGTTTGCGCATGAA-3′, 18S rRNA primers: 5′-CGGCTACCACATCCAAGGAA-3′, 5′-GCTGGAATTACCGCGGCT-3′. The value of the PCR product of the target gene was normalized to that of 18S rRNA. Measurement of locomotor activity rhythm {#s4g} ---------------------------------------- Mice were individually housed in translucent polypropylene cages under 12∶12 LD (200 lux) and DD, and locomotor activity was assessed either by a running wheel ([Fig. 2](#pone-0017655-g002){ref-type="fig"}) or an area sensor (F5B, Omron; [Fig. 3](#pone-0017655-g003){ref-type="fig"}, [4](#pone-0017655-g004){ref-type="fig"}). Activity was continuously monitored and analyzed using ClockLab software (Actimetrics). Schedules for RF {#s4h} ---------------- Individually housed KD mice and WT controls were maintained in a 12∶12 LD cycle under FF for at least 7 days. The RF experiment for the measurement of anticipatory activity was performed as previously described [@pone.0017655-Wakamatsu1]. Briefly, after 1 day of food deprivation starting at ZT12 (day 0), food was restricted to ZT6-ZT10 for 12 days (day 1--12). From day 13 to 14, food was again withdrawn for the entire day to record motor activity under food deprivation. Another RF experiment for the measurement of clock gene expression was performed. After application of the same RF conditions, animals were sacrificed at ZT5, 11, 17, and 23 on day 13 under food deprivation. *In situ* hybridization (ISH) {#s4i} ----------------------------- Riboprobe was labeled with ^35^S-UTP (Amersham/GE Healthcare) by *in vitro* transcription using either T7 or SP6 polymerase (Promega). Frozen mouse brain sections (40 µm thickness) were hybridized with riboprobe for overnight and apposed to Kodak film (BioMax MR). The *Rgs16* cDNA (entire ORF length 606 bp) was amplified by PCR and subcloned into pGEM-T Easy Vector (Promega). The plasmids were linealized with *Nco*I to synthesize riboprobe. Statistics {#s4j} ---------- Results were expressed as the mean ± SEM. One-way ANOVA was applied to evaluate significant difference of circadian rhythmicity of gene expression, and 2-way ANOVA was applied to evaluate significance between KD and control mice. The significance of the differences between groups was determined by the Student\'s *t*-test. Statistical analysis software (StatView version 5.0, SAS Institute) was used. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Schematic representation of the shRNA lentiviral vectors.** The shRNA hairpin sequences (\#41 and \#53) targeting mouse *Rgs16* is cloned into pTZV, a HIV1-based self-inactivating (SIN) lentiviral transfer vector containing the central polypurine tract/termination sequence (FLAP) and the Woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) for enhanced gene expression. A PCR-amplified fragment containing the mouse U6 promoter, shRNA hairpin, and pol III termination sequence (TTTTT) is cloned into pTZV between 5′ -- *Cla*I and 3′ -- *Bam*HI sites. A GFP marker is co-expressed with the shRNA from a CMV promoter to enable visualization of transduced cells and transgenic embryos. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### ***In vitro*** **knockdown of the** ***Rgs16*** **mRNA by shRNAs used in this study.** 2 µg of the pSilencer 1.0 vector expressing either of the two shRNAs (\#41 or \#53) was transfected into the NIH3T3 mouse fibroblast cell line with Lipofectamine 2000. RNA was extracted from each of the transfected cells two days after the transfection, and expression levels of *Rgs16* were compared by qPCR (see [Materials and Methods](#s4){ref-type="sec"}). Both \#41 and \#53 shRNAs significantly reduced the average *Rgs16* mRNA level (\*\**P*\<0.01, n = 4). (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Ralph Mistlberger for critical reading, and Takahiro Moriya, Yuka Miyoshi, Yuka Sugahara, Shinsuke Noso, Naru Babaya, Ikuko Yamada, Tomohiro Suzuki, Tamio Furuse and Hiroshi Takemori for technical support and helpful discussion. **Competing Interests:**Authors include employees at Tranzyme Pharma and have an affiliation to Tranzyme Pharma using the material TranzVector for the study. This does not alter the authors\' adherence to all the PLoS ONE policies on sharing data and materials. **Funding:**This study was supported by Grant-in-Aid for Scientific Research: No. 19590235 and 21590264 from the Japan Society for the Promotion of Science, Grants-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan including the High-Tech Research Center and Anti-aging Center grants, those from the Ministry of Health, Labor and Welfare of Japan for Research on HIV/AIDS, and those from the Japan Health Sciences Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: NH. Performed the experiments: NH KA SK SY STK KH. Analyzed the data: NH HI SW TM SS. Contributed reagents/materials/analysis tools: JKW RR MM. Wrote the paper: NH SS. [^2]: ¤a Current address: Thermo Scientific/Open Biosystems, Huntsville, Alabama, United States of America [^3]: ¤b Current address: Integrated Systems Toxicology Division, NHEERL, ORD, United States Environmental Protection Agency, Research Triangle Park, North Carolina, United States of America

PubMed Central

2024-06-05T04:04:19.248001

2011-3-9

PMC3052373

Introduction {#s1} ============ Many animals, including humans, use vocal signals to communicate with conspecifics. Song is a long, complex vocalization of several acoustic elements arranged in specific sequences [@pone.0017721-chpole1]. [@pone.0017721-Fitch1]. While most mammals, birds [@pone.0017721-Jarvis1], and frogs [@pone.0017721-Feng1] tested show only genetically regulated patterns of vocalizations, several rare groups of birds (songbirds, parrots, hummingbirds) and mammals (whales, bats and humans) also learn vocalizations. They learn them through social imitation, with different degrees of innate constraints depending on the species [@pone.0017721-Janik1], [@pone.0017721-Marler1]. In most species, vocal learning occurs mainly during juvenile development. In zebra finches, for instance, approximately 30 days after hatching, young males start producing unstructured sounds. The onset of vocal learning after exposure to a song model from a tutor, usually the father, is marked by the rapid emergence of structured sounds. To learn a song, the bird has to compare these sounds with a memory template of the song model using auditory feedback [@pone.0017721-Doupe1]. Learning songs is achieved by transforming and differentiating prototype sounds until they resemble the different syllables of the song model. This type of vocal learning for which neural and molecular substrates have been well documented [@pone.0017721-Jarvis1] is similar to human spoken language learning [@pone.0017721-Doupe1]. The mouse, *Mus musculus*, is a genetically and neurochemically well-described mammalian organism. Mice emit ultrasonic vocalizations with frequencies higher than 30 kHz, which is far beyond the human audible range [@pone.0017721-Sewell1]. Mice produce ultrasonic vocalizations in 2 social contexts: first, pups\' production of ''isolation calls'' in cold conditions or when they are separated from the dam [@pone.0017721-Uematsu1], [@pone.0017721-Noirot1]; second, males emitting ''ultrasonic vocalizations'' in the presence of females or when they are stimulated by the female\'s urinary pheromones [@pone.0017721-Nyby1]. Recent studies have demonstrated that ultrasonic song vocalizations of male mice have behavioral features similar to those of bird songs, including discrete syllables with temporal sequencing, repeated phrases, and variability among individuals [@pone.0017721-Holy1]. The B6D2F1strain of male mice showed individual differences in syllable usage and the temporal structure of their songs as reported by Holy and Guo [@pone.0017721-Holy1]. Furthermore, mating has been shown to change the quality and quantity of male ultrasonic vocalization [@pone.0017721-Wang1]. These findings lead to the hypothesis that male mouse songs may have an experience-dependent phenotype. However, the influence of social environments during the early developmental period, in which songbirds learn the prototype of songs from their tutors as clearly shown by cross-fostering studies [@pone.0017721-Clayton1], has not been examined. To elucidate genetic and environmental effects on mouse songs, we conducted a cross-fostering study to understand the effects of the social experience during the juvenile developmental period on song development. First, we compared 2 strains of inbred C57BL/6 and BALB/c males and found that these 2 strains of male mice emitted a different pattern of songs with regard to frequency, inter-syllable intervals, and syllable composition. C57BL/6 males showed a higher peak frequency of syllables and more frequency-modulated syllables with 1 or multiple jumps and short- and upward syllables, whereas BALB/c males produced more chevron-, flat-, and harmonics-syllables. None of these strain-specific parameters were affected by cross-fostering. Therefore, developmental social environments appear to have no significant role in adult male songs of mice. In other words, mouse songs do not seem to involve imitative learning. Results {#s2} ======= Strain differences in ultrasonic songs {#s2a} -------------------------------------- ### Song parameters {#s2a1} When a male subject encountered a female, he emitted complex ultrasounds. Sound spectrograms demonstrated that B6 males showed a peak at 70--80 kHz, and BALB males at 50--60 kHz ([Fig. 1a](#pone-0017721-g001){ref-type="fig"} and [Audio S1](#pone.0017721.s003){ref-type="supplementary-material"} and [S2](#pone.0017721.s004){ref-type="supplementary-material"}). The comparison between B6 and BALB mice revealed that the average peak frequency of syllables was lower in BALB males (Mann--Whitney test, p\<0.005), the average interval between syllables was longer in BALB males (Mann--Whitney test, p\<0.005), but the number and duration of syllables ([Fig. 1b](#pone-0017721-g001){ref-type="fig"}) emitted in the 3-min test did not differ significantly (B6, 240±45 times/min; BALB, 257±32 times/min). ::: {#pone-0017721-g001 .fig} 10.1371/journal.pone.0017721.g001 Figure 1 ::: {.caption} ###### Strain-specific characteristics of male mice songs. \(a) Sound spectrograms of ultrasonic songs in B6 (upper) and BALB (lower) male mice. B6 males showed a higher peak frequency of syllables ranging from 70--110 kHz, shorter intervals between syllables, and more upward frequency modulations with jumps (arrows), whereas BALB males produced more "chevron" and "harmonics" syllables (arrow head). (b) The mean syllable peak frequency and inter-syllable interval significantly differed between B6 and BALB mice, but syllable duration was not. Data are expressed as mean ± SEM; \*p\<0.05 between strains. (c) Pie graphs showing percentages of the 10 categories of song syllables in B6 and BALB mice. Percentages were calculated in each strain as the number of syllables in each category for each subject/total number of syllables analyzed in each subject. The number of total syllables analyzed was: 6179 for B6 mice and 6244 for BALB mice. B6 mice produced more "short," "one jump," and "more jumps" syllables than BALB mice, whereas BALB mice produced more "flat", "chevron", "complex", and "harmonics" syllables; \*p\<0.05 between strains. (d) In the sequential analysis, we divided all syllable types into 2 categories, namely, A (syllables with frequency jumps) and B (syllables without jumps). Z indicates silent gaps longer than 0.25 s. Circles represent the percentage of syllable types, and the thickness of the arrows represents the transition probabilities. The sequential analyses of syllables demonstrated strain-specific patterns; B6 mice showed more transition from A to A, A to B, A to Z, B to A, and Z to A than BALB mice and BALB mice showed more B to B self transition compared to that in B6 mice; \*p\<0.05 between strains. :::  ::: ### Syllable category analysis {#s2a2} According to previous studies [@pone.0017721-Scattoni1], each syllable was identified as 1 of 10 distinct categories: "upward," "flat," "chevron," "complex," "more jumps," "downwards," "short," "wave," "one jump," or "harmonics". The resulting pie graph indicated strain differences in the distribution of syllable categories ([Fig. 1c](#pone-0017721-g001){ref-type="fig"}). MANOVA, with the strain as the main factor and the probabilities of each syllable occurrence (10 in total) as dependent variables, revealed a significant between-group difference (F(9,3) = 69.7, p\<0.0001). A post hoc t-test showed strain differences in 8 of the 10 syllable categories ([Fig. 1c](#pone-0017721-g001){ref-type="fig"}). B6 mice produced more "upward," "short," "one jump," and "more jumps" syllables than BALB/c mice (p\<0.05, t-test). In contrast, BALB/c mice produced more "flat," "chevron," "complex," and "harmonics" syllables (p\<0.05, t-test). ### Sequential analyses of syllables {#s2a3} The sequential patterns of B6 and BALB mice songs are shown in [Fig. S2](#pone.0017721.s002){ref-type="supplementary-material"}. All 10 syllable categories were included in the analysis. Because these patterns were overly complicated for rigorous analysis, the syllable categories were lumped into 2 large types, namely, syllables with jumps (A) and other syllable types (B). The gap (more than 0.25 s) between each syllable bout is represented by Z. B6 and BALB mice showed distinct transitional patterns of the song syllables. MANOVA, with the strain as the main factor and the probabilities of each syllable transition (8 in total) as dependent variables, revealed a significant strain difference in the transition patterns (F(7,5) = 4.92, p\<0.05). Post hoc t-tests showed a greater occurrence of transitions from types A to A, A to B, B to A, A to Z, and Z to A in B6 than in BALB mice, whereas BALB mice showed more B to B self-transition compared to B6 mice ([Fig. 1d](#pone-0017721-g001){ref-type="fig"}). Comparison between the fostered groups and naturally-reared sons {#s2b} ---------------------------------------------------------------- ### Sonograms {#s2b1} Sound spectrograms demonstrated that B6-sons and B6-foster males showed a peak at 70--80 kHz, whereas BALB mice showed a peak at 50--60 kHz ([Fig. 2](#pone-0017721-g002){ref-type="fig"} and [Audio S3](#pone.0017721.s005){ref-type="supplementary-material"}, [S4](#pone.0017721.s006){ref-type="supplementary-material"}, [S5](#pone.0017721.s007){ref-type="supplementary-material"} and [S6](#pone.0017721.s008){ref-type="supplementary-material"}). ::: {#pone-0017721-g002 .fig} 10.1371/journal.pone.0017721.g002 Figure 2 ::: {.caption} ###### Sonograms of ultrasonic songs in fostered males. Sonograms of ultrasonic songs recorded from B6-son, B6-foster, BALB-son, and BALB-foster male mice. Cross-fostered mice showed similar patterns to those of normally reared mice, and the effects of the rearing environment were not obvious. B6-son and B6-foster mice showed a higher peak frequency of syllables, shorter intervals between syllables, and more upward frequency modulations with jumps (arrows), whereas BALB-son and BALB-foster males produced more "chevron" and "harmonics" syllables (arrow head). :::  ::: ### Song parameters {#s2b2} We compared songs between fostered groups, and found that the main strain differences we quantified were not affected by fostering. BALB cross-fostered males still showed a lower peak frequency (F(1,20) = 106.5, p\<0.0001) and longer inter-syllable intervals (F(1,20) = 9.67, p\<0.01) than B6-fostered males ([Fig. 3](#pone-0017721-g003){ref-type="fig"}). The syllable duration and the number of syllables emitted in the 3-min test were equivalent in all groups (B6-son, 225±56 times/min; B6-foster, 242±45 times/min; BALB-son, 225±33 times/min; BALB-foster, 249±37 times/min). ::: {#pone-0017721-g003 .fig} 10.1371/journal.pone.0017721.g003 Figure 3 ::: {.caption} ###### Song parameters of fostered males. Song parameters in B6-son, B6-foster, BALB-son, and BALB-foster male mice. The distribution histogram of the peak frequency (a) and intervals (b), but not the duration (c), of the syllables demonstrated significant strain differences, regardless of the fostering. Mean peak frequency (d) and interval (e) significantly differed between genetic B6 and BALB groups. Data are expressed as mean ± SEM; \*p\<0.05 vs. B6-son and B6-foster mice. :::  ::: ### Syllable category analysis {#s2b3} MANOVA revealed a significant effect of strain (F(9,9) = 25.9, p\<0.0001), but not of fostering (F(9,9) = 0.91, p = 0.55) or an interaction of strain and fostering (F(9,9) = 0.41, p = 0.89). Regardless of fostering experience, B6 mice produced more "short," "one jump," and "more jumps" syllables than BALB mice ([Fig. 4](#pone-0017721-g004){ref-type="fig"}, p\<0.05). In contrast, BALB mice produced more "flat," "chevron," "complex," and "harmonics" syllables ([Fig. 4](#pone-0017721-g004){ref-type="fig"}, p\<0.05). The proportions of syllables within each category are shown in [Fig. 5](#pone-0017721-g005){ref-type="fig"} which indicates that the differences in the appearance of syllable categories were mainly dependent on the strain of the mice. ::: {#pone-0017721-g004 .fig} 10.1371/journal.pone.0017721.g004 Figure 4 ::: {.caption} ###### Appearance ratio of the song syllables in fostered males. The appearance ratio of each of the 10 syllable categories in B6-son, B6-foster, BALB-son, and BALB-foster mice. Genetic B6 groups produced more "short," "one jump," and "more jumps" syllables than BALB/c mice, whereas genetic BALB groups produced more "flat," "chevron," "complex," and "harmonics" syllables. Data are expressed as mean ± SEM; \*p\<0.05 vs B6-son and B6-foster mice. :::  ::: ::: {#pone-0017721-g005 .fig} 10.1371/journal.pone.0017721.g005 Figure 5 ::: {.caption} ###### Distribution pattern of the song syllables in fostered males. Pie graphs showing the percentages of the 10 categories of song syllables in B6-son (a), BALB-son (b), B6-foster (c), and BALB-foster (d) mice. Percentages were calculated in each strain as the number of syllables in each category for each subject/total number of syllables analyzed in each subject. The total syllables determined are as follows: 5487 syllables; B6-son; 6414 syllables, B6-foster; 4973 syllables, BALB-son; 6963 syllables, BALB-foster. :::  ::: ### Sequential analyses of syllables {#s2b4} Regardless of fostering, B6 and BALB mice showed distinct transitional patterns of the song syllables, and these characteristics were displayed by cross-fostered males. MANOVA revealed a strain difference (F(6,12) = 24.6, p\<0.0001), but no fostering effect (F(6,12) = 0.655, p = 0.687) and no interaction between these (F(6,12) = 1.56, p = 0.241). A Bonferonni post hoc test revealed that sons of BALB mice showed a greater occurrence of B to B self-transitions, B to Z and Z to B transitions as well as a lower occurrence of A to A self-transitions, A to B, B to A, A to Z, and Z to A transitions compared to sons of B6 and B6-foster male mice (p\<0.05, [Fig. 6](#pone-0017721-g006){ref-type="fig"}). BALB-foster mice demonstrated a greater occurrence of B to B self-transitions and a lower occurrence of A to A, A to B, B to A, A to Z, and Z to A transitions compared to sons of B6 and B6-foster mice (p\<0.05, [Fig. 6](#pone-0017721-g006){ref-type="fig"}). ::: {#pone-0017721-g006 .fig} 10.1371/journal.pone.0017721.g006 Figure 6 ::: {.caption} ###### Sequential analysis of syllables in fostered males. Sequential analyses of syllables demonstrated strain-specific patterns; BALB-son mice showed a greater occurrence of B to B self transitions, and B to Z and Z to B transitions, as well as a lower occurrence of A to A self-transitions and A to B, B to A, A to Z, and Z to A transitions compared to B6-son and B6-foster mice. BALB-foster mice demonstrated greater occurrence of type B to B self-transitions and a lower occurrence of A to A, A to B, B to A, A to Z, and Z to A transitions compared to B6-son and B6-foster mice. Circles represent the percentage of syllable types, and the thickness of the arrows represents the transition probabilities; \*p\<0.05 vs. B6-son and B6-foster mice. :::  ::: Discussion {#s3} ========== In the present study, we revealed that B6 and BALB male mice showed distinct patterns and sound profiles of songs when encountering a female. Our syllable categories are similar to those reported in earlier studies [@pone.0017721-Holy1], [@pone.0017721-Scattoni1]--[@pone.0017721-Panksepp1]. The average peak frequency of syllables was higher in B6 mice, and the average interval between syllables was shorter in B6 mice. In addition, B6 mice produced more "upward," "short," "one jump," and "more jumps" syllables than BALB mice; whereas BALB mice produced more "flat," "chevron," "complex," and "harmonics" syllables. By using the cross-fostering procedure, we further showed that these strain differences remained even after the pups were cross-fostered to another strain of parents, suggesting that the strain-specific song profile is determined by genetic factors and is independent of the juvenile social auditory environment. Studies of the natural history of mice have demonstrated that a pair of male and female mice lives in a nest together with their juveniles [@pone.0017721-Crowcroft1]. In the laboratory the female goes into estrus around the time of delivery as indicated by an increase in estrogen levels, the so-called postpartum estrus [@pone.0017721-Conner1], [@pone.0017721-Connor1]. The odor of female urine stimulates the males to sing [@pone.0017721-Holy1]; therefore, the pups can be exposed to male songs especially when the mother comes into the round of the reproductive cycle. Since the mouse can hear from at least postnatal day 10 [@pone.0017721-Shnerson1], pups in the juvenile period have sufficient opportunity to be exposed to adult male songs. To be sure, we recorded in the laboratory one pair of B6 and one pair of BALB male and female continuously every day for three weeks after pup delivery, and found that the pair generated over 200 seconds of vocalizations each day, and these were always during the dark period (unpublished data). Thus, although not directly measured, we believe that the male mice in the cross fostering studies sang during the cross-fostering period. In this study, the fostered mice were housed in mixed strains of the same age in the post-weaning period, to standardize the possibility of hearing songs from littermates. If the strain differences reported in [Fig. 1](#pone-0017721-g001){ref-type="fig"} had been the result of learning, our methods would have ensured that this rearing condition was sufficient to establish such strain differences. In fact, this is a general breeding condition utilized by most laboratories [@pone.0017721-Bernstein1]. Therefore, our procedures should have been able to detect the effect of cross-fostering rearing environments, but we did not observe any of such effects. Several studies have demonstrated that female mice show attraction to male songs [@pone.0017721-Hammerschmidt1], [@pone.0017721-Pomerantz1]. In these studies, however, a 2-choice test presenting 2 types of songs was not conducted; therefore, it remains a question whether female mice have a preference for a specific character of songs, as shown in songbirds [@pone.0017721-Miller1]. Furthermore, female mice have been shown to respond to synthetic 70 kHz ultrasounds presented behind a devocalized male mouse [@pone.0017721-Pomerantz1] and to pup vocalizations [@pone.0017721-Uematsu1], in which the observed syllable categories are similar to adult male songs [@pone.0017721-Scattoni2]. In a recent study, female mice were shown to be able to distinguish between a familiar male song and an unfamiliar one based on the social experience of a short-term encounter and showed investigative behavior toward the unfamiliar song, implying that female mice can distinguish the individual profile of the songs [@pone.0017721-Musolf1]. These results suggest that a certain level of ultrasound complexity is sufficient to attract female mice, although the value of learning songs for male mice to achieve reproductive success remains unclear. Recent studies have demonstrated that ultrasonic vocalization of mouse pups is affected by genes related to neuropsychiatric disorders such as Autism [@pone.0017721-Scattoni1], [@pone.0017721-Nakatani1]. These genetic approaches could reveal the genes that regulate ultrasonic vocalization in mice. For example, the function of Foxp2, a transcription factor shown to be related to a human language disorder [@pone.0017721-Enard1], is involved in pup isolation calls. When human-type FoxP2 was inserted into the mice genome, isolation call pitch increased [@pone.0017721-Enard2]. In addition, when FoxP2 was knocked out [@pone.0017721-Shu1] or a FoxP2 mutation corresponding to the human language disorder was knocked in [@pone.0017721-Fujita1], the number of isolation calls decreased. However, these transgenic mice were tested with maternal separation-induced pup ultrasound vocalizations, not with male courtship songs. Therefore, it is of interest to test whether these genetically modified mice would show the quantitative and qualitative differences in adult male songs we observed. The complexity of the song pattern itself raises an interest in understanding the neural and molecular mechanisms controlling song in mammals. Here we showed that imitative vocal learning is not involved in the strain specificity of mouse songs. Vocal learning requires two independent processes. First, the animal must have voluntary control over the vocal output. Second, the animal should be able to match its vocal output with the externally acquired auditory memory. For the first process, the existence of the direct motor pathway connecting the oro-facial motor cortex and the medullar phonatory and respiratory areas, including the nucleus ambiguus, has been suggested as an anatomical substrate responsible for vocal plasticity [@pone.0017721-Jarvis1]. In fact, this cortico-bulber pathway for vocal plasticity exists in humans but not in non-human primates [@pone.0017721-Jurgens1]. This pathway is also found in oscine songbirds such as the zebra finch and the canary but not in pigeons [@pone.0017721-Wild1]. Since humans and oscine songbirds are vocal learners and non-human primates and pigeons are vocal non-learners, the existence of this pathway coincides with vocal learning. Arriaga et al. reported singing-related gene expression in mice cingulated, motor cortex and basal ganglia [@pone.0017721-Arriaga1]. They also reported the existence of the cortico-bulber pathway for vocal plasticity in mice [@pone.0017721-Arriaga2], which should be related to the observed vocal complexity of the mouse song. Our data, which show no effect of the auditory environment by tutors on mouse song, may appear contradictory to these findings. However, vocal plasticity alone does not guarantee vocal learning, since vocal-auditory matching is also required for vocal learning to occur. A certain degree of voluntary vocal plasticity may be necessary in animals with complex vocalizations to maintain a stable performance even without learning. It may be interesting to examine the anatomical pathways in animals with complex vocalizations but without learning abilities, including sub-oscines [@pone.0017721-Kroodsma1] and gibbons [@pone.0017721-Marler2]. Further, even though there was no clear evidence of vocal learning in the mice examined in this study, there may be other factors that modulate the phonetic and sequential variability of male songs. It is often assumed that highly variable songs are suggestive evidence of vocal learning. As seen from our sonograms and sequence analyses, mice songs are highly variable yet we find evidence that they are innate. This variability could be generated by a random pattern generator independent of learning or by some hormonal influence [@pone.0017721-Nunez1]. In either case, our findings indicate that the presence of variability does not automatically mean the presence of vocal learning. Conclusion {#s3a} ---------- Our results show that the auditory environment does not affect song phenotypes in mice, and, thus, vocal learning does not appear to be involved in mouse songs. Nevertheless, mouse song is a very complex behavior, with at least 10 categories of vocal tokens and complex note-to-note transition rules. Even if this phenotype is largely controlled by genetic factors and only limited learning is involved, we can still pose interesting questions regarding the genetic encoding of acoustic categories and the neural mechanisms involved in sequence generation. Thus, the mouse song should remain an important model in which to study the biological basis of complex communicative behavior, including spoken human language. Materials and Methods {#s4} ===================== Animals {#s4a} ------- BALB/cAJcl (BALB) and C57BL/6JJcl (B6) mice were originally obtained from Japan Clea Co. Ltd. (Japan Clea, Yokohama, Japan) and bred in our laboratory. Food and water were given ad libitum, and all the animals were kept at a constant temperature (23±1°C) and humidity (40%±10%) under a 12-h light:dark cycle (light on at 0600). All experiments were conducted in accordance with the guideline of the \"Policies Governing The Use of Live Vertebrate Animals\" by Azabu University, and were approved by The Ethical Committee for Vertebrate Experiments (ID\# 070418). Pairing and cross-fostering {#s4b} --------------------------- A male and a female mouse of the same strain were pair-housed in a cage (17.5 cm × 24.5 cm × 12.5 cm) for breeding. When the female was pregnant, delivery was examined every 6--8 hours. When newly born pups were found at the same time in both strains of parents, a part of the litter was reciprocally cross-fostered to parents of the other strain of mice (B6-foster and BALB-foster). The control mice were handled in the same manner as fostered pups but returned to their own parents (B6-son and BALB-son). All litters were left undisturbed until weaning (postnatal day (PD) 21). After PD21, they were housed with males of the non-cross fostered controls of the different strain until ultrasound recording at 10--20 weeks of age ([Fig.7](#pone-0017721-g007){ref-type="fig"}). The number of animals and litters (animals/litters) used in this experiment were as follows: B6 (6), BALB (7), B6-son (5/4), B6-foster (5/3), BALB-son (5/4), and BALB-foster (6/5). Because it is known that mating can affect the vocal morphology of male songs, we separately analyzed strain differences between B6 and BALB mice in sexually experienced males and strain and environmental effects between cross-fostered and naturally reared B6 and BALB mice in sexually inexperienced males. ::: {#pone-0017721-g007 .fig} 10.1371/journal.pone.0017721.g007 Figure 7 ::: {.caption} ###### Timeline of the cross-fostering procedure. This figure illustrates the case of cross-fostering from BALB to B6. When newly born pups were found at the same time in both strains of parents, a part of the litter was reciprocally cross-fostered to parents of the other strain of mice. The control mice were handled in the same manner as fostered pups but returned to their own parents. All litters were left undisturbed until weaning (PD21). After weaning,they were housed with males of the non-cross fostered controls of the different strain until ultrasound recording at 10--20 weeks of age. :::  ::: Ultrasound recording {#s4c} -------------------- All experiments were carried out in a soundproof chamber (Muromachi Kikai, Tokyo, Japan) under a red dim light, from 1300 to 1700 hours. Ultrasonic sounds were detected using a condenser microphone (UltraSoundGate CM16/CMPA, Avisoft Bioacoustics, Berlin, Germany) designed for recordings between 10 and 200 kHz. The microphone was connected to an A/D converter (UltraSoundGate 116, Avisoft Bioacoustics, Berlin, Germany) with a sampling rate of 300 kHz and acoustic signals were transmitted to a sound analysis system (SASLab Pro, Avisoft Bioacoustics, Berlin, Germany). During the recording, a subject male mouse was individually housed in a test cage (12.5 cm × 20.0 cm × 11.0 cm) and kept there for at least 2 h for habituation. The test cage was placed in the soundproof chamber, and a female mouse, devocalized by unilateral sectioning of the inferior laryngeal nerve [@pone.0017721-Hammerschmidt1], was introduced into the test cage. The ultrasound was recorded for 3 min, and the data were later analyzed. Ultrasound analysis {#s4d} ------------------- Spectrograms were generated with an FFT-length of 1024 points and a time-window overlap of 75% (100% frame, Hamming window). The spectrogram was produced at a frequency resolution of 488 Hz and a time resolution of 1 ms. A lower cut-off frequency of 20 kHz was used to reduce background noise outside the relevant frequency band. Parameters analyzed for each subject included the number of syllables, duration of syllables, and qualitative and quantitative analyses of sound frequencies measured in terms of frequency at the maximum of the spectrum. Waveform patterns of calls collected from every group (B6, 6179 syllables; BALB, 6244 syllables; B6-son, 5487 syllables; B6-foster, 6414 syllables; BALB-son, 4973 syllables; BALB-foster, 6963 syllables) were analyzed in detail. Each syllable was identified as 1 of 10 distinct categories, based on internal pitch change, length, and shape, according to previously reported categories with minor modifications ([Fig. S1](#pone.0017721.s001){ref-type="supplementary-material"}) [@pone.0017721-Clayton1]. The classification of the 10 categories of ultrasonic vocalization syllables is described in the [Results](#s2){ref-type="sec"} section. The frequency of appearance of each category was compared between the groups. In order to confirm the categorization, a likelihood ratio test examining whether there was a systematic difference between the 2 blind experimenters was performed by a generalized linear model that consisted of an explanatory variable (number of syllables) and 3 response variables (2 operators, 11 categories of syllables, and 6 mice). No significant difference was found between the 2 operators (quasi-Poisson error, log link, total number of all syllables in each mouse as offset, *F* ~(1,115)~ = 0.13, *p* = 0.72). The occurrence of each syllable was compared between groups using MANOVA, followed by a Bonferonni post-hoc test. Sequential analysis of syllables {#s4e} -------------------------------- The prevalence of a syllable type was defined as follows on the basis of a previous study [@pone.0017721-Clayton1]: the syllable types with jumps (1 jump, more jumps) were denoted as A, with all other syllable types denoted as B, and the gap (more than 0.25 s) was Z. One-to-one transition probabilities between these 3 categories were analyzed and indicated by diagrams (Eureka version 1.0 <http://sites.google.com/site/eurekawiki/>). The occurrence of each transition type was compared between groups using MANOVA, followed by a Bonferonni post-hoc test. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Song syllable characteristics.** Ten categories were defined as follows. Upward: duration of 5--50 ms, frequency increaseof more than 5 kHz from starting point to end. Downward: duration of 5--50 ms, frequency decrease of more than 5 kHz from starting point to end. Flat: duration of 5--35 ms, frequency difference of less than 5 kHz between starting point and end. Short: duration of less than 5 ms, frequency difference of less than 5 kHz between starting point and end. Chevron: duration of 15--80 ms, frequency increase of more than 5 kHz from starting point to frequency peak and frequency increase or decrease of more than 5 kHz from frequency peak to end (\*; frequency peak). Wave: duration of 15--100 ms, frequency increase or decrease of more than 5 kHz from starting point to the first frequency peak (or bottom) and containing 1 frequency peak and 1 frequency bottom (\*; frequency peak and bottom). Complex: duration of 30--150 ms, frequency increase or decrease of more than 5 kHz from starting point to the first frequency peak (or bottom) and containing more than 3 frequency peaks and/or frequency bottoms that differ from each other by more than 5 kHz in frequency (\*; frequency peak and bottom). One jump: duration of 10--50 ms and containing 1 frequency gap (\#; frequency gap, less than 1 ms and more than 5 kHz frequency difference). More jumps: duration of 15--100 ms and containing more than 2 frequency gaps (\#; frequency gap). Harmonics: duration of 10--100 ms and containing more than 2 Chevron, Wave, Complex, One jump, or More jumps syllables in parallel with a main syllable that has the highest dB count. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Sequential analysis of syllable types in B6 and BALB mice.** The sequential analyses of 10 categories of syllables demonstrated a very complicated transition both in B6 (upper) and BALB (lower) mice. a: upward, b: downward, c: flat, d: short, e: chevron, f: wave, g: complex, h: one jump, i: more jumps, j: harmonics, Z: gap. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Audio S1 ::: {.caption} ###### **B6 male song.** (WAV) ::: ::: {.caption} ###### Click here for additional data file. ::: Audio S2 ::: {.caption} ###### **BALB male song.** (WAV) ::: ::: {.caption} ###### Click here for additional data file. ::: Audio S3 ::: {.caption} ###### **B6-son male song.** (WAV) ::: ::: {.caption} ###### Click here for additional data file. ::: Audio S4 ::: {.caption} ###### **B6-foster male song.** (WAV) ::: ::: {.caption} ###### Click here for additional data file. ::: Audio S5 ::: {.caption} ###### **BALB-son male song.** (WAV) ::: ::: {.caption} ###### Click here for additional data file. ::: Audio S6 ::: {.caption} ###### **BALB-foster male song.** (WAV) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Drs. Erich Jarvis and Gustavo Arriaga of Duke University for generously showing their preliminary data and providing the discussion on this experiment. We are grateful to Professor Björn Brembs for suggestions and encouragement and to Dr. Olga Feher and Benjamin Treuhaft for English proof reading. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by Japan Society for the Promotion of Science (T.K. and R.N.), the Promotion and Mutual Aid Corporation for Private Schools of Japan, Grant-in-Aid for Matching Fund Subsidy for Private Universities (T.K.), RIKEN Brain Science Institute (K.O.), and by "Exploratory Research for Advanced Technology, Okanoya Emotional Information Project" (K.O.) from Japan Science and Technology Corporation. Funders paid for experimental supplies and personnel cost for this research. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: TK KN. Performed the experiments: KN RN. Analyzed the data: KN RN MN KM. Wrote the paper: TK KN RN KO.

PubMed Central

2024-06-05T04:04:19.250627

2011-3-9

PMC3052374

Introduction {#s1} ============ New generations of massively parallel sequencers have revolutionized the field of sequencing. With the dramatic reduction of sequencing cost and the ongoing development of new high throughput sequencers, the possibility of investigating hundreds of gigabases in a single sequencing run has come into effect. The throughput of recently developed sequencers, the HiSeq 2000 (Illumina) and SOLiD 4hq system (Life Technologies), makes it possible to, for instance, RNA-profile a couple of hundred samples in parallel while producing 10 million reads per sample. Targeted enriched cDNA libraries have also been shown to be an effective method for expression profiling [@pone.0017785-Zajac1], [@pone.0017785-Li1], [@pone.0017785-Fan1] where massive sequencing could be used as readout with the possibility to screen thousands of samples in parallel. In addition, targeted highly multiplex amplification methods, such as molecular inversion probes [@pone.0017785-Nilsson1], [@pone.0017785-Hardenbol1], GoldenGate [@pone.0017785-Fan2] and trinucleotide threading [@pone.0017785-Pettersson1], combined with massive sequencing is a powerful way of genotyping large sets of markers and samples at the same time [@pone.0017785-Pettersson2]. In population genetics these platforms make it possible to tag and in parallel sequence highly polymorphic loci such as the MHC gene complex [@pone.0017785-Wegner1], [@pone.0017785-Bentley1], [@pone.0017785-Babik1]. Retrieving genotype data from large cohorts is essential for studies aiming to investigate a chain of occurrences by studying stochastic events, such as mutations. The continuously growing sequencing throughput enables researchers to analyze more and more samples within the same experiment, hence creating a need for scalable and manageable protocols for preparing and typing these large sets of samples. In order to sequence a huge number of samples in a single sequencing run, a stable and reliable method for identifying the generated reads is required. The introduction of DNA barcodes by PCR or by ligation has previously been described as methods to distinguish the origin of each read in a set of mixed sequences [@pone.0017785-Binladen1], [@pone.0017785-Meyer1], [@pone.0017785-Erlich1], [@pone.0017785-Parameswaran1], [@pone.0017785-Galan1]. In this study we demonstrate how the sample pooling and identification procedures can become more reliable and less complicated, enabling larger experimental designs. In addition to the complexity of designing and performing highly multiplex experiments, the cost of reagents and the time of protocol executions often limit the experiments. By setting up an automated protocol utilizing third party reagents, we describe how both time and cost for sample processing prior to sequencing can be substantially reduced. We describe a pooling strategy able to produce an even spread of reads across the barcoded samples, addressing the growing need of multiplex genotyping. The technique described in this work utilizes a combination of two tags to identify the origin of a certain read and thereby reduces the number for unique tags needed for a given number of samples. Instead of incorporating one identification tag at the end of the amplicon, a first tag is incorporated at both ends of PCR fragments and then a second tag is ligated to a pool of 96 samples originating from a 96-well PCR plate. By using this system the complexity of sample handling is greatly reduced and the need for making unique primer combinations for each sample is eliminated. The method was tested on 52 indexed plates, allowing analysis of 4992 samples, by sequencing the 270 bp 2^nd^ exon of the highly polymorphic DLA-DRB1 gene using the 454 GS FLX Titanium Chemistry [@pone.0017785-Margulies1]. By taking advantage of this two tagging system we demonstrate that it is possible to readily sequence very large cohorts in parallel. Materials and Methods {#s2} ===================== Samples and DNA preparation {#s2a} --------------------------- The sample set consisted of a total of 4708 dogs and wolves: Animal Ethics Committee, The State Provincial Office of Southern Finland, ESLH-2009-07827/Ym-23; Federal Fish and Wildlife Permit Number MA053639-1; Endangered/Threatened Species Wildlife and CITES Appendix II Permit No. 09US222450/9; University of New South Wales *Care and Ethics Committee (ACEC)* Project 05/74A. 2059 of the samples were collected by FTA-cards, 819 were hair samples and the remaining 1830 were blood samples. The DNA from hair and blood was extracted as described by Savolainen *et al.* [@pone.0017785-Savolainen1] while the FTA-cards (Whatman, Maidstone, UK) were prepared according to the manufacturer\'s instructions (using a 2 mm punch) with the exception that the washing volumes were adjusted to 100 µl. A Magnatrix™ 1200 (NorDiag ASA, Oslo, Norway) liquid handling robot was used for washing 96 pellets in parallel using sample specific tips and washing solution containers. After the final wash, the pellets were dried at 56°C for 10 minutes. Primers {#s2b} ------- DLA-DRB1 exon 2 primer sequencess were obtained from Kennedy *et al.* [@pone.0017785-Kennedy1]. To distinguish the different samples in each plate, 96 pairs of DLA-DRB1 exon 2 primers with different tags were designed (supplementary [table S1](#pone.0017785.s002){ref-type="supplementary-material"}). The tagging sequences were obtained from Meyer *et al.* [@pone.0017785-Meyer1]. Tags were chosen to not contain any homo polymeric combinations and to differ from each other in at least three positions. 8 bases of identification sequence were used for the position-tags. The primers were synthesized by Thermo Scientific (Wilmington, DE, USA). Tagging by PCR {#s2c} -------------- Touch-down PCR was performed in a total volume of 50 µl containing 1 U Platinum Taq DNA polymerase (Invitrogen, Carlsbad, Ca, USA), 1× Taq Platinum DNA polymerase buffer (Invitrogen), 1.5 mM MgCl~2~, 0.2 mM dNTPs and 0.2 µM of the forward and reverse primers, respectively. After initial activation at 95°C for 2 min the reaction was cycled through 14 cycles of touch-down starting at 95°C for 30 seconds, 62°C for 1 min and 72°C for 1 min., and lowering of annealing temperature by 0.5°C for each cycle making the last cycle anneal at 55°C. Then, depending on template source, 40 or 45 cycles of 95°C for 30 seconds, 55°C for 1 min and 72°C for 1 min followed and the reaction ended in a final extension of 10 min at 72°C. For FTA-card samples the number of post touch-down cycles were 45 and for blood and hair samples it was 40. About half of the PCR-products were examined by gel electrophoresis or capillary electrophoresis using the MCE®-202 MultiNA Microchip Electrophoresis System (Shimadzu Biotech, Kyoto, Japan). Pooling and clean-up {#s2d} -------------------- For each plate, 5 µl of each sample was pooled and 250 µl of the resulting mixture was applied to a QIAquick™ PCR clean up column (Qiagen, Hilden, Germany) and treated according to the manufacturer\'s procedures for PCR clean up and eluted in 30 µl of Qiagen elution buffer (EB). The concentrations of the cleaned pools were measured using Invitrogen\'s Quant-iT™ dsDNA BR kit. The concentration of each pool was then adjusted to 1 µg DNA/34 µl EB. Automated MID ligation and sequencing {#s2e} ------------------------------------- 52 pools were indexed using the extended set of 454 Multiplex Identifier (MID) adaptors for the GS FLX Titanium Chemistry (supplementary [table S2](#pone.0017785.s003){ref-type="supplementary-material"}). Adaptors were synthesized by Thermo Scientific. All steps of the library preparation were automated and performed using a Magnatrix™ 1200 Biomagnetic Workstation (NorDiag) capable of running custom made scripts as described previously [@pone.0017785-Lundin1] with the exception that the library reagents including end repair, ligation and fill in reagents were purchased from New England Biolabs (NEBNext End Repair Module, NEBNext Quick Ligation Module and Fill-in and ssDNA Isolation Modules) (New England Biolabs, Ipswich, MA, USA). The quality of each library was examined after ligation using Experion 1k DNA analysis kit (BioRad, Herculees, CA, USA). Three equimolar pools were made from the 52 adaptor indexed libraries and single stranded DNA was isolated from each pool. The emulsion PCR titration by quantification, amplification and sequencing were performed using the 454 GS FLX Titanium Chemistry according to the manufacturer\'s instructions. Each library pool was sequenced on a separate lane using a total of three lanes in a four-lane setup. Data analysis {#s2f} ------------- The generated reads were sorted by their plate- and position-tags into individual folders and the reading direction were normalized using an in-house developed BioPerl [@pone.0017785-Stajich1] script. All reads corresponding to each of the individuals were separately aligned using MUSCLE [@pone.0017785-Edgar1] generating one alignment per individual. Starting from these sequence alignments, one or two consensus sequences were generated for each individual depending on if the individual was homozygous or heterozygous respectively. To compensate for pyrosequencing errors in homo polymeric regions, the generated consensus sequences were aligned against a reference sequence and deletions in the sequencing data were corrected. Results and Discussion {#s3} ====================== The introduction of massively parallel sequencing platforms has opened up entirely new possibilities in all fields of functional genomics. However, the capacity of these platforms greatly exceeds the needs in most studies involving targeted profiling and typing. In an attempt to better exploit the capacity of these systems and reduce the cost per sequenced sample, indexing procedures has been introduced, allowing multiplex identification and sorting of over 100 samples [@pone.0017785-4541], [@pone.0017785-Applied1]. However, increasing the sample size to thousands necessitates additional tagged adaptors leading to preparation of thousands of libraries, which is time consuming and expensive. This problem has been addressed by Erlich *et al.* [@pone.0017785-Erlich1] by implementing a sample-pooling pattern combined with tagging of the pools and Galan *et al.* [@pone.0017785-Galan1] by combination of tags in the forward and reverse PCR primers. Although these studies demonstrated that the number of unique tags does not necessarily need to match the number of samples, they considerably increase the complexity of sample handling by employing experimentally complicated sample or primer mixing procedures. In this study we present a two tagging strategy that employs a combination of two tags, added in two steps, which not only enables accurate multiplex analysis of thousands of samples in parallel, using a reduced number of tags, but also reduces the complexity of sample handling. The principle of the two tagging strategy is depicted in [figure 1](#pone-0017785-g001){ref-type="fig"}. The tags are added in two steps. First, the target is amplified by tagged PCR primers giving amplicons that are tagged in both ends (denoted position-tags). By using 96 position-tags and strictly handle them in a 96-well PCR-plate, 96 samples are uniquely tagged in a single round. The PCR products from a plate are then pooled and a second tag (denoted plate-tag), which is incorporated next to the sequencing primer, is ligated to the pooled samples. The degree of sample multiplexity is thus substantially increased by adding up indexed plates. The number of required tags can be described as *tags = (N/96)+96*, where N is the total number of samples (so that N/96 expresses the number of required plate-tags while the constant 96 is the number of position tags in the PCR plate). However, if more than one sequencing lane is utilized, the number of tags is reduced to *tags = (N/L\*96)+96* where L is the number of lanes used for sequencing the sample set. In this study, N was set to 4992 and L to 2, giving a total of 122 (26+96) tags. ::: {#pone-0017785-g001 .fig} 10.1371/journal.pone.0017785.g001 Figure 1 ::: {.caption} ###### The tagging procedure. 96 tagged primer pairs are designated to each position in a 96-well PCR plate grid giving one specific tag for each position, referred to as position tags (shown as red). Sets of 96 samples are tagged and amplified hence, giving N/96 plates (where N is the total number of samples). All products within each plate are pooled and a second tag, denoted plate tag (shown as green) together with the sequencing adaptors (shown as blue), is ligated to the pool. The plate-indexed products are then pooled and amplified prior to sequencing. :::  ::: To prove the concept of the two tagging approach, the 2^nd^ exon of the DLA-DRB1 gene in dogs and wolves was amplified using the 96 position-tags in 52 plates. Out of the 4992 wells, samples were added to 4708, while 284 wells were used for negative PCR controls. After gel electrophoresis, the number of successful PCR products was estimated to 3700 (a success rate of about 79%). The success rate of the PCR was largely dependent on the quality of the samples. For example, 59% of the DNA samples originated from hair and 85% from FTA cards resulted in detectable products. The rate of successfully amplified samples would however increase if the quality of all samples were checked prior to amplification. In addition, two of the position-tagged primers (D11 and D12 (supplementary [table S1](#pone.0017785.s002){ref-type="supplementary-material"})) did not give any products at all and position-tagged primers C3 and G9 performed poorly. This is probably due to poor primer synthesis and since the position-tags exist in all 52 plates, over 150 of the failed PCR reactions could be related to these primers. After initial PCR amplification, the library pooling strategy was deployed, where each library was pooled prior to ssDNA isolation, enabling time and cost effective sample processing ([figure 2](#pone-0017785-g002){ref-type="fig"}). Worth mentioning, to save time and reagents, no concentration measurement of the 4992 PCR products was done. The 52 libraries could be processed in two days using an automated setup, and using third party reagents substantially reduced the cost. Uneven read distribution between indexed libraries is a common problem with increasing degree of multiplexing, raising high requirements on concentration determination. By using an automated electrophoresis station it was possible to distinguish the ligated amplicon peak from the non-ligated peak (supplementary [figure S1](#pone.0017785.s001){ref-type="supplementary-material"}), allowing for correct quantification of ligated amplicons and volume adjustment and thereby enabling accurate library pooling. With this set-up an even read distribution was obtained ([figure 3](#pone-0017785-g003){ref-type="fig"}). ::: {#pone-0017785-g002 .fig} 10.1371/journal.pone.0017785.g002 Figure 2 ::: {.caption} ###### The workflow. Figures in circles indicate the number of samples that are handled within each step showing the reduction of sample handling complexity. The first step is conducted plate wise, marking 96 samples at a time with a position (specific) tag. In the second step pooling of the 96 samples in each plate is performed and the pooled samples are subjected to spin column PCR cleanup. This step is followed by the third fully automated step of end polishing, phosphorylation and plate (specific) tag ligation. The concentration of the ligation products are measured and the plate pools are equimolarly pooled into one sequencing library for each lane. The immobilization of sequencing libraries onto paramagnetic beads and fill-in reaction is fully automated. The workflow ends with standard emPCR and sequencing. :::  ::: ::: {#pone-0017785-g003 .fig} 10.1371/journal.pone.0017785.g003 Figure 3 ::: {.caption} ###### Distribution of reads across the MIDs. Pie chart showing an even read distribution of the plate specific tag, illustrating successful equimolar pooling of ligated fragments prior to immobilization, fill-in and emPCR. Since the experiment was designed for two lanes, each plate tag was used for tagging two plates and hence, each sector represents 2\*96 = 192 samples. :::  ::: The 52 libraries were sequenced on the 454 GS FLX Titanium Chemistry using three out of four lanes, generating a total of 700,000 reads. As mentioned above, the project was designed to employ 2 lanes (L = 2) but as we managed to access three lanes, the libraries were divided into three emPCR and sequencing lanes instead of two (see [figure 2](#pone-0017785-g002){ref-type="fig"} and supplementary [table S2](#pone.0017785.s003){ref-type="supplementary-material"}). For 285,000 reads, the read-lengths were long enough to identify the plate-tag and both position-tags. The low rate of useable sequence reads could be explained by the use of a strict criterion not allowing sequence errors in the tags and the fact that the target region harbors a high number of homopolymeric sequences (a well known problem with Pyrosequencing chemistry). 10% of these sequence-reads showed different position-tags in the ends, indicating chimeric formation. These chimers are most probably formed from different single-stranded PCR-products (from different samples) that are recombined when amplicons within each plate are pooled and extended to completion during the end-polish reaction. To minimize this effect PCRs with fewer cycles could be performed (reducing the amount of single stranded amplicons) or the end-polish reaction could be skipped (by employing a polymerase that lacks terminal transferase activity in the PCR). However, the fact that the inner position-tags are incorporated in both ends of the sequencing templates makes it possible to detect chimeric sequences from different samples and exclude them from the data analysis. This demonstrates that our straightforward approach to detect chimeric sequences is needed for obtaining correct results during multiplex sequencing and sorting of polymorphic genes. These erroneous constructs would have gone undetected or at least difficult to detect employing previously mentioned tagging methods [@pone.0017785-Galan1]. To further investigate the extent of multi-sample chimeric sequence formations, a library was created in which the plate-tags were incorporated by a second PCR (using the C and D handles, see primer sequences in supplementary [table S1](#pone.0017785.s002){ref-type="supplementary-material"}) after sample pooling (instead of ligation). This means that there were 96 samples in each plate-tagging PCR and thus the extent of chimeric products was expected to increase. After sequencing, chimeric sequences were observed in 97% of the reads where all tags could be identified (data not shown). We therefore strongly recommend that samples should not be pooled before a PCR reaction when sample sorting and individual genotyping is the aim of the study. Allelotyping of pooled samples [@pone.0017785-Pettersson2] is not included in this recommendation since the aim of these studies is to estimate and compare allele frequencies in different cohorts. A ten-fold difference between the most and the least sequenced position-tags was observed (data not shown). However, since each position-tag is incorporated in 52 samples, there should be an even spread of sample quality across the position-tags, hence less difference would be expected. The difference in sequence depth within the position-tags is therefore more likely an effect of primer qualities in combination with the fact that different tag sequences could affect the primer annealing by forming secondary structures. Still, the majority of the PCR products were correctly genotyped. The minimum requirement for correct genotyping was set to 10 reads per allele resulting in 20 reads per individual [@pone.0017785-Harismendy1]. Out of the estimated 3700 PCR-products, 3465 generated more than 20 reads ([figure 4](#pone-0017785-g004){ref-type="fig"}). To investigate how the polymorphisms are distributed across the sequenced exon, all generated consensus sequences were aligned and the fraction of polymorphisms for each position was plotted ([figure 5](#pone-0017785-g005){ref-type="fig"}). This resulted in a number of hot spot positions where the polymorphisms are concentrated. ::: {#pone-0017785-g004 .fig} 10.1371/journal.pone.0017785.g004 Figure 4 ::: {.caption} ###### Distribution of reads across the samples. Histogram showing the distribution of reads per sample for the 3700 successfully amplified samples. The vertical line indicates the breakpoint of 20 reads required for accurate genotyping. :::  ::: ::: {#pone-0017785-g005 .fig} 10.1371/journal.pone.0017785.g005 Figure 5 ::: {.caption} ###### Polymorphisms. Indicating the polymorphic positions across the target exon. All generated consensus sequences have been aligned together and the diversity of each position in the alignment has been calculated. The bars indicate the fraction of the aligned consensus sequences that differs from the most common allele at a certain position. :::  ::: To conclude, we here demonstrate a simple, robust and reliable method for sequencing thousands of samples in parallel in one single sequencing run. The method is robust enough to omit the time and reagent consuming step of equimolar pooling at the individual level by performing equimolar pooling of 96 samples at a time after introduction of the second plate specific tag. The need for making unique primer combinations for each sample is replaced by having the same 96 position specific primer pairs for all PCR plates. By strictly handling the PCR tagging procedure in 96-wells format and by automating the ligation of plate-tags and library preparation, it is possible to increase the number of samples without affecting the sample handling complexity. This higher control over pipetting errors combined with automated library preparation is one of the main strengths of the presented method since this approach does not require less number of unique primers compared to the competing methods. We have shown that it is possible to obtain sufficient sequence depth for 94% of the successfully amplified samples when running at a multiplexing level of 4992 tag combinations on the 454 sequencing system. Furthermore, incorporation of the same position-tags at both ends of the fragments allows detecting chimeric sequences from different samples that is a prerequisite for accurate identification of samples. We believe that the reliability of the method combined with scalability makes it suitable for sequencing targeted enriched DNA or RNA of even greater sample sizes on platforms such as HiSeq 2000 and SOLiD where the number of reads per run is in the magnitude of 2000 times greater than the 454 system. Supporting Information {#s4} ====================== Figure S1 ::: {.caption} ###### **Electropherogram of ligated product.** Indexed library concentration determination. A characteristic library electropherogram illustrating the main ligated product peak that was used for quantification, separated from the non-ligated smaller peak not taken into account when the concentration was determined. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Indexed PCR primers.** The 96 primer pairs used for position specific tagging by PCR. The gene specific regions are situated in the 3′-ends of the primes, the tag sequence in the middle and at the 5′-end universal handles are included. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Concentration of ligated products.** The 52 pooled PCR plates were indexed using 26 MID adaptors. Due to poor performance, related to primer synthesis, some MIDs were excluded hence, the lack of some numbers in the MID column. Three library pools (1, 2 and 3) were made, each sequenced on a separate lane. Concentrations were determined for the large (ligated) peak. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Steven Fain, Dyan Straughan, Alan Wilton, Pranoy Ghosh, Barbara van Asch, Hannes Lohi, Eija Seppala and Ranja Eklund, for help with sample preparation. We would also like to thank Patrik Ståhl for valuable discussions and Henrik Stranneheim for help with automation of washing procedures. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by a grant from The Swedish Board of Agriculture (Jordbruksverket). Peter Savolainen is a Royal Swedish Academy of Sciences Research Fellow supported by a grant from the Knut and Alice Wallenberg Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MN PS AA. Performed the experiments: MN SL. Analyzed the data: MN SL PS AA. Contributed reagents/materials/analysis tools: PS AA. Wrote the paper: MN SL PS AA. Designed the script for automated library prep: SL. Designed the scripts for data analysis: MN.

PubMed Central

2024-06-05T04:04:19.253337

2011-3-9

PMC3052375

Introduction {#s1} ============ The role that norms play as regulator mechanisms of certain aspects of social, economic and organizational behaviours has been thoroughly studied in the social sciences [@pone.0017661-Coleman1], [@pone.0017661-Kandori1]. Once a norm has been established, it acts as a self-reinforcement mechanism of behaviour. However, the emergence, diffusion and collapse of social norms are, in general, exempt from explicit mechanisms of control. There are different kinds of norms depending on the type of social interaction. Concretely, in the economics field, an important research effort is focused on understanding the emergence of norms that determine the property distribution in a community. Thus, in contrast to the equity norm that leads to "distributive justice" and fair division of goods in some communities, there is also evidence of systematic persistence of discriminatory norms that allocate different shares of a resource according to some individual characteristic or group membership. Evolutionary game theory is a powerful framework to analyse this type of norms formally. In particular, the Nash bargaining game [@pone.0017661-Nash1] is often used as a simple archetypical model of economic interaction and good distribution. Succinctly, the two-player Nash bargaining game consists of two players that have to divide a sum of money among them. The payoff for each player is the amount of money they asked for, but if the sum of the demands exceeds the total, they both obtain nothing. If the game is played repeatedly among an infinite population of players that are randomly paired up and change their strategy according to the replicator dynamics then, given a particular initial condition, it is possible to compute the distribution of strategies in the population over time [@pone.0017661-Skyrms1]. Notwithstanding, the influence of such assumptions has proved very relevant for the results of evolutionary game models [@pone.0017661-Roca1]. This is particularly relevant given that such assumptions are not always easy to justify. In 2001, Axtell, Epstein and Young [@pone.0017661-Axtell1] proposed an agent-based model (henceforth AEY\'s model) to understand the transient and the asymptotic dynamics of the Nash demand game in a finite population. They simplified the analysis considering just three possible demands: low (L), medium (M) and high (H). They proved that different self-reinforcing norms can emerge spontaneously. These emergent norms may be completely different from one another even though all the agents of the population have exactly the same behavioural rule. Which particular norm appears first depends on initial conditions and on purely accidental events, such as the specific pair of agents that happened to be (randomly) paired at a certain time. To obtain these conclusions Axtell et al. conduct their analysis in two parts. Initially they study the dynamics of a population of indistinguishable agents with the capacity to store in their memories the strategies played by their opponents in the last encounters. Each agent uses this information to form an expectation about her opponent\'s strategy, assuming that the probability of the next demand equals the relative frequency of the remembered experiences in the last encounters. Given that belief, each agent responds with a "noisy best reply", i.e. a best reply with a small probability of selecting a random demand. One of the norms that can emerge in this setting is the so-called "equity norm", i.e. a self-fulfilling situation where everyone expects the others to demand M and, as a consequence, everyone demands M; this behaviour, in turn, confirms the expectations that everyone already has, thus closing a self-consistent loop. Axtell et al. point out that the "equity norm" is the unique stochastically stable state of the game (see Young [@pone.0017661-Young1], [@pone.0017661-Young2] for a comprehensive analysis of the required conditions to obtain this conclusion), but they also find other persistent stable fractious states, in which players play repeatedly L or H, but never M. More interestingly, in the second part of their analysis, Axtell et al. endow each individual agent with one of two possible tags (which can be recognised by all, but has no initial meaning) and with the ability to remember both the past behaviour of her opponents and their tag. In this second setting, they find that a new stable state can endogenously emerge, in addition to the ones previously observed. In this new state, agents behave differently within and outside their own tag group, so the state was naturally labelled "segregation". The implications of Axtell et al.\'s finding are astonishing: a discriminatory norm in which property is unequally distributed based on observable characteristics that are initially meaningless, may not only emerge but even perpetuate for long, as a consequence of the self-reinforcing nature of the dynamics. These results are very suggestive from a social point of view when we associate the concept of tag in the model with some social or cultural trait such as race, gender or age, which may condition people\'s behaviour in human societies. Using the model as reference, the emergence of a rich variety of collective outcomes can be explained. An example would be the situation where a divided underclass is oppressed by a unified elite: this would correspond in the model to a state where the elite group systematically plays H against the oppressed group (who responds optimally playing L) and plays equitable (M) among themselves, while the discriminated group is stuck in a fractious state. The replicator dynamics embeds two important assumptions: infinite populations (which is the hypothesis relaxed by Axtell et al. [@pone.0017661-Axtell1]) and random pairings. The assumption that pairings are random can be understood as an abstraction of persistent bargaining interaction with strangers. However, in some contexts this may be unrealistic; agents may interact only with just a small number of other agents with which they are in direct contact [@pone.0017661-Alexander1], [@pone.0017661-Charness1]. In those cases the global interaction assumption can be removed and we can analyze the effect of a given social or spatial structure. Introducing structure in the population implies that the probability of interaction between two agents depends on the specific pair of agents. The structure of the population can be usefully represented by means of a graph or network that describes the interaction connectivity. Ohtsuki et al. [@pone.0017661-Ohtsuki1], [@pone.0017661-Ohtsuki2] argue that in a general case, the structure should be described by two graphs, one representing the interaction of the game played and a second one representing the interaction of the adoption or learning mechanism. Usually both graphs are considered the same. The effect of many different types of graphs in games has been investigated, examples of which include the analysis of iterated 2×2 games such as the Prisoner\'s Dilemma on regular lattices [@pone.0017661-Fort1]--[@pone.0017661-Perc1], Erdos-Renyi [@pone.0017661-Perc1]--[@pone.0017661-Poncela1], small-world [@pone.0017661-Perc1], [@pone.0017661-Fu1], scale-free [@pone.0017661-GmezGardees1], [@pone.0017661-Poncela1], [@pone.0017661-Perc2]--[@pone.0017661-Poncela2] or real networks [@pone.0017661-Lozano1], [@pone.0017661-Lozano2], the analysis of the snowdrift game on lattices [@pone.0017661-Hauert1], [@pone.0017661-SysiAho1], small-world [@pone.0017661-Tomassini1], [@pone.0017661-HanXin1] or scale-free [@pone.0017661-Perc2]--[@pone.0017661-Santos2] networks, and n-person games such as public good games on lattices [@pone.0017661-Helbing1]--[@pone.0017661-Szolnoki2] or the minority game on small world networks [@pone.0017661-Chen1], [@pone.0017661-Kirley1] (some reviews can be found at [@pone.0017661-Roca1], [@pone.0017661-Szabo1], [@pone.0017661-Perc3]). In this article we have extended the analysis of norm diffusion in a population considering AEY\'s model as a framework. We have studied the influence of the topology on the results of the game. To this aim, we have considered the spatial dimension of the game by introducing a regular spatial structure. We have also modified the original model by adding a new behavioural rule that requires less cognitive abilities than those required in the original paper. When agents use this behavioural rule, the segregation norm emerges more frequently, and a richer space of solutions is observed. This work is organized as follows: first, we briefly explain the extensions and modifications that we have performed on AEY\'s original model. Next, we describe the results that we have obtained when agents are randomly assigned a tag. At the end of this section we discuss some cases where several persistent regimes can simultaneously emerge, and their relation with some mesoscopic topological properties. We then finish with the conclusions of this work. Methods {#s2} ======= Agent-based Model of Bargaining in a Regular Lattice {#s2a} ---------------------------------------------------- In this section we describe an agent-based model of bargaining in regular lattices based on the original tag model proposed by Axtell et al. [@pone.0017661-Axtell1]. Our model introduces a spatial restriction in the structure of interactions: agents are embedded on a regular lattice and they can only bargain with their spatial neighbours. In each time period of the model, each agent selects one of her neighbours at random to play the Nash demand game. When playing the game, each agent considers three possible demands of a pie (which is a metaphor of something that is going to be shared between two persons), i.e. *low (L)* or 30%, *medium (M)* or 50% and *high (H)* or 70%. The agents get the chosen demand if the sum of their demands does not exceed 100 percent of the pie; otherwise they both get nothing (see the payoff matrix in the [Table 1](#pone-0017661-t001){ref-type="table"}). The Nash demand game represented in [Table 1](#pone-0017661-t001){ref-type="table"} has exactly three pure-strategy Nash equilibria, corresponding to the pairs (L, H), (M, M), and (H, L). These are called the equilibria of the one-shot bargaining game. ::: {#pone-0017661-t001 .table-wrap} 10.1371/journal.pone.0017661.t001 Table 1 ::: {.caption} ###### Payoff matrix of the Nash demand game. ::: {#pone-0017661-t001-1} H M L --- ------------- ------------- ------------- H (0,0) (0,0) **(70,30)** M (0,0) **(50,50)** (50,30) L **(30,70)** (30,50) (30,30) ::: As in Axtell et al. [@pone.0017661-Axtell1], the population of agents is divided into two groups of equal size whose members share a recognizable characteristic which has no a priori social meaning, i.e. a tag. These tags are used by the agents to select their demand in the one-shot game. To be precise, the agents\' decision rule, which is identical for all individuals, is based on the agents\' capacity to remember their opponent\'s demand in the *m* most recent interactions with opponents with the same tag (i.e. intratype interaction) and the *m* most recent interactions with opponents with the other tag (i.e. intertype interaction). These experiences constitute the agent\'s intratype and intertype memories. In the AEY model, an individual chooses the best reply that maximizes the expected demand considering their past experiences with similar opponents, i.e. those with the same tag. In contrast, in our model we consider a simpler decision rule, henceforth the mode rule, which dictates that individuals choose the best reply against the most frequent demand with similar opponents (ties are resolved randomly without any bias). The mode rule is cognitively less demanding than AEY\'s and, naturally, it induces different results than those obtained with the original rule [@pone.0017661-Poza1]. The stochastic version of the game considers that agents may make mistakes in their decisions (or simply experiment from time to time). Hence, with probability (1−*ε*) an individual chooses the best reply and with probability *ε* she chooses one of the three possible demands at random (*low*, *medium* or *high* with the same probability). The model has been implemented in Netlogo (<http://ccl.northwestern.edu/netlogo/>) and can be downloaded at this url: <http://ingor.ubu.es/models/aeygrid>). We can summarize the model as follows: there is a population of *N* agents randomly distributed in a regular 2-dimensional toroidal lattice of *LxL = N* cells, each one inhabited by one and only one of the *N* agents. The population is divided exactly into two groups whose members have a distinctive tag. The number of agents is chosen satisfying simultaneously two conditions: (1) it is even, so the population can be divided exactly into two groups, (2) and its square root is an integer, so the regular lattice is square too. Each agent is endowed with two memories of length *m* to keep the demands of the two classes of tags. Memories are initialized at random. In each time period *t*, each agent randomly selects one of her 8 neighbours (radius-1 Moore neighbourhood) to play the game. The agent observes her opponent\'s tag and decides the best reply against the most frequent demand in her corresponding memory (i.e. intratype or intertype). However, with a small probability *ε* an agent decides randomly between the three possible demands. Afterwards both agents update their memories. [Figure 1](#pone-0017661-g001){ref-type="fig"} shows the different interaction networks of a particular spatial distribution of tagged agents in a 4×4 lattice. ::: {#pone-0017661-g001 .fig} 10.1371/journal.pone.0017661.g001 Figure 1 ::: {.caption} ###### Example of a spatial distribution and its corresponding interaction networks. The spatial distribution, depicted in the centre, consists of 16 agents randomly distributed in a regular 2-dimensional toroidal lattice of 4×4 cells. The corresponding set of interaction networks are shown in the corners: (i) the complete interaction network (upper left corner), (ii) the intertype interaction network (upper right corner), (iii) the "black tag" intratype interaction network (bottom left corner), and (iv) the "white tag" intratype interaction network (bottom right corner). :::  ::: Results {#s3} ======= Understanding the dynamics of the model {#s3a} --------------------------------------- Before our computational exploration of the model, we have conducted a brief analysis within the framework of Markov Chains [@pone.0017661-Izquierdo1] to gain some insights about the expected dynamics and behaviour of the model. Fortunately, some aspects of this formal analysis have already been carried out for the AEY model by Axtell et al [@pone.0017661-Axtell1], and for the evolutionary model of bargaining by Young [@pone.0017661-Young1], [@pone.0017661-Young2]. In terms of markovian properties, our model shares many characteristics with these models. To represent the model as a time-homogeneous Markov Chain (THMC), we define the state of the system in a time period t as a N-dimensional vector X~t~ = {X^1^ ~t~, X^2^ ~t~,..., X^N^ ~t~} of 2m-tuples X^i^ ~t~, each one corresponding to agent i\'s memory of both intratype and intertype encounters (it is not necessary to use all m values of the agents\' memory to represent the state space, since knowing only the memory length and two of the frequencies of each possible demand {L,M,H} is enough). Note that the spatial distribution of tags conditions the chances of intratype/intertype encounters in each period, but the possible changes that may occur in each interaction are only dependent on the particular form of the two m-tuple memories that are involved in the interaction. The characteristics of the system dynamics are strongly determined by the presence or absence of errors (mutations in evolutionary terminology) in agents\' decisions. In the absence of decision errors, i.e. the *unperturbed model*, the system has absorbing states in which sooner or later it will be trapped (if we run the model for long enough). These absorbing states are directly related with the three pure-strategy Nash equilibria of the Nash demand game, giving rise to the *equity norm* (EQ) and the *inequity norm* (IQ). The former happens when everyone in the population expects the others will demand M, and consequently everyone demands M, so the system ends reaching an absorbing state for both intratype and intertype bargaining processes, which is equitable because all agents get equal payoffs, and is also efficient (in Pareto sense) because no agent can be made better off without making another agent worse off. Apart from this, in the AEY model without spatial restrictions [@pone.0017661-Axtell1], there are also IQ absorbing states for the intertype bargaining game. An IQ equilibrium corresponds to a state in which tagged agents coordinate in one of the two asymmetric pure-strategy Nash equilibria. Whenever agents of one tag expect the others will demand L and hence they will demand H, and simultaneously the others will expect and demand the complementary decisions, the system reaches an absorbing state, which in this case is efficient but not equitable in the proportions obtained by each agent. Interestingly, additional absorbing states show up as a consequence of the imposed spatial structure. For example, a spatial distribution of 4×4 tagged agents like the one depicted in [Figure 2](#pone-0017661-g002){ref-type="fig"} allows an IQ absorbing state in both intratype bargaining games, i.e black-black and white-white. ::: {#pone-0017661-g002 .fig} 10.1371/journal.pone.0017661.g002 Figure 2 ::: {.caption} ###### Toroidal grid of 4×4 cells with 8 white and 8 black agents distributed in the way shown. In this particular spatial distribution the system may reach the IQ state in both intratype games whenever similar tagged agents placed in the same column demand exactly the complementary quantity (L or H) of their neighbours of columns just next to them. This result is true if the intratype network is bipartite, i.e. there are no odd-length cycles. :::  ::: When randomness is introduced in agents\' decisions (motivated by the possibility of mistakes or by a simple desire for exploration), the system becomes ergodic. In this case, there is a unique limiting distribution over the state space which determines the probability of finding the system in each of its states in the long run (e.g. lim~t→∞~P(X~t~ = i)). Such probabilities are strictly positive and independent of the initial conditions. This limiting distribution can be estimated sampling just one simulation run for a sufficiently long time, by computing the fraction of the time that the system spends in each state, i.e. the occupancy distribution [@pone.0017661-Izquierdo1]. In contrast to what one may expect, when the tagged model for a finite population and global interaction is asymptotically analysed, this limiting distribution concentrates only on one of the two absorbing states of the unperturbed model, the EQ. The formal demonstration of this relies on the concept of stochastic stability [@pone.0017661-Young2]. When some small noise exists, the EQ state is stochastically stable while IQs are not. This implies that, in the long run and for sufficiently unlikely perturbations, the system tends to spend most of the time at the EQ state. Nevertheless, Axtell et al. [@pone.0017661-Axtell1] make an interesting contribution turning the attention from the asymptotic to the transient dynamics, and showing that there are other relevant states in which the system spends a considerable fraction of the time, henceforth *persistent regimes*. In the transient evolution of the global interaction model, sometimes the system is temporarily trapped in a particular regime, called *fractious regime* (FR), in which agents alternate their demands between H and L, making the emergence of the equity norm very difficult (We keep the word fractious for consistency with the original AEY\'s model; but it may be worth noting that other names, such as "fluctuating agents" [@pone.0017661-GmezGardees1], [@pone.0017661-Flora1], have been used in the literature for essentially the same concept, i.e. agents that intermittently change their strategy). Moreover, they show that the transition time between this fractious regime FR to the stochastically stable state EQ can be enormously long and this time grows exponentially with the number of agents and their memory length -i.e in their terminology: ergodicity is broken. Formally, the system is completely characterised by the vector X~t~, which can be graphically represented using a 2-simplex of the agents\' states (see [Figure 3](#pone-0017661-g003){ref-type="fig"}). Each of the two agent i\'s memories keeps track of the demands made by her opponents in the m most recent intratype (or intertype) encounters, and can be represented by a vector of the relative frequencies of these demands Xi = {n~L~/m, n~M~/m, n~H~/m}, where n~L~ denotes the number of times that agent i\'s opponent demanded L in the m most recent intratype (or intertype) interactions. This vector corresponds in the simplex with the point . Since the memory of an agent is made by two partitions, corresponding to the past demands of the two classes of opponents, we can use two separated simplexes to represent each one. ::: {#pone-0017661-g003 .fig} 10.1371/journal.pone.0017661.g003 Figure 3 ::: {.caption} ###### 2-simplex representation of the state space used in both intratype and intertype bargaining games. The shaded regions correspond to the state subspaces in which an agent always decides one of the three possible demands {L,M,H}. For example, the light grey area at the bottom-left of the triangle represents a set of states in which the majority of the items in the memory are L, and therfore the agent will demand H. The opposite happens in the dark grey area at the top. Finally the M demand dominates the agents\' memories in the white area, so the response of the agent in that area is also M. Note that with the mode-decision rule, the centre of the triangle, which is equidistant from the three vertices, corresponds to the indifferent state in which any of the three demands is equally possible. :::  ::: Most Frequent Persistent Regimes {#s3b} -------------------------------- If, following the approach in Axtell et al [@pone.0017661-Axtell1], we focus our analysis on the transient dynamics of the spatial model, the complexity of the system makes us to resort to computer simulation as methodology. We have designed a set of experiments to discover and understand the persistent regimes that emerge in the model. The parameterization of all cases mentioned in this paper corresponds to a model of N = 100 agents randomly distributed in a regular lattice of 2-dimensional grid of 10×10 cells, each one keeping one of the N agents. Each agent is endowed with two memories of length 10 for intratype and intertype bargaining games, initialized at random. In each time period t, each agent selects one of her 8 neighbours (Moore neighbourhood) at random and decides the best reply against the most frequent demand in her memory for the type of opponent. However, with a small probability ε = 0.01 an agent decides randomly between the three possible demands {L,M,H}. Note that each time period consists of N matches, and consequently it is probable that an agent bargains more than once in each time period. We have sampled 10.000 simulation runs during T~f~ = 30.000 time periods. The system state at the end of the simulation time can be summarized as a 3-tuple of the regimes reached by the intertype and the two intratype bargaining games {Intra-white~regime~, Intra-black~regime~, Inter~regime~}. Taking into account the characterization of the types of stable and persistent regimes described in the previous section, we may expect that if we let the system run for long enough, it will reach one of the 3^3^ possible combinations, i.e. {EQ,EQ,EQ}, {EQ,EQ,FR}, .... We define a set of simple conditions, henceforth *C1 stop conditions*, for reaching each of the persistent regimes according to their nature: the EQ state is considered reached whenever all agents in the corresponding bargaining process have at least *(1−ε)*×*m* instances of M in their memories (note that the memory vector has a finite number of instances, so we approximate (1−ε)×m to the lower integer and ε×m to the higher integer), the IQ and the FR regime are considered reached whenever all agents have at most *ε*×*m* instances of M and, moreover, in the case of the IQ state a group of agents have *(1−ε)*×*m* instances of L and the rest have *(1−ε)*×*m* of H, and in the case of FR all agents have a combination of *(1−ε)*×*m* instances of both L and H. In short, a simulation run stops when either it satisfies one of the C1 stop conditions or it reaches the final time period T~f~. [Figure 4](#pone-0017661-g004){ref-type="fig"} plots the frequency distribution of the stop conditions reached by all simulations we run. As one may expect, the system reaches one of the persistent regimes previously defined in the majority of the cases (80.98% of the runs). A relevant result is that even when the assumption of regular spatial structure of interaction, all the persistent regimes obtained in the global interaction case are also reached. [Figure 5](#pone-0017661-g005){ref-type="fig"} illustrates graphically the most frequent states and regimes through a set of simplexes of some representative runs. The regime is characterized by the corresponding pair of simplexes of both intertype and intratype bargaining. Some of these states can be interpreted from a social perspective as a divided underclass oppressed by a unified elite, as class distinctions, discriminatory regimes, etc. (see [@pone.0017661-Axtell1] for a deeper insight on some interpretations). ::: {#pone-0017661-g004 .fig} 10.1371/journal.pone.0017661.g004 Figure 4 ::: {.caption} ###### Frequency distribution of the stop conditions reached by 10.000 simulation runs. C1 represents the relative frequency of runs that reached one of the C1 stop conditions defined in section 'Most Frequent Persistent Regimes'. C2 represents the frequency of runs that reached the C2 stop conditions defined in section 'Isolated Bargaining Clusters'. This stop criterion extends the C1 conditions to disconnected interaction components that can randomly appear in the spatial distribution of agents on the grid. Finally C3 gathers the rest of the runs, which are analysed in section 'Other Persistent Regimes'. :::  ::: ::: {#pone-0017661-g005 .fig} 10.1371/journal.pone.0017661.g005 Figure 5 ::: {.caption} ###### Most frequent persistent regimes of the transient dynamics for the intertype and intratype bargaining games. The bargaining between different groups (intertype) can reach the EQ state (top-left simplex), the IQ state (middle-left simplex) or the FR regime (bottom-left simplex). In the bargaining within groups (intratype) we have shown the combination of {EQ,EQ} when both groups coordinate in the EQ state (top-right simplex), {EQ,FR} when one group is in the EQ state but the other is in the FR regime (middle-right simplex), and when both groups stay in the FR regime (bottom-right simplex). :::  ::: Isolated Bargaining Clusters {#s3c} ---------------------------- Although the analysis of the simulation results described in the previous section explains more than 80 percent, it still leaves out a significant set of them. A preliminary visual exploration of some anomalous cases gives us a possible answer: the presence of disconnected groups of agents which play the bargaining game isolated from other groups. In the initialization of the model, agents are randomly distributed and consequently most of spatial distribution samples have agents of both tags dispersed in the lattice, but close enough to make the dynamics interdependent. However, sometimes this randomness produces the formation of two or more isolated groups, i.e. groups of agents who decide their (intertype or intratype) demands without any direct or indirect influence from the agents that belong to other groups. This possibility had not been considered when we defined the C1 stop conditions, so when this event happens the simulation may reach the final time period if groups evolve to different regimes. It is important to notice that the intratype and intertype interaction networks are formed in the random initialization process and are fixed until a stop criterion is reached. Other relevant research in coevolving games does not assume fixed interaction networks but instead the structure dynamically emerges as a consequence of the game. Some of these coevolutionary rules have been used to model mechanisms of learning [@pone.0017661-Skyrms2], conditional dissociation [@pone.0017661-Izquierdo2], unilateral and mutual choice in group dynamics [@pone.0017661-Yamashita1], [@pone.0017661-Yamashita2], reputation-based partner choice [@pone.0017661-Fu2] or the formation and deletion of strategy-independent links [@pone.0017661-Szolnoki3]--[@pone.0017661-Wu1]. We illustrate these cases with one of the runs that exhibits this type of spatial distribution (see [Figure 6](#pone-0017661-g006){ref-type="fig"}). In particular, the example run has two disconnected white-tagged groups that reach different final regimes. In order to discriminate this sort of cases we define the *C2 stop conditions* which are exactly the same conditions as C1 but applied at the level of disconnected groups -or components, in the terminology of network theory that we will use in the next section-, instead of at the level of the whole population, as we do to define C1 conditions. ::: {#pone-0017661-g006 .fig} 10.1371/journal.pone.0017661.g006 Figure 6 ::: {.caption} ###### An example case that shows two disconnected groups in the white tag intratype interaction network. Upper figure: spatial distribution of agents with two disconnected groups within white-tagged agents. Lower figure: the corresponding white tag intratype interaction network in which the partition of the network is easily seen (the 51--53 couple vs the rest of white agents). Although it has not been mentioned, note that it is not difficult to identify two other disconnected groups within black agents in this example (the 60--89 couple vs the rest of the black agents). :::  ::: Introducing the C2 stop conditions, the number of runs that end in some of the persistent regimes described so far increases until it reaches more than 89 percent of them (see [Figure 4](#pone-0017661-g004){ref-type="fig"}). The remaining set of runs, which end without reaching any of the expected regimes, i.e. C3 stop conditions, are analysed in detail in the next section. Introducing the C2 stop conditions, the number of runs that end in some of the persistent regimes described so far increases until it reaches more than 89 percent of them (see [Figure 4](#pone-0017661-g004){ref-type="fig"}). The remaining set of runs, which end without reaching any of the expected regimes, i.e. C3 stop conditions, are analysed in detail in the next section. Other Persistent Regimes {#s3d} ------------------------ The results above do not capture all persistent regimes in the game. As evidence of intensive simulation in the spatial game with random configurations, we find that there are still situations that need a much longer time to stop with one of the two criteria (C1 and C2). This fact could suggest the appearance of some other basins of attraction beyond the original AEY regimes that we have found in section 'Most Frequent Persistent Regimes' and the mentioned combinations of isolated states of section 'Isolated Bargaining Clusters'. This implies that there are additional situations where the transient dynamics of the system differs from the long-run behaviour of the system. A visual inspection of the tag spatial distribution of these cases puts forward some effects of the topology of interaction that could explain additional regimes. This happens when there are connected clusters of agents with the same tag who play different types of intratype coordination in each of the clusters. The key difference with the cases analysed in the previous section is that such clusters are indeed connected. It seems clear that the structure of interaction has an influence on the game dynamics. We can consider the structure of intratype interaction as an undirected network where each player represents a node and there is a link between two nodes if both players can play the intratype game (i.e. they are spatial neighbours and they have the same tag). Our hypothesis is that the behaviour of the system is affected by the topological properties in the mesoscale, between the individual and the whole population, of this underlying interaction network. One of the most relevant mesoscopic characteristics in a network is the property of community structure. Informally, a community in a network consists of a subset of nodes that are relatively densely connected to each other but sparsely connected to other dense groups [@pone.0017661-Porter1]. This type of local structure can be easily identified in a variety of social contexts: families, friendship circles, virtual groups in the Internet, neighbourhoods, etc. In fact, there is a very rich and growing literature of networks that present community structure, going from the networks of committee and subcommittee assignments in the United States House of Representatives [@pone.0017661-Porter2], scientific collaboration networks [@pone.0017661-Girvan1], to networks of e-mail interactions between university employees [@pone.0017661-Guimer1] or the collaboration network of jazz musicians [@pone.0017661-Gleiser1]. We presume that in connected networks that present strong community structure, different communities can reach different persistent regimes, and the spread of one of the regimes to the whole connected group can be obstructed if the inter-community connectivity is low. We illustrate the intuition of this phenomenon in the following idealized case. In [Figure 7](#pone-0017661-g007){ref-type="fig"} we represent a certain configuration of tags and the underlying intratype interaction network of white-tagged players. Intuitively there are two communities in the network (depending on the algorithm used to identify communities, there may be other partitions in communities different to the presented in the example). If we play this game repeatedly a frequent result is showed in [Figure 8](#pone-0017661-g008){ref-type="fig"}. Each community reaches a different regime, stays trapped in it for a long time, and the diffusion of a general homogeneous behaviour in the game is hindered. ::: {#pone-0017661-g007 .fig} 10.1371/journal.pone.0017661.g007 Figure 7 ::: {.caption} ###### Idealized case of white-tagged players and the underlying intratype interaction network. We have analysed the effects that appear in the stylized configuration showed on the left of the figure. On the right, we represent the underlying interaction structure for the intratype game of white-tagged players. :::  ::: ::: {#pone-0017661-g008 .fig} 10.1371/journal.pone.0017661.g008 Figure 8 ::: {.caption} ###### Partition in communities of the idealized case and final regimes. On the right, we can observe two different persistent regimes in the intratype game of white-tagged players depicted in [Figure 7](#pone-0017661-g007){ref-type="fig"}. The different regimes correspond exactly to the different communities showed on the left. :::  ::: If we want to extend these results to more general conditions in the lattice game, we need to specify exactly how to define the concept of community beyond the intuitive and vague idea of some nodes very connected among them and sparsely linked with other communities. As a matter of fact, the problem of detecting communities is very challenging for two reasons: first, the number of possible partitions is huge for non-trivial networks, and second, but no less important, the concept (and hence the preferred definition of communities) may be domain-specific, depending on the field of application. Given this, it is not surprising that nowadays there is a wide plethora of methods based on different techniques and ideas to define and to identify communities in networks (see some recent reviews in [@pone.0017661-Porter1], [@pone.0017661-Danon1]--[@pone.0017661-Lancichinetti1]). In our analysis of the spread or lock-in of the persistent regimes, the idealized case gives us a hint to select the identifying community algorithm. We see that the edges that separate communities act as bottlenecks that enable or put obstacles to the flow of strategies. Based on this idea, Girvan and Newman [@pone.0017661-Girvan1] defined the concept of betweenness of an edge generalizing the concept of betweenness of a node by Freeman [@pone.0017661-Freeman1]. The betweenness of an edge is calculated as the number of geodesic (i.e. shortest) paths between node pairs that run through it, normalized dividing by the number of pairs of nodes. The betweenness of an edge gives us an idea of the importance of the link to stop the flow of information in the network. The algorithm of Girvan and Newman requires calculating the betweenness of all edges in the network and removing the one with the highest betweenness, repeating the whole process until no edges remain (in case of tie, one can be randomly removed, or all can be simultaneously removed). The logic of the algorithm is based on the idea that the edges connecting communities will have comparatively high betweenness and hence, by removing them iteratively, we will separate the different components of the network that reveal the hidden community structure of the graph. The result of this algorithm is a dendrogram where horizontal cross-sections represent different possible community divisions, depending on the desired number of communities. Since the method does not provide the appropriate number of communities to split the network, the same authors [@pone.0017661-Newman2] proposed to evaluate the divisions using the concept of modularity as the fitness function. The modularity of a partition is an index that aims to quantify how good a partition is. Partitions with high values of modularity are those in which there are dense internal connections between the nodes within clusters but only sparse connections between different clusters. Modularity compares the number of links inside a community with the expected number of links that one would find in the community if the network were randomly generated keeping the degree of every node (i.e. the number of links), but linking them randomly. Following Newman [@pone.0017661-Newman3], the modularity Q of an unweighted and undirected network partitioned into communities can be computed as:where *e~ii~* denotes the fraction of all edges that have both ends in community *i*, and *b~i~* is the fraction of edges that have one or two ends in community *i*. Given that this algorithm to partition the network formalizes the idea of information flow, we hypothesize that some additional persistent regimes can appear when each community adopts a coordinated regime except for potentially some border agents with other communities that can present a fluctuating behaviour depending on the community with which they play, and hence act as bottlenecks for the diffusion of norms between communities. In order to check our hypothesis we have analysed the instances where simulations have not reached any of the persistent regimes considered in the previous sections: the simulations that stopped because they reached the final time period T~f~, i.e. stop condition C3. In each one of these cases we have recorded the final state of each player in the intratype game according to the definitions of section 'Isolated Bargaining Clusters'. If the agent did not reach any of the predefined states based on her memory, we classify her as "regime not established". In those games, we have also exported the topology of each component of the underlying intratype interaction network and applied the Girvan-Newman algorithm maximizing the modularity to identify the different communities. We can compare the partition given by the algorithm with the final behavioural state of the players in the game. If the mesoscopic topology conditions the spread and diffusion of strategies in the lattice, the state of the players should be homogeneous in each community except for potentially some nodes that are at the border of the community. We define a node as border in a community if she has a link to another player that belongs to a different community. When two connected communities stabilize in a different regime, the agents that are at the border should present a flipping strategy, as a consequence of their exposure to different regimes. In fact, given the construction of the Girvan-Newman algorithm, the interaction of an agent that is at the border of a community with the neighbour community is done by means of links of high betweenness. In general, the frequency of interaction of those agents with players in the neighbour community is going to be lower than with players in their own community. In terms of diffusion of regimes this fact is crucial, since in order to change their strategy they would need to play very often with players from the other communities, which is against the chances imposed by the topology. Agents at the border act as buffers and stabilizers of the diffusion of regimes. In the 1007 components from simulations that finished with the C3 criterion, we have computed the number of nodes that have a homogeneous strategy with the community where they belong, and the number of border players that have a different behaviour. The number of nodes in this category accounts for 91.7%. We represent in [Figure 9](#pone-0017661-g009){ref-type="fig"} the percentage of nodes explained in the final state of the 1007 components. Our community analysis of the 1007 networks has identified 6,366 communities. We have also computed the number of players within a community proposed by the algorithm that have a discordant behaviour according to our hypothesis. Results are presented in [Figure 10](#pone-0017661-g010){ref-type="fig"}. As we can see more than 60% of the communities have exactly the expected behaviour. ::: {#pone-0017661-g009 .fig} 10.1371/journal.pone.0017661.g009 Figure 9 ::: {.caption} ###### Number of components analysed that have a given percentage of explained nodes (i.e. nodes that have a homogeneous strategy with their communities, or border nodes with a different strategy to that in their community). :::  ::: ::: {#pone-0017661-g010 .fig} 10.1371/journal.pone.0017661.g010 Figure 10 ::: {.caption} ###### Number of communities identified by the Girvan-Newman algorithm which have none or more nodes that are not homogeneous neither border. :::  ::: The 8% of nodes that present a strategy discordant with the expected behaviour can be explained by different reasons. First of all, it is important to notice that the model is stochastic, and hence some randomness is going to be present. This randomness may introduce important inertia in the analysis of the state of the players. Another factor could be that we are stopping the simulations after 30 000 ticks, which for some complex topologies may be insufficient to converge to a complete persistent regime. Apart from that, we should keep in mind that the topologies analysed are obtained from complete random initialization of agents in the lattice which may produce strange topologies. The partition in communities of such topologies can be different depending on the algorithm used. The Girvan-Newman algorithm is very appealing to explain diffusion processes because it is based on a centrality measure but other algorithms are better at maximizing the modularity [@pone.0017661-Danon1]. It is possible, therefore, that other algorithms give us other partitions that improve the explanation based on the Girvan-Newman algorithm. In any case, this study has shown the significant effect of the mesoscopic interaction structure in the spatial diffusion of strategies of the game in the lattice. Discussion {#s4} ========== In this work we have addressed the effect of a regular spatial structure on the Nash bargaining game in a finite population of tagged agents. We have showed that all the transient regimes proposed by Axtell et al [@pone.0017661-Axtell1] can also be present in the lattice game. More interestingly, depending on the particular tag distribution of agents through the grid we have found some topological properties that explain the diffusion of the agent\'s strategies in the lattice. We have showed that isolated clusters of intra or inter type of interaction can reach different persistent regimes. Moreover, we have proved the influence of the topology in understanding new stable regimes different from those found by Axtell et al [@pone.0017661-Axtell1]. To explain their appearance and persistence, we have based our analysis on the mesoscopic properties of the interaction structure, concretely in the community structure of the network of interaction. Using the Girvan-Newman algorithm based on the edge betweenness and the concept of modularity to identify communities, we can understand the behaviour of many of the nodes of the simulation that do not reach any of the previous described regimes. Although the results can be dependent on the rules of the game, they may explain the emergence of different norms of economic interaction and resource allocation among different spatial groups, not only if the groups are isolated and do not communicate among them, but also if the interaction among groups has community structure. The findings of these mesoscopic effects in a property distribution game strongly corroborate the relevance of the arguments previously exposed by authors like Lozano et al. [@pone.0017661-Lozano1], [@pone.0017661-Lozano2] in the evolutionary Prisoner\'s Dilemma, Roca et al. [@pone.0017661-Roca2] and Tomassini and Pestelacci [@pone.0017661-Tomassini2] in cooperation dilemmas, and similar phenomena also described by Castelló et al. [@pone.0017661-Castell1], [@pone.0017661-Toivonen1] in the context of dynamical models of competing options. Although the game played on the spatial substrate is different from the games explored by these authors, the mechanism that prevents the homogenization of a general strategy in the population is very similar. They use the idea of topological traps [@pone.0017661-Roca2], [@pone.0017661-Castell1], [@pone.0017661-Toivonen1] (i.e. links between nodes of different degrees in regions with few or no redundant paths) to explain why "homogeneous strategy waves" do not propagate over the network uniformly. Since we are partitioning the network using a methodology based on the concept of betweenness, we are indirectly detecting the notion of topological traps in the borders between communities, as our results show. We would like to thank Luis R. Izquierdo, Segismundo Izquierdo, Cesáreo Hernández, Javier Pajares and two anonymous reviewers for some advice and comments on this manuscript. We would also like to thank the participants in the SiCoSSys Project workshop held in Valladolid in December 2010. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work stems from the participation of the authors in research projects funded by the Spanish Ministry of Science and Innovation, references TIN2008-06464-C03-02 and CSD2010-00034 (CONSOLIDER-INGENIO 2010), and by the Junta de Castilla y León, references VA006A009, BU034A08 and GREX251-2009. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: DJP JIS JMG ALP. Performed the experiments: DJP JIS JMG ALP. Analyzed the data: DJP JIS JMG ALP. Contributed reagents/materials/analysis tools: DJP JIS JMG ALP. Wrote the paper: DJP JIS JMG ALP.

PubMed Central

2024-06-05T04:04:19.255496

2011-3-9

PMC3052380

Introduction {#s1} ============ Prosocial behavior is a pervasive aspect of human life: We cooperate with others and help them when they are in need. However, diametrically opposed to these behaviors are everyday experiences of people taking advantage of others. The present study is concerned with the question whether compassion training can increase prosocial behavior. Compassion has been defined as the emotion one experiences when one feels concern for another\'s suffering and desires to enhance that person\'s welfare ([@pone.0017798-Baumeister1], see [@pone.0017798-Singer1] and [@pone.0017798-Fehr1] for more detailed definitions). In the present paper, we use the term "compassion" to describe an emotional as well as a motivational state, characterized by feelings of warmth, love, and concern for the other as well as the desire to help and promote the other\'s welfare. The term "empathic concern" has been used in a very similar way in developmental and social psychology [@pone.0017798-Davis1], [@pone.0017798-Batson1]. For example, Batson [@pone.0017798-Batson2] maintains that empathic concern "is an other-oriented emotional response elicited by and congruent with the perceived welfare of someone in need involving feelings *for* the other such as sympathy, compassion, tenderness, and the like." However, while empathic concern mainly denotes a situation-specific, rather short-living emotion, compassion can also be thought of as an attitude [@pone.0017798-Gilbert1]. Empirical evidence suggests that empathic concern is a perpetuator of prosocial behavior [@pone.0017798-Batson3], [@pone.0017798-Eisenberg1]. For example, it has been demonstrated [@pone.0017798-Batson4] that momentarily inducing feelings of empathic concern for a person in need by having participants focus on the person\'s feelings increases their prosocial behavior towards that person. More specifically, participants who were instructed to feel empathic concern for a person receiving painful electric shocks were willing to receive more shocks themselves to alleviate the other person\'s suffering than participants who had been encouraged to remain detached. The effects of this situation-specific induction of empathic concern, however, are probably rather short-lived and might not extend over the particular experimental session. Furthermore, it is not clear whether the induction of empathic concern for a specific person leads to increases of prosocial behavior only for that specific person \[e.g. 6\] or whether it generalizes to different persons as well [@pone.0017798-Greitemeyer1], [@pone.0017798-Twenge1]. While the experimental induction of empathic concern through explicit perspective-taking instructions or listening to songs with prosocial lyrics [@pone.0017798-Greitemeyer1] might temporarily prime people to experience empathy when seeing the distress of others, training of compassion aims at permanently changing people\'s motivation and their feelings towards other people. It strives to develop a more friendly, benevolent, connected and positive attitude towards others. In the long run, compassion training-induced changes at the trait level -- but not at the state level - might even take effect on the opiate- and oxytocin-based affiliative system [@pone.0017798-Gilbert1], [@pone.0017798-Depue1]. We hypothesize that, contrary to a short-term instruction-based induction of empathic concern towards a specific person, compassion training will elicit a longer-lasting enhancement of general compassionate motivation, which in turn may lead to an increase in the general tendency to act prosocially, independent of person and situation. Even though there is a long history of behavioral plasticity research pertaining to the training of cognitive [@pone.0017798-Lustig1], perceptual [@pone.0017798-Fahle1], motor [@pone.0017798-Gentili1] as well as affective skills [@pone.0017798-Golan1], [@pone.0017798-Slaski1], no study to our knowledge has investigated behavioral changes resulting from compassion training. Thus, in one study, for example, empathy for a personal offender was trained over eight 1-hour sessions and an increase in reported empathy and forgiveness but not prosocial behavior was measured [@pone.0017798-McCullough1]. Similarly, the few existing studies on compassion training have examined the effects of compassion training on mood and health but not prosocial behavior [@pone.0017798-Carson1]--[@pone.0017798-Pace1]. In a pilot study, Gilbert and Procter [@pone.0017798-Gilbert2] administered compassionate mind training \[CMT; 24\], which aims at reducing self-criticism by focusing on compassionate images and emotions, to a sample of psychiatric patients with severe long-term difficulties. They reported reductions in depression and anxiety as well as increases in self-soothing abilities and feelings of warmth for oneself. Other studies investigating the effects of compassion training have used meditation-based techniques that involve the development of warm, positive feelings towards a variety of people and ultimately towards all human beings: Six to seven weeks of meditation-based compassion training result in increases in positive mood and life satisfaction [@pone.0017798-Fredrickson1] as well as a reduction of interleukin-6 release in response to a psychosocial stressor [@pone.0017798-Pace1]. The more time participants had actually spent training, the stronger the reduction in interleukin-6 release, suggesting a dose-dependent effect of compassion training. Hutcherson et al. [@pone.0017798-Hutcherson1] report that a very brief (7-min) compassion meditation exercise results in a more positive attitude towards the target of the exercise. Taken together, these studies provide promising support for the health- and positivity-promoting effects of compassion training. However, so far, no study has investigated whether prosocial behavior can actually be increased through compassion training and whether the practice of compassion promotes a generalized tendency for prosocial behavior. Thus, the aim of the present study was to investigate the effect of short-term compassion training on prosocial behavior. In behavioral economics, prosocial behavior is usually studied in the context of well-controlled monetary exchange games [@pone.0017798-Camerer1] and mostly explained in terms of social preferences or norms, such as fairness and reciprocity [@pone.0017798-Fehr2], whose evolution has also been linked to reputation concerns [@pone.0017798-Nowak1], [@pone.0017798-Nowak2]. It could also be shown that observation of prosocial behavior in a public goods game with multiple rounds increases the likelihood of later prosocial behavior of the observer towards another person in the following rounds [@pone.0017798-Fowler1]. However, the influence of compassion or empathy and their training on prosocial behavior has so far never been discussed or studied in the field of economics. In the context of game theoretical paradigms, the dictator game is most commonly used for assessing altruistic acts towards others [@pone.0017798-Hoffman1]--[@pone.0017798-Forsythe1]. In this game, participants are endowed with a sum of money that they can split between themselves and another participant who has no money. Giving in the dictator game is likely driven by fairness norms and not by kindness [@pone.0017798-Bolton1]. While several motives have been discussed as underlying prosocial behavior, only recently a differentiation between norm-based and compassion-based prosocial behavior has been suggested [@pone.0017798-Singer1]. While the former is particularly encountered in "cold", reasoning-driven exchange situation, the latter is often present in "hot", emotion-provoking situations. Compassion training might take its effects on the latter but not the former. Since many of our everyday interactions are not purely rational, but involve emotions, an adequate paradigm that assesses prosocial behavior in an engaging, ecological setting and that is sensitive to affective interventions needs to be developed. This paradigm would moreover allow for future investigation of the proposed differentiation between norm-based and compassion-based prosocial behavior. In social psychology, prosocial behavior is mostly assessed in emotion-provoking one-shot helping situations of high ecological validity, such as dropping pens, soliciting donations for charities, or soliciting help with filling out or scoring questionnaires [@pone.0017798-Batson2], [@pone.0017798-VanLange1]--[@pone.0017798-Levine1]. However, these paradigms as well as the above-mentioned economic paradigms do not allow for the repeated assessment of prosocial behavior within the same person, which is required in intervention studies with multiple measurement time points such as the present study. We therefore developed a new prosocial task -- the Zurich Prosocial Game (ZPG) -- that allows for the repeated assessment of prosocial behavior within the same person while still being ecologically valid, and thus being suitable to investigate changes in prosocial behavior due to compassion training. In addition the new game was developed to simultaneously assess the influence of reciprocity, the cost associated with helping, and distress cues on prosocial behavior. It has been shown that people help more often if they have been helped before [@pone.0017798-Falk1], [@pone.0017798-Wilke1], if the costs of helping are low [@pone.0017798-Dovidio1] and if they are confronted with signs of distress [@pone.0017798-Batson5], [@pone.0017798-Carlo1]. These factors are of interest as evolutionary biologists and anthropologists demonstrated that they are selected for in evolution and provide a biological basis for altruism. Reciprocal altruism evolved as a costly altruistic act which might be repayed at a later time [@pone.0017798-Trivers1], costly helping is mostly directed towards kin as suggested by the model of inclusive fitness [@pone.0017798-Burnstein1] and distress cues, such as crying, evolved to signal the need for help and to sustain close personal bonds [@pone.0017798-Nelson1]. The possibility to distinguish between these helping-related factors within one task allows the investigation of differential effects of context, intervention or personality on different helping conditions in future research. Here, the aim was first to test the effect of a short-term compassion training on prosocial behavior in the ZPG. To validate the newly developed task and to test the effects of compassion training on prosocial behavior, we performed two independent experiments. The first experiment was conducted to validate the newly developed prosocial task, the so-called Zurich Prosocial Game (ZPG) and to test its sensitivity to the influence of reciprocity norms, helping costs and distress cues on helping. We hypothesized that people would help more a) if they had been helped before, b) if the cost of helping was low, and c) if they were confronted with distress cues. The second experiment was conducted to investigate the influences of short-term compassion training on prosocial behavior towards strangers as measured by the ZPG -- a game that is completely unrelated to the training context. We hypothesized that short-term compassion training leads to stronger increases in helping than a short-term memory training, the latter received by a control group. Furthermore, time spent practicing the compassion-enhancing technique should be positively correlated with this increase in helping. Based on the suggested distinction between compassion-based and norm-based prosocial behavior, and on the assumption that the compassion training has effects on the former this correlation could possibly only arise for non-reciprocity trials. Results {#s2} ======= Experiment 1 {#s2a} ------------ To investigate the effects of reciprocity, cost, and distress on the occurrence of prosocial behavior, we computed a 2 (reciprocity: no reciprocity, reciprocity) x 2 (cost: low, high) x 2 (distress: no distress cues, distress cues) within-subjects repeated-measures analysis of variance (ANOVA). This analysis revealed main effects of reciprocity, cost, and distress (see [Table 1](#pone-0017798-t001){ref-type="table"}). As hypothesized, participants helped significantly more in reciprocity trials, in trials with a low cost of helping, and in trials in which the co-player\'s virtual character expressed distress (see [Materials and Methods](#s4){ref-type="sec"} for a detailed description of the ZPG). Thus, the ZPG indeed seems to be sensitive to the three operationalizations of the influencing factors, which suggests that reciprocity, distress cues, and low cost are associated with increased helping behavior (see [Figure 1](#pone-0017798-g001){ref-type="fig"}). ::: {#pone-0017798-g001 .fig} 10.1371/journal.pone.0017798.g001 Figure 1 ::: {.caption} ###### Percent helping in the different conditions of the ZPG. Error bars denote standard errors of mean. :::  ::: ::: {#pone-0017798-t001 .table-wrap} 10.1371/journal.pone.0017798.t001 Table 1 ::: {.caption} ###### ANOVA for the effects of reciprocity, cost and distress cues on prosocial behavior in experiment 1 and 2. ::: {#pone-0017798-t001-1} Source df F partial *η^2^* *p* ------------------------------- ---- ------- ---------------- -------- Experiment 1 Reciprocity 67 73.22 .52 \<.001 Cost 67 73.78 .52 \<.001 Distress 67 7.02 .10 .01 Reciprocity x Cost 67 13.21 .17 .001 Reciprocity x Cost x Distress 67 4.10 .06 \<.05 Experiment 2 Reciprocity 68 51.55 .43 \<.001 Cost 68 66.04 .49 \<.001 Reciprocity x Cost 68 7.96 .11 \<.001 Reciprocity x Distress 68 6.11 .08 \<.05 All main effects and interactions significant on a p\<.05 level are reported. ::: The main effects of reciprocity and cost were qualified by a significant interaction between the two factors (see [Table 1](#pone-0017798-t001){ref-type="table"}). Increasing the cost of helping resulted in a larger decrease in prosocial behavior in the no-reciprocity compared to the reciprocity trials, suggesting that norms such as reciprocity can absorb the decline in prosocial behavior when helping is costly. Furthermore, there was a three-way interaction between the three factors (see [Table 1](#pone-0017798-t001){ref-type="table"}). Separate follow-up ANOVAs for distress and no-distress trials revealed that the reciprocity x cost interaction was only significant in the distress trials, *F*(1,67) = 17.73, *p*\<.001, partial *η^2^* = .21. To confirm that interindividual differences in helping behavior as measured using the ZPG are not brought about by differences in allocation of attention, we calculated the correlation between the total amount of helping and the percentage of stars picked up, that randomly appeared during the game. On average, participants picked up the star in 42.6% of the trials in which a star appeared. A star appeared in two to six (of nine) trials. The non-significant correlation, *ρ*(66) = −.07, *p*\>.05, between total amount of helping and percentage of stars picked up indicates that differences in attention allocation most likely do not account for interindividual differences in prosocial behavior. Furthermore, to control for potential effects of individual differences in risk preferences on helping in the high cost trials, the risk questionnaire and the lottery index were correlated with helping in high cost trials (see [Materials and Methods](#s4){ref-type="sec"} for a description of the risk perception control measures). As both correlations were non-significant, we can preclude that interindividual differences in risk preferences, *ρ*(66) = .10, *p*\>.05 (risk questionnaire) and *ρ*(66) = .02, *p*\>.05 (lottery index), accounted for the difference in prosocial behavior between low and high cost trials. Furthermore, participants had to judge their engagement in the game (see [Materials and Methods](#s4){ref-type="sec"}). The analyses of these subjective engagement scores revealed that on average participants indicated that they were very engaged in the game (range: 2--5; mean: 4.15, SD: 0.74); a result which matches the observation of the experimenter who reported that the subjects were all very immersed in the ZPG. To assess the convergent validity of the ZPG, participants played the dictator game [@pone.0017798-Hoffman1] (see [Materials and Methods](#s4){ref-type="sec"}). Based on our reasoning about norm-based and compassion-based prosocial behavior in the [introduction](#s1){ref-type="sec"}, we did not expect an exceedingly high correlation between the ZPG and the dictator game, but still, as both tasks assess variants of prosocial behavior, a sufficiently high correlation to maintain that the ZPG indeed measures prosocial behavior. In the dictator game, participants gave 36.29% of their endowment on average. Most of the participants (40.3%) gave half of their endowment and 10.4% gave nothing. As expected, giving behavior in the dictator game correlated with helping behavior in the ZPG, *ρ*(65) = .35, *p* = .004, substantiating the validity of our game as a measure of prosocial behavior. To assess the divergent validity of the ZPG, we used a memory task (see [Materials and Methods](#s4){ref-type="sec"}). Participants remembered 18.24 words (standard deviation \[sd\]: 5.86) on average in the memory task. The number of remembered words did not correlate significantly with helping in the ZPG, *ρ*(64) = .06, *p*\>.05, demonstrating divergent validity of the ZPG. Experiment 2 {#s2b} ------------ In the following, we will first present the pre-training data from the newly developed ZPG to ascertain whether the results found in Experiment 1 are robust. We will then report data on the effectiveness of the compassion training workshop and on the effects of compassion training on prosocial behavior in the ZPG. The effects of compassion training were tested one-sided as we had clear hypotheses about the direction of effects (see [Introduction](#s1){ref-type="sec"}). ### Robustness of ZPG {#s2b1} To investigate the robustness of the result pattern in the ZPG, we computed a 2 (reciprocity: no reciprocity, reciprocity) x 2 (cost: low, high) x 2 (distress: no distress cues, distress cues) within-subjects repeated-measures ANOVA for the total sample of Experiment 2 (compassion group and memory group). The analysis again revealed main effects of reciprocity, and cost (see [Table 1](#pone-0017798-t001){ref-type="table"}). Participants helped significantly more in reciprocity trials and in trials with a low cost of helping. As in Experiment 1, these main effects were qualified by a significant interaction between the two factors, again suggesting that norms such as reciprocity can absorb the decline in prosocial behavior when helping is costly. In contrast to Experiment 1, however, no main effect of distress was observed, *F*(1,68) = 1.89, *p*\>.05, partial *η^2^* = .03. The analysis did however yield a reciprocity x distress interaction: Distress cues increased helping in the no-reciprocity but not in the reciprocity trials, which might indicate that no-reciprocity trials are more sensitive to other influencing factors (see [Table 1](#pone-0017798-t001){ref-type="table"}). We again did not observed a significant correlation between total amount of helping and percent of stars picked up, *ρ*(67) = −.12, *p*\>.05, indicating that differences in attention allocation most likely do not account for interindividual differences in helping behavior. As in Experiment 1 the correlations between helping in the high cost trials with both the risk questionnaire, *ρ*(57) = .02, *p*\>.05, and the lottery index, *ρ*(57) = .05, *p*\>.05, were non-significant, precluding that interindividual differences in risk preferences accounted for the difference in prosocial behavior between low and high cost trials. Participants in experiment 2 also reported to be very engaged in the game (range: 1--5; mean: 4.04, SD  = 1.07). There was no differences between participants in the compassion and memory training group in the engagement with the game, t(52) = 1.78, p\>.05, suggesting that potential differences between the groups cannot be accounted for by differences in motivation and degree of being emerged into the game. In the dictator game, on average, participants gave 33.8% of their endowment. Most of the participants (39.4%) gave half of their endowment and 13.6% gave nothing. More importantly, giving behavior as measured with the dictator game again correlated with helping behavior as measured with the ZPG, *ρ*(67) = .45, *p*\<.001. Participants remembered on average 20.53 words (sd: 7.04) in the memory task. And as in Experiment 1, the number of remembered words did not correlate with helping in the ZPG, *ρ*(66) = −.04, *p*\>.05, giving repeated evidence for divergent validity of the ZPG. ### Effectiveness of compassion training {#s2b2} Repeated-measures ANOVAs with time (pre-training, post-training) as a within-subjects factor and training (compassion, memory) as a between-subjects factor were calculated to determine the effectiveness of the compassion training in enhancing self-reported positive (assessed with the Positive and Negative Affect Scale \[PANAS; 47\]) and compassionate (assessed with the Compassionate Love Scale \[CLS; 48\]) feelings and reducing negative feelings (also assessed with the PANAS [@pone.0017798-Watson1]). A significant main effect of time on positive mood, *F*(1,54) = 23.47, *p*\<.001, partial *η^2^* = .30, was revealed, indicating that compassion training as well as memory training increased positive mood. A significant main effect of time, *F*(1,54) = 5.84, *p* = .02, partial *η^2^* = .10, was revealed for compassionate feelings that was qualified by a marginally significant interaction between time and training, *F*(1,54) = 3.61, *p* = .06, partial *η^2^* = .06. Post-hoc t-tests indicated that only the compassion-training group experienced a significant increase in compassionate feelings, *t*(23) = 2.66, *p* = .01. For negative mood, a significant time x training interaction was revealed, *F*(1,68) = 6.11, *p* = .016, partial *η^2^* = .08. While negative mood decreased in the compassion-training group, *t*(23) = −1.94, *p* = .03, one-sided, it marginally significantly increased in the memory-training group, *t*(23) = 2.02, *p* = .05. ### Effect of compassion training on prosocial behavior {#s2b3} To test whether a brief compassion training had an effect on prosocial behavior in the ZPG, we conducted two analyses: First, we performed a 2 (time: pre-training, post-training) x 2 (reciprocity: no reciprocity, reciprocity) x 2 (cost: low, high) x 2 (distress: no distress cues, distress) repeated-measures ANOVA with training (compassion, memory) as a between-subjects factor. Second, we tested for increases in helping as a function of interindividual differences in hours of reported training. To this end, we calculated the correlation between participants\' self-reported time spent praticing outside of the training and the change in helping from pre- to post-training (self-report data could only be obtained from a subset of the samples: *n* ~compassion~ = 19, *n* ~memory~ = 22). In the first analysis, we observed a significant time x training interaction, *F*(1,57) = 4.09, *p* = .05, partial *η^2^* = .07. While there was no reliable change in helping from pre- to post-training for the memory training group, *t*(31) = −1.20, *p* = .24, compassion training significantly increased helping, *t*(26) = 1.85, *p* = .04, one-sided (see [Figure 2](#pone-0017798-g002){ref-type="fig"}). Additionally, a time x cost interaction was observed, *F*(1,57) = 6.76, *p* = .01, partial *η^2^* = .11. These interactions were qualified by a significant three-way interaction between time x cost x training, *F*(1,57) = 4.55, *p* = .04, partial *η^2^* = .07. Follow-up independent t-tests indicated that, at pre-training, helping in the low-cost, *t*(57) = 1.21, *p* = .23, and high-cost trials, *t*(57) = .55, *p* = .58, did not differ between the compassion and the memory group whereas, at post-training, the groups differed significantly in helping in both the low-, *t*(57) = 3.07, *p* = .003, and the high-cost trials, *t*(57) = 2.27, *p* = .03. ::: {#pone-0017798-g002 .fig} 10.1371/journal.pone.0017798.g002 Figure 2 ::: {.caption} ###### Effects of training on overall helping in the ZPG for the compassion-training and memory-training group. Error bars denote standard errors of mean. \* p\<.05, one-sided. :::  ::: The second analysis did not reveal the hypothesized correlation between interindividual differences in reported hours of compassion training and total helping, *ρ*(17) = .27, *p* = .13, one-sided. To test our expectation that interindividual differences in reported training hours are differentially related to the different trial types, affecting more non-reciprocity than reciprocity trials, we calculated the correlation between training hours and helping in the different trial types and found a significant correlation between interindividual differences in reported hours of compassion training and helping in no-reciprocity trials, *ρ*(17) = .39, *p* = .05, one-sided. Correlations with the other trial types as well as all correlations in the memory group were non-significant at *p*\<.05, one-sided. To investigate whether compassion training could also increase giving in the dictator game, we computed a repeated-measures ANOVA with time (pre-training, post-training) as a within-subjects factor and training (compassion, memory) as a between-subjects factor. Interestingly, neither a significant main effect of time, *F*(1,56) = .02, *p* = .89, partial *η^2^*\<.001, nor an interaction between time and training *F*(1,56) = 1.63, *p* = .21, partial *η^2^*\<.001, was observed. Participants in neither group gave more money to the other person after training compared to before training. Discussion {#s3} ========== The present study introduces a newly developed prosocial game -- the Zurich Prosocial Game (ZPG) -- and provides first evidence for the effectiveness of short-term compassion training in enhancing prosocial behavior in this new training-unrelated game towards strangers. The ZPG was developed as previous prosocial tasks from behavioral economics or social psychology are either not particularly ecologically valid or do not allow for the repeated assessment of prosocial behavior which is required in intervention studies with multiple measurement time points. The ZPG extends the prosocial tasks from behavioral economics and social psychology in several aspects. First, the influence of reciprocity, cost, and distress on prosocial behavior has been studied separately before, but the ZPG now has the advantage to allow their simultaneous assessment in the same setting. This is particularly useful when studying the differential influence of experimental manipulations on these factors. Second, while many prosocial tasks are only applicable one time, the ZPG can be played multiple rounds and on different time points with the same subjects thus allowing for more stable estimates of prosocial behavior, the usage in neuroscientific settings where many trial repetitions are needed, and, the assessment of changes in prosocial behavior over time through interventions. Third, the ZPG is more ecologically valid than, for example, monetary exchange games, as it minimizes the influence of strategic considerations, minimizes effects of task-affordances due to explicit instructions and maximizes the influence of emotion-driven, fast decisions, since participants are immersed in the game itself whose explicit goal is to achieve a treasure in short time rather than act prosocially. In the ZPG participants help others by spending ressources (key, time) they might need later on. This type of prosocial behavior that involves uncertainty for oneself can be encountered often in daily life, for example when we run for an important appointment and see someone fall from his bike. Do we stop to help this person without knowing the outcome and how much time it will take or do we refrain from helping and make sure that we reach our appointment on time? And fourth, as the ZPG is very engaging and easy to use, it is also very well suited to study prosocial behavior in children. The present results confirm that the ZPG is sensitive to influences of reciprocity, cost, and distress on prosocial behavior: As predicted, participants of two independent experiments helped more when having been helped before and when costs are low. Interestingly, the drop in prosocial behavior with increasing costs was less pronounced when participants had been helped before suggesting that norms of reciprocity override cost considerations. While in Experiment 1 participants helped more when confronted with distress cues, in Experiment 2, distress cues increased helping only in the no-reciprocity trials but not in the reciprocity trials. This may again suggest that reciprocity norms are so pervasive that they overrule the effect of any other influencing factor, whereas prosocial behavior without reciprocation is more affected by other factors. In both samples, the convergent validity with a well-established economic prosocial task, the dictator game, was confirmed. This supports our claim that the new game does indeed assess prosocial behavior. The correlation between the two tasks, however, is modest, suggesting that the two measures tap into different aspects of prosocial behavior. Furthermore, divergent validity was established with a memory task. In Experiment 2, we were able to demonstrate that compassion training but not memory training significantly increased helping in the ZPG. Previous studies have demonstrated that a momentary instruction-based induction of empathic concern for a specific person increases prosocial behavior towards that person immediately after induction [@pone.0017798-Batson3]. Here we show for the first time that compassion training had longer-lasting effects on prosocial behavior as the post-test was completed two to five days after training. Furthermore, short-term training resulted in transfer to behavior in a novel task that was completely unrelated to the previous affective training. Finally, compassion training increased prosocial behavior towards people who were not specifically targeted during training but complete strangers to the participants. The present results support the notion that similar to situation-specific induction of empathic concern for a specific person in need [@pone.0017798-Batson3], [@pone.0017798-Eisenberg2], the training of compassionate motivation leads to increases in prosocial behavior. In comparison to experimental inductions of empathic concern, however, compassion training has the potential to lead to longer-lasting changes in people\'s attitude and behavior towards other people that transcend the specific situation in which compassionate feelings were evoked and transfer to a much broader range of people and situations. Self-benefiting effects of compassion training such as increases in positive mood, life satisfaction, decreased depressive symptoms [@pone.0017798-Fredrickson1], and less reactivity to psychosocial stress [@pone.0017798-Pace1] have been reported before. The present study adds to these findings by showing that even a short-term compassion training may not only have benefits for the practitioner\'s health and subjective well-being but also for other people and society in general as it increases the propensity to act prosocially even towards people one has never met. Notably, the prosocial behavior observed here was not directed towards a target of the compassion training but to random strangers and was assessed at least two days after the training, which lends further credibility to the societal impact that the implementation of compassion training in schools, organizations, and clinical settings might have (for compassion training in psychotherapy, see [@pone.0017798-Gilbert1]). Another interesting finding of the present study was that helping in no-reciprocity trials, but not helping in reciprocity trials, was related to interindividual differences in reported training hours in the compassion group. This might provide tentative evidence for a differentiation between compassion-based and norm-based prosocial behavior as has been suggested before [@pone.0017798-Singer1]. Accordingly, helping after having been helped may rely on a felt obligation to reciprocate cooperation. In contrast, helping without the possibility for reciprocity may be motivated more by feelings of compassion than by "cold" norms. The pattern of correlations found here suggests that compassion training might have differential effects on both types of underlying motivation. This is further supported by our finding that giving in the dictator game did not change from pre- to post-training in either group and that the modal giving at pre-training was 50%. Moreover, the correlation between helping in the ZPG and giving in the dictator was modest suggesting that the two measures tap into different aspects of prosocial behavior. Giving in the dictator game has previously been shown to depend more on fairness norms than on kindness [@pone.0017798-Bolton1]. These findings suggest a distinction between compassion-based and norm-based prosocial behavior with compassion training possibly exerting a stronger effect on the former than on the latter. As the current study was not designed to test the hypothesis of a distinction between compassion-based and norm-based prosocial behavior, future investigations are needed. For example, using priming of a reciprocity-norm or compassion could be used to show a differential effect of these concepts on different helping settings. The simultaneous assessment of reciprocity and non-reciprocity driven prosocial behavior in the ZPG makes this game ideal for this aim. Similarly, investigations with longer training will be of great interest. As the results suggest, the novel ZPG might be a better measure for assessing training-induced changes in prosocial behavior, specifically compassion-based prosocial behavior, than standard economic games or psychological measures as we were able to show that it is more sensitive to change than, for example, the dictator game. The higher sensitivity to changes in compassion-based prosocial behavior might result from the high emotional engagement participants experience when playing the ZPG. Importantly, as the game is framed as a treasure hunt with monetary gains, demand effects induced by the content of the training should be less strong than in other prosocial tasks. While in economic games the sharing purpose is made explicit, here the instruction focuses on the rules of the game and emphasize that the goal is to reach the treasure in a limited time while having to overcome certain obstacles. Furthermore, the game is very engaging (on average participants rate their involvement in the game with 4 on a 1-to-5-scale and report later that they find the game very enjoyable) and fast, thus making strategic considerations difficult. Compassion training not only increased prosocial behavior but also led to increases in reported compassionate feelings and positive affect and a decrease in negative affect. Interestingly, the memory-training group also evinced an increase in positive mood, suggesting that increases in positive mood are not sufficient for explaining enhanced prosocial behavior. We maintain that compassion training enhanced prosocial behavior through initial changes to participants\' way of feeling and thinking about other people to a more positive, benevolent and friendly attitude. This is in line with participant\'s qualitative post-study reports of being more sensitive to others, feeling more connected, secure and open and having "a bigger and more open heart." The present study provides first evidence for compassion training but not memory training causing increases in prosocial behavior. Future studies should elucidate, which aspects of the training led to the observed effect. Apart from the suggested change in other-related attitudes, increased relaxation or feeling of oneness (perceived self-other overlap; [@pone.0017798-Cialdini1]) could be additional mechanisms through which compassion training increases prosocial behavior. In sum, the present study provides first evidence for the effectiveness of a short compassion training in increasing prosocial behavior in a newly developed computer task, the Zurich Prosocial Game. Using this novel training-unrelated computer task, we found that compassion training that aimed at fostering a friendly, benevolent attitude towards others produced a significant increase in prosocial behavior two to five days after training towards strangers. Interestingly, practicing compassion strategies seems to influence compassion-based prosocial behavior more strongly than norm-based prosocial behavior. The effectiveness of the compassion training was further supported by an increase in positive mood and compassionate feelings and a decrease in negative mood. Future research with longer training and bigger sample sizes needs to ascertain how long lasting these effects are and who is benefitting from compassion training. Clinical research for example suggests that some people find compassion-focused imagery distressing [@pone.0017798-Gilbert4], [@pone.0017798-Rockliff1] and thus do not benefit from it. Furthermore it needs to be investigated whether long-term compassion training leads to stronger increases in specific types of prosocial behavior and whether this effect can also be observed in everyday life behavior. As the interpersonal effects were directed towards total strangers and transferred to situations outside the training context, compassion training could have great societal impact when implemented in institutions of daily life. Materials and Methods {#s4} ===================== Participants {#s4a} ------------ In Experiment 1, that aimed to validate the newly developed ZPG, we investigated 68 healthy female volunteers (aged 18--35 years; mean: 25.18; years of education after the 16^th^ birthday: 2--15 years; mean: 6.54). In Experiment 2, that aimed to assess the effect of a short-term compassion training workshop on prosocial behavior as measured using the newly developed ZPG, we investigated 69 healthy female volunteers (age: 18--34 years; mean: 23.69). Only female participants were included in Experiments 1 and 2 because of better performance in emotional tasks [@pone.0017798-Kring1] and higher self-reported empathy in women [@pone.0017798-BaronCohen1]. All participants came from the University of Zurich and the surrounding communitya and were recruited through local advertisement and internet postings. The advertisements for Experiment 2 asked for people interested in mental training but never mentioned the word compassion. All participants completed the Toronto Alexithymia Scale \[TAS; 55\], the Beck\'s Depression Inventory \[BDI; 56\] and sociodemographic questions online. Only when they met the following inclusion criteria they were contacted via telephone: aged 18--35 years, TAS \<60, BDI \<18, right hander and no contraindication for fMRI. Importantly, possible participants of Experiment 2 were additionally not allowed to have prior experience with mental compassion training or the method of loci. On the phone, participants were given information about the timing but, importantly, in case of Experiment 2, not about the specific content of the study and underwent a structured psychological interview (screening questions for axis-I disorders and psychotic disorders of the Structured Clinical Interview for DSM Disorders \[SCID; german version: 57\]). Woman with current psychiatric illnesses were excluded from the study. For Experiment 2, allocation to the compassion-training and memory-training (control) group depended on slot availability and time of the participants. 35 participants entered the compassion-training group and 34 participants entered the memory-training group. 28 participants from the compassion group and 32 participants from the memory group completed the study. One participant of the compassion group was eliminated from the analysis as data on the ZPG was missing. The majority of the dropout in the compassion group (5/7) occurred before the training. Furthermore, the seven participants that dropped out of the study did not differ in age, *t*(32) = .57, *p*\>.05, years of education, *t*(32) = 1.75, *p*\>.05, empathic concern, *t*(32) = −.81, *p*\>.05, alexithymic symptoms, *t*(32) = 1.70, *p*\>.05, depressive symptoms, *t*(32) = −.75, *p*\>.05, prosocialness, *t*(32) = −.79, *p*\>.05, compassionate feelings, *t*(32) = −.99, *p*\>.05, and general positive, *t*(32) = 1.42, *p*\>.05, and negative affect, *t*(32) = −.06, *p*\>.05, from the participants that finished the study, thus excluding selective dropout in the compassion group. The compassion group and the memory group did not differ in age, *t*(57) = 1.98, *p*\>.05, years of education, *t*(57) = .64, *p*\>.05, prosocialness, *t*(54) = 1.91, *p*\>.05, empathic concern, *t*(54) = .1, *p*\>.05, alexithymic symptoms, *t*(57) = .89, *p*\>.05, or depressive symptoms, *t*(57) = .87, *p*\>.05. There was also no difference in the distribution of type of education between the samples (*χ^2^* = 3.06, *p*\>.05). See [Table 2](#pone-0017798-t002){ref-type="table"} for sample characteristics. The study was approved by the Research Ethics Committee of Zurich ("Kantonale Ethikkommission des Kantons Zürich -- Spezialisierte Unterkommission Psychiatrie, Neurologie, Neurochirurgie"; E-25/2008) and was performed according to the Declaration of Helsinki. All participants gave written informed consent after having received a full description of the study. ::: {#pone-0017798-t002 .table-wrap} 10.1371/journal.pone.0017798.t002 Table 2 ::: {.caption} ###### Sample characteristics. ::: {#pone-0017798-t002-2} Validation sample (N = 68) Compassion training sample (N = 27) Memory training sample (N = 32) -------------------------------------------------- ------------------------------------------------------------------------- ------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------- Age 25.18 (4.08) 24.74 (4.22) 22.66 (3.86) Highest completed education Apprenticeship: 5 (7.5%) High school: 34 (50.8%) University: 28 (41.8%) Apprenticeship: 3 (11.1%) High school: 18 (66.7%) University: 6 (22.2%) Secondary school: 1 (3.1%) Apprenticeship: 1 (6.3%) High school: 24 (75%) University: 4 (12.6%) PhD: 1 (3.1%) Education (years after 16^th^ birthday) 6.54 (2.87) 5.48 (2.44) 5.06 (2.54) Prosocialness[1](#nt102){ref-type="table-fn"} 64.03 60.75 64.50 Empathic concern[2](#nt103){ref-type="table-fn"} 27.64 27.08 27.19 Alexithymia[3](#nt104){ref-type="table-fn"} 41.24 39.41 41.16 Depression[4](#nt105){ref-type="table-fn"} 6.13 6.04 4.53 1 Prosocialness Scale [@pone.0017798-Caprara1] (range: 16--80). 2 Empathic Concern Subscale Interpersonal Reactivity Index [@pone.0017798-Davis1] (IRI; range: 7--35). 3 Toronto Alexithymia Scale [@pone.0017798-Kring1] (TAS;\>60 clinically relevant). 4 Beck\'s Depression Inventory [@pone.0017798-BaronCohen1] (BDI;\>18 clinically relevant). ::: Measures {#s4b} -------- ### Zurich Prosocial Game {#s4b1} A novel game, the Zurich Prosocial Game (ZPG), was developed that allows for repeated assessment of prosocial behavior and for parsing the influence of reciprocity, cost, and distress on prosocial behavior. The participants\' task is to navigate a virtual character through a maze and reach a treasure in a limited amount of time. Each treasure is worth 0.50 Swiss francs (∼\$ 0.50). At the same time, participants see the virtual character of an ostensible co-player from another research institute in Europe who is also trying to reach a treasure. Importantly, the two players do not share the same paths in the maze and do not compete for the same treasure. Thus, in principle, the game can be played while completely ignoring the other player. Participants are told that in each round of the game they are connected via the internet with a new co-player who is sitting in a different research institute in Europe. At the onset of each round, the participant and the ostensible co-player select one of two paths. While the players move their virtual character through the maze, red and blue gates fall on the paths that can block the participant and the co-player. Each of the two players is equipped with red and blue keys with which they can open the corresponding gates. When the co-player runs out of keys, participants can use their own keys to open the gates for them. Importantly, participants cannot delay their help to observe the progression of the game (i.e., whether they need their keys themselves) as the virtual characters become inactive before the next gate falls and thus cannot reach the treasure anymore. During each trial, participants can see how many gates are still going to fall, which and how many keys they and the co-player still possess, and how much time is left (for a screenshot of the game display, see [Figure 3](#pone-0017798-g003){ref-type="fig"}). When playing the game, participants wear headphones as sounds convey distress cues in the distress trials and add emphasis to events on the screen (e.g., sound when a gate is falling). Importantly, to reduce demand effects, participants are never told that the purpose of the game is to help the co-player. Instead, the instructed aim of this computer game is to reach a goal, the treasure, in a short amount of time to optimize monetary winnings. ::: {#pone-0017798-g003 .fig} 10.1371/journal.pone.0017798.g003 Figure 3 ::: {.caption} ###### Labeled screenshot of the ZPG. Participants move their virtual character forward by clicking with the mouse on the field in front of it. Usage of keys in order to open the blocking gates occurs by mouse click on the key matching the gate\'s color. Collection of stars also occurs by clicking on them with the mouse. :::  ::: Different trial types were introduced to probe the effect of different factors on prosocial behavior. First, to assess the influence of reciprocity on prosocial behavior, no-reciprocity and reciprocity trials were created. In the no-reciprocity trials, participants had the opportunity to help the co-player while knowing that the co-player would not have any opportunity to reciprocate as either no gates at all or no gates that the participant could not open with her own keys were still going to fall. In the reciprocity trials, participants had the opportunity to help the co-player after the co-player had helped them earlier in the trial. In these trials, participants could also see that there would not be any opportunity for the co-player to reciprocate. By designing the trials this way and by changing the co-player for each trial, we excluded the possibility of participants helping because they anticipated that they might need the co-player to reciprocate later on. To assess the influence of helping cost on prosocial behavior, there was a low- and high-cost variant of all trial types. In the high-cost variant, participants knew that after they helped the co-player, there would be a 25% chance that they would need the donated key to reach the treasure themselves; in the low-cost variant, players knew that they could donate keys without risking needing them later themselves, the only cost in this condition being loss of time. Finally, to investigate the effect of distress cues, when the co-player\'s virtual character was blocked, it either a) started to cry and sweat, as implemented by visual changes in the virtual character and by crying sounds that participants heard over headphones (distress cues) or b) gave no distress cues. This resulted in a 2×2×2 factorial design with the three factors reciprocity (no reciprocity, reciprocity), cost (low, high) and distress (no distress, distress). One game consisted of nine trials, one of each type, plus one trial in which no helping was necessary to reduce the affordance of the game. Trial types appeared in random order with the restriction that the first reciprocity trial could appear at the earliest as the third trial. This restriction was introduced to reduce the likelihood of an anchoring effect (helping agreed upon behavior) being introduced by experiencing a helpful co-player right away. At the beginning of the game, participants were first given written and verbal instructions and asked five questions probing their comprehension of the game. Then participants completed four practice trials to familiarize themselves with the handling of the game and to determine individual reaction time thresholds. To offset individual differences in speed and proficiency with computer games, the individual time limit for all trials of a given game was set at the average time the individual required to reach the treasure in the four practice trials plus 5 s. Furthermore, to control for the possibility that interindividual differences in helping might be due to differences in participants\' allocation of attention to their own and the co-player\'s path, we let a star appear randomly on some trials. The star yielded 0.20 Swiss francs (∼\$ 0.20) when picked up. This was expected to result in the allocation of attention to the whole display, as the star could appear anywhere. If participants collected stars but refrained from helping, attentional influences on helping behavior could most certainly be ruled out. ### Risk perception control {#s4b2} As interindividual differences in the behavior in the high cost trials could be brought on by differences in risk preferences, i.e., participants differ in their perception of the risk of not reaching the treasure in the high cost trials, we assessed risk preferences. First, we asked participants on an eight-point scale how risk-seeking they are and, second, we presented them with seven lotteries where the amount that can be won (6 Swiss Francs) stays the same but the amount that can be lost varies (1--7 Swiss Francs). Participants can decide for each lottery whether they want to play it or not. The computer then randomly picks one lottery and the outcome of this lottery is paid out to the participants if they had decided to play it. The number of lotteries accepted is an index for risk preferences. ### Engagement with the game {#s4b3} Participants were asked after playing the game to indicate on a five-point scale how engaged they were when playing the game. A high engagement of the participants would indicate that they were emerged in the game and diminish the probability that demand effects and strategic decision-making influenced prosocial behavior in the ZPG. ### Dictator game {#s4b4} To assess the convergent validity of the ZPG, participants played the dictator game [@pone.0017798-Hoffman1]. Based on our reasoning about norm-based and compassion-based prosocial behavior in the [introduction](#s1){ref-type="sec"}, we did not expect an exceedingly high correlation between the ZPG and the dictator game, but still, as both tasks assess variants of prosocial behavior, a sufficiently high correlation to maintain that the ZPG indeed measures prosocial behavior. Participants were again told that they would be paired with another player from another research institute in Europe. In the dictator game, based on random assignment, participants are endowed with 80, 120, or 160 points that they can split between themselves and an ostensible co-player who has no points. Points are later converted to money with a conversion scheme of one point equaling six, four, or three Swiss rappen (or "Swiss penny"), respectively. ### Memory task {#s4b5} To assess the divergent validity of the ZPG, we used a memory task that was later used as an outcome measure for the memory training group in the intervention experiment. Participants were presented with 34 words on the computer screen and were asked to memorize them and their sequence. Each word appeared for four seconds and words were separated with a 2-sec presentation of a crosshair. After the presentation, a wordfile opened and participants had five minutes to remember as many words as possible in the correct sequence. ### Effectiveness measures -- Experiment 2 {#s4b6} To measure the effectiveness of the compassion-training workshop, we assessed the difference in mood and compassionate feelings reported before and after compassion training. Participants completed the Positive and Negative Affective Scale \[PANAS; 47\] and the Compassionate Love Scale \[CLS; 48\]. If effective, compassion training should lead to increases in positive mood, compassionate feelings, and possibly to a decrease in negative mood. Experiment 1 {#s4c} ------------ ### Procedure {#s4c1} All participants gave written informed consent after having received a description of the study. They were told that they would play interactive computer games via the internet with other participants in different research institutes across Europe in order to investigate cross-cultural differences in interpersonal behaviors. In reality, there were no co-players; the ostensible co-players\' behavior was pre-programmed. Participants were seated in front of a computer and the experimenter provided oral and written instructions to the Zurich Prosocial Game (see below). Participants then answered five questions testing their comprehension of the game. The experimenter then checked the answers to ensure that participants fully understood the rules of the game. Then participants put on headphones and the experimenter ostensibly logged the participant into the game network. An abbreviated version of the instructions appeared on the computer screen and participants were asked to enter a freely chosen nickname to play the game. When ready, participants started the game and played four practice rounds after which they again had the opportunity to ask questions before playing the actual game. After finishing the Zurich Prosocial Game, the experimenter provided oral and written instructions explaining the dictator game (see below) and ostensibly logged the participant into the game network again. After playing the dictator game, participants completed a memory task (see below), filled out questionnaires (see below) and completed a lottery task to assess risk preferences (see below). Participants also completed a task in the magnetic resonance (MR) scanner and other unrelated tasks (results to be reported elsewhere). All participants were debriefed after the study was completed. Experiment 2 {#s4d} ------------ Procedure {#s4e} --------- Participants came to the lab one to two weeks prior to the training for their pre-training measurement (pre-test) and two to five days after the training workshop for their post-training measurement (post-test). The pre- and post-training measurements were identical, except that risk-preferences were only assessed at pre-test. A detailed description of the measurement procedure with respect to the ZPG can be found in the documentation of Experiment 1. Briefly, participants first played the ZPG and the dictator game, both under the assumption that they were playing the games with other participants in research institutes across Europe. Afterwards, participants completed the memory task and a lottery task to assess risk preferences (only at pre-test). Then, in contrast to the validation sample in Experiment 1, at post-test participants filled out questionnaires that were to probe the effectiveness of the compassion training (see below). Participants also completed a task in the MR scanner and other unrelated tasks (results to be reported elsewhere). Thus, the newly developed Zurich Prosocial Game was assessed in the context of several other non-helping tasks, which further helped to reduce possible demand effects of the compassion training. Participants were asked to continue praticing after the training (see below for training details) in the days before post-test. To facilitate continuation, we offered a one-hour guided evening training session on each of these days. Participants were debriefed after the end of the study. ### Compassion and memory training {#s4e1} The compassion group attended a one-day training to learn a compassion- enhancing technique developed in Buddhist contemplative traditions. This compassion meditation technique (called "Metta" in Pali) aims to foster an attitude of loving kindness, emotional positivity, benevolence, and friendliness towards oneself and others [@pone.0017798-Carson1], [@pone.0017798-Salzberg1]. An experienced meditation teacher with over ten years of teaching experience led the training workshop. The training involves sitting in an upright position and developing warm, positive feelings sequentially towards oneself, a beloved person, a neutral person, a person one has difficulties with, and all human beings by imagining each while silently repeating sentences like "May you be happy" or "May you be safe" and cultivating these positive emotional attitudes towards the visualized persons. The training day was held in silence and lasted for six hours in which mental training was sometimes done while sitting and sometimes while walking. The mental training periods were usually between 15 and 30 minutes long. There was a 45 minutes lunch break in between. During the course of the training, the target of the compassion meditation changed in the following succession: oneself, beloved person, neutral person, difficult person, all human beings. Ultimately, this should lead to an attitude of emotional positivity, benevolence, and friendliness towards oneself and others [@pone.0017798-Salzberg1]. Thus, as in compassion-focused therapy [@pone.0017798-Gilbert1], compassion here is trained as a skill. In contrast to a momentary induction of empathic concern through instruction to feel for a specific person in distress in a specific situation, compassion training aims at permanently changing one\'s motivation and attitude towards others in general. The memory control group underwent a one-day training workshop in the method of loci, a technique used to memorize items in an ordered sequence [@pone.0017798-Bower1], [@pone.0017798-Verhaeghen1]. An experienced memory technique teacher with over ten years of teaching experience led the training workshop. The method of loci involves linking a series of locations (e.g., a learned route through Zurich) with a series of specific items (a list of words) by creating visual mental images that combine each item with a location. For example, in order to remember the word "egg," one would imagine a big fried egg hanging down from the towers of the cathedral in Zurich. During recall, one recreates the images by mentally walking from one location to the next. This particular mnemonic technique was chosen for the control group as it contains most elements also needed in the compassion training workshops: People need to actively engage in inner mental processing and to create active mental images and specific associations between items. The difference is that the memory group focuses purely on improving cognitive rather than affective skills. Participants of both groups were asked to continue training in the days before post-test (1--3 days) and keep a diary of their practice. They were asked to join the daily offered one-hour evening training sessions or, if this was not feasible, to train at home. We thank Alexandra Freund for helpful comments and suggestions on previous versions of this manuscript, Astrid Hopfensitz for valuable ideas during task development, and George Margaris for programming the ZPG. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by the European Research Council under the European Community\'s Seventh Framework Programme (FP7/2007-2013)/ERC Grant agreement number 205557 \[EMPATHICBRAIN\] (<http://erc.europa.eu>); by the Swiss National Science Foundation through the National Centre of Competence in Research \"Neural Plasticity and Repair,\" Project 5: Cortical Plasticity (<http://www.nccr-neuro.ethz.ch>); by the University of Zurich (<http://www.uzh.ch>); and by the Mind and Life Institute (<http://www.mindandlife.org>). The authors gratefully acknowledge support by the Research Priority Program \"Foundations of Human Social Behavior\" at the University of Zurich. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: TS SL OK. Performed the experiments: SL OK. Analyzed the data: SL. Wrote the paper: SL TS OK.

PubMed Central

2024-06-05T04:04:19.260532

2011-3-9

PMC3052382

Introduction {#s1} ============ Release of the unconventional neurotransmitter nitric oxide (NO) is contingent on the activity of an enzyme, nitric oxide synthase (NOS), rather than on depolarization-dependent vesicle release. NO is generally released because NOS is activated by Ca^2+^ entry into the cell when it spikes [@pone.0017779-Garthwaite1]. However, if NOS were active even without spikes, NO release via the actions of NOS could modulate neurons without neural activity. We have examined the possible control of key neurons affecting *Aplysia* feeding by NO in the absence of spiking. *Aplysia* feeding is a complex behavior that consists of appetitive (food finding) behaviors controlled primarily by the cerebral ganglion, and subsequent consummatory behaviors controlled primarily by the buccal ganglia [@pone.0017779-Kupfermann1], [@pone.0017779-Kupfermann2]. NO is an established transmitter in both the cerebral and buccal ganglia of *Aplysia* and other gastropod molluscs [@pone.0017779-Elphick1]--[@pone.0017779-Straub1], and in these ganglia NO has been shown to be released from nitrergic neurons when they fire. In a number of other systems, in addition to being released in response to a stimulus, NO also acts as a tonic modulator of neural activity, and its tonic modulation is revealed when the actions of NO are blocked [@pone.0017779-Kobayashi1], [@pone.0017779-Bon1]--[@pone.0017779-Scholz1]. We examined the possibility that in addition to being released by stimuli signaling aspects of feeding, NO is also produced in the absence of elicited neural activity in the *Aplysia* buccal ganglia. We found that inhibition of NO actions in the buccal ganglia initiates fictive feeding in the absence of additional stimuli. Thus, as in other systems, NO is a tonic modulator of the central pattern generator (CPG) generating repetitive feeding behaviors. Access to major elements of the CPG organizing *Aplysia* consummatory feeding behaviors [@pone.0017779-Elliott1] allowed us to investigate the loci at which tonic NO production regulates feeding. In particular, we were able to examine possible effects of NO on B31/B32, key neurons having a central role in deciding to initiate consummatory feeding behaviors [@pone.0017779-Dembrow1], [@pone.0017779-Hurwitz1]. We found that B31/B32 is inhibited by NO. The ability to study *Aplysia* CPG neurons cultured in isolation [@pone.0017779-Saada1] allowed us to examine some of the cellular mechanisms by which NO acts on B31/B32. Such experiments showed that these neurons produce NO at rest, and NO contributes to their resting potential. Blockers of NO opened a depolarizing leak current, suggesting that NO acts at rest to block this current. Regulating neural activity via the background production of a neuroactive agent such as NO could potentially act as a mechanism to coordinate different aspects of a behavior that are controlled at different neural sites. For example, NO could be produced in the absence of neural activity by neurons controlling food finding in the cerebral ganglion and food consumption in the buccal ganglia. A circulating metabolite or a hormone could then affect both these sites, and thereby regulate multiple aspects of feeding behavior. This and an additional paper [@pone.0017779-Miller1] demonstrate such regulation. In this paper, we show that nitrergic self-inhibition is found in buccal ganglia neurons B31/B32, which control consummatory behaviors [@pone.0017779-Hurwitz2], [@pone.0017779-Hurwitz3], [@pone.0017779-Hurwitz4], as well as in neuron C-PR, a command neuron for a behavioral state, food arousal [@pone.0017779-Kupfermann3], [@pone.0017779-Teyke1]. Results {#s2} ======= Background NO production in the buccal ganglia inhibits feeding programs {#s2a} ------------------------------------------------------------------------ Isolated buccal ganglia in *Aplysia* contain a central pattern generator (CPG) organizing consummatory feeding behaviors. Activation of the CPG causes fictive feeding consisting of protraction and retraction phases of activity [@pone.0017779-Elliott1]. Fictive feeding can be monitored via extracellular recordings from buccal nerves [@pone.0017779-Morton1], as well as via intracellular recordings from neurons B31/B32 [@pone.0017779-Hurwitz2], [@pone.0017779-Hurwitz3]. Since NO is a modulator of central pattern generators in other systems [@pone.0017779-Kobayashi1], [@pone.0017779-Mahadevan1], [@pone.0017779-McLean1], [@pone.0017779-Scholz1], [@pone.0017779-Kyriakatos1], and since NO affects aspects of *Aplysia* feeding [@pone.0017779-Katzoff1], we examined the possibility that NO production within the buccal ganglia has a role in modulating the buccal ganglia CPG. Treating the isolated ganglia with the NO scavenger PTIO induced fictive feeding consisting of repetitive cycles of protraction and retraction, as monitored via extracellular recordings from buccal nerves ([Fig. 1](#pone-0017779-g001){ref-type="fig"}). In the extracellular recording shown, there were no bursts of fictive feeding in ASW ([Fig. 1A](#pone-0017779-g001){ref-type="fig"}). However, application of PTIO elicited fictive feeding ([Fig. 1B, D](#pone-0017779-g001){ref-type="fig"}). The buccal ganglia can produce ingestion-like or egestion-like activity [@pone.0017779-Morton1], [@pone.0017779-Morgan1], [@pone.0017779-Nargeot1]. In ingestion-like activity, the radula closes during retraction, pulling food into the mouth [@pone.0017779-Morton1]. Firing in the Radula Nerve (RN) is a monitor of radula closing, whereas firing in Buccal Nerve 2 (BN2) is a monitor of radula retraction [@pone.0017779-Morton1]. Motor programs elicited by PTIO were ingestion-like ([Fig. 1C](#pone-0017779-g001){ref-type="fig"}), as shown by the simultaneity of RN and BN2 activity. ::: {#pone-0017779-g001 .fig} 10.1371/journal.pone.0017779.g001 Figure 1 ::: {.caption} ###### The NO scavenger PTIO induces fictive feeding when applied to the isolated buccal ganglia. Fictive feeding was monitored via extracellular recordings from the radula nerve (RN) and from buccal nerve 2 (BN2). Activity in RN is a correlate of radula closing, whereas activity in BN2 is a correlate of retraction. Activity representative of radula retraction was counted as a single burst of fictive feeding. **A**) In ASW, no fictive feeding was seen, although a single unit in BN2 fired. **B**) Application of PTIO (at the start of the trace) elicited repeated bursts of fictive feeding. Recordings similar to those shown were observed in 7 separate isolated buccal ganglia preparations. **C**) Expansion of the boxed area in part B shows overlap between firing in BN2 and RN, indicating that PTIO induced ingestion-like activity. **D**) Means and standard errors of the number of fictive feeding bursts recorded in 10 min in ASW and after the application of PTIO. PTIO caused a significant increase in fictive feeding (*p* = 0.02 *t*(6) = 2.78; two-tailed paired *t*-test). :::  ::: Fictive feeding can also be monitored via intracellular recording from key CPG neurons B31/B32 [@pone.0017779-Hurwitz3]. These neurons depolarize preceding the protraction phase, and remain depolarized throughout protraction. They are repolarized during retraction. The somata of B31/B32 are electrically inexcitable [@pone.0017779-Susswein1], and 10 mV axon spikes are recorded in the soma while B31/B32 is depolarized [@pone.0017779-Hurwitz2]. B31/B32 is excited by neuron B63 via fast and slow cholinergic synapses, as well as via an electrical synapse [@pone.0017779-Saada1], [@pone.0017779-Hurwitz5]. The electrical coupling between B63 and B31/B32 allows B63 to excite B31/B32 as a result of B31/B32 depolarization, which elicits spikes in B63. Application of PTIO to the buccal ganglia caused cyclical depolarizations and repolarizations in B31/B32 ([Fig. 2A](#pone-0017779-g002){ref-type="fig"}) typical of that seen previously in response to food, or in response to activation of command-like neurons responding to food [@pone.0017779-Hurwitz3], [@pone.0017779-Hurwitz6]. ::: {#pone-0017779-g002 .fig} 10.1371/journal.pone.0017779.g002 Figure 2 ::: {.caption} ###### Blocking NO activity elicits fictive feeding as monitored via intracellular recording from neurons B31/B32. **A**) IN ASW, there were no bouts of fictive feeding. Application of the NO scavenger PTIO in 4 of 4 preparations caused fictive consummatory responses, as evidenced by cyclic depolarizations and repolarizations in a B31/B32 neuron. The recordings are a continuous record displaying 6 bouts of fictive feeding over a period of approximately 100 sec. The protraction and retraction phases of fictive feeding are marked. B31/B32 fires during protraction, and is repolarized in retraction. The B31/B32 soma is inexcitable, and axon spikes are recorded in the soma as 5--10 mV spikes, as shown. **B**) PTIO was applied to buccal ganglia in which spiking was blocked by TTX. PTIO elicited cyclical depolarizations of B31/B32 in 4 B31/B32 neurons from 2 preparations. The fast potentials at the top of a depolarization are presumably electrical EPSPs derived from Ca^+2^ spikes in the terminal of the electrically coupled B63 neuron, which fires in response to the depolarization of B31/B32. :::  ::: Because no stimuli other than the NO scavenger were used to elicit ficitive feeding, these experiments indicate that NO within the buccal ganglia in the absence of stimuli that elicit feeding is an inhibitory modulator of the buccal ganglia CPG. Reducing the background NO levels by the NO scavenger presumably initiated cyclical B31/B32 activity and fictive feeding by eliminating the inhibitory modulation. Feeding activity in the buccal ganglia elicited by food or other stimuli presumably elicit feeding against a background presence of NO that inhibits feeding. NO is produced without spiking by neurons B31/B32 {#s2b} ------------------------------------------------- B31/B32 are key components of the CPG organizing feeding [@pone.0017779-Hurwitz2], [@pone.0017779-Susswein1]. Are B31/B32, which are cyclically activated by an NO scavenger, directly affected by the scavenger, or is the effect of the scavenger via activation of neurons that synapse onto B31/B32? To answer this question, we applied PTIO to buccal ganglia that had been treated with tetrodotoxin (TTX), thereby blocking action potentials and synaptic release dependent on Na^+^-dependent spiking ([Fig. 2B](#pone-0017779-g002){ref-type="fig"}). If PTIO directly acts on B31/B32, it should depolarize the cell even in the presence of TTX. In the presence of TTX, PTIO elicited cyclical depolarizations of B31/B32 which were followed by repolarizations. The depolarizations are presumably the result of PTIO, indicating that the effect of PTIO is not dependent on spiking and transmitter release. However, PTIO did not depolarize B31/B32 as strongly as it did in the absence of TTX, indicating that TTX partially blocked the effect of PTIO. Block by TTX of the effects of PTIO need not arise as a result of an effect of NO on sodium or potassium currents underlying spiking. The lack of complete depolarization of B31/B32 is also explained by previous findings [@pone.0017779-Dembrow1], [@pone.0017779-Saada1] that the sustained depolarization of B31/B32 is not dependent on an endogenous current, but rather it is driven by synaptic output, both from neurons electrically coupled to B31/B32 and from B31/B32 onto itself. The self-excitatory depolarization of B31/B32 is blocked by TTX. In the absence of TTX, when B31/B32 is sufficiently depolarized, it begins spiking, and excites itself. In the presence of TTX the self-excitatory transmitter release is absent, and endogenous potassium currents [@pone.0017779-Hurwitz4] can repolarize B31/B32. The cyclical B31/B32 activity is likely to be explained by a direct depolarizing effect of PTIO on B31/B32, combined with repolarizing effects of endogenous K^+^ currents [@pone.0017779-Hurwitz4]. The PTIO presumably depolarized B31/B32, and thereby activated the K^+^ currents, which repolarized B31/B32. The cycle repeats when the K^+^ currents are deactivated by the repolarization. In the absence of TTX, these potassium currents slow the depolarization, and partially brake it [@pone.0017779-Hurwitz4], but do not block the transmitter-dependent depolarization. This experiment alone does not eliminate the possibility NO affects cells presynaptic to B31/B32, or that Ca^+2^ dependent spikes which are not blocked by TTX are releasing NO at rest from neurons presynaptic to B31/B32. To be certain that the effects of PTIO on B31/B32 are direct, we examined the effect of PTIO on isolated, cultured B31/B32 neurons ([Fig. 3](#pone-0017779-g003){ref-type="fig"}). In 5 of 5 preparations, PTIO depolarized and caused firing in isolated B31/B32 neurons ([Fig. 3A](#pone-0017779-g003){ref-type="fig"}). Since no other neuron was present, and no spikes were observed in the absence of stimuli, PTIO must have a direct effect on B31/B32. Since it depolarized the cell in the absence of spiking, this result cannot be explained by Ca^+2^ spikes that were not blocked by TTX. The mean latency to spiking after application of PTIO was 2.4±1.3 (SE) min. ::: {#pone-0017779-g003 .fig} 10.1371/journal.pone.0017779.g003 Figure 3 ::: {.caption} ###### Nitrergic self-inhibition in isolated B31/B32 neurons. In the recordings shown in A--D, pharmacological agents were applied at the arrow, with the resting potential preceding treatment at −60 mV. Note that recordings are minutes in length, reflecting the relatively slow effects caused by the pharmacological agents. **A**) The NO scavenger PTIO depolarized the cell in 5 of 5 preparations. Note that spikes in cultured B31/B32 neurons are larger than in neurons *in situ*. **B**) The NOS inhibitor L-NAME depolarized B31/B32 in 11 of 11 preparations. **C**) In 4 of 5 preparations, D-NAME, the inactive enantiomer of L-NAME, did not depolarize isolated B31/B32 neurons. **D**) A second NOS blocker, L-NNA, depolarized B31/B32 in 4 of 4 preparations. **E**) L-arginine has effects that are opposite to those of blocking NO. In 5 of 5 preparations previously treated with L-NAME, L-arginine (×100 of the L-NAME concentration) reversed the effect of L-NAME. In the example shown the previous treatment with L-NAME (depicted by the arrows to the left) caused a modest depolarization from −60 mV. L-arginine was applied before the cell depolarized fully. The dashed line shows −60 mV. **F**) Summary of the effects of ASW (N = 3), and of D-NAME and of 3 blockers of nitrergic transmission on the membrane potential of B31/B32. For substances causing firing of B31/B32, the potential just before the first spike is shown. Means and standard errors are shown. :::  ::: The depolarization of B31/B32 *in situ*, and the direct depolarization of B31/B32 in culture by PTIO, might have been caused by possible effects of PTIO that are not related to its inhibition of NO. To eliminate this possibility, we examined the effects of blockers of nitrergic transmission that operate via a different mechanism. We examined the effect on isolated B31/B32 neurons of L-NAME, a competitive inhibitor of NOS, which produces NO from L-arginine [@pone.0017779-Garthwaite1]. In 11 of 11 preparations, application of L-NAME depolarized B31/B32 and caused firing ([Fig. 3B](#pone-0017779-g003){ref-type="fig"}). The latency from application of L-NAME to spiking in B31/B32 was 7.65±2.07 (SE) min, somewhat longer than for PTIO, as might be expected for a substance that does not directly act on the already-released NO. By contrast, in 4 of 5 preparations, D-NAME, the enantiomer of L-NAME that does not affect NOS, had no effect on isolated B31/B32 neurons ([Fig. 3C](#pone-0017779-g003){ref-type="fig"}). Application of L-NNA, another competitive inhibitor of NOS, also depolarized and caused B31/B32 firing ([Fig. 3D](#pone-0017779-g003){ref-type="fig"}) in 4 of 4 preparations, with a mean latency to spiking of 7.0±1.9 (SE) min, comparable to that caused by L-NAME. The three blockers of nitrergic transmission depolarized B31/B32 by a mean of 24.9±8.9 (SD) mV ([Fig. 3F](#pone-0017779-g003){ref-type="fig"}) before the neuron began to fire. In these experiments the depolarization and firing were not terminated by the endogenous K^+^ currents, since firing in B31/B32 neurons autaptically excites the neurons [@pone.0017779-Saada1], overcoming the effects of the K^+^ currents. The autaptic excitation is blocked in the recording in TTX shown in [Fig. 2B](#pone-0017779-g002){ref-type="fig"}. A summary of the effects of the NO blockers, and of the effects of ASW and D-NAME controls, is shown in [Fig. 3F](#pone-0017779-g003){ref-type="fig"}. In experiments on cultured B31/B32 which fire after treatment with NO blockers, the firing was generally terminated by washing out the blocker a few seconds after the start of the firing. However, in some experiments the depolarization and firing was allowed to continue for up to 2 min. There was no repolarization of B31/B32, since in situ repolarization is dependent on synaptic input that inhibits B31/B32 [@pone.0017779-Hurwitz7], [@pone.0017779-Sasaki1]. Neurons producing this inhibition were not co-cultured with B31/B32. Since NO is synthesized by NOS from L-arginine, increases in L-arginine concentration should increase the activity of NOS, and thereby overcome the excitatory effect of L-NAME. In cultured, isolated B31/B32 neurons that had been depolarized by treatment with L-NAME, in 5 of 5 preparations subsequent treatment with L-arginine repolarized the B31/B32 neurons ([Fig. 3E](#pone-0017779-g003){ref-type="fig"}). In these experiments, the L-arginine was applied before the L-NAME had completely depolarized B31/B32 and had initiated firing. The ability of nitrergic blockers to depolarize B31/B32 neurons that were cultured in isolation was surprising, since no other cells were present that could be releasing NO. These experiments strongly suggest that B31/B32 neurons are themselves nitrergic, and produce NO at rest, in the absence of firing and Ca^+2^ entry into the cell. Production of NO by B31/B32 at rest causes self-inhibition of B31/B32. L-NAME and L-NNA blocked the tonic self-inhibitory NO production, and thereby excited the neuron. PTIO reduced the NO produced tonically, and thereby also excited the neuron. Block of guanylyl cyclase depolarizes B31/B32 {#s2c} --------------------------------------------- Effects of NO are often mediated via the activation of guanylyl cyclase and the synthesis of cyclic GMP (cGMP), which acts as a second messenger in effecting cellular changes. To strengthen the finding that B31/B32 displays self-inhibitory NO production at rest, we applied inhibitors of guanylyl cyclase to isolated, cultured B31/B32 neurons. The effects of both methylene blue [@pone.0017779-Elphick1], [@pone.0017779-Katzoff1], [@pone.0017779-Arnold1], [@pone.0017779-Lewin1] and 1H-\[1,2,4\]Oxadiazolo\[4,3-a\]quinoxalin-1-one (ODQ) were tested. Application of either methylene blue ([Fig. 4A](#pone-0017779-g004){ref-type="fig"}) or of ODQ ([Fig. 4B](#pone-0017779-g004){ref-type="fig"}) caused depolarization and firing of cultured, isolated B31/B32 neurons similar to that induced by blocking nitrergic transmission. This finding is consistent with a continuous synthesis of cGMP that contributes to the resting potential of B31/B32, since block of guanylyl cyclase in the absence of additional stimuli depolarized the cell. The continuous synthesis of cGMP is consistent with continuous production of NO. ::: {#pone-0017779-g004 .fig} 10.1371/journal.pone.0017779.g004 Figure 4 ::: {.caption} ###### Blockers of guanylyl cyclase depolarize isolated B31/B32 neurons. In the recordings shown, pharmacological agents were applied at the arrow, with the resting potential preceding treatment at −60 mV. **A**) The guanylyl cyclase blocker methylene blue depolarized B31/B32 in 5 of 5 preparations. Mean latency to spiking: 19.6±1.4 (SE) min. Mean amplitude of depolarization: 23.0±3.3 (SE) mV. **B**) The guanylyl cyclase blocker ODQ depolarized B31/B32 in 5 of 5 preparations. Mean amplitude of depolarization: 24.0±6.06 (SE); mean latency to spiking: 3.2±2.94 (SE) min. **C**) Summary of the effects of ASW (N = 3), and of the 2 blockers of guanylyl cyclase on the membrane potential of B31/B32. For the substances causing firing of B31/B32, the potential just before the first spike is shown. Means and standard errors are shown. :::  ::: Currents underlying nitrergic self-inhibition {#s2d} --------------------------------------------- What type of ion channels are present in B31/B32 that respond to NO, and that depolarize the neuron when NO is blocked? To characterize the channels underlying the response of B31/B32 to blocking NO, B31/B32 neurons were voltage-clamped in the presence and absence of either PTIO or L-NAME. As described previously [@pone.0017779-Hurwitz4], B31/B32 neurons were impaled with two electrodes, one for passing current, the other for measuring the voltage. The experiment was performed on B31/B32 neurons *in situ* in response to 3 sec command pulses ranging from −90 to −10 mV. Spikes were blocked with tetrodotoxin (TTX). The effects of PTIO and L-NAME were quantified over the last 100 msec of the pulse ([Fig. 5A](#pone-0017779-g005){ref-type="fig"}), by subtracting currents in the presence of the NO blockers from those in ASW ([Fig. 5C](#pone-0017779-g005){ref-type="fig"}). A previous study [@pone.0017779-Hurwitz4] showed that the B31/B32 somata contain two active outward currents that both have an activation threshold of approximately −40 mV, and which show time-dependent inactivation. Neither current could contribute to the steady-state inhibition of B31/B32, since the currents are not active at rest. Both currents are largely inactivated by end of a 3 sec voltage pulse, and they do not contribute to measurements of the effects of PTIO and L-NAME at the end of the 3 sec voltage pulses. ::: {#pone-0017779-g005 .fig} 10.1371/journal.pone.0017779.g005 Figure 5 ::: {.caption} ###### Block of NO opens an inward current. **A**) Effect of PTIO and of L-NAME on currents recorded during the last few hundred milliseconds of a voltage clamp experiment performed in TTX. Only currents recorded in response to voltage steps to −90, −60 and −10 mV are shown. Note that both PTIO and L-NAME induce inward currents at −90 and −60 mV, with the currents at −90 mV larger than those at −60 mV. Also note the reversal of the currents at −10 mV. **B**) Mean and standard errors (hidden by the points) of current amplitudes recorded during the last 500 msec of voltage pulses with and without PTIO or L-NAME (N = 5 for each group). **C**) The difference in current between values recorded with and without PTIO or L-NAME at the various voltage steps. The data were combined from experiments using the two blockers. Means and standard errors are shown. :::  ::: Similar results were obtained with both PTIO and L-NAME, which block nitrergic transmission via different modes of action. Treatment with either PTIO or L-NAME produced a net inward current ([Fig. 5A](#pone-0017779-g005){ref-type="fig"}). From −90 mV to −10 mV the current was not voltage-dependent ([Figs. 5C](#pone-0017779-g005){ref-type="fig"}). PTIO and L-NAME both caused a net increase in conductance of approximately 5 mS. Combined data from recordings in PTIO and L-NAME ([Fig. 5C](#pone-0017779-g005){ref-type="fig"}) showed that blocking nitrergic transmission unmasked a leak current with a −20 mV reversal potential. Nitrergic self-inhibition is not an artifact of culturing B31/B32 {#s2e} ----------------------------------------------------------------- Although nitrergic background inhibition was observed in intact buccal ganglia (see [Figs. 1](#pone-0017779-g001){ref-type="fig"}, [2](#pone-0017779-g002){ref-type="fig"}), it was important to be certain that our central finding, that background self-inhibition is also found in isolated, cultured B31/B32 neurons ([Fig. 3](#pone-0017779-g003){ref-type="fig"}), did not arise as a result of changes in the properties of the neuron when it is cultured in isolation. To rule out this possibility, we examined whether similar nitrergic self-inhibition is also seen in other isolated, cultured neurons that control aspects of feeding. Neither L-NAME ([Fig. 6A](#pone-0017779-g006){ref-type="fig"}) nor PTIO (not shown) had an effect on buccal ganglia neuron B8. L-NAME also had no effect on buccal ganglia neuron B4 (not shown), or on cerebral ganglion neuron MCC ([Fig. 6B](#pone-0017779-g006){ref-type="fig"}). These data show that not all isolated, cultured *Aplysia* buccal or cerebral ganglia neurons respond to L-NAME. ::: {#pone-0017779-g006 .fig} 10.1371/journal.pone.0017779.g006 Figure 6 ::: {.caption} ###### Effects of NO blockers and donors on other neurons. **A**) L-NAME had no effect on neuron B8 in 6 of 6 preparations. L-NAME also had no effect on B4 in 5 of 5 preparations (not shown). **B**) Treatment of an isolated MCC neuron in culture with L-NAME had no effect in 7 of 7 preparations. **C**) In an isolated, cultured C-PR neuron application of L-NAME caused a depolarization in 5 of 5 preparations. Mean amplitude of depolarization: 13.3±1.07 (SE) mV; Mean latency to spiking: 7.11±1.8 (SE) min. The dashed line marks the −60 mV resting potential. :::  ::: Cerebral ganglion arousal neuron C-PR also displays nitrergic self-inhibition {#s2f} ----------------------------------------------------------------------------- We also tested the effects of L-NAME on neuron C-PR of the cerebral ganglion. C-PR polysynaptically excites the MCC and additional neurons active during head-waving and biting [@pone.0017779-Arnold1], [@pone.0017779-Lewin1], and C-PR has been characterized as a command neuron for a behavioral state, food arousal [@pone.0017779-Teyke1]. Treatment with L-NAME caused depolarization and firing of isolated C-PR neurons ([Fig. 6C](#pone-0017779-g006){ref-type="fig"}). Since no other neuron was present, this experiment demonstrates that C-PR also contains an isoform of NOS that is active and produces NO at rest, in the absence of firing. As in B31/B32, L-NAME blocked the self-inhibition caused by NO production from C-PR onto itself at rest, and thereby excited the neuron. Depolarization of C-PR was also found when recording from the neuron in situ when PTIO was applied to the cerebral ganglion ([Fig. 7](#pone-0017779-g007){ref-type="fig"}), indicating that the excitatory effect on C-PR of blocking NO is not an artifact of culturing. In these experiments, PTIO increased excitatory synaptic outputs onto the cell ([Fig. 7A](#pone-0017779-g007){ref-type="fig"}), indicating that part of the tonic nitrergic inhibition of C-PR is indirect, via inhibition of neurons that synaptically excite C-PR. However, some of the effect of PTIO *in situ* is also direct, since the depolarization is also seen after treatment with TTX ([Fig. 7B](#pone-0017779-g007){ref-type="fig"}), which blocked firing in presynaptic neurons. ::: {#pone-0017779-g007 .fig} 10.1371/journal.pone.0017779.g007 Figure 7 ::: {.caption} ###### Block of NO depolarizes C-PR in situ. **A**) Application of the NO scavenger PTIO in an isolated cerebral ganglion preparation depolarized C-PR and caused an increase in EPSPs. **B**) In the presence of TTX, PTIO still depolarizes C-PR (N = 5 cells in 3 preparation), indicating that part of the effect is direct. :::  ::: Discussion {#s3} ========== We have presented data in support of a novel mechanism by which NO regulates the nervous system. In the absence of action potentials, NO is produced by at least two key neurons that control behavior, C-PR and B31/B32 ([Figs. 3](#pone-0017779-g003){ref-type="fig"}, [6](#pone-0017779-g006){ref-type="fig"}). In these neurons the tonic presence of NO causes self inhibition and a stabilization of the resting potential. C-PR has been characterized as a command neuron eliciting a central motive state, food arousal [@pone.0017779-Teyke1]. C-PR is excited by food on the lips [@pone.0017779-Teyke1], and its firing induces the head lifting component of head-waving [@pone.0017779-Nagahama1], a behavior by which *Aplysia* localize the position of food, before locomoting toward the food [@pone.0017779-Kupfermann1]. C-PR also polysynaptically excites the MCC, a neuron that effects aspects of food arousal [@pone.0017779-Kupfermann3], as well as exciting command-like interneurons [@pone.0017779-Teyke1], [@pone.0017779-Hurwitz8] that recruit the CPG organizing repetitive bites [@pone.0017779-Hurwitz8], [@pone.0017779-Rosen1]. Since inhibiting nitrergic transmission depolarized and caused firing of C-PR, one would predict that in intact, behaving animals inhibitors of nitrergic transmission will induce food-finding behaviors such as head-waving, as well as consummatory behaviors, such as biting. These predictions have been confirmed [@pone.0017779-Miller1]. B31/B32 are key components of the buccal ganglia CPG controlling consummatory feeding behaviors [@pone.0017779-Dembrow1], [@pone.0017779-Hurwitz2]--[@pone.0017779-Hurwitz4]. B31/B32 are active during the protraction phase of consummatory behaviors [@pone.0017779-Hurwitz3], and depolarizing or hyperpolarizing the cells respectively increases or decreases the rate of fictive consummatory responses [@pone.0017779-Hurwitz2]. The properties of the cells indicate that the decision of whether or not to initiate a consummatory behavior is made by these cells [@pone.0017779-Hurwitz4]. The initiation of buccal motor programs ([Figs. 1](#pone-0017779-g001){ref-type="fig"}, [2](#pone-0017779-g002){ref-type="fig"}) after treatment with blockers of nitrergic transmission is consistent with the effects of blocking the stabilization of the B31/B32 resting potential. Data presented elsewhere [@pone.0017779-Miller1] confirm that treating intact animals with blockers of nitrergic transmission also initiates consummatory feeding behaviors. The findings that C-PR and B31/B32 are nitrergic is consistent with previous morphological studies that characterized nitrergic neurons and processes in *Aplysia* [@pone.0017779-Moroz1], [@pone.0017779-Moroz2], [@pone.0017779-Jacklet3]. These studies showed heavy staining of nitrergic processes in the area of B31/B32 and C-PR. Background nitrergic inhibition is not restricted to B31/B32 and C-PR. NO scavengers applied to the cerebral ganglia induced increases synaptic excitation onto C-PR ([Fig. 7](#pone-0017779-g007){ref-type="fig"}), indicating that neurons presynaptic to C-PR are depolarized and fire in response to blocking nitrergic inhibition. In addition, some buccal ganglia motor neurons also contain NO [@pone.0017779-Ye1], suggesting that NO is tonically produced in these neurons. Specificity of pharmacological effects {#s3a} -------------------------------------- Our findings are largely based on the pharmacological effects of nitric oxide blockers, such as L-NAME, L-NNA and PTIO. These could have effects not related to their inhibition of nitrergic transmission. However, a number of points strongly support the likelihood that the effects seen result from changes in nitrergic transmission. First, both competitive inhibitors of NOS and an NO scavenger were used in different experiments, and sometimes in the same experiment ([Figs. 3](#pone-0017779-g003){ref-type="fig"}, [5](#pone-0017779-g005){ref-type="fig"}). Although these agents have different modes of inhibiting nitrergic transmission, their effects were consistent across experiments, or produced similar effects in the same experiment. Second, L-arginine, the precursor of NO, had an inhibitory effect opposite to the excitatory effects a NOS inhibitor ([Fig. 3E](#pone-0017779-g003){ref-type="fig"}). Third, blocking guanylyl cyclase, which in many systems is activated by NO [@pone.0017779-Garthwaite1], caused depolarization of B31/B32 similar to that caused by blocking NO. Fourth, since pharmacological agents blocking NO had similar effects on both B31/B32 and on C-PR ([Figs. 1](#pone-0017779-g001){ref-type="fig"}, [2](#pone-0017779-g002){ref-type="fig"}, [3](#pone-0017779-g003){ref-type="fig"}, [6](#pone-0017779-g006){ref-type="fig"}, [7](#pone-0017779-g007){ref-type="fig"}), it is unlikely that an effect is an anomaly arising from the unusual properties of a specific neuron, or of an artifactual effect of an agent at a specific site. Fifth, blockers and donors of NO had effects on behavior in intact animals that were consistent with the effects documented above on neural activity [@pone.0017779-Miller1]. Cellular mechanism of nitrergic self-inhibition {#s3b} ----------------------------------------------- Neuronal NO production depends on spike-dependent Ca^+2^ entry, which activates NOS [@pone.0017779-Garthwaite1]. However, NO is produced without spikes by B31/B32 and C-PR. In situ, blocking Na^+^-dependent spiking with TTX did not affect the background inhibitory effect of NO, since an NO scavenger was still effective in depolarizing B31/B32 and C-PR. In addition, an NO scavenger and an NOS blocker depolarized B31/B32 and C-PR when they were cultured in isolation and were silent, at rest potential. These neurons may have a Ca^2+^-insensitive NOS isoform (*e.g.*, [@pone.0017779-Moroz2]), there could be Ca^+2^ leak at rest, or resting Ca^2+^ concentrations may be sufficient to activate NOS, as in other tissues [@pone.0017779-Fulton1], [@pone.0017779-Garthwaite2]. Experiments blocking background NO effects in B31/B32 de-suppressed a constitutive inward leak current which depolarizes the cells when NO is blocked ([Fig. 5](#pone-0017779-g005){ref-type="fig"}). A depolarizing leak current has been cloned in mice [@pone.0017779-Lu1], and such a current is present in *Lymnaea* [@pone.0017779-Lu2]. Our results suggest that such a current may be present in *Aplysia*. The finding that blocking guanylyl cyclase also depolarizes B31/B32 suggested that NO acts via guanylyl cyclase, as it does in many other systems. If so, the inward current would be inhibited by NO via the activation of guanylyl cyclase and the production of cGMP. Although a background self-inhibitory nitrergic inhibition in the absence of spikes is to date a unique phenomenon, it has features in common with processes occurring in vertebrate retinal receptor cells. In these cells, the tonic presence of cGMP opens a channel which depolarizes the cell [@pone.0017779-Fesenko1], rather than closing such a channel and thereby preventing depolarization, as in B31/B32. Thus, the control of the membrane potential in B31/B32 and in the retina is via cGMP, but cGMP has opposite effects in the two tissues. Nitrergic self-inhibition also has features in common with nitrergic excitation of the MCC in *Aplysia* [@pone.0017779-Jacklet2], in which NO via cGMP closes an outward leak channel, thereby depolarizing the MCC, rather than closing an inward leak channel and thereby polarizing B31/B32. L-arginine is likely to act by regulating background NO {#s3c} ------------------------------------------------------- The existence of background inhibitory NO production in neurons controlling aspects of feeding raises a question of function. Why actively inhibit feeding-related neural activity via the constant production of NO at rest, in the absence of stimuli that elicit the behavior? What is there to actively inhibit if there are no stimuli eliciting feeding behavior? A trivial potential answer is that nitrergic inhibition of C-PR and B31/B32 is merely an unusual mechanism for generating the resting potential, a general property of all cells. However, the expression of this unusual cellular mechanism in key neurons organizing different aspects of feeding behavior, localized in different ganglia, suggests that it may function to allow a circulating molecule to regulate NO production, and thereby regulate different aspects of feeding behavior. Another study [@pone.0017779-Miller1] is consistent with this suggestion. This study showed that physiological increases in the hemolymph concentration of the amino acid L-arginine can inhibit feeding. The regulation of feeding by L-arginine provides a plausible function for the presence of nitrergic self-inhibition in neurons C-PR and B31/B32. L-arginine is the precursor from which NO is synthesized. L-arginine is generally present in concentrations that are sufficient for the production of NO when NOS is activated. In most nitrergic neurons, NOS will be transiently active in response to Ca^+2^ entry resulting from spiking. Because NOS is transiently active, a tonic increase in L-arginine concentration could transiently lead to a brief increase in NO concentration, which extends the active radius of NO release, as well as more strongly affecting cells already within the active radius. However, in C-PR and in B31/B32, in which NOS is likely to be continuously active, the increased L-arginine will cause a continuous increase in NO synthesis, potentially allowing even a small increase in L-arginine to have a continuous effect. In addition, since the neurons producing NO are themselves sensitive to NO, the maximal effect of the increased NO synthesis will be at the site of the NO synthesis. Thus, building neurons with continuously active, self-affecting NO production is a highly effective way to regulate their activity by changes in L-arginine. Although it is not mediated by NO, background synaptic inhibition is also a feature of the CPG controlling *Lymnaea* feeding [@pone.0017779-Staras1], indicating that inhibitory modulation of a feeding CPG may be a general control feature. In this system, an inhibitory neuron that is part of the CPG tonically fires at rest, and thereby inhibits the CPG. Just as the tonic NO is release is overcome by food-related stimuli in *Aplysia*, the tonic firing in *Lymnaea* is suppressed when feeding is initiated. Guanylyl cyclase as a modulator of feeding strategies {#s3d} ----------------------------------------------------- Inhibitors of guanylyl cyclase depolarize isolated B31/B32 neurons ([Fig. 4](#pone-0017779-g004){ref-type="fig"}), suggesting that nitrergic self-inhibition, and the regulation of feeding via L-arginine [@pone.0017779-Miller1], is via the cGMP second messenger pathway. cGMP is a modulator of feeding in other systems [@pone.0017779-Kaun1]. Variations in genes governing PKG, the kinase responding to cGMP, affects feeding strategies in *Drosophila* [@pone.0017779-Douglas1], and in other insects [@pone.0017779-BenShahar1]--[@pone.0017779-Lucas2], as well as in *C. elegans* [@pone.0017779-Bretscher1] and possibly in vertebrates [@pone.0017779-Kaun1]. Control of feeding by cGMP in different phyla suggests that such control may have been present in early multi-cellular animals, and is preserved with variations in descendents in different phyla. Elicited NO release also has predominantly inhibitory effects on feeding {#s3e} ------------------------------------------------------------------------ In addition to acting as a background factor inhibiting key neurons driving feeding, NO is released in *Aplysia* in response to a number of stimulus conditions. Conditions causing an increase in NO also generally inhibit feeding, suggesting that background and elicited effects of NO are consistent. For example, NO is a transmitter of the L29 neurons that facilitate withdrawal reflexes [@pone.0017779-Antonov1]. NO is also elicited by stimuli causing tissue damage [@pone.0017779-Lewin1]. Noxious stimuli initiating withdrawal or causing tissue damage also inhibit feeding [@pone.0017779-Kupfermann4]. Treating *Aplysia* with an NO donor induces egg-laying [@pone.0017779-Miller2], which also inhibits feeding [@pone.0017779-Stuart1]. NO is released by neuron C2 [@pone.0017779-Jacklet1], which responds to food-related stimuli, particularly to stimuli causing attempts to consume a tough food [@pone.0017779-Chiel1]. An adaptive response to tough food is initially to try hard to consume it, but if such attempts fail, animals then reject the food and eventually to stop feeding. NO release may partially underlie the early facilitation of feeding, as animals try to consume the tough food, as well as the later inhibition of feeding. The early facilitation may be via the connection from cerebral ganglion neuron C2 to the MCC. C2 [@pone.0017779-Weiss1] is a sensory neuron that fires in response to attempts to swallow [@pone.0017779-Chiel1]. It excites the MCC, which effects aspects of food arousal [@pone.0017779-Kupfermann3], by releasing both NO and histamine [@pone.0017779-Jacklet1], [@pone.0017779-Weiss1], [@pone.0017779-McCaman1]. NO also functions in biasing feeding responses toward rejection. Thus, treatment with L-NAME causes rejection responses to become more irregular [@pone.0017779-Katzoff1]. In addition, when animals are fed a tough inedible food, treatment with L-NAME reduces rejections and causes increased attempts to swallow the food [@pone.0017779-Katzoff2]. In learning that food is inedible, a learning paradigm affecting *Aplysia* feeding, NO substitutes for efforts to swallow [@pone.0017779-Katzoff1], [@pone.0017779-Katzoff3], a necessary component for memory formation [@pone.0017779-Katzoff2], suggesting that NO functions in signaling such attempts. Learning is expressed as increased rejection responses and an eventual cessation of attempts to feed [@pone.0017779-Schwarz1], which are consistent with the inhibitory effects of NO on feeding. Inhibition of feeding caused by NO released as a result of efforts to swallow could act at the same sites responding to the background tonic inhibitory NO production. Efforts to swallow will release much more NO, and will produce a much stronger feeding inhibition, than the feeding inhibition caused by background NO production. In *Lymanea* blocking NO blocks responses to food, presumably because taste afferents are nitrergic [@pone.0017779-Elphick1]. *Lymnaea* feeding is characterized by rasps [@pone.0017779-Elliott1], rather than by separate ingestion and rejection feeding responses. Rasps in *Lymnaea* may be excited by factors such as NO that facilitate rejection in *Aplysia*. In addition, in *Lymnaea* increased NO levels inhibit fictive feeding [@pone.0017779-Kobayashi1], [@pone.0017779-Kobayashi2]. Thus, NO may have mixed effects on *Lymnaea* feeding. Background nitrergic modulation in other systems {#s3f} ------------------------------------------------ Background nitrergic modulation of neurons is also found in other systems. However, two aspects of nitrergic inhibition of *Aplysia* feeding are to date unique: 1) NO is produced in neurons in the absence of spiking; 2) NO causes self-inhibition by blocking an inward leak current. Inhibiting NO destabilizes the crab stomatogastric ganglion [@pone.0017779-Scholz1], which contains two networks. After blocking either NO or guanylyl cyclase the networks combine into a single circuit. Many neurons in the ganglion are nitrergic, suggesting that NO released from the CPG modulates it. In the crustacean neurogenic heart, NO released from the heart exerts inhibitory control on the cardiac ganglion CPG [@pone.0017779-Mahadevan1]. NO also inhibits the locomotor pattern generator in tadpoles [@pone.0017779-McLean1], [@pone.0017779-McLean2], [@pone.0017779-Kyriakatos1], where NO increases the effects of inhibitory interneurons, and depolarizes motor neurons by closing a K^+^ channel, which increases their input resistance. Background NO release affecting the nervous system without neural activity is found in the hippocampus, where both elicited and background NO release facilitate long-term potentiation (LTP) [@pone.0017779-Bon1], [@pone.0017779-Schwarz1]. In this system, background and elicited NO are released from different tissues by different mechanisms. Background NO is released from endothelia lining blood vessels that are in proximity to neural targets, whereas elicited NO release occurs as a result of action potentials in neurons [@pone.0017779-Garthwaite2]. Release of NO from capillaries onto neurons also occurs in the optic nerves of mammals [@pone.0017779-Hopper1]. Materials and Methods {#s4} ===================== Animals {#s4a} ------- *Aplysia californica* (5--200 g) were purchased from Marinus Scientific (Garden Grove, CA), Santa Barbara Marine Bio (Santa Barbara, CA) and from the NIH/University of Miami National Resource for *Aplysia*. Animals were maintained on a 12 hours light-dark cycle in 900 liter tanks of aerated, filtered Mediterranean seawater at 18°C. They were fed every 3--4 days with *Ulva lactuca* gathered fresh from the Mediterranean Sea and then kept frozen until needed. Pharmacology {#s4b} ------------ Concentrations used were as follows: for the nitric oxide synthase (NOS) inhibitor Nω-nitro-L-arginine methyl ester (L-NAME), 0.37 mM; for a second competitive NOS inhibitor L-N^G^-nitroarginine (L-NNA), 1 mM; for the enantiomer of L-NAME, Nω-nitro-D-arginine methyl ester (D-NAME), 0.37 mM; for the NO scavenger 2-phenyl-4,4,5,5-tetramethyl-imidazdine-1-oxy-3-oxide (PTIO), 1 mM; for tetrodotoxin (TTX), 60 µM; for the guanylyl cyclase blocker methylene blue, 100 µM; for a second guanylyl cyclase blocker 1H-\[1,2,4\]Oxadiazolo\[4,3-a\]quinoxalin-1-one (ODQ), 20 mM (ODQ was dissolved in DMSO before placing it ASW). The substances were added to solutions of artificial seawater (ASW) whose composition was (in mM): NaCl, 460; KCl, 10; CaCl~2~, 11; MgCl~2~, 55; and NaHCO~3~, 5 mM; pH = 7.64. The concentrations used for L-NAME, D-NAME, L-NNA, and PTIO were chosen because previous studies [@pone.0017779-Katzoff1], [@pone.0017779-Moroz2], [@pone.0017779-Lewin1], [@pone.0017779-Miller2] showed the efficacy of these concentrations. Chemicals were purchased from Sigma, Israel. In isolated cultured neurons, nitrergic transmission was blocked by either an NO scavenger or by competitive NOS inhibitors. In acutely dissected preparations, only an NO scavenger was consistently used. Acute preparations require extensive dissection causing tissue damage, which releases NO [@pone.0017779-Lewin1]. In this condition, preliminary experiments showed that a scavenger is more effective than is a competitive inhibitor of NOS. Acute extracellular recording {#s4c} ----------------------------- Animals were anesthetized with isotonic MgCl~2~ (25--50% of the body weight) prior to dissection. The buccal ganglia were removed and placed in a chamber containing artificial seawater (ASW). Fictive feeding was recorded via suction electrodes that were placed on the cut end of the radula nerve (RN) and buccal nerve 2 (BN2). RN and BN2 were chosen because recordings from these nerves are useful monitors of ingestion versus egestion-like patterns of fictive feeding [@pone.0017779-Morton1]. *In situ* current and voltage clamping {#s4d} -------------------------------------- After the animals were anesthetized, either the cerebral ganglion or the buccal ganglia were removed and placed in a chamber containing artificial seawater (ASW). The connective tissue sheath overlying neurons was surgically removed. Recordings were at room temperature with 1M KCl electrodes (10--20 MΩ on neurons *in situ*; 40--70 MΩ on neurons in cell culture), via an Axoclamp 2 voltage clamp/amplifier (Axon Instruments) used in either current clamp mode or in two-electrode voltage clamp mode, as appropriate to the design of the experiment. Cell culture {#s4e} ------------ Cerebral or buccal ganglia were dissected, and then bathed in protease (Sigma, Israel) for 2 hours. Ganglia were desheathed, the cells of interest were identified via intracellular recording, and then removed from the ganglion and cultured for 3--4 days at 18° in hemolymph and L15 (Sigma) with salts added to adjust the salinity to that of seawater, as described previously [@pone.0017779-Rayport1], [@pone.0017779-Schacher1]. Recordings were in 50% ASW- 50% L-15. We thank Drs. John H. Byrne and Hillel Chiel for comments on earlier versions of the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The research was supported by grant 420/06 from the Israel Science Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: NM RS IH AJS. Performed the experiments: NM RS SF IH. Analyzed the data: NM RS SF AJS. Wrote the paper: AJS.

PubMed Central

2024-06-05T04:04:19.265609

2011-3-9

PMC3052383

Introduction {#s1} ============ All neurons and glial cells in the brain are derived from neural stem cells (NSCs). NSCs maintain their own numbers by self-renewal and also give rise to daughter cells that terminally differentiate into neurons, astrocytes, and oligodendrocytes [@pone.0017736-Gage1], [@pone.0017736-Temple1]. NSCs have been found to persist in the adult brain and generate new neurons throughout adult life, particularly in the subgranular zone (SGZ) of the dentate gyrus and the subventricular zone (SVZ) of the lateral ventricles [@pone.0017736-Zhao1]. This raises the exciting possibility that NSCs may be useful for the therapy of neurodegenerative diseases. The factors that control the division and differentiation of NSCs are of tremendous scientific and medical importance. Geminin (*Gmnn*) is an unstable regulatory protein that is thought to maintain neural progenitor cells in an undifferentiated state while they proliferate [@pone.0017736-Aigner1]. Geminin is expressed in both embryonic and adult mouse neural progenitor cells, and in the Xenopus central nervous system throughout embryonic development [@pone.0017736-Kroll1], [@pone.0017736-Spella1]. Geminin is preferentially expressed in neural precursor cells, and expression is down-regulated before neural differentiation [@pone.0017736-Seo1] [@pone.0017736-Spella1]. Geminin binds to Brg1, the catalytic (ATPase) subunit of a SWI/SNF chromatin remodeling complex, and inhibits its recruitment to neuron-specific promoters by the basic helix-loop-helix (bHLH) transcription factors Neurogenin (Ngn) and NeuroD [@pone.0017736-Seo1]. A complex between Geminin and the transcription factor AP4 represses the transcription of neuronal genes in non-neuronal cell types [@pone.0017736-Kim1]. In addition to its effects on the nervous system, Geminin inhibits tissue differentiation in a variety of other organs by binding and inhibiting various transcription factors and chromatin remodeling proteins, including *sine oculis (Six)* and *Homeobox* (*Hox*) transcription factors and the *Polycomb* protein Scmh1 [@pone.0017736-Seo1], [@pone.0017736-DelBene1], [@pone.0017736-Luo1]. In addition to regulating cell differentiation, Geminin also limits the extent of DNA replication to one round per S phase by binding and inhibiting the essential replication factor Cdt1 [@pone.0017736-Machida1]. The concentration of Geminin is cell-cycle regulated; the protein begins to accumulate at the G1/S transition and persists throughout S and G2 phase. Geminin is destroyed by ubiquitin-dependent proteolysis during M phase, which allows a new round of replication in the next cell cycle [@pone.0017736-McGarry1]. This expression pattern has been documented extensively in developing mouse brains [@pone.0017736-Spella1]. *Six* and *Hox* transcription factors can compete with Cdt1 for binding to Geminin [@pone.0017736-DelBene1], [@pone.0017736-Luo1], raising the possibility that Geminin links exit from the cell cycle with cell differentiation. According to this model, the destruction of Geminin when cells enter G0 phase would relieve the repression of Brg1 and other transcription proteins and trigger terminal differentiation [@pone.0017736-Aigner1], [@pone.0017736-Luo2], [@pone.0017736-Seo2]. In early embryos Geminin can also act as an inducer of nervous tissue. In an unbiased expression-cloning screen, Geminin was identified as a molecule that expands the size of neural plate in Xenopus embryos [@pone.0017736-Kroll1]. These effects are correlated with increased expression of the proneural gene Neurogenin-related 1 (Ngr1) and decreased expression of BMP4, an epidermis-inducing growth factor. Over-expression of Geminin in Drosophila embryos induces ectopic neural cells in the epidermis [@pone.0017736-Quinn1]. The role of Geminin in regulating neural development has been examined by deleting its gene from model organisms. C. elegans embryos treated with Geminin siRNA show gonadal abnormalities and ∼20% of the worms are infertile, but no neural phenotype has been described [@pone.0017736-Yanagi1]. *Geminin^Δ/Δ^* Drosophila embryos die at larval stages with mostly normal neuroanatomy, although a small percentage of them have sharply reduced numbers of peripheral neurons [@pone.0017736-Quinn1]. Geminin-deficient Xenopus and mouse embryos do not develop past the blastula stage because of defects in DNA replication. Geminin-depleted Xenopus embryos arrest cell division in G2 phase at the mid-blastula stage because over-replication activates the DNA replication checkpoint [@pone.0017736-Kerns1], [@pone.0017736-McGarry2]. *Gmnn^Δ/Δ^* mouse embryos arrest development at about the 8-cell stage, as soon as the maternal supply of Geminin is exhausted [@pone.0017736-Gonzalez1], [@pone.0017736-Hara1]. Their cells contain more nuclear DNA than normal, consistent with over-replication of the DNA. Interestingly, the *Gmnn^Δ/Δ^* cells prematurely differentiate as trophoblast cells and none express markers of the embryonic stem cells that form the embryo proper. *Gmnn(−)* Xenopus and mouse embryos arrest development long before neural induction takes place, which has precluded examining the role of Geminin in vertebrate neural development using a rigorous genetic system. To address this question, we constructed a strain of mice in which Geminin was specifically deleted from neural stem cells. To our surprise, we found that neural-specific *Geminin^Δ/Δ^* mice displayed no obvious neurological defects and had apparently normal neurogenesis. We conclude that Geminin is dispensable for normal neurogenesis during most of embryogenesis and in adulthood. Results {#s2} ======= The mouse genome contains a single copy of the Geminin gene, which is composed of seven exons. Exons 5, 6, and 7 encode Geminin\'s dimerization domain and the domains that bind Cdt1 and Brg1 ([Figure S1](#pone.0017736.s001){ref-type="supplementary-material"}). Because these domains are essential for Geminin\'s biological activity [@pone.0017736-Benjamin1], deletion of these exons is predicted to produce a *Geminin^null^* allele. We flanked exons 5, 6, and 7 with loxP sites to create a floxed Geminin allele (*Gmnn^fl^*) and established a line of *Gmnn^fl/fl^* mice, which are completely viable and fertile [@pone.0017736-Shinnick1]. To delete Geminin specifically from nerve cells, *Gmnn^fl/fl^* mice were crossed to *Nestin-Cre* mice. Nestin is a neurofilament protein that is expressed in neural precursor cells and NSCs [@pone.0017736-Lendahl1]. *Nestin-Cre* mediated recombination begins around embryonic day 7.5 (e7.5), the time when the neural plate first forms, and continues throughout adulthood. Recombination is virtually complete in all neurons and glial cells by e15 [@pone.0017736-Dubois1], [@pone.0017736-Yu1]. *Nestin-Cre/Gmnn^fl/fl^* mice were born in the expected Mendelian ratio ([Table 1](#pone-0017736-t001){ref-type="table"}) and were indistinguishable from their control littermates in terms of size, activity, and longevity ([Figure 1B](#pone-0017736-g001){ref-type="fig"}). They moved all four extremities and exhibited normal locomotor activity. Their feeding and avoidance behavior were both normal, and they responded normally to noxious stimuli. Both males and females were able to mate and rear pups until the time of weaning, although the average litter size was slightly reduced when either parent had the genotype *Nestin-Cre/Gmnn^fl/fl^* or if the male had the genotype *Nestin-Cre/Gmnn^fl/+^* ([Table 2](#pone-0017736-t002){ref-type="table"}). The brains of *Nestin-Cre/Gmnn^fl/fl^* mice were grossly normal in appearance and had no obvious neuroanatomical defects upon sectioning ([Figure 1A and 1C](#pone-0017736-g001){ref-type="fig"}). ::: {#pone-0017736-g001 .fig} 10.1371/journal.pone.0017736.g001 Figure 1 ::: {.caption} ###### Nes-Cre/Gmnn^fl/fl^ Mice have Normal Neuroanatomy and Normal Survival. \(A) Whole-mount brains of *Gmnn^fl/fl^* (left) and *Nes-Cre/Gmnn^fl/fl^* (right) mice. (B) Survival of *Gmnn^fl/fl^* (black) and *Nes-Cre/Gmnn^fl/fl^* (red) mice. Numbers above the X axis indicate the number of mice that were at risk at each timepoint (black, control animals; red, *Nes-Cre/Gmnn^fl/fl^* animals). Controls include both *Gmnn^fl/fl^* and *Gmnn^fl/+^* animals. (C) Representative Nissl-stained sections of the cortex and the cerebellum of *Gmnn^fl/fl^* and *Nes-Cre/Gmnn^fl/fl^* mice. Scale bar, 2 mm. :::  ::: ::: {#pone-0017736-t001 .table-wrap} 10.1371/journal.pone.0017736.t001 Table 1 ::: {.caption} ###### Viability of Nestin-Cre/Gem^fl/fl^ Mice. ::: {#pone-0017736-t001-1} Offspring ----------------------------------- ----------- ----- ---- ---- ------ *Nes-Cre/Gem^+/fl^ X Gem^fl/fl^* 22 21 27 28 0.82 *Nes-Cre/Gem^fl/fl^ X Gem^fl/fl^* \-- \-- 42 33 0.75 ::: ::: {#pone-0017736-t002 .table-wrap} 10.1371/journal.pone.0017736.t002 Table 2 ::: {.caption} ###### Fertility of Nestin-Cre/Gmnn^fl/fl^ Mice. ::: {#pone-0017736-t002-2} Genotype Sex Litter size (mean ± SD) \# of Litters \# of Animals P value ----------------------- -------- ------------------------- --------------- --------------- --------- Control M or F 7.8±2.0 19 6 \-- *Nes-Cre/Gnmm(fl/+)* F 7.4±1.3 5 3 0.700 *Nes-Cre/Gnmm(fl/+)* M 5.3±2.0 9 4 0.007 *Nes-Cre/Gnmm(fl/fl)* F 5.2±2.9 15 4 0.002 *Nes-Cre/Gnmm(fl/fl)* M 5.3±2.6 19 3 0.006 ::: To confirm that the Geminin gene had been deleted, we isolated protein and RNA samples from the brains of e14.5 animals. Geminin protein was undetectable in the brains of *Nestin-Cre/Gmnn^fl/fl^* mice compared to the brains of littermate controls ([Figure 2A](#pone-0017736-g002){ref-type="fig"}). The brains of e14.5 and e16.5 *Nestin-Cre/Gmnn^fl/fl^* mice had only ∼20% of the level of intact Geminin mRNA found in littermate controls as judged by quantitative RT-PCR ([Figure 2C](#pone-0017736-g002){ref-type="fig"}). We hypothesized that the residual amount of Geminin message may have come from Geminin expression in vascular or other non-neural cell types. To see if this was the case, we cultured neurospheres from the brains of newborn mice and measured Geminin mRNA levels in these purified NSC populations. The level of intact Geminin RNA in *Nestin-Cre/Gem^fl/fl^* neurospheres was only ∼0.2% of the concentration in control neurospheres, which indicates virtually complete deletion of the Geminin gene ([Figure 2C](#pone-0017736-g002){ref-type="fig"}, bottom panel). As an additional control we crossed our mice to *R26R-LacZ* mice, in which a loxP-flanked transcription/translation STOP cassette is inserted between the ROSA promoter and the β-galactosidase gene [@pone.0017736-Soriano1]. *Cre*-mediated recombination removes the STOP cassette and causes β-galactosidase expression in cells that express *Cre* and their descendants. The brains of adult (6--8 week old) *Nestin-Cre/Gem^fl/fl^/R26R-LacZ* mice showed β-galactosidase expression in virtually all neurons while the brains of littermate *Gmnn^fl/fl^/R26R-LacZ* mice showed no β-galactosidase expression ([Figure 2B](#pone-0017736-g002){ref-type="fig"}). Taken together, our results indicate that deletion of the Geminin gene is mostly complete by e14.5 and virtually complete by birth, consistent with what has been previously reported for other mice that express *Nestin-Cre* [@pone.0017736-Dubois1], [@pone.0017736-Yu1]. The persistence of β-galactosidase-expressing neurons in the brains of adult *Nestin-Cre/Gmnn^fl/fl^/R26R-LacZ* mice excludes the possibility that neurogenesis is rescued by rare cells that escape Cre-mediated Geminin deletion. ::: {#pone-0017736-g002 .fig} 10.1371/journal.pone.0017736.g002 Figure 2 ::: {.caption} ###### Geminin is Deleted from the Brains of Nes-Cre/Gmnn^fl/fl^ Mice. \(A) Immunoblot showing the amount of Geminin protein in e14.5 mouse brain homogenates. (B) Xgal staining of coronal brain sections showing β-galactosidase activity in the brains of *Nes-Cre/Gmnn^fl/fl^/R26R-LacZ* and *Gmnn^fl/fl^/R26R-LacZ* mice. Scale bar, 0.25 mm. (C) Quantitative real-time PCR showing the amount of Geminin mRNA in the brains of e14.5 and e16.5 mice and in neurospheres cultured from newborn (P0) mice. :::  ::: Next we examined the process of neurogenesis in more detail to see if we could detect quantitative defects in neuron formation. First we compared the number of NSCs present in the SGZ of the dentate gyrus of 2-month old *Nestin-Cre/Gmnn^fl/fl^* and control mice. The mice were injected with bromodeoxyuridine (BrdU) and 24 hours later the brains were fixed, sectioned, and stained with anti-BrdU antibodies. There was no significant difference in the number of BrdU^+^ NSCs in the dentate gyri of *Nestin-Cre/Gmnn^fl/fl^* and control mice ([Figure 3A and 3B](#pone-0017736-g003){ref-type="fig"}). We also performed a pulse-chase experiment to monitor the differentiation of NSCs into neurons. Adult 7 week-old mice were injected daily with BrdU for 12 days, then 9 weeks later their brains were fixed, sectioned, and stained with antibodies against BrdU and the neuronal marker NeuN ([Figure 3C](#pone-0017736-g003){ref-type="fig"}). We counted the number of newly born neurons (NeuN^+^BrdU^+^ cells) in the dentate gyrus using confocal microscopy. We found no difference in the number of BrdU^+^NeuN^+^ cells in the dentate gyri of *Nestin-Cre/Gmnn^fl/fl^* or control mice. These results indicate that in *Nestin-Cre/Gem^fl/fl^* brains the number of NSCs is normal and their rate of differentiation into neurons is normal. ::: {#pone-0017736-g003 .fig} 10.1371/journal.pone.0017736.g003 Figure 3 ::: {.caption} ###### Normal Hippocampal Neurogenesis in Nes-Cre/Gmnn^fl/fl^ Mice. \(A) Adult mice were injected with BrdU 24 hours before sacrifice. Adjacent sections of the hippocampus were stained with either anti-BrdU antibodies or with DAPI. Scale bar, 50 µM. (B) Quantification of the number of BrdU(+) cells in the subgranular zone of the hippocampus in *Nes-Cre/Gmnn^fl/fl^* mice and control littermates. n = 4 for controls; n = 3 for *Nes-Cre/Gmnn^fl/fl^* mice. (C) 7 week-old mice were injected with BrdU for 12 days then sacrificed 9 weeks later. Sections of the hippocampus were stained with anti-BrdU antibodies and with anti-NeuN antibodies. The Z-stack image on the right shows BrdU^+^, NeuN^+^, and BrdU^+^NeuN^+^ cells. Scale bar, 50 µM. (D)Quantification of the number of BrdU(+)NeuN(+) cells in the hippocampus of *Nes-Cre/Gmnn^fl/fl^* mice and control littermates. n = 2 for both genotypes. :::  ::: To test whether Geminin regulates the differentiation of NSCs in vitro, we used primary neurosphere cultures from neonatal *Nestin-Cre/Gmnn^fl/fl^* and control brains. Neurosphere cultures could easily be established from both genotypes. To induce differentiation, neurospheres were cultured in a low concentration of growth factors then stained with specific antibodies to identify differentiated neurons (TuJ1^+^ cells) or astrocytes (GFAP^+^ cells). We found that *Nestin-Cre/Gmnn^fl/fl^* and control neurospheres produced equal numbers of GFAP^+^ and TuJ1^+^ cells ([Figure 4](#pone-0017736-g004){ref-type="fig"}). ::: {#pone-0017736-g004 .fig} 10.1371/journal.pone.0017736.g004 Figure 4 ::: {.caption} ###### Normal Differentiation of *Nes-Cre/Gmnn^fl/fl^* Neurosphere Cells. (A)Representative images of TuJ1(+) neurons and GFAP(+) astrocytes derived from *Nes-Cre/Gmnn^fl/fl^* and control neurosphere cells. Scale bar, 50 µM. (B)Quantification of the number of TuJ1(+) neurons and GFAP(+) astrocytes derived from *Nes-Cre/Gmnn^fl/fl^* and control neurosphere cells. HPF, High Powered Field. :::  ::: Finally, we examined the effect of Geminin deletion on the cell cycle in neurospheres. Both *Nestin-Cre/Gmnn^fl/fl^* and control neurospheres cells grew at identical rates ([Figure 5A](#pone-0017736-g005){ref-type="fig"}), and the distribution of cells in different phases of the cell cycle was indistinguishable between the two types of cells ([Figure 5B](#pone-0017736-g005){ref-type="fig"}). In particular, the percentage of *Nestin-Cre/Gmnn^fl/fl^* cells that had excessive DNA contents \>4n was the same as in *Gmnn^fl/fl^* littermate controls, and both genotypes had the same proportion of cells with a G2/M DNA content ([Figure 5C](#pone-0017736-g005){ref-type="fig"}). These results indicate that *Gmnn^Δ/Δ^* neurosphere cells do not over-replicate their DNA and display no obvious cell cycle defects. ::: {#pone-0017736-g005 .fig} 10.1371/journal.pone.0017736.g005 Figure 5 ::: {.caption} ###### *Nes-Cre/Gmnn^fl/fl^* Neurosphere Cells Show Normal DNA Replication and Normal Cell Cycle Kinetics. (A)Doubling time, (B) cell cycle phase distribution, and (C) DNA content of neurosphere cells derived from *Nes-Cre/Gmnn^fl/fl^* mice and control littermates. :::  ::: Discussion {#s3} ========== Geminin is widely expressed in the developing brain and is thought to have two functions in controlling the development of the central nervous system. Early in development, Geminin acts as a neural inducer: over-expression of the protein causes expansion of the neural plate at the expense of the epidermis [@pone.0017736-Kroll1]. Later in development, Geminin inhibits neuronal differentiation by inhibiting the recruitment of a SWI/SNF chromatin-remodeling complex to neuron-specific promoters [@pone.0017736-Seo1]. Geminin also limits the extent of DNA replication during the cell cycle by binding and inhibiting the essential replication factor Cdt1 [@pone.0017736-Arias1]. Because of its effects on both the cell cycle and on gene expression, it has been proposed that Geminin maintains neural stem cells in an undifferentiated state while they proliferate [@pone.0017736-Aigner1], [@pone.0017736-Luo2], [@pone.0017736-Seo2]. In this study we rigorously examined the role of Geminin in vertebrate neurogenesis by constructing a strain of *Nestin-Cre/Gmnn^fl/fl^* mice in which Geminin was deleted from neural stem cells early in embryonic development. To our surprise, we could detect no defects in neurogenesis in *Geminin^Δ/Δ^* brains. *Nestin-Cre/Gmnn^fl/fl^* mice had normal viability and displayed no obvious neurological or neuroanatomical abnormalities. Quantitative assays of neural stem cell division and differentiation patterns also revealed no abnormalities. Previous studies have shown that mice with defects in adult neurogenesis have well defined and unmistakable phenotypes. For example, mice that carry a NSC-specific deletion of the chromatin remodeling factor *Mll1* develop growth retardation and ataxia during the second week of life and die around 4 weeks of age [@pone.0017736-Lim1]. Mice with a postnatal deletion of both *Numb* and *Numb-like* show defects in lateral ventricle integrity and SVZ neuroblast survival [@pone.0017736-Kuo1]. We conclude that Geminin is largely dispensable for adult neurogenesis and for embryonic neurogenesis after e14.5. Our results are consistent with what has been reported for *Geminin^Δ/Δ^* Drosophila larvae, most of which have normal neural architecture [@pone.0017736-Quinn1]. It is unlikely that the mouse genome encodes a protein that can substitute for Geminin, which is a single copy gene. The 41 kD GEMC1 protein ([GEM]{.underline}inin [C]{.underline}oiled-coil domain containing protein [1]{.underline}) includes a 49 amino-acid sequence that is ∼30% identical to Geminin\'s coiled coil, but GEMC1 serves a different function than Geminin and is actually required for DNA replication [@pone.0017736-Balestrini1]. Mice carrying a homozygous deletion of the Geminin gene die at the early blastula stage, as soon as the maternal supply of Geminin is exhausted [@pone.0017736-Gonzalez1], [@pone.0017736-Hara1]. We have also observed strong phenotypes when Geminin is deleted from other types of cells using other types of *Cre* drivers. For example, deleting Geminin profoundly affects the differentiation pattern of bone marrow cells; the production of red blood cells is virtually abolished while the production of megakaryocytes is greatly expanded [@pone.0017736-Shinnick1]. Geminin\'s role in controlling the extent of DNA replication has been well documented by many different laboratories using many different experimental systems [@pone.0017736-Machida1], [@pone.0017736-Arias1]. It is therefore somewhat surprising that DNA replication and cell division appear completely normal in *Gmnn^Δ/Δ^* neurosphere cells. These cells probably cycle normally in the absence of Geminin because of redundant Geminin-independent mechanisms that limit the extent of DNA replication. For example, in addition to being inhibited by Geminin, Cdt1 is also destroyed by ubiquitin-dependent proteolysis during S phase, and the degradation is coupled to the initiation of DNA replication[@pone.0017736-Arias2]. In order to induce over-replication in Xenopus egg extracts, it is necessary to both remove Geminin and inhibit the proteolysis of Cdt1 [@pone.0017736-Kerns1], [@pone.0017736-Arias2], [@pone.0017736-Li1]. The requirement for Geminin to prevent re-replication seems to vary depending upon the type of cell and upon conditions. For example, *Gmnn^Δ/Δ^* white blood cells over-replicate their DNA when stimulated with growth factors [@pone.0017736-Shinnick1], and *Gmnn^Δ/Δ^* T lymphocytes show cell cycle defects when stimulated to proliferate in vitro [@pone.0017736-Karamitros1]. Some of the previous studies that implicated Geminin in neurogenesis were based on over-expressing the protein, which may cause non-physiological effects [@pone.0017736-Kroll1], [@pone.0017736-Quinn1]. Others employed cultured cell lines, which may not reproduce all the characteristics of in vivo NSCs [@pone.0017736-Seo1]. Some of the previously observed effects on neurogenesis may have been caused by cell cycle abnormalities rather than a specific effect of Geminin on neural differentiation. It remains possible that Geminin is required for very early neural induction. The neural plate becomes visible as distinct anatomical structure at e7.5 and the neural tube becomes completely closed by e10 [@pone.0017736-Theiler1]. Both these events occur several days before *Nestin-Cre* mediated recombination is complete. Even so, *Nestin-Cre*-mediated recombination has been detected as early as e7.5, and extensive recombination has occurred in the midbrain and hindbrain by e9.5 [@pone.0017736-Dubois1], [@pone.0017736-Yu1]. Our targeting construct did not delete the sequences encoding the small "neuralizing domain" found near Geminin\'s amino terminus. Expression of this fragment in Xenopus embryos induces neural tissue as efficiently as full-length Geminin [@pone.0017736-Kroll1]. Although we could detect transcripts encoding this domain in *Nestin-Cre/Gmnn^Δ/Δ^* neurospheres, we could not detect a new fusion protein in *Gmnn^Δ/Δ^ brains* ([Figs. S2](#pone.0017736.s002){ref-type="supplementary-material"} and [S3](#pone.0017736.s003){ref-type="supplementary-material"}). Geminin deficiency might cause defective neurogenesis in strains with a different genetic background. It is also possible that *Nestin-Cre/Gem^fl/fl^* mice will prove to have a subtle neurological or neuroanatomical defect than could not be detected by the methods employed here. Such a defect might explain the mild decrease in fertility we observed in *Nestin-Cre/Gmnn^fl/fl^* mice. Nevertheless, our results clearly indicate that Geminin is largely dispensable for mammalian neurogenesis. Materials and Methods {#s4} ===================== Ethics Statement {#s4a} ---------------- All animal work was performed according to the protocol approved by the Nothwestern University Animal Care and Use Committee (Protocol \# 2009-0911). Mouse Breeding {#s4b} -------------- The construction of the *Gmnn^fl^* allele was described previously [@pone.0017736-Shinnick1]. *Nestin-Cre* mice (B6.Cg-Tg(Nes-Cre)1Kln/J, Jackson Labs Strain 3771) were kindly provided by Dr. Anjen Chenn (Northwestern University). For most experiments, *Nes-Cre/Gmnn^fl/fl^* mice were mated to *Gmnn^fl/fl^* mice to produce experimental *Nes-Cre/Gmnn^fl/fl^* mice and control *Gmnn^fl/fl^* littermates. To obtain embryos at defined embryonic stages, timed matings were performed in which male and female mice were caged together overnight then separated immediately after the mating plug appeared. Genotypes were determined by PCR of tail DNA. The following primers were used for amplification: Nes-Cre forward (5′-gcctgcattaccggtcgatgcaacga-3′), Nes-Cre reverse (5′-gtggcagatggcgcggcaacaccatt-3′), Geminin forward (5′-gctcagaggtttcaggg-3′), Geminin^WT^ reverse (5′-catcaggtgttctctcaagtgtctg-3′), and Geminin^fl^ reverse (5′-gctacttccatttgtcacgtcc-3′). Antibodies {#s4c} ---------- Affinity-purified anti-mouse Geminin antibody was obtained by immunizing rabbits with recombinant his-Geminin and purifying antibodies from serum over a column of recombinant his-Geminin coupled to cyanogen bromide sepharose beads [@pone.0017736-Shinnick1], [@pone.0017736-Harlow1]. Real Time PCR {#s4d} ------------- RNA was isolated from fresh tissue using Trizol reagent (Invitrogen). cDNA synthesis was carried using a standard kit (Ambion) and RT-PCR was performed using an Applied Biosystems 7500 Fast Real Time PCR System. Primers and fluorescently labeled probes for RT-PCR were designed using Primer Design software (Applied Biosystems). All RNA levels were normalized to the amount of 18S ribosomal RNA in each sample. For [Figure 2](#pone-0017736-g002){ref-type="fig"}, Geminin RNA primer sequences were 5′-acggatgctaggccgtgtac-3′ (forward), 5′-gcaccgtgtagttagtttaccaagag-3′ (reverse), and 5′-acgcactgccagcgttgccc-3′ (probe). 18S RNA primer sequences were 5′-aacgagactctggcatgctaact-3′ (forward), 5′-cgccacttgtccctctaagaa-3′ (reverse), and 5′-ttacgcgacccccgagcgg-3′ (probe). For PCR of exons 3 and 4 ([Figure S3](#pone.0017736.s003){ref-type="supplementary-material"}), Geminin primers were 5′-gtgaagaatagtcctgtccc-3′ (forward) and 5′-ccacagcttgaagtctgag-3′ (reverse). SYBR green was used instead of a probe. Histology and Immunohistochemistry {#s4e} ---------------------------------- For pulse labeling, mice were injected intra-peritoneally with three doses of bromodeoxyuridine (Invitrogen B23151, 10 mg/ml in PBS, 50 mg/kg) separated by two hours. Brains were fixed 24 hours after the last BrdU dose by transcardial perfusion with a solution of 4% paraformaldehyde in PBS then post-fixed overnight at 4°C in the same solution. For pulse-chase analysis, mice received a single daily dose of BrdU (50 mg/kg) for 12 days and the brains were fixed 9 weeks after the final dose. Free-floating coronal sections (40 µM) were cut using a vibratome (Leica VT 1000S) and collected in cold PBS. For β-galactosidase staining, sections were incubated at room temperature in PBS containing 5 mM K~3~Fe(CN)~6~, 5 mM K~4~Fe(CN)~6~, 2 mM MgCl~2~, and 1 mg/ml X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside, Roche). For Nissl staining, paraformaldehyde-fixed brains were embedded in OCT Medium and cut into frozen sections (10 µM). Nissl staining was performed using standard procedures [@pone.0017736-Prophet1]. To stain for BrdU, sections were incubated in 0.1 N HCl in PBS at 60°C for 20 minutes then washed three times in PBS at room temperature. Sections were blocked in 1X PBS, 10% Normal Donkey Serum (NDS, Jackson Immuno Research), 0.5% Triton X-100 (TX-100) and stained overnight at 4°C with anti-BrdU antibody (Fitzgerald Industries \#20-BS17) diluted 1:500 in 1X PBS, 7.5% NDS, and 0.25% TX-100. After extensive washing in PBS, sections were sequentially stained with biotinylated Donkey anti-Sheep antibody (Jackson \#713-065-003) diluted in 1X PBS, 5% NDS, and 0.25% TX-100 and with Alexa 663-conjugated streptavidin (Invitrogen S-21375) diluted in 1X PBS. For NeuN staining, BrdU-stained sections were blocked in 1X PBS, 4% Normal Goat Serum (NGS, Jackson), and 0.25% TX-100 then stained overnight at 4°C with anti-NeuN antibody (Millipore clone A60, \#MAB377) diluted 1:300 in 1X PBS, 2% NGS, and 0.25% TX-100. After washing in PBS, the sections were stained with Alexa 488-conjugated Donkey anti-Mouse IgG1 (Invitrogen A21121) diluted 1:300 in 1X PBS, 1% NDS, and 0.25% TX-100. Sections were mounted in Vectashield medium containing DAPI (Vector Laboratories). To determine the number of BrdU^+^ cells in the dentate gyrus, we counted every sixth section (8 sections per brain) and multiplied the result by 6. A total of 900 -1400 BrdU^+^ cells were counted per brain. To determine the number of NeuN^+^BrdU^+^ double-positive cells, we took confocal images of representative sections and used Z-stack images to determine the percentage of BrdU^+^ cells that were also NeuN^+^, then multiplied this percentage (∼50% for all mice) by the total number of BrdU^+^ cells. To determine these percentages we counting an average of 250 BrdU^+^ cells per brain. Neurosphere Culture and Differentiation {#s4f} --------------------------------------- Primary neurosphere cultures were obtained from newborn (P0) mice as described [@pone.0017736-Zhu1], [@pone.0017736-Mehler1], [@pone.0017736-Bonaguidi1]. Neurospheres were grown in DMEM/F-12 medium containing N2 supplement, B27 supplement, penicillin, streptomycin, glutamine (all from Invitrogen), and 20 ng/ml Epidermal Growth Factor (EGF, BD Biosciences). After 3 days the cells were dissociated with trypsin, triturated, counted, and re-seeded at a density of 5 X 10^4^ cells/ml. To calculate the doubling time, cells were trypsinized and re-counted three days later. For cell cycle analysis the cells were stained with propidium iodide using standard procedures[@pone.0017736-Darzynkiewicz1]. To induce differentiation, dissociated neurosphere cells were seeded onto poly-D-lysine coated coverslips (Sigma, 20 µg/ml for \>1 hour) in medium containing 1 ng/ml EGF. The medium was changed after 3 days, and after 7 days the coverslips were washed with PBS and fixed in 1X PBS/4% paraformaldehyde. For antibody staining, coverslips were washed with 1X PBS/0.1% TX-100 (PBST) then incubated overnight at 4°C in PBS/5% NGS containing either anti-TuJ1 antibody (Mouse IgG2b, clone SDL.3D10, Sigma \#T8660) or anti-GFAP antibody (Mouse IgG1, clone GA5, Sigma \# G3893). The secondary antibodies were Alexa 488-conjugated goat anti-mouse IgG2b (Invitrogen \# A21141) and Cy3-conjugated goat anti-rabbit IgG (Jackson Immuno Research \#111-165-144) respectively. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Protein Domains Deleted in** ***Gmnn^Δ/Δ^*** **Mice.** (Top) Map of the Geminin locus with color-coded exons. Deleted exons are enclosed by the rectangle. (Bottom) Exon boundaries mapped onto the domains of the Geminin protein. D-box, destruction box; NLS, bipartite Nuclear Localization Signal; Neural, neuralizing domain (underlined) which overlaps the D-box and the NLS. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Geminin Deletion does not Generate a Detectable New Fusion Protein.** The full immunoblot from [Figure 2A](#pone-0017736-g002){ref-type="fig"}. Biotinylated MW markers are shown on the left. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Geminin Deletion does not Cause Over-Expression of Geminin RNA.** RNA was isolated from undifferentiated or differentiated neurospheres from control or Nes-Cre/Gmnn(fl/fl) mice. The amount of exon 3/4-containing RNA was determined by RT-PCR. The location of the amplified fragment is indicated in [Figure S1](#pone.0017736.s001){ref-type="supplementary-material"}. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Dr. Anjen Chenn for the gift of *Nestin-Cre* mice. **Competing Interests:**The authors have read the journal\'s policy and have the following conflicts: TJM is an inventor for the patent \"Geminin Gene and Protein\" (US Patent \#6,548,290) that describes manipulating the Geminin level in cells to treat neurodegenerative diseases. The Assignee for this patent is the President and Fellows of Harvard University. TJM\'s status as an inventor does not alter our adherence to all PLoS guidelines concerning the sharing of data and materials as outlined in the guidelines for authors. **Funding:**This work was supported by grants to TJM from the American Heart Association (\#0630290N, [www.aha.org](http://www.aha.org)), the Illinois Division of the American Cancer Society ([www.cancer.org](http://www.cancer.org)), the National Heart Lung and Blood Institute (R21HL093500), and the Schweppe Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: TJM RJM. Performed the experiments: KMS ROL GB TM. Analyzed the data: TJM KMS ROL GB. Contributed reagents/materials/analysis tools: TJM JAK RJM. Wrote the paper: TJM.

PubMed Central

2024-06-05T04:04:19.269680

2011-3-9

PMC3052384

Introduction {#s1} ============ Microorganisms have large population sizes and show long-distance dispersal, high reproductive rates and remarkable genetic diversity, suggesting that they can cross environmental boundaries, including salinity, more frequently than multicellular organisms [@pone.0017789-Logares1]. These particularities support the Baas-Becking hypothesis formulated in 1934, summed up as follows: "Everything is everywhere, but the environment selects" (revised by Hooper *et al.* [@pone.0017789-Hooper1]). Although this seems logical and plausible, the clustering test performed *in silico* by Lozupone and Knight [@pone.0017789-Lozupone1] using annotated sequences from 202 globally distributed natural environments demonstrates that salinity is the major barrier to microbial communities, showing a strong environment-specific evolution between freshwater and marine bacteria. Until the late 1980\'s, fresh and salt water planktonic bacteria were thought to be ecologically similar, despite minor differences such as some biotic interactions within the food web and sodium requirement. Salt-dependence in marine bacteria was not considered a fundamental ecological difference and species distribution and their physiology were thought to be similar to freshwater bacteria [@pone.0017789-Hobbie1]. Since molecular methods started to be applied to the study of uncultivated microbial communities [@pone.0017789-Pace1], [@pone.0017789-Olsen1], knowledge of microbial ecology in aquatic systems has been significantly increased [@pone.0017789-Fuhrman1]--[@pone.0017789-McCarren1]. The first difference seen in bacterial community composition in fresh and marine water was the dominance of *β-Proteobacteria* in the former, in contrast to the dominance of α- and γ- subdivisions of *Proteobacteria* in the latter [@pone.0017789-Meth1]--[@pone.0017789-Giovannoni1]. Most bacterial sequences retrieved from freshwater environments were neither affiliated with known bacterial species nor with soil and marine relatives but clustered in a habitat-specific manner, leading to the conclusion that these were typical freshwater bacteria. Interestingly, this bacterial cluster presented a cosmopolitan distribution, including habitats located in different climatic zones [@pone.0017789-Hahn1]. Estuarine waters are dynamic environments due to the mixing of sediments, marine and freshwater, resulting in salinity and nutrient gradients. Shifts in physical, chemical, and microbiological properties between freshwater and adjacent coastal marine environments occur in short periods of time, driven by tides and freshwater flow, creating an intense abiotic pressure that influences the composition of bacterioplankton communities [@pone.0017789-Crump1]. The presence and abundance of typical freshwater and marine bacterial taxa are closely related with these gradients and also with growth rates, viral lysis, predation, and retention times [@pone.0017789-Gonzalez1]--[@pone.0017789-Vieira1]. Long-term adaptability to different salinity conditions is also indicated by the ability of some organisms to occur in both marine and freshwater habitats [@pone.0017789-Sigee1]. In spite of a number of published studies of large estuaries and *in silico* comparisons between freshwater and seawater bacterioplankton, very few concerned South American tropical habitats. The Atlantic rain forest, a species diversity hotspot [@pone.0017789-Myers1]--[@pone.0017789-Carnaval1], represents a substantial contribution of organic and inorganic material to the coastal waters of the Southwest Atlantic Ocean. Bacteria and fungi from Atlantic forest habitats have been analyzed mainly by culture-dependent methods [@pone.0017789-Baltazar1]--[@pone.0017789-Carvalho1]. By means of 16S rRNA gene libraries, it has been estimated that millions of new bacterial species exist in the Atlantic rain forest soil and phyllosphere [@pone.0017789-Lambais1]--[@pone.0017789-Faoro1]. As most of the Brazilian population lives in the coast, Atlantic forest habitats are greatly impacted by human activities. The Atlantic rain forest extends along the Brazilian coast from Rio Grande do Norte to Rio Grande do Sul states and has been reduced to less than 8% of its range [@pone.0017789-Morellato1]. The forest has a well-defined dry winter and rainy summers with high precipitation levels, with a mean annual rainfall of 1368 mm [@pone.0017789-Marengo1] that greatly increases river transport. This dynamic hydrology sustains a great biodiversity of flora and fauna which characterizes the Atlantic forest as a diversity hotspot [@pone.0017789-Myers1]--[@pone.0017789-Carnaval1]. One of the few protected areas of the Atlantic rain forest is Ilha Grande island in Rio de Janeiro state, Brazil ([Figure 1](#pone-0017789-g001){ref-type="fig"}). Ilha Grande has some coastal marine and freshwater sites that may be considered as undisturbed. Based on the construction and analyses of 16S rRNA gene libraries, we compared bacterioplankton diversity in six representative habitats of Ilha Grande\'s aquatic ecosystems in the context of a salinity gradient. Here we present results that corroborate the idea of divergent evolution and the lack of transitions between marine and freshwater bacterial communities. ::: {#pone-0017789-g001 .fig} 10.1371/journal.pone.0017789.g001 Figure 1 ::: {.caption} ###### Map of the studied site and the six sampled locations. FWS -- Parnaioca freshwater spring; FWP -- Parnaioca river; FWM -- mangrove; SWP -- Parnaioca beach; SWA -- Aventureiros beach; SWM -- seawater near Meros island. :::  ::: Materials and Methods {#s2} ===================== Sampling {#s2a} -------- The six analysed sites, three freshwater and three marine, are shown in [Figure 1:](#pone-0017789-g001){ref-type="fig"} FWS - a water spring (23°10′57.00″S/44°14′55.19″W); FWR - Parnaioca river (23°11′21.33″S/44°15′11.08″W); SWP - Parnaioca beach (23°11′24.77″S/44°15′15.07″W), just where Parnaioca river flows into; FWM - a mangrove (23°10′26.98″S/44°17′08.49″W) which, at the time of sampling, had the communication to the sea closed by a sand barrier; SWA - Aventureiros beach (23°11′24.53″S/44°18′58.06″W); SWM - two milles west from Ilha Grande island near Meros island (23°12′53.67″S/44°21′55.03″W). Water samples (5.8 Liters) were collected at 1 m depth (except for the water spring) on September 7, 2007 for DNA extraction and for abiotic and microbiological characterization (100 mL). Samples were kept on ice until processed in the laboratory. Chemical and microbiological parameters {#s2b} --------------------------------------- Chemical data were determined in triplicates by standard oceanographic methods. Temperature, salinity, and pH were determined at the moment of sample collection using a field thermometer, a hand-held refractometer (Leica) and pH strips. Ammonia was measured by the indophenol method [@pone.0017789-Parsons1], nitrite by diazotation [@pone.0017789-Grasshoff1] and nitrate by reduction in a Cd-Cu column followed by diazotation [@pone.0017789-Grasshoff1]. Total phosphorus was evaluated by acid digestion to phosphate and silicate by reaction with molibdate [@pone.0017789-Grasshoff1]. Bacterial abundance was determined by flow cytometry [@pone.0017789-Andrade1] and bacterial production by ^3^H-leucine incorporation [@pone.0017789-Kirchman1]--[@pone.0017789-Gonzalez2]. Specific production (SP) is an index calculated as the ratio Microbial Production versus Microbial Abundance [@pone.0017789-Urbach1] that allows comparisons of secondary productivity between environments with differences in prokaryotic counts. DNA extraction {#s2c} -------------- The water samples were filtered through 0.2 µm Sterivex filters (Millipore, Bedford, MA, USA) after filtration through 3.0 µm to separate free-living microbes from larger organisms and particles. Total cellular nucleic acids were isolated by cell lysis with proteinase K and SDS, followed by phenol-chloroform extraction [@pone.0017789-Vieira2]. DNA integrity was checked on a 1% (w/v) agarose gel that was subsequently stained with Syber Green (FMC Bioproducts, Rockland, ME, USA) and the gel image was digitalized with Storm Image Scanner (GE Healthcare, Little Chalfont, UK). Bacterial 16S rRNA gene library construction {#s2d} -------------------------------------------- PCR was performed in 50 µl reaction mixtures (2.5 mM MgCl~2~, 0.2 mM deoxynucleoside triphosphates, 1 ng of each primer.µl^−1^, 2.5 U of High Fidelity *Taq* DNA polymerase \[Promega\], 1× PCR buffer and 200 ng of each environmental DNA sample, using the universal bacterial primers 27BF (5′-AGAGTTTGATCCTGGCTCAG-3′) [@pone.0017789-Lane1] and 907RAB (5′-TTTGAGTTT MCTTAACTGCC-3′) [@pone.0017789-Weisburg1]. PCR amplification began with a 5 min denaturing step at 94°C; this was followed by 25 cycles of 94°C for 90 seconds, 50°C for 90 seconds, and 72°C for 2 min. The final cycle was an extension at 72°C for 5 min. PCR products were concentrated and purified with a GFx PCR DNA and Gel Band Purification Kit (GE Healthcare) after electrophoresis on a 1% (w/v) agarose gel. PCR products were cloned into the pGEM-T cloning vector (Promega) and used to transform competent *E. coli* DH10B cells. Positive colonies for the blue-white colony screen used for this vector were picked and frozen at −70°C. Six 16S rRNA gene libraries were constructed from different environmental DNA samples. Sequence analyses and taxa identification {#s2e} ----------------------------------------- Approximately 192 clones from each clone library were submitted to sequence analysis. Plasmidial DNA from each clone (400 ng) was prepared and PCR-sequencing reactions with primer 27BF were carried out using the DYEnamic ET terminator cycle-sequencing kit (GE Healthcare). Partial 16S rRNA sequences were obtained by capillary electrophoresis on a MegaBace1000 DNA analysis system (GE Healthcare). Chromatograms were transformed into Fasta format with Phred software [@pone.0017789-Edwing1] and sequences with less than 300 bp and chimeras were removed prior to further analysis using MOTHUR. A total of 831 valid sequences with approximately 642 bp were compared with sequences in the Ribosomal Database Project II [@pone.0017789-Cole1]. Sequences were also analyzed by BLAST [@pone.0017789-Altschul1] searches in GenBank database (<http://www.ncbi.nlm.nih.gov>) and were aligned with representative bacterial sequences obtained from the public databases using ClustalX software [@pone.0017789-Thompson1]. The partial 16S rRNA gene sequences generated in this study have been deposited in GenBank under accession numbers FJ717864-FJ718690. All submissions conform to the "Minimum information standards" recommended by the Genomic Standards Consortium [@pone.0017789-Yilmaz1]. Biodiversity and phylogenetic analyses {#s2f} -------------------------------------- Re-sampling and adjustment of the total number of sequence reads to identical sequencing depth was done before analysis [@pone.0017789-Gilbert1]. Sequences were clustered as OTUs at an overlap identity cutoff of 97% or 80% by MOTHUR software [@pone.0017789-Schloss1]. Richness and diversity statistics including the nonparametric richness estimators ACE, Chao1 and the Shannon diversity index were calculated. The diversity of OTUs and community overlap were also examined using rarefaction analysis and Venn diagrams. Phylogenetic trees were constructed for marine and freshwater libraries with reference sequences from GenBank by the neighbor-joining algorithm based on distances calculated by the Kimura-2 method. This analysis was performed with the MEGA4 program [@pone.0017789-Kumar1] and bootstrap analysis with 1000 replications was used. Tree topology and distribution of hits along the tree were uploaded to the UniFrac computational platform [@pone.0017789-Lozupone1], [@pone.0017789-Lozupone2]. UniFrac is a beta diversity metric analysis that quantifies community similarity based on phylogenetic relatedness. In order to visualize distribution patterns of bacterial communities we used the UniFrac metric to perform PCA highlighted by significance. Libraries were sub-sampled randomly to test the consistency of the results. Statistical comparison between 16S rRNA libraries {#s2g} ------------------------------------------------- In an attempt to determine the differences between clone libraries, we applied LIBSHUFF statistics [@pone.0017789-Schloss2] that uses Monte Carlo methods to generate homologous and heterologous coverage curves. Sequences were randomly shuffled 999 times between samples prior to the distance between the curves being calculated using the Cramér-von Mise statistic test. The DNADIST program of the PHYLIP package, using the Jukes-Cantor model for nucleotide substitution was used to generate the distance matrix analyzed by LIBSHUFF. Results {#s3} ======= Abiotic and microbiological parameters {#s3a} -------------------------------------- Abiotic and microbiological parameters from each sampling site are shown in [Table 1](#pone-0017789-t001){ref-type="table"}. Temperatures varied from 22 to 28°C. The low salinity found at Parnaioca beach (SWP) is explained by the input of freshwater from Parnaioca River to this site. In the same way, salinity in the mangrove (FWM) was typical of a freshwater environment due to strong rainfall that fell a few days before sampling which increased river input and blocked the communication of the mangrove with the sea by a sand barrier. For further analysis, the water spring, river and mangrove habitats were considered as freshwater environments, and Parnaioca, Aventureiros beach and Meros Island as marine environments. All are representative samples of the dynamic environmental conditions which characterize the Atlantic rain forest. Analysis of nitrogenated compounds showed the highest ammonia concentration at the mangrove site, FWM, while nitrate was the main compound in Parnaioca river, FWP. Nitrite concentrations ranged between 0.33 and 0.54 µM and silicate concentrations reached high values in the mangrove. Freshwater samples were more acidic than marine ones, with pH values ranging from 5.5 to 6.5 ([Table 1](#pone-0017789-t001){ref-type="table"}). ::: {#pone-0017789-t001 .table-wrap} 10.1371/journal.pone.0017789.t001 Table 1 ::: {.caption} ###### Abiotic and microbiological parameters. ::: {#pone-0017789-t001-1} FRESHWATER SEAWATER --------------------------------------------------------------- ------------ ---------- ------- ------- ------- ------- [a](#nt101){ref-type="table-fn"} **Sal (S)** 0.09 0.83 0.73 26.67 33.64 32.63 [b](#nt102){ref-type="table-fn"} **T (°C)** 22 22 28 25 25 26 [c](#nt103){ref-type="table-fn"} **TP (µM)** 0.54 0.22 0.78 0.32 0.49 0.33 [d](#nt104){ref-type="table-fn"} **NH~3~ (µM)** 0.48 1.11 7.17 1.40 0.90 0.73 [e](#nt105){ref-type="table-fn"} **NO~2~^−^ (µM)** 0.44 0.33 0.54 0.39 0.38 0.41 [f](#nt106){ref-type="table-fn"} **NO~3~^−^ (µM)** 1.80 8.20 nd 0.95 0.90 0.73 [g](#nt107){ref-type="table-fn"} **SiO~2~ (µM)** 20.45 28.03 44.98 22.85 2.44 1.53 **pH** 5.5 5.5 6.5 7.0 7.5 7.0 [h](#nt108){ref-type="table-fn"} **MA (10^6^cells.mL^−1^)** 0.15 0.23 1.36 0.30 0.26 0.12 [i](#nt109){ref-type="table-fn"} **MP (µg C.L^−1^.h^−1^)** 0.26 1.97 3.44 1.08 0.76 0.44 [j](#nt110){ref-type="table-fn"} **SP (fg C.cell^−1^.h^−1^)** 1.69 8.51 2.53 3.54 2.88 3.43 a Sal, salinity; b T, temperature; c TP, total phosphorous; d NH~3~, ammonia; e NO~2~ ^−^, nitrite; f NO~3~ ^−^, nitrate; g SiO~2~, silicon; h MA, microbial abundance; i MP, microbial production; and j SP, specific production. **FWS** -- Parnaioca freshwater spring; **FWP** -- Parnaioca river; **FWM** -- mangrove; **SWP** -- Parnaioca beach; **SWA** -- Aventureiros beach; **SWM** -- seawater near Meros island. ::: Prokaryotic counts were in the range of 10^6^ cells per mL, being most abundant in the mangrove. Bacterial production values, which mean the heterotrophic activity, varied from 0.26 to 3.44 µg C.L^−1^.h^−1^. Although the highest heterotrophic activity was found in the mangrove, the bacterial production versus bacterial counts ratio (specific productivity - SP) was higher in the river. Marine samples presented SP values varying from 2.88 to 3.54 ag C.cell^−1^h^−1^ ([Table 1](#pone-0017789-t001){ref-type="table"}). Clone library coverage, richness and diversity {#s3b} ---------------------------------------------- The number of OTUs from each site as well as richness and diversity indexes calculated by MOTHUR [@pone.0017789-Schloss1] are shown in [Table 2](#pone-0017789-t002){ref-type="table"}. The coverage of each library was calculated using the abundance-based coverage estimator (ACE). We also grouped freshwater (FWS, FWR, FWM) and marine sites (SWP, SWA, SWM) to perform these calculations. In order to account for uneven sampling efforts, the same number of sequences was randomly selected from each sample. The Parnaioca water spring, FWS, library had higher richness based on ACE, Chao1 and H′. Parnaioca river, the mangrove and Meros island libraries had the lowest richness values, but the H′ values were not far from the other libraries. Although no major differences among marine samples were found, SWP was the richest sample. Interestingly, the comparison between marine and freshwater libraries showed that, at 97% similarity level, bacterial richness and diversity of fresh and seawater communities are similar. ::: {#pone-0017789-t002 .table-wrap} 10.1371/journal.pone.0017789.t002 Table 2 ::: {.caption} ###### Species richness estimates and diversity of 16S rRNA gene sequences as determined by MOTHUR software. ::: {#pone-0017789-t002-2} FRESHWATER SEAWATER ------------------------------------------- ------------ ---------- ------ ------ ------ ------ ------ ------ [a](#nt112){ref-type="table-fn"} **OTUs** 269 89 56 58 219 90 63 57 [b](#nt113){ref-type="table-fn"} **ACE** 2457 1024 184 101 762 233 187 296 **Chao1** 1018 564 130 85 543 220 134 252 [c](#nt114){ref-type="table-fn"} **H′** 5.33 4.43 3.72 3.83 5.07 4.42 3.90 3.69 a Number of unique OTUs defined by using the furthest neighbor algorithm in MOTHUR at 97% similarity. b Abundance based coverage estimator (ACE). c Shannon-weaver index of diversity (H′). **FWS** -- Parnaioca freshwater spring; **FWP** -- Parnaioca river; **FWM** -- mangrove; **SWP** -- Parnaioca beach; **SWA** -- Aventureiros beach; **SWM** -- seawater near Meros island. **FW** and **SW** were calculated by merging the respective libraries. ::: All rarefaction curves at a high cutoff phylogeny resolution (97%) show that the diversity is very high and the total coverage of bacterial richness was not achieved. A decline in the rate of OTU detection at 80% cutoff indicates that the most dominant bacterial phyla have been detected for freshwater and marine samples. Rarefaction analysis at this cut-off revealed that freshwater environments were more diverse than marine ones, as well as at 97% cutoff ([Figure 2](#pone-0017789-g002){ref-type="fig"}). Additionally, Venn diagram shows that no OTUs are shared between fresh and marine water samples at species level (97%) indicating that the bacterial communities are completely different in these two kinds of environment. ::: {#pone-0017789-g002 .fig} 10.1371/journal.pone.0017789.g002 Figure 2 ::: {.caption} ###### Rarefaction analysis of 16S rDNA clone libraries from Ilha Grande using a distance level of 80% (A, B and E) and 97% (C, D and F). In A and B or C and D each freshwater or marine water libraries are plotted, respectively. In E and F the three samples of seawater and the three samples of the freshwater were joined. FWS -- Parnaioca freshwater spring; FWP -- Parnaioca river; FWM -- mangrove; SWP -- Parnaioca beach; SWA -- Aventureiros beach; SWM -- seawater near Meros island. :::  ::: Bacterial Groups {#s3c} ---------------- In order to reveal bacterial phyla composition in such diverse communities, sequences from each library were classified with the RPD classifier tool (<http://rdp.cme.msu.edu/classifier>). Marine samples showed a higher abundance of *Cyanobacteria*, *Alphaproteobacteria* while freshwater samples were dominated by *Betaproteobacteria* ([Figure 3](#pone-0017789-g003){ref-type="fig"}). *Gammaproteobacteria* were found mainly in the river (FWP) and Meros island (SWM) sites. A minor proportion of *Deltaproteobacteria* was observed in the FWP and mangrove (FWM) libraries. *Actinobacteria* were seen only in the river and mangrove environments, being more abundant in the latter one. *Bacteroidetes* were present in all the sites, except at the water spring. The newly described group OD1 was only found at the water spring and mangrove sites. A greater percentage of unclassified sequences were found in marine samples. Freshwater samples were richer at the phylum level than marine ones, with nine and four phyla represented, respectively. ::: {#pone-0017789-g003 .fig} 10.1371/journal.pone.0017789.g003 Figure 3 ::: {.caption} ###### Distribution of sequences in bacterial phyla classified by the Classifier tool at RDP Database. Clones from freshwater libraries are shown in A and from seawater are shown in B. FWS -- Parnaioca freshwater spring; FWP -- Parnaioca river; FWM -- mangrove; SWP -- Parnaioca beach; SWA -- Aventureiros beach; SWM -- seawater near Meros island. :::  ::: Phylogenetic Analysis {#s3d} --------------------- The phylogenetic tree allowed us to recognize the bacterial phylotypes that compose the groups listed above ([Figure 4](#pone-0017789-g004){ref-type="fig"}). The tree shows that most of our sequences were affiliated to environmental uncultured bacterial species. In freshwater samples, *Betaproteobacteria* sequences were affiliated to uncultured bacteria from lakes, freshwater ponds, aquifers, rivers, and subsurface freshwater. A great number of sequences from the river site were closely related to *Acidovorax* sp. The *Acinetobacter* was the most represented group among *Gammaproteobacteria*. Members of *Bacteroidetes* were not found in the water spring while they occurred in high percentage in the mangrove and river sites. Among all freshwater sequences, only two mangrove clones fell into the *Alphaproteobacteria* clade, being related to *Rhodobacteriaceae* retrieved from a Taiwan mangrove and river sediments, and two other OTUs fell into the *Deltaproteobacteria* group. At the mangrove, *Actinobacteria* were mainly represented by *Microbacteriaceae*. Additionally, in the mangrove and river libraries we found members of the recently proposed OD1 group, affiliated with a eutrophic lake bacterium. The *Cyanobacteria* found in the mangrove were related to marine species, different from those of the water spring site which were more related to drinking water system bacteria. ::: {#pone-0017789-g004 .fig} 10.1371/journal.pone.0017789.g004 Figure 4 ::: {.caption} ###### Phylogenetic tree of bacterial clones obtained in the freshwater or seawater locations. Reference sequences from GenBank (**in bold**). OTUs were defined by using a distance level of 3% by using the furthest neighbor algorithm in MOTHUR. One access number from each OTU is displayed. The tree topology is based on neighbor joining and bootstrap analysis was performed with 1000 replications. Bootstrap value \>50 and representative OTUs are shown. More detailed trees can be found in [Figures S1](#pone.0017789.s001){ref-type="supplementary-material"} and [S2](#pone.0017789.s002){ref-type="supplementary-material"}. :::  ::: Phylogenetic analysis of the marine libraries revealed that *Cyanobacteria* were well represented by *Prochlorococcus* and *Synecchococus*, which is expected for coastal marine samples. Sequences from marine samples were mainly represented by *Alphaproteobacteria*. In this group, a representative clade with OTUs related to uncultured bacteria from Chesapeake Bay (USA), Mallorca Island (Spain), and Guanabara Bay (Brazil) and other clades with OTUs related to genera commonly found in marine waters, like *Roseobacter* and *Ruegeria*, were observed. The distribution of OTUs within *Gammaproteobacteria* followed this pattern, with a representative clade formed by uncultured bacteria from marine samples and by *Neptuniibacter* and *Oceanospirillum* species and another clade related to *Alteromonas*. Library Comparison {#s3e} ------------------ The comparison by LIBSHUFF statistics revealed that bacterial community composition differed significantly between marine and freshwater sampling sites. We obtained p\<0.0001 for the comparisons of each marine library to each freshwater ones and also for the comparison of all marine sequences against all freshwater ones. Nevertheless, freshwater libraries were different among themselves whereas marine libraries were statistically similar (p = 0.0003 for the comparison between Parnaioca and Aventureiros, p = 0.0004 for Parnaioca and Meros, and p = 0.1718 for Aventureiros and Meros). Through a scatter plot of the first two principal coordinates by the UniFrac analysis ([Figure 5](#pone-0017789-g005){ref-type="fig"}), PC1 and PC2 explained 9.5% and 7.4% of the data variation, respectively. The randomly constructed sub-libraries were grouped according to the original libraries. Marine libraries were separated from freshwater ones in the plot by PC1. The three marine libraries grouped together showing a high similarity with each other, whereas freshwater samples were dispersed in the plot and seem to be different among them. Additionally, the mangrove FWM clustered between freshwater and marine samples along the PC1 axis, which divides saline and other freshwater environments. This result corroborates the LIBSHUFF analysis, wherein only marine libraries reached high p values. ::: {#pone-0017789-g005 .fig} 10.1371/journal.pone.0017789.g005 Figure 5 ::: {.caption} ###### Match between bacterial communities in freshwater and seawater samples. Principal coordinates plots (PCA) were generated using the pair wise unweighted UniFrac distances. Freshwater in open symbols: FWS (**△**) -- water spring; FWP (○) -- Parnaioca river, FWM (□) - mangrove. Marine samples in filled symbols: SWP (•) -- Parnaioca beach, SWA (▪) -- Aventureiros beach, SWM (▴) -- Meros island. :::  ::: Discussion {#s4} ========== In this work we investigated for the first time the bacterioplankton diversity in the tropical island, Ilha Grande. This environment suffers very low anthropogenic impact and is located in the Brazilian coast at the South Atlantic Ocean. The differences found in community composition add new knowledge to planktonic bacteria distribution in freshwater and coastal marine ecosystems. Many abiotic parameters, such as nutrient concentration and organic matter, are thought to influence the composition of natural bacterioplankton communities [@pone.0017789-Jickells1]--[@pone.0017789-Carlson1]. In the same manner, autochthonous biological activity can modify water chemical features [@pone.0017789-Strom1]. In this study, nutrient concentrations in marine samples were similar to Sepetiba Bay values but lower than in the highly eutrophic Guanabara Bay [@pone.0017789-Almeida1], [@pone.0017789-Vieira2]. Both are economically important water bodies which lie geographically close to Ilha Grande. In the mangrove environment, high bacterial production contrasts with low specific productivity. A possible explanation is that many marine cells that entered into the mangrove are not active anymore because of the change in salinity. In opposition, the river community, that reached higher specific productivity values, seems to be a well-adapted community, which probably has a large supply of oxygen available for aerobic metabolism. In estuaries, shifts in bacterioplankton community composition along salinity gradients are related to residence and community doubling times [@pone.0017789-Crump1], [@pone.0017789-Crump2]. Specific productivity and bacterial abundance estimates allow microbial communities to be compared and can be used to measure the metabolic status of the planktonic microbes [@pone.0017789-Urbach1]. A particular estuarine community is formed in intermediate salinities when average metabolic status and, consequently, the doubling times are shorter than residence times. Although specific productivity values for Parnaioca river and all marine samples are around one order of magnitude higher when compared to a previous study in Guanabara bay, an urban, pollution impacted Brazilian bay [@pone.0017789-Vieira3], there is no water residence time as the river water flows directly into the sea without a transition area, causing an abrupt change in salinity, and giving no time for the development of local bacterial species. The consequence is a complete shift in community composition when Parnaioca river and Parnaioca beach are compared, despite the close proximity (50 m) of these two sites. Typical marine clades, such as *Cyanobacteria* and the *Alpha* and *Gamma* subdivisions of *Proteobacteria* were more represented in marine coastal and open-sea samples, not just in our data but also in the literature [@pone.0017789-Crump1], [@pone.0017789-Brown1]. However, in contrast to previous studies that found a low relative abundance of phototrophic *Cyanobacteria* compared to heterotrophic bacteria [@pone.0017789-Giovannoni2]--[@pone.0017789-Pommier1], members of *Synechococcus* and *Prochlorococcus* were one of the most abundant groups in Ilha Grande marine samples. The most abundant group in water spring, river, and mangrove sites were the *Betaproteobacteria*, a typical freshwater clade [@pone.0017789-Meth1] that was not recorded in marine samples. Recovery of 16S rRNA gene clones affiliated to *Betaproteobacteria* is common in libraries constructed from coastal samples, but few to no *Betaproteobacteria* have been reported by open ocean surveys [@pone.0017789-Crump1], [@pone.0017789-Brown1], [@pone.0017789-Riemann1]--[@pone.0017789-Huber1]. These findings lead to the idea that bacterioplankton represented by these lineages have a probable freshwater origin and are adapted to coastal marine environments and could be representative of bacterioplanckton phylotypes that transit between freshwater and marine habitats [@pone.0017789-Rappe1]. However, the present data clearly do not support this proposal, since no *Betaproteobacteria* was retrieved from our marine libraries. The *Gammaproteobacteria* and *Bacteroidetes* clades were well represented in both saline and freshwater environments. This might be a consequence of the presence of closely related marine phylotypes of common freshwater taxa [@pone.0017789-Pommier1]. In fact, the bacterial phylotypes belonging to these two clades encompass distantly related organisms in freshwater and marine samples, as seen in the phylogenetic trees, indicating an evolutionary separation between these marine and freshwater lineages [@pone.0017789-Logares1]. In the marine sites, several *Gammaproteobacteria* and *Bacteroidetes* related OTUs were affiliated to sequences from marine habitats of different geographic areas, indicating that these are worldwide distributed bacteria. Our data show a strong spatial heterogeneity of bacterial community composition in Ilha Grande. Most libraries, except when the three marine libraries are compared among themselves, are statistically different to each other. This most likely reflects the remarkable abiotic differences of these environments, especially salinity. This was also observed by Vieira et al [@pone.0017789-Vieira3] in Guanabara Bay, but contrasts to the results found for Chesapeake Bay (USA), where only temporal variation was significant [@pone.0017789-Kan1]. The water spring is an interesting case, as it is highly different from the other environments, including other freshwater habitats. This may be explained by a strong influence of soil, plant-associated and underground water bacterial communities. As seen by Lozupone [@pone.0017789-Lozupone1], our data showed a clear separation between freshwater and marine libraries. The PC1 axis represented the saline barrier which segregates marine and freshwater bacterial communities. In fact, salinity is pointed out as the major environmental determinant of aquatic microbial community composition, rather than extremes of temperature, pH, or other physical and chemical factors by the global pattern of the bacterial diversity [@pone.0017789-Lozupone1]. Recently, deep evolutionary divergence between marine and freshwater SAR11 lineages was seen not only by means of 16S phylogenetic constructions and Unifrac analysis, but also by Fragment Recruitment Analysis using metagenomic libraries from environments of different salinities [@pone.0017789-Logares2]. Although our marine samples clustered together in the PCA analysis, freshwater ones were dispersed in the plot, showing a higher heterogeneity among these environments. Interestingly, mangrove communities cluster along the PC1 axis, between saline and other freshwater environments. This could be a result of the recent changes in salinity due to a sand barrier formation and the intense rainfall that brought a large input of freshwater to this habitat. The dispersion seen among the freshwater environments has been observed in other studies [@pone.0017789-Lozupone1], [@pone.0017789-Logares2] and is probably the result of complex interactions between biotic and abiotic factors, not only salinity, which ultimately shape communities in natural habitats. Community composition changes across salinity gradients probably lead to changes in expression patterns that can modify the way in which organisms interact with each other and with the environment. In fact, seasonal changes in bacterial gene expression patterns across the salinity gradient in the Columbia river was recently observed by microarrays [@pone.0017789-Smith2]. In summary, our results support the notion of ecologically defined bacterial species and processes and increase our knowledge about the relationships between bacterial diversity and environmental parameters in a tropical region. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Phylogenetic tree of bacterial clones obtained in the freshwater locations.** Reference sequences from GenBank (**in bold**). OTUs were defined by using a distance level of 3% by using the furthest neighbor algorithm in MOTHUR. The tree topology is based on neighbor joining and bootstrap analysis was performed with 1000 replications. Bootstrap value \<50 and singletons are not shown. FWS (**△**) -- Parnaioca freshwater spring; FWP (○) -- Parnaioca river; FWM (□) -- mangrove. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Phylogenetic tree of bacterial clones obtained in seawater locations.** Reference sequences from GenBank (**in bold**). OTUs were defined by using a distance level of 3% by using the furthest neighbor algorithm in MOTHUR. The tree topology is based on neighbor joining and bootstrap analysis was performed with 1000 replications. Bootstrap value \<50 and singletons are not shown. SWP (•) -- Parnaioca beach; SWA (▪) -- Aventureiros beach; SWM (▴) -- seawater near Meros island. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We acknowledge Genome Sequencing facilities core Johanna Döbereiner IBqM/UFRJ. We are grateful to Aline Turque, Denise Oliveira, and Vivian Monteiro for library construction and sequencing, and Alessandra Gonzalez for bacterial abundance and production measurements. Special thanks to Barbara Ignacio for enthusiastic discussions. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by grants from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Fundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro (FAPERJ). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: CBS RPV OBM. Performed the experiments: CBS RPV RP. Analyzed the data: CBS RPV AMC RMA. Contributed reagents/materials/analysis tools: AMC RP RMA OBM. Wrote the paper: CBS RPV AMC RMA.

PubMed Central

2024-06-05T04:04:19.272542

2011-3-9

PMC3052385

Introduction {#s1} ============ Human kidneys filter about 180 liters of blood per day, retaining most of the plasma proteins in blood and allowing passage of water and small molecules into urine. The filtering unit of the kidney, the glomerulus, consists of a network of capillaries covered by specialized visceral epithelial cells, the podocytes, and supported by the mesangial cells in the interstitium. Glomerular capillaries are lined by very specialized glomerular endothelial cells (GEnCs) that possess transendothelial pores or "fenestrations" and an overlying negatively charged glycocalyx [@pone.0017838-Ballermann1]. GEnCs and podocytes secrete basement membranes which fuse as glomeruli mature to form the glomerular basement membrane (GBM) [@pone.0017838-Mundel1], [@pone.0017838-Abrahamson1]. Collectively, GEnCs, the GBM and podocytes together constitute the glomerular filtration barrier [@pone.0017838-Haraldsson1]--[@pone.0017838-Jarad1]. Injury to or disruption of any of its constituents compromises the integrity of the glomerular filtration barrier resulting in proteinuria. In the recent years, several critical molecular components of podocytes that are indispensable for glomerular filtration have been identified. A number of human diseases with mutated podocyte components have also been identified and several mouse models have been generated that have facilitated our understanding of podocyte biology [@pone.0017838-Kerjaschki1]--[@pone.0017838-Patrakka2]. In comparison, our knowledge of molecular components of GEnCs has lagged behind. This is largely due to lack of information on functionally-critical glomerular endothelium-restricted proteins and the unavailability of suitable mouse models that exhibit glomerular disease upon deletion of endothelial-expressed genes. Members of the EHD protein family (EHD1-4) have emerged as critical regulators of endocytic traffic of membrane as well as cell surface receptors [@pone.0017838-Grant1]. The reported glomerular endothelium-restricted expression of EHD3 within the kidney [@pone.0017838-Patrakka3], [@pone.0017838-Brunskill1] suggested the possibility that EHD protein-mediated endocytic recycling might contribute to GEnC-specific functions. This possibility was strengthened by recent studies that have implicated EHD proteins in regulating physiological functions in other cell types, including adipocytes [@pone.0017838-Guilherme1], cardiomyocytes [@pone.0017838-Gudmundsson1], myocytes [@pone.0017838-Doherty1], hepatic sinusoidal endothelial cells [@pone.0017838-Jia1] and neuronal cells [@pone.0017838-Yap1]. Though EHD3 was the first protein shown to be specifically expressed in the glomerular endothelium, to date, its potential role in this context has not been investigated. EHD proteins are characterized by an EH domain at the C-terminus, a nucleotide binding P-loop near the N-terminus and a central coiled-coiled region. The EH domain mediates protein--protein interactions by binding to the Asn-Pro-Phe (NPF) tri-peptide motif present in interacting proteins and facilitates membrane binding through association with membrane phospholipids [@pone.0017838-Naslavsky1], [@pone.0017838-Blume1]. The P-loop binds and hydrolyzes ATP at a very slow rate *in vitro* [@pone.0017838-Lee1], while the coiled-coiled region participates in oligomerization [@pone.0017838-Naslavsky2]. To investigate the *in vivo* functional roles of EHD proteins we have previously employed a gene knockout approach and such studies of *Ehd1* and *Ehd4* have demonstrated their roles in normal murine development and physiology [@pone.0017838-Rainey1]--[@pone.0017838-George1]. Deletion of *Ehd1* resulted in small testis with male infertility (27), while *Ehd4* deletion resulted in small testis with moderate reduction in sperm count (28), indicating a role for EHD1 and EHD4 in male germ cell development/differentiation. Deletion of *Ehd1* also leads to embryonic lethality and ocular defects whose severity varies with the genetic background of the mouse strains used (Rainey MA *et al*., manuscript in preparation). Here, we extended this gene knockout approach to *Ehd3*; however, *Ehd3* ^--/--^ mice showed no discernable pathology. Upregulation of the expression of EHD4, a related family member, in *Ehd3* ^--/--^ GEnCs led us to hypothesize that EHD4 functionally compensates for loss of EHD3. We generated *Ehd3^--/--^; Ehd4^--/--^* mice to test this hypothesis and observed severe glomerular disease in these mice, demonstrating a critical role for EHD3 and EHD4 in glomerular health. This is the first report of a knockout mouse model where deletion of genes with GEnC-enriched expression results in phenotypes similar to human diseases with glomerular endothelial injury as the initiating event. Results {#s2} ======= Generation and characterization of Ehd3^--/--^ mice {#s2a} --------------------------------------------------- A recombineering strategy similar to that successfully used to generate *Ehd1* and *Ehd4* null mice in our earlier studies [@pone.0017838-Rainey1], [@pone.0017838-George1], was employed to target the 5′untranslated region and the first exon of *Ehd3* such that targeted animals would not express EHD3 ([Text S1](#pone.0017838.s003){ref-type="supplementary-material"} and [Figure S1A](#pone.0017838.s001){ref-type="supplementary-material"}). Genotyping by PCR ([Figure S1B](#pone.0017838.s001){ref-type="supplementary-material"}) and Western blotting of organ lysates ([Figure S1C](#pone.0017838.s001){ref-type="supplementary-material"}) from wild-type (*Ehd3* ^+/+^), heterozygote (E*hd3* ^+/--^) and homozygous deleted (*Ehd3* ^--/--^) mice confirmed successful targeting. *Ehd3* ^--/--^ mice were born at expected Mendelian ratios, were healthy and fertile with body weights comparable to their wild-type littermates ([Figure S1D & S1E](#pone.0017838.s001){ref-type="supplementary-material"}). Developmental regulation of EHD3 expression in renal glomerular endothelium {#s2b} --------------------------------------------------------------------------- As lack of a renal phenotype in *Ehd3* ^--/--^ mice was unexpected, we first confirmed that EHD3 is indeed expressed in the wild-type kidney using *Ehd3* ^--/--^ kidney as a negative control. Consistent with a previous report [@pone.0017838-Patrakka3], our analysis of adult *Ehd3* ^+/+^ mouse kidney sections confirmed predominant EHD3 expression ([Figure 1A](#pone-0017838-g001){ref-type="fig"}) in GEnCs as confirmed by co-staining of these cells with endothelial markers tomato lectin (from *Lycopersicon esculentum)* ([Figure 1C](#pone-0017838-g001){ref-type="fig"}) and CD31 (data not shown)*;* tomato lectin recognizes N-acetyl glucosamine (GlcNAc) and poly-N-acetyllactosamine on the glomerular endothelium [@pone.0017838-Porter1]. EHD3 staining was not seen in podocytes which stained strongly with the podocyte protein nephrin, ([Figure 1E](#pone-0017838-g001){ref-type="fig"}); nephrin stained podocytes form a discrete rim surrounding EHD3-positive cells, further confirming EHD3 expression in GEnCs. The absence of staining in *Ehd3* ^--/--^ glomeruli validated the specificity of GEnC-specific EHD3 expression ([Figure 1B, D and F](#pone-0017838-g001){ref-type="fig"}). ::: {#pone-0017838-g001 .fig} 10.1371/journal.pone.0017838.g001 Figure 1 ::: {.caption} ###### EHD3 expression in the glomerular endothelium. (A--B) Three µm thick kidney sections from three month old male *Ehd3^+/+^* and *Ehd3^--/--^* mice were stained using antibodies to EHD3 and confocal images acquired as described. EHD3 staining (red) and DAPI staining of nuclei (blue) are shown. White circles demarcate glomeruli. (C--D) Triple staining of kidney sections with labeled tomato lectin (green) and EHD3 (red) and DAPI (blue) shows colocalization between EHD3 and tomato lectin staining (yellow in panel C). (E--F) Triple staining with the podocyte marker, nephrin (green), EHD3 (red) and DAPI (blue) shows nephrin staining around endothelial EHD3 staining in panel E. EHD3 staining is absent in *Ehd3^--/--^* kidney sections as expected in panels B, D and F. (G--H) Kidney sections from 2 day old *Ehd3^+/+^* mice were double stained with EHD3 (red) and lectin (green), yellow in panel H shows colocalization. The white line demarcates the kidney capsule in panel G and white arrows point to EHD3 expression in capillary loop stage glomeruli. S-shaped and comma shaped glomeruli that lack EHD3 staining are clearly visible in the nephrogenic zone in panel H. Scale bar = 10 µm. :::  ::: We next assessed if EHD3 expression in GEnCs is developmentally regulated. Nephrogenesis in mice begins by embryonic day 11 [@pone.0017838-Abrahamson1] and glomerular development proceeds through a series of nephric structures morphologically seen as vesicle, comma- and S-shaped, developing capillary loop, and maturing glomerulus stages. Newborn mice show a sub-cortical "nephrogenic zone" with easily discernable vesicular, comma and S-shaped nephric structures and an expanding cortex with capillary loop and maturing glomerular stages [@pone.0017838-Quaggin1]. Analyses of *Ehd3* ^+/+^ kidneys at postnatal day 2 showed EHD3 expression in the capillary loop stage and mature glomeruli ([Figure 1G and H](#pone-0017838-g001){ref-type="fig"}) but not in the earlier nephric stages, indicating that its expression is indeed developmentally regulated. In view of the endothelium-restricted and developmentally-regulated expression of EHD3 in the kidney, we undertook detailed analyses of *Ehd3* ^--/--^ kidneys to characterize potentially subtle glomerular pathology. However, no detectable glomerular abnormalities were observed in H&E and Periodic Acid Schiff (PAS) stained *Ehd3* ^--/--^ kidney sections ([Figure S2A--D](#pone.0017838.s002){ref-type="supplementary-material"}). Immunofluorescence microscopy showed similar patterns of staining of endothelial, mesangial and podocyte markers in both *Ehd3* ^+/+^ and *Ehd3* ^--/--^ glomeruli ([Figure S2F--M](#pone.0017838.s002){ref-type="supplementary-material"}) and transmission electron microscopic (TEM) analysis of *Ehd3* ^--/--^ kidneys from 8-month old mice showed glomerular endothelial cells with normal fenestrations and intact podocyte foot processes ([Figure S2N--O](#pone.0017838.s002){ref-type="supplementary-material"}). Finally, no proteinuria was detected in urine samples from *Ehd3* ^--/--^ mice on SDS-PAGE ([Figure S2E](#pone.0017838.s002){ref-type="supplementary-material"}) indicating that the functional integrity of the GBM was unaffected in *Ehd3* ^--/--^ kidneys. Compensatory increase in EHD4 expression in Ehd3^--/--^ glomerular endothelium {#s2c} ------------------------------------------------------------------------------ Complete absence of glomerular pathology in *Ehd3* ^--/--^ kidneys despite GEnC-enriched EHD3 expression could reflect a lack of requirement of EHD3 in glomerular development and function, or compensatory upregulation of other EHD proteins in the *Ehd3* ^--/--^ GEnCs. As no information is available on the relative expression of other EHD proteins in the kidney, we undertook immunostaining analyses to assess if other EHD family members are also expressed in the glomerular endothelium. Immunostaining for EHD1, EHD2 and EHD4 showed relatively low signals in the glomerular endothelium but markedly more prominent staining in spatially distinct regions of the kidney: EHD1 was expressed in the brush-border epithelium of proximal tubules (white arrows in [Figure 2A](#pone-0017838-g002){ref-type="fig"}); EHD2 in the endothelium and smooth muscle cells of interlobular arteries (white arrows, [Figure 2B](#pone-0017838-g002){ref-type="fig"}) and afferent arterioles (white arrow heads in [Figure 2B](#pone-0017838-g002){ref-type="fig"}); and EHD4 in the peritubular capillary endothelium (white arrows in panel [Figure 2C](#pone-0017838-g002){ref-type="fig"}). *Ehd1* ^--/--^ and *Ehd4* ^--/--^ kidney sections were used as controls for EHD1 and EHD4 staining, respectively (data not shown); staining without a primary antibody was used as a control for EHD2 staining. Thus, as EHD3 appears to be the predominant GEnC-enriched EHD protein, it is unlikely that lack of glomerular defects in *Ehd3* ^--/--^ mice is due to basal co-expression of other EHD family members. ::: {#pone-0017838-g002 .fig} 10.1371/journal.pone.0017838.g002 Figure 2 ::: {.caption} ###### Spatially distinct expression of EHD proteins in the kidney. (A--C) Kidney sections from three month old male *Ehd3^+/+^* and *Ehd3^--/--^* mice were processed and stained using antibodies to indicated EHD proteins and confocal images acquired as described in [Materials and Methods](#s4){ref-type="sec"}. EHD staining is shown in green. Scale bar = 10 µm. White arrows point to EHD1 expression in brush border epithelium (panel A), EHD2 expression in the interlobular arteries (panel B), and EHD4 expression in the peritubular capillaries (panel C); the white arrow head points to EHD2 expression in afferent arterioles (panel B). :::  ::: As previous studies have shown compensatory changes in EHD protein expression upon knockout of individual family members [@pone.0017838-Rainey1]--[@pone.0017838-George1], [@pone.0017838-Sengupta1] we examined *Ehd3* ^--/--^ kidneys for upregulation of other EHD protein expression. Immunofluorescence analyses showed lack of appreciable differences in EHD1 or EHD2 expression between *Ehd3* ^+/+^ and *Ehd3* ^--/--^ kidney sections ([Figure 3A--B and 3C--D](#pone-0017838-g003){ref-type="fig"}). Notably, while EHD4 expression in the peritubular capillaries was comparable in *Ehd3* ^+/+^ and *Ehd3* ^--/--^ kidney sections, a marked increase in EHD4 expression was observed in *Ehd3* ^--/--^ glomeruli ([Figure 3E--F](#pone-0017838-g003){ref-type="fig"}). Co-staining with labeled tomato lectin showed that EHD4 was specifically upregulated in the *Ehd3* ^--/--^ GEnCs ([Figure 3G--J](#pone-0017838-g003){ref-type="fig"}). The selective upregulation of EHD4 in *Ehd3* ^--/--^ GEnCs suggested that EHD4 might functionally compensate for *Ehd3* deletion and account for lack of glomerular pathology. ::: {#pone-0017838-g003 .fig} 10.1371/journal.pone.0017838.g003 Figure 3 ::: {.caption} ###### Compensatory increase in EHD4 in the *Ehd3^--/--^* glomerular endothelium. (A--B) Confocal images of formalin-fixed, paraffin-embedded 3 µm thick kidney sections from 5 month old male *Ehd3* ^+/*+*^ and *Ehd3^--/--^* mice were immunostained with antibodies to EHD1, (C--D) EHD2 and (E--F) EHD4. EHD protein staining is shown in green. One glomerulus in each panel is denoted by a white arrow. Scale bar = 20 µm. (G--J) Sections triple stained for EHD4 (red, panel G--J), labeled tomato lectin (green) and DAPI (blue) are shown. Yellow in panel J shows colocalization between lectin and EHD4. White circles demarcate glomeruli. Scale bar = 10 µm. :::  ::: Combined deletion of Ehd3 and Ehd4 in mice leads to renal pathology with proteinuria {#s2d} ------------------------------------------------------------------------------------ To test the hypothesis that a compensatory increase in EHD4 expression prevents the appearance of glomerular pathology in *Ehd3* ^--/--^mice, we intercrossed *Ehd3* ^--/--^ mice with the previously described *Ehd4* ^--/--^ mice [@pone.0017838-George1], [@pone.0017838-Sengupta1] to generate *Ehd3* ^--/--^ *; Ehd4* ^--/--^ double-null mice. *Ehd3* ^--/--^ *; Ehd4* ^--/--^ mice were smaller ([Figure 4A](#pone-0017838-g004){ref-type="fig"}), displayed markedly pale ([Figure 4B](#pone-0017838-g004){ref-type="fig"}) and smaller kidneys ([Figure 4C](#pone-0017838-g004){ref-type="fig"}), developed severe proteinuria ([Figure 4D](#pone-0017838-g004){ref-type="fig"}) and died between 3--24 weeks of age. As expected, both EHD3 and EHD4 staining was absent in *Ehd3* ^--/--^ *; Ehd4* ^--/--^ kidneys ([Figure 4E--H](#pone-0017838-g004){ref-type="fig"}). Interestingly, EHD2 expression was upregulated in the glomerular and peritubular capillary endothelium of *Ehd3* ^--/--^ *; Ehd4* ^--/--^ kidneys ([Figure 4K--L](#pone-0017838-g004){ref-type="fig"}). *Ehd3* ^--/--^ *; Ehd4* ^--/--^ mice did not show an upregulation of EHD1 in the glomerular or peritubular capillary endothelium ([Figure 4I--J](#pone-0017838-g004){ref-type="fig"}), but showed increased EHD1 expression at the proximal tubule brush border epithelium. Overall, these results indicate that concurrent deletion of *Ehd3* and *Ehd4* leads to renal pathology. ::: {#pone-0017838-g004 .fig} 10.1371/journal.pone.0017838.g004 Figure 4 ::: {.caption} ###### Proteinuria in *Ehd3* ^--/--^ *; Ehd4* ^--/--^ mice. \(A) Seventeen day-old *Ehd3* ^--/--^ *; Ehd4* ^--/--^ and littermate *Ehd3* ^+/--^ *; Ehd4* ^+/--^ mice were euthanized and photographed. (B) Kidneys dissected from mice shown in (A) were photographed, note the smaller and paler *Ehd3* ^--/--^ *; Ehd4* ^--/--^ kidneys. (C) Kidneys weights from 3-week old *Ehd3* ^--/--^ *; Ehd4* ^--/--^ mice and littermate controls were plotted, error bars indicate standard deviation. (D) Urine samples from 17 day old mice of indicated genotypes (lane 2 & 3) and a 3 month old *Ehd3* ^--/--^ mouse (lane 4) were run on a 7.5% SDS-PAGE and stained with Coomassie Blue. Gels were scanned following de-staining. Bovine serum albumin (BSA) was used as a positive control (lanes 5 and 6). MWM, molecular weight marker (lane 1). Lanes 1--6 were run on the same gel but were noncontiguous. (E--H) 3 µm kidney sections from mice shown in (A) were stained for EHD3 and EHD4 (green) and confocal images acquired as described. White arrows point to expression of EHD3 and EHD4 in *Ehd3* ^+/--^ *; Ehd4* ^+/--^ kidney sections (panel E and G) and absence of expression in *Ehd3* ^--/--^ *; Ehd4* ^--/--^ kidney sections (panel F and H). (I--L) Kidney sections from mice shown in (A) were stained for EHD1 (green, panels I--J) and EHD2 (green, panels K--L) and confocal images acquired as described. White arrows point to expression of EHD1 and EHD2 in kidney sections, while white arrow heads point to glomeruli. Scale bar = 20 µm. :::  ::: Ehd3^--/--^; Ehd4^--/--^ mice develop thrombotic microangiopathy {#s2e} ---------------------------------------------------------------- H&E and PAS staining of kidney sections from *Ehd3^--/--^; Ehd4^--/--^* mice showed lesions characteristic of thrombotic microangiopathy (TMA). TMA is a lesion observed in a number of human diseases including pre-eclampsia, hemolytic uremic syndrome and malignant hypertension and is defined by the primary locus of injury in the endothelium [@pone.0017838-Stillman1]. Although the majority of human cases of TMA exhibit variable numbers of thrombi, a percentage of cases do not. For example, pre-eclampsia, which has the characteristic lesions of TMA, rarely exhibits thrombi. A diagnosis of TMA is based on a set of characteristic morphologic findings irrespective of the presence of glomerular or arteriolar thrombi. These lesions include endothelial cell swelling, expansion of the subendothelial zone, duplication of the GBM, mesangial cell interposition and evidence of red cell destruction. PAS stained *Ehd3^--/--^; Ehd4^--/--^* mouse kidney sections showed characteristic glomerulomegaly, thickening and duplication of the GBM, expanded and lytic-appearing mesangium, variable degrees of mesangial interposition and abnormal capillary loops ([Figure 5A--B](#pone-0017838-g005){ref-type="fig"}). Other changes included tubular dilation and protein reabsorption droplets in proximal tubules. The glomerular changes were more easily visible with the Jones Methenamine silver stain ([Figure 5C--D](#pone-0017838-g005){ref-type="fig"}). Notably, we did not observe any thrombi in the *Ehd3^--/--^; Ehd4^--/--^* mouse kidney sections analyzed despite other hallmarks of TMA. ::: {#pone-0017838-g005 .fig} 10.1371/journal.pone.0017838.g005 Figure 5 ::: {.caption} ###### *Ehd3^--/--^; Ehd4^--/--^* mice develop thrombotic microangiopathy. (A--B) Kidney sections from 22 day old *Ehd3^--/--^; Ehd4^--/--^* mice and littermate controls were stained with PAS to visualize tubular and glomerular basement membranes. Healthy capillary loops are seen in the *Ehd3* ^+/--^; *Ehd4* ^+/--^ glomeruli (panel A) while *Ehd3^--/--^; Ehd4^--/--^* kidney sections showed endothelial swelling (e), enlarged glomeruli with segmental thickening (t), duplication of basement membranes (d) and mesangiolysis (m). Some tubular dilation is also seen (asterisks). (C--D) Jones methenamine silver (JMS) staining of kidney sections from *Ehd3^--/--^; Ehd4^--/--^* mice (panel D) and littermate controls (panel C) show duplicated and thickened glomerular basement membrane, expanded mesangium and relatively avascular glomeruli in *Ehd3^--/--^; Ehd4^--/--^* kidney sections. Thick black arrows point to glomeruli. :::  ::: In order to assess the ultrastructural alterations in glomerular endothelial cells of *Ehd3^--/--^; Ehd4^--/--^* mice, we performed TEM of kidney sections. Endothelial cells with a normal pattern of fenestrations (black arrow heads, [Figure 6A](#pone-0017838-g006){ref-type="fig"} and Ai), podocytes with intact foot processes (white arrows, [Figure 6A](#pone-0017838-g006){ref-type="fig"} and Ai) and normal GBM were identified in *Ehd3^+/--^; Ehd4^+/--^* kidneys ([Figure 6A](#pone-0017838-g006){ref-type="fig"} and Ai) and in kidney sections from *Ehd3^--/--^; Ehd4^+/--^ and Ehd3^+/--^; Ehd4^--/--^* mice (data not shown), indicating that the presence of at least one copy of *Ehd3* or *Ehd4* gene was sufficient to assure ultrastructurally normal glomerular capillaries. ::: {#pone-0017838-g006 .fig} 10.1371/journal.pone.0017838.g006 Figure 6 ::: {.caption} ###### Ultrastructural changes in *Ehd3^--/--^; Ehd4^--/--^* glomeruli. \(A) Endothelial cells with fenestrations (black arrow heads) and podocytes with intact foot processes (white arrows) are evident in the *Ehd3* ^+/--^; *Ehd4* ^+/--^ section. (Ai) Inset shows a representative higher magnification image, inset scale bar = 500 nm. (B) Abnormally thickened and lamellated basement membranes (t), swollen endothelium without fenestrations and effacement of podocytes foot processes were seen in the *Ehd3^--/--^; Ehd4^--/--^* kidney section. Multiple small "holes" presumed to be abnormal endosomes were seen in the endothelium (black arrows); these structures are clear in the higher magnification image in B. Inset scale bar = 500 nm. Scale bars = 2 µm. (C--D) Images from *Ehd3^--/--^; Ehd4^--/--^* kidneys also show additional changes including mesangiolysis (m), mesangial interposition (i), flocculent subendothelial material (f) and platelet (p) accumulation in a capillary loop. GBM, glomerular basement membrane. Scale bars = 5 µm. :::  ::: In contrast, TEM of *Ehd3^--/--^; Ehd4^--/--^* kidney sections ([Figure 6B--D](#pone-0017838-g006){ref-type="fig"}) revealed swollen glomerular endothelial cells without fenestrations, podocytes with variable segmental foot process effacement and thickening and lamellation of GBM (t, [Figure 6B](#pone-0017838-g006){ref-type="fig"}, Bi and D) with widening of subendothelial zones containing "flocculent material" (f, [Figure 6C](#pone-0017838-g006){ref-type="fig"}), lucencies within mesangial areas suggesting lysis (m, [Figure 6C](#pone-0017838-g006){ref-type="fig"}) and variable mesangial interposition (i, [Figure 6C](#pone-0017838-g006){ref-type="fig"}). Many glomerular endothelial cells contained numerous small structures that looked like abnormal endosomes or vacuoles ([Figure 6B](#pone-0017838-g006){ref-type="fig"}, Bi and [Figure 6D](#pone-0017838-g006){ref-type="fig"}, black arrows). These structures were never observed in kidneys from littermate controls. Other changes observed included electron dense deposits in mesangial areas (data not shown), vacuolation within the mesangium and platelet aggregation (p, [Figure 6D](#pone-0017838-g006){ref-type="fig"}) within capillary lumens. It is interesting to note that marked pathology was seen not only in endothelial cells but also in podocytes and mesangial cells within the glomeruli. Characteristic lesions of TMA were seen in all *Ehd3^--/--^; Ehd4^--/--^* kidneys analyzed (at ages of day 17 (d17), d22, d23, d39, d45 and d180), while the only 6 month old animal showed cellular infiltration and collagen deposition as additional renal changes (data not shown). The observed lesions were strikingly similar to that seen in human TMAs like pre-eclampsia and in mouse models with reduction of podocyte-expressed VEGF [@pone.0017838-Stillman1]--[@pone.0017838-Eremina2]. Staining and distribution of endothelial, podocytic and mesangial markers were drastically altered in Ehd3^--/--^; Ehd4^--/--^ glomeruli {#s2f} ---------------------------------------------------------------------------------------------------------------------------------------- To further characterize the lesions observed by light microscopy and TEM, we performed immunostaining for glomerular cell type markers. The staining patterns of podocyte markers nephrin and synaptopodin were severely altered in *Ehd3^--/--^; Ehd4^--/--^* glomeruli with loss of the lobulated staining pattern seen in control *Ehd3^+/--^; Ehd4^+/--^* glomeruli ([Figure 7A--D](#pone-0017838-g007){ref-type="fig"}). Tomato lectin staining of GEnCs and desmin staining of mesangial cells showed markedly altered expression patterns; *Ehd3^--/--^; Ehd4^--/--^* glomeruli had fewer lectin-positive endothelial cells ([Figure 7E--F](#pone-0017838-g007){ref-type="fig"}) and desmin expression was greatly upregulated in the glomeruli and interstitium when compared to littermate controls ([Figure 7G--H](#pone-0017838-g007){ref-type="fig"}). These studies indicate that concurrent deletion of *Ehd3* and *Ehd4* adversely affects all glomerular cell types. ::: {#pone-0017838-g007 .fig} 10.1371/journal.pone.0017838.g007 Figure 7 ::: {.caption} ###### Alteration in endothelial, podocytic and mesangial cells in *Ehd3^--/--^; Ehd4^--/--^* glomeruli. (A--D, G--H) Kidney sections from 39 day old *Ehd3* ^+/--^; *Ehd4* ^+/--^ and *Ehd3^--/--^; Ehd4^--/--^* mice were immunostained with antibodies to synaptopodin (A--B), nephrin (C--D) and desmin (G--H) as described in [Materials and Methods](#s4){ref-type="sec"}. (E--F) Endothelial cells were visualized by labeled tomato lectin staining (green) and DAPI (blue) was used to counterstain nuclei. Markedly altered staining patterns were observed in the *Ehd3^--/--^; Ehd4^--/--^* kidney sections in each case. Scale bar = 20 µm in panels A--F and 10 µm in panels G--H. White lines demarcate glomeruli in panels G--H. :::  ::: Altered localization of VEGFR2 and increased apoptosis in Ehd3^--/--^; Ehd4^--/--^ glomeruli {#s2g} -------------------------------------------------------------------------------------------- Given the similarity of glomerular lesions in *Ehd3^--/--^; Ehd4^--/--^* mice with those seen in mice with 50% reduced podocytic VEGF expression [@pone.0017838-Eremina1], [@pone.0017838-Eremina2] and in pre-eclampsia in humans [@pone.0017838-Stillman1], we analyzed if VEGF, VEGFR1 or VEGFR2 expression was altered in *Ehd3^--/--^; Ehd4^--/--^* glomeruli. While the expression pattern of VEGF and VEGFR1 in *Ehd3^--/--^; Ehd4^--/--^* glomeruli were comparable to those in *Ehd3^+/--^; Ehd4^+/--^* littermate control glomeruli (data not shown), they displayed a dramatically altered expression of VEGFR2 on endothelial cells, as seen upon co-staining for VEGFR2 and tomato lectin ([Figure 8A--F](#pone-0017838-g008){ref-type="fig"}). Tomato lectin staining was very diffuse and fewer GEnCs stained positive for VEGFR2 in *Ehd3^--/--^; Ehd4^--/--^* glomeruli; in contrast, nearly all GEnCs showed VEGFR2 staining in control *Ehd3^+/--^; Ehd4^+/--^* glomeruli. Notably, some *Ehd3^--/--^; Ehd4^--/--^* GEnCs showed dramatically more intense intracellular VEGFR2 staining compared to control glomeruli (white arrows, [Figure 8F](#pone-0017838-g008){ref-type="fig"}), suggesting potential defects in VEGFR2 trafficking. ::: {#pone-0017838-g008 .fig} 10.1371/journal.pone.0017838.g008 Figure 8 ::: {.caption} ###### Altered VEGFR2 localization and increased apoptosis in *Ehd3^--/--^*; *Ehd4^--/--^* glomeruli. (A--F) Kidney sections from *Ehd3^+/--^*; *Ehd4^+/--^* and *Ehd3^--/--^*; *Ehd4^--/--^* mice were immunostained with antibodies to VEGFR2 (red). Labeled lectin (green) was used to mark endothelial cells (B and E), while DAPI (blue) stained nuclei (A and D). Uniform low level of VEGFR2 staining in seen in lectin positive glomerular endothelial cells in the *Ehd3^+/--^*; *Ehd4^+/--^* glomeruli (panel C), while diffuse and abnormal lectin staining was seen in the *Ehd3^--/--^*; *Ehd4^--/--^* glomeruli (panel F), few lectin positive cells showed intense VEGFR2 staining (white arrows, panel F). (G--I) Kidney sections from *Ehd3^+/--^*; *Ehd4^+/--^* and *Ehd3^--/--^*; *Ehd4^--/--^* mice were subjected to TUNEL assay as described in [Materials and Methods](#s4){ref-type="sec"}. Confocal images of TUNEL staining (green) are shown (panels G--H), white arrow points to apoptotic nuclei (panel H). About 45 glomeruli each were counted in 12 kidney sections to arrive at the graph (panel I). Error bars indicate standard deviation (\*\*indicates P\<0.05 using two-tailed analysis). :::  ::: In view of the altered expression and distribution of VEGFR2 in *Ehd3^--/--^; Ehd4^--/--^* glomeruli and the known role of VEGF signaling in regulating glomerular physiology [@pone.0017838-Eremina2]--[@pone.0017838-Foster1], we assessed if *Ehd3^--/--^; Ehd4^--/--^* glomeruli showed increased apoptosis. Indeed TUNEL detected a significantly large increase in apoptotic cells in *Ehd3^--/--^; Ehd4^--/--^* glomeruli compared to *Ehd3^+/--^; Ehd4^+/--^* glomeruli ([Figure 8G--I](#pone-0017838-g008){ref-type="fig"}). Discussion {#s3} ========== Studies reported here describe the generation and characterization of a mouse model in which concurrent deletion of two endocytic recycling regulatory proteins, EHD3 and EHD4, results in renal thrombotic microangiopathy (TMA), a pathological lesion commonly seen in diverse clinical conditions. TMA is also seen in pre-eclampsia, a condition that affects about 5% of all pregnancies and is thus the most common glomerular disease in the world [@pone.0017838-Stillman1]. Irrespective of the clinical condition, the initiating event in TMA is thought to be an insult or injury to the endothelium. Though research in recent years has remarkably increased our understanding of the cell types in the glomeruli, much is still unknown about the glomerular endothelial cell. Whole-body deletion of endothelial-expressed genes in mice often results in embryonic lethality precluding analyses of endothelial cell function in the maturing glomerulus. A limited set of genes have recently been shown to be specifically expressed in the glomerular endothelium [@pone.0017838-Patrakka3], [@pone.0017838-Brunskill1], however, knockout mouse models of these genes have not been described. The lack of suitable animal models describing glomerular endothelial disease upon genetic deletion of genes specifically expressed in the glomerular endothelium has hampered research into this important cell type in the renal glomerulus. Thus, our studies reporting a gene knockout mouse model that recapitulates features of human TMA represents a major step forward that should facilitate mechanistic and pharmacological explorations of glomerular endothelial injury. Since we were unable to find unaltered endothelium in the double null mice of various ages analyzed, our results suggest that processes regulated by EHD proteins are critical to glomerular endothelial health. In view of the extensive changes in the endothelial compartment, together with endothelial-specific expression of EHD3 (and increased endothelial EHD4 in *Ehd3^--/--^* glomeruli), it is reasonable to suggest that the changes seen in podocytes and mesangial cells are likely to be secondary to endothelial defects caused by deletion of *Ehd3* and *Ehd4*. This is not unexpected as cell types within the glomerulus communicate via soluble mediators; elegant studies have shown that injury to any one cell type could compromise the functional integrity of other cell types and the glomerulus as a whole [@pone.0017838-Eremina3], [@pone.0017838-Eremina4]--[@pone.0017838-Lindahl1]. For example, soluble factors such as platelet derived growth factor (PDGF) BB elaborated by healthy endothelial cells have been shown to be critical to mesangial health [@pone.0017838-Eremina4] and this might explain the mesangial changes seen in *Ehd3^--/--^; Ehd4^--/--^* mice. It is also possible that the observed alterations in VEGFR2 expression in glomerular endothelial cells ([Figure 8](#pone-0017838-g008){ref-type="fig"}) might affect their health and secondarily contribute to the podocyte pathology seen. Future studies to delete *Ehd3* and *Ehd4* in a cell type-specific manner (endothelium or podocytes) using specific Cre-recombinase expressing transgenic mouse lines will help unequivocally distinguish the direct versus indirect effects of endothelial EHD3 and EHD4 loss on the podocyte. Towards this end, we have generated double *Ehd3 ^fl/fl^*; *Ehd4 ^fl/fl^* mice that should allow such experiments to be carried out in the future. The TMA observed in *Ehd3^--/--^; Ehd4^--/--^* glomeruli bear a striking resemblance to that seen in mouse models with reduced podocyte VEGF expression and a recent mouse model with inducible whole-body deletion of VEGFR2 [@pone.0017838-Sison1]. This is especially significant since our genetic manipulations were not aimed at altering the levels of VEGF or its receptors, VEGFR1 and VEGFR2 expressed in the glomerular endothelium. The marked alterations in VEGFR2 expression in glomerular endothelial cells and the increased apoptosis seen in *Ehd3^--/--^; Ehd4^--/--^*glomeruli, however point to a possible role for EHD proteins in the endocytic traffic of VEGFR2 and subsequent regulation of VEGF mediated signaling in the glomeruli. It has been demonstrated that at steady state VEGFR2 localizes to the cell surface and intracellular vesicles presumed to be sorting endosomes [@pone.0017838-Scott1]. In response to VEGF addition, recycling of intracellular VEGFR2 to the cell surface increases to allow higher VEGF binding and signaling [@pone.0017838-Scott1], [@pone.0017838-Gampel1]. Since EHD3 [@pone.0017838-Naslavsky2] and EHD4 [@pone.0017838-George2] are both known to regulate trafficking out of sorting endosomes, and in view of the appearance of abnormal vesicular structures in endothelial cells as seen by TEM, it is possible that altered recycling of VEGFR2 out of sorting endosomes in *Ehd3^--/--^; Ehd4^--/--^* glomerular endothelial cells might impair VEGF signaling resulting in the phenotypes we observe. It is noteworthy that no differences in VEGFR1 staining were noticed between *Ehd3^--/--^; Ehd4^--/--^* and control glomeruli. This is not surprising as VEGFR1 trafficking has not been shown to be regulated by endocytic regulators in the early or recycling pathway; hence we would not expect *Ehd3* and *Ehd4* deletion to affect trafficking of VEGFR1. While we have not yet found a relevant endothelial cell system to analyze whether VEGFR2 distribution and expression pattern in *Ehd3^--/--^; Ehd4^--/--^* mice might reflect its altered endocytic traffic, studies of individual siRNA-mediated knock-down of EHD3 [@pone.0017838-Naslavsky2] or EHD4 [@pone.0017838-George2] in HeLa cells indicate that they control traffic of transferrin receptor out of sorting endosomes to a recycling compartment and double knock-down of EHD3 and EHD4 has an identical effect (data not shown). As VEGFR2 has been shown to reside in a sorting endosomal compartment and to recycle to the cell surface upon VEGF stimulation [@pone.0017838-Scott1], [@pone.0017838-Gampel1], absence of EHD3 and EHD4 in glomerular endothelial cells might be expected to impose a block to its recycling and thus disrupt its signaling. Given the roles of EHD proteins in regulating endocytic recycling, it is likely that loss of EHD3 and EHD4 causes aberrant trafficking of key receptors such as VEGFR2, that are critical for maintenance of glomerular endothelial cell function and endothelial-cell dependent inter-cellular signaling critical for glomerular health. Thus, further studies using the experimental model described here should help link the basic cell biological processes of endocytic traffic, sorting and recycling of such receptors to the pathogenesis of endothelial injury which is an integral part of a number diseases of substantial human health importance. Materials and Methods {#s4} ===================== Breeding and maintenance of mice colonies {#s4a} ----------------------------------------- Male and female *Ehd3* ^+/--^ mice were crossed to generate *Ehd3* ^+/+^, *Ehd3* ^+/--^ and *Ehd3* ^--/--^ mice. To generate *Ehd3* ^--/--^; *Ehd4* ^--/--^ mice, the previously described *Ehd4* ^--/--^ mice [@pone.0017838-George1] were mated to *Ehd3* ^--/--^ mice to generate *Ehd3* ^+/--^; *Ehd4* ^+/--^ mice and these were intercrossed to give rise to *Ehd3* ^--/--^; *Ehd4* ^--/--^ mice. Animals were genotyped by tail PCR and ink-tattooed on toes for identification. Primer sequences for PCR genotyping are available on request. All experiments involving animals were approved by the University of Nebraska Medical Center Institutional Animal Care and Use Committee and carried out under the approved IACUC protocol number 07-061-FC12. All animals were treated humanely in accordance with institutional guidelines and that of the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. Tissue staining {#s4b} --------------- Formalin fixed, paraffin embedded, 3 µm kidney sections from mice of the indicated genotype and ages were processed as described previously [@pone.0017838-George1] and H&E, PAS and JMS staining was performed using standard procedures by the Tissue sciences and Pathology and Microbiology core facilities. Immunostaining was performed as described previously [@pone.0017838-George1]. The following antibodies were purchased commercially and used: CD31 (1∶50), WT-1 (1∶50), nephrin (1∶50), synaptopodin (1∶50), desmin (1∶50) and VEGFR2 (1∶50). Fluorescein labeled tomato lectin (from *L.esculentum*) was commercially purchased (Vector labs, FL-1171). For EHD protein staining, formalin fixed, 3 µm kidney sections were deparaffinized and immunostained as described previously. Polyclonal rabbit anti-EHD3 primary antibody was used at a 1∶200 dilution, while polyclonal rabbit anti-EHD1, EHD2 and EHD4 primary antibodies were used at a 1∶50 dilution in PBS/5% fetal bovine serum and a goat anti-rabbit Alexa Fluor 488 (or 594, Invitrogen) secondary antibody was used at 1∶200 dilution in PBS. The slides were mounted in Vectashield (Vector Labs) containing DAPI. Confocal images were acquired with a LSM510 fluorescence confocal microscope (Carl Zeiss, Thornwood, NY). Analyses of proteinuria {#s4c} ----------------------- Urine from mice of indicated genotypes was assayed for proteinuria using a dip-stick method with Albustix (Siemens). Urine samples were boiled for 5 minutes in equal volumes of 2X sample buffer and 0.8 to 2 µl was loaded on a 7.5% SDS-PAGE gel. Bovine serum albumin (BSA) was used as a positive control. Following separation of proteins, the gel was stained using Coomassie Brilliant Blue and images were scanned following de-staining in water. Electron Microscopy {#s4d} ------------------- Kidneys were fixed in a 2.0% paraformaldehyde/2.5% glutaraldehyde phosphate buffered fixative, secondarily fixed with 1% osmium tetroxide, dehydrated using a graded acetone series and infiltrated using Polybed 812 epoxy resin. Blocks were polymerized at 60°C. Sections were thinned at 70 nm and stained using uranyl acetate and lead stains followed by scanning under a JEOL 1230 transmission electron microscope. Digital images were taken using a KeenView high-resolution camera and Soft Imaging Solutions AnalySIS ITEM digital software. TUNEL {#s4e} ----- For TUNEL assay, 3 µm neutral buffered formalin fixed kidney sections from d17, d23 and d45 *Ehd3^--/--^*; *Ehd4^--/--^* mice and littermate controls were deparaffinized and following antigen retrieval, an In Situ Cell Death Detection kit, POD (Roche) was used for TUNEL assay which was performed following the manufacturer\'s instructions. Appropriate negative and positive controls were included in each experiment. Confocal images of TUNEL were acquired with a LSM510 fluorescence confocal microscope (Carl Zeiss, Thornwood, NY) under either a 63X objective. TUNEL positive cells were counted from 12 sections (1 section/kidney, 2 kidneys/mice, 3 mice/genotype) and plotted. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Generation and characterization of** ***Ehd3^--/--^*** **mice.** (A) Partial restriction map of the murine *Ehd3* locus, the targeting vector and the mutated *Ehd3* loci is depicted. *LoxP* sequences were inserted to flank the first exon such that it could be deleted by Cre/*loxP*-mediated recombination. Black rectangles represent exons; black and grey triangles represent *loxP* and *FRT* sequences, respectively. RI, *EcoR*I; H, HindIII. (B) Samples of tail DNA from 10 day old mice were genotyped by PCR. Three primers were used in a single duplex PCR reaction to amplify the WT *Ehd3* allele (377 bp) and the deleted allele (488 bp), and the products were separated by agarose gel electrophoresis to determine one of three genotypes of mice carrying various *Ehd3* alleles. (C) Western blotting of organ lysates from *Ehd3* mice. Fifty µg aliquots of organ lysates from three month-old *Ehd3* wild-type (*Ehd3* ^+/+^), heterozygote (E*hd3* ^+/--^) and null (*Ehd3* ^--/--^) male mice were subjected to Western blotting with antisera raised against human EHD proteins as described under [Materials and Methods](#s4){ref-type="sec"}. The membrane was serially probed beginning with EHD3, followed by EHD1 and EHD4 and then EHD2 antibodies. The \* denotes bleed-through from the previous blot. In the kidney and brain lysates, the anti-EHD3 antibody recognizes smaller sized products that may represent tissue-specific alternate spliced products of EHD3 lacking the first exon or they may be non-specific bands detected by the antibody. MWM, Molecular weight marker. (D and E) Quantitative growth curves of female (D) and male (E) littermate *Ehd3* mice (numbers in parentheses indicate number of mice of each genotype weighed). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Lack of glomerular phenotypes in** ***Ehd3^--/--^*** **mice.** (A--D) Formalin-fixed, paraffin- embedded, three µm thick kidney sections from 3--5 month old male *Ehd3^+/+^* and *Ehd3^--/--^* mice were stained with H&E and PAS. (F--M) Immunostaining was performed on kidney sections using antibodies to synaptopodin (F--G), nephrin (H--I), WT-1(J--K) and desmin (L--M) as described in [Materials and Methods](#s4){ref-type="sec"}. Scale bar = 10 µm. (E) Two µl of boiled urine samples from 8 month old mice of indicated genotypes were run on a 7.5% SDS-PAGE and stained using Coomassie Blue. Gels were scanned following de-staining. Bovine serum albumin (BSA) was used as a positive control (lanes 2 and 3). MWM  =  molecular weight marker (lane 1). Lanes 1--6 were run on the same gel but were noncontiguous. (N--O) Electron micrographs of glomeruli from 8 month old *Ehd3^+/+^* and *Ehd3^--/--^* mice are shown. Endothelium with fenestrations (black arrows) and podocytes with intact foot processes are seen in *Ehd3^+/+^* and *Ehd3^--/--^* mice. Scale bar = 2 µm. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S1 ::: {.caption} ###### **Supporting Information Materials and Methods** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank the UNMC Comparative Medicine Core Facility for animal husbandry and veterinary care; Maureen Harmon of Tissue sciences core facility and Janice Taylor and James Talaska of the Confocal Core Facility at the University of Nebraska Medical Center (supported by the Nebraska Research Initiative and the Eppley Cancer Center Core Grant from the NCI) for technical assistance; and Dr. Babu Padanilam, Dr. Dale Abrahamson and the Band Lab members for helpful discussions. **Competing Interests:**Ehd3 null mice and Ehd4 null mice were generated while MG, MAR, MN, GGY, RMG, CBG, VB and HB were at Evanston Northwestern Healthcare Research Institute, Evanston IL. **Funding:**NIH Grants CA105489, CA099163, CA87986, CA116552 (HB), CA096844 (VB); Department of Defense Breast Cancer Research Grant DAMD W81XWH-07-1-0351,W81XWH-11-1-0171 (VB), W81XWH-11-1-0166 (HB), W81 XWH-10-1-0740 (MN), W81XWH-08-1-0621 (MG); CoBRE Pilot funding (MG) from NIH grant 2P20 RR018788 from the CoBRE Program of the National Center for Research Resources (NCRR); UNMC-Eppley Cancer Center Pilot Grant (MN). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MG. Performed the experiments: MG MAR CBG. Analyzed the data: MG KWF SCS. Contributed reagents/materials/analysis tools: KWF. Wrote the paper: MG. Conceived and designed Ehd3 null mouse: HB VB MN GGY. Secured funding: HB VB. Edited manuscript: MAR MN KWF CBG VB SCS HB. Technical assistance: LLW RMG MSH.

PubMed Central

2024-06-05T04:04:19.275359

2011-3-9

PMC3052386

Introduction {#s1} ============ Upon entering the cell, glucose is phosphorylated to glucose-6-phosphate (G-6-P), which is used catabolically in glycolysis, or anabolically in glycogen synthesis and lipid synthesis via the pentose phosphate shunt. In both cases, this first step is catalyzed by hexokinases (HKs), which comprise a family of four isoforms. HKI and HKII are the most abundant isoforms, with HKI ("the brain HK") ubiquitous in most tissues but especially brain and red blood cells [@pone.0017674-Lowry1], [@pone.0017674-Purich1] where glycolysis plays a critical role in energy production. In contrast, HKII ("the muscle HK") is found primarily in insulin-sensitive tissues such as adipocytes and adult skeletal and cardiac muscle, where it accounts for 80% of total HK activity [@pone.0017674-Mandarino1]. At birth and after weaning, the latter tissues switch from expression of GLUT1 and HKI, to GLUT4 and HKII. The expression of GLUT4 and HKII coincides with the development of insulin sensitivity as muscle switches from a straight carbohydrate to a mixed fat-carbohydrate diet [@pone.0017674-Postic1]. In adult muscle, fatty acids as well as glucose and glycogen are available as substrates to support oxidative metabolism [@pone.0017674-Depre1]. G-6-P facilitates glycogen synthesis by reciprocally activating glycogen synthase (GS) and inhibiting glycogen phosphorylase (GP) [@pone.0017674-Schulz1], [@pone.0017674-Syed1], and possibly by stimulating translocation of HK from mitochondria to the cytoplasm. In liver cells where glycogen synthesis is driven by glucokinase (GK or HKIV) rather than HKI, G-6-P stimulates glycogen synthesis by causing a redistribution of GK and GS to the cell periphery [@pone.0017674-Ferrer1]. These data support the idea of compartmentalized metabolic channeling, with HKI feeding glycolysis, and HKII and HKIV feeding glycogen synthesis. The idea that differing subcellular locations of HKI and HKII are important was postulated by Wilson when he stated that "the Type I isozyme bound to actively phosphorylating mitochondria facilitates introduction of glucose into glycolysis, with the final stages of glucose metabolism occurring in the mitochondria. In contrast, Type II and to some extent Type III isozymes serve primarily anabolic function to provide G-6-P for glycogen synthesis or lipid synthesis via the pentose phosphate pathway" (see review [@pone.0017674-Wilson1]). However, it was not clear at that time how independent was the anabolic function of the Type II isozyme and binding to mitochondria were related. Since then, Wilson and others [@pone.0017674-Galluzzi1], [@pone.0017674-GimenezCassina1], [@pone.0017674-Gottlob1], [@pone.0017674-Majewski1], [@pone.0017674-Miyamoto1], [@pone.0017674-Pastorino1], [@pone.0017674-Rose1], [@pone.0017674-Skaff1], [@pone.0017674-Sui1] have shown that the interaction of HKs with mitochondria is not static, but is regulated by factors such as glucose, G-6-P and kinases such as Akt and GSK-3. Thus, a picture is emerging that HKII may play a dual role: channeling G-6-P into the glycogen and the pentose phosphate pathways when localized in the cytoplasm, and preferentially shuttling G-6-P to glycolysis and oxidative phosphorylation when bound to mitochondria [@pone.0017674-Jurczak1]. In contrast, HKI generally facilitates glycolysis; although under some specific non-physiological conditions may contribute to glycogen synthesis [@pone.0017674-Cifuentes1]. HKI and HKII are inhibited allosterically by their product, G-6-P, and this sensitivity to G-6-P decreases when HKs are bound to mitochondria [@pone.0017674-AzoulayZohar1], [@pone.0017674-deCerqueiraCesar1]. Physiological levels of orthophosphate (Pi) counter the G-6-P inhibition of HKI [@pone.0017674-Purich1], [@pone.0017674-Ellison1], [@pone.0017674-Fang1], but not HKII. In fact, Pi may cause further HKII inhibition. Based on these observations, Wilson [@pone.0017674-Wilson1] suggested that "reciprocal changes in intracellular levels of G-6-P and Pi are closely associated with cellular energy status, and that the response of HKI to these effectors adapts it for catalytic function, introducing glucose into glycolytic metabolism. In contrast, HKII serves primarily anabolic functions." In the present study, we have expressed HKI and HKII tagged with YFP in Chinese Hamster Ovary (CHO) cells to track their subcellular location in real time and their mobilization in response to substrates. Concomitantly, we measured changes in intracellular glucose using a genetically-encoded intracellular glucose biosensor, FLIPglu-600 µM, which undergoes changes in FRET upon binding glucose [@pone.0017674-Fehr1] and assessed glycogen formation, also in live cells, using a glycogen targeting protein, PTG, tagged with GFP. In this study our goal was three-fold: 1) To further investigate the roles of HKI and HKII in directing the metabolic fate of glucose towards catabolic versus anabolic uses; 2) To assess how the metabolic roles of HKI and HKII are related to their subcellular localization; 3) To determine the signaling pathways regulating subcellular distributions of HKI and HKII. Our findings support the hypothesis that in response to changes in glucose, subcellular translocation of HKII dynamically directs the metabolic fate of glucose between catabolic (glycolysis) and anabolic (glycogen synthesis and pentose phosphate shunt) uses, while HKI remains associated with mitochondria to promote glycolysis. Factors such as G-6-P and Akt play a central role in the regulation of HKII activity and localization in response to changes in glucose. Results {#s2} ======= HK interaction with mitochondria and translocation to the cytosol {#s2a} ----------------------------------------------------------------- To test the hypothesis that HKI interacts preferentially with mitochondria to facilitate entry of G-6-P into the glycolytic pathway, while HKII translocates into the cytosol to channel G-6-P into the glycogen formation pathway, we carried out two types of optical imaging experiments in CHO cells and HEK293 cells. In the first, we linked YFP to the C terminus of HKI or HKII (HKI-YFP and HKII-YFP) to track subcellular location. In the second, we used a genetically-encoded low affinity intracellular glucose sensor Flipglu-600 µM to monitor intracellular glucose metabolism. In [Fig. 1A](#pone-0017674-g001){ref-type="fig"}, data obtained with HKI-YFP demonstrates that HKI is bound to mitochondria with little or no fluorescence found in the cytosol. Co-labeling with HKI-CFP and MitoTracker-YFP ([Fig. S1](#pone.0017674.s001){ref-type="supplementary-material"}) supports the interpretation that HKI, even when overexpressed, is predominantly bound to mitochondria. In contrast, the distribution of HKII-YFP showed both mitochondrial labeling and diffuse fluorescence in the cytoplasm ([Fig. 1C](#pone-0017674-g001){ref-type="fig"}). To further characterize the interaction of HKs with mitochondria, we cotransfected HKI-YFP together with wild-type (wt) HKII (ratio 1∶3), and conversely, HKII-YFP together with wt HKI (ratio 1∶3). HKII did not displace HKI-YFP from mitochondria ([Fig. 1B](#pone-0017674-g001){ref-type="fig"}), while HKII-YFP was no longer associated with mitochondria in the presence of excess HKI ([Fig. 1D](#pone-0017674-g001){ref-type="fig"}). These data suggest that HKI has a strong affinity for a mitochondria-binding site which cannot be easily displaced by HKII. In contrast, HKII binding to mitochondria was much weaker and readily displaced by excess HKI. ::: {#pone-0017674-g001 .fig} 10.1371/journal.pone.0017674.g001 Figure 1 ::: {.caption} ###### Subcellular distribution of HKI and HKII linked to YFP in CHO cells. In the left panel HKI-YFP is expressed alone (A) or with HKII (B) (ratio 1∶3). In this case HKI-YFP is bound to mitochondria and is not displaced by overexpressed HKII. In the right panel HKII-YFP is expressed alone and shows a mixed distribution between mitochondria and cytosol (C). In this case overexpression of HKI with HKII-YFP (ratio 1∶3) prevents HKII-YFP interaction with mitochondria (D). Similar subcellular distribution of HKI-YFP and HKII-YFP was observed in HEK293 cells. :::  ::: Effects of hexokinase overexpression in CHO cells transfected with GLUT1 {#s2b} ------------------------------------------------------------------------ Next, we investigated how HKI and HKII affect glucose metabolism by expressing a genetically encoded low affinity intracellular glucose sensor Flipglu-600 µM in native CHO cells. CHO cells express low levels of glucose transporters (GLUT), however, they have low rates of glucose uptake and metabolism compared to muscle cells [@pone.0017674-John1]. Thus we overexpressed GLUT1 or GLUT4, to increase glucose uptake ([Figs S2](#pone.0017674.s002){ref-type="supplementary-material"} and [S3](#pone.0017674.s003){ref-type="supplementary-material"}). GLUT overexpression both improved FRET signals and, facilitated glycogen synthesis, which is activated by high intracellular glucose. To characterize intracellular glucose handling, we used the GLUT inhibitor Cyto B (20 µM) to block GLUT-sensitive transmembrane glucose transport. We had previously shown that in native CHO cells, exposure to Cyto B prior to the addition of glucose fully blocked subsequent glucose uptake [@pone.0017674-John1]. In contrast, Cyto B had almost no effect on the rate of glucose clearance following glucose removal ([Fig. S3A](#pone.0017674.s003){ref-type="supplementary-material"} [@pone.0017674-Lowry1] vs. [Fig. S3A](#pone.0017674.s003){ref-type="supplementary-material"} [@pone.0017674-Purich1]). The latter observation suggests that glucose metabolism, rather than glucose efflux, dominates glucose clearance in native CHO cells with a slow rate of glucose transport (τd = 33 s). On the other hand, in CHO cells overexpressing GLUT1 ([Fig. S3B](#pone.0017674.s003){ref-type="supplementary-material"} [@pone.0017674-Postic1]), with a high rate of glucose transport (uptake half-time 3.5 s), Cyto B markedly slowed the rate of glucose clearance upon glucose removal, indicating that in CHO cells overexpressing GLUT1, glucose efflux, rather than glycolysis, is the dominant factor regulating glucose clearance. Together, these observations indicate that in CHO cells overexpressing GLUT1 or GLUT4, glucose clearance is dominated by glucose efflux via GLUTs rather than glucose utilization by glycolysis. Thus, to study the metabolic fate of glucose in CHO cells overexpressing GLUTs, subsequent experiments were performed in the presence of Cyto B to block glucose efflux following glucose removal. Data in [Fig. 2](#pone-0017674-g002){ref-type="fig"}, obtained with and without HK overexpression, show how expression of HKs alters the rate of glucose clearance during glucose removal in the presence of Cyto B. Panels B and D show that HKI overexpression decreased the average half-time of glucose clearance (t~1/2~) from 114±12 s to 72±7 s (n = 9) ([Fig. 2D](#pone-0017674-g002){ref-type="fig"}). In contrast, HKII overexpression (panels C and D) prolonged the half-time to 217+/−20 s (n = 9). Panels E and F show the distribution of overexpressed HKI or HKII in CHO cells. These data clearly demonstrate that increased HKI activity, which is primarily associated with mitochondria, increases glucose metabolism, while increased HKII activity, which is found at least in part in the cytosol, slows glucose metabolism. ::: {#pone-0017674-g002 .fig} 10.1371/journal.pone.0017674.g002 Figure 2 ::: {.caption} ###### Effects of HKI and HKII on glucose utilization in CHO cells. These data illustrates how overexpression of HKI (Panel B) and HKII (Panel C) affects glucose clearance measured in the presence of CytoB. Comparison of A and B indicates that HKI increases the rate of glucose clearance and thus metabolism, while comparison of A and C demonstrates that HKII has the opposite effect and decreases this rate. Panel E and F show images similar to those in [Fig. 1](#pone-0017674-g001){ref-type="fig"}, which illustrate once more the interaction of HKI-YFP with mitochondria and the differential distribution of HKII-YFP, these differential distribution can be correlated with the glucose clearance data in B and C, respectively. The bar graph in D quantifies these data. :::  ::: To further assess the role of endogenous HKI and HKII in glucose clearance we transfected CHO cells with 30 to 40 nM of HKI or HKII siRNA. Knockdown of either HKI or HKII caused a decrease in the rate of glucose clearance ([Fig S4](#pone.0017674.s004){ref-type="supplementary-material"}). These results are consistent with our finding from RT-PCR analysis that the levels of endogenous HKI and HKII are very similar in CHO cells. Glycogen-derived G-6-P mediates glucose-dependent HK inhibition {#s2c} --------------------------------------------------------------- We hypothesized that when HKII is overexpressed, the prolonged t~1/2~ of glucose clearance following a glucose pulse is related to glycogen synthesis, which occurs during the glucose pulse. Thus, when glucose is removed at the end of the pulse, the cell mobilizes the newly synthesized glycogen for glycolysis, rather than metabolizing the residual intracellular glucose. This explanation is plausible only if glycogen mobilization also inhibits HK enzymatic activity, such that the conversion of residual intracellular glucose to G-6-P is suppressed. Since glycogenolysis involves production of G-6-P, which inhibits HKs [@pone.0017674-AzoulayZohar1], [@pone.0017674-deCerqueiraCesar1], we postulated that elevated G-6-P levels during glycogenolysis might be responsible for inhibiting HK activity, so that intracellular glucose clearance is delayed. Consistent with this hypothesis, [Fig. 3A](#pone-0017674-g003){ref-type="fig"} shows after a 30 s exposure to glucose ([Fig. 3A](#pone-0017674-g003){ref-type="fig"} [@pone.0017674-Purich1]), glucose clearance upon removing extracellular glucose (in the presence of Cyto B) was rapid, and occurred at a rate similar to that observed in the absence of GLUT overexpression. However, after a 75 s exposure to 10 mM glucose ([Fig. 3A](#pone-0017674-g003){ref-type="fig"} [@pone.0017674-Lowry1]), to allow more time for glycogen synthesis, glucose clearance by metabolism was delayed, occurring very slow initially, followed by rapid clearance similar to that after the short exposure to glucose. The t~1/2~ increased on average from 88±14 s after a 30 s exposure to glucose, to 285±29 s (n = 15) after a 75 s exposure. In contrast, exposure to 1 mM extracellular glucose for several minutes, which increased intracellular glucose level by less than half compared to 10 mM glucose, did not cause a delay in glucose metabolism ([Fig. 3B](#pone-0017674-g003){ref-type="fig"} [@pone.0017674-Postic1] and [@pone.0017674-Depre1]). ::: {#pone-0017674-g003 .fig} 10.1371/journal.pone.0017674.g003 Figure 3 ::: {.caption} ###### The time of exposure to extracellular glucose controls the rate of intracellular glucose clearance. Panel A illustrates how exposure to 10 mM extracellular glucose for 75 s (1), 30 s (2) and 50 s (3) affects intracellular glucose clearance. In the 3 cases CytoB was applied for 15 s prior to removal of extracellular glucose. The 3 rates are compared in the right hand side panel and show that the final rate of glucose clearance is similar in the 3 cases and that long exposure to extracellular glucose delays the clearance. In B a comparison of data obtained with 10 mM and 1 mM extracellular glucose show that application for up to two minutes of 1 mM glucose had no effect on the rate of intracellular glucose clearance, demonstrating that intracellular glucose must reach a threshold to induce this effect. The right hand side panel shows again that the final rate of glucose clearance measured in the presence of 10 mM glucose is similar to the maximum rate measure with 1 mM and the effect of high glucose is thus to delay clearance. :::  ::: To test this hypothesis further, we imaged glycogen stores directly by using probes linking either mCherry or GFP to PTG and G~L~, both of which are part of the family of glycogen targeting subunits of PP-1 [@pone.0017674-Lerin1]. CHO cells were transfected with one of these probes, either with or without GLUT1 co-transfection. One day after transfection, cells were incubated for 2 to 3 hours in the absence of glucose to deplete glycogen stores. In the absence of GLUT1 overexpression, some CHO cells exhibited dim homogenous fluorescence, while others had a few small and bright punctuate deposits ([Fig. 4A](#pone-0017674-g004){ref-type="fig"}). After re-addition of glucose (10 mM) for 30--60 min, the number, size and brightness of small glycogen deposits began to increase. This phenomenon intensified over 24 hrs, until the deposits fused and partially filled the cell ([Fig. 4C](#pone-0017674-g004){ref-type="fig"}). After a two hour exposure, removal of glucose resulted in a rapid disappearance of glycogen deposits ([Fig. 4B](#pone-0017674-g004){ref-type="fig"}). With overexpression of GLUT1, the rate of appearance of glycogen dramatically increased, such that small deposits were observed within a few minutes of exposure to 10 mM glucose ([Fig. 4D](#pone-0017674-g004){ref-type="fig"}). Thus, in CHO cells overexpressing GLUT1, the rates of glycogen synthesis and depletion are rapid enough to plausibly support our hypothesis that relatively short glucose pulses could induce enough glycogen synthesis to progressively delay the t~1/2~ of glucose clearance following glucose removal. ::: {#pone-0017674-g004 .fig} 10.1371/journal.pone.0017674.g004 Figure 4 ::: {.caption} ###### Effects of GLUT1 overexpression on glycogen synthesis and breakdown in CHO cells. The rate of glycogen formation without (A) and with (D) overexpression of GLUT1 and the rate of glycogen degradation (B) was studied using a protein targeted to glycogen linked to YFP (PTG-YFP). In Panel A and D the cells were incubated in the absence of glucose for 2 to 3 hours prior to beginning cell imaging. This incubation in the absence of glucose was carried out to deplete preformed glycogen. Without GLUT1 overexpression glycogen build up was slow, occurring over several hours. This build up proceeded for up to 24 h filling almost completely the cell in some cases (C). With GLUT1 overexpression the rate of glycogen synthesis increased and glycogen deposits could be observed 5 min after exposure to 10 mM glucose outside. It is interesting to note that glycogen deposition occurred at least initially near the nucleus where mitochondria aggregate, suggesting that G-6-P generated by glycogen degradation may be directly fed on to mitochondria. :::  ::: The effects of PTG overexpression on the t~1/2~ of glucose clearance following glucose removal are shown in [Fig. 5](#pone-0017674-g005){ref-type="fig"}. In CHO cells transfected with PTG alone, PTG had no effect on the t~1/2~ of glucose clearance ([Fig. 5C](#pone-0017674-g005){ref-type="fig"}), even after prolonged exposure to glucose, which we attributed to intracellular glucose never reaching an adequate level to induce glycogen synthesis (as indicated by a low FRET ratio). However, in CHO cells transfected with both GLUT1 and PTG, the lag phase was pronounced after even a brief exposure to 10 mM glucose ([Fig. 5B](#pone-0017674-g005){ref-type="fig"}). Thus, PTG, which stimulates glycogen synthesis by facilitating the interaction of the regulatory enzyme PP-1 (protein phosphatase-1) with GS, GP and phosphorylase kinase, leads to greater glycogen synthesis during brief exposures to 10 mM glucose. According to our hypothesis, this elevates G-6-P levels for longer after glucose removal and delays glucose clearance. ::: {#pone-0017674-g005 .fig} 10.1371/journal.pone.0017674.g005 Figure 5 ::: {.caption} ###### Overexpression of PTG delays glucose clearance. Panels A and B obtained with CHO cells overexpressing GLUT1 illustrate how exogenous PTG delays glucose clearance. In both cases the cells were exposed to extracellular glucose for 50 s. Comparison of data in Panel B (overexpressed GLUT1) and C (no GLUT1 overexpression) illustrate the lack of effect of exogenous PTG in the absence of GLUT1 overexpression, supporting the hypothesis that GLUT1 is necessary for rapid build up of the glycogen store and inhibition of glucose utilization. Compilation of traces in D illustrates once more that the final rate of glucose clearance is similar in all cases and that increased glycogen synthesis only delays glucose utilization. :::  ::: In summary, these data are consistent with removal of glucose causing a switch from glycogen synthesis to glycogen breakdown, thereby increasing G-6-P which blocks the utilization of glucose by mitochondria-bound HK, in effect diverting G-6-P to glycolysis. Regulation of the metabolic fate of glucose by the subcellular distribution of HK {#s2d} --------------------------------------------------------------------------------- The observations so far indicate that HK1 overexpression affects catabolism, accelerating the t~1/2~ of glucose clearance, whereas HKII overexpression is anabolic, delaying the t~1/2~ by promoting glycogen synthesis, with subsequent inhibition of HK by glycogenolysis when extracellular glucose is removed. To determine how these metabolic roles are related to the subcellular localization of HKI and HKII, we studied the effects of extracellular glucose on subcellular HK distribution in CHO cells overexpressing HKI or HKII linked to YFP. Whether the cells were incubated in the presence or absence of glucose (10 mM), HKI-YFP always remained associated with mitochondria, and addition or removal of glucose had no effect on its location (n = 10) ([Fig. 6A](#pone-0017674-g006){ref-type="fig"}). In contrast, the subcellular distribution of HKII-YFP was sensitive to extracellular glucose. In the presence of glucose, a large fraction of HKII was bound to mitochondria, and upon removal of extracellular glucose, HKII rapidly translocated to the cytosol with a time constant averaging 8.5 min (n = 12) ([Fig. 6B and 6C1](#pone-0017674-g006){ref-type="fig"}). This effect was fully reversible, such that upon re-addition of glucose, HKII re-associated with mitochondria with a time constant of 15.5 min (n = 4) ([Fig. 6C2](#pone-0017674-g006){ref-type="fig"}). These data support the hypothesis that HKI binds strongly to mitochondria, while the distribution of HKII between cytoplasm and mitochondria is labile, dynamically regulated by glucose availability. ::: {#pone-0017674-g006 .fig} 10.1371/journal.pone.0017674.g006 Figure 6 ::: {.caption} ###### Removal of glucose causes translocation of HKII, but not HKI, into the cytosol. Images in Panel A obtained with HKI-YFP expressed in CHO cells show that removal of glucose did not affect HKI interaction with mitochondria. In contrast, HKII begun to dissociate from mitochondria 5 min after glucose removal (B). Panel C shows the rates measured as ratio of fluorescence intensity obtained from intracellular domains without and with mitochondria. In almost all cases the region without mitochondria was selected at the cell periphery and that with mitochondria was selected near the nucleus. C1 is a comparison of the rates of HKI and HKII dissociation in response to glucose removal. C2 shows the rate of HKII reassociation with mitochondria after glucose readdition. :::  ::: To evaluate the functional consequences of HKII redistribution on glucose metabolism, cells were subjected to a 20 to 30 min pre-incubation in the absence of glucose to maximize dissociation of HKII from mitochondria ([Fig. 6B](#pone-0017674-g006){ref-type="fig"}), and compared to cells with glucose present throughout (in which HKII remained mostly associated with mitochondria). Preincubation in zero glucose prolonged the t~1/2~ of glucose clearance from 102±11 s to 210±19 s (n = 7) ([Fig. 7C2 and C4](#pone-0017674-g007){ref-type="fig"}). This finding demonstrates that the subcellular distribution of HKII strongly influences its ability to promote anabolic use of glucose for glycogen synthesis. ::: {#pone-0017674-g007 .fig} 10.1371/journal.pone.0017674.g007 Figure 7 ::: {.caption} ###### Constitutively active Akt prevents the effect of glucose removal on HKII dissociation from mitochondria. Images in (A) obtained with overexpression of a constitutively active Akt illustrate how Akt prevents HKII dissociation evoked by removal of glucose. Panel B shows a quantification of this effect, comparing the rate of HKII dissociation from mitochondria in response to glucose removal in the presence and absence of exogenous Akt. Data in Panel C show how glucose clearance is affected by preincubation in the absence of glucose and how constitutively active Akt prevents this effect. Previous data ([Fig. 6B](#pone-0017674-g006){ref-type="fig"}) have shown that HKII dissociates from mitochondria 15 to 30 min after removal of extracellular glucose; comparison of traces C1 and C2 (in C2 cells were incubated for 30 min in the absence of glucose) indicates that this effect is accompanied by a decrease in glucose clearance. Thus, glucose-induced HK dissociation slows the rate of glucose utilization. The recording in C3 obtained with cells overexpressing constitutively active Akt indicates that Akt prevents the decrease in glucose clearance induced by incubation in the absence of glucose for 30 min. This effect is very likely related to the effect of Akt on HKII translocation. The bar graph in C4 quantifies the effects of glucose removal and Akt on glucose clearance. :::  ::: Since Akt facilitates HK interaction with mitochondria [@pone.0017674-Miyamoto1], we examined whether translocation of HKII from mitochondria to cytoplasm in response to zero glucose pre-incubation was suppressed by overexpression of constitutively activate Akt. With Akt overexpression, the mixed pattern of mitochondria-bound and cytosol-associated HKII-YFP did not change ([Fig. 7A](#pone-0017674-g007){ref-type="fig"}), and prolongation of the t~1/2~ of glucose clearance by zero glucose preincubation was abolished 107.0±9.5 s versus 210±19 s (n = 8) with and without Akt) ([Fig. 7A](#pone-0017674-g007){ref-type="fig"} and [Fig. 7C3 and C4](#pone-0017674-g007){ref-type="fig"}). Thus, both glucose and Akt signaling promote the binding of HKII to mitochondria, favoring glucose catabolism over glycogen synthesis. Role of G-6-P in regulating HK activity and subcellular distribution {#s2e} -------------------------------------------------------------------- We have hypothesized that HKII translocation to the cytosol delays the t~1/2~ of glucose clearance when glucose is removed because newly synthesized glycogen is mobilized and in the process keeps G-6-P levels elevated, suppressing HK activity until glycogenolysis ceases. We tested this hypothesis using both direct and indirect approaches. First, in the presence of glucose, we applied iodoacetate (IAA, 0.25 mM) to elevate G-6-P levels by inhibiting glycolysis. As shown in [Fig. 8A and C1](#pone-0017674-g008){ref-type="fig"}, IAA caused translocation of HKII into the cytosol, in the continued presence of 10 mM glucose, with a time constant of 7.5 s (n = 10). This effect was fully reversible and HKII re-associated with mitochondria with a time constant of 16 s (n = 9) following removal of IAA ([Fig. 8C2](#pone-0017674-g008){ref-type="fig"}). To rule out ATP depletion, rather than G-6-P accumulation, as the cause of HKII translocation, we also tested the mitochondrial uncoupler FCCP (5 µM) to deplete mitochondrial ATP without elevating G-6-P levels. In 6 cells, FCCP had no effect on HKII translocation ([Fig. 8B](#pone-0017674-g008){ref-type="fig"}). These data support our hypothesis that the effect of glucose on HKII translocation and inhibition is mediated via G-6-P. ::: {#pone-0017674-g008 .fig} 10.1371/journal.pone.0017674.g008 Figure 8 ::: {.caption} ###### Rates of association and dissociation of HKI and HKII in response to glucose, IAA and FCCP. IAA, but not FCCP, facilitates HKII dissociation from mitochondria. Images in Panel A obtained with HKII-YFP expressed in CHO cells show that addition of the glycolysis inhibitor, IAA, in the presence of 10 mM glucose causes HKII dissociation from mitochondria within 5 to 15 min. In contrast, addition of FCCP, which depletes mitochondrial ATP, had no effect on HKII interaction with mitochondria (B). These data indicate that G-6-P, rather than ATP mediates the effect of glucose removal on HKII translocation. Panel C shows the rates measured as ratio of fluorescence intensity obtained from intracellular domains without and with mitochondria. C1 is a comparison of the rates of HKII dissociation from mitochondria in response to IAA and FCCP addition in the presence of 10 mM glucose outside. C2 illustrates the rate of reassociation HKII with mitochondria upon removal of IAA. :::  ::: To obtain further evidence that G-6-P causes dissociation of HKs from mitochondria, we used permeabilized CHO cells overexpressing HKI-YFP or HKII-YFP. Both HKI and HKII spontaneously dissociated slowly from mitochondria when the plasma membrane was permeabilized using 50 µM β escin ([Fig. 9](#pone-0017674-g009){ref-type="fig"}), with time constants of 19.5 min (n = 6) and 17 min (n = 5) ([Fig. 9C](#pone-0017674-g009){ref-type="fig"}), respectively although the difference was not significant. Furthermore addition of 100 nM G-6-P at the time of cell membrane permeabilization accelerated the rates of both HKI and HKII dissociation (τ = 6.2 min (n = 4) for HKI and τ = 5.5 min (n = 5) for HKII). Our data obtained with permeabilized cells corroborate previous findings that G-6-P displaces both HKI and HKII from isolated mitochondria [@pone.0017674-Skaff1]. However when combined with our data obtained from intact cells, indicate that an intracellular factor is present in intact cells that prevents the dissociation of HKI induced by G-6-P in isolated mitochondria. Importantly, our data suggest that, in the absence of any factor other than G-6-P the two HKs have similar affinity for mitochondrial membrane and dissociate at similar rates. ::: {#pone-0017674-g009 .fig} 10.1371/journal.pone.0017674.g009 Figure 9 ::: {.caption} ###### HKs dissociation from mitochondria in permeabilized cells. Panels A and B illustrate the spontaneous dissociation of HKI-YFP in the absence and presence of G-6-P, respectively, after cell membrane permeabilization with 50 µM β-escin. Panel C1 and C2 show that HKI and HKII dissociate for mitochondria at very similar rates, and G-6-P increases these rates by approximately 4 fold (20 min vs. 5 min). The triangles represent measurements obtained with G-6-P (100 nM) and the circles are for values obtained in the absence of G-6-P. Panel C3 shows the change in intracellular G-6-P levels following glucose removal. :::  ::: Finally, we directly measured intracellular G-6-P levels following removal of 10 mM glucose in CHO cells transfected with GLUT1. Upon removal of glucose, G-6-P levels remained elevated for about ∼10 min and then decreased in a sigmoidal fashion with 50% decay reached after ∼15 min ([Fig. 9C3](#pone-0017674-g009){ref-type="fig"} each data point in the graph represents an average of 8 to 26 values). Together, these data corroborate our glucose metabolism measurements, and support the hypothesis that upon glucose removal G-6-P elevation inhibits glucose phosphorylation by hexokinases. The decay in G-6-P that follows, accounts for the gradual reactivation of hexokinase activity and resumption of glucose phosphorylation. Discussion {#s3} ========== Catabolic and anabolic glucose utilization are both directed by hexokinases, which channel G-6-P to glycolysis or glycogen and lipid synthesis. While HKs coexist in many cell types, cells that generate glycogen in response to insulin, such as adult muscle, express primarily HKII, whereas cells that rely primarily on glycolysis for energy production, such as the brain, express high levels of HKI. The specificity of these enzymes is not related to functional differences, since both phosphorylate glucose to G-6-P, but may instead depend on subcellular location, reflecting spatial compartmentalization of glucose metabolism. Thus, it has been suggested that the association of HKI with mitochondria channels glucose to the glycolytic pathway, whereas HKII, when translocated to the cytoplasm, controls glycogen formation [@pone.0017674-Wilson1]. Our results using live cell imaging to track the subcellular distributions of HKI and HKII provide direct support for this hypothesis. Glycogen breakdown generates G-6-P that inhibits HK enzymatic activity {#s3a} ---------------------------------------------------------------------- Our data show that HKI is strongly associated with mitochondria, while HKII dynamically translocates between mitochondria and the cytoplasm. This differential localization is associated with a fast rate of glucose utilization upon removal of extracellular glucose in cells expressing HKI, but inhibition of glucose utilization in cells expressing high levels of HKII. We propose that cytosolic HK II channel G-6-P towards the glycogen synthesis pathway during exposure to 10 mM glucose, such that upon glucose removal, mobilization of the newly-synthesized glycogen regenerates G-6-P (or another glycolytic intermediate) which inhibits HK activity. This phenomenon is primarily observed in cells with cytosolic HKs that favor glycogen synthesis. In cells with mitochondria-associated HKs, G-6-P is channeled towards the glycolytic pathway and there is no G-6-P-induced inhibition. The following evidence supports these hypotheses: 1) Conditions which favor glycogen formation, such as PTG overexpression and long exposure to glucose, enhances the duration of the inhibitory phase; 2) The effect of glucose is only observed with overexpression of GLUT1 in CHO cells, which raises intracellular glucose enough to rapidly stimulate glycogen synthesis. Indeed, only with overexpression of GLUT1 can glycogen synthesis be stimulated by glucose within minutes. Without GLUT1 overexpression, significant glycogen synthesis requires more than 30 min of glucose exposure and there is no HK inhibition upon glucose removal; 3) Akt, which facilitates HKII interaction with mitochondria, reduces the inhibitory phase and removal of glucose, which causes HKII dissociation from mitochondria, enhances the inhibitory phase. G-6-P regulates HK enzymatic activity as well as subcellular localization {#s3b} ------------------------------------------------------------------------- In HKI, the C terminal domain binds ATP and exhibits catalytic activity, while the N terminal domain facilitates HK interaction with the mitochondria outer membrane. Both domains bind glucose and G-6-P synergistically. Binding of G-6-P to the C and N terminal domains inhibits enzymatic activity and causes dissociation from mitochondria either directly or via allosteric interaction between the two domains. Pi prevents HKI inhibition and dissociation by competing with G-6-P at the N terminal domain [@pone.0017674-Purich1], [@pone.0017674-Skaff1], [@pone.0017674-Sui1], [@pone.0017674-Jurczak1], [@pone.0017674-Ellison1], [@pone.0017674-Fang1], [@pone.0017674-Aleshin1]. On the other hand, Pi does not bind to HKII and both the N- and C- domains exhibit catalytic activity. We have shown that HKII dissociation from mitochondria is insensitive to FCCP, which induces ATP depletion, but is facilitated by the glycolytic inhibitor IAA, which elevates G-6-P. These results indicate that the effect of glucose on HKII translocation is mediated by G-6-P and not by ATP. Thus, ATP regulates the catalytic activity of HKs, but not their association with mitochondria. The lack of effect of glucose removal and IAA on HKI association with mitochondria supports the hypothesis that Pi competes with G-6-P to strengthen the interaction with mitochondria. We have shown that upon plasma membrane permeabilization, both HKI and HKII dissociate from mitochondria with a time constant of approximately 20 min, which is accelerated 4-fold by 100 nM G-6-P. These data are consistent with results obtained with isolated mitochondria demonstrating that G-6-P facilitates dissociation of both HKI and HKII from mitochondria. Furthermore, our finding that HKI does not dissociate from mitochondria upon glucose removal in intact cells supports the hypothesis that an additional factor, perhaps Pi, prevents the effect of G-6-P on HKI in intact cells. Phosphorylation of residue T473 by Akt in the linker region between the N and C terminal domain of HKII may facilitate its interaction with mitochondria [@pone.0017674-Miyamoto1]. It has been proposed that this effect of Akt requires the presence of glucose, based on the observation that the protective effect of Akt against apoptosis, which is mediated by HKs, requires glucose [@pone.0017674-Majewski1]. Our results demonstrate that Akt does indeed regulate the interaction of HKII with mitochondria, but its main effect is to prevent dissociation from mitochondria upon glucose removal. To account for these results, we suggest that Akt phosphorylates mitochondria-bound HKII, thereby preventing the ability of G-6-P to reduce HK mitochondrial interaction. Physiological relevance of preferential expression of HKI and HKII {#s3c} ------------------------------------------------------------------ In most native tissues such as muscle, liver and fat, hexokinases coexist, but in muscle HKI is most abundant in fetal tissues, while HKII represent 80% of total hexokinase activity in adult muscle [@pone.0017674-Mandarino1], [@pone.0017674-Postic1]. In adult muscle the increased expression of HKII is associated with increased expression of GLUT4 and the development of insulin sensitivity. In this context, HKII is both cytosolic as well as bound to mitochondria. In light of our data, it is interesting to speculate how this switch in HK and GLUT affects glucose metabolism and cell function. In adult muscle when extracellular glucose is high, for instance during a meal, G-6-P is channeled towards both glycogen synthesis and glycolysis, so that G-6-P levels remain low. However, between meals, when extracellular glucose decreases and more importantly when insulin is no longer released by pancreatic β cells, GLUT4 is internalized and intracellular glucose drops dramatically. The ensuing G-6-P production resulting from glycogen breakdown inhibits mitochondria-bound HKII-mediated glycolysis within seconds. Elevated G-6-P also causes dissociation of HKII from mitochondria further inhibiting glucose utilization. G-6-P generated by glycogen breakdown is then channeled through the glycolytic pathway to be further metabolized. In contrast, in fetal tissue with high level of HKI and GLUT1 most HKs are bound to mitochondria and the low levels of HKII found in the cytosol only generates little glycogen during meals. Moreover, mitochondria-bound HKI being poorly sensitive to G-6-P it follows that the low level of G-6-P produced in between meals will have minimal effect on HK activity and translocation. Finally, GLUT1 insertion into the plasma membrane is not regulated by insulin, as a result glucose uptake and intracellular glucose remain high even when blood glucose decreases, further preventing glycogen degradation. These properties ensure that glycolysis is the main source of energy in these fetal tissues. Such difference between HKI and HKII would not only affect differentially glucose metabolism in fetal and adult muscle, but also cell fate. There is a large body of evidence that mitochondria-bound HKs protect against apoptosis via a direct interaction with VDAC, which prevents the mitochondrial permeability transition pore opening. Alternatively, the anti-apoptotic effect of mitochondria-bound HKs may require glucose metabolism [@pone.0017674-Gottlob1], [@pone.0017674-Sun1]. It follows that during stress fetal cells preferentially expressing HKI may cope better than adult cells expressing HKII as HKI remains associated with mitochondria, thus maintaining glycolysis activity and stability of VDAC in a low permeability state. In summary, our data suggest that in cells that preferentially express HKII and GLUT4, such as adult muscle, removal of glucose causes the glycogen pathway to switch from synthesis to degradation. The ensuing elevation of G-6-P inhibits and then displaces HKII from mitochondria, causing the cell to switch from glucose to glycogen utilization as a source of energy. In contrast, in cells like neonatal muscle cells, which express preferentially HKI and GLUT1, the strong association of HKI with mitochondria implies that the cells primarily rely on glucose utilization as a source of energy and glycogen has a minor role. In this case, the strong association of HKI with mitochondria may also protect against metabolic stress. Materials and Methods {#s4} ===================== Solutions and experimental techniques {#s4a} ------------------------------------- The bath solution for cell imaging consisted of (in mM) 140 NaCl, 5 KCl, 1.1 MgCl~2~, 2.5 CaCl~2~, 10 HEPES, with the pH adjusted to 7.2 with KOH. Glucose was added to this solution and NMDG was added to maintain the solution\'s osmolarity in the absence of glucose. Solutions were perfused directly over the cells using a gravity fed eight way perfusion device (Warner Instruments, Hamden, CT) with electrically controlled solenoids (The Lee Company, Westbrook, CT). Input and output of solution volumes to the recording chamber (glass bottomed Petri dish) were equilibrated to maintain constant flow rates and pressures within the recording chamber. Experiments were carried out at room temperature (25°C). Cytochalasin B (Cyto B) and other chemicals were purchased from Sigma-Aldrich (St Louis, MO). Molecular biology and cell culture {#s4b} ---------------------------------- We obtained the FLIPglu-600 µM cDNA from Dr. W.B. Frommer, GLUT from Dr. D. Abel, GLUT4-GFP from Dr. S. Cushman, rat HKI and HKII from Dr. J. Wilson, PTG from Dr. M Brady and GL-GFP and PTG-GFP from Dr. J. Guinovart. Fusion of PTG or GL to mCherry was carried out by subcloning into the mCherry N-1 vector. Constitutively active Akt was obtained from Dr. W. Sellers (Addgene plasmid 9009). Fusion of rat HK\'s to YFP was accomplished by inserting a Bam H1 site at the last amino acid of the coding sequence and subcloning into a modified pEYFPN-1vector (Clontech, Mountain View, CA). The modified pEYFPN-1 carried the mutations Q86K and A206K. Similarly, GLUT1-GFP was constructed by inserting a Kpn site at the last amino acid of the coding sequence and subcloning into pEGFPN-1 vector (Clontech, Mountain View, CA). All constructs were subcloned into the mammalian expression vectors utilizing the CMV promoter. CHO cells were transfected with 1 µg DNA plus 2.5 µl lipofectamine 2000 (Invitrogen, Carlsbad, CA) per 1.5 cm^3^ Petri dish. Expression of FLIPglu-600 µM was sufficiently high after 24 h to perform FRET experiments. Cells were cultured in DMEM high (25 mM) glucose medium supplemented with 10% (v/v) fetal bovine serum (FBS), penicillin (100 units/ml), streptomycin (100 units/ml) and 2 mM glutamine and divided once a week by treatment with trypsin. siRNA-mediated depletion and quantitative PCR {#s4c} --------------------------------------------- CHO cells (0.5--2×10^6^) were transfected with 30 to 40 nM siRNAs using lipofectamine 2000 reagent (Invitrogen, Carlsbad CA). Cells were used for experiments 2 days after transfection. The sequences of siRNAs (DsRNA from IDT, San Diego, CA) were: HKI AACGTGAATCCCACAGGTAACTTCTTG and CGGATGTCTTCTAATGATCCATCGTC HKII GTATCCAATTCAATAGTTACATCCCTC and CTTTGGTTTCCTTTGCTTAACATCCCA RT--PCR analysis. Purified RNA from CHO cells (2--8×10^6^) was isolated using the RNAeasy Kit (Qiagen, Valencia, CA). First-strand cDNA synthesis was primed with oligo (dT) (Superscript III kit, Invitrogen, Carlsbad, CA). Real-time PCR was performed using an iCycler IQ5 (Bio-Rad, Hercules, CA) with the iQ SYBR Green Supermix (Bio-Rad, Hercules, CA). The cDNA levels were normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) or actin. Values shown in Supplementary [Table S1](#pone.0017674.s005){ref-type="supplementary-material"} are C~T~ test -- C~T~ GAPDH. Sequences of the primers were: HKI GAAGAATGGCCTCTCCCGGG and GCCATGCACGATGTTCTCTGG HKII AGAGCATCCTCCTCAAGTGGAC and ACCAAGTGCAGAAGGTTGACCA GAPDH GCGCCAGCATCACCCCATTTGATG and GGTCGGCGTGAACGGATTTGGCCG Actin TGGCATCCACGAAACTACAT AND TGGTACCACCAGACAGCACT Cell membrane permeabilization {#s4d} ------------------------------ The cells were permeabilized with 50 µM β-escin dissolved in an intracellular-type solution containing 140 mM KCl, 5 mM NaCl, 0.5 mM MgCl~2~, 100 nM Ca^2+^, 5 mM glucose and 1 mM DTT. To monitor cell membrane permeabilization we expressed the fluorophore CFP together with HKx-YFP. After exposure to β-escin for about 45 s CFP leaked slowly out of the cell over a period of 30 to 45 min, indicating cell membrane permeabilzation. At the same time HKs begun to slowly dissociate from mitochondria. FRET imaging {#s4e} ------------ All cells were imaged live without fixation. Images (16-bit) were acquired using a Nikon Eclipse TE300 microscope fitted with a 60× (N.A. 1.4) oil immersion lens (Nikon, Melville, NY) and equipped with a filter cube comprising a CFP bandpass excitation filter: 436/20b, together with a longpass dichroic mirror: 455DCLP (Chroma Technology Corp, Rockingham, VT). LED\'s (Lumileds; San Jose, CA) were used as light sources: one emitting at 455±20 nm (royal blue) and the other emitting at 505±15 nm (cyan). LED\'s and camera exposure were controlled by MetaFluor Imaging 6.1 software (Molecular Devices Corp., Sunnyvale, CA). Ratiometric FRET measurements were performed by simultaneously monitoring CFP and YFP emissions of the sample when excited at the wavelengths for CFP (royal blue LED), as described previously [@pone.0017674-John1]. The ratio between YFP and CFP emission were measured online in real time using MetaFluor Imaging software. For analysis, background light intensity was subtracted from the individual YFP and CFP emission. YFP and CFP images were acquired simultaneously using a Dual View image splitter (Optical Insights, Tucson, AZ) equipped with a 505 nm long-pass dichroic filter to separate the CFP and YFP signals, a CFP emission filter (480/30) and a YFP emission filter (535/40). Superposition of the CFP and YFP images was carried out using the imaging software. Images were captured with a Cascade 512B digital camera (Photometrics, Tucson, AZ). Exposure times were optimized in each case but varied between 300--500 ms and were recorded at a constant rate for each cell between 0.2--0.33 Hz. Many experiments lasted more than one hour leading to a slow drift in the FRET ratio baseline in some cases. In most figures an initial drift in FRET ratio was corrected using exponential curve fitting. In most cases changes in the FRET ratio measured upon addition of glucose could be fitted to a single exponential. FRET ratio recovery following glucose removal could not be fitted to a single exponential due to a lag in glucose clearance following removal of extracellular glucose. In this case the rate of glucose clearance was satisfactorily fitted to a sigmoidal function y = y~0~+a/1+e−(x~0~/s), where ×0, an estimate of the time requires to reach half of the change in FRET ratio, accounts for the lag period, while s, an estimate of the rate of change is more or less independent of the lag period. Translocation imaging of HKs between mitochondria and cytoplasm was quantified as ratio of light intensity between two intracellular regions of interest with and without a high concentration of mitochondria. Background light was subtracted in both cases. Images were acquired using an Olympus IX70 inverted microscope (Olympus America Inc., Center Valley, PA) fitted with an Olympus plan apo ×60. 1.4 N.A. oil immersion objective and a cooled CCD camera (Model Quantix, Photometrics, Tucson, AZ). Imaging Workbench software was used for data acquisition and analysis. YFP (XF104-2) and CFP (XF130-2) filter cubes were purchased from Omega Optical Inc. (Brattleboro, VT). Glucose-6-Phosphate assay {#s4f} ------------------------- We used a commercial G-6-P assay kit (Biovision Research Products, Mountain View, CA). CHO cells (0.5--2×10^6^) were transfected with GLUT1 (∼3 µg). Two days later cells were washed in ice cold PBS and frozen in situ in dryice ethanol. 200 µl of 6% (v/v) perchloric acid added and the cells scraped while frozen. Cells were homogenized via a "Qiashredder" column at 4°C and the homogenate neutralized with 500 mM ethanolamine and 10 M KOH [@pone.0017674-Taylor1]. Samples were then centrifuged to remove insoluble material (10 mins 15000 g 4°C) 50 µl samples were transferred into wells of a 96-well plate and brought to a volume of 100 µl with assay buffer. Absorbance at 450 nm was measured, using a POLARstar Omega microplate reader (BMG LABTECH GmbH, Offenburg/Germany). Background samples (without enzyme mix) were introduced to estimate the NADH/NADPH levels prior to G-6-P conversion. Finally, to estimate the G-6-P levels a curve was generated with G-6-P standards containing 0, 2, 4, 6, 8, 10, 20 nmol/well. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Subcellular distribution of HKI linked to YFP in CHO cells.** To investigate the subcellular localization of HK in CHO cells we co-expressed HKI-CFP (A) together with the mitochondria marker MitoTracker-YFP (B). In panel C superimposition of the two images shows that the localization of HKI and MitoTracker overlap, indicating HKI association with mitochondria. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Glucose uptake and clearance in CHO cells over expressing GLUT1 and GLUT4.** To monitor GLUT insertion into the plasma membrane, we used constructs of GLUT1 and GLUT4 linked to GFP. In both cases, there was only faint GFP fluorescence detected in the plasma membrane, with most of the fluorescence associated with intracellular membrane compartments (results not shown). However, as shown in Panels B and C, even this low level of plasma membrane GLUT insertion had dramatic effects on glucose transport. With GLUT1 over expression, the rate of glucose entry, illustrated by a decrease in FRET ratio, increased dramatically, with the time constant τ dropping from 34.5+/−5 s (n = 7) to 1.9+/−0.6 s (n = 8) (Panels 1 and 3). The rate of glucose clearance, indicated by an increase in FRET ratio also increased, with a t~1/2~ approaching 10 s (9.8+/−3.5 s). With GLUT4 over expression, the rate of glucose uptake decreased by almost 14-fold, from 34.5 s to 2.5 s (Panels 1 and 5). The rate of glucose clearance also increased in this case, although to a lesser level than with GLUT1, the time to reach half clearance (t~1/2~) being around 12.5+/−6.5 s (n = 7). In panels 1, 3 and 5 the rate of glucose uptake was fitted to an exponential decay, in panels 2, 4 and 6 glucose clearance was fitted to a sigmoidal function (see [Materials and Methods](#s4){ref-type="sec"} for details). These data suggest that increased insertion of GLUTs in the plasma membrane increase both glucose influx and efflux. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Effects of GLUT1 on glucose utilization in the presence of Cyto B in CHO cells.** Panel A shows data obtained with a CHO cell exposed to Cyto B (20 µM) in the continuous presence of 10 mM glucose. The right hand side plots show the rate of glucose clearance following removal of glucose in the absence of Cyto B (A1) and in the presence of Cyto B (B1). The closeness of the two τs suggests that, with cells exhibiting a slow glucose uptake phenotype, GLUT-mediated glucose efflux has little effect in glucose clearance. This trace also illustrates that the effect of Cyto B is readily reversible, as glucose entry resumed, albeit at a lower rate, 30 s after Cyto B removal. In B data were obtained with a CHO cell transfected with GLUT1, which exhibited a high rate of glucose uptake. In this case, Cyto B dramatically reduced the rate of clearance as compared to that recorded in the absence of Cyto B and after outside glucose removal (plots B3 and B4). These data suggest that GLUT-mediated efflux plays a role in glucose clearance in cells overloaded with glucose. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **siRNA-mediated depletion of HKI and HKII.** To study the role of HKs in glucose clearance we transfected CHO cells with 30 to 40 nM of HKI or HKII siRNA. (A) Inhibition of HKI expression caused a decrease in the rate of glucose clearance with the time to reach half clearance (t~1/2~) increasing from 75+/−15 s to 135+/−22 s (n = 9). (B) Similarly, inhibition of HKII expression decreased glucose clearance with t~1/2~ reaching 152+/−20 s (n = 9). These results indicate that endogenous HKI and HKII equally contribute to glucose clearance in CHO cells. The data obtained with RT-PCR corroborate this result and show that the levels of endogenous HKI and HKII are very similar in these cells. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Quantification of HKI and HKII levels in CHO cells.** We have used the RT-PCR method to estimate the levels of HKI and HKII in native cells and in cells over expressing HKI and HKII. Our results indicate that the level of these two enzymes is similar. This is consistent with our observation where depletion of endogenous HKI and HKII decrease glucose clearance to similar extend. This quantitative analysis also shows that cell culture in the absence of glucose causes a slight increase in HK levels. This is important when considering the effect of glucose removal on glucose utilization. Indeed, we have shown that incubation in the absence of glucose decreases glucose clearance and we hypothesized that this effect was related to HK translocation away from mitochondria. Our quantitative analysis supports this hypothesis and shows that the effect on glucose clearance is not due to a decrease in HK level. Finally, we show, as expected, that over expression of HKs is accompanied by an increased level of intracellular HK. Thus, the decrease in glucose clearance observed with HKII over expression may be related to increased level of the enzyme in the cytoplasm, favoring glycogen synthesis. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by NIH grants R37HL60025 and P01 HL071870, Laubisch and Kawata Endowments to J.N.W. and a Laubsich endowment to S.J. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SJ JNW BR. Performed the experiments: SJ BR. Analyzed the data: SJ BR. Wrote the paper: SJ JNW BR.

PubMed Central

2024-06-05T04:04:19.279075

2011-3-9

PMC3052387

Introduction {#s1} ============ Pramipexole is a dopamine D~2~ receptor (D~2~R) subfamily agonist. It was introduced for treating motor symptoms in patients with idiopathic Parkinson\'s disease (PD) [@pone.0017723-Bennett1] and has been shown to be effective by various clinical trials [@pone.0017723-Lieberman1], [@pone.0017723-Shannon1]. In addition, various studies have recently found antidepressive effects of pramipexole not only in patients with PD complicated by depressive state [@pone.0017723-Barone1], [@pone.0017723-Lemke1], [@pone.0017723-Reichmann1], [@pone.0017723-Rektorova1], but also in depressive patients without parkinsonian symptoms [@pone.0017723-Lattanzi1], [@pone.0017723-Corrigan1], [@pone.0017723-Goldberg1], [@pone.0017723-Zarate1], [@pone.0017723-DeBattista1], [@pone.0017723-Ostow1], [@pone.0017723-Sporn1], [@pone.0017723-Perugi1], [@pone.0017723-Cassano1]. Its antidepressive effects have been also shown in animal experiments [@pone.0017723-Maj1], [@pone.0017723-Willner1]. The D~2~R subfamily consists of D~2~, D~3~, and D~4~ receptor subtypes [@pone.0017723-Missale1]. Pramipexole is active mainly at D~2~ and D~3~ receptors, and compared with other dopamine agonists, it is unique in that the binding affinity for D~3~ receptors is higher than that for D~2~ receptors [@pone.0017723-Millan1], [@pone.0017723-Mierau1], [@pone.0017723-Piercey1], [@pone.0017723-Kvernmo1]. Kvernmo et al. reported that binding affinities (inhibition constant; K~i~) for cloned human D~2~ and D~3~ receptors of pramipexole were 3.9 and 0.5 nmol/L, respectively [@pone.0017723-Kvernmo1]. The distribution of D~3~ receptors in the brain is different from that of D~2~ receptors [@pone.0017723-Gurevich1], [@pone.0017723-Murray1], [@pone.0017723-Sokoloff1], [@pone.0017723-Bouthenet1], [@pone.0017723-Landwehrmeyer1], [@pone.0017723-Diaz1]. Compared with D~2~ receptors, D~3~ receptors are predominantly located in extrastriatal regions including the mesolimbic dopamine system involved in mood and behavior. On the other hand, although both D~2~ and D~3~ receptors in the striatum are much more abundant than those in other regions, the D~2~ receptor density is higher than the D~3~ receptor density in the striatum. Stimulation of D~2~ and D~3~ receptors appears to induce different effects. There are growing evidences that D~3~ receptors may play a role in the pathogenesis of depression because of their pharmacology and distribution in the brain [@pone.0017723-Sokoloff1], [@pone.0017723-Joyce1], [@pone.0017723-Levant1], although the exact mechanism remains unknown. The mechanism of antidepressive effects and extrastriatal binding sites of pramipexole are also unknown, and no study has investigated this issue. These effects have been speculated to occur by means of activation of D~2~R subfamily, especially the D~3~ receptor subtype, in the mesolimbic dopamine system [@pone.0017723-Bennett1]. Therefore, we aimed to determine the binding sites of pramipexole in the extrastriatal dopaminergic regions by using ^11^C-FLB 457 positron emission tomography (PET) scanning for quantification of D~2~/D~3~ receptors in extrastriatal brain regions. In addition, we discussed whether the regional sites occupied by pramipexole may be target sites where pramipexole exerts its antidepressive effects on the basis of previous anatomical and functional reports on depression. Materials and Methods {#s2} ===================== Subjects {#s2a} -------- This study protocol was approved by the Ethics Committee of the Tokyo Metropolitan Institute of Gerontology. Written informed consent was obtained from all participants. A total of 15 healthy volunteers (7 men and 8 women; mean age  =  50.2 years, SD = 11.7, range  =  30--77) participated in the study. All subjects underwent two ^11^C-FLB 457 PET scans and magnetic resonance imaging (MRI) of the brain. They were classified into 3 groups according to the dose of pramipexole (0.25, 0.125, or 0 mg). The 5 subjects in the high-dose group (2 men and 3 women, 57.2±12.8 years) received a single oral 0.25 mg dose of pramipexole. Another 5 subjects in the low-dose group (2 men and 3 women, 51.4±9.4 years) received a single oral 0.125 mg dose of pramipexole. The drug was administered between the two PET scans. The other 5 subjects were in the control group (3 men and 2 women, 42.0±6.3 years) and received no medication. Significant difference was not found in age between the 3 groups with one-way ANOVA test. All volunteers were free of any current or past mental disorders, and defined as healthy on the basis of their medical history, the results of their physical and neurological examinations and routine mental health interview performed by a neurologist, and the findings of the MRI. None had been receiving any other medications at the time of this study. Doses of pramipexole {#s2b} -------------------- Previous studies have shown that administration of more than approximately 1 mg of pramipexole exerts antidepressive effects [@pone.0017723-Barone1], [@pone.0017723-Lemke1], [@pone.0017723-Reichmann1], [@pone.0017723-Rektorova1], [@pone.0017723-Lattanzi1], [@pone.0017723-Corrigan1], [@pone.0017723-Goldberg1], [@pone.0017723-Zarate1], [@pone.0017723-DeBattista1], [@pone.0017723-Ostow1], [@pone.0017723-Sporn1], [@pone.0017723-Perugi1], [@pone.0017723-Cassano1]; compared with these studies, the doses of pramipexole used in our study were low. We chose these low doses to ensure the safety of the participants in this study. According to unpublished data on file in Nippon Boehringer Ingelheim (Tokyo, Japan), a single administration of pramipexole 0.4 mg caused orthostatic hypotension in an early clinical trial of German volunteers. On the basis of these results, the doses of pramipexole were set at 0.1, 0.2, and 0.3 mg in the Phase one clinical trial of Japanese volunteers, and no one developed more than moderate adverse effects. After oral administration of a single dose of pramipexole 0.1 mg, C~max~, T~max,~ and t~1/2~ were 294.6±46.3 pg/mL, 1.5±0.5 h and 7.71±1.90 h (mean ± SD), respectively. After administration of pramipexole 0.2 mg, the values were 583.2±69.9 pg/mL, 1.4±0.5 h, and 6.36±1.46 h, respectively; after a dose of 0.3 mg, the values were 766.3±88.8 pg/mL, 2.3±1.2 h, and 6.94±1.09 h, respectively. One tablet of pramipexole equals 0.125 mg. Therefore, the doses of pramipexole were set at 0.25 and 0.125 mg in this study. ^11^C-FLB 457 PET imaging {#s2c} ------------------------- Each volunteer participated in two ^11^C-FLB 457 PET scans on the same day---one in the morning and another in the afternoon. Of 15 subjects, 10 were administered with either 0.25 or 0.125 mg of pramipexole after the first PET scan; the second PET scan took place 1--1.5 h later because the concentration of pramipexole in plasma reaches its peak in approximately 1--2 hours as described above. PET imaging was performed at the Positron Medical Center, Tokyo Metropolitan Institute of Gerontology, with a SET-2400W scanner (Shimadzu, Kyoto, Japan). The spatial resolution was 4.4 mm full width at half maximum in the transverse direction and 6.5 mm full width at half maximum in the axial direction. Images with 50 slices were obtained with a 2×2×3.125-mm voxel size and a 128×128 matrix size. The transmission data were acquired by using a rotating ^68^Ga/^68^Ge rod as a source for attenuation correction. ^11^C-FLB 457 was prepared as described previously [@pone.0017723-Halldin1]. In the first PET experiment, the injected dose, specific activity, and mass of injected ligand were 283±24 MBq, 118±45 GBq/µmol, and 3.0±1.7 µg (mean ± SD), respectively. The respective values in the second PET experiment were 285±19 MBq, 110±38 GBq/µmol, and 3.2±2.0 µg. The time interval between the first and second injections of ^11^C-FLB 457 was 4--4.5 hours. The mass of injected ligand in each second scan was carefully adjusted to that in each first scan because of potential occupancy effects by unlabelled ligand itself [@pone.0017723-Olsson1], [@pone.0017723-Suhara1], and no significant difference was found in the mass of injected ligand as well as the injected dose and specific activity between the first and second scans in each group, using paired Student *t* test. A dynamic series of decay-corrected PET data acquisition was performed in the 3D mode for 90 minutes starting at the time of the intravenous injection of ^11^C-FLB 457. The frame arrangement was 20 s ×6 frames, 60 s ×2 frames, 180 s ×2 frames, and 300 s ×16 frames. For the 5 subjects in the high-dose group, arterial blood samples were also obtained. Immediately after the intravenous injection of ^11^C-FLB 457, 18 arterial blood samples were collected at 10-s intervals over 3 min; the next 2 samples were collected at 60-s intervals over 2 min, and the remaining 10 samples were collected at longer intervals, for a total of 30 samples. All samples were manually drawn. Plasma was separated, weighed, and measured for radioactivity with a sodium iodide (Tl) well scintillation counter. Six samples collected at 3, 10, 20, 30, 40, and 60 minutes were further processed by high-performance liquid chromatography to determine the fractions of plasma radioactivity corresponding to unchanged ^11^C-FLB 457 and labeled metabolites, as described previously [@pone.0017723-Halldin1]. Data analysis {#s2d} ------------- Image manipulations were performed using Dr. View version R2.0 (AJS, Tokyo, Japan) and statistical parametric mapping 2 (SPM2; Functional Imaging Laboratory, London, UK) implemented in MATLAB version 7.0.1 (The MathWorks, Natick, MA). First, individual two dynamic ^11^C-FLB 457 images and MRI images were coregistered. Next, regions of interest (ROIs) were defined over the prefrontal, parietal, lateral temporal and anterior cingulate cortices, medial and lateral parts of the thalamus, amygdala, hippocampus, and cerebellum on the individual coregistered MRI. These ROIs were spatially moved on the corresponding coregistered dynamic ^11^C-FLB 457 images. ^11^C-FLB 457 binding to extrastriatal D~2~/D~3~ receptors was calculated as the binding potential (BP) by the simplified reference tissue model (SRTM) using cerebellum as a reference tissue [@pone.0017723-Lammertsma1]. BP derived with this method is referred to as BP~ND\_SRTM~ (ND: nondisplaceable). For the high-dose group with arterial blood samples, the binding was also analyzed by using the linear graphic analysis by Logan et al. [@pone.0017723-Logan1]. The slope of the linear phase of the obtained plot corresponds to the total distribution volume (*V* ~T~) of the ligand plus the plasma volume. The regional *V* ~T~ was determined from the slope, and the BP with this method was calculated as follows using cerebellum as a reference region: BP~ND\_Logan~  =  (*V* ~T~ on ROI/*V* ~T~ on cerebellum) - 1. D~2~/D~3~ occupancy rate by pramipexole was calculated for each ROI by using the following equation: occupancy rate (%) = 100× (BP at baseline -- BP at pramipexole-loading)/BP at baseline. BP at baseline and BP at pramipexole-loading are obtained from first and second PET scans, respectively. Data were expressed as mean ± SD. Statistical Analysis {#s2e} -------------------- The differences between first and second PET scans were tested by paired Student\'s *t*-test. Correlations between BP~ND\_SRTM~ and BP~ND\_Logan~ in high-dose group were assessed by means of linear regression analysis with Pearson\'s correlation test. *P* values\<0.05 were considered statistically significant. Results {#s3} ======= For the high-dose group, BP~ND\_SRTM~ was found to be significantly correlated with BP~ND\_Logan~ (r = 0.97; *P*\<0.01) using data of both the first and second experiments, as shown in [Figure 1](#pone-0017723-g001){ref-type="fig"}. Although the slope of the regression line was slightly less than one, the BP~ND\_SRTM~ and BP~ND\_Logan~ had almost a one-to-one relationship. ::: {#pone-0017723-g001 .fig} 10.1371/journal.pone.0017723.g001 Figure 1 ::: {.caption} ###### Correlation between BP~ND\_Logan~ and BP~ND\_SRTM~. The binding potential estimated by Logan plot method and that estimated by the simplified reference tissue model method are represented as BP~ND\_Logan~ and BP~ND\_SRTM~, respectively. The solid line represents the regression line. Linear correlation is significant (r = 0.97, *P*\<0.01, y = 0.95x + 0.17). The dotted line represents the line of "y = x" for reference. ND: nondisplaceable. :::  ::: Each regional BP~ND~ of first and second PET scans for each dose of pramipexole is shown in [Figure 2](#pone-0017723-g002){ref-type="fig"}. After administration of pramipexole 0.25 mg, BP~ND\_Logan~ in the prefrontal cortex (*P* = 0.03, *t* = 3.15), medial (*P* = 0.01, *t* = 4.56) and lateral (*P* = 0.01, *t* = 3.78) thalamus, and amygdala (*P* = 0.02, *t* = 3.32) and BP~ND\_SRTM~ in medial (*P* = 0.01, *t* = 4.51) and lateral (*P* = 0.02, *t* = 3.33) thalamus decreased significantly. D~2~/D~3~ occupancy rates estimated with BP~ND\_Logan~ in the prefrontal cortex, medial and lateral thalamus, and amygdala were 10.3%±6.8%, 16.7%±6.9%, 14.9%±8.9%, and 20.4%±8.6%, respectively. Occupancy rates estimated with BP~ND\_SRTM~ in the medial and lateral thalamus were 10.3%±5.0% and 10.8%±6.4%, respectively. No significant difference was found in occupancy rates estimated with either BP~ND\_Logan~ or BP~ND\_SRTM~ between the medial and lateral thalamus, using paired Student *t* test. In the low-dose group with 0.125 mg of pramipexole and in the control group, there was no significant correlation between BP~ND\_SRTM~ of the two PET scans in all regions. ::: {#pone-0017723-g002 .fig} 10.1371/journal.pone.0017723.g002 Figure 2 ::: {.caption} ###### Changes in BP~ND~ between first and second PET scans in each extrastriatal region. For 0.25 mg dose group, each BP~ND~ was estimated by the Logan plot method and simplified reference tissue model method. For 0.125 mg and 0 mg dose groups, each BP~ND~ was estimated only by the simplified reference tissue model method. Pramipexole was orally administered 1--1.5 h before second PET scanning at doses of 0.25 mg, 0.125 mg. Each *P* value was estimated by paired Student\'s *t*-test between first and second PET scans. Significant differences were found only in the high-dose group (\* *P*\<0.05). BP: binding potential, ND: nondisplaceable. :::  ::: The time-activity curves of representative regions before and after administration of pramipexole 0.25 mg are displayed in [Figure 3](#pone-0017723-g003){ref-type="fig"}. Visually, the radioactivity levels of the putamen, entire thalamus and amygdala seemed to decrease after administration of pramipexole 0.25 mg. On the other hand, the radioactively level of the cerebellum seemed mostly unchanged between the first and second scans. Actually, *V* ~T~ on the cerebellum estimated by Logan plot method in the first and second PET scans was 4.55±0.68 and 4.67±0.97, respectively, and no significant difference was found between the two. ::: {#pone-0017723-g003 .fig} 10.1371/journal.pone.0017723.g003 Figure 3 ::: {.caption} ###### Average time-activity curves of representative regions of five subjects before (left) and after (right) administration of pramipexole 0.25 mg. Each point was normalized to the radioactivity of 185 MBq. The time-activity curve of the putamen is displayed for reference. Thalamus (▪) represents the entire thalamus. :::  ::: Discussion {#s4} ========== Pramipexole is a synthetic aminobenzothiazole derivative with selective actions mainly on D~2~ and D~3~ receptors, and it binds with the highest affinity to D~3~ receptors [@pone.0017723-Millan1], [@pone.0017723-Mierau1], [@pone.0017723-Piercey1], [@pone.0017723-Kvernmo1]. In the brain, the distribution of D~3~ receptors is known to be different from that of D~2~ receptors [@pone.0017723-Gurevich1], [@pone.0017723-Murray1], [@pone.0017723-Sokoloff1], [@pone.0017723-Bouthenet1], [@pone.0017723-Landwehrmeyer1], [@pone.0017723-Diaz1], although there are some differences in the relative proportion of D~2~ and D~3~ receptors between the previous studies. The D~2~ binding sites are widely detected with the highest concentration found in the striatum, followed by the nucleus accumbens, external segment of the globus pallidus, substantia nigra and ventral tegmental area. The distribution of D~3~ receptors is relatively restricted and D~3~ binding sites are enriched in the amygdala, nucleus accumbens, ventral striatum, substantia nigra, anteroventral nucleus of the thalamus and internal segment of the globus pallidus. Thus, D~2~ or D~3~ binding sites are located in the synapse of the afferent structures as well as the neurons of the efferent structures such as substantia nigra and ventral tegmental area. D~2~ or D~3~ receptors in the efferent structures are thought to act as autoreceptors, which could play an important role in regulating the activity of dopaminergic neurons [@pone.0017723-Gurevich1], [@pone.0017723-Diaz1]. On the other hand, in this study, the BP~ND~ in the substantia nigra or ventral tegmental area could not be quantified with reliability, because the structures were too small for usual ROI analysis of the dynamic data on the basis of the resolution of the PET scanner and the number of subjects was relatively small. For the same reasons, the ROIs were drawn over not each small nucleus but medial and lateral parts of the thalamus although the thalamus is known to have a great deal of regional heterogeneity in D~2~ and D~3~ expression. Also, the BP~ND~ in the striatal and its closely neighbor regions such as the ventral striatum and globus pallidus could not be quantified using ^11^C-FLB 457 because a long time more than a few hours is needed for reaching equilibration in the striatum [@pone.0017723-Loch1]. Based on the distribution of D~2~ and D~3~ receptors and the technical matters as described above, we investigated the binding sites of pramipexole especially in the limbic system, thalamus and cortical regions. To our knowledge, this is the first in vivo study that has investigated the relationship between a dopamine agonist and its binding sites in extrastriatal regions. This study showed that a single dose of pramipexole 0.25 mg decreased BP~ND\_Logan~ significantly in the prefrontal cortex, amygdala, and thalamus. The mesolimbic pathway begins in the ventral tegmental area of the midbrain and projects to the limbic areas, including the nucleus accumbens in the ventral striatum, amygdala, and hippocampus; it also projects to the cortical areas, including the prefrontal and cingulate cortices. The latter cortical pathway is called as the mesocortical pathway. Both pathways are known to be involved in the depressive state [@pone.0017723-Nestler1], [@pone.0017723-DiChiara1], [@pone.0017723-Nikolaus1], [@pone.0017723-Cabib1]. Indeed, the amygdala is important for emotional processing and its functional abnormalities are associated with depression [@pone.0017723-LeDoux1], [@pone.0017723-Drevets1], [@pone.0017723-Remy1], and frontal cortical dopamine function has been reported to be involved in depression [@pone.0017723-Espejo1], [@pone.0017723-Ohmori1], [@pone.0017723-Agren1]. In the mesolimbic pathway, D~2~ and D~3~ binding sites are predominant in the hippocampus and amygdala, respectively [@pone.0017723-Gurevich1], [@pone.0017723-Murray1], [@pone.0017723-Bouthenet1], and this difference may be one of the reasons that significant decrease of BP~ND\_Logan~ was not found in the hippocampus. The mesocortical pathway also has D~3~ receptors as well as D~2~ receptors [@pone.0017723-Bouthenet1], [@pone.0017723-Diaz1], although there is no detail report on the relative proportion of D~2~ and D~3~ receptors in that area. The thalamic dopaminergic system has been recently identified; this system is speculated to have a prominent role in depression, especially in regard to emotion, attention, cognition, and complex somatosensory and visual processing [@pone.0017723-Remy1], [@pone.0017723-GarciaCabezas1], [@pone.0017723-SanchezGonzalez1], [@pone.0017723-GarciaCabezas2]. Although this study could not find the significant difference between medial and lateral parts of the thalamus, D~3~ binding sites are relatively abundant and especially tend to be concentrated along the midline in the thalamus, while D~2~ binding sites are more homogeneously distributed [@pone.0017723-Gurevich1], [@pone.0017723-Bouthenet1]. On the basis of the relationship between the occupied sites by pramipexole, the distribution of D~2~ and D~3~ receptors and previous anatomical and functional reports on depression, it is reasonable to suggest that pramipexole may exert its antidepressive effects by activating D~2~R subfamily, especially the D~3~ receptor subtype, in these regions (prefrontal cortex, amygdala, and thalamus). With regard to the BP~ND\_SRTM~ method, after administration of a single dose of pramipexole 0.25 mg, binding in the medial and lateral thalamus decreased significantly as shown in the BP~ND\_Logan~ method, and bindings in both the prefrontal cortex and amygdala showed the tendency to decrease without significant difference. On the other hand, there was no significant difference between first and second PET scans in both the low-dose group and the control group. On the basis of the difference between the high-dose group and the other two groups, we speculate that the effects of pramipexole may be dose dependent, although it is impossible to confirm this finding only with our data. Despite the high correlation between BP~ND\_Logan~ and BP~ND\_SRTM~ methods as shown in [Figure 1](#pone-0017723-g001){ref-type="fig"}, the discrepancies between the both methods found in the high-dose group could be explained by the interindividual and intraindividual variability of each analytical method [@pone.0017723-Vilkman1], [@pone.0017723-Olsson2], [@pone.0017723-Sudo1]. However, BP~ND\_Logan~ estimated with artery blood samples should be regarded as more reliable and accurate than BP~ND\_SRTM~ estimated without artery blood samples, and the findings for the prefrontal cortex and amygdala, where only the BP~ND\_Logan~ method showed significance, were considered to be meaningful. Vilkman et al. conducted a test-retest analysis of ^11^C-FLB 457 PET scanning with 7 healthy volunteers (mean age ± SD = 29.0±6.9) and suggested that coefficient of variation (COV) of each extrastriatal region in BP~ND\_Logan~ and BP~ND\_SRTM~ methods was about 20% and the reproducibility of both methods was good [@pone.0017723-Vilkman1]. Consistent with Vilkman et al., our results of the control group corresponding to a test-retest analysis showed no significant difference in BP~ND\_SRTM~ of each region between first and second PET scans. In the control group, the COV of each region between the first and second experiments was similar, and the ranges of COV in the first and second experiments were 19.3%--33.9% and 16.3%--38.5%, respectively. Thus, the reproducibility of the BP~ND\_SRTM~ method in this study was good. Compared with previous studies [@pone.0017723-Vilkman1], [@pone.0017723-Olsson2], [@pone.0017723-Sudo1], the COV in our control group was somewhat larger. A possible explanation for this difference could be given by following two reasons. One is the relatively smaller number of subjects, and the other is the relatively larger variability in age because D~2~R subfamily in each extrastriatal region is known to show age-related decline [@pone.0017723-Kaasinen1], [@pone.0017723-Inoue1]. In vitro brain homogenate binding studies have demonstrated that D~2~R subfamily exists in two affinity states, i.e., high and low affinity states [@pone.0017723-DeLean1], [@pone.0017723-Sibley1], [@pone.0017723-George1]. The high affinity state is thought to represent the functional state, and agonists bind preferentially to D~2~R subfamily in the high affinity state, while antagonists have equal affinity for D~2~R subfamily in the high and low affinity states. In vivo competition studies between endogenous dopamine and a labeled agonist or antagonist ligand estimated the percentage of high affinity state to be about 60--70% [@pone.0017723-Seneca1], [@pone.0017723-Narendran1]. On the other hand, some recent in vivo studies indicated that most D~2~R subfamily is in the high affinity state at living conditions because the binding of exogenous unlabeled agonist to D~2~R subfamily in high or low affinity states could not be differentiated with either a labeled agonist or antagonist ligand [@pone.0017723-Finnema1], [@pone.0017723-Peng1]. Thus, the accurate proportion of the two states remains controversial. In this study, relatively low D~2~/D~3~ occupancy rates by pramipexole were mainly due to low dose of pramipexole. However, based on the two states theory, another reason may be because D~2~/D~3~ occupancy rates by agonist pramipexole were estimated by antagonist ligand ^11^C-FLB 457. One of the drawbacks of this study may be that we used the cerebellum as a reference region, in order to gain smaller variability and better reproducibility for the analysis of the PET data, compared with the two-tissue compartment four-rate constant model [@pone.0017723-Vilkman1]. Asselin et al. reported that using the cerebellum as a reference region could lead to underestimation of BP~ND~ and occupancy rate [@pone.0017723-Asselin1]. However, we showed no statistical difference in *V* ~T~ on the cerebellum estimated by Logan plot method before and after administration of pramipexole 0.25 mg, and at least our data, especially in the high-dose group, would be appropriate for the purpose of confirming the extrastriatal effects of a dopamine agonist. Other drawbacks of this study may be that we collected arterial blood samples only from the high-dose group, the number of subjects was relatively small and a dose of pramipexole was relatively low for safety, as described previously. In conclusion, we demonstrated that pramipexole binds to D~2~/D~3~ receptors in the prefrontal cortex, amygdala, and medial and lateral thalamus. These regions have been indicated to have some relation to depression and may be part of the target sites where pramipexole exerts its antidepressive effects. The authors are thankful to Ms. Hiroko Tsukinari and Mr. Kunpei Hayashi for their technical assistance. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Funding for this work was supported by a Grant-in-Aid for Scientific Research (B) No. 20390334 (Ishiwata K) by the Japan Society for the Promotion of Science (JSPS). JSPS had no further role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: K. Ishibashi K. Ishii K. Ishiwata. Performed the experiments: K. Ishibashi K. Ishii KO K. Ishiwata. Analyzed the data: K. Ishibashi KO K. Ishiwat. Contributed reagents/materials/analysis tools: K. Ishibashi K. Ishii KO HM K. Ishiwata. Wrote the paper: K. Ishibashi K. Ishiwata. Discussed the results: K. Ishibashi K. Ishii KO HM K. Ishiwata.

PubMed Central

2024-06-05T04:04:19.283319

2011-3-9

PMC3052388

Introduction {#s1} ============ Pancreatic cancer (PC) is one of the most lethal malignant diseases with worst prognosis, which is ranked as the fourth leading cause of cancer-related deaths in the United States [@pone.0017850-Jemal1]. Over the past two decades, numerous efforts have been made in improving treatment and survival PC patients but the outcome has been disappointing. This disappointing outcome is due to many factors among which *de novo* resistance (intrinsic) and acquired (extrinsic) resistance to conventional therapeutics (chemotherapy and radiation therapy) including gemcitabine alone or in-combination with other cytotoxic or targeted agents. Emerging evidence suggest that the resistance could in fact be due to the enriched existence of tumor initiating cells, also classified as cancer stem-like cells (CSC) in a tumor mass [@pone.0017850-Gaviraghi1]--[@pone.0017850-Mueller1]. The CSCs have the capacity of self-renewal and the potential to regenerate into all types of differentiated cells giving rise to heterogeneous tumor cell populations in a tumor mass, which contributes to tumor aggressiveness [@pone.0017850-Gaviraghi1]--[@pone.0017850-Mueller1]. Thus, the failure to eliminate these special cells is considered to be one of the underlying causes of poor treatment outcome with conventional therapeutics, suggesting that newer and novel therapeutic strategies must be developed for the targeted killing of drug resistant CSCs in order to eradicate the risk of tumor recurrence for improving the survival of patients diagnosed with PC. In search of novel yet non-toxic agents, attention has been focused on natural agents for several years. One such agent is curcumin (diferuloylmethane), which is derived from the plant Curcuma longa (Linn) grown in tropical Southeast Asia [@pone.0017850-Abas1]--[@pone.0017850-Narayan1]. Curcumin has been shown to inhibit the growth of a variety of tumor cells; however, the poor bioavailability of curcumin limits its application in the clinic. Recently, we have developed a novel synthetic analogue of curcumin, 3,4-difluoro-benzo-curcumin \[we named it as Difluorinated-Curcumin or in short CDF [@pone.0017850-Padhye1], [@pone.0017850-Padhye2]\], which showed greater bioavailability in pancreatic tissues, and also inhibited cell growth, DNA-binding activity of NF-κB, Akt, COX-2, and the production of PGE~2~ and VEGF, and caused induction of miR-200 and inactivation of miR-21 in PC cells [@pone.0017850-Ali1]. Since miR-200 is associated with the acquisition of epithelial to mesenchymal transition (EMT), which is also believed to be associated with CSCs or cancer stem-like cells, here we investigated the effects of CDF on CSC function. Here we report, for the first time, that CDF could inactivate many functions of CSCs including self-renewal capacity as demonstrated by the inhibition of sphere-forming (pancreatospheres) ability of drug-resistant PC cells, which was consistent with inactivation of CSC biomarkers such as CD44 and EpCAM. We also showed anti-tumor activity of CDF alone and in-combination with gemcitabine, which was consistent with inactivation of miR-21, and consequently increased expression of PTEN, attenuation of the DNA binding activity of NF-κB inhibition in the expression of COX-2, and activation in the expression of miR-200 in tumor remnants of a xenograft mouse model of human PC, all of which provide convincing *in vivo* activity of CDF which is consistent with *in vitro* findings. Results {#s2} ======= AsPc-1 and MIAPaCa-2 cell lines and their clones were chosen for this study because of their relatively resistant nature. The CSC characteristics of these cell lines using stem cell markers\' Lin28B and Nanog by RT-PCR, and EpCAM and CD44 by western blot showed an increase in expression level in the PC-GTR cell lines compared to their parental cell lines ([Figure 1](#pone-0017850-g001){ref-type="fig"}). Hence we chose these to test our hypothesis that CDF is more effective than curcumin even in resistant cell lines and also their resistant clones--GTR. ::: {#pone-0017850-g001 .fig} 10.1371/journal.pone.0017850.g001 Figure 1 ::: {.caption} ###### Comparative expression of Lin28B (A) and Nanog (B) mRNA by qRT-PCR showed increased expression in resistant cell lines compared to parental cell lines, supporting the CSC characteristics of these cell lines. The characteristics of CSCs were further confirmed by the protein expression of EpCAM and CD44 (**C**). :::  ::: CDF strongly prevents clonogenicity and invasion of PC cells compared to gemcitabine and curcumin {#s2a} ------------------------------------------------------------------------------------------------- We selected the concentration of 20 nmol/L of gemcitabine and 4 µmol/L of curcumin or CDF to conduct clonogenic assay following our previous publication [@pone.0017850-Ali1]. The results demonstrated that there was a significant reduction in clonogenicity of AsPC-1 and MIAPaCa-2 cells treated with curcumin and CDF, but not with gemcitabine ([Figure 2A](#pone-0017850-g002){ref-type="fig"}). However, CDF treatment had a much greater and significant reduction in colony formation compared to curcumin. AsPC-1-GTR and MIAPaCa-2-GTR cells had an 80% reduction of clonogenicity with CDF treatment, whereas, only 20--30% reduction of clonogenicity was observed with gemcitabine or curcumin treatment ([Figure 2A](#pone-0017850-g002){ref-type="fig"}). Overall, CDF treatment showed a significant reduction in clonogenicity of human PC cells, suggesting the superiority of CDF. ::: {#pone-0017850-g002 .fig} 10.1371/journal.pone.0017850.g002 Figure 2 ::: {.caption} ###### CDF and Curcumin decreased clonogenicity and invasion in AsPC-1, AsPC1-GTR, MIAPaCa-2, and MIAPaCa-2-GTR cells. Clonogenic assay (A) Invasion assay (B). Fluorescence of the invaded cells was read using ULTRA Multifunctional microplate reader (TECAN) at excitation/emission wavelengths of 530/590 nm. Basal level of ABCG2 expression showing relatively higher expression in drug resistant cell lines (**C**). :::  ::: CDF or curcumin treatment decreased PC cell migration and invasion. The results showed that 4 µmol/L of curcumin had minimal inhibition of invasion whereas similar concentration of CDF showed significant inhibition of invasion ([Figure 2B](#pone-0017850-g002){ref-type="fig"}). The basal level of ABCG2 expression was found in parental cell lines (*de novo* drug resistant cells); however, the level of expression of ABCG2 was further increased in drug resistant (acquired drug resistant) cell lines ([Figure 2C](#pone-0017850-g002){ref-type="fig"}). CDF inhibited viability of human PC cells more than curcumin and gemcitabine as evaluated by MTT assay {#s2b} ------------------------------------------------------------------------------------------------------ Initially MTT assay was conducted to examine the effect of different concentrations of gemcitabine (1 to 50 nmol/L), and curcumin or CDF (2--6 µmol/L) on cell survival after 72h of treatment (data not shown). Subsequently, 4 µmol/L of CDF or curcumin, and 20 nmol/L of gemcitabine were used to treat individually as well as in combination with gemcitabine for 72h. The results showed that CDF treatment in combination with gemcitabine caused a remarkable reduction of cell survival in all four cell lines compared to curcumin and gemcitabine combination treatment ([Figure 3](#pone-0017850-g003){ref-type="fig"}). Furthermore, analysis of drug combination treatment showed that the combination index after treatment with CDF in combination with gemcitabine was less than 1.00 ([Figure 3](#pone-0017850-g003){ref-type="fig"}), suggesting the synergistic effect of CDF combination. In contrast, the combination index with curcumin and gemcitabine was more than 1.00 ([Figure 3](#pone-0017850-g003){ref-type="fig"}), showing non-synergistic effect. Overall, these results suggest that CDF caused a much more significant reduction of cell survival in PC cells, compared to gemcitabine/curcumin alone or their combinations compared to CDF and gemcitabine combination. ::: {#pone-0017850-g003 .fig} 10.1371/journal.pone.0017850.g003 Figure 3 ::: {.caption} ###### CDF and its combination with gemcitabine inhibited cell viability. MTT assay was conducted in all four cell lines after 72h of treatment with CDF, curcumin, or its combination with gemcitabine. Untreated control has been assigned a value of 100%. The p value shown represents comparisons between single agent and their combinations by using paired t-test. The combination Index (CI) \<1 for CDF and Gemcitabine combination indicates synergism. :::  ::: CDF remarkably increased pancreatosphere disintegration of PC cells {#s2c} ------------------------------------------------------------------- To examine the effect of treatments on the sphere forming ability of PC cells (pancreatosphere) and disintegration of pancreatospheres, we conducted sphere disintegration assay for 10 days to generate the formation of pancreatospheres, followed by 5 days of drug treatment. The results show that there was a remarkable increase of sphere disintegration by curcumin and CDF treatment, not by gemcitabine treatment ([Figure 4](#pone-0017850-g004){ref-type="fig"}). However, the greatest effect on disintegration was observed in response to the CDF treatment ([Figure 4](#pone-0017850-g004){ref-type="fig"}), once again suggesting that CDF is much more superior in inhibiting the functions of cancer stem-like cells. ::: {#pone-0017850-g004 .fig} 10.1371/journal.pone.0017850.g004 Figure 4 ::: {.caption} ###### CDF remarkably increased disintegration of pancreatospheres in AsPC-1, AsPC1-GTR, MIAPaCa-2, and MIAPaCa-2-GTR cells. P values were calculated by the paired *t* test. :::  ::: CDF inhibited pancreatospheres formation in PC cells {#s2d} ---------------------------------------------------- To examine the effect of drugs on CSC self-renewal capacity in PC cells, we conducted sphere formation assay for 1 week and four weeks ([Figure 5A and B](#pone-0017850-g005){ref-type="fig"}). The results indicated that CDF in combination with gemcitabine completely eliminated pancreatospheres formation after four weeks of treatment compared to gemcitabine and curcumin combination in PC cells even in gemcitabine-resistant PC cells, suggesting that CDF may cause the pancreatospheres more sensitive to gemcitabine than that of curcumin treatment, and could be useful for targeted killing of CSCs. [Figure 5C](#pone-0017850-g005){ref-type="fig"} shows the effect of different concentration of gemcitabine and CDF on 2^nd^ passage of pancreatospheres in pre-treated primary pancreatospheres of AsPC-1 cells. CDF treatment remarkably inhibited 2^nd^ passage of pancreatospheres in a dose-dependent manner. Furthermore, CDF-pre-treated cells exhibited a greater effect than non-CDF-pre-treated cells. ::: {#pone-0017850-g005 .fig} 10.1371/journal.pone.0017850.g005 Figure 5 ::: {.caption} ###### CDF decreased the formation of pancreatospheres in AsPC-1 and AsPC-1-GTR cells 1 week treatment (A); 4 weeks treatment (B); AsPC-1 and CDF-pre-treated AsPC-1 cells treated with gemcitabine and CDF (C). A significant reduction in pancreatospheres was observed in cells treated with CDF shown by asterisk :::  ::: CDF decreased CD44 and EpCAM expression in pancreatospheres of PC cells {#s2e} ----------------------------------------------------------------------- We examined the effect of drugs on CSC biomarkers, CD44 and EpCAM in pancreatospheres of AsPC-1 and AsPC-1-GTR cells by confocal microscopy ([Figure 6](#pone-0017850-g006){ref-type="fig"}). The results indicate that CDF decreased CD44 and EpCAM expression in pancreatospheres, suggesting the inhibitory effect of CDF on pancreatosphere formation may be associated with the inhibition of CD44 and EpCAM expression. ::: {#pone-0017850-g006 .fig} 10.1371/journal.pone.0017850.g006 Figure 6 ::: {.caption} ###### CDF treatment decreased the expression of CD44 and EpCAM, the known markers of CSCs. Expression in pancreatospheres of AsPC-1 and AsPC-1-GTR cells was assessed by confocal microscopy (Magnification X250). :::  ::: CDF in combination with Gemcitabine inhibited Pancreatic Tumor Growth in vivo much more than curcumin combination {#s2f} ----------------------------------------------------------------------------------------------------------------- We have used a subcutaneous xenograft tumor model where the tumor was induced by MIAPaCa-2 cells in CB17-SCID mice. CDF treatment in combination with gemcitabine significantly inhibited tumor growth in MIAPaCa-2 tumors much more than curcumin and gemcitabine combination ([Figure 7A](#pone-0017850-g007){ref-type="fig"}) as well as compared to either untreated controls or those treated with a single drug. The mice did not show any weight loss during the treatment period (30 days), suggesting that these treatment had no major adverse effects on animals. ::: {#pone-0017850-g007 .fig} 10.1371/journal.pone.0017850.g007 Figure 7 ::: {.caption} ###### CDF exhibited anti-tumor activity in MIAPaCa-2 cells induced tumors in a xenograft mouse model, which was consistent with inhibition of NF-κB DNA binding, COX-2, miR-21, and caused re-expression of miR200 in tumor remnants. Anti-tumor activity and changes in tumor weight from each group of animals (**A**). The arrow indicates starting day of the treatment. NF-κB DNA binding activity of tumor tissues; and NF-κB competition control study with unlabeled NF-κB oligonucleotide (**B**). Western blots analysis of COX-2, PTEN and β-actin expression in tumor remnants (**C**); miR-21, miR-200b and miR-200c expression in tumor remnants as measured by real-time RT-PCR (**D**). P values were calculated by the paired *t* test. :::  ::: CDF with Gemcitabine significantly decreased NF-κB Activation *in vivo* {#s2g} ----------------------------------------------------------------------- NF-κB activation was determined in the CDF or curcumin, and/or gemcitabine treated tumor remnants derived from MIAPaCa-2 cells induced tumors as shown above. CDF and curcumin as single agent down-regulated NF-κB activation whereas gemcitabine activated NF-κB level, which was abrogated in combination treatment with CDF. The combination treatment of CDF with gemcitabine showed a significant decrease in NF-κB level compared to curcumin and gemcitabine treatment ([Figure 7B](#pone-0017850-g007){ref-type="fig"}), suggesting that the inactivation of NF-κB could be one of the molecular mechanisms by which CDF elicits its anti-tumor activity against PC tumors. CDF effects on protein expression *in vivo* {#s2h} ------------------------------------------- The COX-2, PTEN, and β-actin expression was determined by Western blot. A significant down-regulation in the expression of COX-2 was observed in both the combination, but the effect was more pronounced in CDF combination group. The expression of phosphatase and tension homolog (PTEN), a tumor suppressor gene was found to be decreased in MIAPaCa-2 cells; however, the expression of PTEN was up-regulated when treated with CDF ([Figure 7C](#pone-0017850-g007){ref-type="fig"}). These results suggest that CDF is much more effective than curcumin. Since PTEN is a known target of miR-21, which has been reported to be up-regulated in PC [@pone.0017850-Chan1]--[@pone.0017850-Meng1], we assessed the expression levels of miR-21 in tumor remnants as shown below. Modulation in the expression of miR-21 and miR-200 family *in vivo* {#s2i} ------------------------------------------------------------------- We determined the expression levels of miR-21, miR-200b and miR-200c in MIAPaCa-2 tumors by real time RT-PCR. Over-expression of miR-21 was observed in MIAPaCa-2 tumors whereas we found a significant reduction in the expression of miR-21 in tumors treated with either CDF alone or in combination with CDF and gemcitabine ([Figure 7D](#pone-0017850-g007){ref-type="fig"}). We further determined the expression levels of miRNA-200b and miR-200c in tumor tissues which are known regulators of EMT and found to be significantly low in MIAPaCa-2 cells ([Figure 7D](#pone-0017850-g007){ref-type="fig"}). In contrast, we found that the CDF treatment with or without gemcitabine combination showed increased expression of both miR-200b, and miR-200c, but the effect with curcumin or its combination was minimal, suggesting the superiority of CDF in suppressing the expression of miR-21, resulting in the re-expression of PTEN, and re-expression of miR-200 which could be responsible for the reversal of EMT phenotype in cells treated with CDF. Overall, these results suggest that the phenotypic characteristics of MIAPaCa-2 tumors are consistent with enriched population of CSCs and EMT characteristics, and these drug resistant cells could be killed either by CDF alone or in combination with gemcitabine. Pancreatospheres enhanced tumor growth *in vivo* {#s2j} ------------------------------------------------ Under traditional experimental conditions, we normally inject one million cells for assessing tumor growth; however, for investigating the greater potential of tumor growth by pancreatospheres, we injected only 5,000 cells in mice as a proof-of-concept study. The tumor weight was remarkably increased as the days progressed ([Figure 8A](#pone-0017850-g008){ref-type="fig"}). The level of miR-21 was increased between tumors implanted with one million of parental cells compared to pancreatospheres ([Figure 8B](#pone-0017850-g008){ref-type="fig"}). The animal was euthanized after 30 days because of tumor burden, and gross tumors are shown in [Figure 8C](#pone-0017850-g008){ref-type="fig"} indicating larger tumors as well as loco-regional lymph node metastasis whereas tumors derived from parental cells did not show any metastasis over a period of 30 days. The tumor-derived cells showed significant inhibition of pancreatospheres when treated with CDF ([Figure 8D](#pone-0017850-g008){ref-type="fig"}). Overall, these results suggest that CSCs (pancreatospheres) can be grown in mice and CDF could be useful for the killing of these drug resistant cells ([Figure. 8](#pone-0017850-g008){ref-type="fig"}). ::: {#pone-0017850-g008 .fig} 10.1371/journal.pone.0017850.g008 Figure 8 ::: {.caption} ###### Tumor growth pattern of pancreatospheres derived from MIAPaCa-2 cells. (**A**). 5,000 pancreatospheres were inoculated in mice using 1∶1 matrigel, progressive tumor growth over a period of 30 days. Moderate increase in the expression of miR-21 as measured by real-time RT-PCR was observed in tumors derived from pancreatospheres compared to tumors derived from parental cells by injecting one million cells and tumor was assessed over the same period of time (**B**). Photographs showing tumor growth, arrow points to tumor and asterisk (\*) refers to loco-regional lymph node metastasis whereas we did not find any metastasis when one million parental cells were injected (**C**). Tumor cells harvested from the tumors derived from pancreatospheres were treated with CDF showed significant inhibition in the formation of pancreatospheres (**D**). :::  ::: Discussion {#s3} ========== In this study, we have demonstrated that a synthetic analogue of curcumin, CDF, is significantly more effective compared to curcumin in the killing of gemcitabine-resistant pancreatic cancer (PC) cells that consists high proportion of cells with cancer stem cells (CSCs) or cancer stem-like cells characteristics. The inhibition of cell growth could in part be due to better cellular uptake, retention and reduced metabolic inactivation of CDF by PC cells, which is consistent with our published findings on cellular and animal pharmacokinetics data [@pone.0017850-Padhye1], [@pone.0017850-Padhye2]. Our previous reports indicate that CDF inhibits NF-κB and COX-2 activity in PC cells *in vitro* [@pone.0017850-Ali1]. Here we confirm these observations *in vivo* using a mouse xenograft model. Thus, the killing of gemcitabine-resistant PC cells by CDF is associated with inactivation of NF-κB and COX-2 signaling pathway which is very important because these pathways are known to contributes to drug-resistance of PC cells to chemotherapeutic agents [@pone.0017850-Arlt1]--[@pone.0017850-Zhang1]. CSCs comprises only a very small proportion of cells in a tumor mass and posses the ability to self-renew and give rise to differentiated tumor cells [@pone.0017850-Hermann1]--[@pone.0017850-Lee1], [@pone.0017850-Liu1]. The CSC theory has fundamental clinical implications especially because CSC has been identified in many malignant tumor tissues including pancreatic cancers and considered to be highly resistant to chemo-radiation therapy than differentiated daughter cells [@pone.0017850-Hermann1]--[@pone.0017850-Lee1], [@pone.0017850-Sarkar1]; however, CSCs isolated from human tumors are usually insufficient for further mechanistic studies. The existence of CSCs provides an explanation for the clinical observation that tumor regression alone may not correlate with patient survival [@pone.0017850-Creighton1] because of tumor recurrence, which is in part due to the presence of CSCs. Therefore, targeting self-renewal pathways and the killing of CSCs might provide more specific approach for eliminating cells that are the root cause of tumor recurrence. A potential challenge in this regard is the development of therapies that selectively affect CSCs while sparing normal stem cells that may rely on similar mechanisms for self-renewal. In this study, we have demonstrated that CDF not only inhibit cell growth of PC cells, but also inhibit CSC self-renewal capacity as assessed by sphere formation (pancreatospheres) assays. Therefore, CDF could have a greater potential to inhibit cancer growth as documented by our xenograft mouse model of gemcitabine resistant PC cells, which appears to be mediated via inhibition of CSC self-renewal capacity. Emerging evidence suggests the role of microRNA (miRNA) in many biological processes [@pone.0017850-Li1]--[@pone.0017850-VandenboomIi1]. Among many miRNAs, miR-21, commonly considered as an oncogene, is over-expressed in many solid tumors including PC and has been reported to be associated with tumor progression, poor survival and drug resistance [@pone.0017850-Chan1], [@pone.0017850-Dillhoff1], [@pone.0017850-Krichevsky1], [@pone.0017850-Shimosegawa1]. In our previous report, we have demonstrated that the expression of miR-21 is up-regulated in gemcitabine-resistant PC cells and that its expression can be significantly down-regulated by CDF treatment *in vitro* [@pone.0017850-Ali1]. The increased expression of miR-21 is known to be associated with inactivation of PTEN, a know tumor suppressor gene, resulting in activation of PI3K/Akt/mTOR signaling pathway, leading to aggressive tumor growth [@pone.0017850-Meng1], [@pone.0017850-Bunney1]. In this study, we confirmed that CDF treatment could results in the down-regulation of miR-21, resulting in the up-regulation of PTEN *in vivo*, suggesting that the anti-tumor activity of CDF is associated with up-regulation of PTEN resulting from the inactivation of miR-21 expression. In contrast to miR-21, miR-200 family is known as tumor suppressor and they are usually down-regulated in some tumors including PC and the loss of expression of miR-200 family contribute to the acquisition of EMT phenotype and drug resistance. Down-regulation of miR-200 by siRNA technique has been shown to be associated with EMT phenotype while re-expression of miR-200 can result in the reversal of EMT phenotype [@pone.0017850-Kong1], [@pone.0017850-Li3]. In our previous publication [@pone.0017850-Ali1], we demonstrated that CDF treatment could re-express miR-200 in PC cells. Here we showed, for the first time, that CDF can up-regulate miR-200b and miR-200c in tumor remnants *in vivo*, consistent with significantly greater inhibition of tumor growth in the xenograft mouse model when CDF was used in combination with gemcitabine. These results suggest that the anti-tumor activity of CDF is mediated via re-expression of miR-200 which may potentially results in the reversal of EMT phenotype and could also lead to overcome drug resistance in PC. In conclusion, CDF showed much more pronounced growth inhibitory effect, inhibited CSC self-renewal consistent with inactivation of CSC biomarkers (CD44 and EpCAM) in PC cells especially in gemcitabine-resistant PC cells compared to curcumin. In xenograft mouse model of human PC tumors induced by MIAPaCa-2 cells, CDF exhibits anti-tumor activity by regulating COX-2, PTEN, miR-21, miR-200, and NF-κB *in vivo*. These results strongly suggest that CDF could be a novel agent for the treatment of PC in general but gemcitabine-resistant PC in particular by attenuating the behavior of CSCs. Materials and Methods {#s4} ===================== Ethics Statement {#s4a} ---------------- This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Any animal found unhealthy or sick were promptly euthanized. The protocol was approved by the Committee on the Ethics of Animal Experiments of Wayne State University institutional Users Animal Care Committee (Permit Number: A-10-03-08). Cell Culture, Drugs and Reagents {#s4b} -------------------------------- Human pancreatic cancer cell lines AsPC-1, and MIAPaCa-2 were purchased from ATCC (Manassas, VA). These two cell lines AsPC-1 and MIAPaCa-2 were exposed to 200 nmol/L of gemcitabine and 5 µmol/L of tarceva (erlotinib) every other week for about 6 months to create gemcitabine and tarceva resistant (GTR) cell lines, named as AsPC-1-GTR and MIAPaCa-2-GTR, respectively. As a result, AsPC-1, AsPC-1-GTR, MIAPaCa-2, and MIAPaCa-2-GTR were chosen for this study based on their differential sensitivities to gemcitabine. All the cell lines have been authenticated (Applied Genomics Technology Center at Wayne State University) on March 13, 2009 and these authenticated cells were frozen for subsequent use. The method used for testing was short tandem repeat profiling using the PowerPlex 16 System from Promega. Gemcitabine and curcumin were purchased from Eli Lilly (Indianapolis, IN) and Sigma-Aldrich (St. Louis, MO), respectively. CDF was synthesized as described in our earlier publication [@pone.0017850-Padhye1], [@pone.0017850-Padhye2]. Gemcitabine was dissolved in water, whereas CDF and curcumin were dissolved in DMSO with a final concentration of 0.1% DMSO in medium. Antibodies {#s4c} ---------- Antibodies against ABCG2, and PTEN were purchased from Santa Cruz (Santa Cruz, CA). Antibody to COX-2 and β-actin was acquired from Cayman Chemicals (Ann Arbor, MI), and Sigma Chemicals (St. Louis, MO). Clonogenic assay {#s4d} ---------------- Clonogenic assay was conducted to examine the effect of drugs on cell growth in PC cells, as described previously [@pone.0017850-Ali1]. 5×10^4^ cells were plated in a six-well plate. After 72h of exposure to 20 nmol/L of gemcitabine, 4 µmol/L of CDF or curcumin, the cells were trypsinized, and 1,000 single viable cells were plated in 100-mm Petri dishes. The cells were then incubated for 10 to 12 days at 37°C in a 5% CO~2~/5% O~2~/90% N~2~ incubator. Colonies were stained with 2% crystal violet and counted. Invasion assay {#s4e} -------------- The invasive activity of cells was tested by using BD BioCoat Tumor Invasion Assay System (BD Biosciences, Bedford, MA) according to the manufacturer\'s protocol as described previously [@pone.0017850-Wang1]. Briefly 5×10^4^ cells were seeded with serum free medium supplemented with curcumin or CDF into the upper chamber and bottom wells were filled with complete medium in the system. The, fluorescence was read using Microplate Reader (TECAN) at 530/590 nm and were photographed. Cell survival assay {#s4f} ------------------- MTT assay was conducted using AsPC-1, AsPC-1-GTR, MIAPaCa-2, and MIAPaCa-2-GTR as described previously [@pone.0017850-Ali1] after 72 h of treatment. Combination index and Isobologram for combination treatment were also calculated and plotted using CalcuSyn software (Biosoft, Cambridge, United Kingdom) to determine synergy based on the method of Chou and Talalay [@pone.0017850-Chou1]. Sphere formation/disintegration assay {#s4g} ------------------------------------- Single cell suspensions of cells were plated on ultra low adherent wells of 6-well plate at 1,000 cells/well in sphere formation medium [@pone.0017850-Yu1]. After 7 days, the spheres termed as "pancreatospheres" were collected by centrifugation and counted [@pone.0017850-Yu1]. For sphere disintegration assay, 1,000 cells/well were incubated for 10 days, following 5 days of drug treatment, which examined the effect of drug treatment on disintegration of pancreatospheres as described previously [@pone.0017850-Yu1]. The pancreatospheres were collected by centrifugation and counted under a microscope. Confocal microscopy {#s4h} ------------------- Single cell suspensions of AsPC-1 and AsPC-1-GTR cells were plated using ultra low adherent wells of 6-well plate at 3,000 cells/well in sphere formation medium. After 7 days of treatment, the pancreatospheres were collected by centrifugation, washed with 1xPBS, and fixed with 3.7% parformaldehyde. CD44 and EpCAM antibodies were used for immunostaining assay, as described previously [@pone.0017850-Kong1]. The CD44 or EpCAM-labeled pancreatospheres were photographed by confocal microscopy (Leica TCS SP5) using software LAS AF 1.2.0 Build 4316. Protein extraction and Western blot analysis {#s4i} -------------------------------------------- Proteins were extracted from all four cell lines and also from animal tumor tissues as described previously [@pone.0017850-Ali1]. Relative level of ABCG2 was evaluated for all four cell lines. The effects on COX-2, PTEN and β-actin expression were evaluated on tumor tissues by Western blot analysis. as described previously [@pone.0017850-Ali1]. Animal Experiments {#s4j} ------------------ The animal protocol was approved by the Animal Investigation Committee, Wayne State University, Detroit, MI. Female CB17 SCID mice 4 wks old were purchased from Taconic Farms (Germantown, NY) and fed Lab Diet 5021 (Purina Mills, Inc., Richmond, IN). Small fragments of the MIAPaCa-2 xenograft were implanted subcutaneously and bilaterally into mice for the drug-efficacy trials. Once the mice developed palpable tumors, they were randomly selected into the following treatment groups (n  =  5/group): (1) untreated control; (2) CDF (5 mg/mouse/day), intragastric once daily for 12 days; (3) curcumin (5 mg/mouse/day), intragastric once daily for 12 days; (4) gemcitabine (1 mg/mouse/day), intravenous every third day for a total of three doses; (5) CDF and gemcitabine using the doses indicated above; (6) curcumin and gemcitabine using the doses as indicated above. Tumor measurements and changes in weight were performed and tissue was stored at −70°C for RNA and protein extraction. Electrophoretic Mobility Shift Assay (EMSA) assay for assessing the DNA binding activity of NF- κB {#s4k} -------------------------------------------------------------------------------------------------- Nuclear proteins were prepared from tumors tissue using a Dounce homogenizer with 400 µl of ice cold lysis buffer extracted as described earlier [@pone.0017850-ElRayes1]. EMSA was performed using the Odyssey Infrared Imaging System with NF-κB IRDye labeled oligonucleotide from LI-COR, Inc. (Lincoln, NE). Ten µg of the nuclear protein extract was used as described earlier [@pone.0017850-ElRayes1]. The NF-κB competition control study was conducted using unlabeled NF-κB consensus oligonucleotide. The samples were loaded and run at 30 mA for 1 hour. The gel was scanned using Odyssey Infrared Imaging System (LI-COR, Inc.). TaqMan miRNA Real-Time Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR) {#s4l} ------------------------------------------------------------------------------- To determine the expression of miRNAs (miRNA-200b, miR-200c, and miR-21) in MIAPaCa-2 tumors, we used TaqMan MicroRNA Assay kit (Applied Biosystems) following manufacturer\'s protocol. 5 ng of total RNA was reverse transcribed and real-time PCR reactions were carried as described earlier [@pone.0017850-Li3], using Smart Cycler II (Cepheid). Data were analyzed using C~t~ method and were normalized by RNU6B expression. Growth of CSC in xenograft model {#s4m} -------------------------------- 5,000 pancreatospheres were isolated and implanted in mice with 1:1 matrigel. The growth rate was observed for a period of 30 days. RNA was extracted from the tumor tissue for subsequent molecular assays as presented under figure legend. Statistical Analysis {#s4n} -------------------- Comparisons of treatment outcome were tested for statistical difference by the paired *t* tests. Statistical significance was assumed at a *P* value of \<0.05. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**National Cancer Institute, NIH grants 5R01CA131151, 3R01CA131151-02S1, and 5R01CA132794 (F.H. Sarkar). We thank Puschelberg and Guido foundations for their generous financial contribution. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: BB SA DK SHS ZW SB AA SP PAP FHS. Performed the experiments: BB SA DK SHS ZW AA. Analyzed the data: BB SA ZW AA FHS. Contributed reagents/materials/analysis tools: PAP SP FHS. Wrote the paper: BB SA FHS.

PubMed Central

2024-06-05T04:04:19.285976

2011-3-9

PMC3052389

Introduction {#s1} ============ Excitation-contraction coupling (ECC) in cardiac ventricular muscle occurs via the calcium induced calcium release (CICR) mechanism (for review see [@pone.0017901-Bers1]). In CICR, the action potential opens L-type calcium channels (dihydropyridine receptors -DHPRs) in the surface membrane and t-tubular membranes to activate the sarcoplasmic reticulum (SR) ryanodine receptors (RyRs) to cause release of Ca^2+^ from the SR. The SR release forms Ca^2+^sparks which summate to produce the cell-wide increase in Ca^2+^ which regulates force [@pone.0017901-Cannell1]. The close and precise alignment of DHPRs and RYRs in a structure called the "couplon" is thought to be critical to efficient CICR [@pone.0017901-Cannell2], [@pone.0017901-FranziniArmstrong1]. The transverse tubular system (t-tubules) of cardiac ventricular myocytes enables highly synchronized calcium release by CICR at couplons [@pone.0017901-Cheng1]. The t-tubules form a network of plasma membrane invaginations [@pone.0017901-Fawcett1], [@pone.0017901-Forbes1], [@pone.0017901-Forssmann1] and in human ventricular myocytes, couplons have a transverse spacing of ∼0.8 µm [@pone.0017901-Soeller1]. In a dog model of heart failure, a reduction in t-tubule density was observed [@pone.0017901-Balijepalli1], [@pone.0017901-He1] and similar alterations have been reported in other heart failure animal models (see below). It is unclear whether these alterations in t-tubule density result in a reduction in couplon density, although such an effect might help explain reduced contractility in heart failure. The idea that microscopic structural alterations in the organization of DHPRs and RyRs within the couplon might play a role in heart failure was first raised in a computer modelling study [@pone.0017901-Cannell3] and this idea has gained traction from rodent studies that have detected a reduced ability of the L-type Ca^2+^ current to trigger CICR [@pone.0017901-Gomez1], [@pone.0017901-Song1]. Dyssynchronous or non-uniform CICR associated with a change in t-tubule organisation has also been observed in murine [@pone.0017901-Louch1], canine [@pone.0017901-Meethal1] and porcine models [@pone.0017901-Heinzel1] of heart failure. Ca^2+^ handling abnormalities have been described in various forms of human heart failure including idiopathic dilated cardiomyopathy (IDCM) [@pone.0017901-Beuckelmann1], [@pone.0017901-Lindner1], [@pone.0017901-Louch2]. Despite the importance of the t-system in calcium handling, few studies have examined the organization of the t-system in normal and failing human myocardium [@pone.0017901-Cannell4] although t-tubule-related structures are known to be changed [@pone.0017901-Heling1], [@pone.0017901-Kostin1]. Recently the work of Lyon *et-al* (2009) using scanning ion conductance microscopy and confocal imaging of di-8-ANNEPs labeling suggested a loss of t-system structure in isolated failing human cardiomyocytes [@pone.0017901-Lyon1]. In contrast, the work of Ohler *et-al* (2009) showed no significant change in the structure of t-system between isolated normal and failing cardiomyocytes by using 2-photon imaging of di-8-ANNEPs labeling [@pone.0017901-Ohler1]. A potential limitation in using isolated myocytes is those cells most affected by the disease process may be preferentially destroyed during the isolation procedure, introducing a selective bias into the cells that are analyzed [@pone.0017901-Ohler1]. To gain greater insight into the possible changes in t-system structure and changes in EC coupling proteins, we have used intact tissue and immunohistochemistry combined with high resolution confocal microscopy and quantitative analysis of labelling patterns in ventricular muscle from both healthy human hearts and from patients with idiopathic (non-ischemic) dilated cardiomyopathy (IDCM). This avoids the potential for a selection bias due to cell isolation but also has the advantage that multiple components of the t-system can be labelled. Importantly it also allows a detail assessment of DHPR and RyR labelling, for which no data on the [human]{.underline} heart currently exists. Our results demonstrate that disarrangement in ECC structures occurs in [human]{.underline} heart failure. The results underpin related animal studies of heart failure but also reveal some unexpected differences in the geometric organization of key structures and proteins in the normal and diseased human heart. Methods {#s2} ======= Myocardial tissue {#s2a} ----------------- Human cardiac tissue was obtained with the written informed consent of heart transplant recipients and from families of organ donors for normal tissue as approved by the New Zealand Health and Disability Ethics Committee (NTY/05/08/050). Normal tissue samples were obtained from 4 donors that could not be matched to recipients they had a mean age of 60 (54--68), and had normal echo and ECG. Medications for donors included inotropic support in intensive care unit, and one patient was recieving felodipine. Diseased tissue came from 7 end-stage failing hearts from patients diagnosed with idiopathic dilated cardiomyopathy which were class III or IV on the New York Heart Association heart failure scale. These patients had a mean age of 33 years (16--56), mean ejection fraction (by echo) of 17% (5--30), mean heart rate of 84 bpm (70--100), and mean systolic blood pressure 95 mm Hg (90--105). Common drug therapies included renin-angiontensin-aldosterone system modulators (cilazapril, spironolactone, quinapril, losartan), diuretics (furosemide, bendroflumethiazide), beta blockers (carvedilol, metoprolol) and anticoagulants (warfrin, asprin). Other cardiac medications included digitalis, amiodarone, simvastatin, and dobutamine. Reduced contraction strength in three failing hearts was confirmed by tagged MRI and tissue was obtained directly from the operating theatre (e.g. [Figure 1](#pone-0017901-g001){ref-type="fig"}). Samples were taken from the midline region of the left ventricle, fixed with 1% paraformaldehyde in PBS overnight at 4°C, cryoprotected in 30% sucrose, frozen with liquid nitrogen chilled isopentane and stored at −80°C until further processing. 30 µm thick frozen sections were cut on a Leica CM 1900 cryostat and mounted on either glass slides or coverslips for subsequent antibody labelling. ::: {#pone-0017901-g001 .fig} 10.1371/journal.pone.0017901.g001 Figure 1 ::: {.caption} ###### Panel A shows tagged MRI short axis image through the middle of a failing heart in end diastole. Tracked grids are shown as yellow overlay with sampled region marked in red. Panel B shows circumferential shortening (CC%) of the region indicated in the panel A compared to a similar region in normal human heart. Panel C shows RyRs (green) and WGA (red) labeling of diseased myocytes from the region shown in panel A. Image is z projections of 10 slices with z depth of 2.5 µm from deconvolved image stack. Scale bar is 10 µm. :::  ::: Indirect immunofluoresence labeling {#s2b} ----------------------------------- Labelling with wheat germ agglutinin (WGA) was carried out by hydrating and incubating with WGA conjugated to Alexa Fluor 594 (Molecular Probes/Invitrogen). Double labelling for membrane proteins and WGA included blocking with FX signal enhancer (Molecular Probes/Invitrogen) for 1 hour. The sections were then incubated with the following antibodies: dihydropyridine receptor (DHPR, ACC-003 (a Cav 1.2 II-III loop specific antibody shown here) and ACC-013 (raised against Cav 1.2 N terminal residues 1-46 see supplementary [Figure S1](#pone.0017901.s001){ref-type="supplementary-material"}), Alomone Labs), ryanodine receptor 2 (RyR, MA3-916, Affinity Bioreagents), sodium calcium exchanger (NCX, AB2869, Abcam) or caveolin-3 (Cav-3, AB2912, Abcam) overnight at 4°C. Sections were washed three times in PBS and incubated with secondary antibodies labelled with Alexa Fluor 488 and wheat germ agglutinin Alexa Fluor 594 (Molecular Probes/Invitrogen). *f*-actin labelling followed the protocol for double labelling of membrane proteins and WGA, except WGA was replaced with phalloidin conjugated to Alexa Fluor 594 (Molecular Probes/Invitrogen). Controls consisted of omitting primary antibodies, WGA and phalloidin to check for non-specific labelling of secondaries and cross-talk. For DHPR labelling, the specificity of antibody was further tested by pre-incubation with the supplied blocking peptide (Cav1.2~848-865~ peptide) and running a Western blot against human cardiac tissue homogenate which gave major band at 240 kda and minor band 100 kDa, confirming results provided in the manufacturer\'s data sheet. Imaging {#s2c} ------- Fluorescent images of single and double labelled tissue sections were recorded with a Zeiss LSM410 confocal microscope using a Zeiss 63x NA 1.25 oil-immersion objective. Images of tripled labelled tissue sections were recorded with a Olympus FV1000 confocal microscope using a Olympus 60x NA 1.35 oil-immersion objective. Three-dimensional images stacks were acquired (with lateral and axial pixel spacings of 0.1 µm and 0.2 µm respectively). T-tubule correlation analysis {#s2d} ----------------------------- To quantify changes in t-tubule orientation, a 10×10 µm region of WGA labelled sections were taken perpendicular to the plasma membrane and z-projected over 4 image planes (0.8 µm). The analysis involved rotating the images in a clockwise direction in 1° increments and calculation of the autocorrelation. The autocorrelation was then cross-correlated with a line (1 pixel wide) to give the amount of signal in the direction set by the image rotation. To test this analysis, two model t-tubule images were created and correlated to the reference image, one with tubules running at 90° angle ([Figure 2A](#pone-0017901-g002){ref-type="fig"}) and the other with tubules running at 45° angle ([Figure 2B](#pone-0017901-g002){ref-type="fig"}). ::: {#pone-0017901-g002 .fig} 10.1371/journal.pone.0017901.g002 Figure 2 ::: {.caption} ###### WGA labelling of t-tubules in normal and failing human ventricular myocytes. The top row shows images from normal cells in longitudinal and transverse sections (left to right) and corresponding images from diseased tissue is shown in the lower two rows. (**A**) Longitudinal sections of normal tissue shows uniformly spaced t-tubules. Occasional axial elements can also be seen. (**B**) A magnified view of the region shown by the box in **A**. (**C**) Normal myocyte in transverse section. A radial "spoke-like" organization of t-tubules is apparent. (**D**) Enlarged view of the region shown by the box in **C**. (**E**, **I**, **K**) Longitudinal sections from three different diseased cells, demonstrating the range of t-tubular morphologies found in disease with corresponding (**F**, **J**, **L**) magnified views. Note that while the enlarged view in L appears relatively normal, other regions with the same cell (**K**) are clearly abnormal. (**G**) Transverse section showing that, while the general direction of diseased tubules is radial, tubules are more disorganized. (**H**) Magnified view of the region shown by the box in **G**. Images are projections of 5 slices with z depth of 1 µm. Scale bars in overview images are 10 µm and in close up images 2 µm. :::  ::: Image deconvolution {#s2e} ------------------- The PSF of the microscope was determined using 100 nm fluorescent latex beads (Molecular Probes - Invitrogen, NZ) in the mounting medium. Iterative constrained deconvolution was performed using the Richardson-Lucy maximum likelihood algorithm as described elsewhere [@pone.0017901-Soeller2]. RyR and f-actin analysis {#s2f} ------------------------ To quantify the volume of the cell occupied by contractile apparatus an 8 µm×8 µm×4 µm volume of *f-actin* labelling was converted into binary mask using an automatic thresholding algorithm [@pone.0017901-Ridler1]. The volume of the binary mask was then determine and expressed as percentage of labelled pixels over total. The number of RyR clusters were then measured for the same volume by using a detection algorithm [@pone.0017901-Soeller1] and expressed per unit volume of *f-actin*. RyR and DHPR colocalization {#s2g} --------------------------- The amount of RyR and DHPR colocalization was determined in 8 µm×8 µm×2.5 µm volumes using Pearson correlation coefficient and Manders coefficients for each label [@pone.0017901-Zinchuk1]. For the calculation of Manders coefficients an automated threshold detection was used [@pone.0017901-Ridler1]. To provide an estimate of the actual overlap between DHPR and RyR signals along the t-tubule, the intensity plot of RyR and DHPR labelling were measured along WGA labelled tubule. A mask of RyR labelling was created by determining full width half max of RyR intensity peaks. DHPR labelling within the RyR mask was considered colocalised. Statistical analysis {#s2h} -------------------- Data was expressed as mean and standard error or median and interquartile range for non-parametric data. Differences between normal and diseased cells were tested with Students t-test or the Mann-Whitney U test where data did not have equality of variance as determined by Levene\'s test. *P*\<0.05 was considered statistically significant. Results {#s3} ======= WGA labelling of normal and failing human cardiac myocytes {#s3a} ---------------------------------------------------------- WGA was used to visualize t-tubules of ventricular myocytes since it binds to sialic acid residues of the cell membrane glycocalyx [@pone.0017901-Peters1]. The longitudinal section of healthy heart ([Figures 2A and 2B](#pone-0017901-g002){ref-type="fig"}) shows a regular spacing of parallel t-tubules that run approximately perpendicular to the cell surface, although there are t-tubules that bifurcate to connect adjacent z-lines. In transverse sections of healthy myocytes ([Figures 2C and 2D](#pone-0017901-g002){ref-type="fig"}) the t-tubules run in a radial direction giving rise to a "spoke-like" appearance, however the t-tubules tend to curve as they approach the centre of the cell. In longitudinal sections of failing heart tissue, there is a marked reduction in the degree of order within the t-system ([Figures 2E and 2F](#pone-0017901-g002){ref-type="fig"}). In diseased cells, t-tubules are often seen to run in an oblique direction and can cross several sarcomeres. In diseased cells there is also a less uniform distribution of tubules with axial tubules becoming more common. In transverse sections of diseased cells ([Figures 2G and 2H](#pone-0017901-g002){ref-type="fig"}), the direction of the t-tubule was generally toward the cell centre in a spoke-like pattern but appears more disjointed and tortuous. A correlation method was used to determine the distribution of angles made by WGA labelled t-tubules relative to the cell membrane (see [Methods](#s2){ref-type="sec"} for further details). As shown in [Figure 3A and 3B](#pone-0017901-g003){ref-type="fig"}, the analysis could detect the predominant angle of tubular structures. [Figure 3C](#pone-0017901-g003){ref-type="fig"} shows that the median angle of t-tubules relative to the cell surface was 95±8° in normal cells (n = 12 cells from 4 hearts), indicating that in normal myocytes the t-tubules are organized in a radial fashion, supporting the visual impression gained from [Figure 2](#pone-0017901-g002){ref-type="fig"}. In diseased tissue, the alteration in t-tubule organization resulted in an increase in the width of the distribution of t-tubule angles and the median angle was reduced to 57±9° (P\<0.001, Mann-Whitney U, n = 18 cells from 6 hearts). In addition, there was a much greater variability in the t-tubule direction as seen by the marked increase in the width of the distribution of t-tubule angles. ::: {#pone-0017901-g003 .fig} 10.1371/journal.pone.0017901.g003 Figure 3 ::: {.caption} ###### Analysis of changes in ventricular myocyte t-tubule geometry in human heart failure. Correlation analysis of t-tubule angle: To illustrate the analysis, two model t-tubule images were created and correlated to the reference image, one with tubules running at 90° angle (panel **A**) and the other with tubules running at 45° angle (panel **B**), demonstrating that the angle with peak correlation intensity matches the direction of tubules in the model images. (**C**) Correlation analysis of longitudinal sections to measure t-tubule orientation referred to the cell surface. Analyses of normal cells are shown in the left panel and diseased tissue are shown on the right. Note the change in predominant tubule angle and increased spread of t-tubule angles in the diseased tissue. (**D**) Full width at half maximum (FWHM) measurements of control and diseased tubules corrected for optical blurring. In normal myocytes, tubule widths are smaller and have a narrower size distribution compared to diseased myocytes (p = 0.0003 see text). :::  ::: T-tubule diameters were measured by the full width at half maximum (FWHM) of the intensity profile across their images. The size distribution of 120 control t-tubules (from 12 control cells and 4 hearts) and 150 diseased tubules (from 15 cells and 5 hearts) are shown in [Figure 3D](#pone-0017901-g003){ref-type="fig"}. The mean diameter of WGA labelled t-tubules of diseased cells was 0.47±0.02 µm (n = 15) and was significantly larger (P\<0.001, Mann-Whitney U) than the mean control tubule size of 0.38±0.01 µm (n = 12). Although close to the diffraction limit, these values are larger than reported for rat ventricular myocytes [@pone.0017901-Soeller2] but similar to those reported for rabbit [@pone.0017901-SavioGalimberti1]. In addition, the variance of t-tubule diameters was larger in the diseased tissue (P\<0.001, Levene\'s test) so that the change in t-tubule diameter was not simply due to an increase in diameter of all tubules. These data show that diseased myocytes can undergo significant remodelling of t-tubular structure with a loss of radial directed t-tubules and an increase in oblique and longitudinally oriented t-tubules. In addition, a wide variation is t-tubule morphology is apparent with some t-tubules appearing almost normal while others flatten, fragment and/or become more filigree in appearance (even within a single diseased cell). T-tubules without WGA labeling {#s3b} ------------------------------ There appeared to be large regions of the cell without obvious t-tubule labelling but without a suitable lipid marker (for fixed tissue) it is possible that some t-tubules might not be WGA labelled due to a change in glycosylation and therefore WGA labelling. To examine this possibility, tissue from 3 control and 3 diseased hearts were dual labelled for WGA and various sarcolemmal membrane proteins: dihydropyridine receptors (DHPR), sodium calcium exchangers (NCX), and caveolin 3 (Cav3). This labelling suggested that the t-system was more extensive than indicated by WGA alone (see [Figure 4](#pone-0017901-g004){ref-type="fig"}). For all labels examined, the apparently wider t-tubules, which contained strong WGA labelling, are extended and connected by thinner tubules which contain DHPR, NCX and Cav3. These thinner tubules formed ring-like structures which would be consistent with these tubules surrounding myofibrils (see [Figure 5](#pone-0017901-g005){ref-type="fig"}). In diseased myocytes, similar labelling patterns are seen but the organization of the thinner tubules was generally disrupted. ::: {#pone-0017901-g004 .fig} 10.1371/journal.pone.0017901.g004 Figure 4 ::: {.caption} ###### Sarcolemmal protein labelling with WGA in normal and failing human cardiac myocytes. Panels are arranged so that the left column corresponds to normal and the right column diseased tissue. **A** and **B** show longitudinal sections of control and diseased myocytes respectively, labelled for DHPRs (green) and WGA (red). The DHPR labelling is more continuous across the cell than the WGA label. Panels **C** and **D** show transverse images labelled for NCX (green) and WGA (red). In control myocytes (**C**), finer tubules containing only the NCX label can be seen to connect to larger spoke like tubules containing both labels. Diseased (**D**) myocytes have similar labelling pattern but many connections between larger tubules appear broken. Panels **E** and **F** show transverse sections labelled for Cav3 (green) and WGA (red). (**E**) Again, finer tubules labelled with only Cav3 connect larger tubules containing both labels. Diseased myocytes (**F**) have similar labelling pattern but with a reduction in fine Cav3 labelling between larger tubules. Images are z projections of 4 slices with z depth of 0.8 µm from deconvolved image stacks. Scale bars are 2 µm. :::  ::: ::: {#pone-0017901-g005 .fig} 10.1371/journal.pone.0017901.g005 Figure 5 ::: {.caption} ###### Comparison of myofilaments, t-tubules and RyR localization. Panels **A** and **B** show transverse sections of normal and diseased myocytes respectively, labelled for Cav3 and *f*-actin. The contractile machinery is surrounded by Cav3 labelling at this plane which was selected to be centred at the z-line. Gaps in the centre of contractile bundles contain isolated Cav3 labelling (small white arrow) and inspection of adjacent sections show that this arises from axial tubules that join transverse tubules outside the presented image plane. A connection between adjacent radial t-tubules is indicated by the large arrow. Panels **C** and **D** show transverse sections of normal and diseased tissue (respectively) labelled for RyR and *f*-actin. It is apparent that each contractile bundle is surrounded by several RyR clusters. All images are projections of 4 slices with z depth of 0.8 µm. Panel **E** shows analysis of cell volume occupied by *f*-actin and no significant difference (P = 0.81) exists between normal and diseased tissue. Panel F shows analysis of the number of RyR clusters per µm^3^ of *f*-actin and there was a significant reduction in the density of RyR clusters in disease cells (\*P = 0.03). Scale bars are 2 µm. :::  ::: Relationship of t-tubular network to myofibrils {#s3c} ----------------------------------------------- Since Cav3 provided a reliable marker of surface membrane, we compared the distribution of Cav3 labelling to that of myofibrillar *f*-actin ([Figure 5A and 5B](#pone-0017901-g005){ref-type="fig"}). Myofibrils generally followed the spoke-like arrangement of t-tubules so the contractile machinery appeared as flattened wedges or plates in cross section. In places, Cav3 labeled t-tubules formed anastamoses between the radial t-system spokes (large arrow) and some isolated Cav3 labeling could be seen surrounded by myofibrils (small arrow). Examination of adjacent sections (not shown) showed that these tubules were axially oriented and joined radial t-tubules at the next z-line. A similar relationship between Cav3 and *f-*actin was seen in diseased cells ([Figure 5B](#pone-0017901-g005){ref-type="fig"}) but the overall pattern of labelling was more disjointed with the wedge-like pattern of contractile machinery becoming less obvious. Relationship between ryanodine receptors and contractile machinery {#s3d} ------------------------------------------------------------------ Sections were labelled for RyR and *f-*actin ([Figure 5C and 5D](#pone-0017901-g005){ref-type="fig"}) and analysed in three dimensions (3D). This labelling showed that each myofilament bundle was surrounded by several RyR clusters and there was no obvious change in the labelling between control and diseased cells (beyond a subtle reorganization of the myofibrils). To assess the volume of the cell occupied by contractile apparatus, a binary mask of *f-*actin labelling was created and expressed as percentage of total area. This analysis ([Figure 5E](#pone-0017901-g005){ref-type="fig"}) showed no significant change (P = 0.81, Student\'s t-test) in the fraction of cell volume occupied by the contractile apparatus in control myocytes (55±1%, n = 9 cells, 3 hearts) compared to diseased myocytes (55±1%, n = 18 cells, 6 hearts). The number of RyR clusters in the analysis volume was counted using a cluster detection algorithm [@pone.0017901-Soeller1] which showed a small but significant (P = 0.03, Student\'s t-test) reduction in the number of RyR clusters per unit volume of contractile apparatus in diseased myocytes ([Figure 5F](#pone-0017901-g005){ref-type="fig"}). In control, there were approximately 0.94 clusters per µm^3^ *f*-actin which would equate to a cell wide cluster density of 0.52 RyR clusters per µm^3^ cell volume (given 55% of cell volume occupied by myofilaments). RyR and DHPR labeling {#s3e} --------------------- [Figure 6](#pone-0017901-g006){ref-type="fig"} shows tissue dual labeled for DHPRs and RyRs. The RyR labelling was more punctate than the DHPR labelling and in longitudinally orientated normal myocytes, DHPR labelling occurred predominantly at the Z-line. In diseased cells, some DHPR label was displaced from the Z-line. This visual observation was supported by colocalisation analysis by Pearson\'s and Mander\'s coefficients (in 3D) which showed a significant reduction in DHPR colocalization ([Table 1](#pone-0017901-t001){ref-type="table"}). This may reflect the axial distortion of the t-system in diseased myocytes as shown above and/or the reduction in number of RyR clusters shown in [Figure 5F.](#pone-0017901-g005){ref-type="fig"} ::: {#pone-0017901-g006 .fig} 10.1371/journal.pone.0017901.g006 Figure 6 ::: {.caption} ###### RyR and DHPR cluster colocalization is reduced in failing human cardiac myocytes. Panels **A** and **B** show images of control myocytes in transverse and longitudinal orientation respectively, labelled for RyR (red) and DHPR (green). Panels **C** and **D** show images of diseased myocytes in transverse and longitudinal orientation respectively, labelled for RyR (red) and DHPR (green). Images are z projections of 4 optical sections with z depth of 1 µm. Scale bars are 2 µm. Panel **E** shows an exemplar region of DHPR and RyR labelling in a normal cell, this image is a single optical section. The white line indicates the position of intensity readings along a WGA labelled t-tubule (label not shown for clarity). Panel **F** shows the intensity profile of DHPR and RyR labelling along the white line shown in panel **E**. Similar results were seen with an alternative DHPR antibody (see [methods](#s2){ref-type="sec"} and supplementary [Figure S1](#pone.0017901.s001){ref-type="supplementary-material"}). :::  ::: ::: {#pone-0017901-t001 .table-wrap} 10.1371/journal.pone.0017901.t001 Table 1 ::: {.caption} ###### Colocalisation analysis of DHPR and RyR labeling. ::: {#pone-0017901-t001-1} Tissue Cell orientation number Pearson\'s correlation DHPR colocalisation(Manders) RyR colocalisation(Manders) ---------- ------------------ -------- ------------------------ -------------------------------------------- ----------------------------- ------------------------------------------- ----------- ---------- Normal Longitudinal 9 0.33±0.03 [\*](#nt101){ref-type="table-fn"}P = 0.001 0.46±0.03 [\*](#nt101){ref-type="table-fn"}P\<0.001 0.24±0.03 P = 0.09 Diseased Longitudinal 18 0.22±0.01 0.24±0.02 0.18±0.02 Normal Transverse 9 0.33±0.06 P = 0.72 0.26±0.03 P = 0.34 0.30±0.02 P = 0.14 Diseased Transverse 18 0.32±0.03 0.37±0.05 0.29±0.03 \*Significantly different at p\<0.05. ::: The distribution of RyR and DHPR labels along a WGA-labelled t-tubule from a normal myocyte is illustrated in [Figure 6E and 6F](#pone-0017901-g006){ref-type="fig"} (see also Supplementary [Figure S1](#pone.0017901.s001){ref-type="supplementary-material"}). Both labels were non-uniform, and a density plot ([Figure 6F](#pone-0017901-g006){ref-type="fig"}) shows that local concentrations of DHPR were not perfectly aligned with the RyR clusters although considerable overlap between both labels was present. To quantify these data, we constructed a mask by thresholding the RyR label at 50% maximum local intensity and integrated the fraction of DHPR label signal within the mask boundaries. In these transverse sections, we found that 51±3% (9 t-tubules from 9 cells) of the DHPR labelling was within the region defined by RyR labeling, which was not significantly different in disease (50±4%, p = 0.81), in agreement with the co-localization measured by Pearson\'s and Manders coefficients ([Table 1](#pone-0017901-t001){ref-type="table"}). Given the smaller fraction of membrane area occupied by jSR (e.g. ∼21% in rabbit [@pone.0017901-Page1]), suggests that there was a preferential localization of DHPRs near RyRs, albeit less pronounced than in other species (see [Discussion](#s4){ref-type="sec"}). Labelling with an alternative antibody directed toward the N-terminus of the DHPR showed similar labeling patterns (see Supplementary [Figure S1](#pone.0017901.s001){ref-type="supplementary-material"}). Discussion {#s4} ========== This report presents a detailed analysis of the architecture of cellular structures critical to excitation-contraction coupling in normal and failing human myocardium. Quantitative analysis of WGA labelled t-tubules in ventricular myocytes demonstrated major alterations in the tubular network in diseased cells compared to normal cardiac myocytes. Although some changes in the pattern of WGA labelling have been previously noted [@pone.0017901-Cannell4] these changes have not been quantified. An unexpected finding was that WGA labelled only a portion of the t-system as labels for the sarcolemmal membrane proteins DHPR, NCX and Cav3 revealed a more extensive network of fine tubules. Disruption of t-tubule architecture in diseased cells was associated with a small loss of RyR clusters and reduced colocalization between RyRs and DHPRs. This indicates remodelling of the t-tubular network in the failing human heart may contribute to calcium handling abnormalities seen in heart failure and underpins many functional studies in animal models. Structure of the human t-system {#s4a} ------------------------------- Two previous studies that have examined the t-system in human cardiac myocytes, although not at the level of detail shown here. One of the papers described major changes [@pone.0017901-Lyon1] while no major changes were found in a second study [@pone.0017901-Ohler1] Our results show that major changes in the structure of t-tubules can occur in human heart failure, in general agreement with conclusions from animal models of heart failure. However, it is important to note that there is a greater variation in t-tubule pathology seen in human myocytes (see [Figure 2](#pone-0017901-g002){ref-type="fig"}) than the generally simpler loss of t-tubules described for animal models. We suspect that the high degree of t-tubule morphology shown here may reflect the long(er) duration of the disease process in humans. Our results also largely agree with previous immunohistochemical examination of cytoskeletal proteins [@pone.0017901-Heling1], [@pone.0017901-Kostin1], [@pone.0017901-Kaprielian1], [@pone.0017901-Kostin2], [@pone.0017901-Vatta1]. A potential explanation for the lack of changes reported by Ohler *et-al* (2009) [@pone.0017901-Ohler1] might be arise from selection bias associated with myocyte isolation, which would be expected to "purify" healthier myocytes. We have found that quite normal t-tubular structure can be found adjacent to severely disrupted t-tubules. Importantly, our results show that WGA labels only a portion of the t-tubular network in human myocytes since DHPRs, NCX and Cav3 labels revealed an additional fine network of sub-resolution structures that connected the larger WGA-labelled t-tubules ([Figure 4](#pone-0017901-g004){ref-type="fig"}). This indicates that the t-system in humans is more extensive than might be deduced from WGA labelling alone. This is a noticeable difference between human and animal cells, as, for example there is a high degree of overlap between Cav3 and WGA labeling in rat myocytes [@pone.0017901-Jayasinghe1]. Although the cause and role of the changes in glycocalyx as reported here within t-tubules is unclear, it is possible that the expansion of the glycocalyx drives the remodelling of parts of the t-tubule network. It is also possible that some hypertrophic signals, which lead to the expansion of cytoskeletal proteins [@pone.0017901-Aquila1] in heart failure, also drive t-system remodelling. In contrast to our results, Kaprielian et al. [@pone.0017901-Kaprielian1] suggested there was a proliferation of t-tubules in diseased myocytes but the extent of this proliferation was not quantified. However, we suggest that their different conclusion may have arisen from the dilation of t-tubules in disease reported here which would have made them more obvious. In connection with this point, Figure 8D in their paper showed a dilated non-radial t-tubule in EM similar to the observations reported here. We conclude that our results are more in line with studies of the t-system in animal models and humans which have described either a loss and/or disorganisation of the t-system [@pone.0017901-Balijepalli1], [@pone.0017901-He1], [@pone.0017901-Cannell4], [@pone.0017901-Lyon1], [@pone.0017901-Dibb1] rather than an expansion. It should also be noted that, even in normal tissue, the general architecture of the human t-system is different (i.e. predominantly spoke-like) to the t-system rete described for rat [@pone.0017901-Soeller2] and we did not detect frequent varicosities along the length of t-tubules as recently described for rabbit [@pone.0017901-SavioGalimberti1]. T-tubule disorganisation and E-C coupling proteins {#s4b} -------------------------------------------------- In addition to a marked reorganisation of the t-system within the failing human heart, we found a small reduction in the number of RyR clusters per unit myofibril area and a decrease in colocalization between DHPRs and RyRs. The importance of the local geometric arrangement of DHPRs and RyRs has been highlighted in modelling studies where EC coupling gain is diminished with small distortions in junctional geometry [@pone.0017901-Cannell3]. Our results support this idea, although the level of colocalization observed between DHPR and RyR clusters in control human ventricular myocytes suggests that DHPRs are not almost exclusively restricted to the couplon as has been suggested to be the case in rat [@pone.0017901-Scriven1]. Nevertheless on average, approximately 51% of the DHPR label was closely associated with the RyR clusters so that these data still support the idea of tight regulation of RyRs by DHPRs. Our analysis of RyR and DHPR longitudinal colocalization demonstrated an apparent weakening of the overlap between DHPR and RyR clusters in failing human heart cells. That such a change was not observed in transverse section may be explained by the poorer axial resolution of the microscope and the change in t-tubule orientation towards a axial direction. Even if the couplons themselves did not move away from z-lines, the presence of non junctional DHPRs in displaced t-tubules would still lead to a reduction in colocalisation. In a rat model of heart failure, Song *et. al* [@pone.0017901-Song1] reported a similar (∼20%) loss of colocalization between RyRs and DHPRs coupled with t-tubule remodelling and "dysynchronous" calcium spark formation. This effective displacement of mean DHPR location from RyRs would result in a reduced ability of the Ca^2+^ current to activate SR release and therefore a loss of E-C coupling "gain" [@pone.0017901-Cannell2], [@pone.0017901-Gomez1] While we were unable to measure the functional consequences of the changes in key EC coupling proteins/structures measured here, the possibility that the failing human myocyte may suffer reduced EC coupling fidelity seems very likely. However, it should also be noted that the spatial reorganization of the t-tubule network in human heart failure that we report here is quite different to the loss of transverse tubules and an increase axial tubules reported for the SHR model [@pone.0017901-Song1]. We suggest that future study of the mechanisms underlying the control of t-tubule remodelling may lead to new directions for treating heart failure. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **DHPR immunohistochemistry with an alternative antibody gives generally similar DHPR labelling patterns relative to RyR.** This antibody was raised against a peptide corresponding to residues 1--46 of rabbit Cav1.2a N terminus (ACC-013, Alomone). Panels **A** and **B** show longitudinal images of control and diseased myocytes respectively, labelled for RyR (red) and DHPR (green). Images are z projections of 4 optical sections with z depth of 1 µm. Scale bars are 2 µm. It is apparent that RyR and DHPR labelling is non-uniform with only moderate colocalisation. In disease cells greater proportion of DHPR appears between Z-lines. The white line indicates the position of intensity readings along a WGA (blue) labelled t-tubule. Panel C shows the WGA labeling (blue) together with RyR (red) and DHPR (green) in a single transverse section. As shown for the other antibody (ACC-003) in the manuscript, there was no significant difference in the measured overlap between RyR and DHPR labelling (as a fraction of total DHPR labeling) between normal and diseased samples (46%±7 vs. 55%±4 respectively n = 6 in each group, p = 0.46). Panel **D** shows an intensity profile of DHPR and RyR labelling along the white line following the t-tubule shown in panel **C**. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Isuru Dilshan and Cherrie Kong for laboratory assistance, Professor Stephen Munn, Helen Gibbs and Janice Langlands for assistance in obtaining donor tissue and transplant recipients and donor families for donating tissue. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the Health Research Council of New Zealand (05/049) <http://hrc.govt.nz/>. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MBC DJC PRR. Performed the experiments: DJC. Analyzed the data: MBC DJC CS. Contributed reagents/materials/analysis tools: MBC CS PRR. Wrote the paper: DJC MBC.

PubMed Central

2024-06-05T04:04:19.288789

2011-3-9

PMC3052398

References 60 and 71 contained errors. The correct references are: \"60. Rasheed MA, Unsworth RKF (2011) Long-term climate associated dynamics of a tropical seagrass meadow: implications for the future. Marine Ecology Progress Series. 422: 93-103\" and \"71. Unsworth RKF, Cullen LC (2010) Recognising the necessity for Indo-Pacific seagrass conservation. Conservation Letters 3: 63-73\" **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.291657

2011-3-08

PMC3052399

The 11th author\'s name was entered incorrectly. The correct name is: Julio Viera **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.292429

2011-3-08

PMC3052419

Introduction {#s1} ============ Unilateral movements are mainly controlled by the primary motor cortex (M1) of the contralateral hemisphere. However, previous studies have reported that also primary motor cortex ipsilateral (M1~ipsi~) to the moving body side can undergo task-related modulations of activity. Using transcranial magnetic stimulation (TMS) it was shown that performing a forceful, isometric contraction with one hand induced a significant increase of corticomotor excitability in M1~ipsi~, even when the other hand was at rest such that no overt electromyographic (EMG) activity was observed [@pone.0017742-Hortobagyi1]--[@pone.0017742-Hess1]. Even though facilitation of M1~ipsi~ has been shown repeatedly for strong, isometric contractions, somewhat inconsistent results were obtained during phasic hand or finger movements: Brief, phasic movements requiring only low forces, induced rather inhibition than excitation of M1~ipsi~ [@pone.0017742-Liepert1], [@pone.0017742-Sohn1]. By contrast, rhythmical flexion-extension movements of one wrist increased corticomotor excitability of M1~ipsi~ such that this facilitation mirrored the phasic activity of homologous muscles of the moving hand [@pone.0017742-Carson1]. Ziemann and Hallett [@pone.0017742-Ziemann1] and Tinazzi and Zanette [@pone.0017742-Tinazzi1] reported increased corticomotor excitability for M1~ipsi~ which was larger when subjects performed complex finger sequences as compared to simple movements. Functional imaging studies have revealed that the activation of motor areas ipsilateral to the moving hand is asymmetric such that the left hemisphere is more activated when a complex movement task is executed with the ipsilateral, left body side than the right hemisphere during movements with the right body side, or when simple tasks are executed [@pone.0017742-Callaert1]--[@pone.0017742-Verstynen1] for a review see [@pone.0017742-Serrien1]. However, these asymmetries were most consistently reported for areas upstream from M1 and, particularly, for parietal and premotor regions, probably because functional imaging offers only limited sensitivity for studying M1. Only a few studies tested hemispheric asymmetries of ipsilateral M1 facilitation using TMS. Stinear et al [@pone.0017742-Stinear1], applied TMS to both hemispheres while the ipsilateral hand performed isometric contractions at different force levels, however, no hemispheric differences were observed. By contrast, Ziemann and Hallett [@pone.0017742-Ziemann1] applied TMS to M1~ipsi~ of each hemisphere while right-handed subjects performed a thumb-to-middle-finger opposition task (simple task) versus a sequence of opposition movements from the thumb to index, middle, ring or little finger. They found significant hemispheric asymmetries such that M1~ipsi~ facilitation was larger when the task was executed with the left compared to the right hand and particularly, when subjects had to perform the complex sequencing task. In the present study we performed two experiments to further investigate behavioural and neural determinants of hemispheric asymmetries in M1~ipsi~ facilitation. In the first experiment we investigated whether ipsilateral facilitation would be modulated when rhythmical wrist movements were executed in isolation or in the context of a simple or difficult hand-foot coordination pattern, and whether this modulation would differ for the left versus right hemisphere. It has been shown previously that task complexity of these multilimb coordination tasks depends on the spatiotemporal pattern between hand and foot movements such that coordination control is easier when both limbs move into the same direction (in-phase) than when limbs move into opposite directions (anti-phase)[@pone.0017742-Baldissera1]--[@pone.0017742-Swinnen2]. Importantly, using this paradigm, subjects perform identical wrist movements across all conditions, such that the output of the investigated muscles can be kept constant while task complexity is systematically varied. In the second experiment we tested whether hemispheric asymmetries in the modulation of M1~ipsi~ excitability due to coordinative task complexity might result from reduced intracortical inhibition. Previous studies indicated that the M1~ipsi~ facilitation emerges, at least partly, at the cortical level: First, Tinazzi and Zanette [@pone.0017742-Tinazzi1] reported increased M1~ipsi~ excitability only for TMS which activates corticospinal neurons transsynaptically but not for transcranial electric stimulation (TES) that activates corticospinal axons directly, probably within white matter structures. Second, it was shown that M1~ipsi~ is facilitated while responses to cervicomedullary stimulation of the descending tracts were unchanged [@pone.0017742-Hortobagyi1], [@pone.0017742-Carson1]. Third, paired-pulse short-interval intracortical inhibition (SICI), which is mediated by GABAergic, cortical interneurons [@pone.0017742-Kujirai1] was decreased due to forceful isometric contractions of the opposite hand [@pone.0017742-Muellbacher1] and this decrease became stronger the more force was applied [@pone.0017742-Perez1]. Moreover, at high force levels, the decrease of SICI was correlated to an increase in interhemispheric inhibition suggesting that intracortical and transcallosal pathways interact to control M1~ipsi~ [@pone.0017742-Perez1]. However, it is currently unknown whether SICI of the ipsilateral M1 also changes during phasic movements and whether these potential modulations depend on the complexity of the task and/or which body side performs the task. Methods {#s2} ======= 2.1 Subjects {#s2a} ------------ Twelve healthy volunteers (age 20--23 yrs, 12 female) participated in *experiment 1* and eight subjects (age 19--24 yrs, 3 female) in *experiment 2*. They were all right-handed, as assessed by the Edinburgh Handedness Inventory [@pone.0017742-Oldfield1] and naïve to the task. Subjects were screened for contra-indications for TMS such as epilepsy, migraine, implants in the head as well as for overt sensorimotor and other major physical deficits. The study was approved by the Medical Ethics committee of the University Hospital at the K.U. Leuven in accordance to the Declaration of Helsinki (1964) and each subject read and signed a written informed consent prior to the experiment. 2.2 General setup {#s2b} ----------------- Measurements were performed while either the right arm and/or leg were active whereas the left arm and leg were resting, or vice versa. Subjects were comfortably seated in a low chair with their legs outstretched on a soft support. The resting arm and leg were fully supported such that subjects could completely relax. The leg of the actively moving foot was positioned such that the calf was supported but the ankle could move without restrictions. The elbow of the actively moving arm was supported such that the forearm was held upright and the wrist could move freely ([figure 1](#pone-0017742-g001){ref-type="fig"}). Subjects were instructed to fully extend their wrist, to ensure that both the wrist flexor and extensor had to be activated to move the wrist against gravity in the respective part of the movement cycle. Displacement data of the moving hand and foot were measured by shaft encoders (HP Hewlett-Packard, Malaysia) mounted to custom made wrist and ankle orthoses. The axis of each orthosis was aligned to the axis of rotation of the wrist/ankle joint and flexion and movements were measured with a frequency of 100 Hz and a spatial resolution of 0.18°. ::: {#pone-0017742-g001 .fig} 10.1371/journal.pone.0017742.g001 Figure 1 ::: {.caption} ###### Experimental setup (A) and illustration of performance of the anti-phase coordination pattern (B). Abbreviations: flexor carpi radialis (FCR); extensor carpi radialis (ECR); tibialis anterior (TA); triceps Surae (TS). Consent to publication was obtained from the participant shown at the photograph. :::  ::: The electromygraphic (EMG) activity of the Tibialis Anterior (TA) and Triceps Surae (TS) of the resting leg and the Extensor Carpi Radialis (ECR) and Flexor Carpi Radialis (FCR) of both arms were recorded throughout the experiment (Mespec 8000, Mega Electronics Ltd., Kuopio, Finland) with two disposable Ag--AgCl surface electrodes (Blue Sensor P-00-S, Ambu, Ølstykke, Denmark) placed over the muscle belly and one reference electrode placed over a bony structure. Each EMG channel was measured with a frequency of 5 kHz, amplified, filtered (30--1500 Hz) and displayed on a computer screen in front of the subject (CED Power 1401, Cambridge Electronic Design, Cambridge, UK and Signal 3.03 software). Using this online EMG feedback, subjects were trained to fully relax their resting limbs even when movements were performed with the other body side ([figure 2](#pone-0017742-g002){ref-type="fig"}). Also during the experiment, EMG activity of the resting limbs was closely monitored by the subjects and the experimenters and trials were repeated when overt EMG activity was observed. Due to this setup they had only partial vision of their limb movements and only in the periphery of their field of view. Thus, the movement tasks were mainly executed under proprioceptive control. ::: {#pone-0017742-g002 .fig} 10.1371/journal.pone.0017742.g002 Figure 2 ::: {.caption} ###### Typical example of the EMG signals registered from the moving wrist muscles (upper two panels), the resting wrist muscles (middle two panels) and the resting foot muscles (lower two panels) during rhythmical hand flexion and extension movements. Arrows indicate TMS stimulation timed such that the first one was positioned within the extension and the second within the flexion burst. TMS was applied in the hemisphere ipsilateral to the moving hand and evoked clear MEPs in the contralateral resting FCR and ECR. Abbreviations are identical to [figure 1](#pone-0017742-g001){ref-type="fig"}. :::  ::: 2.3 Experiment 1 {#s2c} ---------------- ### 2.3.1 Task {#s2c1} The participants were instructed to execute five different experimental conditions: (1) rhythmical plantarflexion and dorsiflexion movements with the foot (called "foot flexion" and "foot extension", respectively, in the remainder of the manuscript), (2) rhythmical flexion and extension movements with the hand while the foot was resting, (3) rhythmical in-phase coordination, i.e. hand and foot were simultaneously flexed and extended, (4) rhythmical anti-phase coordination, i.e. the hand was extended when the foot was flexed and vice versa, and (5) a rest condition. All movements were produced rhythmically as paced by an auditory metronome at 1 Hz. Each trial lasted 20 seconds and started with the wrist and ankle in neutral position. Subjects were instructed to synchronize their movements to the metronome such that the wrist was flexed on the beat in the hand (HAND), in-phase (IN) and anti-phase condition (ANTI). When only the foot was moved (FOOT condition), ankle flexion was performed on the beat. Additionally, there was a rest condition in which subjects did not move, but remained completely relaxed while the metronome produced a 1 Hz rhythm. For each condition, several training trials were performed prior to the TMS measurements to practice the different tasks while relaxing the non-involved body side. During the experiment, each of the movement conditions was performed 8 times and rest 4 times in pseudo-random order. The experiment consisted of two separate sessions that took place at two different days. Half of the subjects performed all tasks first with their left and then with their right body side while the order was reversed for the other half of the subjects. ### 2.3.2 TMS-procedure {#s2c2} Single-pulse transcranial magnetic stimulation (TMS) was delivered through a figure-of-eight shaped stimulation coil (mean diameter of each wing, 70mm) connected to a Magstim 200 (Magstim Company Ltd., Carmarthenshire, UK) and motor evoked potentials (MEPs) were recorded in the contralateral, resting wrist muscles. The subject wore a tight-fitting cotton cap with a 1-cm grid. The coil was placed tangentially to the scalp over the primary motor cortex with the handle pointing backward and 45° away from the midline. The stimulator produced a near monophasic wave form and, with this coil orientation, the induced current was directed from posterior-lateral to anterior-medial which activates corticospinal neurons predominantly via horizontal corticocortical connections [@pone.0017742-DiLazzaro1]. The hotspot of the FCR, i.e. the optimal position to elicit maximal MEPs in the contralateral limb, was determined and marked on the swimming cap. The rest motor threshold (RMT) was determined as the stimulation intensity that elicited a MEP peak-to-peak amplitude \>0.05 mV in the relaxed FCR in at least five out of ten consecutive stimuli [@pone.0017742-Rossini1]. Even though the parameter setting procedures focused on the FCR, ECR parameters were assumed to be satisfactorily similar, due to the overlapping representations of forearm flexors and extensors [@pone.0017742-Schieber1]. During the experiment, the stimulation intensity was set at 125% of the FCR RMT and the coil was placed over the primary motor cortex ipsilateral to the moving body side (M1~ipsi~) to record MEPs in the contralateral, resting muscle. The participants performed 36 trials in total and four stimulations (on average 5s apart) were applied during each trial. The absolute timing of the stimulation was randomized, but the pulses were applied either 150ms before the beat of the metronome (i.e. targeting the flexion burst of the moving hand) or 400 ms after (i.e. targeting the extension burst). Each movement condition was repeated 8 times (2 stimulations during extension burst, 2 during flexion burst per trial) and the rest condition was repeated 4 times (4 stimulations during rest). Thus, there were 16 stimulations for each condition and each flexion/extension phase. ### 2.3.3 Data-analysis of EMG and kinematics of the moving limbs {#s2c3} For the hand-foot coordination conditions, the relative phase angle between limbs was calculated from the displacement data of hand and foot by: where θ~hand~ is the phase of the hand movement at each sample; X~hand~ is the position of the hand after rescaling to the interval \[−1,1\] for each movement cycle, and dX~hand~/dt is the normalized instantaneous velocity. The mean continuous relative phase and its standard deviation was determined for each trial and the absolute phase error was calculated as the absolute difference between the mean relative phase and the target phase (i.e 0 deg for the IN and 180 deg for the ANTI condition). Additionally we determined the mean movement amplitude and cycle duration for each limb and condition. For the ECR and FCR of the actively moving hand, EMG was determined as the root-mean-square (RMS) value of the EMG signal during the last 50 ms prior to stimulation. ### 2.3.4 Data-analysis of corticomotor excitability in M1 ipsilateral to the moving limbs {#s2c4} Corticomotor excitability of the resting FCR and ECR, i.e. elicited via the hemisphere ipsilateral to the moving limbs, was determined by the peak-to-peak amplitude of the MEPs. All EMG traces were visually inspected and MEPs were removed from subsequent analysis (9.2% in total) (a) when overt EMG activity emerged in the resting hand or foot muscles 50 ms prior to stimulation, (b) when the stimulation fell outside the targeted EMG burst of the active, homologous muscle or (c) when subjects performed the wrong coordination pattern. MEP-amplitudes of the FCR and ECR were averaged within each subject such that one mean value was calculated for each muscle, condition and phase (only for the movement conditions). Finally, the MEP-amplitudes were normalized relative to the rest MEPs (MEP~norm~ = MEP~movement\ condition~/MEP~rest~). ### 2.3.5 Statistical analysis {#s2c5} All statistical analyses were performed with Statistica 8.0 (StatSoft, Inc., Tulsa, USA). Differences between RMT of the left versus right hemisphere were tested by dependent t-tests. The absolute phase error and the standard deviation of the relative phase were subjected to an analyses of variance for repeated measurements (repeated measures ANOVA) with the within factors *moving body side* (left, right) and *coordination pattern* (IN, ANTI). Movement amplitude and cycle duration were analyzed by a repeated measures ANOVA with the factors *moving body side* (left, right), *condition* (single limb, IN, ANTI) and *limb* (hand, foot). The active EMG of the moving limbs as well as the normalized MEP amplitudes generated from the hemisphere ipsilateral to the moving body side were subjected to a repeated measures ANOVA with the factors *hemisphere* (left, right), *condition* (HAND, FOOT, IN, ANTI), *muscle* (FCR, ECR) and *contraction phase* of the homologous, contralateral muscle (active, passive). Significant effects were further tested with Fisher LSD posthoc tests. The criterion for statistical significance was α = 0.05. Descriptive statistics will be reported as mean and standard error in text and figures. 2.4 Experiment 2 {#s2d} ---------------- ### 2.4.1. Task {#s2d1} The task was identical to experiment 1, however subjects performed only 3 different conditions: HAND, ANTI and REST, each lasting 12 s. Prior to testing, subjects familiarized themselves with the task and practiced the coordination pattern while the resting body side was completely relaxed. In the main experiment each of the REST, HAND and ANTI conditions was tested 14 times, in a randomised order and, per trial, two TMS stimulations were applied that were at least 4 s apart. ### 2.4.2 TMS procedure {#s2d2} TMS was applied over M1~ipsi~ with a 70 mm figure eight coil connected to a Magstim 200 stimulator through the BiStim Module (Magstim, Whitland, Dyfed UK). After the hotspot of the ECR was located, RMT and the active motor threshold (AMT) were determined, defined as the minimal stimulus intensity necessary to produce MEPs larger than 0.1 mV in at least five out of ten consecutive trials while subjects maintained a slight voluntary contraction of the ECR at approx. 3% of the maximal voluntary contraction. SICI was measured by a double-pulse paradigm such that in half of the trials a superthreshold test stimulus (TS) was preceded by a subthreshold conditioning stimulus (CS) [@pone.0017742-Kujirai1]. The interstimulus interval was set to 2.5 ms, the optimal interval to induce intracortical inhibition [@pone.0017742-Fisher1]--[@pone.0017742-Stinear2]. As experiment 1 has shown that excitability of M1~ipsi~ differs depending on the task and hemisphere tested, we attempted to adjust the test stimulus (TS) intensities such that MEP responses with an amplitude of 0.7--1.0 mV were evoked by single pulse stimulation for all conditions. The conditioning stimulus (CS) intensity was set such that MEP amplitude was reduced by approximately 50% when the TS was preceded by the CS at REST. Intensities were adjusted in the beginning of each session and kept constant for the remainder of the experiment. In each trial one single (TS) and one double pulse stimulation (CS+TS) were delivered. In the REST condition, this occurred at random time points and with an interval between single and double pulse stimulation of at least 4 seconds. In the HAND and ANTI condition the EMG burst of the moving ECR was detected online and used to trigger the stimulation over M1~ipsi~. Stimulation was provided during the burst because experiment 1 showed M1~ipsi~ facilitation to be larger when the homologous muscle of the other body side is active. The interval between the first and second stimulation was at least 4 seconds. A trial was repeated (a) if the experimenters observed overt EMG activity in the resting body side, (b) if only one stimulation pulse was given, or (c) in case the subject did not perform the coordination task correctly. ### 2.4.3 Data-analysis {#s2d3} EMG and kinematics of the actively moving body side were analyzed analogous to experiment 1. For the TMS data collected from M1~ipsi~, MEP amplitudes were determined as described above and the percentage of intracortical inhibition was calculated by %SICI = (MEP~TS~-MEP~CS+TS~)/MEP~TS~\*100 (note that large values indicate a high level of inhibition). ### 2.4.4 Statistics {#s2d4} Differences between hemispheres for RMT, AMT, CS and the coordination performance of ANTI were analysed by dependent t-tests. Active EMG of the ECR of the moving hand was analysed by a repeated measures ANOVA with the factors *moving body side* (left, right), *condition* (HAND, ANTI) and *stimulation* (TS, CS+TS). Finally, TS intensities, MEP amplitudes and %SICI were analyzed by a repeated measures ANOVA with the factors *moving body side* (left, right), *condition* (REST, HAND, ANTI) and gender (male, female) as a covariate of no interest. Results {#s3} ======= 3.1 Experiment 1 {#s3a} ---------------- RMT was similar between the hemispheres and ranged from 38--55% of the maximum stimulator output (average of 46±1.6%) for the left hemisphere and from 36--59%, (average of 45.6±1.8%) for the right hemisphere (p\>0.05). ### 3.1.1 Movement performance of the active body side {#s3a1} Subjects complied well with the required cycling frequency for all conditions even though small but significant differences were found such that movements were slightly faster for ANTI (0.980±0.005 s) than IN or the single limb conditions (both 0.983±0.003 s) and also when moving with the right body side (0.980±0.004 s) as compared with the left (0.984±0.004 s) (F(2, 22)≥4.4184, p\<0.01). Movement amplitudes were generally larger for hand (115±23 deg) than foot movements (41±11 deg) (F(1,11) = 74, p\<0.0001). The ANOVA revealed also a significant *condition* x *limb* interaction (F(2,22) = 6.4401, p\<0.01) which was driven by a significantly larger hand amplitude for HAND only (123±15 deg) than for IN (111±14 deg) and ANTI (109±13 deg), while the foot amplitude remained virtually unchanged. The absolute relative phase error did not differ between IN and ANTI coordination or between the body sides (overall 16±1.1 deg; F(1,11) = 2.0609, p = 0.18). However, the standard deviation of the relative phase was slightly higher for ANTI (26.0±1.3 deg) than for IN (24.3±2.1 deg) and when movements were performed with the left body side (26.3±1.4 deg) than with the right (24.0±2.1) \[F(1,11) ≥5.5, p = 0.03 and p = 0.04, respectively\]. ### 3.1.2 EMG of the actively moving hand {#s3a2} Only for the HAND, IN and ANTI conditions, EMG activity of the wrist muscles was substantially larger during the active than the passive phase ([figure 3](#pone-0017742-g003){ref-type="fig"}). This indicates that subjects complied well with the imposed timing and that our stimulation fell reliably within the burst of the ECR when TMS was applied 400 ms after the beep of the metronome and the burst of the FCR when stimulation was applied 150 ms before the beep of the metronome. During the active phase, EMG activity was significantly higher for the ECR (which had to move the hand during most of the extension phase against gravity) than for the FCR (which was activated only during the beginning of the flexion phase as anti-gravitational muscle) (muscle × condition × phase interaction F(3,33) = 38.512, p\<0.00001). Muscle activity was slightly stronger when the right than when the left body side was moved (hemisphere × condition interaction F(3,33) = 3.0129, p = 0.04387). ::: {#pone-0017742-g003 .fig} 10.1371/journal.pone.0017742.g003 Figure 3 ::: {.caption} ###### Mean EMG activity of the ECR and FCR of the actively moving hand during TMS in experiment 1. Data are shown for all movement conditions executed with the left or right body side and for the active phase (square) and passive phase (circle) of each wrist muscle. :::  ::: ### 3.1.3 Corticospinal excitability of M1~ipsi~ as indicated by the normalized MEP-amplitude {#s3a3} Performing movements with one body side had a significant influence on the corticospinal excitability of M1~ipsi~. Statistics of the normalized MEP amplitudes revealed a significant effect of *condition* (F(3,33) = 3.9105, p\<0.05), *contraction phase* (F(1,11) = 16.365, p\<0.005), *hemisphere* x *contraction phase* (F(1,11) = 5.8727, p\<0.05) and *condition* × *contraction phase* (F(3, 33) = 11.216, p\<0.00005). However, these effects can best be understood in light of the significant *hemisphere* × *condition* × *contraction phase* interaction (F(3,33) = 4.4635, p\<0.01) shown in [figure 4](#pone-0017742-g004){ref-type="fig"}. For all conditions involving hand movements (i.e. HAND, IN, ANTI), MEP amplitudes of M1~ipsi~ were significantly larger when the contralateral homologous muscle was in the active phase than in the passive (p\<0.0001). Importantly, during the active phase, there was a substantial hemispheric asymmetry such that M1~ipsi~ responses were significantly larger in the left than in the right hemisphere for HAND, IN and ANTI (p\<0.0001). ::: {#pone-0017742-g004 .fig} 10.1371/journal.pone.0017742.g004 Figure 4 ::: {.caption} ###### Mean normalized amplitudes of MEPs evoked over the left or right primary motor cortex ipsilateral to the moving body side in experiment 1. Data are shown for all movement conditions and when the homologous muscle of the moving hand was either active (squares) or passive (circles). Significant differences between conditions are indicated by \* (p\<0.05), \*\* (p\<0.01), \*\*\* (p\<0.001), \*\*\*\* (p\<0.0001). :::  ::: In the left hemisphere ([figure 4, left panel](#pone-0017742-g004){ref-type="fig"}), corticomotor excitability of M1~ipsi~ during the active phase was lowest in the FOOT condition, which differed significantly from all other conditions (p\<0.0001). NormMEP was highest for the ANTI condition and differed significantly from HAND (p\<0.01). Similar, albeit smaller modulations for the active phase were found in the right hemisphere ([figure 4, right panel](#pone-0017742-g004){ref-type="fig"}) where normMEP was smallest for FOOT (which differed significantly from IN and ANTI, p\<0.01) and largest for ANTI that differed also significantly from HAND (p = 0.03). Finally, we performed an extra analysis on the data during the active phase only to test whether hemispheric asymmetries of M1~ipsi~ facilitation (i.e. higher responses for the left than the right hemisphere) would be larger for the difficult ANTI than the easy HAND condition. However, statistics revealed no significant interaction effect (F(2,22) = 0.28303, p = 0.75620) indicating the hemispheric asymmetries in ipsilateral facilitation were similar across all tasks involving hand movement. In summary, M1~ipsi~ was mirrored the activity of the moving wrist muscles, such that facilitation was stronger when the stimulation fell into the active phase of the homologous muscle. This effect was more pronounced in the left than in the right M1~ipsi~. Additionally, facilitation was modulated by coordination complexity, being significantly stronger when subjects performed a complex hand-foot coordination pattern (ANTI) as compared to simple hand movements. In the next experiment, we specifically tested whether consistent effects would be observed for intracortical inhibition in M1~ipsi~. Results {#s4} ======= 3.2 Experiment 2 {#s4a} ---------------- RMT and AMT were similar and not significantly different between hemispheres (RMT~left~ = 42.9±1.7%, RMT~right~ = 45.1±1.6%; AMT~left~ = 38.4±1.9%; AMT~right~ = 39.1±1.7% of maximal stimulator output) (p≥0.17). Also CS intensity expressed as %AMT was largely comparable between hemispheres (CS~left~ = 86.2±5.1, CS~right~ = 94.8±3.0) (p = 0.09). ### 3.2.1 Movement pattern of the active body side and EMG of the actively moving hand {#s4a1} Subjects complied well with the required movement frequency and there were no significant differences between the HAND and ANTI condition or between hemispheres (overall movement frequency = 1.03±.01 Hz, p≥0.19). Movement amplitude was slightly larger for HAND (67±9 deg) than for ANTI (62±9 deg) (F(1,7) = 42.4; p\<0.001) but did not differ between moving with the left versus right body side (p≥0.68). For ANTI the overall relative phase error was 9.6±1.5 deg and the relative phase variability was 30.6±4.2 deg but there were no significant differences between hemispheres (p≥0.09). Mean active burst EMG of the ECR was similar for HAND (0.197±0.11 mV) and ANTI (0.189±0.1 mV) (F(1,6) = 2.5; p≥0.16). ### 3.2.2 Corticospinal excitability and SICI of M1~ipsi~ {#s4a2} We aimed to adjust TS intensities such that MEP~TS~ amplitudes were matched across all conditions, ranging between 0.7--1 mV. However, post-hoc analyses revealed that the MEP~TS~ amplitude evoked in left M1~ipsi~ during REST was significantly smaller than for all other conditions, which were well matched ([table 1](#pone-0017742-t001){ref-type="table"}) (main effect in [condition]{.smallcaps}: F(2,12) = 4.6 p\<0.05; [hemisphere]{.smallcaps} × [condition]{.smallcaps}: F(2,12) = 4.3; p\<0.05). %SICI varied across tasks and hemispheres ([figure 5](#pone-0017742-g005){ref-type="fig"}). When stimulating the left hemisphere, intracortical inhibition remained relatively constant across conditions. By contrast, for the right hemisphere, inhibition was increasingly released such that %SICI was highest at REST and lowest for ANTI. Statistics confirmed this differential behavior by means of a significant interaction between [hemisphere]{.smallcaps} and [condition]{.smallcaps} (F(2,12) = 4.5 p\<0.05). Post hoc analyses indicated a significant difference between the REST and ANTI condition in the right hemisphere (p\<0.05), whereas, the difference between HAND and ANTI just failed to reach significance (p = 0.063). ::: {#pone-0017742-g005 .fig} 10.1371/journal.pone.0017742.g005 Figure 5 ::: {.caption} ###### Intracortical inhibition in the left (open bars) and right primary motor cortex (grey bars) ipsilateral to the moving body side is shown for all conditions of experiment 2. Significant differences between conditions are indicated by \* (p\<0.05). :::  ::: ::: {#pone-0017742-t001 .table-wrap} 10.1371/journal.pone.0017742.t001 Table 1 ::: {.caption} ###### MEP peak-to-peak amplitude measured in experiment 2 in response to the test stimulus (TS). ::: {#pone-0017742-t001-1} Left MEP~TS~ (mV) Right MEP~TS~ (mV) ------ ------------------------------------------ -------------------- Rest .60[\*](#nt101){ref-type="table-fn"}±.21 .93±.33 Hand .97±.44 .99±.33 Anti .93±.31 1.02±.30 \* indicates that the MEP amplitude evoked via the left hemisphere during rest was significantly smaller than for the other condition (p\<0.05). ::: Discussion {#s5} ========== Here we measured corticomotor excitability of the primary motor cortex ipsilateral to the moving body side when subjects performed rhythmical flexion and extension movements of the wrist either in isolation or as part of a simple (in-phase) or more demanding hand-foot coordination pattern (anti-phase). As a novel result we showed that ipsilateral facilitation of wrist representations in M1 increased significantly when the wrist movements were performed in the context of a demanding hand-foot coordination task. Our data revealed a clear hemispheric asymmetry such that MEP responses were significantly larger when elicited from the left M1~ipsi~ than from the right. Moreover, we found that intracortical inhibition as quantified by SICI was reduced in right but not left M1~ipsi~. Corticomotor excitability of M1~ipsi~ is higher when homologous muscles of the other body side are moved {#s5a} -------------------------------------------------------------------------------------------------------- In agreement with previous studies [@pone.0017742-Carson1], corticomotor excitability of M1~ipsi~ changed substantially as a function of the muscular activity of the homologous muscles of the moving body side. This finding was further supported by the FOOT condition that influenced the excitability of wrist muscles in M1~ipsi~, however, this effect was significantly smaller than when homologous wrist muscles were activated, particularly, during the coordination tasks. The finding that ipsilateral facilitation is strongest in homologous muscles is also interesting in the context of interlimb coordination. It has been shown that rhythmical movements are tightly coupled when they are performed with the same effectors of both body sides (i.e. hand~left~-hand~right~ or foot~left~-foot~right~ coordination) [@pone.0017742-Carson2], [@pone.0017742-Kelso1], [@pone.0017742-Swinnen3] or with different effectors of the same body side (i.e. hand~left~-foot~left~ or hand~right~-foot~right~ movements) [@pone.0017742-Baldissera2]--[@pone.0017742-Byblow1]. By contrast, motor actions can be performed with remarkable independence when the wrist of one body side is moved together with the foot of the other body side (i.e. hand~left~-foot~right~ or hand~right~-foot~left~). Our finding that M1~ipsi~ facilitation was strongest when the homologous hand muscle was activated as compared to a non-homologous hand or foot muscle is very much in line with the behavioural results and suggests that M1~ipsi~ facilitation and interlimb coordination might reflect the same physiological phenomenon and probably arise via callosal pathways that are slightly denser between homologous than non-homologous motor areas [@pone.0017742-Rouiller1]. Corticomotor excitability of M1~ipsi~ depends on movement complexity and is stronger in the left than in the right hemisphere {#s5b} ----------------------------------------------------------------------------------------------------------------------------- Subjects had to move their wrist in isolation or together with foot movements either in accordance to the simple IN or more complex ANTI pattern. Importantly, wrist movements during ANTI were performed with the same speed and the same or slightly smaller movement amplitude than during HAND/IN and also the active EMG of the wrist muscles did not differ between movement conditions and was only slightly larger when the right than when the left body side performed the task. Thus, hand movements were largely similar across conditions and moving body side such that only the demanding coordination context could have induced the high facilitation of M1~ipsi~ during the ANTI task. This notion is also in line with functional imaging studies indicating that hand-foot coordination according to the ANTI pattern activates areas upstream from M1 more strongly than the IN pattern [@pone.0017742-Debaere1]--[@pone.0017742-Caeyenberghs1]. Our results extend previous findings of Ziemann and Hallett [@pone.0017742-Ziemann1] who reported a similar increase of ipsilateral facilitation when subjects performed complex compared to simple finger sequences. However, the advantage of our paradigm is that movement characteristics of the investigated muscles were kept constant while task complexity was varied in a systematic way. In line with Ziemann and Hallett [@pone.0017742-Ziemann1] we found that M1~ipsi~ facilitation was stronger in the left than the right hemisphere. Importantly, subjects were able to perform the motor tasks nearly equally well with their left and right body side as there were no significant differences in mean coordination performance between hemispheres (as measured by the relative phase error) and only minor differences in coordination stability indicating that ANTI was somewhat harder to control than IN and that the left body side moved less consistently than the right. Moreover, the stronger facilitation of left M1~ipsi~ compared to right M1~ipsi~ was found to an equal extent for HAND, IN, or ANTI movements. These results indicate that task complexity modulated the extent of ipsilateral facilitation, but not the extent of hemispheric asymmetries. Hemispheric asymmetries of SICI {#s5c} ------------------------------- In experiment 2, we measured changes of intracortical inhibition of the ECR in M1~ipsi~ while the homologous muscle was either at REST or was activated during HAND and ANTI movements. We found that %SICI decreased significantly with complexity but only for right M1~ipsi~ and not for left M1~ipsi~. The right hemisphere result is in line with Muellbacher et al [@pone.0017742-Muellbacher1], who only tested the right hemisphere and found that voluntary activation of the right APB decreased SICI in the right M1~ipsi~. Similarly, Perez et al [@pone.0017742-Perez1] measured SICI in the right hemisphere while subjects performed isometric contractions at increasing force levels with their right hand. SICI varied in a task specific way and was increasingly released and significantly lower for a strong isometric contraction at 70% as compared to 10%. Our data extend these results, by indicating that %SICI of right M1~ipsi~ tended to change parametrically when subjects performed hand movements either in isolation or in the context of a demanding coordination task. It is important to note that the produced wrist movements were well matched with respect to movement frequency, movement amplitude, EMG activity of the moving ECR as well as relative phase error and variability for ANTI, that did not differ between hemispheres. We experienced difficulties in matching MEP~TS~ amplitudes across conditions, however, recent experiments have indicated that this has no influence on the results when SICI is expressed relative to the unconditioned MEP amplitude (i.e. %SICI as reported in our present study) [@pone.0017742-Rosenkranz1], [@pone.0017742-Rosenkranz2]. In our study, %SICI was very similar and not significantly different for the REST condition of the left versus right hemisphere. Thus, it is justified to argue that the differential modulation of %SICI across conditions and in the right versus left hemispheres can not be explained by sub-optimally matched MEP~TS~ amplitudes. Instead, our data suggest that also intracortical inhibition of M1~ipsi~ differs between the right versus left hemisphere. Potential mechanisms underlying hemispheric asymmetries of M1~ipsi~ facilitation and %SICI {#s5d} ------------------------------------------------------------------------------------------ Even though facilitation of the hemisphere ipsilateral to a moving limb has been demonstrated repeatedly and with different methods [@pone.0017742-Ziemann1], [@pone.0017742-Verstynen1], [@pone.0017742-Chen1], [@pone.0017742-Kobayashi1] for a review see Serrien et al [@pone.0017742-Serrien1], it is not completely clear which mechanisms or anatomical pathways cause this effect and why ipsilateral activity is usually larger in the left than the right hemisphere when tested in right-handed subjects. There are three potential levels of the nervous system (which are not mutually exclusive) that might contribute to M1~ipsi~ facilitation and/or to hemispheric asymmetries of this effect: 1) spinal cord physiology, 2) M1-M1 interactions via transcallosal pathways or 3) functional asymmetries in M1 or areas upstream from M1. First, it has been shown that strong isometric contractions or rhythmical movements of one hand lead to a depression [@pone.0017742-Hortobagyi1], [@pone.0017742-Carson1] and an additional rhythmic modulation [@pone.0017742-Carson1] of H-reflexes of the resting hand. This suggests that movements with one hand might modulate segmental inputs to spinal motorneurons controlling the contralateral hand, probably via (presynaptic) inhibition of Ia afferents [@pone.0017742-Hortobagyi1], [@pone.0017742-Carson1]. However, the same studies have shown that M1~ipsi~ was facilitated while responses to cervicomedullary stimulation of the descending tracts were unchanged, indicating that excitability of the spinal motorneuron pool was not affected by movements of the opposite limb [@pone.0017742-Hortobagyi1], [@pone.0017742-Carson1]. Thus, even though a spinal contribution can not be ruled out completely, it appears that the facilitation of responses from M1 ipsilateral to a moving limb emerges to a large part at the cortical level. Moreover, previous studies comparing H-reflexes between the left and right body side did not find asymmetries in healthy subjects [@pone.0017742-Aymard1], making it unlikely that the strong left-right differences of M1~ipsi~ facilitation found in our study emerged at the spinal level. Second, at the cortical level it has been shown that corticomotor excitability is strongly influenced by inhibitory and facilitatory circuits that act either locally within M1 or via transcallosal M1-M1 projections. Results from Perez et al [@pone.0017742-Perez1] and our own findings indicate that SICI was reduced in the right hemisphere which can explain the increased corticomotor excitability of right M1~ipsi~. Interestingly, Perez at al [@pone.0017742-Perez1] have also shown that for strong isometric contractions with the right hand, interhemispheric inhibition exerted from the contralateral left to the ipsilateral right hemisphere interacted with SICI in M1~ipsi~, such that SICI was weak when IHI was strong and vice versa. Applied to our data, this would suggest that interhemispheric inhibition might have been asymmetric between hemispheres, being larger from left to right than vice versa. This is generally consistent with the view that, during motor task preparation and/or execution the left, motor-dominant hemisphere (in right handed subjects) has a stronger influence on the right, motor non-dominant hemisphere than vice versa [@pone.0017742-Serrien1], [@pone.0017742-Koch1]. It is possible that this asymmetry reflects structural features of the corpus callosum [@pone.0017742-Moes1]. However, studies measuring IHI which is related to the structural integrity of the corpus callosum [@pone.0017742-Wahl1], revealed inconsistent results: Some experiments showed that IHI measured at rest was stronger from the dominant to the non-dominant hemisphere [@pone.0017742-Baumer1]--[@pone.0017742-Netz2], while others revealed no asymmetries [@pone.0017742-DeGennaro1], [@pone.0017742-Salerno1]. Also when IHI was correlated with brain activity in M1~ipsi~ either a positive [@pone.0017742-Kobayashi1], negative [@pone.0017742-Talelli1] or no correlation [@pone.0017742-Callaert1] was found. Thus, more research is needed to establish a convincing link between corpus callosum structure and hemispheric asymmetries in M1~ipsi~ facilitation. Alternatively, it is possible that there are hemispheric asymmetries concerning the function of M1 or upstream motor areas. There is ample evidence that, in right handed subjects, the left hemisphere is involved in the control of complex motor tasks performed with either body side (for a review see [@pone.0017742-Serrien1]). This asymmetry is particularly pronounced for parietal and premotor areas that are believed to contain "movement representations" which are effector independent [@pone.0017742-Haaland1], [@pone.0017742-Swinnen1], [@pone.0017742-Serrien1], [@pone.0017742-Bohlhalter1], [@pone.0017742-Rijntjes1]. Consequently, activity in left M1~ipsi~ might reflect input deriving from the left parietal-premotor networks, that are involved in controlling complex movements even when these are executed with the ipsilateral body side. This view is supported by several studies showing that disrupting M1 activity by repetitive TMS leads to performance decrements when the ipsilateral hand executes a *complex* motor task [@pone.0017742-Chen1], [@pone.0017742-Davare1], [@pone.0017742-Avanzino1]. In line with our results it has been shown that the disruptive effect during demanding motor control was stronger when left than when right M1~ipsi~ was disrupted [@pone.0017742-Chen1]. An important area that can modulate activity in both hemispheres is the premotor cortex [@pone.0017742-Bestmann1]. Premotor areas of the left hemisphere are specifically involved in the preparation and execution of motor actions of either hand [@pone.0017742-Bestmann1], [@pone.0017742-OShea1] whereas right premotor areas are important to prevent unwanted mirror activity in M1~ipsi~ [@pone.0017742-Cincotta1]. Thus, hemispheric differences in the facilitation and %SICI of M1~ipsi~ are likely to reflect the differential involvement of the left versus right premotor cortex in complex motor control. However, more research is needed to delineate the differential contributions of mechanisms acting at the level of spinal cord, transcallosal M1-M1 interaction or premotor-parietal circuits. We thank M. Janssen and J. De Smedt who helped with the data collection in the context of their master thesis. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Support for this study was provided through a grant from the Flanders Fund for Scientific Research (Project G.0758.10). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: FEvdB NW SPS. Performed the experiments: FEvdB NW. Analyzed the data: FEvdB NW. Contributed reagents/materials/analysis tools: FEvdB NW. Wrote the paper: FEvdB SPS NW.

PubMed Central

2024-06-05T04:04:19.292768

2011-3-9

PMC3052420

INTRODUCTION ============ Krabbe disease, also known as globoid-cell leukodystrophy (GLD; MIM\# 245200), is an autosomal recessive disorder resulting from the deficiency of galactocerebrosidase (GALC; E.C. 3.2.1.46), a lysosomal enzyme involved in the catabolism of galactosylceramide, a cerebroside located mainly in the myelin sheath. GALC defects lead to the accumulation of a cytotoxic metabolite (galactosylsphingosine or psychosine) which results in the apoptosis of myelin-forming cells \[[@b31]\]. The majority of patients with Krabbe disease (around 85%-90% of cases) have the infantile form of the condition, presenting with extreme irritability, spasticity and developmental delay before the age of six months, with death occurring before the age of two. The remaining 10-15% of patients have a late-onset form of the disease, the onset of symptoms occurring between the age of 6 months and the fifth decade, and with slower disease progression. In the infantile form of Krabbe disease, the child, who appears normal for the first few months post-partum, presents with frequent crying and irritability, feeding difficulties, arrest/regression of motor and mental development, as well as seizures in some cases. Severe motor and mental deterioration follows, leading to decerebration. Peripheral neuropathy is typically present. In the late-onset forms of the disease, individuals can be clinically normal until weakness, vision loss and intellectual regression become evident; however, adult onset patients may show spastic paraparesis as the only symptom. Disease severity is variable, even within families. Further differentiation between late-infantile (onset age, six months to three years, severe clinical course), juvenile (onset age, three to eight years; more gradual progression lasting for years), and the adult-onset forms, is commonly employed (<http://www.genetests.org>) \[[@b29]\]. Nerve conduction velocity studies have been reported to be normal in some adults with an enzymatically confirmed diagnosis. Brain magnetic resonance imaging (MRI) usually reveals deep white matter abnormalities consistent with a demyelinating process that involves the brainstem and cerebellum. Computed Tomography (CT) scans may reveal hyperdensity indicative of calcification involving the cerebellum, thalami, caudate, corona radiata and brainstem. The galactosylceramidase gene *(GALC*; MIM\# 606890\], spanning 60 kb of genomic DNA on chromosome 14q31, comprises 17 exons and encodes a 3.8 kb mRNA ([Figure 1](#fig01){ref-type="fig"}). Two in-frame ATG translational start sites, located in exon 1 ([Figure 1](#fig01){ref-type="fig"}), give rise to distinct isoforms of GALC which are both translated. Although translational initiation at these alternative ATG codons yields protein precursors with 42-residue (NP\_000144.2) or 26-residue leader sequences respectively, both precursors are processed to the 669-residue mature enzyme \[[@b2]; [@b25]; [@b14]\]. The precursor protein, estimated to have a molecular mass of 80-kDa, is proteolytically processed into 30- and 50-kDa fragments in the extracellular medium. Neither the 50-nor the 30-kDa fragment, expressed *in vitro* in COS cells, have been demonstrated to possess GALC activity, an indication that the entire structure is probably necessary for enzymatic activity and that fragments expressed separately cannot subsequently associate to form an active enzyme \[[@b2]; [@b3]; [@b25], [@b19]\]. ::: {#fig01 .fig} Figure 1 ::: {.caption} ###### Distribution of the detected mutations and polymorphisms in relation to the *GALC* gene and GALC polypeptide. In the top half of the diagram (A), the map of the *GALC* gene depicts the positions of the seventeen exons (dark green boxes numbered 1 to 17) and associated unnumbered introns (gray lines). In the bottom half of the diagram (B), the schematic representation of the polypeptide shows the 26-amino acid signal (leader) sequence, and the two protein subunits (50-52 kDa and 30 kDa) predicted to be proteolytically processed from the precursor protein. All mutations encountered in this patient series are given in (A) above the gene schema, all polymorphisms below. Novel mutations and polymorphisms reported here are denoted in red and violet respectively. HGVS nomenclature guidelines stipulate amino acid numbering from the first methionine of the 42-residue signal sequence. Hence, HGVS nomenclature was used for the mutations and the polymorphisms reported here (with the traditional designations given in parentheses). :::  ::: Although the *GALC* gene was identified more than 15 years ago, fewer than 80 mutations have been reported as a cause of Krabbe disease \[<http://www.hgmd.org>; [@b28]\] and relatively few papers have addressed the issue of a genotype-phenotype relationship in this disorder. Clinical phenotypes can however differ quite markedly between individuals with the later-onset forms of the disease, including siblings harbouring the same *GALC* genotypes. The *GALC* mutational profile differs between European and Japanese Krabbe disease patients; thus, whilst the most common large deletion, c.1161+6532\_polyA+9kbdel (IVS10del30kb) plus three other mutations \[c.1586C\>T, c.1700C\>T, c.1472delA (1538C\>T, 1652A\>C, 1424delA)\] together account for about 60% of alleles in patients of European ancestry with the classic infantile form \[[@b11]; [@b30]\], these lesions are absent in their Japanese counterparts. By contrast, ∼30% of Japanese *GALC* alleles associated with the infantile form of the disease possess either c.683\_694del12insCTC or c.2002A\>C (635\_646del12insCTC and 1954A\>C) \[[@b32]\]. It should be noted that *GALC* mutations have traditionally been described on the basis of their amino acid position in the mature enzyme, with p.M17 being designated as the first residue. Current HGVS nomenclature recommendations however require proteins to be numbered from the first methionine of the complete 42-residue signal sequence. Hence, throughout this article, we ascribe HGVS designations to specific amino acid residues, with the traditional designations given in parentheses. The median prevalence of Krabbe disease is estimated to be ∼1 in 100,000 births (1.00 × 10^−5^) with wide variations between countries: 1.35 in Netherlands, 1.21 in Portugal, 1.00 in Turkey, 0.71 in Australia, and 0.40 in the Czech Republic \[[@b22]\]. However, two quite separate inbred communities in Israel (the Druze and Moslem Arab populations) have an extremely high prevalence of the infantile form (about 1 in 100-150 live births), due to two non-identical mutations, c.1796\>G and c.1630G\>A (1748T\>G and 1582G\>A), respectively \[[@b24]\]. The founder mutation c.169G\>A (121G\>A) is responsible for the other known Krabbe disease 'hotspot' in Catania (Sicily), this time associated with the late-onset form of the disease \[[@b13]\]. In Italy, accurate epidemiological data for lysosomal disorders (LSDs) are not available owing to both the geographically wide dispersion of the patients between the various collaborating clinical/laboratory centres and the lack of any national register of rare diseases. However, the Gaslini laboratory has extensive experience in the analysis and diagnosis (both antenatal and postnatal) of LSDs, including Krabbe disease, going back over 35 years. Additionally, biobanked cell lines ("Cell line and DNA Biobank from patients affected by Genetic Diseases" at <http://dppm.gaslini.org/biobank/>) have enabled us to put together, between 1976 and 2009, unique collections of clinical samples and related data from a large series of families involving a total of 970 patients with 41 diverse LSD types/subtypes. Using the Biobank database, we have estimated the relative prevalence of Krabbe disease as ∼4% of the total number of LSD patients (data not shown). Although the examined time period and number of LSD types/subtypes differ slightly between published studies, our prevalence data are comparable to those reported from the Czech Republic and Australia (3.3% and 5%, respectively) \[Poupetova et al., 2010; [@b17]\]. To date, the Krabbe disease collection within the Biobank includes 36 families. We have previously reported eight unrelated patients, in five of whom the second allele had remained undetermined \[[@b27]\]. Here, we have characterized the underlying *GALC* gene lesions in a total of 30 unrelated Krabbe disease patients, in an analysis which included the complete genotyping of two patients from the previous study \[[@b27]\]. This study therefore reports one of the largest mutational analyses of the *GALC* gene so far performed in a Caucasian population affected by Krabbe disease. It is apparent that the Italian mutational *GALC* profile differs significantly from other populations of European origin (http://www.genetests.org) \[[@b29]\]. MATERIALS AND METHODS ===================== Patients -------- The present series comprises a total of 30 unrelated patients with Krabbe disease. The diagnosis was suspected upon clinical evaluation and was supported by neuroimaging and neurophysiological findings. Most (27/30) patients underwent neuroradiological studies (brain MRI in 22/27 and CT scan in 13/27); nerve conduction velocity studies were performed in 14/30 patients. The characteristics of these patients and the main clinical findings are summarized in [Table 1](#tbl1){ref-type="table"}. ::: {#tbl1 .table-wrap} Table 1 ::: {.caption} ###### Clinical, instrumental and molecular data encountered in 30 unrelated GLD patients ::: Neurological findings Neuroradiologic findings ---- --- ------ -------------------------------------- ----------------------- ------------------------------------------- --------------- ------ -------- ----- ------ ------------------------------------------------- 1 F 1m Failure to thrive 7m +/ Truncal hypotonia \+ ND Slowed 4.4 \[c.1161+6532\_polyA+9kbdel\]+\[?\] 2 F 3m Psychomotor regression 5m +/ Truncal hypotonia \+ (Hypomyel) ND Slowed 8.2 \[p.L469YfsX22\]+\[p.L469YfsX22\]\# 3 M 3m Muscular hypertonia 6m +/Nystagmus \+ +/ + ND 0.3 \[ p.G553R\]+\[p.G553R\]\# 4 M 3m Irritability, muscular hypertonia 5m +/Truncal hypotonia \+ ND ND 14.7 \[p.S303F\]+\[p.G553R\] 5 M 3m Muscular hypertonia. seizures 6m +/ Truncal hypotonia ND +/ + ND yes 18.5 \[p.E136EfsX35\]+\[p.E136EfsX35\] 6 F 3m Irritability, psychomotor regression 7m +/ Truncal hypotonia \+ ND Slowed 21.7 \[p.K359AfsX3\]+\[p.Y567S 7 F 3m Nystagmus, psychomotor regression 6m +/Truncal hypotonia, poor tendon reflexes \+ ND Slowed 3.2 \[ p.Y174LfsX3\]+\[c.1161+6532\_polyA+9kbdel\] 8 M 4m Irritability, muscular hypertonia 6m +/Nystagmus \+ ND Slowed yes 0 \[p.E130K\]+\[p.N295T\]\# 9 M 4m Muscular hypertonia 6m +/Truncal hypotonia, poor tendon reflexes ND +/ + ND 11.1 \[p.H391 IfsX65\]+\[p.H391 IfsX65\] 10 M 4m Muscular hypertonia 8m +/ Truncal hypotonia ND -/ + ND yes 5 \[p.T529M\]+\[c.1161+6532\_ polyA+9kbdel\] 11 F 4m Muscular hypertonia 6m +/ Truncal hypotonia \+ (Hypomyel) ND Slowed 8.4 \[p.A21RfsX5\]+\[p.G553R\] 12 F 4m Irritability, muscular hypertonia 5m +/ Truncal hypotonia \+ (Hypomyel) +/ + ND yes 0 \[p.Y314C\]+\[c.1161+6532\_polyA+9kbdel\]\# 13 M 5m Psychomotor regression 6m +/Truncal hypotonia \+ ND ND yes 0 \[ p.F596SfsX1 6\] +\[p.F596SfsX1 6\]\# 14 F 5m Psychomotor regression, seizures 7m +/ Truncal hypotonia \+ (Hypomyel) ND Slowed 0 \[p.G553R\]+\[p.G553R\] 15 F 5m Irritability, muscular hypertonia 7m +/ Truncal hypotonia \+ (Hypomyel) +/ + ND 0 \[p.N295T\]+\[c.1161+6532\_polyA+9kbdel\] 16 F 5m Psychomotor regression 7m +/ Truncal hypotonia ND ND ND 8.3 \[p.R127X\]+\[p.G553R\]\#\# 17 F 5m Muscular hypertonia 10m +/ Truncal hypotonia ND ND ND 4.8 \[p.K88X\]+\[p.Y490N\]\# 18 M 5m Psychomotor regression 9m +/ Truncal hypotonia \+ +/ + ND 8.8 \[p.R396L\]+\[c.1161+6532\_ polyA+9kbdel\] 19 M 5m Psychomotor regression 8m +/ Truncal hypotonia \+ ND ND yes 18.2 \[p.G553R\]+\[p.G553R\]\# 20 M 5m Muscular hypertonia 7m +/ Truncal hypotonia \+ ND ND yes 9.7 \[p.L634X\]+\[c.1489+1G\>A\] 21 F 5m Psychomotor regression 8m +/ Truncal hypotonia \+ ND ND yes 0.71 \[p.E130K\]+\[p.Y490N\] 22 M 8m Psychomotor regression 11m +/ Truncal hypotonia \+ -/ + ND yes 13.2 \[p.G102GfsX5\]+\[?\] 23 M 8m Psychomotor regression 11m +/ Truncal hypotonia ND -/ + ND yes 0.26 \[p.D187V\]+\[p.G323R\] 24 F 10m Psychomotor regression 12m +/ Truncal hypotonia \+ -/ + Slowed yes 0 \[p.G59R\]+\[?\]\# 25 M 11m Irritability, muscular hypertonia 13m +/ Truncal hypotonia ND -/ + Slowed 10.1 \[p.I250T\]+\[p.R396W\] 26 F 3y6m Gait disturbances and frequent falls 4y +/Ataxia, poortendon reflexes \+ ND Slowed 5 \[ p.G286D\]+\[c. 1161 +6532\_ polyA+9kbdel\]\# 27 F 3y6m Gait disturbances and frequent falls 3y8m +/Ataxia \+ -/ + Normal NA \[p.R69X\]+\[p.I384T\] 28 M 4y Gait disturbances and frequent falls 4y +/Nystagmus ND ND Slowed 13 \[p.N295T\]+\[p.G609GfsX6\] 29 F 4y yReduced visual acuity 5y9m +/- \+ -/ + Normal yes 0 \[p.R79H\]+\[p.G553R\] 30 M 26y Gait disturbances and frequent falls 30y +/- \+ ND Slowed 5 \[p.G286D\]+\[p.P318R\] Legend: Pt=Patient; m= month(s); y=year(s); \#Genotype confirmation by parental DNAanalysis; \#\# only mother\'s DNAavailable; ND: Not Done; N=Normal; NA: Not Available \*\*GenBank-EMBLaccession no. NM\_000153.2 and no. NP\_000144.2; Hypomyel= hypomyel inat ion; MRI= Magnetic resonance imaging, WM=white matter; CT= Computed tomography; NCVs= nerve conduction velocity study; ♣ indicates presence of other affected patients in the family; §Owingto the use of different assay met hodsand tissue samples, as reported intheMaterialsand Methods, enzyme act ivity valuesare expressed asa percentage of average control values. ::: A diagnosis of Krabbe disease was confirmed by enzymatic assay in homogenates of leukocytes or fibroblast cell lines using either tritium-labelled \[H^3^\] galactosylceramide (until 2004) or 6-hexadecanoylamino-4-methylumbelliferone-beta-D-galactoside (post-2004) as substrates. GALC activity values are provided for each patient in [Table 1](#tbl1){ref-type="table"}. The patients in this series were diagnosed over a period of more than 30 years. The present study has been made possible thanks to the availability of the corresponding patient fibroblast/lymphoblast cell lines and/or DNA samples, cryopreserved within the laboratory "Cell Line and DNA Biobank from Patients affected by Genetic Diseases". Following ethical guidelines, all samples obtained for analysis and storage required prior written informed consent using a form approved by the Local Ethics Committee. Cell culture ------------ Fibroblast and lymphoblast cells were cultured according to standard procedures. The cell lines were cultured and maintained in RPMI medium (EuroClone, Gibco, Paisley, UK) containing 15% FCS and penicillin/streptomycin, in a humidified atmosphere containing 5% CO2 at 37°C. Molecular analysis ------------------ Genomic DNA was extracted using standard methods from cultured fibroblasts or lymphoblasts derived from the affected individuals and from peripheral blood leukocytes of the available family members. *GALC* gene exons and exon-intron boundaries were PCR amplified using specific primers designed by reference to the genomic sequence (GenBank-EMBL Accession No. NC\_000014.8). We initially screened for the large common Caucasian 30-kb *GALC* gene deletion (c.1161+6532\_polyA+9kbdel) in all DNA samples, using three PCR primers in accordance with the previously reported method \[[@b15]\]. In the next step, a total of 16 PCR amplimers were tested on each patient, exons 2-3 being amplified simultaneously, as reported in [Supp. Table S1](#tbl3){ref-type="table"}. PCR amplification conditions were as follows: initial denaturation 2 min at 94°C, followed by 35 cycles amplification, denaturation at 94°C for 30 sec, annealing from 58°C to 62°C for 30 sec, and extension at 72°C for 40 sec using AmpliTaq DNA Polymerase (Applied Biosystems, Foster City, CA). Details of PCR-product sizes and annealing temperatures are reported in [Supp. Table S1](#tbl3){ref-type="table"}. Total RNA was extracted from patient fibroblasts/lymphoblasts using an RNeasy mini kit (QIAGEN, Courtaboeuf, France) and reverse transcribed by means of an Advantage RT-for-PCR kit (BD Biosciences Clontech, Mountain View, CA, USA). RT-PCR was performed using sets of primers designed by reference to the GALC mRNA sequence (GenBank accession No. NM\_000153.2). The RT-PCR set of primers employed is given in [Supp. Table S1](#tbl3){ref-type="table"} together with the temperature profiles and expected product sizes. RT-PCR amplification conditions were as follows: initial denaturation 2 min at 98°C, followed by 30 cycles of amplification, denaturation at 98°C for 20 sec, annealing from 59°C to 65°C for 30 sec, and extension at 72°C for 30 sec, using Phusion High-Fidelity DNA Polymerase (Finnzymes, Keilaranta, Finland). These PCR products were cloned into the TOPO TA Cloning KIT (with pCR2.1-TOPO vector) (Invitrogen, San Diego, CA) according to the manufacturer\'s instructions. Sequence analysis of PCR and RT-PCR products was performed in the forward and reverse directions using the ABI PRISM Big Dye Terminator Cycle Sequencing kit (Applied Biosystems, Foster City, CA). Sequences were analyzed on an ABI PRISM 3700 DNA Analyzer. Sequence alterations were confirmed by sequencing duplicate PCR products and/or by digesting PCR products with the specific restriction endonuclease whose recognition site had been concomitantly altered. If a given alteration had neither created nor destroyed a restriction site, PCR amplification was carried out by PCR-mediated site-directed mutagenesis that artificially introduced a new restriction enzyme cleavage site ([Supp. Table S2](#tbl4){ref-type="table"} \[[@b26]\]. The issue of whether the novel *GALC* sequence alterations detected were causative mutations or neutral polymorphisms was addressed by (i) searching dbSNP (<http://www.ncbi.nlm.nih.gov/SNP>) for their presence, (ii) screening 100 alleles from healthy control subjects for each alteration, and (iii) employing the MutPred program \[[@b12]; [@b18]\] (see below). *MutPred* analysis of *GALC* mutations and polymorphisms -------------------------------------------------------- The likely pathogenicity of missense mutations (and non-synonymous polymorphisms) identified in the human *GALC* gene was assessed by means of a computational model termed MutPred \[[@b12]; [@b18]\]. MutPred was designed to model changes of structural and functional sites between wild-type and mutant protein sequences. Hence, MutPred can be used to generate hypotheses regarding the underlying molecular mechanism(s) responsible for disease pathogenesis Haplotype analysis ------------------ Segregation analysis of the identified polymorphisms was performed on all patients in the series including all patients coming from the same geographic area (Campania region). Mutation nomenclature --------------------- All mutations are described according to current mutation nomenclature guidelines (<http://www.hgvs.org/mutnomen>), ascribing the A of the first ATG translational initiation codon as nucleotide +1 \[[@b6]; [@b7]\]. Traditional amino acid residue numbering has nevertheless also been provided in parentheses. RESULTS AND DISCUSSION ====================== A comprehensive clinical evaluation, facilitated by the results of a range of diagnostic procedures including neuroradiological, neurophysiological and enzymatic testing, was performed on 30 unrelated patients affected by Krabbe disease prior to characterization at the molecular genetic level. The clinical characteristics of these 30 patients and the main neuroradiological/ neurophysiological findings are summarized in [Table 1](#tbl1){ref-type="table"}. Apart from patient \#2, of Moroccan origin, and patient \#9, belonging to an itinerant Roma family, all patients were of Italian origin. Clinical aspects ---------------- As reported in [Table 1](#tbl1){ref-type="table"}, 21 patients (pts) were classified as having the infantile form of Krabbe disease (age at onset ranging from 1 to 5 months) while 9 were considered to have late-onset forms of the disease. In this latter group were 4 patients with the late-infantile form (age at onset ranging from 8 to 11 months), 4 patients with the juvenile form (age at onset ranging from 3 years 6 months to 4 years) and one adult-onset patient (onset at age 26 years). Except for the youngest patient who presented at 1 month of age simply with failure to thrive, muscular hypertonia (11/21) and psychomotor regression (9/21) variably associated with irritability (5/21) were the main presenting symptoms in the infantile-onset patients. Seizures and nystagmus were also observed at disease onset in two infantile patients (\#5 and \#7). Psychomotor regression was the presenting symptom in 3/4 patients with the late-infantile form, whereas irritability and muscular hypertonia occurred in the remaining patients from this group. Gait disturbances and frequent falls were the presenting symptoms in 4/5 patients with the juvenile (pts \#26, \#27, \#28) and adult forms (pt \#30) of the disease, whereas reduced visual acuity was the presenting symptom in the remaining case (pt \#29). [Table 1](#tbl1){ref-type="table"} also reports the findings at neurological examination for all subjects, at variable age after disease onset, who invariably exhibited spasticity irrespective of the age at onset, variably associated with truncal hypotonia, nystagmus, ataxia and poor tendon reflexes. CT scans revealed white matter hypodensity in all 13 patients who underwent examination; this was associated with calcification in 6 patients (5/6 in the thalami, in one case in the periventricular white matter). Brain MRI revealed white matter changes in all studied patients which were consistent with hypomyelination in 5 cases ([Table 1](#tbl1){ref-type="table"}). Interestingly, this latter finding may represent a feature of lysosomal storage disorders with onset in the first months of life, when the process of myelination is particularly active, indicating that neuronal storage disorders may be primarily responsible for central nervous system hypomyelination \[[@b8]\]. Nerve conduction velocity studies were slowed in 12/14 analyzed patients, but normal in 2 subjects with the juvenile form (pts \#27 and \#29) of the disease. *GALC* gene mutations in the Italian population ----------------------------------------------- The *GALC* gene was investigated by sequencing analysis in all 30 unrelated Krabbe disease patients. As reported in [Supp. Table S1](#tbl3){ref-type="table"}, all 17 exons and most of the flanking intronic regions were analyzed. Additionally, the sequence analysis was extended to part of the 5′ untranslated region as well as of the untranslated region ([Supp. Table S1](#tbl3){ref-type="table"}). [Table 1](#tbl1){ref-type="table"} summarizes the 27 distinct mutant genotypes encountered in this 30-patient series. Apart from 7 (23%) of the patients found to be homozygous for rare \[p.G553R (G537R), p. F596SfsX16 (F580Sfs)\] or novel \[p.E136EfsX35, p.H391IfsX65, p.L469YfsX22\] mutations, the remainder were rare compound heterozygotes, including 3 patients in whom the second mutant allele could not be identified. Among this latter group, 7 patients possessed the common large deletion of European origin (c.1161+6532\_polyA+9kbdel) in *trans* to rare missense mutations p.G286D (G270D), p.N295T (N279T), p.Y314C (Y298C), p.R396L (R380L), p.T529M (T513M), a new frameshift (p.Y174LfsX3), or a still-unidentified mutation. As shown in [Table 2](#tbl2){ref-type="table"}, the *GALC* mutational profile was highly heterogeneous, being characterized by a total of 33 distinct mutant alleles including 15 previously unreported alleles; three mutant alleles still remain to be identified. Apart from 4 novel missense mutations (p.P318R, p.G323R, p.I384T, p.Y490N), 73% of the newly described mutations were expected to affect mRNA processing. These comprised 7 frameshift mutations resulting from 3 microdeletions (c.61delG, c.408delA, c.521delA), 2 indels (c.1171\_1175delCATTCinsA, c.1405\_1407delCTCinsT) and 2 microduplications (c.302\_308dupAAATAGG, c.1819\_1826dupGTTACAGG), respectively; 3 nonsense mutations (p.R69X, p.K88X, p.R127X); and a splicing mutation at the donor site of intron 13 (c.1489+1G\>A) which induced the partial skipping of exon 13. [Figure 1](#fig01){ref-type="fig"} depicts the location of the various mutations detected in relation to the *GALC* gene and its protein product. It is evident that the *GALC* mutations associated with severe clinical phenotypes are spread throughout the gene without any preferential clustering within the sequence encoding the 30-kD subunit of the mature GALC protein, thought to be critical for the synthesis of catalytically active enzyme \[[@b23] and [@b24]\]. ::: {#tbl2 .table-wrap} Table 2 ::: {.caption} ###### Characteristics of the *GALC* gene gene mutations identified in the 30 Krabbe disease patients and MutPred analysis of the missense mutations ::: MutPred analysisof missense mutations\# ---------- ----------------------------- ------------------- ------------------------------------- ---------------- ----------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------- --------------- Ex. 1 **c.61delG** **p.A21RfsX51** **Frameshift** \- Present study c.175G\>C p.G59R 127G\>C (G43R) Missense 0.92 None [@b9] **c.205T\>C** **p.R69X** **Nonsense** \- Present study Ex.2 c.236G\>A p.R79H 188G\>A(R63H) Missense 0.85 None [@b5] **c.262A\>T** **p.K88X** **Nonsense** \- Present study Ex.3 **c.302\_308dupAAATAGG** **p.G102GfsX5** **Frameshift** \- Present study **c.379C\>T** **p.R127X** **Nonsense** \- Present study Ex. 4 c.388G\>A p.E130K 340G\>A(E114K) Missense 0.80 Gain of molecular recognition feature (MoRF) binding (P=0.0047), Gain of methylation at E130 (P=0.0114), Gain of ubiquitination at E130 (P=0.0269) [@b13] **c.408delA** **p.E136EfsX35** **Frameshift** \- Present study Ex. 5 **c.521delA** **p.Y174LfsX3** **Frameshift** \- Present study c.560A\>T p.D187V 512A\>T(D171V) Missense 0.71 None [@b16] Ex. 7 c.749T\>C p.I250T 701T\>C(I234T) Missense 0.79 Gain of protein disorder (P=0.0309), Loss of beta sheet secondary structure (P=0.0392) [@b5] c.857G\>A p.G286D 809G\>A(G270D) Missense 0.98 None [@b10] Ex. 8 c.884A\>C p.N295T 836A\>C(N279T) Missense 0.80 None [@b30] c.918C\>T p.S303F 870C\>T(S287F) Missense 0.83 None [@b30] c.941A\>G p.Y314C 893A\>G(Y298C) Missense 0.89 None [@b5] Ex. 9 **c.953C\>G** **p.P318R** **Missense** 0.79 Increased solvent accessibility (P=0.0179), Loop \> Helix secondary structure change (P=0.0259) Present study **c.967G\>A** **p.G323R** **Missense** 0.91 None Present study Ex. 10 c.1075\_1084delAAGACAGTTG p.K359AfsX3 1027\_1036delAAGACAGTTG§(K343AfsX3) Frameshift \- [@b30] **c.1151T\>C** **p.I384T** **Missense** 0.74 None Present study Intr. 10 c.1161+6532\_polyA+9Kbdel IVS10del30kb Deletion \- [@b23] **d 171\_1175delCATTCinsA** **p.H391IfsX65** **Frameshift** \- Present study Ex.11 c.1186C\>T p.R396W 1138C\>T(R380W) Missense 0.96 Loss of protein disorder (P=0.0371) [@b30] c.1187G\>T p.R396L 1139G\>T(R380L) Missense 0.95 Gain of ubiquitination at K393 (P=0.0452) [@b27] Ex. 13 **c.1405\_1407delCTCinsT** **p.L469YfsX22** **Frameshift** \- Present study **c.1468T\>A** **p.Y490N** **Missense** 0.68 Loss of phosphorylation at Y490 (P=0.0245) Present study Intr. 13 **c.1489+1G\>A** **p.S488NfsX200** **Splicing** \- Present study Ex. 14 c.1586C\>T p.T529M 1538C\>T(T513M) Missense 0.68 None [@b30] c.1657G\>A p.G553R 1609G\>A(G537R) Missense 0.91 Sheet \> Helix secondary structure change (P=0.0151) [@b4] c.1700A\>C p.Y567S 1652A\>C(Y551S) Missense 0.73 None [@b30] Ex. 15 c.1787delT p.F596SfsX16 1739delT(F580SfsX16) Frameshift \- [@b27] **c.1819\_1826dupGTTACAGG** **p.G609GfsX6** **Frameshift** \- Present study Ex. 16 c.1901delT p.L634X 1853delT(L618X) Nonsense \- [@b30] Legend: Ex=exon; Intr=intron; \**GALC* gene GenBank-EMBL accession no. NM\_000153.2; \*\**GALC* gene GenBank-EMBL accession no. NP\_000144.2; the novel mutations are given in bold; §mutation reported as c.1026del10 by Rafi et al., unpubished observations \[Wenger et al., 2007\]. \#[@b12] and [@b18]. ::: Relative frequencies of the *GALC* mutations and evidence for a founder effect involving the p.G553R (G537R) mutation --------------------------------------------------------------------------------------------------------------------- In agreement with previous studies on individuals of European ancestry (<http://www.genetests.org>) \[[@b29]\], the large deletion (c.1161+6532\_polyA+9kbdel) was confirmed to be the most frequent *GALC* mutation in the Italian population (taking together 6 previously reported deletion alleles \[[@b27]\] with the 7 deletion alleles from the present series). However, while in other populations this large deletion has been shown to have a higher frequency, being respectively 52% of disease alleles in Dutch patients \[[@b11]\] and 44% of disease alleles in patients with Northern European ancestry \[[@b15]; [@b23]\], in our series we found that this same deletion \[c.1161+6532\_polyA+9kbdel\] barely accounts for 18% of disease alleles in our own series. Moreover, our findings contrast with the three other common mutations \[p.T529M (T513M), p.Y567S (Y551S) and c.1472delA (1424delA)\], expected to account for ∼15% of *GALC* alleles in the European population (<http://www.genetests.org>) \[[@b29]\]; in our series, the two missense mutations were noted only once each (corresponding to 3.2% of alleles) whereas the microdeletion c.1472delA was not detected in our series at all. Instead, the second most frequent allele was p.G553R (G537R), previously only reported in one other family originating in southern Italy \[[@b4]\]. Since all seven patients harbouring p.G553R (G537R) ([Table 1](#tbl1){ref-type="table"}, pts \#3, \#4, \#11, \#14, \#16, \#19, \#29) shared a common geographical origin around Naples (Campania region, southern Italy), we analysed their haplotype backgrounds. Among the various identified polymorphisms, it was evident that two very rare SNP alleles, namely rs74073730 \[c.1072C\>T (1024C\>T)\] and rs74076317 \[c.1338+23T\>C (1290+23T\>C)\], were exclusively present in all 7 p.G553R-bearing patients ([Supp. Table S3](#tbl5){ref-type="table"}). Both SNP alleles, originally identified by massively parallel sequencing \[[@b1]\], are extremely rare; dbSNP gives their heterozygosity as \'not known\' (<http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=rs74073730>). Hence the degree of linkage disequilibrium between the rare SNP alleles and the p.G553R disease allele could not be formally measured. However, we did exclude the presence of the rare SNP alleles in the remainder of the Krabbe disease patients, including three other individuals from the Campania region, who carried different *GALC* mutant genotypes ([Table 1](#tbl1){ref-type="table"}, pts \#13, \#22, \#24.). It is therefore likely that the p.G553R (G537R) mutation occurred on the rs74073730(T) -rs74076317(C) haplotype background of a common founder ancestor. Identity-by-descent would account for the unexpectedly high number of an otherwise extremely rare *GALC* mutation in Italian Krabbe disease patients. In addition to the putative pathological mutations, a total of 24 *GALC* polymorphisms were identified within either the exons, introns or 5′ untranslated region (5′ UTR). Novel SNPs were noted in the 5′ UTR (1), within the coding region (2) and in the introns (3) ([Figure 1](#fig01){ref-type="fig"}; [Supp. Table S3](#tbl5){ref-type="table"}). The allele frequencies of the new variants, derived from 200 normal Italian controls, are given in [Supp. Table S3](#tbl5){ref-type="table"}. *MutPred* analysis ------------------ In an attempt to establish the functional relevance or otherwise of the novel *GALC* missense mutations detected, we employed the *in silico* analysis tool, *MutPred* \[available at <http://mutdb.org/profile/>; [@b12]; [@b18]\]. To assess the predictive power of this bioinformatic approach, the analysis was performed on both the newly identified missense mutations and the previously reported deleterious missense mutations. For the known pathological missense mutations listed in [Table 2](#tbl2){ref-type="table"}, *MutPred* successfully predicted all 18 (probability threshold \>0.65) to be deleterious and generated confident *in silico* hypotheses for 39% (7 out of 18) of these mutations. Further investigation of the 4 novel missense mutations (p.P318R, p.G323R, p.I384T, p.Y490N) identified in the present study, indicated all four to have a high probability (\>0.68) of being deleterious; *in silico* hypotheses for the underlying mechanism were generated for two of them. Thus, the novel missense mutation p.P318R is predicted to impact upon protein structure in terms of both solvent accessibility (P = 0.0179) and secondary structure (loop → helix; P = 0.0259) whereas the novel missense mutation p.Y490N is predicted to disrupt a phosphorylation site at this residue (P = 0.0245). By contrast, none of the missense polymorphisms listed in [Supp. Table S3](#tbl5){ref-type="table"}, were predicted to be deleterious (probability threshold \< 0.65). However, it should be noted that, according to *MutPred*, both the novel missense polymorphism p.A21P and the previously identified missense polymorphism \[p.T641A (T625A)\] have probabilities of being deleterious that lie close to the notional threshold (0.62 and 0.58 respectively) suggesting that they could be of functional significance. None of the missense polymorphisms were however predicted to disrupt any structural or functional sites in the GALC protein. Analysis of evolutionary conservation of amino acid residues affected by missense mutations ------------------------------------------------------------------------------------------- Additional support for the pathological/functional significance of the missense mutations (and potentially some of the polymorphisms) identified in the present study, came from the analysis of the extent of evolutionary conservation of the mutated residues in 9 orthologous (vertebrate) GALC proteins. The computational analysis, carried out at <http://www.ensembl.org/>, revealed that all 18 missense mutations occurred at amino acid residues which were evolutionarily conserved in chicken as well as in various mammals. Further, eleven of the residues involved (including those harbouring the novel p.P318R, p.G323R and p.Y490N mutations) were invariant even when zebrafish was considered ([Supp. Table S4](#SD1){ref-type="supplementary-material"}). When a similar analysis was performed on the missense polymorphisms ([Supp. Table S5](#SD2){ref-type="supplementary-material"}), one of the two SNPs predicted by *MutPred* to be potentially deleterious (p.A21P) was found to occur in a residue which was conserved in 5 vertebrates and hence could therefore be of functional significance. However, residue 641, which harbours the other potentially functional p.T641A (T625A) SNP, was not evolutionarily conserved. The relationship between *MutPred* values and residue evolutionary conservation therefore appears to be rather more tenuous in the case of the missense polymorphisms than it is for the missense mutations. Thus, the residue harbouring the D248N (D232N) SNP was conserved in 7 vertebrates including the three-spined stickleback, *Gasterosteus aculeatus*, even although its *MutPred* score was 0.52 (whilst residue 562, harbouring the p.I562T SNP allele (*MutPred* score 0.53), was not evolutionarily conserved). Intriguingly, this latter SNP was present in all three patients with an unknown *GALC* allele but, since it is also present in 14 other patients, it is not possible to draw any firm conclusions as to its possible functional significance. Consequences of *GALC* mutations for mRNA processing ---------------------------------------------------- Reverse transcript-polymerase chain reaction (RT-PCR) analysis was used, wherever possible, to investigate frameshift, nonsense and splicing mutations to assess their effect on *GALC* mRNA processing. ### Frameshifts and stop codons A total of 6 microdeletions, 2 insertion/deletions (indels) and 2 microduplications were identified during the course of this study. Most were explicable in terms of slipped mispairing between direct repeats or through the deletion of a single base within a mononucleotide tract (see [Supp. Table S6](#SD3){ref-type="supplementary-material"}). Nine of these mutant alleles (seven previously unreported) resulted in frameshifts and were therefore analysed by RT-PCR. Although the analyses revealed RT-PCR products of a size equivalent to that expected of the wild-type in the case of p.E136EfsX35, p.Y174LfsX3, p.H391IfsX65 and p.L469YfsX22, no product was present when the results were evaluated for p.A21RfsX51, p.G102GfsX5 and p.G609GfsX6 (data not shown). One explanation for these findings could be that the latter three premature stop codons elicited nonsense-mediated mRNA decay (NMD) and hence the corresponding abnormal transcripts would have undergone degradation \[[@b20]\]. RT-PCR analyses were also performed to analyse transcripts bearing the three novel stop codons (p.R69X, p.K88X and p.R127X). Whereas the transcript harbouring p.R127X appears to be unstable (since the *GALC* mutation *in trans* and the associated SNPs were invariably found in apparent homozygosity in patient \#16), the transcript harbouring p.R69X was found to be expressed. The situation pertaining with the p.K88X mutation (occurring in *trans* to p.Y490N in patient \#17) was however found to be more complex in that it also impacts on splicing. An RT-PCR-fragment spanning exons 1-6 of the *GALC* gene yielded two products in this patient, one of the expected size (630 bp), probably carrying the p.Y490N missense mutation, and one smaller (561 bp). Both fragments were cloned and sequenced; although the 603 bp fragment did not exhibit any alteration within the region analysed, the smaller 561 bp fragment was found to contain a 69 bp in-frame deletion corresponding to the entire length of exon 2 ([Figure 2](#fig02){ref-type="fig"}). It would appear that the c.262A\>T transversion, occurring in the first base of the last codon (AAG) in this exon, affects one of the variant donor splice sites of exon 2, introducing a premature stop codon at its last residue (K88X). Although it is likely that the mutation directly abrogated the canonical exon 2 donor splice site, it is also possible that the splicing machinery acted so as to restore the open reading frame simply by removing exon 2 in its entirety \[[@b21]\]. ::: {#fig02 .fig} Figure 2 ::: {.caption} ###### RT-PCR on patient (Pt) \#17 \[p.K88X\]+\[p.Y490N\]. **A:** RT-PCR analysis performed on the RNA sample using primers encompassing exons 1-6 ([Supp. Table S1](#tbl3){ref-type="table"}), revealing the presence of an abnormally shorter *GALC* transcript (561-bp) in addition to the normally-sized product (630-bp). **B:** The graphical representation of the result, confirmed by sequencing the two cloned products. The analysis demonstrated normal splicing (unbroken red line) of the 630-bp fragment and abnormal splicing (dotted red line) of the 561-bp product resulting in a 69-bp in-frame deletion corresponding to the entire exon 2 of *GALC* in the shorter product. The [A]{.underline}AG\>[T]{.underline}AG transversion, affecting the last codon of exon 2, is given. Note that only exons 1, 2 and 3 are graphically shown. M, marker = ϕX 174 DNA *Hae*III-digested; C1, C2, C3 = control samples. :::  ::: ### Splicing mutation To confirm the pathological authenticity of the intronic mutation c. 1489+1G\>A, RT-PCR analysis was performed on the mRNA of patient \#20. In this patient, the mutation occurred in *trans* to the rare microdeletion c.1901delT (1853delT) that was predicted to introduce a premature stop codon at residue 634 (p.L634X) (L618X) ([Table 1](#tbl1){ref-type="table"}). The c.1489+1G\>A transition, occurring within the invariant GT dinucleotide of the intron 13 donor splice-site, was expected to lead to the skipping of exon 13 by abolishing the canonical splice site. However, contrary to expectation, RT-PCR analysis of *GALC* RNA from the patient (using a primer set spanning exons 10-17, as reported in [Supp. Table S1](#tbl3){ref-type="table"}) revealed the presence of an abnormally short transcript in addition to the normally-sized 1089 bp products (data not shown), both of which were cloned and sequenced. Although sequence analysis of the two RT-PCR products confirmed the presence of the microdeletion c.1901delT in the apparently normally-sized cDNA product, only partial skipping of exon 13 was evident, with the shorter transcript lacking the last 27 nucleotides (c.1463\_1489) ([Figure 3](#fig03){ref-type="fig"}). The junction between nucleotides c.1462 (exon 13) and c.1490 (exon 14) at the new donor splice-site (probably mediated by the use of the cryptic exonic GT dinucleotide at c. 1463-1464), gave rise to a serine to asparagine substitution at residue 495 followed by a frameshift that is predicted to lead to the premature termination of translation (p.S488NfsX200). ::: {#fig03 .fig} Figure 3 ::: {.caption} ###### Schematic representation of the abnormal *GALC* splice variant in patient \#20. In the top half of the diagram (A), red dotted lines represent the aberrant splicing event which results from the c.1489+G\>A mutation within the invariant gt dinucleotide at the 5′ splice site (ss) (marked by a red arrow). As a consequence of the aberrant splicing event, the junction between nucleotides c.1462 (exon 13) and c.1490 (exon 14) occurs at a novel donor splice site, almost certainly mediated by the use of the exonic (underlined) GT dinucleotide at c.1463-1464. In the bottom half of the diagram (B), a sequence analysis chromatograph of an RT-PCR product from the patient shows the aberrantly non-canonical exon 13-14 junction at c.1462-c.1490 instead of c.1489-c.1490, consequent to the loss of 27 nucleotides (red type). Green boxes denote the portion of the genomic sequence of exons 13 and 14 that is presented in the chromatograph (B); the pink box denotes the skipped 27 nucleotides of exon 13. The black dotted line represents intron 13. :::  ::: Insights into the genotype-phenotype relationship in Krabbe disease ------------------------------------------------------------------- [Table 1](#tbl1){ref-type="table"} summarizes the clinical phenotypes of the 30 Krabbe disease patients, most of whom presented with the clinically severe infantile form. Owing to the highly heterogeneous *GALC* mutational profile, it is difficult to discern any general trends in terms of a genotype-phenotype relationship. In addition to the mutational heterogeneity, the highly variable polymorphic background manifested by each patient ([Supp. Table S3](#tbl5){ref-type="table"}) could also play a role in modulating the genotype-phenotype relationship. Consistent with previous studies, the common 30kb deletion (c. 1161+6532\_polyA+9kbdel) was invariably found in *cis* to c.550T (502T), and frequently in patients with the infantile form of the disease (but sometimes also with the juvenile form, depending on the precise combination with the other mutant alleles). Our series of patients also provided support for previous speculation that the p.G286D (G270D) mutation might be associated specifically with the juvenile/mild forms of Krabbe disease \[[@b10]; [@b4]\]. Indeed, the juvenile form of the disease in patient \#26 ([Table 1](#tbl1){ref-type="table"}), carrying the large 30kb deletion (c.1161+6532\_polyA+9kbdel) on one *GALC* allele, may have been consequent to the contribution of the putatively milder p.G286D (G270D) lesion on the second allele. Further evidence to support the less deleterious nature of this missense mutation was provided by the adult form of the disease in patient \#30 ([Table 1](#tbl1){ref-type="table"}) in whom p.G286D (G270D) occurred in *trans* to the novel (and probably highly deleterious) p.P318R mutation. The replacement of a cyclic uncharged proline (P) with a basic charged arginine (R) in position 318 was predicted by *MutPred* to impact upon protein structure both in terms of solvent accessibility (P = 0.0179) and secondary structure (loop → helix; P = 0.0259) ([Table 2](#tbl2){ref-type="table"}). Another *in silico* hypothesis was available for the novel missense mutation p.Y490N: the non-conservative substitution of tyrosine (Y) by asparagine (N) is predicted to lead to the loss of a phosphorylation site at residue 490 (P = 0.0245), consistent with the severe infantile form of the patient \#17 ([Table 1](#tbl1){ref-type="table"}). The second allele in this patient (\#17) is the nonsense mutation (p.K88X), discussed above, that predicts premature termination of translation at codon 88. In accordance with *in vitro* expression studies which indicated that no GALC enzymatic activity was evident in association with p.G553R (G537R) \[[@b4]\], *MutPred* analysis postulated a change in the alpha-helical secondary structure as consequence of the replacement of a glycine with an arginine at residue 553. Consistent with this prediction, we found this mutation in association with the severe infantile form of the disease either in homozygosity ([Table 1](#tbl1){ref-type="table"}, pts \#3, \#14, \#19) or compound heterozygosity ([Table 1](#tbl1){ref-type="table"}, pts \#4, \#11, \#16). Only once was p.G553R (G537R) found in association with a late-onset form of the disease ([Table 1](#tbl1){ref-type="table"}, pt \#29); in this case, the less severe phenotype was probably due to the contribution of p.R79H (R63H) on the second allele, confirming an already reported association with the late-onset form of the disease \[[@b4]\]. The third most frequent mutation, p.N295T (N279T), was found in two infantile onset patients (\#8 and \#15) in *trans* to p.E130K (E114K) and the common large 30kb deletion, respectively. The same p.N295T (N279T) mutation was also found in compound heterozygosity in a patient (\#28) with the juvenile form of the disease, in association with the novel frameshift mutation, p.G609GfsX6. These observations are not only suggestive of a highly detrimental effect for the p.E130K (E114K) mutation (first reported by [@b13]) but also allow us to postulate a less severe phenotype in association with the novel frameshift mutation, p.G609GfsX6. Finally, examination of our patient series revealed a possible association with the juvenile form of the disease for the novel missense mutation p.I384T, found in patient \#27 ([Table 1](#tbl1){ref-type="table"}) in *trans* to the novel nonsense p.R69X mutation (expected to introduce a premature termination codon; see above). To conclude, despite the highly heterogeneous mutational and polymorphic profiles of the *GALC* gene, studies on our large series of Krabbe disease patients have provided support for various (previously somewhat tentative) genotype-phenotype correlations that were based on relatively small numbers of patients with extremely rare mutations. In addition, the present data suggest that the *GALC* mutational spectrum underlying Krabbe disease in the Italian population is somewhat different from that reported in other patient cohorts with European ancestry. This appears to be due, at least in part, to a prevalent *GALC* missense substitution p.G553R (G537R) whose high frequency appears to be due to a founder mutation. Other known examples of *GALC* founder mutations include the large common deletion (c.1161+6532\_polyA+9kbdel) thought to have originated in Sweden \[[@b30]\], the two mutations c.1796\>G and c.1630G\>A (1748T\>G and 1582G\>A) present in the Druze and Moslem Arab populations \[[@b24]\] and c.169G\>A (121G\>A), which has been reported as being responsible for the high incidence of Krabbe disease in a restricted geographical area of southern Italy (Catania, Sicily) \[[@b13]\]. The samples were obtained from the "Cell Line and DNA Biobank from Patients Affected by Genetic Diseases" (G. Gaslini Institute) - Telethon Genetic Biobank Network (Project No. GTB07001A). Supplementary material ====================== ::: {#tbl3 .table-wrap} Supp. Table S1 ::: {.caption} ###### List of primers used for polymerase chain reaction (PCR) amplification of *GALC* gene exons and exon--intron boundaries ::: Exons amplified 5′ sequence flanking exon (bp) Forward (5′\>3′)[\*](#tf3-3){ref-type="table-fn"} Reverse (5′\>3′)[\*](#tf3-3){ref-type="table-fn"} 3′ sequence flanking exon (bp) Annealing Temperature (°C) PCR product size (bp) --------------------------------------------------------------- -------------------------------- --------------------------------------------------- --------------------------------------------------- -------------------------------- ---------------------------- ---------------------------------------- **Genomic DNA Amplification**[§](#tf3-1){ref-type="table-fn"} 1 86 \# GTCAGCATCAGCGGCCTCCT ACTGGCACCCTAGGGGAAT 134 62 415 2-3 169 CAATGATTGGCACAGAAAGG TCACAGTCCATATGCTGAGGT 70 58 619[\*\*](#tf3-4){ref-type="table-fn"} 4 72 GGTGGGGAGTGAGATGGTC CAAGGGCAAAGAAAGGATCA 172 58 359 5 64 TATTTTCAATAGCGCCAGCA GCAAAGGGGAGCAATTAAAG 131 50 381 6 71 TCGTAACGATAATCTGCTTTCTG GGTATTTCCAACACAAATTTC 101 60 211 7 77 ATCTTGGTCATAAATTCAACAGC GAGAATGTAATCAAATGGGGAGA 98 60 305 8 182 TGCTGAGACAAAAGGGCATA GATGACGCTAACAAGGCAAA 133 58 471 9 131 TTGGGTGATCCTTTTATTGTCA CACTGGCAAATCTTGCTTAAAA 140 58 397 10 106 GTTTGGATGAATCAGACTCAAAT TGGCATCTGTCTGTATGCTTATG 116 58 350 11 97 AAATTTCTGTTAATCTTGGGCATT GAACTACTGGCCTGTGACAGAA 73 60 259 12 81 TCTTGCTGGTACTGATTTTGGA TGACATTTCTGTGCCCTTTT 71 60 239 13 63 TTCTGTCCACATGAGATGAGC CATCATGCACCCAGTTTGAC 86 60 300 14 60 AGCAAGGAGAGCTTCTGAAGGA AGGTTCTTGAAATAGGAGGACCA 107 60 347 15 154 TCTTGAAGCCCATCTCTGCT TGCTTACATCCCTTCCCAAT 108 60 456 16 86 CCACTCAAGAACCCCACTGA GTCACACTTTCCCCCTCCTA 94 60 256 17 140 GGAATTGTGTTTTGCTGTGG ACCAGTTTTCCCCTTGGAAT 200 \#\# 60 487 **cDNA Amplification**[°](#tf3-2){ref-type="table-fn"} 1-6 GAGTCATGTGACCCACACAA TTGGCATTATATGACCTCTCAT 62 630 1-10 GAGTCATGTGACCCACACAA TAAGCCATCAGTCAGAGCTACG 65 1150 10-17 AGTTTACTCAACCTGGCTGG AACAAGAATTGGCTCTGAACC 59 1089 § *GALC* gene: GenBank-EMBLaccession no. NC\_000014.8 ° *GALC* cDNA GenBank-EMBL accession no. NM\_000153.2 \* the primersare located 5′ (upstream) and 3′ (downstream) to the intronic sequences flanking each exon \*\* in this case, the size also includes intron 2; part of genomic region at 5′ UTR(\#)and 3′ UTR(\#\#) ::: ::: {#tbl4 .table-wrap} Supp. Table S2 ::: {.caption} ###### Oligonucleotide primers used for PCR amplification of the novel missense mutations and polymorphisms ::: Location Forward (5′\>3′)\# Reverse (5′\>3′)\# Annealing Temperatur e (°C) PCR product size (bp) Restriction enzyme -------------------------------------------------- ------------ ------------------------------------------------------- --------------------------------------------------------- ----------------------------- ----------------------- -------------------- **Mutation**[§](#tf4-1){ref-type="table-fn"} Ex. 9 c.953C\>G TTGGGTGATCCTTTTATTGTCA CAACCCGCATCTCCCATGA[\*](#tf4-2){ref-type="table-fn"} 58 195 Mnl I Ex. 9 c.967G\>A AGTTGCCTTATGGGAGCTGC[\*](#tf4-2){ref-type="table-fn"} CACTGGCAAATCTTGCTTAAAA 60 229 Pst I Ex. 10 c.1151T\>C GTTTGGATGAATCAGACTCAAAT ACAAAAGTTTACCATGGTTCCA[\*](#tf4-2){ref-type="table-fn"} 56 246 Bsr I Ex. 13 c.1468T\>A TTCTGTCCACATGAGATGAGC CCAACATTGAAATCCACCTTAT[\*](#tf4-2){ref-type="table-fn"} 59 215 Bstx I **Polymorphism**[§](#tf4-1){ref-type="table-fn"} UTR 5′ c.42GC GCTTCCTGGCAACGCCGATC[\*](#tf4-2){ref-type="table-fn"} ACTCATGGCCCTCTTCCTTT 62 309 Taq I Ex. 1 c.61GC GTCAGCATCAGCGGCCTCCT ACTCATGGCCCTCTTCCTTT 62 415 Sma I Ex. 1 c.75CA GTCAGCATCAGCGGCCTCCT ACTCATGGCCCTCTTCCTTT 62 415 Hga I Intr. 1 c.195+34GT GTCAGCATCAGCGGCCTCCT ACTCATGGCCCTCTTCCTTT 62 415 Bbv I Intr. 7 c.752+56TC ATCTTGGTCATAAATTCAACAGC GAGAATGTAATCAAATGGGGAGA 60 305 Taq I Legend: Ex=exon, Intr=intron; § *GALC* gene GenBank-EM BL accession no. NM\_000153.2 and no. NC\_000014.8 \* primer modified so asto introduce a new restriction enzyme cleavage site according to [@b26] ::: ::: {#tbl5 .table-wrap} Supp. Table S3 ::: {.caption} ###### Characteristics of putatively neutral SNPs identified in the *GALC* gene ::: MutPred analysis ♦ -------------- ------------ ---------------- --------------------- ---------- -------------------- ---------- ----------------------------------------------------------------- **UTR5′** **c.42GC (p.A14A)** **0.22** \- 10,11,16 **Ex. 1** **c.61GC (p.A21P)** **0.15** **0.62** **None** 10,16 **Ex. 1** **c.75CA (p.G25G)** **0.11** \- 10,11,16 **Intr. 1** **c.195+34GT** **0.17** \- 10,16 **Intr. 2** rs2245387 216+108GA c.264+108GA 0.48 \- 5,6,7,8,15,18,21,22,24,25,27,28 **Intr. 3** **c.328+19TA** NA \- 10 **Ex. 4** rs11552556 282CT (D94D) c.330CT (p.D110D) NA \- 23,30 **Ex. 5** rs1805078 502CT (R168C) c.550CT (p.R184C) 0.47 0.38 None 1,7,9,10,12,15,18,23,24,26 **Ex. 7** rs34362748 694GA (D232N) c.742GA (p.D248N) 0.09 0.52 None 10,16 **Intr. 7** **c.752+56TC** **0.13** \- 16 **Ex. 9** rs12888666 936GA (Q312Q) c.984GA (p.Q328Q) 0.48 \- 5,6,7,8,15,18,21,22,24,25,27 **Ex. 10** rs74073730 1024CT (L342L) c.1072CT (p.L358L) NA \- 3,4,11,14,16,19,29 **Intr. 10** rs17687109 1113+38TC c.1161+38TC 0.08 \- 10 **Intr. 10** rs11300320 1114-4delT c.1162-4delT NA \- 1,4,7,8,13,15,16,17,18,21,24,30 **Intr. 12** rs74076317 1290+23TC c.1338+23TC NA \- 3,4,11,14,16,19,29 **Ex. 13** rs398076 1302CT (S434S) c.1350CT (p.S450S) 0.44 \- 1,2,6,11,12,13,17,18,21,23,24,25,26,27,28,29,30 **Ex. 14** rs367327 1572GA (T524T) c.1620GA (p.T540T) 0.15 \- 1,2,5,6,7,8,9,10,11,13,15,16,17,18,19,20,21,22,23,24,25,28,29 **Ex. 14** rs9672064 1584TC (D528D) c.1632TC (p.D544D) NA \- 19 **Intr. 14** rs366615 1622+60CT c.1670+60CT 0.48 \- 1,4,9,11,19,21,22,24,27,29 **Intr. 14** rs12432149 1623-15CT c.1671-15CT 0.48 \- 5,7,8,10,15,16,17,20,21,22,24,25,27,28,29,30 **Ex. 15** rs398607 1637TC (I546T) c.1685TC (p.I562T) 0.47 0.53 None 1,2,4,6,9,12,13,17,21,22,23,24,25,27,28,29,30 **Ex. 15** rs421466 1650AT (V550V) c.1698AT (p.V566V) 0.17 \- 1,2,4,5,6,7,8,9,10,11,12,13,15,16,17,18,20,21,22,23,25,28,29,30 **Intr. 15** rs448805 1786+5CG c.1834+5CG 0.19 \- 1,4,8,11,16,28,29,30 **Ex. 17** rs421262 1873AG (T625A) c.1921AG (p.T641A) 0.17 0.58 None 1,2,4,5,6,7,8,9,10,11,12,13,15,17,18,20,21,22,24,25,28,29,30 Legend: dbSNP, the Single Nucleotide Polymorphism database, is available at <http://www.ncbi.nlm.nih.gov/projects/SNP/> \* GenBank-EMBL accession no. NM 000153.2 \*\* GenBank-EMBL accession no. NP 000144.2 § the minor allele frequencies were available at dbSNP or calculated from 200 Italian control alleles (novel polymorphisms, in bold) NA: not available \# patient number according to [Table 1](#tbl1){ref-type="table"}. ♦ [@b12] and [@b18]. ::: Supp. Table S4. Evolutionary comparison of the protein sequences flanking the missense mutations identified in the human GALC protein with their orthologous counterparts in nine eukaryotes Supp. Table S5. Evolutionary comparison of the protein sequences flanking five missense SNPs identified in the human GALC protein with their orthologous counterparts in nine eukaryote Supp. Table S6. Different types of repeats (underlined and highlighted in yellow) that could have mediated micro-deletions, micro-insertion/deletions (indel) and micro-duplications in the *GALC* gene. Mutated nucleotides are shown in bold upper-case red letters. [^1]: Communicated by William S. Sly

PubMed Central

2024-06-05T04:04:19.296513

2010-12-01

PMC3052422

Introduction {#Sec1} ============ Rockwood's classification of acromioclavicular dislocation is based on the degree and direction of clavicular displacement \[[@CR1]\]. Grades I and II are benign and are widely regarded as best managed conservatively \[[@CR2], [@CR3]\]. There is a general consensus that type V and VI lesions should be treated operatively \[[@CR2], [@CR4]\]. However, there remains controversy over the optimal management strategy for grade III and IV injuries \[[@CR4]--[@CR7]\]. Grade III is classified as a superior displacement of the lateral end of the clavicle of one clavicular diameter or 1 cm on the anteroposterior radiograph, whilst grade IV is described as a separation of the acromioclavicular joint with the distal clavicle displaced posterior into the trapezial fascia \[[@CR6], [@CR7]\]. In both grades the acromioclavicular and coracoclavicular ligaments are torn. Advocates of non-operative treatment suggest that patients often regain excellent clinical results and painless shoulder function, although for some there is the potential for chronic instability and pain \[[@CR8], [@CR9]\]. Alternatively, operative treatment strategies are able to address these shortcomings, but occasionally compromise shoulder function \[[@CR2], [@CR8]\]. Given this degree of equipoise, the purpose of this study was to compare the clinical outcomes of patients managed operatively and non-operatively following grade III acromioclavicular dislocation. Materials and methods {#Sec2} ===================== Study eligibility {#Sec3} ----------------- To be eligible for inclusion in the systematic review, studies had to compare operative to non-operative management following an acute, closed grade III acromioclavicular dislocation. Studies had to report at least one outcome of interest (see below). All randomised controlled trials (RCTs) and non-randomised controlled trials (nRCT) were included. All included studies reported that all patients recruited gave informed consent prior to being included. All studies were authorized by a local ethical committee, and performed in accordance with the ethical standards of the 1964 Declaration of Helsinki, as revised in 2000. Search strategy {#Sec4} --------------- The electronic databases: MEDLINE, Embase, Cinahl, Ahmed, Cochrane library and Scopus were searched from their inception to 1st May 2010 in accordance to PRISMA guidelines \[[@CR10]\]. A secondary search was conducted reviewing unpublished literature databases including: Greynet, SIGLE, National Technological Information Service, British Library Integrated catalogue, Current Controlled Trials and the Cochrane Central Register of Controlled Trials. In order not to omit any important papers, a broad search was initially undertaken using the MeSH terms and Boolean operators ("acromi\$" OR "acromioclavicular") AND ("injur\$" OR "disrupt\$" OR "dislocat" OR "subluxat\$" OR "ruptur\$") AND ("operat\$" OR "surg") AND ("conservat\$" OR "non-surg\$" OR "immobilis\$" OR "rehabilit\$" OR "physical therapy" OR "physiotherapy"). The reference lists of all potentially eligible studies were reviewed. Finally, the corresponding authors of all eligible studies were contacted and asked to review the search results to identify any studies which may have been initially missed. Study identification {#Sec5} -------------------- Two reviewers (TS, CH) independently screened the titles and/or abstracts of all identified citations against the inclusion and exclusion criteria. The full texts of all potentially eligible studies were obtained. These were then reviewed against the eligibility criteria before inclusion in the review. Data extraction {#Sec6} --------------- One reviewer extracted all the data onto a pre-defined database (CH). This was then independently verified by a second reviewer for accuracy (RC). Data collected included patients' characteristics, study design, interventions, follow-up periods and relevant outcomes. Methodological appraisal {#Sec7} ------------------------ Study methodological assessment was evaluated using the PEDro score. This is an eleven-item critical appraisal tool which assesses documentation of eligibility, subject allocation and randomisation, subject assessment and blinding, subject follow-up, data assessment and analysis. This has previously been demonstrated to be a reliable and valid scoring system \[[@CR11], [@CR12]\]. The critical appraisal was conducted by one reviewer (CH), and independently verified by a second reviewer (RC). Outcomes of interest {#Sec8} -------------------- The primary outcome was the Constant score \[[@CR13]\]. Secondary outcomes included: duration of sick leave, strength, pain, cosmetic outcome, implant failure, infection rate, throwing ability, loss of reduction of anatomical position, ossification of the coracoclavicular ligament, range of motion, and the incidence of acromioclavicular joint osteoarthritis (OA). Data analysis {#Sec9} ------------- An assessment of study heterogeneity was made by observing for population or interventional differences between the studies from the data extraction tables. Secondly, statistical heterogeneity was evaluated using the chi^2^ (*χ*^2^) test and *I*^2^ statistics. For outcomes when *I*^2^ and*χ*^2^ were less than 20% or *P* \< 0.05, a fixed-effects model was adopted. When these assumptions were not met, a random-effects model was adopted. A meta-analysis was conducted where appropriate to pool outcomes. For dichotomous outcomes, the effects measure was the risk difference (RD). For continuous outcome measures, the effect measure was mean difference (MD) or standardised mean difference (Std MD). In each case, a *P* \< 0.05 was considered statistically significant, and 95% confidence intervals (CI) were calculated. The principal analysis was to compare outcomes between operative and non-operative management of acromioclavicular joint grade III dislocations. A secondary analysis included a sensitivity analysis to compare outcomes for RCTs only. Publication and small study bias was assessed using a funnel plot. All meta-analyses were performed using the Review Manager software (RevMan Version 5.0; Nordic Cochrane Centre, Copenhagen, Denmark) and the Mantel--Haenszel method \[[@CR14]\]. Results {#Sec10} ======= Search strategy results {#Sec11} ----------------------- A total of 724 citations were identified (Fig. [1](#Fig1){ref-type="fig"}). Twenty-four were identified as potentially relevant. On second review, thirteen were deemed not appropriate, whilst one study reported the outcomes of the same cohort in two publications \[[@CR15], [@CR16]\]. The most recent version of this paper was included in the review \[[@CR15]\]. Four studies did not clearly define the grade of acromioclavicular displacement \[[@CR17]--[@CR20]\]. To minimise review heterogeneity, these studies were excluded, leaving six eligible studies. All were retrospective case series. The funnel plot of infection rate indicated mild evidence of small study exclusion and publication bias (Fig. [2](#Fig2){ref-type="fig"}).Fig. 1PRISMA chart illustrating the results of the search strategyFig. 2Funnel plot illustrating publication bias using the cosmetic results outcome measure Methodological quality {#Sec12} ---------------------- The findings of the PEDro critical appraisal indicated that the methodological quality of the current evidence base was poor (Table [1](#Tab1){ref-type="table"}). Although all studies clearly defined their study participants, only two studies demonstrated baseline comparability between the operative and non-operative groups \[[@CR21]\]. Furthermore, no study randomised their patients to the allocated intervention. No study based their sample size on a power calculation. Whilst it may have been impractical to blind subjects or clinicians to treatment allocation, no study blinded their assessors during the investigations. Although subject drop-out was more than 85% in all but two studies, no study analysed their results by intention-to-treat principles, or adjusted their results to estimate this missing data. Nonetheless, all clearly described their results and appropriately used descriptive and inferential statistical tests to analyse their cohorts.Table 1PEDro scoreCalvo et al. \[[@CR8]\]Fremerey et al. \[[@CR15]\]Galpin et al. \[[@CR26]\]Gstettner et al. \[[@CR22]\]Press et al. \[[@CR21]\]Taft et al. \[[@CR9]\]Eligibility criteria111111Random allocation000000Concealed allocation000000Baseline comparability111100Blind subject000000Blind clinician000000Blind assessor000100Adequate follow-up (≥85%)100001Intention-to treat analysis000000Between-group analysis111111Point estimates and variability111111Total score5445341: criterion satisfied; 0: criterion not satisfied Study characteristics {#Sec13} --------------------- In total, 380 patients were included in the review (Table [2](#Tab2){ref-type="table"}). The operative management cohort consisted of 195 shoulders, 125 males and 15 females with a mean age of 24.4 \[standard deviation (SD) = 4.5\] years. The non-operative group consisted of 185 shoulders, 96 males and 13 females with a mean age of 27.8 (SD = 6.1) years. One study did not document the cohort age or gender, and therefore total numbers for age and gender are incomplete \[[@CR9]\]. Five studies solely evaluated outcomes in patients with Grade III Rockwood injuries. One study's cohort consisted of 78% grade III injuries, and 22% grade V injuries \[[@CR15]\]. Given this high proportion, and since this study provided some outcomes based on grade of injury separately, this study was included in the review.Table 2Study characteristicsStudyStudySampleAge (years)Gender (m/f)Gd of ACJ dis.Surgical mgmtNon-surgical mgmtMean follow up (years)OpNOpOpNOpOpNOpCalvo et al. \[[@CR8]\]Retro3211403527/511/0Type IIIPhemister techniqueSling---type not specified. Physiotherapy---type not specifiedOp: 123 mNOp: 41 mFremerey et al. \[[@CR15]\]Retro514633.735.948/339/777Gd III20Gd VPDSPhysiotherapy exercisesOp: 6.1NOp: 6.5Galpin et al. \[[@CR26]\]Retro1621293716/017/3Gd IIIBosworthSling---type not specified. Strength and ROM exercisesOp: 35.0 mNOp: 33.7 mGstettner et al. \[[@CR22]\]Retro282237.236.225/320/2Gd IIIHook plateSling---type not specified. Physiotherapy---type not specified34 mPress et al. \[[@CR21]\]Retro161030.749.612/49/1Gd IIIWeaver--Dunn procedureSling---type not specified. Rehabilitation---not specified32.3 mTaft et al. \[[@CR9]\]Retro5275N/SN/SN/SN/SGd IIIBosworth screw or 1--3 Steinmann pin fixationSling, Kenny--Howard splint, taping or cast. Mobilisation exercisesOp: 10.8Nop: 9.5*av* average, *Co* conservative management, *Comp* complete, *Dis* disruption, *exs* exercises, *Gd* grade, *gp* group, *immob* immobilised, *inj* injury, *K wire* Kirschner wire, *Lig* ligament, *m* months, *Mgmt* management, *ND* Not documented, *Op* operative management, *physio* physiotherapy, *Pros* Prospective, *rec* recreational, *recon* reconstructed, *Ret* retrospective, *RTA* road traffic accident, *Sed* sedentary, *sh* shoulder, *wks* weeks The operative procedures were clearly described in all papers. Five studies included fixation using Kirschner wire or screw fixation methods, whilst one study used hook plates as the form of fixation \[[@CR22]\]. All studies reported repair of coracoclavicular and acromioclavicular ligaments using sutures. The non-operative management adopted was poorly described. All subjects were immobilised using a sling of some description. However, Taft et al. \[[@CR9]\] reported immobilising patients using a taping technique or cast, but did not specify the method of application. Immobilisation varied between the studies from 2 weeks \[[@CR8]\] to 4 weeks \[[@CR9]\]. The remaining papers reported immobilising patients until pain and symptoms had resolved. Following this, subjects commenced range of motion and/or strength rehabilitation programmes, but this was not described in detail in the studies. The follow-up period ranged from 32 months \[[@CR21]\] to 10.8 years \[[@CR9]\]. Meta-analysis {#Sec14} ------------- The results of the meta-analysis are shown in Table [3](#Tab3){ref-type="table"}. The primary outcome of this study was the Constant score. This revealed that there was a significantly better functional outcome following operative compared to non-operative management of grade III acromioclavicular separation (MD = 9.70; 95% CI: 1.00, 18.40; *P* = 0.03; Fig. [3](#Fig3){ref-type="fig"}). However, this is based on the complete data from one study \[[@CR22]\]. There was no statistically significant difference between the interventions in respect to strength, pain, throwing ability, loss of anatomical reduction, ossification of the coracoclavicular ligament or acromioclavicular joint osteoarthritis (*P* \> 0.05). There were significantly poorer cosmetic results following non-operative management (RD = 0.64; 95% CI: 1.09, 0.19; *P* \< 0.0001; Fig. [4](#Fig4){ref-type="fig"}). The results also suggested that there was a significantly greater duration of sick leave following operative management compared to non-operative management (MD = 3.3; 95% CI: 2.10, 4.50; *P* \< 0.001; Fig. [5](#Fig5){ref-type="fig"}). Although a relatively low incidence, unsurprisingly, the infection rate was significantly higher in the operative compared to the non-operative group (RD = 0.05; 95% CI: 0.01, 0.09; Fig. [6](#Fig6){ref-type="fig"}).Fig. 3Forest plot illustrating constant scoreFig. 4Forest plot illustrating cosmetic outcomeFig. 5Forest plot illustrating duration of sick leaveFig. 6Forest plot illustrating infection rateTable 3Results of the meta-analysisOutcomeStudiesEffect estimate*P*-valueHeterogeneity*I* ^2^Chi^2^ (*P* value)Duration of sick leave23.30 (2.10, 4.50)\<0.0001NENEConstant score29.70 (1.00, −18.50)0.03NENEThrowing ability3−0.00 (−0.15, 0.15)0.9800.57Strength (≥90% normal)2−0.01 (−0.12, 0.11)0.9000.82Strength (≤70% normal)20.35 (0.04, 3.51)0.3700.88No pain20.90 (0.33, 2.41)0.8300.60Severe pain2−0.00 (−0.06, 0.06)0.9500.97Poor cosmetic outcome4−0.79 (−0.92, −0.66)\<0.0001520.10Tenderness over the acromioclavicular joint20.08 (−0.23, 0.40)0.61750.05Implant failure20.04 (−0.05, 0.13)0.4200.91Infection50.05 (0.01, 0.09)0.0300.66Loss of anatomical reduction20.50 (−1.07, 0.52)0.5098\<0.0001Ossification of the coracoclavicular ligament20.17 (−0.32, 0.66)0.50820.02Acromioclavicular joint osteoarthritis30.12 (−0.21, 0.46)0.46890.001*NE* not estimable One study assessed the effect of range of motion. Fremerey et al. \[[@CR15]\] reported no substantial difference between the interventions, with two patients demonstrating a loss of abduction and external rotation following operative management compared to one patient following non-operative rehabilitation (*P* \> 0.05). Finally, Press et al. \[[@CR21]\] reported loss of reduction from the anatomical position. They documented that two patients following operative management presented with loss of reduction, compared to no cases following non-operative management. Sensitivity analysis {#Sec15} -------------------- It was not possible to undertake a sensitivity analysis since none of the studies included in the meta-analysis were randomised controlled trials. Discussion {#Sec16} ========== The principal finding of this study was that, for the majority of outcomes, there was no statistically significant difference in clinical or radiological outcomes between operative and non-operative management for this patient group. Nonetheless, there was some evidence to suggest that operative management provided a significantly better Constant score compared to non-operative following grade III acromioclavicular dislocation, but this was based on the results from a single study. Non-operative management was associated with significantly poorer cosmetic outcome but less sick leave compared to operative management (*P* \< 0.001). The current evidence base presented with a number of methodological limitations, including not randomising patients to group allocation, permitting allocation bias \[[@CR23]\], and not blinding assessors to subject groups, therefore increasing the risk of assessment bias \[[@CR24]\]. Finally, the studies did not base their sample sizes on power calculations, increasing the risk of a type II statistical error due to an insufficient sample size \[[@CR25]\]. Accordingly, future robust, well-designed RCTs are required to improve the currently poor evidence base in order to determine the optimal management strategy for patients following grade III acromioclavicular separation. A previous meta-analysis by Philips et al. \[[@CR6]\] ultimately advised against surgical treatment following grade III acromioclavicular separation. This review differed to the previous review as it specifically included only those studies with cohorts of predominantly grade III acromicoclavicular separation. Furthermore, with the advantage of time, we have been able to include a number of studies which have recently been published on this topic. Whilst there is agreement with some of Philips et al.'s \[[@CR6]\] conclusions, this study concludes, with some reservations, that there is little difference in the outcome of operative and non-operative management for patients following grade III acromioclavicular separation, with the exception that non-operative management provides cosmetically poorer outcomes. A more recent paper \[[@CR22]\] has shown that maintenance of reduction is possible with the operative group having a statistically better outcome than the non-operative group. As operative techniques improve, there may be a paradigm shift from the historically poor results of fixation with K-wires. The mechanism of injury appeared similar among the studies, with a combination of sporting, accidental and occupation trauma as the associated factor. Few studies distinguished whether upper limb dominance was a factor in outcome. This may have a been particularly important confounding variable for functional-based outcomes and return to sports measures, where those with a dominant limb injury may present with poorer outcomes---particularly during early review---compared to non-dominant limb injury. A further confounding factor which may have affected outcome was time from injury to surgery. Rolf et al. \[[@CR2]\] reported that those patients who had an acute acromioclavicular reconstruction after trauma reported significantly better functional outcomes and patient satisfaction rates as well as lower complication rates compared to patients with delayed reconstruction. Whilst the four studies reported that all operations were acute, the duration from injury to surgical reconstruction was not clearly stated in the papers of Galphin et al. \[[@CR26]\] or Taft et al. \[[@CR9]\]. Finally, to the study's credit, the follow-up period of the evidence base was reasonable, providing some evidence for detecting late failures and longer-term outcomes. The results of this meta-analysis indicate that there was no significant difference in respect to maintenance of anatomical reduction between operative and non-operative management of grade III acromioclavicular separation (*P* = 0.15). Calvo et al. \[[@CR8]\] acknowledged that complete reduction may not necessarily be a pre-requisite for optimal functional outcome \[[@CR6], [@CR8], [@CR27]\]. They suggested that the rationale of surgical reconstruction to achieve anatomic alignment for full functional recovery may not always be achieved following grade III acromioclavicular separation \[[@CR8]\]. Thus, anatomical reduction alone cannot justify operative intervention. However, the method of assessing anatomical alignment was unclear from the included studies. Previous authors have argued that only by assessing the acromioclavicular joint with stress radiography can anatomical position be determined \[[@CR28]\]. Accordingly, future study is recommended to determine the optimal method of radiographic evaluation of acromioclavicular displacement following operative and non-operative management strategies. The current meta-analysis suggests that there was no difference in the incidence of OA or ossification of the coraclavicular ligament between the two management strategies. Authors such as Calvo et al. \[[@CR8]\] have suggested that the incidence of OA changes may be related to the surgical manipulation and inability to maintain reduction, whilst ossification of the coracoclavicular ligaments has been associated with the manipulation of ligament tissue when attempting to repair it \[[@CR8], [@CR29]\]. Fremerey et al. \[[@CR15]\] and Taft et al. \[[@CR9]\] suggested that post-traumatic OA in surgically managed patients is related to the unphysiological contact of traumatised joint surface and subsequent joint cartilage injury. Several authors have suggested that surgical reconstruction should be advocated for those patients who have physically demanding occupations or sporting interests. However, since the mean age of each study's cohort was under 28 years, and the mechanism of injury was largely sporting or occupationally related, there was little evidence to substantiate this claim based on clinical outcomes. Furthermore, since this study suggested that duration of sick leave was significantly higher following non-operative procedures, and that there was no significant difference in strength outcomes, then non-operative management may be seen as superior to manage this patient group. For those patients who carry heavy weights on their shoulders, such as soldiers carrying rucksacks, operative intervention may be indicated to prevent anatomical deformities from affecting return to normal activities. The literature poorly described the non-operative management strategies used. Historically, various straps, harnesses, casting techniques and traction methods have been used as part of closed reduction \[[@CR30]--[@CR34]\]. Currently, there appears greater support for the use of internal rotation slings. Since non-operative management strategies were not clearly defined, it remains unclear as to whether there was a variation in these strategies between the studies. Furthermore, it also remains unclear as to whether clinical outcomes are affected by the type of rehabilitation programme adopted, immobilisation method or period of immobilisation. As Gstettner et al. \[[@CR22]\] acknowledged, the disadvantage of all operative strategies is the risk of complications. This was mirrored by our study, which demonstrated a significantly higher risk of infection following surgical management compared to non-operative treatment (*P* = 0.03). However, the incidence of infection was relatively low following acromioclavicular surgery. There has been a paradigm shift in clinical practice. Earlier studies adopted Phemister fixation methods. This developed into a consensus of using Bosworth screw and then later Hook plate fixation methods \[[@CR35]--[@CR38]\]. Currently, TightRope fixation methods and biodegradable slings have been introduced \[[@CR39], [@CR40]\]. Whilst clinical differences between operative and non-operative strategies have evaluated previous surgical interventions, the comparison to biodegradable sling fixation is yet to be evaluated using a large, well-designed RCT. Finally, no study compared cost-effectiveness with a formal economical evaluation. Since the meta-analysis indicated that whilst patients reported a shorter duration of sick leave following non-operative management, and that higher costs of hospitalisation, the operative procedure and prolonged rehabilitation are associated with this strategy, there initially appears to be greater support from an economic perspective for adopting a non-operative management strategy for this patient group. Formal health economical assessment is therefore imperative to assess the differences in this and clinical outcomes when developing the evidence base with well-designed, sufficiently powerful RCTs. To conclude, based on the current evidence base, operative management of grade III acromioclavicular dislocations results in a better cosmetic outcome (*P* \< 0.0001) but a greater duration of sick leave (*P* \< 0.001) compared to non-operative management. There was no difference between the two interventions in terms of strength, pain and throwing ability (*P* \> 0.05). Study performed at the University of East Anglia and St George's Hospital, UK. We would like to thank the staff at the Sir Thomas Browne Library, Norfolk and Norwich University Hospital for their assistance in gathering the papers required for this review. We also thank Ms Leigh Davies, Senior Orthopaedic Physiotherapist, Norfolk and Norwich University Hospital, for her assistance in the preparation of this paper. Conflict of interest {#d32e1754} ==================== None. Open Access {#d32e1759} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.305393

2011-2-23

PMC3052424

Introduction {#Sec1} ============ Venous thromboembolism (VTE) represents a problem of relevant clinical and social impact. Recent data indicate an incidence of VTE of approximately 900,000 cases per year in the USA and of approximately 770,000 in Europe; in addition, pulmonary embolism (PE) is the direct cause of almost 10% of in-hospital deaths \[[@CR1]\]. Anticoagulant prophylaxis for preventing VTE is a well-established procedure in hip (HR) and knee (KR) replacement surgery and in treating femoral neck fractures (FNF). Several meta-analyses indicate that in patients undergoing this kind of prophylactic treatment, an important reduction in symptomatic VTE is seen, with no relevant increase in major bleeding events \[[@CR1]--[@CR5]\]. These observations led the American College of Chest Physicians to generate universally recognized grade Ia recommendations on the need to initiate prophylaxis before all HR surgical interventions and to prolong treatment during the following 5 weeks. At present, pharmacological and/or mechanical prophylaxis is started in all cases of major orthopedic surgery (MOS), including elective HR and KR, and FNF surgery, as well as in several other cases of fracture (high-impact trauma, multiple fractures, multiple trauma) \[[@CR1]\]. The Italian Society for Studies on Haemostasis and Thrombosis (SISET) has been focussing its research efforts on this topic for many years \[[@CR2], [@CR3]\]. When the need to formulate practical recommendations arose in the world of orthopedics and traumatology, the Italian Society of Orthopaedics and Traumatology (SIOT) and the association of Orthopaedists and Traumatologists of Italian Hospitals (OTODI) identified SISET and the Italian Society of Anesthesia, Analgesia, Resuscitation and Intensive Care (SIAARTI) as their natural counterparts. This intersociety consensus statement aims at simplifying the grading system reported in the literature, and its goal is to improve its clinical application. For this reason, we believed that there was no need to define the strength of recommendations provided, as national and international dedicated guidelines already exist \[[@CR1]--[@CR5]\]. This statement is therefore addressed to the Italian scientific community and institutions with the aim of attaining good clinical practice in the profession. The present statement will be published in the Journals of the different Societies participating in this consensus. Purpose {#Sec2} ======= Four purposes have been identified: Keeping patients as safe as possible concerning the possibility of a thromboembolic event as a potential sequela in case of HR, KR, or FNF surgery in adults.Reducing the possible complications linked to antithrombotic prophylaxis following HR, KR, or FNF surgery as much as possible.Providing all specialists involved with unequivocal indications on the types of antithrombotic prophylaxis to be followed, in keeping with data reported in the national and international literature and with the laws in force in Italy.Supplying useful suggestions on daily clinical practice in all situations in which no clear evidence is provided. Patients {#Sec3} ======== Patients were subdivided into three groups: Patients at high risk of VTE;Patients at high risk of bleeding;Particular or "fragile" patients requiring individualized treatment. Patients at high risk of VTE {#Sec4} ---------------------------- All patients undergoing HR, KR, or FNF surgery are at high risk of VTE and must follow an antithrombotic prophylaxis protocol. In-depth hematological and instrumental screening in quest of additional risk factors for thromboembolism is not believed to be essential, as knowledge of these factors would not alter the prophylactic strategies. An exception is made for patients with past episodes of deep vein thrombosis (DVT) and/or pulmonary embolism, who require an individualized preventive/curative approach. Patients at high risk of bleeding {#Sec5} --------------------------------- Patients at high risk of bleeding are described in Table [1](#Tab1){ref-type="table"}.Table 1Patients at high risk of bleeding and patients who need careful evaluation for possible risk of bleedingPatients at high risk of bleedingPatients to be carefully evaluated for possible risk of bleedingProlonged PT (INR \> 1.5)Prolonged APTT (except antiphospholipid antibody syndrome)Thrombocytopenia \< 50,000/µlKnown bleeding diathesisSevere CRF (creatinine clearance \< 30 ml/min)Chronic liver disease with prior bleeding episodesFamily or personal history of major bleedingMultiple trauma (ISS ≥ 15)Concomitant use of drugs affecting hemostasis (e.g., antiplatelet drugs, anti-inflammatory drugs)*PT* prothrombin time,*INR* International normalized ratio,*ISS* injury severity score,*APTT* antiplatelet treatment,*CRF* chronic renal failure Fragile patients {#Sec6} ---------------- Fragile patients requiring individualized treatment are those who present with:Body weight \<50 kgAge \>75 yearsModerate chronic renal failure (CRF) (creatinine clearance 30--50 ml/min) The creation of a personalized, shared folder for thrombotic and hemorrhagic risk assessment and initiation of adequate thromboprophylaxis is suggested in all hospital settings. Furthermore, we recommend that the creation of this document be suggested by all administrations involved (hospital directorate, local health authority, regional administration, etc.). Type of prophylaxis {#Sec7} =================== PharmacologicalLMWH, FON, NOA, VKA, UHMechanicalActive (IPC, VFP)Passive (GCS)CombinedPharmacological + mechanical*LMWH* low-molecular-weight heparin,*FON* fondaparinux,*NOA* new oral anticoagulants,*UH* unfractionated heparin,*VKA* vitamin K antagonists, *VFP* venous foot pump,*IPC* intermittent pneumatic compression,*GCS* graduated compression stockings Pharmacological prophylaxis {#Sec8} --------------------------- Pharmacological prophylaxis is based on low-molecular-weight heparin (LMWH), fondaparinux (FON), and new oral anticoagulants (NOA).Aspirin must not be used for VTE prophylaxis, as indicated by its label and by current guidelines.Unfractionated heparin (UH) must not be used considering that its efficacy is lower than that of LMWH, it has a short half-life, and it more frequently induces thrombocytopenia.Vitamin K antagonists (VKA) should not be administered because they are difficult to manage and maintain within a range of therapeutic anticoagulation \[International normalized ratio (INR) ranging between 2 and 3\]. Exceptions are possible but must be evaluated on an individualized basis with the consultant cardiologist or an expert in hemostasis and thrombosis. ### Low-molecular-weight heparin (LMWH) {#Sec9} Concerning HR and KR, no differences in efficacy and safety have been reported between LMWH preoperative and postoperative first administration (Table [2](#Tab2){ref-type="table"}) \[[@CR6], [@CR7]\]. LMWH labels in Italy, however, require a preoperative first administration except for bemiparin and dalteparin (for the latter only in hip surgery).Table 2Dosage and time of administration of low-molecular-weight heparin (LMHW) available in ItalyActive principleBrand nameDosage and time of administrationEnoxaparinClexane^®^4,000 IU 12 h before surgery, then 4,000 IU/dayNadroparinFraxiparine^®^ Seleparin^®^38 IU/kg 12 h before surgery and 12 h after, 38 IU/kg every 24 h during the 3 days following surgery, thereafter increasing the dose to 57 IU/kg/dayDalteparinFragmin^®^5,000 IU 8--12 h before surgery, then 5,000 IU/day. Alternatively 2 h, 500 IU 1--2 before surgery^a^ and 2,500 IU 8--12 h after, thereafter either 5,000 IU/day or (only in hip surgery) 2,500 IU 4--8 h after surgery then 5,000 IU/dayBemiparinIvor^®^3,500 IU 6 h after surgery, then 3,500 IU/day. Alternatively 3,500 IU 2 h before surgery^a^, then 3,500 IU/dayParnaparinFluxum^®^0.4 ml (4,250 anti-Xa IU) 12 h before surgery, then 0.4 ml (4,250 anti-Xa IU)/dayReviparinClivarin^®^0.4 ml (4,200 anti-Xa IU) 12 h before surgery, then 0.4 ml (4,200 anti-Xa IU)/day^a^Although reported by the product label, this type of prophylaxis is not recommended ### Fondaparinux (FON) {#Sec10} Fondaparinux has proved to be effective and safe in VTE prevention in HR, KR, and FNF (Table [3](#Tab3){ref-type="table"}) \[[@CR8]\]. In particular, FON has been reported to be more effective than LMWH (only demonstrated by decrease in phlebography-proven asymptomatic DVT) with modest, although statistically significant, increase in bleeding and need for transfusions (with no related increase in fatal hemorrhage, in critical organs, or need for reintervention).Table 3Dosage and time of administration of FondaparinuxActive principleBrand nameDosage and time of administrationFondaparinuxARIXTRA^®^2.5 mg at least 6 h after surgery, then 2.5 mg/day^a^If creatinine clearance 20--50 ml/min 1.5 mg^ba^In agreement with the latest edition of the American College of Chest Physicians (ACCP) guidelines \[[@CR1]\], initiation may be postponed up to 24 h after the end of the intervention \[[@CR9]\], although this has not been included in the label as yet^b^According to the recent guidelines of the European Society of Anaesthesiology \[[@CR10]\], FON is contraindicated if creatinine clearance \< 30 ml/min ### New oral anticoagulants (NOA) {#Sec11} New oral anticoagulants (dabigatran and rivaroxaban) have proved to be effective and safe in VTE prevention in HR and KR (Table [4](#Tab4){ref-type="table"}) \[[@CR11]--[@CR16]\]. On the other hand, no direct comparison has ever been made between the two drugs, allowing for a definite confirmation of any different efficacy and safety. There is no evidence in the literature concerning the use of NOA in patients undergoing FNF surgery and concerning prolonged prophylaxis after KR; furthermore, experience in fragile patients is limited. Although these drugs do not require laboratory monitoring, they have been shown to prolong PT and APTT.Table 4Dosage and time of administration of available new oral anticoagulants (NOA)Active principleBrand nameDosage and time of administrationDabigatran^a^(antifactor IIa)Pradaxa^®^110 mg 1--4 h after surgery, then 220 mg/dayIf age \> 75 years or creatinine clearance 30--50 ml/min or amiodarone intake, 75 mg 1--4 h after surgery, then 150 mg/dayRivaroxaban^b^(antifactor Xa)Xarelto^®^10 mg 6--10 h after surgery, then 10 mg/day^a^Dabigatran has proved not to be inferior to low-molecular-weight heparin (LMWH) both in terms of efficacy and safety. As concerns dabigatran, in the literature, there is no information available on patients undergoing regional anaesthesia \[[@CR11], [@CR12]\]^b^Rivaroxaban has shown to have greater efficacy than LMWH, with overlapping safety \[[@CR13]--[@CR16]\]. An analysis performed after publication of rivaroxaban registration study confirmed its safety in patients undergoing neuraxial anesthesia Mechanical prophylaxis {#Sec12} ---------------------- Mechanical prophylaxis is based on the use of graduated compression stockings (GCS) and on intermittent pneumatic compression (IPC) devices \[[@CR17]\]. GCS (thigh-to-foot or knee-to-foot) increase the effectiveness of pharmacological prophylaxis, must be used until recovery of good mobility with autonomous de-ambulation (best if used on both legs), must be correctly positioned avoiding the "tourniquet effect," and must not be used in the presence of peripheral arterial disease or diabetic neuropathy. IPC devices (either sural or plantar) have a high efficacy and enhance the action of anticoagulant drugs, but there is a low compliance by nurses and patients as concerns their management. When should prophylaxis be started? {#Sec13} =================================== Patients with femoral neck fracture (FNF) {#Sec14} ----------------------------------------- Selection and initiation of the prophylactic treatment to be followed strongly depend on the adopted schedule:If surgery is performed on an emergency basis (within 24 h), LMWH may be used (starting 12 h before or 12 h after) or, alternatively, FON (starting at least 6 h after the end of the intervention and, in any case, within 24 h).If surgery is postponed, LMWH must be administered early. In this case, there is no information available on the possibility of initiating FON 6--8 h after the end of the intervention, thus producing a shift between the two anticoagulant drugs. At present, no recommendation can be made on this subject.NOA must not be used, as no study pertaining to FNF has been published. Patients candidate for hip (HR) and knee (KR) replacement {#Sec15} --------------------------------------------------------- In the literature, no significant difference in efficacy and safety has been reported between preoperative and postoperative initiation of LMWH in HR and KR \[[@CR1], [@CR6], [@CR7]\]. Consequently, the choice must be based on evidence reported in published studies as well as on what is indicated on LMWH labels, which in Italy require initiation of prophylaxis 12 h before surgery, except for dalteparin and bemiparin (Table [2](#Tab2){ref-type="table"}). Both FON and NOA must always be started postoperatively (Tables [3](#Tab3){ref-type="table"}, [4](#Tab4){ref-type="table"}). How long should pharmacological prophylaxis last? {#Sec16} ================================================= Concerning the duration of pharmacological prophylactic treatment, if LMWH is used, therapy should last a minimum of 10 days in all patients, with a strong recommendation to protract prophylaxis for 35 days after HR and FNF surgery and the suggestion -- with a lower level of evidence -- to protract treatment similarly in patients undergoing KR surgery \[[@CR1], [@CR18]\]. Regardless, in Italy, it is standard procedure to protract prophylaxis for 35 days even after KR surgery. This approach is also suggested for FON therapy. As far as NOA are concerned, indications on duration of treatment derive from registration studies and are reported on the labels of dabigatran and rivaroxaban:With dabigatran, duration is 4--5 weeks in HR and 10 days in KR surgery;With rivaroxaban, duration is 5 weeks in HR and 2 weeks in KR surgery. The safety of 5-week treatment with NOA has been proven in HR studies, which suggest the reliability and feasibility of this prophylaxis duration after KR as well. Lastly, it must be remembered that further protraction of prophylaxis (longer than suggested duration) has to be addressed in patients who, for different reasons (prolonged recumbence, additional risk factors), are at risk of developing VTE complications for a longer period than usual. Anesthesia techniques and initiation of pharmacological prophylaxis {#Sec17} =================================================================== No particular problem is identified in relation to general anesthesia (GA). On the other hand, regarding regional anesthesia (RA), timing must be carefully respected with epidural or intrathecal anesthesia, whereas there are no contraindications in perineural block \[[@CR19], [@CR20]\]. It is widely accepted that RA reduces the risk of VTE and that the correct timing (prophylaxis/RA administration and, if present, catheter removal) is crucial to prevent complications. Actually, all anticoagulants used in VTE prevention in HR, KR, and FNF are closely related to the risk of developing epidural hematoma. In particular, upon removal of the epidural catheter, drug effectiveness, half-life (T~1/2~), and time to maximum concentration (T~max~) must be assessed: as a general rule, the recommendation is made to wait at least 2 half-lives before removal, resuming pharmacological prophylaxis after 8 h (period required for clot formation) minus T~max~. To simplify:*LMWH and RA* \[[@CR19], [@CR20]\]T~1/2~: 4 hT~max~: 4 h Last administration before catheter removal: at least 12 h. First administration after catheter removal: at least after 6--8 h. If LMWH is administered twice daily, either at the prophylactic or therapeutic dosage, 24 h must pass after catheter removal before proceeding with the following dose. If traumatic puncture, consider the possibility of initiating prophylaxis after 24 h.*FON and RA* \[[@CR21]\]T~1/2~: 17 hT~max~: 1 h If FON is administered at the therapeutic dosage, no central block must be performed. Last administration before catheter removal: at least 36 h. First administration after catheter removal: at least after 12 h. If traumatic puncture, consider the possibility of initiating prophylaxis after 24 h.*NOA and RA* \[[@CR10]\] As concerns the relationship between NOA and RA, there is no information available (randomized clinical studies) concerning timing and method of use; therefore, refer to what is reported on the product label:Dabigatran---not recommended in patients who must undergo anesthesia requiring the use of postoperative permanent epidural catheters, as no information is reported in the literature.Rivaroxaban---last administration 18 h before removal, resume administration 6 h after removal; recent guidelines of the European Society of Anaesthesiology suggest a longer period between last rivaroxaban dose and epidural catheter removal (22--26 h) \[[@CR10]\]. Anesthesia/patient correlation in antiplatelet treatment {#Sec18} ======================================================== See Table [5](#Tab5){ref-type="table"}. Table 5Correlation between anesthesia and antiplatelet treatment (APT)Regional anesthesia^a^General anesthesiaPatients on APT withPatients on APTAcetylsalicylic acid (ASA): do not interrupt in case of secondary prevention (75--100 mg/day)Ticlopidine---interrupt 10 days pre-opGA always feasibleIIb/IIIa inhibitors  Abciximab---RA contraindicated  Eptifibatide---interrupt 8 h pre-op  Tirofiban---interrupt 4 h pre-opClopidogrel---interrupt 7 days pre-opRisk of surgical bleeding must always be considered before surgery^a^APT, if no bleeding occurs, must be resumed the day following the intervention and, in the presence of epidural catheterization, after catheter removal Management of vitamin K antagonist (VKA) patients {#Sec19} ================================================= The main purpose is leading patients to surgical intervention with an adequate hemostasis and reducing the risk of thromboembolism as much as possible. Femoral neck fracture (FNF) patients {#Sec20} ------------------------------------ Intervention should be delayed and INR measured: If INR \> 2, administer vitamin K 10 mg in 100 ml of saline or glucose solution i.v. and measure INR every 6/8 h until INR \< 2 is attained. If INR \< 2, start LMWH administration at the prophylactic dose and timing (4,000--5,000 IU/day), plan surgical intervention as soon as possible, and request consultation by a cardiologist and/or by an expert in hemostasis and thrombosis to plan VKA resumption after surgery. Patients candidate for elective hip (HR) and knee (KR) replacement {#Sec21} ------------------------------------------------------------------ Each hospital should have a written and shared protocol concerning the management of VKA patients who have to undergo major lower-limb orthopedic surgery; consultation by a cardiologist and/or an expert in hemostasis and thrombosis should be requested to prepare a personalized schedule addressing VKA interruption and resumption; the timing of surgery must be respected, and the procedure should not be delayed. Management of antiplatelet treatment (APT) patients {#Sec22} =================================================== Aspirin administered as primary prevention must be interrupted 7 days before elective surgery, whereas it must be interrupted upon hospital admission in patients with FNF planned for surgery. Aspirin administration as secondary prevention (in patients with prior cardiovascular events) must be continued at the dose of 75--100 mg/day. Femoral neck fracture (FNF) patients {#Sec23} ------------------------------------ For FNF, APT patients should undergo surgery as soon as possible. For patients on clopidogrel or ticlopidine (or dual anti-aggregation), request consultation by a cardiologist and/or an expert in hemostasis and thrombosis. Patients candidate for hip (HR) and knee (KR) replacement {#Sec24} --------------------------------------------------------- Administration of clopidogrel or ticlopidine must be interrupted 7 and 10 days before surgery, respectively, whereas in patients receiving dual anti-aggregation (aspirin and clopidogrel), surgery must be delayed if clopidogrel interruption is expected during the following months; if interruption is not expected, request consultation by a cardiologist and/or an expert in hemostasis and thrombosis. In all such patients, as a general rule, resume APT as soon as possible and regardless, once hemostasis is achieved. Patients at high risk of bleeding {#Sec25} ================================= In patients at high risk of bleeding, the absolute contraindication to pharmacological prophylaxis is represented by ongoing major bleeding. In this case, mechanical prophylaxis is indicated. The relative contraindication is applied to all the other conditions for which patients are at high risk of bleeding reported in Table [1](#Tab1){ref-type="table"}. In these cases, pharmacological or transfusional correction of the hemostatic defect is recommended whenever indicated and feasible, considering mechanical and/or dedicated pharmacological prophylaxis (dose reduction, postoperative initiation). Special cases or fragile patients {#Sec26} ================================= Providing precise directives for managing such patients goes beyond the purpose of this consensus statement; the only indication is to pay extreme attention to them and to request consultation by an expert on hemostasis and thrombosis. Management of anticoagulant drugs in obese patients \[body mass index (BMI) \> 30\] is not considered to be different from what occurs with other patients. In patients with renal failure, labels of the single drugs administered must be referred to, and careful clinical monitoring must follow. General considerations {#Sec27} ====================== Postoperative mobilization must be started as soon as possible.Bed-rest patients should receive lower-limb mobilization exercises.General practitioners and patients should be informed how to recognize signs and symptoms of DVT and PE, how to correctly manage home prophylaxis, and about the risks of omitting it.Pharmaceutical companies and regulatory authorities (Italian Medicines Agency, etc.) should keep labels updated in agreement with scientific evidence reported in the literature. Conclusions {#Sec28} =========== This document represents a consensus statement of Italian experts, with information based on scientific knowledge and labels available during the summer of 2010, and it will be disseminated by the four societies via different modalities (society journals, society Web sites, symposia organized within national congresses, etc.). A periodical revision of this document is expected, which will be of particular importance for the use of new anticoagulant drugs currently undergoing clinical development, some of which (edoxaban, betrixaban, and others) are still undergoing a preliminary trial phase. For other drugs (apixaban), studies are already available \[[@CR22], [@CR23]\] that prove their efficacy and safety in VTE prevention in HR and KR surgery. It is therefore likely that the number of drugs available for this type of prophylaxis will increase in the near future. F. Randelli---on behalf of Società Italiana di Ortopedia e Traumatologia (SIOT). F. Biggi---on behalf of Ortopedici Traumatologi Ospedalieri di Italia (OTODI). G. Della Rocca, P. Grossi---on behalf of Società Italiana di Anestesia, Analgesia, Rianimazione e Terapia Intensiva (SIAARTI). D. Imberti, R. Landolfi, G. Palareti, D. Prisco---on behalf of Società Italiana per lo Studio dell'Emostasi e della Trombosi (SISET). Claudio Cimminiello worked in cooperation with Filippo Randelli, Davide Imberti, and Paolo Grossi in drawing up the first SIOT document in 2009. Open Access {#d32e1185} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.306890

2011-1-19

PMC3052425

Introduction {#Sec1} ============ Acromioclavicular (AC) joint injuries may occur as a result of a direct force applied to the tip of the shoulder with the arm adducted or due to indirect trauma such as a fall on the outstretched hand. These injuries were classified according to Post \[[@CR20]\] and later according to Rockwood et al. and Williams et al. \[[@CR21], [@CR28]\] into six types. Surgical treatment is advocated for types IV, V, and VI and in some cases type III where there is complete AC joint dissociation. This is due to the increased incidence of unsatisfactory results of conservative treatment of such cases \[[@CR21]\]. Most of the surgical techniques described in literature involve the use of metallic devices to reduce and maintain reduction of the AC joint. Yet, the results, though good, are not totally satisfactory \[[@CR1], [@CR2], [@CR8], [@CR14], [@CR18], [@CR22], [@CR24]\]. The metallic devices used are usually difficult to place and may be associated with complications. They require removal before the patient can return to normal activities \[[@CR15]\]. Another option for surgical treatment is reconstruction of the coracoclavicular (CC) ligament \[[@CR7], [@CR10], [@CR26]\]. This paper discusses the results of a simple, new procedure of surgical repair for acute complete AC joint dissociation (IV, V, and a few type III cases). The procedure involves the use of a loop of no. 5 nonabsorbable sutures passed under the base of the coracoid (knuckle) and through a tunnel drilled in the flat lateral end of the clavicle, with direct repair of the CC ligament. This maintains the AC joint in the reduced position until the CC ligament heals. This technique does not involve the use of metallic implants for fixation. Materials and methods {#Sec2} ===================== Twenty-one cases of acute AC joint dissociation were assessed in this prospective study. All patients had type IV or V AC joint injuries according to Rockwood classification. All surgeries were performed within 1 week of injury by the same surgeon between February 2000 and June 2003. All patients gave their written consent to undergo this type of repair. Also, we received the approval of our local ethical committee (Orthopedic Department, Cairo University) to undergo this study. There were 16 men and five women. The average age at the time of surgery was 31.8 (range 22.3--39.5) years. The duration between the time of injury and the date of surgery varied between 1 and 5 days, with an average of 2.14 days. The nondominant side was affected in six cases. Ten patients were injured while participating in sports (Table [1](#Tab1){ref-type="table"}).Table 1Patient demographyCase .noAge (years)SexDominant sideSportDuration between injury and surgery (days)Mechanism of injuryDuration of follow-up (years)137.2♂DominantBasketball2Hitting the ball-post^a^9.5223.4♂DominantNo2Fall on shoulder tip9.2328.3♂NondominantProfessional hurdles3Fall on shoulder tip^a^9.0636.2♀DominantNo1Car accident8.9523.6♂DominantNo5Fall on shoulder tip8.8638.4♀NondominantNo2Fall down the stairs8.5729.8♂DominantProfessional volleyball1Fall on shoulder tip^a^8.3835.7♀DominantNo2Car accident8.2939.5♂DominantSquash1Hitting wall with shoulder tip^a^8.01026.5♂DominantHandball2Hitting goal-post^a^7.81136.2♂DominantSoccer1Fall on shoulder tip^a^7.61227.8♂NondominantNo2Fall on shoulder tip7.51322.3♀DominantVolleyball2Fall on shoulder tip^a^7.31433.5♂NondominantNo3Car accident7.21529.8♂DominantNo1Fall down the stairs7.01634.7♂NondominantSoccer3Fall on shoulder tip^a^6.91737.5♂DominantNo2Fall on shoulder tip6.81836.3♂DominantProfessional handball4Fall on shoulder tip^a^6.61936.2♂DominantBasketball1Fall on shoulder tip^a^6.42029.4♂NondominantNo3Fall on shoulder tip6.22125.7♀DominantNo2Car accident6.0^a^ Participating in sports Injuries were documented by preoperative plain X-rays of the affected shoulder in the anteroposterior (AP), lateral scapular, outlet, and axillary views. Similar plain X-rays were performed for the opposite shoulder. Patients were evaluated preoperatively using the University of California Los Angeles (UCLA), Constant, and American Shoulder and Elbow Surgeons Shoulder (ASES) scoring systems. This study was conducted according to the principles established in Helsinki and approved by the ethical committee of the orthopedic department of Cairo University. Operative technique {#Sec3} ------------------- Under general anesthesia, the patient was placed in the beach-chair position. A slightly curved 3- to 4-cm linear incision was performed over the AC joint and along the distal half of the clavicle. The deltoid muscle (with the attached periosteum) was elevated off the anterior edge of the distal third of the clavicle. In most cases, it was already stripped off the clavicle. The deltoid was slightly inferiorly retracted until the coracoid process was exposed. A 3.2-mm drill bit was used to make a tunnel in the middle of the flat surface of the distal end of the clavicle about 2.0 cm medial to AC joint (Fig. [1](#Fig1){ref-type="fig"}). A more lateral drill hole will cause forward pull of the distal clavicle, preventing anatomical reduction, which will interfere with the AC joint capsule resuturing.Fig. 1A 3.2-mm drill bit is used to make a tunnel in the middle of the flat surface of the distal end of the clavicle Two strands of no. 5 nonabsorbable Ethibond sutures were passed under the knuckle of the coracoid process using a curved aneurismal needle as a suture passer. It was then passed through the drill hole in the clavicle, in a figure of 8 manner (Fig. [2](#Fig2){ref-type="fig"}).Fig. 2Curved aneurismal needle is used to pass two strands of the Ethibond sutures under the knuckle of the coracoid process The sutures are then passed under the knuckle of the coracoid and through the drill hole in the distal end of the clavicle (Fig. [3](#Fig3){ref-type="fig"}).Fig. 3The two strands of the Ethibond sutures are passed through the drill hole in the lateral flat end of the clavicle in a figure of 8 fashion Stay sutures were placed into both edges of the torn CC ligament using a no. 1 absorbable suture. In many cases, it was difficult to identify the conoid from the trapezoid parts. In such cases, both parts were sutured en masse. With the help of the assistant, the patient's arm was pushed up while the distal clavicle was pushed down to achieve AC joint reduction. Then, while keeping the AC joint in the reduced position, the two ends of the no. 5 Ethibond suture were tied securely to each other, forming a knot under the distal end of the clavicle (between it and the coracoid) (Fig. [4](#Fig4){ref-type="fig"}).Fig. 4Acromioclavicular (AC) joint is reduced and the two ends of the suture material are tied securely to each other, forming a knot under the distal end of the clavicle. The torn coracoclavicular ligament is then repaired The edges of the CC ligament were then repaired by tying the previously placed stay sutures together (Fig. [5](#Fig5){ref-type="fig"}). If the CC ligament was stripped off the undersurface of the clavicale, it was reattached with two transosseous sutures using a 2.5-mm drill bit. This occurred in eight cases.Fig. 5Conoid and trapezoid resutured The ruptured AC joint capsule was then resutured. If it was avulsed off the clavicle (in seven cases), it was resutured to the clavicle by transosseous sutures using sharp bone-cutting needles (Fig. [6](#Fig6){ref-type="fig"}).Fig. 6Resuturing acromioclavicular (AC) joint capsule If (as in five of our cases) there was AP instability in the AC joint, another transosseous suture was placed in front of the AC joint to prevent anterior translation of the clavicle. The deltoid and the trapezius were then fixed back onto the clavicle using a no. 1 absorbable suture. The skin and subcutaneous tissues were closed in the usual manner. ### Postoperative care {#Sec4} The patient was placed in an arm-pouch sling for 4--6 weeks postoperatively, during which light use of the hand was allowed. This aimed at supporting the elbow and arm to reduce the downward pull on the repair. Plain X-rays (AP and outlet views) were performed the first day postoperatively to document AC joint reduced position. Physiotherapy was started 6 weeks postoperatively with gradual return to normal daily activities. Sporting activities were allowed only when AC joint stability was established both clinically and radiologically. The latter was performed by doing stress views 6, 12, and 20 weeks postoperatively. After that, patients were followed every 6 months for the next 2 years and then on annual basis with special emphasis on recurrence of the deformity or osteoarthretic changes in the AC joint. Statistical analysis of the results was performed using the Wilcoxon signed-ranks test. At the final follow-up, all patients were reevaluated using UCLA, Constant, and ASES scores. Results {#Sec5} ======= Twenty-one consecutive patients underwent surgical repair for complete AC joint dissociation (types IV and V) injuries between February 2000 and June 2003. All patients underwent the previously described loop repair technique for stabilizing the AC joint. Preoperatively, all patients had aching shoulder pain, weakness, and deformity that interfered with their daily activities. They all complained of painful limitation of the active range of motion and weakness resisted forward elevation of the affected arm in contrast with the contralateral one (Figs. [7](#Fig7){ref-type="fig"}, [8](#Fig8){ref-type="fig"}, [9](#Fig9){ref-type="fig"}).Fig. 7: Pre and postoperative outlet viewFig. 8Pre and postoperative antroposterior viewFig. 9Pre and postoperative antroposterior view More than 50% superior displacement of the distal clavicle defined recurrence of deformity. Follow-up radiographs revealed maintenance of reduction in 20 of the 21 patients (95.24%). The one patient who sustained a recurrence was seen 6 weeks postoperatively without any deformity. He subsequently went back to playing squash, against medical advice, but his pain prevented him from continuing. Three months postoperatively, he presented with a recurrence of the deformity. He was followed clinically and did well by other objective outcomes measures. Other than this case, there were no intra- or postoperative complications (neurological or other types), except in one case. That patient suffered from superficial wound infection that led to wound gaping. This was successfully managed by prompt administration of systemic antibiotics for 7 days with repeated dressings. The wound finally healed by secondary intention. This patient complained about the appearance of the wound scar and sought the advice of a plastic surgeon. Although there was some anterior deltoid wasting in all patients during the first 4--6 weeks postoperatively, this almost completely disappeared in all cases by the 16th week (after 10 weeks of physiotherapy). At the final follow-up, none showed deltoid atrophy or detachment. During the long follow-up period, no patient developed clinical or roentgenographic evidence of AC arthritis. None showed distal clavicular osteolysis. Three patients developed ossification of the CC interval 3--5 years postoperatively. This was discovered in follow-up roentgenograms (all three did not participate in sports) but with no clinical effect. All patients were able to resume their daily activities. Seven who enjoyed preinjury recreational sports were able to return to sport at the same level. All three patients who were professional athletes were able to return to their preinjury sports performance level. Results according to the UCLA score \[[@CR6]\] showed a significant improvement in pain, function, active forward elevation, strength of forward flexion, and patient satisfaction. The mean total score improved from 52.8 to 95.0 at the final follow-up (Table [2](#Tab2){ref-type="table"}).Table 2Pre and postoperative University of California Los Angeles (UCLA) scoreItemsPreoperative mean ± SDPostoperative mean ± SD*P* valuePain (10)3.3 ± 1.09.1 ± 1.0\<0.001\*Function (10)7.1 ± 1.09.6 ± 0.8\<0.001\*Active FE (5)3.7 ± 0.54.9 ± 0.4\<0.001\*Strength of FF (5)3.4 ± 0.54.9 ± 0.4\<0.001\*Overall patient satisfaction (5)0.0 ± 04.8 ± 1.1\<0.001\*Total score (35)18.5 ± 2.633.2 ± 2.9\<0.001\*Total score (100)52.8 ± 7.595.0 ± 8.2\<0.001\**SD* standard deviation, *FE* forward elevation,*FF* forward flexion\* Wilcoxon signed-rank test The Constant score \[[@CR5]\] showed a significant improvement in pain, activity level, arm positioning, strength of abduction, and range of motion (in all directions). The mean preoperative score was 63.3 and the mean final score was 97.8 (Table [3](#Tab3){ref-type="table"}).Table 3Pre- and postoperative Constant scorePreoperative mean ± SDPostoperative mean ± SD*P* valuePain (15)8.3 ± 2.414.8 ± 1.1\<0.001\*Activity level (10)4.0 ± 09.7 ± 1.3\<0.001\*Arm positioning (10)7.3 ± 1.09.7 ± 0.7\<0.001\*Strength of abduction 90º or highest level patient can achieve (Pounds) (25)18.3 ± 3.424.6 ± 1.2\<0.001\*Range of motionForward flexion (10)7.3 ± 1.09.7 ± 0.7\<0.001\*Lateral elevation (10)7.3 ± 1.09.9 ± 0.4\<0.001\*External rotation (10)5.3 ± 1.09.7 ± 0.7\<0.001\*Internal rotation (10)5.3 ± 1.09.7 ± 0.7\<0.001\*Total score (100)63.3 ± 9.397.8 ± 6.2\<0.001\**SD* Standard deviation\* Wilcoxon signed ranks test The ASES score \[[@CR11]\] showed a significant improvement in both pain and activity level, as well as the total score, which was 57.2 preoperatively and 95.0 at the final follow-up (Table [4](#Tab4){ref-type="table"}).Table 4Pre- and postoperative American Shoulder and Elbow Surgeons Shoulder (ASES) scorePreoperative mean ± SDPostoperative mean ± SD*P* valuePain (50)27.4 ± 4.146.7 ± 4.8\<0.001\*Activities of daily living (50)29.8 ± 6.548.3 ± 5.5\<0.001\*Total score (100)57.2 ± 8.395.0 ± 8.2\<0.001\**SD* standard deviation\* Wilcoxon signed ranks test Discussion {#Sec6} ========== Surgical management of AC joint injuries is indicated in types IV, V, and VI injuries and in some cases type III. Many surgical techniques described in literature involve the use of metallic implants for internal fixation. These implants include Kirschner wires \[[@CR14]\], smooth or Knowles pins \[[@CR8], [@CR27]\], fully threaded pins \[[@CR1]\], Arbeitsgemeinschaft fur osteosynthesefragen (AO) screws \[[@CR18]\], and Balser hook--plate \[[@CR22]\] with or without superior AC joint wiring, which acts as a tension band. Others place an AO screw \[[@CR24]\] or a tension band \[[@CR2]\] between the clavicle and the coracoid process. Others use modified Weaver--Dunn procedure with resuturing the released coracoacromial (CA) ligament to the clavicle via transosseous sutures in addition to placing a Bosworth CC screw that is removed after 8 weeks. In one study, there was a loss of reduction in two (11.8%) of 17 cases \[[@CR19]\]. Many of these metallic implants are difficult to place and are accompanied by a high level of serious complications, such as pin migration \[[@CR15]\]. They also need to be removed before the patient can return to normal activities, with the hazards of another surgery. Other surgical techniques to stabilize the lateral clavicle include reconstruction of the CC ligament by transferring the AC ligament or using autograft or allograft tissues. Morrison and Lemos reported good and excellent results using a synthetic loop passed through drill holes in the base of the coracoid and the lateral third of the clavicle. They had good results in 12 of 14 cases \[[@CR17]\]. Chen and coworkers achieved similar results, having overall satisfactory outcome in 41 of 48 (86%) patients \[[@CR4]\]. Baumgarten and coworkers used an arthroscopcally assisted technique utilizing a semitendinosus allograft. This required a 3-cm incision and was technically tedious \[[@CR3]\]. Others reconstructed the CC and AC ligaments for both acute and chronic cases using tendon grafts and supplemented the repair by temporary fixation (Kirschner wires) \[[@CR13], [@CR15], [@CR29], [@CR30]\]. Some surgeons resected the distal clavicle and reinserted the CA ligament intramedullary (docking technique). They used it mainly in chronic cases, achieving good results with loss of reduction in one (6%) of 16 cases \[[@CR16]\]. A technique was described involving transfer of the lateral half of the conjoined tendon to the distal clavicle in a proximally based fashion with loss of reduction in 11% of cases. This technique required additional CC fixation that later needed removal \[[@CR9]\]. Other surgeons dissected the CC ligament and refixed it to the undersurface of the clavicle using transosseous sutures or anchors. This was done by open technique, with loss of reduction in five (17.2%) of 29 patients \[[@CR23]\]. It was also done via an all-arthroscopic procedure \[[@CR12], [@CR19]\]. Tienen and associates transferred the CA ligament to the clavicle and fixed it by absorbable, braided suture cord. There was failure to achieve anatomical reduction in three (14.4%) of 21 cases \[[@CR25]\]. All these techniques aim at providing a stable AC joint without the need for metal fixation. However, some techniques required additional fixation to avoid loss of reduction. The loop technique described here is a novel technique that provides a simple yet stable method for reducing and maintaining reduction in cases of AC joint disruption. In addition, it does not involve the use of metal implants, which require later removal. Also, it does not involve expensive synthetic grafts. The incision used in this technique is small (3--4 cm). The duration of postoperative immobilization is relatively short. One of the difficulties of this technique is the very small size of the CC ligament (about 10--13 mm). This makes its handling quite difficult. In most cases, the CC ligament edges are resutured en masse after failure to identify the conoid from the trapezoid. A drawback of this study is that we could not attribute the good results achieved to the healing of the CC ligament or to the suture material used to maintain the reduction, or to both. Only a magnetic resonance imaging (MRI) study could clarify this issue, but this could not be done, as it was extremely difficult to convince a patient after recovery to perform the expensive MRI study. In conclusion, we believe this technique should be used only in acute cases. Our results show a high incidence of good outcome, with loss of reduction in a single case only, during the long follow-up period. We could not prove that the good results are due to CC ligament healing. The patients were able to return to their daily activities and even to contact sports without any noticeable deformity, feeling of weakness, pain, or limitation of range of motion (compared with the contralateral side). The technique does not involve the use of metallic implants, which require a second surgery to remove them, or the use of expensive synthetic graft or a graft harvested from a distant donor site. The authors declare that they have no conflict of interest. Open Access {#d32e1499} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.308980

2011-2-17

PMC3052426

Introduction {#Sec1} ============ Total dislocation and extrusion of the talus is a rare injury produced by an excessive tibiotalar dorsiflexion or plantarflexion in combination with subtalar supination or pronation in high-energy traumas \[[@CR1]\]. Talar dislocation is a well-known lesion, but the majority of these cases consists of open fractures and dislocations in which the talus preserves at least one capsular bond. In these injuries, the talus loses its anatomical relationships with tibia, calcaneus, and navicular bone, thus impeding all weight-bearing movements \[[@CR2]\]. Recently, large case series suggest talar reimplantation after total extrusion with variable preservation of ligaments, showing acceptable infection rates and rare talar body collapses \[[@CR3]\]. When the talus loses all soft-tissue attachments, it is called missing talus. Although several cases of total talar extrusion have been reported in literature, we found only six cases of complete talar extrusion with loss of all ligaments \[[@CR4]--[@CR8]\]. Considering the absence of guidelines, managing this condition is supported by surgical experience and few cases of class V evidence in the literature. On the other hand, the complications of these injuries could be compared with those of complete talar extrusion without loss of talar ligaments and to talar neck fractures. Complications could be divided in two categories: short-term (infection) and long-term complications. Although no data about infection rates in missing talus injuries are available, the likelihood of infection is strictly correlated with open fracture and extrusion. Marsh reports infection in seven patients (38%) in the treatment of 18 open fracture dislocations \[[@CR9]\], whereas considerably lower rates are reported by Smith and colleagues \[[@CR3]\], who report one case of infection out of 27 cases of extruded talus (4%). Long-term complications could be observed somewhere about 12 months after trauma and consist of bone collapse, stiffness, arthritis, and bone necrosis. In consideration of the loss of all ligaments and vascular supply, bone necrosis is the most dreaded complication in missing talus accidents. In this case report, we propose an original approach characterized by the use of antibiotic cement, followed by harvesting of a sural fasciocutaneous flap and reimplantation of the original talus. The sural fasciocutaneous flap is able to bring excellent trophism to the acceptor site and may be successfully employed to treat defects of the proximal third of the foot and the lower leg \[[@CR10]--[@CR13]\]. Case report {#Sec2} =========== A 27-year-old man was brought into the emergency department after a motorbike accident. No wounds were reported other than the total extrusion of the right talus (Fig. [1](#Fig1){ref-type="fig"}). The bone pedicle appeared severely damaged, whereas the talar bone was intact and was stored in a bone bank. Irradiation was employed to preserve and sterilize the bone.Fig. 1X-ray shows the absence of talus without any fracture of the surrounding bones A culture was obtained and empiric antibiotic therapy with teicoplanin and amikacin was started. The lacerated wound was copiously irrigated, cleansed with povidone iodine, and debrided. The talar void was filled with properly modeled gentamicin/clindamycin-loaded cement spacer in order to reduce the risk of articular talar space loss and infection while waiting the cultural results (Figs. [2](#Fig2){ref-type="fig"}, [3](#Fig3){ref-type="fig"}). An external fixator, with pins inserted in the tibia and calcaneus, was applied to stabilize the joint. Vascularization of the foot appeared intact; tendons and muscles did not display damage; no neurological deficit was reported; and the skin defect was closed primarily.Fig. 2Properly modeled antibiotic cement was employed to fill the talar void (*left*). An external fixator was employed to stabilize the joint; the wound was closed primarily (*right*)Fig. 3Intraoperative X-rays after application of antibiotic cement After 2 weeks, the cutaneous margins of the wound and the surrounding skin displayed necrosis. The patient was referred to our department of plastic surgery. The cultures gave negative results. After debridement of the necrotic tissue, the cutaneous defect measured 6 × 5 cm. The antibiotic cement was removed, and two polyvinyl alcohol sponges were placed in the articular void in order to prevent any damage to the articular cartilages. During the same surgical time, a distally based sural fasciocutaneous flap measuring 8 × 6 cm was harvested and applied to the defect. Split-thickness skin grafts were employed to cover the flap pedicle. Twenty-five days later, the external fixator was removed and the original talus was placed, with a dorsal incision 2 cm above the flap margin. Arthrodesis was performed percutaneously using two screws in the anterior subtalar joint and two in the posterior subtalar joint (Fig. [4](#Fig4){ref-type="fig"}).Fig. 4Arthrodesis with two screws in the anterior subtalar joint and in the posterior subtalar joint. Anteroposterior and lateral X-ray projections after 2 years of follow-up. No signs of avascular necrosis are observed Mobilization of the ankle without weight bearing was allowed after removal of the external fixator. Literature is rather unforthcoming in suggesting guidelines due to the rarity of missing talus lesions. We opted to precociously allow careful movements without weight bearing in order to avoid atrophy and to restore talus and surrounding tissues. Weight bearing was then gradually introduced at 3--4 months after the final surgery, and after 6 months, full weight bearing was achieved. At the 4-year follow-up, the joint showed no signs of avascular necrosis, neither plantarflexion nor dorsiflexion showed impairment, and ambulation was regular (Fig. [5](#Fig5){ref-type="fig"}).Fig. 5Final result after 4 years. Movements are not impaired; ambulation is regular Discussion {#Sec3} ========== Talar extrusion is a rare injury, and only few cases have been reported in literature. Management of total talar extrusion could be divided in two types. In the first type, the reconstructive surgeon has to avoid infection and prevent permanent damage to the articulation. Lee \[[@CR2]\] suggests immediate reimplantation of the talus in the event that vascular supply is not permanently damaged. Often, this strategy is recommended owing to the possibility of infection, which can affect 38--85.7% of cases \[[@CR14]\]. In consideration of infective and necrotic risks, which are higher in cases of immediate talus reimplantation, we delayed reimplantation. The talus-shaped antibiotic cement was employed while waiting for cultural results and to reduce the risk of articular talar space loss. In a second phase, reconstruction was performed. Several reconstructive options should be considered, including replacing the talus, tibiocalcaneal arthrodesis, and pseudoarthrodesis. Marsch, in his retrospective study of 27 open extrusions of the talus (without interruption of vascular pedicles), observed that the favorable outcomes could be explained by a protocol including serial irrigation, debridement, and rigid fixation \[[@CR9]\], whereas more recent studies demonstrate that reimplantation has proven to be a safe treatment \[[@CR3]\]. In our case, although the vascular supply was completely interrupted, reimplantation of the talus with arthrodesis of the subtalar joint was performed. This treatment was supported by the use of a sural fasciocutaneous flap which increased the vascular supply to the traumatized area and reduced the otherwise high risk of avascular necrosis of the reimplanted talus. Although the vascular supply was interrupted, no signs of avascular necrosis of the bone were observed in 4 years. Managing such cases remains controversial, and good options are talus replacement, tibiocalcaneal arthrodesis, and pseudoarthrodesis. Notwithstanding, the successful outcome in our case strongly supports our therapeutic choices. The patient gave his informed consent prior being included in this study; the study was authorized by the local ethical committee and performed in accordance with the ethical standards of the 1964 Declaration of Helsinki, as revised in 2000. None. Open Access {#d32e368} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.310678

2011-2-25

PMC3052427

Introduction {#Sec1} ============ Simultaneous palmar dislocation of the scaphoid and lunate is rare \[[@CR1]--[@CR7]\] and has been classified into two types depending on whether or not the scapholunate ligament is intact. Ten patients with dislocation of the scaphoid and lunate as a unit have been described to date, as well as six patients with divergent dislocation \[[@CR1]--[@CR6]\]. The patient described here is therefore the seventh with palmar-divergent dislocation of the scaphoid and lunate. In this patient, the scapholunate and lunotriquetral ligaments were sutured through the dorsal approach, the anterior capsule was sutured through the palmar approach, and the scapholunate and lunotriquetral joints were fixed with Kirschner wires. To our knowledge, this is the first report in which interosseous ligaments were sutured by open surgery for divergent dislocation of the scaphoid and lunate. Case report {#Sec2} =========== A 46-year-old man who fell from a height of 1.5 m onto his left hand was brought to the emergency center of our hospital and underwent a medical examination. Radiography of the wrist revealed palmar-divergent dislocation of the scaphoid and lunate (Fig. [1](#Fig1){ref-type="fig"}) but with no neurovascular disturbance in the hand. Two hours after the injury, we performed closed reduction under local anesthesia. Although closed reduction was successful, severe carpal instability was observed. Seven days after the injury, open surgery was performed through the palmar and dorsal approaches. The dorsal approach showed ruptures of the scapholunate and lunotriquetral ligaments, which were sutured with anchors. The palmar approach showed an oblique tear of the anterior capsule, which was sutured with absorbable threads. Finally, the scapholunate and lunotriquetral joints were fixed with two Kirschner wires, inserted from the scaphoid to the lunate and from the triquetrum to the lunate, respectively, and the wires were buried under the skin (Fig. [2](#Fig2){ref-type="fig"}). A short arm plaster splint was applied postoperatively; 2 weeks later, it was changed to a removable splint and rehabilitation was started. As Kirschner wires remained in the carpal bones, range of motion (ROM) exercises of the wrist were restricted to avoid wire failure. At 7 weeks, the Kirschner wires and splint were removed, and the patient was started on intensive rehabilitation for an additional 3 months. At the 1-year follow-up, the patient had returned to normal life and work and had no pain in his wrist, although wrist motion was still restricted. Measurements of wrist and forearm ROM showed that right/left extension was 60/50°, flexion was 70/40°, supination was 90/80°, and pronation was 90/90°. A hand dynamometer showed that grip strength in his left hand was 16 kg compared with 27 kg on the contralateral (dominant) side. Although we observed no evidence of dorsal or volar intercalated segment instability pattern deformity, radiography showed a break in arc II of Gilula's line between the lunate and triquetrum, as well as flexion deformity of the scaphoid (Fig. [3](#Fig3){ref-type="fig"}) \[[@CR8]\]. Magnetic resonance imaging showed no evidence of avascular necrosis of the scaphoid and lunate (Fig. [4](#Fig4){ref-type="fig"}).Fig. 1Radiographs at initial diagnosis showing palmar-divergent dislocation of the scaphoid and lunateFig. 2Postoperative radiographs showing good alignment of the carpal bones. The scapholunate angle was 54° and the radiolunate angle 6°. Gilula's line was well-regulatedFig. 3Radiographs at the 1-year follow-up. The scapholunate angle was 67° and the radiolunate angle 0°. Dorsal intercalated segment instability (DISI) deformity was not observed, although there was flexion of the scaphoid and a break in arc II of Gilula's line at neutral and ulnar deviationFig. 4Magnetic resonance imaging at the 1-year follow-up showing no evidence of avascular necrosis of the scaphoid and lunate The patient provided written informed consent prior to inclusion in this study, which was authorized by the local ethics committee and performed in accordance with the ethical standards of the 1964 Declaration of Helsinki as revised in 2000. Discussion {#Sec3} ========== As palmar-divergent dislocation of the scaphoid and lunate is rare, its optimal treatment remains unclear. In previous reports, two patients were treated by open reduction and cast immobilization \[[@CR1], [@CR2]\], two by open reduction and percutaneous pinning of the carpal bones and cast immobilization \[[@CR3], [@CR4]\], one by open reduction and suture of the anterior capsule and cast immobilization \[[@CR5]\], and one by proximal row carpectomy (PRC) (Table [1](#Tab1){ref-type="table"}). Carpal instability is severe in divergent dislocation due to ruptures of both the scapholunate and lunotriquetral ligaments. Therefore, it is difficult to stabilize the carpal bones and still retain sufficient wrist motion.Table 1Review of previous patients with divergent dislocation of the scaphoid and lunateAuthorFollow-up (months)Surgical procedureApproachImmobilization (duration)K-wire removalRange of motionComplicationsCampbell \[[@CR2]\]12Only open reductionPalmarCastNR--Ext 1/2Flex 1/3 of healthy sideNoneGordon \[[@CR1]\]12Only open reductionPalmarCast4 weeks--Ext 15°Flex 25°DISIKupfer \[[@CR3]\]42Open reduction K-wire pinning (S-L)Palmar & dorsalCast4 months4 monthsExt 25°Flex 0°CRPS DISI AN (scaphoid, lunate)Baulot \[[@CR5]\]42Open reduction Anterior capsule suturePalmarCast6 weeksAlmost fullDISIKang \[[@CR4]\]18Open reduction K-wire pinning (S-L/S-C)PalmarCast6 weeks6 weeksAlmost fullNoneDomeshek \[[@CR6]\]1Proximal row carpectomyPalmar & dorsalSplint1 month--NRNR*AN* avascular necrosis, *NR* not recorded, *K*-*wire* Kirschner wire, *S* scaphoid, *L* lunate, *T* triquetrum, *C* capitate, *DISI* dorsal intercalated segment instability, *CRPS* complex regional pain syndrome, *Ext* extension, *Flex* flexion Among the methods recommended to repair, the anterior and posterior ligaments on both sides of the lunate are combined palmar and dorsal approaches \[[@CR5]\], and open reduction and percutaneous pinning of the scapholunate and scaphocapitate joints without suture of the interosseous ligaments \[[@CR4]\]. Although we found that suturing of the dorsal scapholunate and lunotriquetral ligaments provided a satisfactory outcome in our patient, wrist stiffness, carpal malalignment due to a break in arc II of Gilula's line between the lunate and triquetrum, and flexion of the scaphoid still remained. Several problems arose during surgery and postoperative management. First, we should have sutured the palmar, not the dorsal, lunotriquetral ligament because the palmar ligament is stronger. This may have prevented the break in Gilula's line. Moreover, in addition to fixing the scapholunate and lunotriquetral joints with Kirschner wires, we should have fixed the scaphocapitate joint to maximize anatomical carpal alignment. Fixation of the scaphocapitate joint may have prevented flexion deformity of the scaphoid. Thus, for reliable carpal stability, we recommend ligament repair and temporary joint fixation of the carpal bones. Subsequent wrist stiffness may be prevented by early removal of Kirschner wires after surgery and starting wrist exercises. Indeed, it may be possible to remove Kirschner wires earlier than 6 weeks when interosseous ligaments are sutured \[[@CR4]\]. The injury to our patient may have been accompanied by avascular necrosis of the scaphoid and lunate \[[@CR3]\]. PRC on a patient with a scapholunate dislocation and complete scaphoid extrusion resulted in a good clinical outcome \[[@CR6]\], suggesting that PRC may eliminate avascular necrosis and avoid additional surgery in patients with this type of injury. However, although PRC has shown satisfactory clinical outcomes, postoperative ROM and grip strength averaged 50--70% and 60--90%, respectively, compared with the healthy side \[[@CR9]\], outcomes similar to those observed in our patient. Therefore, except when unavoidable, we recommend surgical repair, especially for active young people and manual workers, with PRC considered a salvage procedure. The authors declare that they have no conflict of interest. Open Access {#d32e601} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.312096

2011-2-22

PMC3052428

Introduction {#Sec1} ============ Bladder injuries occur in up to 8% of cases with pelvic trauma \[[@CR1]\]. Many clinical indicators are suggestive and should prompt the clinician to undertake further investigations. Mortality rates range from 11% to 44% \[[@CR2]\], but this is commonly attributed to the associated polytrauma \[[@CR3]\]. Conventional retrograde cystography has been considered the gold standard for investigating such patients \[[@CR1], [@CR4]\] and has a reported sensitivity of 100% \[[@CR1], [@CR3]--[@CR6]\]. More recently, computed tomography (CT) cystography has been used as an adjunct to trauma CT to evaluate bladder trauma, showing similar accuracy rates \[[@CR6]\]. This procedure involves retrograde administration of contrast into the bladder through a catheter and performing a further scan in addition to the trauma CT. \[[@CR6]\]. We describe two cases in which conventional retrograde cystography failed to detect a bladder rupture in patients with an associated pelvic fracture. In both cases, informed consent was obtained for inclusion in this report. The purpose of this case report is to highlight such occurrences, the possible reason for failure and the implications occurring as a result. Case reports {#Sec2} ============ Case 1 {#Sec3} ------ A 26-year-old woman sustained multiple injuries, including a pelvic fracture, after falling 20 feet (Fig. [1](#Fig1){ref-type="fig"}). The patient was transferred to a trauma centre for primary treatment. Initial assessment revealed macroscopic haematuria. A CT scan did not demonstrate contrast extravasation from the bladder. Urologists were consulted, and as the degree haematuria was settling, conservative treatment was chosen. Subsequently, the patient was transferred to a regional pelvic trauma unit for further treatment.Fig. 1Initial pelvic injuries seen on an anteroposterior pelvic radiograph for case 1. The pelvic injury comprised right superior and inferior rami fractures with an associated pubic symphysis diastasis and a left sacral fracture At 7 days following injury, a retrograde cystogram was performed prior to definitive pelvic fracture stabilisation, which is routine practice at the hospital. The cystogram was performed using 250 ml of iodinated contrast diluted with 100 ml of normal saline. No leak was identified on full and postdrainage bladder films (Fig. [2](#Fig2){ref-type="fig"}a, b). A Pfannenstiel incision was then used to approach the pelvic fractures. On dividing the rectus muscle, significant amounts of clear fluid were encountered, raising the suspicion of bladder injury. Direct visualisation of the bladder revealed a tear approximating 5 cm and extending from behind the pubic symphysis (Fig. [3](#Fig3){ref-type="fig"}). The edges were rounded and fibrous, and much of the bladder wall was adherent to the surrounding tissues.Fig. 2Retrograde cystography in case 1 demonstrating full (**a**) and postdrainage (**b**) films in assessing bladder injury in case 1. No leak was identified using 350 ml of contrastFig. 3Intraoperative image in case 1 demonstrating the bladder rupture through a Pfannenstiel incision. The rupture was approximately 5 cm long extending from behind the pubic symphysis. Balloon of the urethral catheter is visible inside the bladder. Note the pubic symphysis diastasis plate in situ After copious irrigation, the fractures were stabilised as previously planned, and the bladder was repaired by a urologist. Urethral and suprapubic catheters were left in situ. Follow-up urethrocystograms revealed no remaining leaks, and catheters were sequentially removed. Case 2 {#Sec4} ------ A 36-year-old man was involved in a 45-mph car accident. Multiple injuries were sustained, including a lateral compression-type pelvic injury (Fig. [4](#Fig4){ref-type="fig"}) and a perineal laceration.Fig. 4Initial pelvic injuries seen on an anteroposterior pelvic radiograph in case 2 include pubic symphysis diastasis, right sacroiliac diastasis and right sacral ala fracture After initial trauma surveys, he was catheterised without difficulty and had haematuria. A urology review was undertaken and a decision made to manage conservatively. A contrast-enhanced CT scan did not identify kidney or bladder injury. The patient was transferred to a regional pelvic trauma unit for pelvic stabilisation 10 days after the initial injury. A routine preoperative urethrocystogram revealed a small bladder-neck tear. Urology input was sought to confirm this, and a urethral catheter was left in situ. A laparotomy was performed to create a defunctioning colostomy. On entering the peritoneal cavity, further intraperitoneal and extraperitoneal ruptures were seen in addition to the bladder-neck tear. Both injuries were repaired by urologists and a suprapubic catheter inserted. An external fixator was applied to stabilise the pelvis. A urethrocystogram 4 weeks after bladder repair revealed no further leaks, and catheters were sequentially removed. Discussion {#Sec5} ========== Bladder ruptures can be categorised into extraperitoneal (60%), intraperitoneal (30%) and combined (10--12%) \[[@CR7]\]. Extraperitoneal ruptures are usually associated with pelvic fractures, most commonly occurring along the anterolateral aspect or bladder base. These injuries are due either to a shearing effect with disruption of the pelvis or direct bony penetration \[[@CR3]\]. Intraperitoneal ruptures result from severe blunt lower-abdominal or pelvic trauma to a distended bladder. This leads to sudden intravesical pressure increase and rupture, often creating horizontal tears at the dome being the weakest point. Lateral compression fractures and pubic symphysis diastasis injuries are more commonly associated with bladder injuries \[[@CR1]\]. Despite occurring relatively commonly, bladder injuries are still missed. In a recent study, 1% of cases were missed due to misinterpretation of cystogram images \[[@CR8]\]. The clinical indicators suggestive of bladder trauma include suprapubic pain or tenderness, inability to void and low urine output. Unfortunately, in the presence of a pelvic fracture, these signs are almost always present. The investigation of choice in our institution for suspected bladder injuries is retrograde cystography. Increasingly, retrograde CT cystography is being used with similar accuracy rates \[[@CR1], [@CR3]--[@CR6]\]. To prevent false negatives, the bladder should be adequately distended with at least 250--300ml of fluid; radiographs should be taken in at least two projections and include a postdrainage film to visualise extravasated contrast \[[@CR9]\]. No postdrainage films are needed for CT cystography. In case 2 in this report, a correctly identified simple bladder-neck tear could have been treated with a urethral catheter alone. However, the cystogram failed to detect additional intraperitoneal and extraperitoneal ruptures, which represent more significant injuries requiring formal repair and a suprapubic catheter. One of the possible reasons the cystogram failed to detect the full extent of the leaks is the time span from injury to investigation---7 days in case 1 and 11 in case 2. The average time to surgery for pelvic/acetabular fractures in our unit is 10 days, comparing favourably with other units around the country. It is possible that the bladder injury had become walled off with fibrous adhesions to the overlying rectus and was disturbed by the surgical approach. Although a trauma CT scan failed to detect a bladder rupture at day 1, no formal cystogram was performed in either case, and perhaps that could have identified the injury sooner. Failure to diagnose a bladder rupture prior to pelvic stabilisation surgery has a number of implications. Not all hospitals in which pelvic surgery is undertaken have a urologist available at all times to assist with bladder repair. From a surgical planning perspective, ideally, the patient should be made aware of the possibility of a suprapubic catheter prior to surgery and the possible associated increased risk of metal-work infection. A medicolegal implication also exists in that a negative preoperative retrograde cystogram does not absolutely imply that a rupture found intraoperatively, or even postoperatively, was caused by the surgeon during fracture repair. Patients and surgeons should be aware that even contrast imaging of the bladder may not detect all injuries. Conflict of interest {#d32e359} ==================== None. Open Access {#d32e364} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.313062

2010-12-8

PMC3052429

Introduction {#Sec1} ============ A lot has been written about minimally invasive stabilization of sacral fractures with percutaneous ilio-sacral screws \[[@CR1]--[@CR5]\] or posterior sacral plates \[[@CR6]\], but few articles give sufficient details on the reduction methods. The achievement of satisfactory reduction is the first hot point in the treatment of pelvic ring disruption \[[@CR7]\]. Despite its importance, this step may be a real challenge even for the expert pelvic surgeon, because of several issues including the frequent finding of multi-planar and rotational displacement components even in so-called vertically displaced fracture. The classic traction method is often insufficient to deal with these difficulties, because of its exclusively axial effect, the large force required and the need to fix the intact hemi-pelvis in a strong and safe manner. Only some of these problems seem to be solved by use of special pelvic frames \[[@CR8], [@CR9]\], and moreover, radiolucent traction tables are expensive devices, available only in few hospitals. Otherwise, open reduction techniques can provide good results, but are expensive for these patients in terms of blood loss \[[@CR7], [@CR10]\]. Furthermore, surgical access has to be achieved via a skin area often damaged by trauma and in zones that are subject to bedsores. The aim of this study is radiological and functional evaluation of a minimally invasive reduction technique for treatment of sacral fracture, which could be effective, tissue sparing and economic in terms of equipment needed. Materials and methods {#Sec2} ===================== From November 2002 to March 2009, 82 patients suffering from sacral fracture were surgically treated at our institution. Among these, 51 patients presented 61-C1.3 fracture according to the Orthopaedic Trauma Association (OTA) \[[@CR11]\]. The method of reduction was predominantly closed with traction (28 cases), then open with posterior surgical access. Starting from April 2007, we began to perform the minimally invasive technique described herein. For this retrospective study, inclusion criteria were: 61-C1.3 fracture pattern according to the OTA \[[@CR11]\] (type C pelvic ring disruption with monolateral sacral fracture), availability of complete clinical and radiological documentation and a minimum 12-month follow-up time. Exclusion criteria were presence of cognitive deficits, major head trauma, neurologic deficits related to extrapelvic lesions, major injuries to the upper and lower limbs, open fractures and pathological fractures. Moreover, patients with significantly impaired mobility or pain during gait already present before the trauma were excluded. We finally included in this study 11 patients with average age of 40.2 years (range 24--59 years), who were referred to our institution from April 2007 to March 2008 and surgically treated by the same surgeons (A.M., A.N.) using the technique indicated in Table [1](#Tab1){ref-type="table"}.Table 1Patients' data, procedures and outcomesCaseType of\ traumaAge (years)Time from trauma to surgery (days)DenisFollow-up time (months)Majeed scoreDisplacement (mm)Posterior\ fixationAnterior\ fixationAssociated lesionsPre-surgicalPost-surgicalFollow-up1Industrial accident247118462166Two IS screws, tension bandEx-fixUrethral tear2Industrial accident591011988734One IS screwFour-hole plateL4-L5 fracture(internal fixation)3Sports injury357117991444Two IS screwsEx-fix4Car accident571312581844Two IS screwsSix-hole plate, ex-fix5Sports injury37221598965One IS screwNoneTwo-column acetabular fracture6Car accident28222281855One IS screwEx-fix7Motorbike accident39622482734One IS screwFour-hole plateBladder tear8Motorbike accident43221871866One IS screwFour-hole plate, ex-fix9Industrial accident334214941278One IS screwEx-fixSplenic rupture10Sports injury481021884844One IS screwFour-hole plateL1 fracture (conservative)11Sports injury371631866171213Two IS screwsFour-hole plateTibial plateau fracture, tibial shaft fractureMean40.07.18--18.9180.9110.825.455.73------SD11.024.68--3.4515.164.662.542.72------*SD* standard deviation, *IS* ilio-sacral The trauma was caused by road accident in 36% of cases, by sports injury in 36% and by industrial accident in 28%. Seven patients had associated injuries, six of which required surgery: two urological lesions, one spleen rupture, one two-column acetabular fracture, one L4-L5 vertebral fracture and one bilateral lower limb fracture. Every patient was submitted to accurate clinical examination and pre-operative imaging planning, including at least antero-posterior (AP) radiogram of the pelvis, inlet and outlet views (Fig. [1](#Fig1){ref-type="fig"}) and a computed tomography scan. Every fracture was classified according to Denis classification \[[@CR12]\]; the most common were type II (54%) and type I (36%). Pre-operative displacements were measured to the nearest millimetre as the maximum point-to-point distance between the fragments of the sacral fracture on each of the three views of the pelvis; all displacements were recorded.Fig. 1Illustrative case: pre-operative radiograms (antero-posterior, outlet, inlet) showing a multi-planar displacement with maximum value of 21 mm The patients were brought to the operating room as soon as permitted by general health conditions, since it is well known that reduction becomes progressively more difficult with time; mean trauma to surgery time was 7.18 days \[range 2--16 days, standard deviation (SD) 4.68 days\]. Each fracture was reduced by the minimally invasive approach described below. Fixation of the posterior pelvic ring was achieved by one cannulated ilio-sacral screw in 64%, by two ilio-sacral screws in 28% and by two ilio-sacral screws plus tension-band plating in one case; all screws were placed in the body of the first sacral segment. Fixation of the anterior pelvic ring was achieved with a symphyseal plate in 36% and with an external anterior fixator in 36%; in two cases they were used together, while in one case no anterior fixation was performed. After surgery, AP, inlet and outlet view radiograms of the pelvis were taken for every patient and post-operative displacements were measured on all three views (Fig. [2](#Fig2){ref-type="fig"}); we considered the highest value as an index of quality of reduction according to Matta's criteria \[[@CR7]\].Fig. 2Illustrative case: post-operative radiograms. Fixation is achieved by two ilio-sacral screws, a posterior tension-band plate and anterior external fixation. Maximum residual displacement was 6 mm Patients were not allowed weight-bearing for 60 days. The external fixator, when used, was removed 2 months after its placement. They were then directed towards walking rehabilitation, with complete weight-bearing allowed 90 days after injury. Clinical and radiographic examination was carried out at least at 1, 2, 3, 6 and 12 months after surgery. Finally, all patients were evaluated functionally at least 1 year after surgery using Majeed's grading scale for pelvic fracture \[[@CR13]\], with mean follow-up time of 18.9 months (range 14--25 months, SD 3.45 months). The study conforms to the 1964 Declaration of Helsinki as revised in 2000 and was approved by our institutional ethical committee. All enrolled patients provided informed consent. Surgical technique {#Sec3} ------------------ The patient is under general anaesthesia, in prone position, on a standard radiolucent orthopaedic table. The patient position is exactly symmetrical, with forward tilting of the pelvis achieved by insertion of a thoraco-pelvic support, the knees being flexed at about 30° to release the sciatic nerves and sacral roots. The C-arm fluoroscope is placed on the uninjured side of the patient, and adequate image rendering is verified before starting the operation. The patient's body has to be placed as caudal as possible, to avoid impingement of the C-arm and table's pedestal; this can be achieved by assembling two standard leg attachments. The posterior pelvis is then prepped and draped in usual fashion. The posterior superior iliac spines (PSIS) are individuated bilaterally by palpation, and two incisions (about 1.2 inches each) are performed just lateral to them (Fig. [3](#Fig3){ref-type="fig"}); then a 3.2-mm drill is used to insert a cortical 4.5 screw in each side (Fig. [4](#Fig4){ref-type="fig"}). The drill is started on the posterior superior iliac spine, angling lateral approximately 40° in relation to the sagittal plane and slightly cranially to achieve placement in the direction of the iliac wing (Fig. [5](#Fig5){ref-type="fig"}a).Fig. 3Bony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the *middle*, the sacral spinous lineFig. 4Surgical access to the PSIS (*right side*) and 3.2-mm drilling to insert the 4.5-mm screw (*left side*)Fig. 5**a** Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. **b** Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction Next, a large Jungbluth reduction clamp (Matta Pelvic System--Stryker Trauma AG, Bohnackerweg 1, CH-2545 Selzach, Switzerland) is connected to these screws (Fig. [5](#Fig5){ref-type="fig"}b). If an important rotational dislocation is detected, a standard 6.0-mm Shanz screw placed proximal to the 4.5 screw in the iliac wing can be used as a joystick. When the fracture displacement is satisfactory reduced, the clamp is tightened to maintain the reduction (Fig. [6](#Fig6){ref-type="fig"}). The reduction is checked by fluoroscopy in AP, inlet and outlet views on the posterior ring, and when obtained, it is common to see also the anterior part of the ring in a reduced position.Fig. 6When the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views Fixation is usually achieved with ilio-sacral stainless-steel 6.5- or 8.0-mm cannulated screws (Asnis III; Stryker Orthopaedics, 325 Corporate Drive, Mahwah, NJ 07430, USA) under fluoroscopic control \[[@CR1]--[@CR5]\]. Most times, two screws are placed in the first sacral segment, even if it is possible to place one screw in S1 and one in S2, or a single screw may be used if the area of safe placement is limited \[[@CR14]\]. At the end, screw position has to be confirmed on antero-posterior, inlet and outlet views \[[@CR15]--[@CR19]\]. In very unstable disruptions, and whenever possible and safe, use of a trans-sacral screw is suggested, as this type of fixation seems to provide greater stability \[[@CR20]\]. After removal of the Jungbluth clamp, it is possible to implant a slide-insertion posterior plate through the same incisions \[[@CR21], [@CR22]\] to improve fracture stabilization. This is particularly suggested in case of large displacement, long trauma to intervention time (15 days or more) or poor bone quality. After fixation of the posterior pelvic ring and closure of wounds, the patient is placed in supine position for anterior pelvic ring fixation, if necessary, by internal or external devices \[[@CR23]--[@CR25]\]. Results {#Sec4} ======= Pre-operative displacements averaged 10.8 mm (range 7--21 mm, SD 4.66 mm) (Table [1](#Tab1){ref-type="table"}). The largest displacement was seen on the 40° caudal view in 63% of cases. Post-operative displacements averaged 5.4 mm (range 3--12 mm, SD 2.54 mm). Using the grading criteria described by Matta \[[@CR7]\], there were five excellent (45.5%), five good (45.5%), one fair and no poor reductions. All patients healed, with average displacement at 1-year follow-up of 5.73 mm (range 4--13 mm, SD 2.72 mm). The improvement obtained with surgery was strongly significant (paired-sample *t*-test: *P* \< 0.0009), while the difference between post-operative and follow-up displacement was nonsignificant (paired-sample *t*-test: *P* = 0.192). There were no operative complications; regarding nonsurgical peri-operative complications, there was one urinary tract infection. In one case, aseptic loosening of the symphyseal plate was detected at 2-month follow-up; the hardware was removed 10 days later, without need for further fixation. In two cases, ilio-sacral screw removal, due to screw-related pain, was performed 9 and 11 months after surgery. In one further case, treated with two ilio-sacral screws and a sacral plate, the hardware was removed 9 months after the accident (Fig. [7](#Fig7){ref-type="fig"}) because it generated discomfort while the patient was using sports equipment (scuba diving gear with air cylinder). We do not consider this to be a device-related complication but rather a sign of surgical success.Fig. 7Illustrative case: after removal of external and internal devices, performed about 9 months after the accident, there was no sign of fracture displacement On functional evaluation, performed using the Majeed score system, 82% of patients obtained good or excellent results, with only one fair and one poor result; the average score was 80.91 points (range 46--99 points, SD 15.16 points). All but one patients returned to work, but the majority of them complained of reduced performance. A 23-year-old male affected by urethral disruption, treated with anterograde and retrograde endoscopic repair by a specialized urologic team, complained of reduced sexual function; at 12-month follow-up he reported fair improvement but not complete remission yet. In this case it is difficult to state whether this reflects a real neurologic issue, and whether it could be related to damage to periurethral neural fibres or sacral roots. We registered slight erectile dysfunction in two further males without urologic tears, who had almost complete recovery at 12-month follow-up. There were no further perineal deficits. There were no major lower limb neurologic impairments, but in two cases we observed transitory lateral femoral cutaneous nerve hypoestesy, probably related to anterior external fixation; in one case recovery was complete within 6 months, in another case a partial sensory deficit remained at 12-month follow-up. Discussion {#Sec5} ========== Reduction of vertically displaced sacral fracture can be a difficult challenge for the pelvic surgeon. Open techniques can obtain good reduction but are costly for the patient in terms of blood loss and soft tissue damage. Tornetta and Matta \[[@CR7]\] stated that 10 mm is an acceptable reduction for injury to the posterior pelvic ring, as they suggested that greater anatomic reduction of posterior injuries did not result in less posterior pain; those authors, performing open reduction and internal fixation of the injured pelvis, reported excellent or good results in 95% of 107 patients with 38 Bucholz type II (type B) injuries and 69 Bucholz type III (type C) injuries, with 1 case of loss of reduction and 3 cases of deep infection. Van den Bosch et al. \[[@CR26]\] evaluated 37 patients (16 type B, 21 type C) treated with internal fixation, obtaining a mean score of 78.6 of 100 on Majeed functional evaluation. Lindahl and Hirvensalo \[[@CR27]\] obtained excellent-good radiographic results in 90% of 101 patients treated with open reduction; the overall functional results, measured with a modified version \[[@CR28]\] of the original Majeed scoring system \[[@CR13]\], were good or excellent in 83% of patients. Percutaneous technique is becoming increasingly popular because it can reduce wound-related problems and blood loss \[[@CR1], [@CR2], [@CR4], [@CR15]\]. On the other hand, closed reduction using vertical traction can sometimes be insufficient, even with the adjunct of dedicated pelvic frames \[[@CR8], [@CR9]\]; moreover, it needs a radiolucent traction table. Routt and Simonian \[[@CR4]\] defined reductions that showed less than 1 cm residual displacement in any plane as acceptable. They obtained only 3 malreductions among 60 sacral fractures treated with closed reduction by manipulation and traction method, but reported 5 cases of failure of fixation and 2 cases of non-union. Schweitzer et al. \[[@CR2]\], revising 71 pelvic ring fractures (10 type B, 61 type C) treated with closed reduction and percutaneous screw fixation, obtained 69 satisfactory reductions and 62 good or excellent functional results according to the Majeed scoring system \[[@CR13]\]. Nevertheless, they reported a 9% rate of surgical-related complications, a rate similar to previous reports regarding this technique \[[@CR5]\]. Minimally invasive transiliac plate osteosynthesis \[[@CR21], [@CR22]\] has been recently described, and its early results appear encouraging. However, this technique has usually been associated with the closed reduction by traction method. The method described herein can obtain good fracture reduction with a limited surgical approach and with a standard radiolucent table; it is particularly indicated in simple, monolateral vertical sacral fractures (61-C1.3). Otherwise, we suggest reduction by traditional techniques for bilateral or complex sacral fractures, because of the extreme posterior instability, as well as for those disruptions which require direct vision of the fracture site. The position of the patient allows traditional open exposure if minimally invasive reduction fails and allows internal fixation of the posterior ring and optimal placement of percutaneous ilio-sacral screws. Using this technique we achieved good to excellent reductions in 91% of patients, coherent with most of the studies regarding sacral fractures which can be found in literature \[[@CR2], [@CR7], [@CR27]\]. In only one patient was a fair result (\>10 mm) obtained; he presented wide pre-operative displacement, and because of the long trauma to surgery time, interposition of soft tissue and fibrous callus formation prevented a better result; nevertheless, we decided not to perform open reduction because the skin and subcutaneous tissues were compromised. Unfortunately, he reported one of the lower functional scores at 1-year follow-up. We suggest that open reduction should be preferred, whenever possible, where the trauma to surgery time is extended beyond 15 days. The main limitations of this study are the small sample size and the short follow-up; furthermore, its strength is blunted by all the implications of its retrospective design and the lack of a control group. In conclusion, we found a minimally invasive reduction technique to be a satisfactory procedure for management of vertically displaced sacral fracture. Axial traction remains a good method, and more cases need to be operated using this technique to confirm its effectiveness in terms of reduction and determine possible complications. The authors declare that they have no conflicts of interest related to the publication of this manuscript. Open Access {#d32e1182} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.314046

2011-2-24

PMC3052430

Introduction {#Sec1} ============ Femoral-shaft fractures are among the most common fractures of the lower extremity in children, with an annual incidence of up to 1 per 5,000 \[[@CR1], [@CR2]\]. There are several different options for treating femoral-shaft fractures in children, including skeletal or skin traction, early or immediate application of a hip spica cast, pontoon spica, closed reduction and minimally invasive plate osteosynthesis, external fixation, plate fixation, and internal fixation with the insertion of intramedullary nails \[[@CR3], [@CR4]\]. Selecting the management strategy is dependent on factors such as the presence of other associated injuries or multiple trauma, fracture properties, age, and socioeconomic factors. Because of its clinical effectiveness and low rate of complications, elastic stable intramedullary nailing for fractures of long bones in the skeletally immature patient (e.g., children) has gained widespread popularity. Titanium elastic nailing (TEN) is commonly used to stabilize femoral fractures in school-aged children, but there have been few controlled studies and with only relatively short-term follow-ups assessing the risks and benefits of this procedure compared with those of the traditional traction and application of a spica cast. The results of previous prospective and retrospective studies comparing TEN with traction and a spica cast were mostly in favor of TEN, considering recovery time, complication rate, and in some cases hospital charges \[[@CR2], [@CR5], [@CR6]\]. According to the lack of data in this regard, we designed a prospective randomized controlled study to compare TEN with traction and a spica cast in treating femoral fractures in children in terms of recovery and complications. Materials and methods {#Sec2} ===================== This randomized controlled trial was conducted from February 2009 to January 2010 in the Department of Orthopedic Surgery of two university hospitals in Isfahan, Iran. Children 6--12 years of age with simple femoral-shaft fractures participated in the study consecutively. Exclusion criteria were segmental Winquist types III and IV comminuted fractures, previously diagnosed neuromuscular disease (e.g., cerebral palsy), metabolic bone disorders (e.g., osteomalacia), or pathological fractures. Considering α = 0.05, study power = 80%, and d = 5 days as the minimal expected difference between the two groups, a sample size of 22 patients was considered for each group. Parents of all children gave informed consent prior to the study, which was authorized by the local Scientific Ethical Committee of Isfahan University of Medical Sciences, Isfahan, Iran, and was performed in accordance with the ethical standards of the 1964 Declaration of Helsinki as revised in 2000. Also, the study has been registered at <http://www.clinicaltrials.gov> (NCT01190696). Using random allocation software, patients were divided into two groups of TEN and spica cast and were treated by a single orthopedic surgeon. Hip-supported long-limb casting splints without skeletal traction was applied for all patients at admission for controlling pain and preventing deformities. For patients in the TEN group, the standard TEN technique was applied according to the method described by Flynn and colleagues \[[@CR6]\]. Operation was done under general anesthesia on a fracture table. After a linear incision, opening the fascia, and passing the muscle fibers, a hole was opened in the bone and enlarged. Then, each titanium elastic nail was retrogradely placed through the distal part of the femur. Each nail was 40% of the canal diameter at the narrowest site of the femoral shaft. Reduction and fixation was done under C-arm image intensifier. All patients received first-generation cephalosporin prophylaxis, which was initiated 12 h preoperatively and continued 24--48 h postoperatively \[[@CR6]\]. Patients in the spica cast group were treated with skeletal traction for about 3 weeks and then with a spica cast. The traction pin was inserted in the distal part of the femur in the operating room and under general anesthesia. Control radiography was carried out after the traction and later at 1-week intervals. The pin was removed after sufficient callus consolidation had been achieved, and a one-and-a-half hip spica was applied (with the hips at 20--30°of flexion and the limb in 10--15° external rotation) in the operating room under general anesthesia. The cast was maintained for about 1 month; After cast removal, patients were referred for physical therapy for initial gait training and additional physical therapy if a satisfactory range of motion was not achieved. The length of hospital stay was recorded, and follow-up visits were made at 2, 4, 12, and 24 weeks after discharge. Limb alignment and rotation, range of knee motion, and incision and skin infections were assessed at each visit. Recovery milestones were time to start walking with aids, time to start independent walking, time absent from school, and parental satisfaction, which ranged from weak = 0 to excellent = 4. Major complications were defined as those leading to unscheduled operative treatment, malunion, or nonunion. Nonunion was defined as the absence of osseous union \>6 months after the injury. Data were analyzed using SPSS software (windows version 16.0) by independent sample*t* test and chi-square test for comparing means and categorical data, respectively, between groups. Results {#Sec3} ======= During the study period, 55 children were presented with femoral-shaft fractures to the centers. From among these patients, 51 met the inclusion criteria (four patients had open fractures). Five patients did not agree to participate in the study protocol, so 46 children with simple closed femoral fractures (23 in each group) entered and completed the study. There was no significant difference between the two groups in terms of age and gender (*P* \> 0.05). Compared with children treated with spica cast, those treated with TEN had shorter hospital stay (*P* \< 0.001) and took a shorter time to start walking with support or independently (*P* \< 0.001), returned to school sooner (*P* \< 0.001), and had better parent satisfaction (*P* = 0.003). The range of knee motion was 138.7 ± 3.4° in the spica cast group and 133.5 ± 13.4° in the TEN group, with no significant difference (*P* = 0.078) (Table [1](#Tab1){ref-type="table"}).Table 1Comparison of outcomes between groupsTEN *n* = 23Spica cast *n* = 23*P* valueAge7.1 ± 1.86.5 ± 1.50.225\*Male/female15 (65.2%)/8 (34.7%)16 (69.5%)/7 (30.4%)0.500\*\*Hospital stay (days)6.9 ± 2.920.5 ± 5.8\<0.001\*Time to start walking with aids (days)17.6 ± 10.265.6 ± 10.7\<0.001\*Time to start walking independently (days)35.2 ± 13.280.0 ± 10.1\<0.001\*Time to return to school (days)31.5 ± 13.464.3 ± 19.6\<0.001\*Parent satisfaction Excellent12 (52.1%)2 (8.6%)0.003\*\* Good11 (47.8%)15 (65.2%) Moderate02 (8.6%) Weak04 (17.3%)Knee range of motion (degree)133.5 ± 13.4138.7 ± 3.40.078\*Malunion03 (13.0%)0.117\*\*Infection3 (13.0%)00.117\*\*Data are presented as mean ± standard deviation or number (%)\* Independent sample *t* -test\*\* Chi-square test Three patients (13.0%) in the spica cast group had malunion, whereas none occurred in the TEN group (*P* = 0.117). Three patients had postoperative infection (13.0%), all in the TEN group, but none was obsered in the spica cast group (*P* = 0.117) (Table [1](#Tab1){ref-type="table"}). There was also one transitional proneal nerve injury after TEN that repaired spontaneously. No arterial injury occurred in any patients during the procedures. Discussion {#Sec4} ========== Although spica casting with skeletal traction is traditionally used for femoral-shaft fractures in children, recent studies have shown its possible effects on social, economic, educational, and emotional costs. In contrast, elastic intramedullary nailing of femoral-shaft fractures has gained extensive popularity because of its better clinical and psycho-socioeconomic outcomes with lower risk of complications \[[@CR5]--[@CR7]\]. In our study, we showed the benefits of the TEN surgical method versus traction and spica casting with respect to hospital stay, time to start walking with support or independently, returning to school, and parent satisfaction. Our findings were in agreement with the results of many studies that showed the efficacy and benefits of elastic nails for treating femoral-shaft fractures. Ligier et al. \[[@CR8]\] used elastic intramedullary nail (anterograde or retrograde) with Kirschner wires or pins. They reported more desirable outcomes in \>120 femoral-shaft fractures treated with TEN. In Reeve et al.'s study \[[@CR9]\], 41 patients with femoral fractures were treated with traction and casting, and 49 cases underwent intramedullary nailing surgery. They showed complications were higher in the traction and casting group in comparison with the group undergoing surgery. In our study, the duration of hospital stay was significantly longer in the traction and spica cast group than in the TEN group. This is in conformity with other studies \[[@CR6], [@CR9]--[@CR11]\], which reported shorter hospital stays with TEN, but is in contrast to Saseendar's study \[[@CR12]\]. This difference was due to the fact in Saseendar's study, patients in the TEN group were discharged only after suture removal to have a closer follow-up for the presence of early postoperative complications (if any), and the spica patients were usually discharged a day or two following spica casting after assessing for the presence of plaster-of-Paris-related complications. Our findings showed shorter time to start walking with support or independently and sooner return to school in the TEN group compared with the spica casting group. It is probably because of better contact of the fracture surfaces and anatomical reduction in patients who underwent TEN surgery. Such earlier recovery milestones have also been shown by Greisberg et al. \[[@CR10]\] and Flynn et al. \[[@CR6]\]. In our study, a higher rate of malunion was observed in the traction and spica group compared with the TEN group. This finding conforms to the results of a similar study conducted by Kirby et al., which compared traction and cast with intramedullary nailing and reported malunion only in the traction and casting group \[[@CR13]\]. In other studies, the rate of malunion in the traction and cast group was higher than that in the TEN group \[[@CR11], [@CR14]\]. Our study had certain limitations. Treatment cost, limb length, and angulation degree were not measured in either group. As with any other new procedure, we had a small sample size, and thus the results could show falsely high complication rates. Conclusion {#Sec5} ========== The results showed significant benefits for TEN compared with traction and hip spica cast in treating femoral-shaft fractures in children. Complication rates associated with hip spica cast was also higher than that associated with TEN. Further trials with longer follow-ups and comparison of TEN with other methods, such as external fixation, in children's femoral-shaft fractures are warranted. This study was supported by Isfahan University of Medical Sciences. The authors are thankful to Ali Gholamrezaei, who helped in statistical analyses and editing the manuscript. Conflict of interest {#d32e633} ==================== None. Open Access {#d32e638} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.315953

2011-2-22

PMC3052431

Introduction {#Sec1} ============ Subtalar dislocation (StD) is an uncommon type of injury that involves concomitant loss of normal anatomical relations between talus, navicular, and calcaneus, while the tibiotalar and calcaneocuboid joints remain congruent \[[@CR1], [@CR2]\]. The mechanism of StD is trauma to a plantar-flexed foot either in inversion, resulting in medial subtalar joint dislocation (85%), or in eversion, resulting in lateral dislocation (15%). Anterior and posterior dislocations have also been described but are exceedingly rare \[[@CR3], [@CR4]\]. Immediate reduction is of paramount importance and is usually provided under sedation. It is followed by a period of immobilization necessary for the healing of the soft tissues. The majority of studies specify this period of immobilization between 4 and 8 weeks \[[@CR5]--[@CR10]\]. However, it has also been stated that subtalar joint stiffness in uncomplicated StD can be minimized by avoiding immobilization longer than 4 weeks \[[@CR7]--[@CR10]\]. In general, the duration of postreduction immobilization, which correlates with the amount of stiffness and remaining functionality, is a controversial subject. We conducted this study to reassess the optimal duration of the immobilization period, after uncomplicated medial StD, able to provide both subtalar joint stability and avoidance of stiffness. Our working hypothesis was that a period of 2--3 weeks of immobilization in a cast, followed by range-of-motion (ROM) exercises and partial weight bearing (PWB), could provide better functional results than those achieved by longer periods of immobilization. To increase the validity of our outcome we chose to use a prespecified treatment and rehabilitation protocol, creating a prospective study. To the best of our knowledge, previous researchers have only retrospectively examined this type of injury. Materials and methods {#Sec2} ===================== This prospective study concerned a period from June 2004 through March 2008. The research was approved by the local Ethics Committee, and all patients signed informed consent. The study was performed in accordance with the ethical standards of the 1964 Declaration of Helsinki as revised in 2000. Men or women of any age admitted for StD were considered for inclusion in the study. The inclusion criteria included: (1) medial subtalar dislocation, and (2) open or closed injuries. Patients were excluded if they had: (1) peritalar fracture accompanying the dislocation, (2) subtalar dislocation of other type (lateral, anterior, posterior), or (3) other comorbidities that would influence or delay the rehabilitation protocol (e.g., brain damage, bilateral lower limbs injuries). Of the 14 patients examined in the Accidents and Emergency (A & E) department, 8 fulfilled the inclusion criteria and were enrolled in the study (Fig. [1](#Fig1){ref-type="fig"}). Of those eight patients, six were male and two were female (male-to-female ratio of 3:1). The mean age of the patients was 37.25 years (range 25--54 years). Five of the patients (62.5%) had undergone a motor vehicle accident (MVA), two patients (20%) had sustained a fall from height, and another one (12.5%) an inversion injury when his foot was trapped in a gap. All injuries were neurovascular intact, including seven closed and one open medial subtalar dislocation.Fig. 1Flow diagram of patients enrolled in the study Treatment was provided by immediate closed reduction in the A & E department under sedation. Axial traction on the foot and heel in the line of deformity was combined with countertraction with the knee in flexion to relax the gastrocnemius muscle. Abduction of the foot and dorsiflexion of the ankle followed. After reduction, the ankle and foot were immobilized in a below-the-knee backslab followed by administration of analgesia and foot elevation. The backslab was transformed to a below-the-knee jigsaw cast 3--4 days post reduction as long as the swelling had subsided and non-weight-bearing (NWB) mobilization was allowed. Active ankle and foot ROM exercises were initiated by the beginning of the third week from reduction. Partial weight-bearing began after the third week, progressing to full weight-bearing (FWB) mobilization by the fifth week. Weight-bearing mobilization was assisted by the use of a below-the-knee functional brace that allowed plantarflexion and dorsiflexion but restricted inversion and eversion movements. Muscle-strengthening physiotherapy was implemented with the beginning of ROM exercises. The time intervals between injury, reduction, and initiation of the several mobilization stages were recorded. All patients followed the same early mobilization protocol. Clinical results were evaluated using the American Orthopaedic Foot and Ankle Society (AOFAS) Ankle--Hindfoot scale \[[@CR11]\], which assigns 50 points to function, 40 to pain, and 10 to alignment of the foot. Moreover, a relative ankle ROM score was created for every patient to assess the post injury and immobilization remaining stiffness. This score consisted of a percentage resulting from the ratio of the ankle ROM on the injured leg in comparison with the contralateral healthy limb. The clinical measurements were accomplished by use of a goniometer. Since we were unable to obtain a perfectly accurate measurement, we preferred to create approximate ratios using steps of 5% (Table [1](#Tab1){ref-type="table"}).Table 1Demographics and analytical results of the eight cases presentedCase no.Age/genderMechanism of injuryOpen/closedTime from injury to reduction (min)Time from injury to ROM exercises (days)Time from injury to PWB mobilization (days)Time from injury to FWB mobilization (days)AOFAS score (function)AOFAS score (pain)AOFAS score (alignment)AOFAS score (total)Relative ankle ROM\*Follow-up (months)1M/25MVAClosed9015233544/5040/408/109295%472M/38Inversion injuryClosed12016213444/5040/4010/109490%493M/42Fall from heightClosed15018223541/5040/408/108990%364F/33MVAOpen12017233744/5040/4010/1094100%375M/54MVAClosed15018243844/5030/4010/108490%356M/49MVAClosed9015223647/5040/4010/109795%347F/28Fall from heightClosed12016243644/5040/4010/109495%268M/29MVAClosed18018264244/5030/408/108285%24Mean SD37.25 ± 10.44127.5 ± 31.0516.625 ± 1.323.125 ± 1.5536.625 ± 2.544 ± 1.637.5 ± 4.629.25 ± 1.0390.75 ± 5.3192.5% ± 4.62%36 ± 8.78\* Percentage of ROM of the injured ankle in relation to the contralateral healthy ankle Results {#Sec3} ======= Successful closed immediate reduction under sedation was achieved for all cases (Figs. [2](#Fig2){ref-type="fig"}, [3](#Fig3){ref-type="fig"}). The mean time between injury and reduction was recorded to be 127.5 (range 90--180) min. In all cases the hindfoot was found to be well aligned with no signs of secondary instability under stress and a below-the-knee backslab was applied. No neurovascular damage was recorded pre or post reduction. For the case of the open dislocation, adequate washout was performed prior to reduction. The wound was left open under chemoprophylaxis and was secondary closed 3 days later during the jigsaw cast application. There was no need for a skin graft, since the wound was successfully sutured after debridement. Wound check was accomplished by periodical removal of the jigsaw cast.Fig. 2Pre- and postreduction anteroposterior and lateral radiographs of cases 2, 6, and 8Fig. 3**a**, **b** Pre- and postreduction clinical views of the open medial subtalar dislocation of case 4. **c**--**f** Pre- and postreduction anteroposterior and lateral radiographs of the open medial subtalar dislocation of case 4. **g**, **h** Postreduction computed tomography with three-dimensional (3D) reconstruction views for the detection of any occult fracture Follow-up assessment was conducted by clinical and radiographic examination every 2 months for the first 6 months post patient discharge. Thereafter, follow-up continued with yearly routine checks for a mean period of 36 months (range 24--49 months). The mean percentage of ankle ROM between the injured and the healthy lower limb (Fig. [4](#Fig4){ref-type="fig"} a--d) was 92.5% (range 85--100%), which was considered as very satisfactory by both physicians and patients (Table [1](#Tab1){ref-type="table"}). No radiographic evidence of arthritis or avascular necrosis of the talus was detected in any patient until the final follow-up appointment (Fig. [4](#Fig4){ref-type="fig"}e, f). The mean AOFAS score was 90.75 points (range 82--97) (Table [1](#Tab1){ref-type="table"}). Two out of eight patients complained of transient mild pain which did not restrict them from their daily activities. One female patient correlated this pain with protracted walking on flat shoes, whereas heeled shoes did not seem to annoy her. None of the patients was keen on sports, preventing evaluation of ankle and foot functionality under conditions of repetitive stress. All patients returned to their previous professional occupation and were happy with the outcome.Fig. 4**a**, **b** Range of motion of the ankle joint 2 months post injury. **c**, **d** Range of motion of the ankle joint 3 years post injury. **e**, **f** Radiographic control 3 years post injury without evidence of arthritis or avascular necrosis Discussion {#Sec4} ========== Management of StD requires immediate reduction under sedation to avoid soft-tissue and vascular complications \[[@CR1]\]. Closed reduction is usually successful \[[@CR1], [@CR12], [@CR13]\]. In nonreducible cases, however, multiple reduction attempts using force should not be undertaken, and open reduction should be performed without delay \[[@CR6]\]. Frequently, associated lesions occur in the ankle and foot such as osteochondral fractures of the dislocated articular surfaces and fractures of neighboring bones (malleoli, the base of the fifth metatarsal, the cuboid, and the navicular tuberosities) \[[@CR14]\]. If such concomitant injuries are suspected from routine radiological images, further investigation should be undertaken with computed tomography (CT) scan. We believe that the prognosis of an isolated medial subtalar dislocation basically relies on three parameters: (1) immediate reduction (the necessity of which has already been underlined), (2) the amount of energy absorbed by the soft tissues at the moment of the violent impact, and (3) the period of postreduction immobilization. The need for immediate reduction has already been discussed earlier. The mechanism of injury is an important factor in predicting long-term results. The results are worse after more violent mechanisms \[[@CR14]\]. Simple inversion rarely produces dislocation with long-term morbidity, while more violent injuries, e.g., those incurred in motor vehicle accidents or after a fall from height, are more likely associated with persistent symptoms \[[@CR14]\]. Concerning the immobilization period, previous studies have set the length of non-weight-bearing plaster use to 5 or 6 weeks \[[@CR5], [@CR6]\]. According to other researchers \[[@CR7]--[@CR10]\], subtalar joint stiffness in uncomplicated StD can be minimized by avoiding immobilization longer than 4 weeks, or 6 weeks in case of StD associated with fractures. In general, the duration of postreduction immobilization is a controversial subject. Christiensen et al. \[[@CR15]\] immobilized their patients in a long cast for 8 weeks after reduction of the dislocation. Twenty-one of their 30 patients faced pain when walking. Buckingham \[[@CR16]\] reported on five patients, of whom four had decreased ROM following immobilization for 6 weeks. In other previous large series \[[@CR5]\] of uncomplicated medial StD that incorporated immobilization for 5 weeks, subtalar joint ROM was decreased by 30--50% compared with the contralateral side and the tibiotalar ROM was moderately reduced. Moreover long immobilization has been correlated with high percentages of arthritis and decreased function rating between 50--80% \[[@CR6], [@CR15], [@CR17]\]. McKeever \[[@CR18]\], on the other hand, strongly encouraged measures to prevent fibrosis by early mobilization of the foot. He stated that, since the reduction of a subtalar dislocation is extremely stable, early mobilization is possible. He began ROM exercises, consisting of active assisted motion of the subtalar and midtarsal joints, after 3 weeks of immobilization. Of the eight patients he treated, five were immobilized for 3 weeks and had a normal range of subtalar motion and no complaint of pain. Forty-five years after McKeever's study, our research confirms the advantages of early mobilization protocols for uncomplicated medial StD, using an even shorter period of immobilization than McKeever did. The prospective nature of our study, which allowed for the use of a prespecified protocol, ensured better control of the patient cohort during the follow-up period and a more reliable survey of the protocol itself. The application of certain inclusion and exclusion criteria created a homogeneous group of patients (those who had sustained uncomplicated medial StD), excluding bias due to nonsimilar injury patterns. We preferred to focus solely on uncomplicated medial StD, since this is the most usual type of peritalar dislocation that a clinician may need to manage. Homogeneity was also ensured by the fact that all patients received the same reduction method, which consisted of manipulation under sedation. It should be noted that none of the patients needed to receive general anesthesia, as usually suggested in literature \[[@CR1], [@CR12], [@CR13]\]. Sedation proved sufficient, allowing immediate reduction in the A & E department and saving valuable time that would be lost if the patient had to be transferred to theater and receive general anesthesia. Another strong point of our study was the existence of pre-assessed intervals for clinical and radiographic follow-up examination, which allowed for analogue timescale comparisons to be made. In contrast, the limited number of cases was a drawback of our research; this was expected, since subtalar dislocation is a rare injury. The small series of patients was the price to pay for the prospective character of the study. We believe, however, that it was worthwhile since the results, albeit from a small sample, were not controversial. Nonetheless, the study is still ongoing, thus further patients and results will be added in the future. Additional credibility could have been provided to the study if a control group of patients undergoing longer periods of immobilization had been included. The researchers were limited by two factors: the already mentioned small number of cases, and the ethical hesitation to provide patients with a type of treatment (long periods of immobilization) that we did not believe to be optimal. Based on the aforementioned, it would be safe to say that our working hypothesis was confirmed: Early ankle ROM exercises and PWB mobilization after uncomplicated medial StD seem to provide better functional results than those achieved by longer periods of immobilization, as already mentioned in literature. As joint instability, following closed reduction of medial subtalar dislocation, is not a possible complication, protracted immobilization only adds to joint stiffness and minimizes ankle and foot functionality. In contrast, early active ROM exercises may help the ligaments and the tendons of the site to heal without compromising proprioception of the joint. As basic practical recommendations extracted from this study, we could state the following: (a) immediate closed reduction should be applied in the case of any kind of StD, (b) sedation of the patient is usually sufficient for reduction of uncomplicated medial StD, (c) time-consuming general anesthesia should be used only in case of irreducible closed StD requiring open surgical reduction procedures, and (d) early mobilization protocols are indicated as beneficial for ankle and foot functionality after medial uncomplicated StD. Continuation of this study will add further credibility to its usefulness. Nonetheless, for definite results to be drawn, multicenter clinical trials will be required, and the creation of collaborative databanks of patients between multiple centers and countries may be necessary. AOFAS : American Orthopaedic Foot and Ankle Society (AOFAS) FWB : Full weight bearing MVA : Motor vehicle accident PWB : Partial weight bearing ROM : Range of motion SD : Standard deviation StD : Subtalar dislocation None. Open Access {#d32e985} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

PubMed Central

2024-06-05T04:04:19.317270

2011-2-10

PMC3052432

Introduction {#s0005} ============ 1 Recent successes in malaria control ([@b0015 b0020 b0025 b0085 b0190]) have resulted in optimism about the possibility of eliminating malaria in many areas where the disease is currently endemic ([@b0170]). Transmission reducing interventions are now acknowledged as key components of malaria control and elimination efforts ([@b0160 b0165 b0355]). The transmission of malaria depends on the presence of infectious sexual stage malaria parasites, gametocytes, in the human peripheral blood. These gametocytes do not cause clinical disease but once ingested by mosquitoes taking a blood meal, can develop into ookinetes, oocysts and ultimately sporozoites, thereby rendering the mosquito infectious to human beings. The infectiousness of gametocytes is influenced by their concentration ([@b0185 b0325 b0280]), degree of maturity ([@b0310 b0175]) and by mosquito ([@b0360]) and human immune responses ([@b0045]). The development of a human immune response to gametocytes is not surprising given that the vast majority of gametocytes are not taken up by mosquitoes but are cleared by the host immune system. There is indirect evidence that human immune responses may actively clear circulating gametocytes after recognising antigens on the gametocyte-infected erythrocyte ([@b0010 b0320 b0260]). A distinct human immune response may also reduce the infectiousness of gametocytes. Naturally occurring transmission reducing activity (TRA) has been associated with antibodies against antigens that are internally expressed in gametocytes but appear on the surface of gametes after gametocytes have been ingested by mosquitoes, notably Pfs48/45 and Pfs230 ([@b0075 b0250 b0045]). TRA forms the basis for the development of transmission blocking vaccines ([@b0080 b0240 b0270]) that could play a key role in malaria elimination efforts ([@b0265 b0315 b0155]) in particular by removing the asymptomatic reservoir from which mosquitoes can be infected. Two types of assays are commonly used to detect TRA: the standard membrane feeding assay (SMFA) and the direct membrane feeding assay (DMFA) ([@b0045]). In the SMFA, cultured gametocytes are fed to *Anopheles* mosquitoes in the presence of an (endemic) test serum or plasma or non-malaria control serum ([@b0235]); in the DMFA, which can be conducted in the field, blood samples from naturally infected gametocyte carriers are fed to mosquitoes in the presence of autologous plasma (AP) or control serum (CS), after a washing step ([@b0325]). Advantages of the DMFA are that it uses parasite strains that are naturally circulating in the study population, gametocyte densities that are representative of the natural situation and locally caught and reared mosquitoes. The DMFA may therefore resemble the natural situation better than the SMFA. However, due to the labour intensiveness of the assay, depending on the dissection of typically 20--60 mosquitoes per experiment, studies using DMFA are often too small to reliably confirm the existence of TRA in endemic populations, let alone to explore factors associated with TRA. Consequently, several fundamental questions about the nature of TRA remain. TRA is thought to be rapidly induced ([@b0050]) but short-lived ([@b0045 b0060 b0120]) but both of these assertions are yet to be confirmed in field studies. To investigate the induction, duration and efficacy of anti-gamete antibodies in natural infections, we determined the presence of TRA and associated factors in combined data from eight membrane-feeding studies conducted in The Gambia, Kenya and Cameroon. Materials and methods {#s0010} ===================== 2 Field studies {#s0015} ------------- 2.1 Data from eight trials with naturally infected individuals from The Gambia, Kenya and Cameroon were included in the current study ([Table 1](#t0005){ref-type="table"}). With the exception of a subset of the experiments from Cameroon, these data were not previously analysed to determine TRA. Experiments from The Gambia and Kenya involved feeds on blood samples obtained from children after anti-malarial treatment for a clinical malaria episode; samples from Cameroon were collected prior to treatment from both symptomatically and asymptomatically- infected individuals. ### Data from The Gambia {#s0020} 2.1.1 The study site, recruitment process, treatment regimes and feeding experiments have been described previously ([@b0310 b0115 b0305]). Briefly, malaria is seasonal and most transmission occurs between August and December. The transmission intensity at the time of the studies was of the order of 10--20 infective bites/person/year. Children attending the health centre at Farafenni were recruited in the transmission season. Eligible children were aged 0.5--15 years with a history of fever and *Plasmodium falciparum* asexual parasitaemia \>500/μL of blood in the absence of other species of *Plasmodium*. Exclusion criteria included anaemia (packed cell volume (PCV) \< 20%); signs of severe malaria; inability to take drugs orally; reported treatment with any anti-malarial within the past 2 weeks; and any evidence of chronic disease or other acute infection. After obtaining consent from parents or guardians, children were randomly assigned to different treatment regimens ([Table 1](#t0005){ref-type="table"}). Following treatment, patients were brought to the Medical Research Council (MRC) laboratory in Farafenni for membrane-feeding experiments on days 4 (*n* = 45), 7 (*n* = 126), 10 (*n* = 6) and 14 (*n* = 9) after treatment. The study protocols were approved by both the Joint Gambia Government/MRC Ethics Committee and the London School of Hygiene and Tropical Medicine Ethics Committee. ### Data from Kenya {#s0025} 2.1.2 The study in Kenya was conducted in Mbita, western Kenya, on the shores of Lake Victoria. Symptomatic children aged 6 months-10 years with uncomplicated malaria and a *P. falciparum* mono-infection with a density of at least 1000 parasites/μL were recruited. Exclusion criteria were similar to those described for the studies in The Gambia; anaemia in this case was defined as haemoglobin concentration lower than 5 g/dL. Children were randomized to receive either artemether--lumefantrine (AL) or dihydroartemisinin--piperaquine. Seven days after enrolment, individuals aged 2--10 years with and without microscopically confirmed gametocytes were recruited for membrane-feeding experiments; however, the current analyses were restricted to microscopically confirmed gametocyte carriers. The study protocol received ethical approval from the Ethical Review Committee of the Kenya Medical Research Institute and the Ethics Committee of the London School of Hygiene and Tropical Medicine The trial was registered online at [\<http://clinicaltrials.gov/ct2/show/NCT00868465\>](http://www.clinicaltrials.gov). ### Data from Cameroon {#s0030} 2.1.3 Two separate studies were conducted in Cameroon in 1995--1998 (Cameroon 1 ([@b0030 b0135])) and in 1995--1996 (Cameroon 2 ([@b0210])). In the first study, gametocyte carriers (aged 4--38 years) were recruited during community-wide cross-sectional surveys in the district of Mengang, where annual malaria transmission intensity is around 100 infective bites/person/year ([@b0035]). In the second study, 55 gametocyte carriers (aged 1--63 years) were recruited among patients of the Messa dispensary in an urban quarter of Yaoundé and exposed to a transmission intensity of ∼34 infectious bites per person per year ([@b0335]). Of these, 5.5% (3/55) presented with a temperature ⩾37.5 °C ([@b0210]). In both studies, individuals with gametocytes by microscopy were selected for membrane-feeding experiments. Individuals with asexual parasites received anti-malarial treatment following national guidelines after sampling for membrane feeds was completed. The projects were approved by the National Ethical Clearance Committee for Cameroon. Membrane feeding and mosquito dissection {#s0035} ---------------------------------------- 2.2 Venous blood samples (2--4 ml) were obtained from children whose parent or guardian had given specific consent for the procedure. Venous blood in citrate--phosphate dextrose (The Gambia) or heparin (Cameroon, Kenya) was centrifuged, and the plasma was removed. After being washed, the red blood cell pellet was split into two aliquots of 300--500 μL each. These were resuspended to a PCV of 33% in, respectively, the original AP and in pooled AB serum from European donors with no history of malaria exposure (CS). Each suspension then was fed to 50--100 3--5 day old female *Anopheles gambiae* sensu stricto mosquitoes. In studies from The Gambia, the next generation progeny of wild-caught gravid female mosquitoes were used; locally reared laboratory strain mosquitoes that were adapted to feeding on a membrane feeder were used in Cameroon ([@b0325]) and Kenya ([@b0055]). In all studies, starved mosquitoes were allowed to feed for 15--30 min via an artificial membrane attached to a water-jacketed glass feeder maintained at 37 °C. After feeding, blood-fed mosquitoes were kept at 26--28 °C with permanent access to a 10% sucrose solution without further blood meals. Mosquito midguts were dissected out 7--8 days later in PBS (The Gambia) or 2% (Cameroon, The Gambia) or 0.5% mercurochrome (Kenya) in distiled water; the number of oocysts -- a developmental stage of the parasite found on the insect midgut -- was recorded. Data analysis {#s0040} ------------- 2.3 Data were entered using Epi-Info or MS-Access and analysed using Stata version 11 (Stata Corporation, Texas, USA). Analyses were restricted to experiments on microscopically confirmed gametocyte carriers that had at least 10 mosquitoes dissected and resulted in at least one infected mosquito in the CS and/or AP experiment. The latter criterion was invoked to rule out technical problems with the DMFA. The relationship between gametocyte density and mosquito infection rates was visualised for all data combined by grouping mosquito feeding experiments according to the density of gametocytes in the blood ingested (into 20 evenly spaced groups on the log-scale). A Bland--Altman (difference) plot was created to visualise the difference between the proportion of mosquitoes that was infected after feeding on a blood sample with CS or AP. Oocyst counts were highly over-dispersed (mean = 2.19; S.D. = 11.51) and were presented in categories: 0, 1--2, 3--5, 6--15, 16--50 and \>50 oocysts. The studies in The Gambia recruited individuals from a wide age range (0.5--15 years) who were treated with artemisinin-combination therapy (ACT) and non-ACT treatment, recruited at the start and through the transmission season and of whom a proportion presented with gametocytes at enrolment, i.e. 4--14 days before membrane feeding. We therefore used the combined datasets from The Gambia to determine factors associated with TRA. The influence of host factors was examined in two ways. Firstly, the prevalence of infectiousness among donors, defined as the proportion of individuals that provided a gametocyte-positive blood sample which resulted in at least one infected mosquito, was compared between CS and AP feeds for: different age-groups (\<5 compared with ⩾5 years); season of enrolment (peak compared with start of transmission season); treatment (non-ACT compared with ACT); history of microscopic gametocytaemia (gametocyte-free versus gametocytes present at enrolment); gametocyte density at the day of membrane feeding above or below the median value (50 gametocytes/μL). Different feeding days were combined. The McNemar test was used to test for a difference between paired CS and AP experiments. Subsequently, the proportion of infected mosquitoes was compared between CS and AP feeds for the same categories of participants. A multilevel logistic regression model was used for this purpose using a multilevel generalised linear model (GLLAMM, Stata version 11; Stata Corporation, Texas, USA). This model incorporated clustering per patient and random effects to account for differences between studies. Because mosquito infection rates were strongly associated with gametocyte density, all analyses except those directly testing the influence of having a high gametocyte density (⩾50 gametocytes/μL) at the time of membrane feeding were adjusted for log (ln) transformed gametocyte density. The GLLAMM model was also used for multivariate analyses where experiments with CS were used as reference category. Interaction terms were included in the model; variables were selected for the multivariate model if *P* \< 0.05 in the univariate analyses and retained in the model if *P* \< 0.10. Results {#s0045} ======= 3 The association between gametocyte density and mosquito infection rates {#s0050} ----------------------------------------------------------------------- 3.1 There was a positive association between gametocyte density and the proportion of infected mosquitoes in CS and AP experiments when all studies were combined ([Fig. 1](#f0005){ref-type="fig"}). Mosquito infection rates were consistently higher in CS experiments compared with AP experiments ([Figs. 1 and 2](#f0005 f0010){ref-type="fig"}). The difference between mosquito infection rates in paired CS-AP experiments was plotted against the average proportion of infected mosquitoes ([Fig. 2](#f0010){ref-type="fig"}). This Bland--Altman plot indicated that in 68.2% of the paired experiments the proportion of infected mosquitoes in the CS experiment was higher than that in the AP experiment (*P* = 0.006). The proportion of infected mosquitoes was statistically significantly higher in CS experiments compared with AP experiments for The Gambia ([Fig. 3](#f0015){ref-type="fig"}, *P* \< 0.001), Cameroon 1 (*P* = 0.03) and Cameroon 2 (*P* = 0.004), after adjustment for the correlation between observations from the same individual and study-year (only applicable for The Gambia). The increased proportion of infected mosquitoes in the CS experiments compared with the AP experiments was not statistically significant for the trial in Kenya (*P* = 0.29). Among infected mosquitoes, there was little evidence for a difference in oocyst burden between CS and AP feeds. There was no difference in intensity of infection (oocyst counts among infected mosquitoes) between CS and AP feeds in the combined Gambia data (*P* = 0.92), the Kenya data (*P* = 0.11) and the first trial in Cameroon (*P* = 0.12), after adjusting for correlation between observations from the same individual and, where applicable, study-year. Oocyst counts in infected mosquitoes were higher in CS compared with AP feeds in the second trial in Cameroon (*P* = 0.01). Factors associated with TRA {#s0055} --------------------------- 3.2 The dataset from the combined Gambian trials was the largest dataset and provided most details of human host factors. Because these trials showed no evidence for differing oocyst counts among infected mosquitoes between CS and AP feeds, the analyses were done on the prevalence of infectiousness (i.e. the proportion of samples infecting at least one mosquito) and the proportion of infected mosquitoes. The proportion of individual donor samples that infected at least one mosquito was higher after serum replacement, although at borderline significance (*P* = 0.05, [Table 2](#t0010){ref-type="table"}). This difference was more pronounced when the proportion of infected mosquitoes was considered (*P* \< 0.001). A higher infectivity in CS compared with AP feeds was found for samples from children older than 5 years of age, samples from individuals who presented with gametocytes at enrolment and samples from individuals who were sampled at the end of the transmission season. The proportion of infected mosquitoes increased after serum replacement for individuals with low and high gametocyte densities but the effect of serum on transmissibility was more pronounced in those with higher gametocyte densities. The type of treatment (ACT versus non-ACT) did not influence the transmission-reducing capacity of serum. To determine independent predictors of TRA, a multivariate model was built with CS experiments as the reference group ([Table 3](#t0015){ref-type="table"}). This model adjusted estimates for gametocyte density at the time of feeding and correlations between observations from the same individual, and incorporated a random effect for study-year. Factors associated with TRA in the univariate analysis were identical to those in [Table 2](#t0010){ref-type="table"}: gametocytes present at the preceding enrolment visit, older age and sampling late in the season. In the multivariate model, however, only the presence of gametocytes at enrolment and sampling late in the season were significantly associated with lower mosquito infection rates after adjustment for gametocyte density, correlations between observations from the same individual and study year. Discussion {#s0060} ========== 4 In this study, we describe the prevalence of transmission reducing immune responses in naturally infected individuals from three malaria endemic countries. All studies showed evidence for TRA that was associated with several indicators of increased recent exposure to gametocytes. When combining successful membrane-feeding experiments of 201 naturally infected individuals, we observed a clear association between mosquito infection rates and gametocyte density. Although this association was evident for both experiments with permissive CS from non-exposed donors and experiments using AP, mosquito infection rates were consistently higher in CS experiments. Replacing AP by CS resulted in significantly higher infection rates in two-thirds of all membrane-feeding experiments. The higher infectivity in CS feeds was well captured by the prevalence scale (i.e. the presence or absence of oocysts) and there was little to be gained by analysis of oocyst densities in infected mosquitoes in most of the datasets. In univariate analyses, higher gametocyte densities at the time of membrane feeding, gametocyte carriage during the week prior to the day of membrane feeding, older age of children and sampling later in the transmission season were all associated with TRA, defined as either a higher likelihood of transmission-success in paired CS-AP experiments or a higher proportion of infected mosquitoes in CS experiments. A multivariate model indicated that the presence of gametocytes during the week prior to the experiment and sampling late in the transmission season were independently associated with increased TRA. Increased TRA in children that carried gametocytes during the week prior to the feeding experiment suggests that recent exposure to gametocytes is associated with TRA. A rapid acquisition of antibody responses to Pfs48/45 and Pfs230, both associated with naturally acquired TRA ([@b0045]), was previously shown in individuals newly exposed to malaria ([@b0215 b0050]). It is therefore plausible that transmission-reducing antibodies were acquired or boosted in response to recent exposure to gametocyte antigens. Some exposure to gametocyte antigens will have remained undetected by relying on microscopy that may underestimate gametocyte carriage considerably ([@b0005 b0055 b0275 b0285 b0200]). It is plausible that some of the individuals who developed gametocytes after treatment had sub-patent gametocytes at enrolment ([@b0005 b0275]). These individuals, however, were less likely to show transmission reducing immune responses at the time of feeding than those who had microscopically detectable gametocytes at enrolment. This suggests that exposure to relatively high density (i.e. microscopically detectable) gametocytes several days before the experiment is needed to elicit functional TRA. This observation is in line with the hypothesis that transmission reducing immunity is rapidly induced and depends on recent exposure to gametocytes ([@b0045]). A similar biological mechanism could be responsible for the observed increase in TRA towards the end of the transmission season. Exposure to gametocytes shows considerable seasonal fluctuation with a higher prevalence and density of gametocyte carriage in the transmission season compared with the dry season ([@b0100 b0220]). The proportion of children who have been exposed to gametocytes in the weeks before the membrane-feeding experiments will therefore be highest towards the end of the transmission season, resulting in higher levels of TRA. These findings also indirectly support the short-lived nature of transmission reducing immune responses. If cumulative life-long exposure to gametocytes was key to the acquisition and maintenance of transmission reducing immune responses, as observed for pre-erythrocytic and blood-stage immune responses ([@b0095 b0365]), one would expect age to be the dominant factor predicting TRA. An effect of age on infectiousness that is independent of gametocyte density was found in some ([@b0130 b0330 b0280]) but not all ([@b0145 b0325 b0105]) studies. Our experiments from The Gambia only included children with a median age of 4 years (range 0.5--15 years) and we therefore cannot extrapolate our findings to adults, who appear to have a markedly lower exposure to gametocytes than children ([@b0100]). In multivariate models, age lost its influence on TRA after adjustment for recent exposure to gametocytes. This suggests the association between age and transmission reducing immune responses is complex. Age could be an indicator of better developed clinical immune responses that change malaria disease progression and treatment seeking behaviour. We hypothesise that in our population of children, older age allowed infections to remain asymptomatic for a longer period, thereby postponing treatment seeking ([@b0350 b0205]) and allowing for a longer period of potential gametocyte production. This would again indicate an influence of higher recent exposure to gametocytes in older children. When considering our individually paired AP--CS experiments, we found evidence for an increased transmission in AP compared with CS experiments in 3.4% of the experiments (4/116; data not shown). The phenomenon of transmission enhancement has been reported in other field studies ([@b0150 b0245]) and may be a real biological phenomenon rather than an artefact resulting from variation in membrane feeding assays ([@b0345]). The biological mechanism is unclear. In some studies ([@b0230 b0125 b0180]) transmission enhancement was linked to the presence of very low concentrations of anti-gamete antibodies but this was not confirmed in the largest study on transmission enhancement in *P. falciparum* ([@b0345]). The importance of this phenomenon at a population level will require further study and our limited number of samples showing enhancement did not allow us to determine associated factors. Our findings illustrate some of the difficulties encountered when doing transmission experiments under field conditions. Mosquito infection rates differed between study sites ([@b0300]) and the association between gametocyte densities and mosquito infection rates can be highly variable ([@b0140 b0070 b0290 b0040 b0090 b0280 b0225]). Even when cultured gametocytes are used under highly standardised laboratory conditions, the proportion of infected mosquitoes and the number of oocysts resulting from a given gametocyte density can vary considerably ([@b0340]). This variation is largest at low gametocyte concentrations ([@b0340]) and is partly a consequence of the narrow age range over which gametocytes are infectious ([@b0195 b0175]). The type of anticoagulant used may affect oocyst development ([@b0235 b0295]), an effect that has been best described for EDTA, which may reduce oocyst numbers ([@b0295]). Such an effect will not have influenced our conclusions since our analyses were based on pair-wise comparisons of samples from the same individual (i.e. with the same anticoagulant) or restricted to data from the Gambia where only citrate phosphate dextrose was used. Similar to variation in the highly standardised SMFA, we expect there will be considerable variation in TRA measured by DMFA ([@b0210 b0340]). This makes it complicated to draw firm conclusions from individual cross-sectional observations on TRA. Longitudinal studies are therefore needed to test the causality of our findings. In future longitudinal studies the acquisition rate and longevity of TRA should be determined in repeated membrane feeding assays on naturally infected individuals. The current findings confirm the presence of TRA in naturally-exposed populations and provide biologically plausible support for the hypothesis that TRA depends on recent exposure to gametocytes. It is well established that transmission can be reduced by antibodies to the gametocyte/gamete antigens Pfs 48/45, Pfs 230 and PfHAP2 ([@b0265 b0315]) and the ELISA titer of antibodies against these antigens is related to the level of TRA ([@b0255 b0110 b0065]). Further dissection of the specific contribution to TRA of these antibodies, and of other immune responses, is required. The data presented here strongly suggest that recent gametocyte exposure is a requirement for effective TRA. Thus vaccines designed to block transmission by mimicking natural TRA will have to overcome this requirement for boosting by antigen exposure, if they are to be effective tools for reducing or interrupting transmission of *P. falciparum.* The authors wish to thank the communities from The Gambia, Cameroon and Kenya for their willingness to participate in the study. We also acknowledge the assistance of entomology assistants at the various sites and the help of Lucy Okell at Imperial College, UK during data analysis. Teun Bousema was supported by the FIGHTMAL project, funded by the European Community's Seventh Framework Programme \[FP7/2007-2013\] under grant agreement PIAP-GA-2008-218164; Chris Drakeley by a research fellowship in tropical medicine (\#063516) from the Wellcome Trust, UK. ::: {#f0005 .fig} Fig. 1 ::: {.caption} ###### The relationship between gametocyte density by microscopy and the proportion of infected mosquitoes. Light grey circles indicate the proportion of infected mosquitoes after feeding on blood samples with control serum (CS); dark grey circles indicate autologous plasma (AP). The sizes of the circles are proportional to the number of mosquitoes dissected for a given range of gametocyte densities. The dataset combines observations from studies in The Gambia (1998--2002; *n* = 106), Kenya (2009; *n* = 11) and Cameroon (1995--1998; *n* = 74). :::  ::: ::: {#f0010 .fig} Fig. 2 ::: {.caption} ###### Bland--Altman (difference) plot comparing paired experiments with control serum (CS) and autologous plasma (AP). Each dot represents a paired CS-AP experiment. The mean proportion of infected mosquitoes is given on the *X*-axis and the difference between AP and CS experiments on the *Y*-axis. A positive value indicates a higher mosquito infection rate for CS feeds. In 68.2% (150/220) of the paired experiments the mosquito infection rate was higher in the CS experiment compared with the AP experiment (positive values); in 29.5% (65/220) of the experiments the infection rate was higher in the AP experiment (negative values); in 2.3% (5/220) of the experiments the proportion of infected mosquitoes was identical for CS and AP feeds. :::  ::: ::: {#f0015 .fig} Fig. 3 ::: {.caption} ###### Oocyst burdens in mosquitoes after feeding on blood samples with control serum (CS) or autologous plasma (AP) in different experiments. Bars indicate the proportion of mosquitoes with the indicated oocyst burden after feeding on blood samples with control serum (CS) or autologous plasma (AP) in studies in The Gambia (1998--2002), Kenya (2009), Cameroon 1 (1997) and Cameroon 2 (1995). The asterisk indicates a statistically significant difference between CS and AP experiments in the proportion of mosquitoes with ⩾1 oocyst. :::  ::: ::: {#t0005 .table-wrap} Table 1 ::: {.caption} ###### Summary membrane-feeding experiments on microscopically confirmed gametocyte carriers with paired Autologous Plasma (AP) and Control Serum (CS) observations. Only gametocyte carriers who had a minimum of 10 mosquitoes dissected in both AP and CS feeds were included in the analyses. ::: Country Year Timing of membrane feed experiments Median gametocyte density (IQR) Number of combined AP-CS feeds (total mosquitoes CS;AP) \% successful feeds (*n*/*N*)[a](#tblfn1){ref-type="table-fn"} --------------------------- ------------------------------- ---------------------------------------------------------------------------------------------- --------------------------------- --------------------------------------------------------- ---------------------------------------------------------------- Farafenni, The Gambia 1998 ([@b0310]) Four (*n* = 45) or 7 days (*n* = 10) after treatment with CQ, SP, SP + AS1 or SP + AS3 120 (48--376) 55 (1439;1272) 56.4 (31/55) Farafenni, The Gambia 1999 ([@b0310]) Seven days after treatment with SP or SP + AS3 20 (10--220) 33(701;715) 100.0 (33/33) Farafenni, The Gambia 2000 ([@b0110 b0115]) Seven days after treatment with CQ or CQ + AS3 50 (10--245) 38 (652;622) 73.7 (28/38) Farafenni, The Gambia 2001 ([@b0175]) Seven (*n* *= *27), 10 (*n* = 6) or 14 (*n* = 8) days after treatment with CQ, SP or CQ + SP 100 (35--615) 41(782;750) 48.8 (20/41) Farafenni, The Gambia 2002 ([@b0305]) Seven days after treatment with CQ + SP or AL 53 (20--160) 19(368;373) 21.1 (4/19) **Farafenni, The Gambia** **1998--2002** **After treatment with anti-malarials** **65(20--320)** **186 (3942; 3732)** **62.4 (116/186)** **Mbita, Kenya** **2009** **After treatment with AL or DP** **200 (120--440)** **12 (360;360)** **91.6 (11/12)** 

 **Mengang, Cameroon 1** **1997 (**[@b0030 b0135]**)** **Prior to treatment** **192 (80--608)** **19 (519;540)** **100 (19/19)** **Yaoundé, Cameroon 2** **1995 (**[@b0210]**)** **Prior to treatment** **296 (88--536)** **55 (1465;1086)** **100 (55/55)** Bold lines indicate the summary figures per country. IQR, interquartile range (25th--75th percentile); CQ, chloroquine; SP, sulphadoxine-pyrimethamine; AS1, one dose of artesunate given together with SP; AS3, three doses of artesunate; AL, artemether-lumefantrine; DP, dihydroartemisinin-piperaquine. a Defined as at least one infected mosquito in feeding experiment using AP and CS. ::: ::: {#t0010 .table-wrap} Table 2 ::: {.caption} ###### Factors associated with transmission reducing activity (TRA) in The Gambian dataset. ::: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Variable Prevalence of infectiousness *P*-value[a](#tblfn2){ref-type="table-fn"} Proportion infected mosquitoes OR, *P*-value[b](#tblfn3){ref-type="table-fn"} ----------------------------------------- ------------------------------ -------------------------------------------- -------------------------------- ------------------------------------------------ All data AP: 63.8 (74/116)\ *p* = 0.05 AP: 11.6 (278/2391)\ 0.60 (0.50--0.72)\ CS: 77.6 (90/116) CS: 17.2 (433/2518) *P *\< 0.001 

 *Gametocyte density at feeding*  \<50 gametocytes/μL AP: 62.2 (31/50)\ *p* = 0.49 AP: 7.2 (76/1049)\ 0.72 (0.52--0.99)\ CS: 70.0 (35/50 CS: 10.2 (112/1101 *P = *0.045  ⩾50 gametocytes/μL AP: 65.2 (43/66)\ *p* = 0.04 AP: 15.0 (201/1341)\ 0.54 (0.43--0.68)\ CS: 83.3 (66/66) CS: 22.7 (321/1417) *P \< *0.001 

 *Gametocyte prevalence at presentation*  Yes AP: 56.1 (23/41)\ *p* = 0.007 AP: 10.8 (88/818)\ 0.36 (0.26--0.49)\ CS: 87.8 (36/41) CS: 22.4 (183/817) *P *\< 0.001  No AP: 64.7 (44/68)\ *p* = 0.27 AP: 11.6 (168/1449)\ 0.87 (0.68--1.11)\ CS: 75.0 (51/68) CS: 13.2 (203/1541) *P *= 0.26 

 *Age*  Under 5 years AP: 65.4 (34/52)\ *p* = 0.87 AP: 9.5 (101/1061)\ 0.84 (0.62--1.14)\ CS: 67.3 (35/52) CS: 10.5 (116/1101) *P *= 0.27  Over 5 years AP: 60.0 (33/55)\ *p* = 0.003 AP: 12.5 (140/1123)\ 0.55 (0.43--0.71)\ CS: 89.1 (49/55) CS: 19.5 (232/1193) *P *\< 0.001 

 *Drug*  Non-ACT AP: 63.6 (49/77)\ *p* = 0.10 AP: 11.3 (180/1595)\ 0.59 (0.47--0.74)\ CS: 77.9 (60/77) CS: 16.7 (270/1616) *P *\< 0.001  ACT AP: 64.1 (25/39)\ *p* = 0.30 AP: 12.3 (98/796)\ 0.62 (0.45--0.86)\ CS: 76.9 (30/39) CS: 18.1 (163/902) *P *= 0.004 

 *Season*  Early AP: 69.6 (32/46)\ *p* = 0.69 AP: 11.7 (101/865)\ 0.99 (0.72--1.38)\ CS: 73.9 (34/46) CS: 11.8 (105/888) *P *= 0.97  Late AP: 57.1 (40/70)\ *p*\<0.001 AP: 11.6 (177/1526)\ 0.47 (0.37--0.59)\ CS: 82.9 (58/70) CS: 20.1 (328/1630) *P *\< 0.001 ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- a By McNemar test for paired control serum-autologous plasma (CS--AP) observations. b By Generalised Linear Latent and Mixed Models (GLAMM), adjusting for gametocyte density at the time of feeding (except for the variable 'gametocyte density') and correlations between observations from the same individual and study-year. OR, odds ratio. ::: ::: {#t0015 .table-wrap} Table 3 ::: {.caption} ###### Independent predictors of mosquito infection prevalence in the Gambian dataset. Results of Generalised Linear Latent and Mixed Models (GLAMM) on the proportion of infected mosquitoes. Estimates were adjusted for gametocyte density, the correlation between observations from the same individuals and a random effect was added for study year. Variables were added to the model if *P \< *0.05 in the univariate model (adjusting for gametocyte density) and retained in the model if *P \< *0.05 in the multivariate model through backward elimination of non-significant variables. ::: Univariate OR (95% CI) *P*-value Multivariate OR (95% CI) *P*-value ---------------------------- ------------------------ ----------- -------------------------- ----------- Control serum 1.0 (ref) 1.0 (ref) 

 *Autologous plasma*  Gametocytes at enrolment 0.36 (0.26--0.49) \<0.001 0.52 (0.37--0.74) \<0.001  Over 5 years of age 0.58 (0.45--0.75) \<0.001 --  Late in the season 0.49 (0.39--0.61) \<0.001 0.62 (0.48--0.80) \<0.001 Gametocyte density (ln)/μL 1.34 (1.17--1.53) \<0.001 1.40 (1.20--1.62) \<0.001 OR, odds ratio; CI, confidence interval; ref, reference category. :::

PubMed Central

2024-06-05T04:04:19.318726

2011-3-1

PMC3052433

Published online: February 3, 2011 Introduction {#sec1} ============ Cytotoxic T cells that express CD8 and recognize peptide-class I major histocompatibility complexes play a key role in immunity. The transcriptional program that determines the effector or memory fate of CTL is controlled by the TCR and cytokines. In particular, interleukin-2 (IL-2), a member of the common cytokine receptor gamma-chain (γ~c~) family of cytokines, has a key role to maintain antigen-specific effector and memory CD8^+^ T cells during viral infections ([@bib3 bib12 bib25 bib32 bib50]). IL-2 is important for CTL because it can upregulate glucose and nutrient uptake in a response that increases cellular energy production in order to support the biosynthetic demands of rapid cell division and effector function ([@bib10 bib31]). However, IL-2 also directs the transcriptional program of CTL and promotes effector CTL differentiation at the expense of memory cell formation ([@bib25 bib35 bib38]). IL-2 acts via a receptor complex consisting of the common gamma chain (γ~c~), a β subunit (CD122), and CD25. Triggering of the IL-2 receptor stimulates the Janus family kinases Jak1 and Jak3 and induces the tyrosine phosphorylation and DNA binding of STAT (signal transducer and activator of transcription) family transcription factors STAT5 and STAT3. The activation of STAT5 is important for IL-2 signaling ([@bib32]) but is not sufficient to mimic the effects of IL-2 on T cell biology. In this context, IL-2 induces sustained accumulation of phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P~3~), the product of phosphatidylinositol 3-kinases (PI3K) ([@bib10 bib45]), and activates a cascade of serine-threonine kinases. These include Akt (protein kinase B \[PKB\]) ([@bib48]) and other members of the protein kinase A/G/C-related (AGC kinase) family including p70 ribosomal S6 kinases (S6Ks) and p90 ribosomal S6 kinase (RSK) ([@bib29]). The proper balance of Akt and STAT5 signaling seems to be critical for virus-specific CD8^+^ T cells ([@bib18]). However, the precise role of Akt in the transcriptional programs that direct effector CD8^+^ T cell differentiation has not been explored. The coordination of glucose metabolism is essential for T cell activation ([@bib9 bib31]) and this function might be controlled by PI3K-Akt. This assumption stems partly from extrapolating data from thymus studies, where it is clear that PI3K and Akt control glucose uptake and the expression of amino acid transporters and transferrin receptors in T cell progenitors ([@bib23 bib26 bib34]). Further support for a tropic role for PI3K-Akt in peripheral T cells come from experiments with PI3K inhibitors, which prevent Akt activation and prevent increases in both glucose uptake and amino acid uptake by activated peripheral T cells ([@bib10]). However, there are caveats because some of the PI3K inhibitors used in early T lymphocyte studies ([@bib10]) have off-target effects ([@bib4]). In this context, evidence for Akt-independent pathways that control T cell metabolism and proliferation are emerging ([@bib5 bib14 bib20]). Moreover, constitutively active Akt can stimulate promote cell growth and survival of CD4^+^ T cells but not CD8^+^ T cells ([@bib18 bib42 bib44]), indicating that Akt may not have an obligate role to control metabolism in all T cell subpopulations. Accordingly, the focus of the present study is the role of PI3K-Akt in TCR and IL-2 regulation of CTL biology. The data have established that PI3K- and Akt-independent mechanisms control CTL metabolism, survival, and proliferation but there is a seminal role for Akt signaling in CTL to direct a diverse transcriptional program that determines effector versus memory T cell fate. Results {#sec2} ======= p110δ Couples the IL-2 Receptor to Akt but IL-2-Mediated T Cell Proliferation Is p110δ Independent {#sec2.1} -------------------------------------------------------------------------------------------------- TCR-primed CD8^+^ T cells cultured in IL-2 clonally expand and differentiate to effector cytotoxic T cells (CTL) ([@bib33]). This model mimics the in vivo situation where sustained IL-2 signaling promotes the production of terminally differentiated effector cytotoxic T cells ([@bib25 bib32 bib38]). IL-2 strongly activates Akt and antigen-experienced T cells cultured in IL-2 contain high amounts of Akt phosphorylated on the 3-phosphoinositide-dependent kinase 1 (PDK1) substrate site T308 ([Figure 1](#fig1){ref-type="fig"}A). The TCR activates Akt via a PI3K complex that contains the p110δ catalytic subunit ([@bib16]). [Figure 1](#fig1){ref-type="fig"}A shows that p110δ PI3K also coupled the IL-2 receptor to Akt. Hence, T cells with wild-type p110δ substituted with a catalytically inactive mutant, p110δ^D910A^, do not phosphorylate Akt T308 in response to IL-2 and thus lack Akt activity as judged by the loss in phosphorylation of Foxo transcription factors ([Figure 1](#fig1){ref-type="fig"}A). The sustained proliferation and survival of antigen-primed CTL is dependent on exogenous IL-2 ([Figure 1](#fig1){ref-type="fig"}B). However, activated p110δ^D910A^-expressing T cells proliferated normally to IL-2 ([Figure 1](#fig1){ref-type="fig"}C). Further experiments examined the impact of a p110δ inhibitor, IC87114, on IL-2-induced Akt activity in antigen-activated P14 TCR transgenic T cells. Triggering of the TCR expressed on these cells with a peptide from the lymphocytic choriomeningitis virus (LCMV), glycoprotein gp33-41, presented by the MHC class I molecule H-2D^b^, followed by clonal expansion in IL-2, generated a homogenous population of CTL. The exposure of IL-2-maintained P14 CTL to IC87114 caused loss of Akt T308 phosphorylation ([Figure 1](#fig1){ref-type="fig"}D) but did not prevent T cell proliferation ([Figure 1](#fig1){ref-type="fig"}E). To explore further the role of Akt in IL-2 proliferative responses, a set of experiments with AktI, a selective allosteric inhibitor of all three isoforms of Akt, were performed. AktI prevents the essential PDK1-mediated phosphorylation of Akt on T308 ([@bib17]). The addition of AktI to CTL caused loss of Akt T308 phosphorylation and a resultant loss of Akt activity as judged by the loss in phosphorylation of the Akt substrates Foxo1 and Foxo3a ([Figure 1](#fig1){ref-type="fig"}F). However, AktI did not prevent IL-2-induced proliferation of the CTL nor did it cause cell death ([Figure 1](#fig1){ref-type="fig"}G). Hence, CTL do not need active Akt to proliferate and survive. PDK1 Is Required for T Cell Proliferation {#sec2.2} ----------------------------------------- One explanation for the Akt independence of IL-2-induced proliferation is that other members of the AGC family of kinases might compensate for loss of Akt function. We therefore examined the impact of deleting the gene encoding PDK1, *Pdpk1,* on IL-2 signal transduction. PDK1 phosphorylates and activates multiple AGC kinases such as RSK and S6K1 in a response that is independent of PI(3,4,5)P~3~ production ([@bib5]). *Pdpk1* deletion is thus a strategy that simultaneously inactivates Akt- and PI(3,4,5)P~3~-independent AGC kinases in T cells ([@bib26]). Accordingly, mice expressing floxed *Pdpk1* alleles (*Pdpk1*^fl/fl^) were backcrossed with C57BL/6-GT(ROSA)26^tm9\ (creEsr1)\ Arte^ mice (TamoxCre) that express a tamoxifen-regulated Cre recombinase in the ROSA locus. *Pdpk1*^f*l/fl*^ Tamox*Cre* CTL were then generated and treated with 4-hydroxytamoxifen (4OHT) to delete the floxed *Pdpk1* alleles. *Pdpk1* deletion in 4OHT-treated *Pdpk1^fl/fl^* Tamox*Cre* CTL was confirmed ([Figure 2](#fig2){ref-type="fig"}A). Importantly, immunoblot analysis also revealed that 4OHT-treated *Pdpk1^fl/fl^* Tamox*Cre* CTL had lost phosphorylation of RSK on the PDK1 target site S227 and had lost phosphorylation of Akt on the PDK1 target site T308. The loss of Akt and S6K1 catalytic activity was judged by the loss of phosphorylated Foxos and the ribosomal S6 subunit ([Figure 2](#fig2){ref-type="fig"}B). *Pdpk1* deletion in CTL thus eliminated the activity of multiple AGC kinases. We therefore assessed the ability of IL-2 to sustain survival and proliferation of *Pdpk1*-null cells. [Figure 2](#fig2){ref-type="fig"}C shows that *Pdpk1*-null CTL failed to proliferate in response to IL-2. However, whereas IL-2 deprivation resulted in rapid cell death of CTL ([Figure 1](#fig1){ref-type="fig"}B), the loss of PDK1 did not cause cell death ([Figure 2](#fig2){ref-type="fig"}D). PDK1 is thus essential for IL-2-induced T cell proliferation but not for cell survival. IL-2 and PDK1 but Not Akt Sustain Glucose Uptake in CTL {#sec2.3} ------------------------------------------------------- In many cells Akt controls cell proliferation because it maintains glucose metabolism. In this context, as naive CD8^+^ T cells respond to cognate antigen and differentiate to CTL, they increase glucose uptake and become highly glycolytic. Glucose uptake in CTL was not autonomous but depended on the provision of exogenous IL-2 ([Figure 3](#fig3){ref-type="fig"}A). Moreover, the importance of glucose uptake for T cell proliferation was easily demonstrated: lowering of exogenous glucose concentrations severely impaired IL-2-induced proliferative responses ([Figure 3](#fig3){ref-type="fig"}B). That glucose deprivation but not Akt inhibition prevented IL-2-induced CTL proliferation argues that Akt activity is not required for IL-2-driven glucose uptake. [Figure 3](#fig3){ref-type="fig"}C addresses this point and shows that loss of Akt activity had no impact on the ability of IL-2-cultured CTL to take up glucose. The Akt independence of IL-2-induced T cell proliferation can thus be explained by the ability of IL-2 to maintain glucose uptake via Akt-independent pathways. We then assessed whether PDK1 plays a role in IL-2-induced glucose uptake. [Figure 3](#fig3){ref-type="fig"}D shows that PDK1-null T cells had a strikingly reduced rate of glucose uptake but no defects in amino acid uptake ([Figure 3](#fig3){ref-type="fig"}E). Hence PDK1, but not Akt, is essential for IL-2 control of glucose uptake. The Akt independence of IL-2-induced glucose uptake raised the question whether Akt activity was required for the initial increase in glucose metabolism initiated by the TCR in CD8^+^ T cells. [Figure 3](#fig3){ref-type="fig"}F shows that TCR triggering stimulated a large increased glucose uptake and this was unimpaired in cells treated with the Akt inhibitor AktI or the PI3K p110δ inhibitor IC87114. The regulation of glucose uptake via the T cell antigen receptor in CD8 T cells is thus not mediated by Akt. PDK1 and Akt Transcriptional Programs in CTL {#sec2.4} -------------------------------------------- If Akt is not essential for glucose uptake or IL-2-mediated T cell survival or proliferation, what does it do? This question is important because sustained IL-2 signaling is required to maintain CTL effector function both in vitro and in vivo ([@bib25 bib38]) and there is clearly the potential for Akt to mediate the effects of IL-2 on CTL function. Accordingly, we used Affymetrix microarray analysis to transcriptionally profile *Pdpk1*-null and Akt-inhibited CTL. Approximately 9400 annotated genes were expressed in CTL, and the impact of PDK1 loss or Akt inhibition was revealed by a decrease in the expression of less than 2%--3% of these transcripts and an increase in the expression of another 3%--4% ([Tables S1 and S2](#app3){ref-type="sec"} available online). We first looked for non-Akt-dependent, PDK1-regulated genes to understand why PDK1 but not Akt is required for IL-2-induced T cell proliferation. Selected genes controlled only by PDK1 but not Akt are highlighted in [Figure 4](#fig4){ref-type="fig"}A ([Table S3](#app3){ref-type="sec"}). Of particular note, PDK1 controlled expression of hexokinase 2, and hexokinases play a key role in effective uptake of exogenous glucose. We also used the NIAID DAVID website (<http://david.abcc.ncifcrf.gov>) to mine the gene ontology terms (GO) associated with the PDK1-controlled genes. We noted that GO terms related to lipid and cholesterol metabolism ([Figure 4](#fig4){ref-type="fig"}B) were strongly overrepresented in this list, indicating that PDK1 but not Akt controls lipid metabolism in CTL. We next interrogated the data to characterize genes coregulated by PDK1 and Akt. One clear result was that Akt inhibition or *Pdpk1* deletion resulted in the re-expression of Foxo target genes ([Figure 4](#fig4){ref-type="fig"}C), including the IL-7 receptor α subunit (CD127), the transcription factor Klf2, and its targets L-selectin (CD62L), CC motif chemokine receptor 7 (CCR7), and sphingosine-1-phosphate receptor 1 (S1P1) ([@bib7 bib13 bib27 bib36]). [Figures 4](#fig4){ref-type="fig"}D and 4E used quantitative polymerase chain reaction (PCR) analysis to verify these changes. IL-2 activation of Akt is dependent on PI3K p110δ ([Figure 1](#fig1){ref-type="fig"}), so it would be predicted that inactivating p110δ would also restore expression of the Foxo-regulated genes in CTL. Indeed, IL-2-cultured CTL expressing catalytically inactive p110δ (*Pik3cd*^D910A^) maintained high expression of Klf2 and CD62L, CCR7, and S1P1 ([Figure 4](#fig4){ref-type="fig"}F). Moreover, treatment of CTL with the p110δ inhibitor IC87114 drove re-expression of Klf2 and its targets ([Figure 4](#fig4){ref-type="fig"}G). Inhibition of PI3K p110δ and Akt Reprograms CTL Trafficking {#sec2.5} ----------------------------------------------------------- Deletion of *Pdpk1* or the inhibition of Akt caused CTL to re-express CD62L and CCR7 ([Figures 4](#fig4){ref-type="fig"}C--4E). CD62L controls T cell adhesion to the endothelium of high endothelial venules and CCR7 directs migration of T cells into lymphoid organs. Moreover, the loss of CD62L and CCR7 by CTL is part of the program that directs the trafficking of CTL away from lymphoid tissue. The re-expression of CD62L and CCR7 in CTL deleted of *Pdpk1* or exposed to AktI could argue that inhibition of Akt might reprogram the trafficking of CTL and restore their ability to transmigrate from the blood into secondary lymphoid tissue. In this context, strong activation of PI3K and Akt is required and sufficient to downregulate the ability of naive T cells to home to secondary lymphoid tissue in vivo ([@bib14 bib48]). It is also known that p110δ activity is required for T cells to exit lymphoid tissue and home to sites of infection ([@bib30]). T cells thus appear to need to activate Akt to exit lymphoid tissues. However, it is not known whether inhibition of Akt is sufficient to reprogram the ability of CTL to home to lymphoid tissue. In this context, inhibition of Akt had pleiotropic effects on the expression of multiple adhesion molecules and chemokine receptors by CTL and was not limited to controlling CD62L and CCR7 expression ([Figure 5](#fig5){ref-type="fig"}A). To directly test the impact of Akt inhibition on CTL trafficking, in vivo adoptive transfer experiments were performed comparing the ability of wild-type CTL or CTL pretreated with either AktI or the p110δ inhibitor IC87114 to home from the blood to lymphoid tissue. Strikingly, CTL pretreated with either AktI or IC87114 showed strong preferential homing to lymph nodes and spleen compared to wild-type CTL ([Figures 5](#fig5){ref-type="fig"}B and 5C). Sustained activation of Akt is thus required to maintain the trafficking characteristics of CTL: Loss of p110δ catalytic activity or Akt inhibition reprograms the ability of CTL to traffic to lymphoid tissue. To further explore the in vivo role of Akt in CD8 T cell trafficking responses, we compared in vivo immune responses of wild-type and *Pdpk1^K465E/K465E^* T cells. The K465E mutation creates a PDK1 molecule with a mutated PH domain that cannot bind PI(3,4,5)P~3~ and can support only a low amount of Akt activity ([@bib48]). [Figure 5](#fig5){ref-type="fig"}D shows that both wild-type and K465E T cells proliferated in vivo in response to immunization with cognate antigen and lipopolysaccharide (LPS). However, K465E T cells failed to downregulate CD62L ([Figure 5](#fig5){ref-type="fig"}E) and were retained in lymphoid tissue ([Figure 5](#fig5){ref-type="fig"}F) compared to control effector CD8^+^ T cells. Akt Is Necessary to Induce and Sustain Expression of CTL Effector Molecules {#sec2.6} --------------------------------------------------------------------------- A key role for IL-2 in CD8^+^ T cells is to promote effector differentiation of CTL by driving expression of cytolytic effector molecules and by controlling the cytokine receptor profile of CTL ([@bib21 bib38]). It was therefore striking that inhibition of Akt caused a decrease in the expression of mRNA encoding several molecules critical for CTL effector function ([Figure 6](#fig6){ref-type="fig"}A), notably multiple granzymes, Fas ligand, the cytolytic effector perforin, and interferon gamma (IFN-γ). It was equally notable that Akt inhibition had a considerable impact on the transcription of key cytokine receptors. Akt inhibition thus decreased expression of mRNA encoding IL-12 receptor β-chains while simultaneously increasing expression of mRNA encoding IL-6 receptors and CD27 (TNFRSF7), the CD70 coreceptor ([Figure 6](#fig6){ref-type="fig"}B). The role of Akt in perforin and IFN-γ expression was verified by real-time PCR analysis, ELISA, and immunoblot analysis ([Figure 6](#fig6){ref-type="fig"}C). Akt inhibition thus caused loss of IFN-γ mRNA expression and prevented IFN-γ protein production. Similar results were obtained after the deletion of *Pdpk1* from CTL ([Figure 6](#fig6){ref-type="fig"}D), providing genetic evidence that PDK1-Akt signaling sustains perforin and IFN-γ protein expression in CTL. Akt is also rapidly activated in CD8^+^ T cells in response to triggering of the TCR. Akt is not required for TCR-induced metabolic programs ([Figure 3](#fig3){ref-type="fig"}F), but is it necessary for the TCR to initiate expression of CTL effector molecules? In these experiments we focused on the induction of IFN-γ because this cytokine can be induced rapidly in response to triggering of the TCR in both naive and effector CD8^+^ T cells. Moreover, it has been shown that p110δ is required for TCR-induced IFN-γ production during immune activation but the role of Akt has not been addressed ([@bib30 bib46]). We found that inhibition of Akt suppressed TCR induction of IFN-γ mRNA and protein in both naive CD8^+^ T cells and effector CTL ([Figures 7](#fig7){ref-type="fig"}A and 7B). This reflects that Akt activity is important for IFN-γ gene transcription. [Figure 7](#fig7){ref-type="fig"}C thus shows that the loss of Akt activity reduced the amount of RNA polymerase II (Pol II) recruitment to the IFN-γ transcription start site and distal exons. Why is Akt needed for IFN-γ production? We considered the possibility that Akt controlled the activity of mTOR (mammalian target of rapamycin), which is known to control the expression of T effector molecules ([@bib41]). However, although *Pdpk1* deletion prevented phosphorylation of the mTOR target S6 ([Figure 2](#fig2){ref-type="fig"}B), inhibition of Akt only weakly suppressed the phosphorylation of S6 and S6K1 ([Figure S1](#app3){ref-type="sec"}). Akt activity is thus dispensable for mTOR activation in CTL. Accordingly, an inability to activate mTOR does not explain why Akt is required for IFN-γ production. However, an important insight as to a possible mechanism for Akt control of IFN-γ production came from experiments with *Pdpk1^K465E/K465E^* T cells that support only a low level of Akt activity. These T cells have a selective defect in the phosphorylation and inactivation of Foxo family transcription factors and fail to downregulate expression of Foxo gene targets ([@bib48]). Foxos can both transactivate and repress gene expression ([@bib13]), and in this context, TCR-induced production of IFN-γ is severely attenuated in *Pdpk1^K465E/K465E^* T cells ([Figure 7](#fig7){ref-type="fig"}D). This is consistent with a model whereby the defects in IFN-γ production caused by inhibition of Akt activity in CTL might be caused by relocation of Foxo transcription factors in the nuclei of T cells that then repress the IFN-γ gene. To assess whether the restoration of Foxos to the nuclei of TCR-activated T cells would impact on IFN-γ gene expression, we examined IFN-γ production in T cells expressing a GFP-tagged Foxo3a mutant with alanine substitutions at its Akt substrate sequences, T32, S252, and S314 (FoxoAAA). This phospho mutant could restore Foxo transcriptional function in cells expressing active Akt. [Figures 7](#fig7){ref-type="fig"}E and 7F show that activated T cells expressing the FoxoAAA mutant could not produce IFN-γ in response to peptide triggering of TCR complexes. Akt-regulated nuclear export of Foxo transcription factors is thus required for IFN-γ production. Discussion {#sec3} ========== The objective of the present study was to gain a more in-depth understanding of the function of PDK1 and Akt in the context of TCR and IL-2 signal transduction in CD8^+^ T cells. The accepted view of Akt is that it controls T cell metabolism. However, the data herein showed that Akt is not essential for CD8^+^ T cell metabolism, survival, or proliferation. Nevertheless, Akt does have a critical role in CTL function. Akt permits TCR and IL-2 signaling to maintain the expression of cytolytic effector molecules and to determine the repertoire of cytokine receptors, adhesion molecules, chemokine receptors, and effector molecules that distinguish effector and memory CTL populations. The present results thus force a shift in the accepted paradigm that Akt functions as an obligate controller of T cell metabolism. Rather, Akt simultaneously induces and represses expression of key genes to create an effector CTL, with the Foxo transcription factors being at the center of this process. The TCR and IL-2 control the metabolic programs of CTL by driving glucose uptake, amino acid uptake, and protein synthesis. The current results show that this metabolic role is not mediated by Akt but is controlled by PDK1. This kinase phosphorylates and activates many AGC family serine kinases including Akt, PKCs, S6Ks, RSK, and the SGKs ([@bib5 bib49]). These all share similar substrate specificities and there are numerous examples of redundant functions between different family members ([@bib23 bib26]). Importantly, the ability of PDK1 to phosphorylate the RSKs and the SGKs is independent of PI3K signaling ([@bib6 bib48]). Accordingly, PDK1 loss is more global than inhibition of PI(3,4,5)P~3~ production for terminating AGC kinase activity. The fact that glucose uptake and proliferation of CTL is PDK1 dependent rather than Akt dependent thus reveals that there is redundancy between AGC family kinases in CD8^+^ T cells. These data also indicate that PI3K and Akt do not have obligatory functions as metabolic switches in lymphocytes because other molecules can assume this role. These include other PDK1-controlled molecules, but it is also noteworthy that deletion of the serine-threonine master kinase LKB1 causes death of activated T cells ([@bib47]). There is also evidence that PIM kinases are important regulators of T cell survival ([@bib15]). The control of T cell metabolism may thus be determined by multiple kinases. Akt may not control CTL metabolism but the data herein shown that Akt does control a fundamental part of the CTL transcriptional program. Here it is relevant that cytotoxic T cells have two fates: they continue to become terminally differentiated "exhausted" effector T cells destined to die during the contraction phase of the immune response or a few may divert to become memory T cells ([@bib1]). TCR triggering initiates CTL differentiation and sustained IL-2 signaling promotes effector CTL differentiation. The present data now show that the ability of these receptors to sustain Akt activity is needed to maintain the transcriptional program of CTL. In particular, Akt signaling maintained CTL effector function and controlled expression of multiple biomarkers that distinguish central memory T cells from effector T cells. The loss of Akt activity thus reprogrammed CTL to assume a "memory-naive" phenotype by synchronizing a decrease in the expression of cytolytic effector molecules such as perforin and IFN-γ while causing an increase in expression of genes encoding the cytokine receptors for IL-6 and IL-7 along with tumor necrosis factor receptor family member CD27. These latter molecules all play key roles in the homeostasis and survival of long-lived memory CD8^+^ T cells ([@bib8 bib19 bib24]). Akt also controlled the repertoire of chemokine and adhesion molecules expressed by T cells that define whether these cells can traffic to secondary lymphoid tissue and progress to memory cells. Importantly, Akt inhibition allowed CTL to reacquire the ability to home to secondary lymphoid tissue. Collectively, these results reveal that sustained Akt signaling is required to maintain CTL effector function; the loss or reduction of Akt signaling does not impact T cell survival or proliferation but causes differentiated CTL to transcriptionally reprogram from an effector to a memory phenotype. Hence the simple view of Akt as a metabolic regulator is likely to be incorrect; rather, in CTL the role of Akt is to control the transcriptional programs that direct effector versus memory T cell fate. In particular, the data regarding Akt and T cell migration afford an explanation for the defects in the ability of antigen-primed PI3K p110δ-deficient T cells in vivo to migrate to antigenic sites ([@bib22 bib30]). The data also argue that the in vivo role of IL-2 to sustain effector CTL function may rely on Akt-dependent expression of differentiation and trafficking molecules. Finally, an emerging concept in T cell biology is that limiting T cell metabolism may accelerate the conversion of effector CTL into a memory subset ([@bib2 bib37 bib40]). This idea was based on studies linking mammalian target of rapamycin (mTOR) and adenosine monophosphate-activated protein kinase (AMPK) to the production of memory T cells during an immune response ([@bib2 bib37 bib41]). AMPK and mTOR are known to regulate cell metabolism but it was not directly proven that the role of either enzyme was caused by effects on T cell metabolism. The role of Akt in CTL suggests an alternative possibility: enzymes that originally evolved to control cell metabolism have evolved the ability to control other aspects of T cell function and can do so independently of effects on T cell metabolism. Experimental Procedures {#sec4} ======================= Mice {#sec4.1} ---- C57BL/6GT(ROSA)26^tm9(creEsr1)Arte^ (Tamox*Cre*) mice were obtained from Taconic Artemis Pharmaceuticals. Mice carrying floxed *Pdpk1* alleles *Pdpk1^flΔneo^* (PDK1^fl^); OT1 and P14 TCR transgenic mice; P14 TCR transgenic mice carrying a knockin mutation for a substitution of lysine for glutamic acid at residue 465 in the PH domain of PDK1 (*Pdpk1^K465E^*) and *Pik3cd*^D910A^ mice containing a knockin mutation of PI3K wherein wild-type alleles of the p110δ catalytic subunit of PI3K were substituted with a point mutation (D910A), which is a catalytically inactive form of p110δ (p110δ^D910A^), have been described ([@bib26 bib28 bib39 bib48]). Mice were produced in the Biological Resource Unit at the University of Dundee in compliance with UK Home Office Animals (Scientific Procedures) Act 1986 guidelines. Cell Culture {#sec4.2} ------------ CD8^+^ T cells were isolated from spleens and/or lymph nodes with an AutoMACs magnetic cell sorter (Miltenyi Biotec). Lymphocytes were suspended in RPMI 1640 plus 10% heat-inactivated fetal calf serum (FCS), penicillin, streptomycin (GIBCO), and 50 μM β-mercaptoethanol (β-ME) (Sigma) and activated for 48 hr in either 0.5 μg/ml CD3 antibody (2C11) or 100 ng/ml LCMV TCR-specific peptide gp33-41, KAVYNFATM. Cells were then washed free of activating agent and thereafter maintained in 20 ng/ml IL-2 at a density of approximately 0.3 × 10^6^/ml at 37°C, 5% CO~2~. 4-hydroxytamoxifen (4OHT, Sigma) was used at 0.6 μM. AktI-1/2 (AktI) (Calbiochem) was used at 1 μM, the p110δ inhibitor IC87114 (made in-house) and LY294002 (Promega) were used at 10 μM. Where indicated cells were cultured in media of defined glucose concentrations with glucose-free RPMI containing L-glutamine (GIBCO) with 10% dialyzed FCS (GIBCO), penicillin, streptomycin, 50 μM β-ME (glucose-free media) plus D(+)glucose (Sigma). *Pdpk1*-deleted CTL were prepared by activating *Pdpk1*^fl/fl^ Tamox*Cre* splenocytes with 2C11 for 2 days and then culturing the cells in IL-2 for 5 days, with 0.6 μM 4OHT being added for the final 72 hr of culture. Flow Cytometry {#sec4.3} -------------- Cells were labeled in RPMI plus 0.5% FCS with Phycoerythrin-conjugated anti-CD62L, Alexa750-conjugated anti-CD4, PerCP.Cy5.5-conjugated anti-TCR Vα2, allophycocyanin-conjugated anti-CD45.2, Horizon V450-conjugated anti-CD45.1, and phycoerythrin.Cy7-conjugated anti-CD8. For FoxoAAA, IFN-γ double-staining cells were stimulated in the presence of 3 μg/ml Golgiplug (BD), fixed and permeabilized as per manufacturer\'s instructions (eBioscience 00-8222-49, 00-8333-56), and then stained with Alexa488-conjugated anti-GFP (Invitrogen) plus allophycocyanin-conjugated anti-IFN-γ (BD Biosciences). Data were acquired on a FACS Calibur (BD Biosciences) and analyzed with FlowJo software (Treestar). Viable cells were gated according to their forward scatter (FSC) and side scatter (SSC) profiles. GFP^+/−^ cell isolation was performed with a Vantage cell sorter (BD Biosciences). Immunoblot Analysis {#sec4.4} ------------------- Cells were lysed on ice in HEPES lysis buffer: 100 mM HEPES (pH 7.4), 150 mM NaCl, 20 mM NaF, 20 mM iodoacetamide, 2 mM EDTA, 1% NP-40, 1 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, 40 mM β-glycerophosphate (Sigma), and protease inhibitors (Roche). Lysates were centrifuged (4°C, 1600 × g, 15 min) and then samples separated via sodium dodecyl sulfate 4%--12% polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. Blots were probed for pT24 Foxo1 + p32Foxo3a; pT308 and total Akt; pT235 + pT236 and total small ribosomal subunit S6; pS389, pS421 + pS424, and total S6K; PKCθ (all Cell Signaling Technologies); pS227 p90RSK (Santa-Cruz Biotechnology); total Foxo3a (made in house); PDK1 (Upstate); or perforin (gift of G. Griffith). Glucose and Phenylalanine Uptake {#sec4.5} -------------------------------- 1 × 10^6^ cells were suspended in 500 μl glucose-free media for 10 min. 500 μl glucose-free media containing 1 μCi/ml 2-Deoxy-D-\[1-^3^H\]glucose (\[^3^H\] 2DOG) (GE healthcare) was then added and the cells incubated for a further 10 min. Cells were pelleted, washed in PBS, and then lysed in water. Lysate ^3^H content was then measured via liquid scintillation counting. For phenylalanine uptake, the assay was performed essentially as for glucose uptake, except that assays were performed in HBSS, 50 μM β-ME, 1 × MEM vitamins (GIBCO), with 10% dialyzed FCS. \[^3^H\]Phe (GE Healthcare) was added and cells incubated for 1 hr prior to pelleting. Real-Time PCR {#sec4.6} ------------- RNA was extracted from cells with the RNeasy RNA purification minikit (QIAGEN). Reverse-transcription PCR was performed with qScript cDNA synthesis kit (Quanta). Real-time PCR was performed with iQ SYBR Green detection chemistry (BioRad) on an iCycler (BioRad). Relative mRNA levels of genes of interest were normalized to hypoxanthine guanine phosphoribosyltransferase (HPRT). Primers are detailed in the [Supplemental Experimental Procedures](#app3){ref-type="sec"}. ChIP {#sec4.7} ---- Real-time PCR-based chromatin immunoprecipitation (ChIP) analysis to measure RNA Pol II binding to the IFN-γ locus was performed described ([@bib11]) with minor modifications. Chromatin was immunoprecipitated with anti-Pol II (Santa-Cruz Biotechnology) or normal rabbit IgG (Cell Signaling) from 5 × 10^6^ cells in presence of 0.2 mg/ml BSA. ChIP Grade Protein G Magnetic Beads (Cell Signaling) were used to collect the immunocomplexes. Chromatin was purified via a NucleoSpin Extract II kit (Macherey-Nagel) and resuspended in TE buffer. Real-time PCR was performed in a Biorad iQ5 with Perfecta SYBR green FastMix for iQ (Quanta BioSciences). Primers are detailed in the [Supplemental Experimental Procedures](#app3){ref-type="sec"}. Retroviral Transduction {#sec4.8} ----------------------- The production of FoxoAAA-GFP (Foxo3a T32A S252A S314A) and control retrovirus has been described previously ([@bib48]). The retroviral infection protocol was performed as described previously ([@bib48]). IFN-γ Quantification {#sec4.9} -------------------- IFN-γ was assayed in culture supernatants with the femto-HS high-sensitivity IFN-γ ELISA kit (eBiosciences). Microarray Analysis {#sec4.10} ------------------- Microarray analysis was carried out by the Finnish DNA Microarray Centre at the Centre for Biotechnology, Turku, Finland via 430\_2.0 mouse expression arrays (Affymetrix) and the manufacturer\'s recommended protocol. Statistically significant differences in gene expression were identified with Multiple Experiment Viewer v4.3 ([@bib43]). Overrepresented gene ontology terms were identified with the NIAID DAVID website (<http://www.david.abcc.ncifcrf.gov>). Full details are given in [Supplemental Experimental Procedures](#app3){ref-type="sec"}. Statistical Analyses {#sec4.11} -------------------- With the exception of the microarray data, statistical analyses were performed with Prism 4.00 for Macintosh (GraphPad). A nonparametric Mann Whitney test was used where the number of experiments performed was not sufficient to prove normal distribution. Statistically significant results (p \< 0.05) are indicated by an asterisk. Accession Numbers {#app1} ================= The microarray data are available in the Gene Expression Omnibus (GEO) database (<http://www.ncbi.nlm.nih.gov/gds>) under the accession number GSE26290. Supplemental Information {#app3} ======================== Document S1. Supplemental Experimental Procedures and One FigureTable S1. Genes Changing Significantly in Expression in CTL after Pdpk1 DeletionTable S2. Genes Changing Significantly in Expression in CTL after Akt InhibitionTable S3. Genes Changing Significantly in Expression in CTL after Pdpk1 Deletion but Not After Akt Inhibition We thank members of the Biological Services Unit; R. Clarke of the Flow Cytometry Facility; and members of the D.A.C. laboratory for critical reading of the manuscript. We thank D. Alessi at the University of Dundee for providing mice carrying PDK1 floxed alleles and the Finnish DNA Microarray Centre at the Centre for Biotechnology (Turku, Finland) for the microarray analysis. This project was supported by a Wellcome Trust Principal Research Fellowship and Program grant (065975/Z/01/A). A.N.M was supported by a Wellcome Trust PhD studentship. Supplemental Information includes Supplemental Experimental Procedures, one figure, and three tables and can be found with this article online at [doi:10.1016/j.immuni.2011.01.012](10.1016/j.immuni.2011.01.012). ::: {#fig1 .fig} Figure 1 ::: {.caption} ###### Akt Is Not Necessary for IL-2-Driven Survival or Proliferation in CTL \(A) Immunoblot analysis of Akt T308 and Foxo1 T24, Foxo3a T32 phosphorylation in wild-type (C57/BL6) and p110δ^D910A^-expressing (*Pik3cd*^D910A^) splenic T cells activated with 2C11 for 48 hr then cultured in IL-2 for 3 days. \(B) Wild-type splenic T cells were activated with 2C11 for 48 hr then cultured in IL-2 for 5 days, with some cells being deprived of IL-2 for the last 24 hr of culture. Data show fold change in cell number over the last 24 hr of culture and the cell FSC and SSC profiles at the end of the experiment. Gates indicate dead (left) and viable (right) cell populations. \(C) Splenic T cells from wild-type and p110δ^D910A^-expressing mice (*Pik3cd*^D910A^) were cultured as in (A). Data show fold change in cell number over the last 48 hr of culture. \(D) Western blot analysis of Akt T308 phosphorylation in P14 T cells activated with gp33 for 48 hr then cultured in IL-2 for 5 days, with IC87114 or LY294002 being added for the last 48 hr of culture. \(E) P14 T cells were activated with gp33 for 48 hr then cultured in IL-2 for 5 days, with IC87114 added for the last 48 hr of culture. Alternatively, IL-2 was removed from the culture for the last 48 hr. Data show fold change in cell number over last 48 hr of culture. (F and G) Wild-type splenic T cells activated with 2C11 for 48 hr cultured in IL-2 for 5 days, with AktI being added for the last 48 hr of culture. Data show (F) immunoblot of Akt T308 and Foxo1 T24, Foxo3a T32 phosphorylations and (G) FSC and SSC profiles of cells at the end of the culture plus fold change in cell number during the last 48 hr of culture. Data are representative of 3 (A, C--F), 7 (B), or 11 (G) experiments; mean ± SEM of biological replicates. :::  ::: ::: {#fig2 .fig} Figure 2 ::: {.caption} ###### PDK1 Maintains IL-2-Driven Proliferation (A and B) Immunoblot analysis of (A) PDK1 expression and (B) phosphorylation of Akt, Foxo1 and Foxo3a, RSK, and small ribosomal subunit S6 in *Pdpk11^fl/fl\ /tm9(creEsr1)Arte^* (*Pdpk1^fl/fl^* Tamox*Cre*) and *Pdpk1^+/+/\ tm9(creEsr1)Arte^* (*Pdpk1^WT/WT^* Tamox*Cre*) T cells activated with 2C11 for 2 days and then cultured in IL-2 for 5 days, supplemented as indicated with 4OHT for the final 72 hr of culture. (C and D) Fold change in cell number over the last 48 hr of culture (C) and FSC and SSC profiles (D) of 4OHT-treated cells from (A), gates indicate dead (left) and viable (right) cell populations. Data are representative of three (A, B) or five (C, D) experiments; mean ± SEM of biological replicates. :::  ::: ::: {#fig3 .fig} Figure 3 ::: {.caption} ###### IL-2 and PDK1 but Not Akt Sustain Glucose Uptake in CTL \(A) 2-deoxyglucose (2DOG) uptake by wild-type splenic T cells activated with 2C11 for 48 hr and then cultured in IL-2 for 5 days, with some cells being deprived of IL-2 for the last 24 hr of culture. \(B) Wild-type splenocytes were activated with 2C11 and IL-2 for 2 days and then cultured in IL-2 in standard media for 72 hr. For the final 48 hr of culture, cells were placed in media containing indicated glucose concentrations and the fold change in cell number assessed. \(C) Relative 2DOG uptake in wild-type splenic T cells activated with 2C11 for 48 hr then cultured in IL-2 for an additional 5 days, with 1 μM AktI being added for the last 48 hr of culture. (D and E) Relative 2DOG uptake (D) and phenylalanine uptake (E) by Cre control (*Pdpk1^+/+/\ tm9(creEsr1)Arte^* +4OHT) and *Pdpk1*-null (*Pdpk1^fl/fl/\ tm9(creEsr1)Arte^* +4OHT) CTL. \(F) Naive P14 T cells were maintained in IL-7 or stimulated for 18 hr with gp33 in the presence or absence of AktI or IC87114 and then their ability to take up 2DOG was assayed. Data are representative of 3 (B, D, F), 5 (A), or 11 (C) experiments; mean ± SEM of biological replicates, except (F), representative data from 2 experiments; mean ± SD of technical replicates. :::  ::: ::: {#fig4 .fig} Figure 4 ::: {.caption} ###### PDK1 and Akt Regulation of the CTL Transcriptional Program *Pdpk1*-null (*Pdpk1^fl/fl/\ tm9(creEsr1)Arte^* +4OHT) and Cre control (*Pdpk1^+/+/\ tm9(creEsr1)Arte^* +4OHT) CTL were generated and their transcriptional profiles compared via microarray. Analysis was performed with cells from three age- and sex-matched mice for each genotype. In a separate experiment, three wild-type (C57/BL6) age- and sex-matched mice were used to generate CTL by activating wild-type splenocytes with 2C11 for 48 hr with IL-2 then culturing the cells in IL-2 for a further 5 days, with 1 μM AktI being added to half of each culture for the last 48 hr. Comparison of the transcriptional profile in control versus AktI-treated cells was then performed by microarray. \(A) Heat map showing the relative normalized expression of selected genes changing in expression after *Pdpk1* loss but not Akt inhibition. \(B) Gene ontology analysis of genes changing in expression after *Pdpk1* loss but not Akt inhibition. \(C) Heat maps showing the relative normalized expression of Foxo-regulated genes that are significantly different in expression between untreated and AktI-treated CTL and between Cre control and *Pdpk1*-null CTL. (D--G) Real-time PCR measurements of relative gene expression levels in (D) Cre control versus *Pdpk1*-null CTL, (E) untreated versus 48 hr AktI-treated wild-type CTL, (F) wild-type versus PI3K p110δ^D910A^-expressing CTL (*Pik3cd*^D910A^), and (G) untreated versus 48 hr IC87114-treated CTL. Data representative of a minimum of three experiments; mean ± SEM of biological replicates. :::  ::: ::: {#fig5 .fig} Figure 5 ::: {.caption} ###### Suppression of PI3K p110δ and Akt Reprograms Lymph Node Homing of CTL \(A) Heat map showing the relative normalized expression of selected genes that are significantly different in expression between untreated and AktI-treated CTL, as determined by microarray. \(B) Lymph node P14 T cells were activated for 2 days with cognate peptide and then cultured for 5 days in IL-2, with IC87114 being added to half of each culture for the last 48 hr. Cells were then labeled with carboxyfluorescein succinimidyl ester (CFSE) or 5-(and-6) (((4chloromethyl)benzoyl) amino) tetramethylrhodamine (CMTMR) and mixed at a ratio of 1:1 before being injected into C57/BL6 host mice. Values indicate recovery of inhibitor-treated or untreated cells as a percentage of the total recovered transferred cells from the blood and lymph nodes 18 hr after transfer. Each dot indicates a mouse; horizontal bars indicate mean. \(C) As for (B) with cells being treated with AktI rather than IC87114 for the last 48 hr of culture. (D--F) CD8^+^ T cells expressing Vα2Vβ5 TCR were purified from *Pdpk1^WT/WT^* OT1 (CD45.1^+/+^) and *Pdpk1^K465E/K465E^* OT1 (CD45.1^+^ CD45.2^+^) mice, mixed at a 1:1 ratio, and stained with CFSE. 5 × 10^6^ mixed T cells were injected into the tail vein of C57/BL6 mice (CD45.2^+/+^). 24 hr later the host mice were injected intraperitoneally with either 25 mg LPS or LPS plus 40 mg of SIINFEKL peptide. The ratio of *Pdpk1^WT/WT^* OT1 and *Pdpk1^K465E/K465E^* OT1 donor cells in lymph nodes and proliferation (CFSE dilution) was analyzed after 6 days, with CD62L expression assayed after 2, 4, and 6 days. :::  ::: ::: {#fig6 .fig} Figure 6 ::: {.caption} ###### Suppression of PI3K p110δ and Akt Alters the Expression of Key CTL Effector Molecules (A and B) Heat maps showing the relative normalized expression of selected genes that are significantly different in expression between untreated and AktI-treated CTL, as determined by microarray. \(C) Wild-type splenocytes were activated with 2C11 for 48 hr with IL-2 and then cultured in IL-2 for a further 5 days, with half the cells being placed in AktI for the last 48 hr of culture. Shown is IFN-γ release by 1 million cells in 6 hr, relative IFN-γ mRNA content, immunoblot analysis of total perforin protein content, and perforin mRNA content of cells cultured with and without AktI. \(D) *Pdpk1*-null (*Pdpk1^fl/fl/\ tm9(creEsr1)Arte^* +4OHT) and Cre control (*Pdpk1^+/+/\ tm9(creEsr1)Arte^* +4OHT) CTL were generated. Shown are IFN-γ release by 1 million cells in 6 hr, relative IFN-γ mRNA content, immunoblot analysis of total perforin protein content, and perforin mRNA content of Cre control versus *Pdpk1*-null CTL. Data representative of a minimum of three experiments; mean ± SEM of biological replicates. :::  ::: ::: {#fig7 .fig} Figure 7 ::: {.caption} ###### PDK1 Controls Antigen Receptor-Driven IFN-γ Expression via a Foxo-Mediated Mechanism \(A) Naive P14 lymph node T cells were stimulated for 4 hr with LCMV peptide gp33 in the presence or absence of AktI and the IFN-γ produced assayed by ELISA. \(B) P14 T cells were activated for 48 hr with gp33 and then cultured for 3 days in IL-2. Left: Cells were then stimulated for 4 hr with 100 ng/ml gp33 in the presence of AktI or IC87114 and IFN-γ production assayed by flow cytometry. Right: Alternatively, cells were stimulated for 4 hr with 125 ng/ml gp33 in the presence of AktI and IFN-γ production assayed by ELISA. \(C) P14 T cells were activated for 48 hr with gp33 and then cultured for 3 days in IL-2. Cells were stimulated for 3 hr with 62.5 ng/ml gp33 in the presence or absence of AktI. ChIP was performed with anti-PolII, and the changes in PolII binding to (left) the IFN-γ-proximal promoter region and (right) the IFN-γ fourth exon were quantified by real-time PCR. Data are normalized to input DNA amounts and plotted as fold over the values for PolII binding to the HPRT-proximal promoter. \(D) P14 TCR transgenic T cells from lymph nodes of *Pdpk1^WT/WT^* and *Pdpk1^K465E/K465E^* P14 transgenic mice were activated for 48 hr with gp33 and then cultured for 3 days in IL-2. Cells were then stimulated for 4 hr with gp33 and IFN-γ produced assayed by ELISA. (E and F) P14 T cells were activated for 24 hr with gp33 and then spinfected with retrovirus encoding either GFP or GFP-tagged Foxo3a T32A S252A S314A (FoxAAA). Three days after infection, cells were (E) stimulated for 4 hr with 100 ng/ml gp33 in the presence or absence of AktI or IC87114 and then stained for intracellular IFN-γ or (F) sorted into GFP^+/−^ populations, cultured for a further 24 hr, and then stimulated for 4 hr with 100 ng/ml gp33 and IFN-γ production assayed by ELISA. Data are representative of a minimum of two (E, F) or three (A--C, D) experiments; mean ± SEM of biological replicates. :::  :::

PubMed Central

2024-06-05T04:04:19.322927

2011-2-25

PMC3052434

pmc Introduction {#s0005} ============ 1 The distinctive class of T cells, known as invariant Natural Killer T (*i*NKT) cells, display characteristics of both T cells and NK cells and play a crucial role in diverse immune responses and other pathologic conditions.[@b0005] The synthetic glycolipid α-galactosyl ceramide (α-GalCer),[@b0010] known as KRN7000 (**1**) ([Fig. 1](#f0005){ref-type="fig"}), is a powerful agonist, which when presented by CD1d, activates *i*NKT cells to release diverse cytokines, including both Th1- and Th2-cytokines.[@b0015] Once stimulated, *i*NKT cells also activate other cells such as dendritic cells, T cells and B cells.[@b0035] It is believed that the release of proinflammatory Th1 cytokines such as interferon-γ (INF-γ) may contribute to antitumour and antimicrobial functions while that of immunomodulatory Th2-cytokines such as interleukin 4 (IL-4) may help alleviate autoimmune diseases[@b0040] such as multiple sclerosis[@b0060] (MS) and arthritis.[@b0065] Maintaining the right balance between Th1- and Th2-cytokines is of utmost importance as over activation of Th1 cells or suppression of Th2 ones can lead to autoimmune diseases.[@b0070]Figure 1CD1d agonist KRN7000, OCH, and analogues. Moreover, skewing of the cytokine release profile towards a Th2 one can help in the treatment of autoimmune and inflammatory conditions.[@b0075], [@b0080] For example, compound **5**, commonly referred to as OCH ([Fig. 1](#f0005){ref-type="fig"}), has been shown to protect mice against experimental encephalomyelitis (an animal model for MS) by favouring the release of the Th2-cytokine IL-4 and suppressing the myelin antigen-specific Th1 responses.[@b0060], [@b0085] Crystal structures of both mouse[@b0090] and human CD1d[@b0095] have identified the antigen-binding site as consisting of two channels or pockets; A′ and F′, lined with hydrophobic residues. While the A′ channel can accommodate an alkyl chain consisting of up to 26 carbons (the acyl chain in α-GalCer), the F′ channel can accommodate an alkyl chain of 18 carbons (the sphingosine chain in α-GalCer).[@b0095], [@b0100] The stability of the bound glycolipid/CD1d complex and in turn the binding affinity of the latter to T cell receptors (TCR) are believed to largely influence the immunological response.[@b0060], [@b0100] The apparent skewing towards a Th2 response in the case of OCH (**5**) is allegedly due to a less stable bound CD1d complex resulting in a less prolonged TCR stimulation.[@b0060], [@b0100], [@b0105] α-Galactosyl ceramides with variations in the acyl chain have been extensively studied, and have also exhibited similar effects on the cytokine release profile.[@b0075] However, the properties of analogues of α-GalCer with variations in the sphingosine chain, such as OCH (**5**) have yet to be fully explored. α-GalCer and its counterparts have proved to be and remain invaluable tools in understanding the functioning of CD1d and NKT cells in a wide range of immune responses. Specifically, compounds such as OCH (**5**), that are able to alter the polarisation of Th1 or Th2, have potential therapeutic values for certain diseases. As such, we have synthesised compounds **2**, **3** and **4** ([Fig. 1](#f0005){ref-type="fig"}), with varying phytosphingosine chain lengths consisting of 9--15 carbons, to investigate their effect on the Th1/Th2 balance as well as to study their overall biological properties. Results and discussion {#s0010} ====================== 2 Various methods, including dihydroxylation reactions[@b0110] and Sharpless asymmetric epoxidation[@b0130] have been described in the literature for the synthesis of sphingolipids.[@b0150] Yet, the preparation of such compounds remains nontrivial. In contrast to many reported syntheses, the strategy employed by Lin et al.[@b0170] is quite concise and affords a relatively high yield of the final phytosphingosine from the readily available [d]{.smallcaps}-lyxose. Their method is particularly attractive as [d]{.smallcaps}-lyxose already possesses the required stereogenic centres and an S~N~2 displacement of a suitable leaving group at C-4 allows the introduction of the amine functionality in the molecule. It is noteworthy that Plettenburg et al.[@b0185] also used a similar strategy, involving conversion of their substrate to phytosphingosine via a Wittig condensation, in their reported synthesis. We have therefore adapted both these methods to synthesise our analogous sphingosines. We first embarked on the synthesis of the appropriate phosphonium salts required for the chain extending Wittig olefination reaction. This was easily achieved by refluxing triphenylphoshine with the corresponding alkyl halides in toluene overnight. For the C-9, C-12, and C-15 phytosphingosine chain lengths, 1-iodobutane, 1-bromooctane and 1-bromodecane were used, respectively. Once isolated, they were resuspended in THF and treated with *n*-BuLi at −78 °C to generate the Wittig ylids, as described by Plettenburg et al.[@b0185] Subsequently, the protected lyxose derivative **6**, which was synthesised as previously described,[@b0170] was condensed with the ylids to afford the desired olefins **7**, **8** and **9** as a mixture of *cis* and *trans* isomers ([Scheme 1](#f0025){ref-type="fig"}). In their preparation of the C-18 phytosphingosine, Lin et al.[@b0170] proceeded to the reduction of the double bond after the Wittig condensation via catalytic hydrogenation using palladium hydroxide (Pd(OH)~2~). However, in our hands, after several attempts the hydrogenation failed to go to completion after prolonged reaction periods. In an attempt to drive the reaction to completion, the catalyst was filtered off and replaced with fresh one. Although the reaction was eventually completed after 48 h, the trityl protecting group was also cleaved, thereby leaving the primary alcohol unprotected. An alternative strategy where the protecting groups would be removed prior to hydrogenation of the double bond was therefore envisaged. Hence, the remaining free secondary hydroxyl group was mesylated by reaction with methanesulfonyl chloride in dichloromethane. The mesyl group acts both as a temporary protecting group and a good leaving group for the following inversion of stereochemistry at this position. Removal of the trityl and acetonide protecting groups by acid treatment then yielded compounds **13**, **14** and **15** in quantitative yields. The reduction of the double bond by hydrogenation, catalysed by 5% palladium on barium sulphate proceeded smoothly to give compounds **16**, **17** and **18**. We substituted the Pd(OH)~2~ with palladium on barium sulphate because the latter is less active and is more suitable for use in the presence of the mesylate group. Finally, the amine functional group was introduced into the molecule by an S~N~2 displacement of the mesylate with sodium azide in DMF. Sphingosine analogues **19**, **20** and **21** were thus obtained in reasonable yields for further use as glycosyl acceptors.Scheme 1Reagents: (a) *n*-BuLi, THF, phosphonium salts (C~4~H~9~PPh~3~^+^I^−^, C~7~H~15~PPh~3~^+^Br^−^, C~10~H~21~PPh~3~^+^Br^−^; (b) MsCl, pyridine, CH~2~Cl~2~, quant; (c) HCl, MeOH/CH~2~Cl~2~; (d) H~2~, Pd--BaSO~4~, THF; (e) NaN~3~, DMF. Hence, following the standard conditions described previously[@b0190] and summarised in [Scheme 2](#f0030){ref-type="fig"}, the sphingosine acceptors **22**--**24** were synthesized. Consistent with our previous work, benzoate esters, rather than benzyl ethers, were adopted as protecting groups to circumvent the hydrogenolysis reaction.Scheme 2Reagents: (a) TBDPSCl, pyridine, quant; (b) BzCl, Pyr, 86%; (c) TBAF, THF, 80%. With respect to the glycosyl donor, we have previously successfully employed the bulky 4, 6-*O*-di-*tert*-butylsilylene (DTBS) group as α-directing in galactosylation donors.[@b0195] DTBS ensures the exclusive formation of an α-glycosidic linkage, irrespective of the nature of the acceptor, and remains the protecting group of choice in challenging glycosylation reactions. However with our glycosyl acceptors, we were inclined to adopt Gervay-Hague's rather simplified glycosylation strategy employing glycosyl iodides and promoted by tetrabutyl ammonium iodide (TBAI).[@b0200] Indeed, the latter's group recently reported both excellent stereoselectivity and yields for the synthesis of α-GalCer and other similar compounds.[@b0205], [@b0210] It is hypothesised that the α-selectivity is due to the TBAI-catalysed isomerization of the α-glycosyliodide to the β-anomer.[@b0215] This strategy represents a straightforward approach to the synthesis of our target compounds given that the donor can be easily obtained in large quantities in contrast to the multi-step preparation of other galactosyl donors. The per-*O*-tetramethylsilyl-α-[d]{.smallcaps}-galactopyranosyl iodide **25** was therefore generated by the reaction of the per-*O*-pentamethylsilyl-α-[d]{.smallcaps}-galactose with 1 equiv of iodotrimethylsilane[@b0215] and then added to the respective phytosphingosine acceptors **22**--**24** which were premixed with diisopropylethylamine (DIPEA) and TBAI ([Scheme 3](#f0035){ref-type="fig"}). After two days at room temperature, the solvent was evaporated and the TMS protecting groups were removed by treatment with an acidic resin in MeOH. The glycosylated compounds **26**--**28** were obtained as the α-anomer exclusively in good overall yields of over 60%. The formation of the desired α-linkage in compounds **26**--**28** was confirmed by the H-1 and C-1 signals in ^1^H and ^13^C NMR, respectively. Finally, Zemplen's deprotection of the benzoate protecting groups, followed by hydrogenation of the azido group in methanol provided the amines which were acylated with the fully saturated fatty acid, hexacosanoic acid. This was accomplished by reaction of the corresponding acid chloride with the free amine in a 1:1 mixture of THF and saturated sodium acetate solution. Glycosphingolipds (GSL) **2** (OCH9), **3** (OCH12) and **4** (OCH15) were obtained as white solids after concentration of the organic phase and purification of the residue by flash chromatography.Scheme 3Reagents and conditions: (a)TMSI, CH~2~Cl~2~, 0 °C, quant; (b) TBAI, DIPEA, **22**, **23** and **24**, benzene, rt; (c) Dowex 50WX8-200, MeOH, rt, 62% over two-steps; (d) NaOMe/MeOH, quantitative; (d) H~2~, Pd, MeOH, 80%; (e) C~25~H~51~COCl, THF/NaOAc (1:1), 71%. To evaluate the biological activity of compounds OCH9 (**2**), OCH12 (**3**) and OCH15 (**4**), human B cells expressing CD1d (C1R CD1d) were pulsed with different concentrations of lipids and incubated with a human *i*NKT cell line. *i*NKT cell activation (as determined by both IL-4 and IFN-γ secretion in the culture supernatant) elicited by OCH15 (**4**) was comparable to that obtained with α-GalCer **1** ([Fig. 2](#f0010){ref-type="fig"}). On the other hand, compounds OCH9 (**2**) and OCH12 (**3**) were found to be weaker stimulants of *i*NKT cells as less IL-4 and IFN-γ were observed. These data clearly indicate that the *i*NKT cells' activation by the GSLs OCH9 (**2**), OCH12 (**3**) and OCH15 (**4**) correlate with the length of their sphingosine chains, with the longest chain inducing the greatest cytokine response. However, unlike what was previously observed with mouse *i*NKT cells,[@b0230] the cytokine profile of human *i*NKT cells stimulated by OCH9 (**2**) and OCH12 (**3**) was not as strongly biased towards a Th2 response ([Fig. 2](#f0010){ref-type="fig"}A).[@b0100] Increasing amounts of IFN-γ were observed with increasing concentrations of OCH9 (**2**) and OCH12 (**3**), as opposed to what was observed with mouse *i*NKT cells ([Fig. 3](#f0015){ref-type="fig"}).[@b0230]Figure 2CIR-hCD1d cells were pulsed with GSL and used to stimulate *i*NKT cells. Supernatant was assayed for IL-4 (**A**) and IFN-γ (**B**) by ELISA.Figure 3A (upper and lower pannel) IL4/IFNγ and IFNγ/IL4 ratio (human *i*NKT cells), B (upper and lower pannel) IL4/IFNγ and IFNγ/IL4 ratio (mice *i*NKT cells); two mice per group were injected with 1 μg of lipid ip and sera was assayed at 2 h (IL4) or 24 h (IFNγ) by ELISA. We previously performed kinetic and affinity experiments to investigate the effect of the phytosphingosine chain length on the stability of the bound glycolipid/CD1d complex and on TCR binding affinity.[@b0100] We reported that shortening the phytosphingosine chain increased the rate of dissociation of the GSLs from hCD1d molecules and led to a decrease of the binding affinity of the *i*NKT TCR for hCD1d--GSL complexes.[@b0100] The increase in the *K*~d~ value as the phytosphingosine chain got shorter was attributed to both a decrease in *k*~on~ and an increase in *k*~off~. Furthermore, our results indicated an important role of the phytosphingosine chain in controlling the formation of the immunological synapse and *i*NKT cell activation.[@b0100] As a control, α-GalCer, OCH9, OCH12 and OCH15 were compared with C20:2 ([Fig. 4](#f0020){ref-type="fig"}), a compound known to induce a Th2 skewing in mice.[@b0075], [@b0080], While no apparent change in IL4/IFNγ ratio was observed between α-GalCer and the various GSLs with human *i*NKT cells ([Fig. 3](#f0015){ref-type="fig"}A, upper panel), a net increase was observed with mice *i*NKT cells ([Fig. 3](#f0015){ref-type="fig"}B, upper pannel). In mice C20:2 and the truncated GSLs OCH9, OCH12 and OCH15 favour the release of the Th2-cytokine IL4. Similarly, the IFNγ/IL4 ratio observed with α-GalCer with mice *i*NKT cells ([Fig. 3](#f0015){ref-type="fig"}B, lower pannel) demonstrates that the latter favours the release of the Th1 cytokine IFNγ, confirming that in mice, compounds C20:2, OCH9, OCH12 and OCH15 induce varying extents of Th2 skewing.Figure 4Analogue of α-GalCer inducing a Th2 skewing. Conclusions {#s0015} =========== 3 In conclusion we have developed a synthetic strategy for the successful syntheses of three truncated analogues of α-GalCer, adapted from examples present in the literature. ELISA assays of these compounds with hCD1d and human *i*NKT cells showed different results than those observed with mice *i*NKT cells. The truncated analogues OCH9 (**2**), OCH12 (**3**) and OCH15 (**4**) were found to be generally less potent than α-GalCer (**1**). Also there was no Th2 skewing observed in this present study with the shorter phytosphingosine chain lengths. Furthermore, control comparison experiments showed that GSLs OCH9 (**2**), OCH12 (**3**), OCH15 (**4**) and C20:2 cause a Th2 skewing in mice *i*NKT cells, thereby confirming the viability of our assays. Experimental {#s0020} ============ 4 ^1^H NMR spectra were recorded at 400 MHz or 300 MHz, using Bruker AMX 400, Bruker AV 400, Bruker AV 300 and Bruker AC 300 spectrometers. ^13^C NMR spectra were recorded at 100 MHz or 75 MHz, respectively, using Bruker AMX 400, Bruker AV 400, Bruker AV 300 and Bruker AC 300 spectrometers. Chemical shifts are reported as *δ* values (ppm) referenced to the following solvent signals: CHCl~3~, *δ*~H~ 7.26; CDCl~3~, *δ*~C~ 77.0; CH~3~OH, *δ*~H~ 3.34; CD~3~OD, *δ*~C~ 49.9. Quaternary carbons were reported only when observed. Optical rotations were measured at 23 °C and reported in degdm^−1^ g^−1^ cm^3^. HRMS were recorded on a Micromass LCT spectrometer using a lock mass incorporated into the mobile phase. All reagents were obtained from commercial sources, and were used without further purification unless stated otherwise. Anhydrous solvents were purchased from Sigma--Aldrich, UK, stored over 4 Å molecular sieves and under an Ar atmosphere. Analytical thin layer chromatography (TLC) was performed on aluminium plates pre-coated with Merck silica gel 60A F-254 as adsorbent. The developed plates were air dried, visualised by UV detection (at 254 nm) and/or stained with 5% phosphomolybdic acid in EtOH (MPA spray). Compounds were purified by flash column chromatography on Merck silica gel (particle size 40--63 lm mesh) or Fluka 60 (40--60 lm mesh) silica gel. General procedure for the synthesis of phosphonium salts (a) {#s0025} ------------------------------------------------------------ 4.1 A mixture of triphenyl phosphine (3 equiv) and the alkyl halide, 1-iodobutane, 1-bromoheptane and 1-bromodecane, respectively, (1 equiv) in toluene was refluxed overnight. The mixture was then allowed to cool and then filtered. The precipitate was washed with cold toluene and dried under vacuo to give the phosphonium salts which were used in the next step without further purification. General procedure for the synthesis of Wittig salts (ylids) and subsequent olefination (b) {#s0030} ------------------------------------------------------------------------------------------ 4.2 The phosphonium salt was resuspended in THF and cooled to −10 °C. *n*BuLi (2.5 M solution in hexanes, 2.8 equiv) was added dropwise and the reaction mixture was stirred for 30 min, before a solution of 6 (1 equiv) in dry THF was added. The resulting solution was stirred overnight at room temperature. Upon completion of the reaction, the reaction was quenched with MeOH, followed by 80% MeOH in H~2~O, and the mixture was extracted with heptane (4 × 20 mL). The combined organic layers were then washed with brine (30 mL), dried over anhydrous Na~2~SO~4~ and concentrated in vacuo. The residue was purified by purified by flash chromatography using hexanes/EtOAc (10:1) to give the desired products. (2*R*,3*S*,4*R*,5*E*/*Z*)-3,4-*O*-Isopropylidene-1-*O*-trityloxy-non-5-en-2-ol **7** {#s0035} ------------------------------------------------------------------------------------ 4.3 Prepared following general procedures (a) and (b) using triphenylphosphine (18.36 g, 70.0 mmol) and iodobutane (4.29 g, 23.3 mmol), *n*-BuLi (26.8 mL, 65.2 mmol) and **6** (10.10 g, 23.3 mmol) to afford compound **7** as a white solid in 82% yield as a mixture of *cis* and *trans* isomer in a 1:2.3 ratio (9.03 g, 19.1 mmol). ^1^H NMR (CDCl~3~): *δ* 7.19--7.49 (15H, m, Ar--H), 5.53--5.57 (2H, m, H-5, H-6), 4.90--4.94 (0.7 H, m, H-4^trans^), 4.45 (0.3H, dd, *J*~4,5~ = *J*~3,4~ = 7.4 Hz, H-4^cis^), 4.28--4.30 (0.3H, m, H-3^cis^), 4.22--4.25 (0.7H, m, H-3^trans^), 3.77--3.81 (0.7H, m, H-2^trans^), 3.69--3.73 (0.3H, m, H-2^cis^), 3.25 (0.3H, dd, *J*~1a,2~ = 5.1, *J*~1a,1b~ = 9.5 Hz, H-1a^cis^), 3.18--3.20 (0.7H, m, H-1a^trans^), 3.07--3.15 (1H, m, H-1b^trans^, H-1b^cis^), 1.75--2.06 (2H, m, H-7a^cis^, H7b^cis^, H-7a^trans^, H-7b^trans^), 1.58, 1.52, 1.39, 1.37 (6H, 4s, 4 × C(CH~3~)~2~), 1.22--1.32 (2H, m, CH~2~), 0.87 (3H, t, *J* = 7.3 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 144.1 (O~2~C(CH3)~2~), 135.8 (C-5), 127.3, 128.0 (C~Ar~), 125.5 (C-6), 79.5 (C-2), 73.3 (C-3), 69.4 (C-4), 65.2 (C-1), 30.6, 30.1 (C-7, C-8), 28.5, 25.7 (2 × C(CH~3~)~3~), 14.9 (C-9); HRMS calcd for C~31~H~36~O~4~ \[M+Na\]^+^: 495.2511, found 495.2511. (2*R*,3*S*,4*R*,5*E*/*Z*)-3,4-*O*-Isopropylidene-1-*O*-trityloxy-dodec-5-en-2-ol **8** {#s0040} -------------------------------------------------------------------------------------- 4.4 Prepared following general procedures (a) and (b) using triphenylphosphine (18.00 g, 68.7 mmol) and 1-bromoheptane (12.24 g, 22.9 mmol), *n*-BuLi (26.4 mL, 64.2 mmol) and **6** (9.90 g, 22.9 mmol) to afford compound **8** as a colourless syrup in 63% yield as a mixture of *cis* and *trans* isomer in a 1:2.3 ratio (5.88 g, 13.68 mmol). ^1^H NMR (CDCl~3~): *δ* 7.19--7.48 (15H, m, Ar--H), 5.47--5.58 (2H, m, H-5, H-6), 4.91--4.93 (0.7 H, m, H-4^trans^), 4.43--4.46 (0.3H, m, H-4^cis^), 4.25 (0.3H, dd, *J*~2,3~ = *J*~3,4~ = 4.6 Hz, H-3^cis^), 4.21 (0.7H, dd, *J*~2,3~ = *J*~3,4~ = 4.4 Hz, H-3^trans^), 3.68--3.81 (0.7H, m, H-2^trans^), 3.51--3.56 (0.3H, m, H-2^cis^), 3.22 (0.3H, dd, *J*~1a,2~ = 5.1, *J*~1a,1b~ = 5.3 Hz, H-1a^cis^), 3.18 (0.7H, dd, *J*~1a,2~ = 9.5, *J*~1a,1b~ = 5.3 Hz, H-1a^trans^), 3.07--3.17 (1H, m, H-1b^trans^, H-1b^cis^), 1.91--2.09 (2H, m, H-7a^cis^, H7b^cis^, H-7a^trans^, H-7b^trans^), 1.54, 1.49, 1.39, 1.38 (6H, 4s, 4 × C(CH~3~)~2~), 1.19--1.39 (8H, m, CH~2~), 0.88 (3H, t, *J* = 5.7 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 145.9, (O~2~C(CH3)~2~), 137.3 (C-5), 129.1, 129.9, 130.8 (C~Ar~), 127.1 (C-6), 79.5 (C-2), 75.1 (C-3), 71.3 (C-4), 67.1 (C-1), 34.2, 31.3, 31.1, 29.7, 25.0 (C-7--C-12), 29.8, 27.1 (2 × C(CH~3~)~3~), 17.1 (C-12); HRMS calcd for C~34~H~42~O~4~ \[M+Na\]^+^: 537.2980, found 537.2982. (2*R*,3*S*,4*R*,5*E*/*Z*)-3,4-*O*-Isopropylidene-1-*O*-trityloxy-pentadec-5-en-2-ol **9** {#s0045} ----------------------------------------------------------------------------------------- 4.5 Prepared following general procedures (a) and (b) using triphenylphosphine (54. 6 g, 0.20 mol, 3 equiv) and 1-bromodecane (15.4 g, 69.5 mmol), *n*-BuLi (74.0 mL, 0.18 mol) and **6** (10.00 g, 23.2 mmol) to afford compound **9** as a colourless syrup in 55% yield as a mixture of *cis* and *trans* isomer in a 1:2.3 ratio (7.12 g, 12.8 mmol). ^1^H NMR (CDCl~3~): *δ* 7.19--7.48 (15H, m, Ar--H), 5.47--5.58 (2H, m, H-5, H-6), 4.91--4.93 (0.7 H, m, H-4^trans^), 4.40--4.46 (0.3H, m, H-4^cis^), 4.25 (0.3H, dd, *J*~2,3~ = *J*~3,4~ = 4.6 Hz, H-3^cis^), 4.21 (0.7H, dd, *J*~2,3~ = *J*~3,4~ = 4.4 Hz, H-3^trans^), 3.72--3.79 (0.7H, m, H-2^trans^), 3.68--3.70 (0.3H, m, H-2^cis^), 3.22 (0.3H, dd, *J*~1a,2~ = 5.1, *J*~1a,1b~ = 5.3 Hz, H-1a^cis^), 3.15 (0.7H, dd, *J*~1a,2~ = 9.5, *J*~1a,1b~ = 5.3 Hz, H-1a^trans^), 3.08--3.13 (1H, m, H-1b^trans^, H-1b^cis^), 1.71--2.03 (2H, m, H-7a^cis^, H7b^cis^, H-7a^trans^, H-7b^trans^), 1.56, 1.47, 1.39, 1.38 (6H, 4s, 4 × C(CH~3~)~2~), 1.11--1.35 (14H, m, CH~2~), 0.87 (3H, t, *J* = 6.7 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 145.7, (O~2~C(CH3)~2~), 137.2 (C-5), 130.6, 129.7, 128.9 (C~Ar~), 126.8 (C-6), 79.5 (C-2), 74.9 (C-3), 70.8 (C-4), 66.5 (C-1), 34.0, 31.3, 31.2, 30.0 (C-7--C-14), 29.9, 27.0 (2 × C(CH~3~)~3~), 16.0 (C-15); HRMS calcd for C~37~H~48~O~4~ \[M+Na\]^+^: 579.3450, found 579.3454. General procedure for mesylation reaction (c) {#s0050} --------------------------------------------- 4.6 Compounds **7**, **8** and **9** (1 equiv) were, respectively, dissolved in a mixture of anhydrous CH~2~Cl~2~ and dry pyridine (3:1) and cooled to 0 °C. Methanesulfonylchloride (1.6 equiv) was then added dropwise and the reaction mixture stirred for 4 h at room temperature. Upon completion of the reaction, the reaction was quenched with sodium bicarbonate solution, diluted with CH~2~Cl~2~ (20 mL) and washed successively with water (20 mL) and brine (20 mL). The organic layer was dried over anhydrous Na~2~SO~4~ and evaporated under vacuo to give the mesylated compounds **10**, **11** and **12** as thick syrups in quantitative yields. (2*R*,3*S*,*R*,5*E*/*Z*)-3,4-*O*-Isopropylidene-2-methanesulfonyloxy-1-*O*-trityloxy-non-5-enol **10** {#s0055} ------------------------------------------------------------------------------------------------------ 4.7 Prepared following general procedure (c) using **7** (6.46 g, 13.7 mmol) and methanesulfonyl chloride (1.8 mL, 22.7 mmol) in a mixture of CH~2~Cl~2~ (45 mL) and pyridine (20 mL) to afford compound **10** (6.13 g, 11.2 mmol) as an off-white solid in 82% yield as a mixture of *cis* and *trans* isomer in a 1:2.3 ratio. ^1^H NMR (CDCl~3~): *δ* 7.21--7.49 (15H, m, Ar--H), 5.89--5.48 (2H, m, H-5, H-6), 4.78--5.07 (1H, m, H-2), 4.34--4.50 (2H, m, H-3, H-4), 3.92--4.26 (2H, m, H-1a, H-1b), 3.14, 3.12 (3H, 2s, OSO~2~CH~3~), 1.85--2.12 (2H, m, H-7a^cis^, H7b^cis^, H-7a^trans^, H-7b^trans^), 1.58, 1.52, 1.39, 1.37 (6H, 4s, 4 × C(CH~3~)~2~), 1.22--1.31 (2H, m, CH~2~), 0.85 (3H, t, *J* = 7.2 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 141.4 (O~2~C(CH3)~2~), 135.3 (C-5), 126.9, 126.8, 126.1 (C~Ar~), 121.9 (C-6), 79.2 (C-2), 74.3 (C-3), 70.2 (C-4), 61.3 (C-1), 37.2 (COSO~2~CH~3~), 34.6, 28.7 (C-7, C-8), 28.5, 25.7 (2 × C(CH~3~)~3~), 14.9 (C-9); HRMS calcd for C~32~H~38~SO~6~ \[M+Na\]^+^: 573.2287, found 573.2290. (2*R*,3*S*,4*R*,5*E*/*Z*)-3,4-*O*-Isopropylidene-2-methanesulfonyloxy-1-*O*-trityloxy-dodec-5-enol **11** {#s0060} --------------------------------------------------------------------------------------------------------- 4.8 Prepared following general procedure (c) using **8** (2.59 g, 5.0 mmol) and methanesulfonyl chloride (0.62 mL, 8.06 mmol) in a mixture of CH~2~Cl~2~ (25 mL) and pyridine (8 mL) to afford compound **11** (2.80 g, 4.7 mmol) as a colourless oil in 94% yield as a mixture of *cis* and *trans* isomer in a 1:2.3 ratio. ^1^H NMR (CDCl~3~): *δ* 7.17--7.52 (15H, m, Ar--H), 5.03--5.52 (2H, m, H-5, H-6), 4.79--4.82 (1H, m, H-2), 4.58--4.75 (2H, m, H-3, H-4), 4.48 (0.7H, dd, *J*~1a,2~ = 5.9, *J*~1a,1b~ = 8.4 Hz, H-1a^trans^), 4.24 (0.3H, dd, *J*~1a,2~ = 6.9, *J*~1a,1b~ = 14.0 Hz, H-1a^cis^), 3.54--3.57 (0.3H, m, H-1b^trans^), 3.45 (0.7H, dd, H-1b^cis^), 2.99, 3.18 (3H, 2s, OSO~2~CH~3~), 1.57--1.82 (2H, m, H-7a^cis^, H7b^cis^, H-7a^trans^, H-7b^trans^), 1.48, 1.41, 1.39, 1.38 (6H, 4s, 4 × C(CH~3~)~2~), 1.10--1.33 (8H, m, CH~2~), 0.85 (3H, t, *J* = 6.7 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 144.8 (O~2~C(CH3)~2~), 137.8 (C-5), 130.2, 129.4, 128.7 (C~Ar~), 125.1 (C-6), 82.5 (C-2), 77.6 (C-3), 73.7 (C-4), 64.6 (C-1), 39.9 (COSO~2~CH~3~), 35.1, 30.9, 30.2 (C-7, C-8, C-9), 29.5, 28.1 (2 × C(CH~3~)~3~), 29.0 (C-10), 24.5 (C-11), 16.8 (C-9); HRMS calcd for C~35~H~44~SO~6~ \[M+Na\]^+^: 615.2757, found 615.2560. (2*R*,3*S*,4*R*,5*E*/*Z*)-3,4-*O*-Isopropylidene-2-methanesulfonyloxy-1-*O*-trityloxy-pentadec-5-enol **12** {#s0065} ------------------------------------------------------------------------------------------------------------ 4.9 Prepared following general procedure (c) using **9** (5.39 g, 9.67 mmol) and methanesulfonyl chloride (1.3 mL, 16.1 mmol) in a mixture of CH~2~Cl~2~ (45 mL) and pyridine (15 mL) to afford compound **12** (5.52 g, 8.70 mmol) as an opaque oil in 90% yield as a mixture of *cis* and *trans* isomer in a 1:2.3 ratio. ^1^H NMR (CDCl~3~): *δ* 7.19--7.48 (15H, m, Ar--H), 5.40--5.48 (0.7H, m, H-6^trans^), 5.28--5.37 (0.7H, m, H-5^trans^), 5.20--5.25 (0.3H, m, H-6^cis^), 4.97--5.06 (0.3H, M, H-5^cis^), 4.79--4.85 (0.7H, m, H-2^trans^), 4.72--4.75 (0.7H, m, H-4^trans^), 4.62--4.66 (0.3H, m, H-2^cis^), 4.57--4.61(0.3H, m, H-4^cis^), 4.49 (0.7H, dd, *J*~3,4~ = 8.7, *J*~2,3~ = 5.6 Hz, H-3^trans^), 4.20 (0.3H, dd, *J*~3,4~ = 9.4, *J*~3,2~ = 4.2 Hz, H-3^cis^), 3.45--3.56 (2H, m, H-1a, H-1b), 3.04, 3.10 (3H, 2s, OSO~2~CH~3~), 1.85--2.12 (2H, m, H-7a^cis^, H7b^cis^, H-7a^trans^, H-7b^trans^), 1.59--1.82 (2H, m, H-7), 1.48, 1.47, 1.38, 1.36 (6H, 4s, 4 × C(CH~3~)~2~), 1.08--1.32 (14H, m, CH~2~), 0.88 (3H, t, *J* = 6.5 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 144.3 (O~2~C(CH3)~2~), 137.3 (C-5), 129.7, 128.9, 128.2 (C~Ar~), 124.6 (C-6), 82.0 (C-2), 77.1 (C-3), 73.2 (C-4), 64.1 (C-1), 40.1 (COSO~2~CH~3~), 33.1, 32.4, 30.5, 30.3, 30.2 (C-7, C-14), 26.8, 26.6 (2 × C(CH~3~)~3~), 16.9 (C-15); HRMS calcd for C~38~H~50~SO~6~ \[M+Na\]^+^: 657.3226, found 657.3223. General procedure for deprotection and reduction of double bond (d) {#s0070} ------------------------------------------------------------------- 4.10 The mesylated compounds **10**, **11** and **12** were, respectively, dissolved in a mixture of dry CH~2~Cl~2~ and MeOH (2:1) (20 mL) and concentrated hydrochloric acid (3 mL) was added dropwise and the mixture stirred at room temperature for 2 h, after which time TLC analysis indicated that the reaction was complete. Solid sodium bicarbonate was then added to quench the reaction until the solution was neutral. The mixture was then filtered and the filtrate concentrated. The residue was dissolved again in EtOAc and the organic solution washed consecutively with water (2 × 20 mL), brine (20 mL), dried over anhydrous Na~2~SO~4~ and concentrated in vacuo. The residues were purified by flash chromatography (gradient from to hexanes/EtOAc (4:1) to neat EtOAc) and dissolved in THF (10 mL). Five percentage of Pd--Ba(SO~4~)~2~ (0.1 equiv) was added to the solution and the mixture was stirred under H~2~ overnight, after which time it was filtered through a pad of Celite, which was subsequently washed with CHCl~3~/MeOH (1:1). The combined filtrates were concentrated to yield compounds **13**, **14** and **15**. (2*R*,3*S*,4*R*)-2-Methanesulfonyloxy-nonane-1,3,4-triol **16** {#s0075} --------------------------------------------------------------- 4.11 Prepared following general procedure (d) using **10** (5.06 g, 9.18 mmol) to give **16** as an off-white wax (1.91 g, 7.07 mmol) in 77% yield. ^23^\[α\]~D~ = −80.0 (*c* 1.00, CH~3~OH). ^1^H NMR (CD~3~OD): *δ* 4.84--4.88 (1H, m, H-2), 3.78--3.88 (2H, m, H1a, H-1b), 3.49--3.55 (1H, m, H-4), 3.43--3.47 (1H, dd, *J*~2,3~ = 8.3, *J*~3,4~ = 2.2 Hz, H-3), 3.15 (3H, s, OSO~2~CH~3~), 2.76 (3H, br s, OH), 1.67--1.75 (1H, m, H-5a), 1.46--1.53 (1H, m, H-7a), 1.19--1.38 (6H, m, H-5b, H-6a, H-6b, H-8a, H-8b), 0.88 (3H, t, *J* = 6.8 Hz, CH~3~); ^13^C NMR (75 MHz, CD~3~OD): *δ* 82.9 (C-2), 73.6 (C-3), 70.5 (C-4), 62.3 (C-1), 38.4 (COSO~2~CH~3~), 33.1, 32.0, 25.2, 22.8 (C-5--C-8), 14.0 (C-9); HRMS calcd for C~10~H~22~SO~6~ \[M+Na\]^+^: 293.1035, found 293.1026. (2*R*,3*S*,4*R*)-2-Methanesulfonyloxy-dodecane-1,3,4-triol **17** {#s0080} ----------------------------------------------------------------- 4.12 Prepared following general procedure (d) using **11** (2.46 g, 4.15 mmol) to give **17** as a colourless syrup (0.946 g, 3.03 mmol) in 73% yield. ^23^\[α\]~D~ = −32.2 (*c* 1.00, CH~3~OH). ^1^H NMR (CD~3~OD): *δ* 4.74--4.78 (1H, m, H-2), 3.62--3.78 (2H, m, H1a, H-1b), 3.31--3.47 (1H, m, H-4), 3.18--3.21 (1H, m, H-3), 3.06 (3H, s, OSO~2~CH~3~), 1.48--1.56 (2H, m, H-5a, H-7a), 1.22--1.43 (12H, m, CH~2~), 0.73 (3H, t, *J* = 6.5 Hz, CH~3~); ^13^C NMR (75 MHz, CD~3~OD): *δ* 83.9 (C-2), 74.7 (C-3), 71.5 (C-4), 63.3 (C-1), 40.1 (COSO~2~CH~3~), 34.2, 33.1, 30.9, 30.8 30.5, 26.5, 23.8 (C-5--C-11), 15.1 (C-12); HRMS calcd for C~13~H~28~SO~6~ \[M+Na\]^+^: 335.1504, found 335.1511. (2*R*,3*S*,4*R*)-2-Methanesulfonyloxy-pentadecane-1,3,4-triol **18** {#s0085} -------------------------------------------------------------------- 4.13 Prepared following general procedure (d) using **12** (6.93 g, 10.9 mmol) to give **18** as a white solid (2.79 g, 7.87 mmol) in 72% yield. ^23^\[α\]~D~ = −180.0 (*c* 1.00, CH~3~OH). ^1^H NMR (CD~3~OD): *δ* 4.99--5.05 (1H, m, H-2), 3.98--4.04 (2H, m, H1a, H-1b), 3.55--3.62 (1H, m, H-4), 3.45--3.49 (1H, m, H-3), 3.15 (3H, s, OSO~2~CH~3~), 1.48--1.86 (2H, m, H-5a, H-7a), 1.22--1.43 (18H, m, CH~2~), 0.89 (3H, t, *J* = 6.8 Hz, CH~3~); ^13^C NMR (75 MHz, CD~3~OD): *δ* 85.4 (C-2), 75.6 (C-3), 72.8 (C-4), 64.1 (C-1), 40.1 (COSO~2~CH~3~), 34.3, 33.1, 32.0, 30.8, 30.4, 29.2, 28.7, 26.2, 25.2, 22.8 (C-5--C-14), 16.2 (C-15); HRMS calcd for C~16~H~34~SO~6~ \[M+Na\]^+^: 377.1974, found 377.1980. General procedure for displacement of mesylate group with sodium azide (e) {#s0090} -------------------------------------------------------------------------- 4.14 The mesylated compounds were dissolved in DMF (10 mL) and sodium azide (2 equiv) was added. The reaction was stirred overnight at 60 °C, after which, it was taken up in water (50 mL) and extracted with EtOAc (3 × 15 mL). The combined organic layers were then washed with brine (30 mL), dried over anhydrous Na~2~SO~4~ and evaporated. The residue was then purified by flash chromatography using hexanes/EtOAc (4:1). (2*S*,3*S*,4*R*)-2-Azido-nonane-1,3,4-triol **19** {#s0095} -------------------------------------------------- 4.15 Prepared following general procedure (e) using **16** (0.74 g, 2.75 mmol), sodium azide (0.36 g, 5.50 mmol) to afford a white solid (0.38 g, 1.73 mmol, 63%). ^23^\[α\]~D~ = +144.0 (*c* 1.00, CH~3~OH). ^1^H NMR (CD~3~OD): *δ* 3.82--3.86 (1H, dd, *J*~1a,1b~ = 11.5, *J*~1a,2~ = 4.9 Hz, H-1a), 3.71--3.77 (1H, m, H-1b), 3.58--3.64 (2H, m, H-3, H-4), 3.44--3.47 (1H, ddd, *J*~1b,2~ = 9.9, *J*~2,3~ = 4.5 Hz), 1.45--1.56 (2H, m, H-5a, H-7a), 1.38 (1H, m, H-5b), 1.27--1.14 (5H, m, CH~2~), 0.79 (3H, t, *J* = 6.1 Hz, CH~3~); ^13^C NMR (75 MHz, CD~3~OD): *δ* 74.0 (C-3), 71.9 (C-4), 62.7 (C-2), 61.0 (C-1), 31.5 (C-5), 31.3 (C-6), 24.9 (C-7), 22.1 (C-8), 13.5 (C-9); HRMS calcd for C~9~H~19~N~3~O~3~\[M+Na\]^+^: 240.1324, found 240.1315. (2*S*,3*S*,4*R*)-2-Azido-dodecane-1,3,4-triol **20** {#s0100} ---------------------------------------------------- 4.16 Prepared following general procedure (e) using **17** (0.95 g, 3.03 mmol), sodium azide (0.39 g, 6.06 mmol) to afford a colourless syrup (0.69 g, 2.64 mmol, 87%). ^23^\[α\]~D~ = +45.0 (*c* 1.00, CH~3~OH). ^1^H NMR (CD~3~OD): *δ* 3.88--3.95 (1H, dd, *J*~1a,1b~ = 11.1, *J*~1a,2~ = 3.3 Hz, H-1a), 3.71--3.80 (1H, m, H-1b), 3.54--3.63 (1H, m, H-4), 3.48--3.53 (1H, m, H-3), 3.28--3.32 (1H, m, H-2), 1.20--1.45 (11H, m, CH~2~), 0.89 (3H, t, *J* = 6.2 Hz, CH~3~); ^13^C NMR (75 MHz, CD~3~OD): *δ* 75.9 (C-3), 72.8 (C-4), 66.2 (C-2), 62.4 (C-1), 33.8, 32.9, 30.7, 30.3, 26.6, 23.6 (C-5--C-11), 13.8 (C-12); HRMS calcd for C~12~H~25~N~3~O~3~\[M+Na\]^+^: 282.1794, found 282.1794. (2*S*,3*S*,4*R*)-2-Azido-pentadecane-1,3,4-triol **21** {#s0105} ------------------------------------------------------- 4.17 Prepared following general procedure (e) using **18** (2.64 g, 7.45 mmol), sodium azide (0.97 g, 14.91 mmol) to afford a colourless oil (1.61 g, 5.36 mmol, 72%).^23^\[α\]~D~ = +31.4 (*c* 1.00, CH~3~OH). ^1^H NMR (CD~3~OD): *δ* 3.92--3.98 (1H, dd, *J*~1a,1b~ = 11.4, *J*~1a,2~ = 3.3 Hz, H-1a), 3.74--3.88 (1H, m, H-1b), 3.58--3.66 (1H, m, H-4), 3.41--3.58 (1H, m, H-3), 3.32--3.36 (1H, m, H-2), 1.51--1.76(3H, m, CH~2~), 1.18--1.47 (17H, m, CH~2~), 0.91 (3H, t, *J* = 6.6 Hz, CH~3~); ^13^C NMR (75 MHz, CD~3~OD): *δ* 73.9 (C-3), 70.8 (C-4), 64.6 (C-2), 60.4 (C-1), 31.8, 31.0, 28.7, 28.4, 24.7, 21.7 (C-5--C-14), 13.6 (C-15); HRMS calcd for C~12~H~25~N~3~O~3~\[M+Na\]^+^: 324.2263, found 324.2275. Preparation of glycosyl acceptors **22**, **23** and **24** {#s0110} ----------------------------------------------------------- 4.18 Compounds **19**, **20** and **21** were subjected to the same standard reaction conditions described by Akimoto et al.[@b0190] and depicted in [Scheme 2](#f0030){ref-type="fig"} for the preparation of (2*R*, 3*S*, 4*R*)-2-azido-3, 4-di-*O*-(benzoyloxy)-octadecan-1-ol. (2*S*,3*S*,4*R*)-2-Azido-3,4-di-*O*-(benzoyloxy)-nonan-1-ol **22** {#s0115} ------------------------------------------------------------------ 4.19 ^23^\[α\]~D~ = +81.4 (*c* 1.00, CHCl~3~). ^1^H NMR (CDCl~3~): *δ* 7.48--8.07 (10H, m, Ar--H), 5.49--5.46 (2H, m, H-3, H-4), 3.98--4.01 (1H, m, H-1a), 3.75--3.82 (2H, m, H-1b, H-2), 1.82--1.99 (2H, m, H-5a, H-5b), 1.18--1.51 (6H, m, CH~2~), 0.89 (3H, t, *J* = 6.8 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 166.1, 165.7 (C\<svg xmlns=\"http://www.w3.org/2000/svg\" version=\"1.0\" width=\"20.666667pt\" height=\"16.000000pt\" viewBox=\"0 0 20.666667 16.000000\" preserveAspectRatio=\"xMidYMid meet\"\>\<metadata\> Created by potrace 1.16, written by Peter Selinger 2001-2019 \</metadata\>\<g transform=\"translate(1.000000,15.000000) scale(0.019444,-0.019444)\" fill=\"currentColor\" stroke=\"none\"\>\<path d=\"M0 440 l0 -40 480 0 480 0 0 40 0 40 -480 0 -480 0 0 -40z M0 280 l0 -40 480 0 480 0 0 40 0 40 -480 0 -480 0 0 -40z\"/\>\</g\>\</svg\>O), 133.7--129.7 (C~Ar~), 129.3, 129.1, 128.6, 128.4 (C~Ar~), 73.4 (C-3), 72.9 (C-4), 63.1 (C-2), 62.1 (C-1), 31.5 (C-7), 29.7 (C-5), 25.2 (C-6), 22.4 (C-8), 13.9 (C-9); HRMS calcd for C~23~H~27~N~3~O~5~\[M+Na\]^+^: 448.1848, found 448.1829. (2*S*,3*S*,4*R*)-2-Azido-3,4-di-*O*-(benzoyloxy)-dodecan-1-ol **23** {#s0120} -------------------------------------------------------------------- 4.20 ^23^\[α\]~D~ = +8.8 (*c* 1.00, CHCl~3~). ^1^H NMR (CDCl~3~): *δ* 7.26--7.89 (10H, m, Ar--H), 5.32--5.39 (2H, m, H-3, H-4), 3.65--3.82 (1H, m, H-1a), 3.54--3.60 (2H, m, H-1b, H-2), 1.62--1.82 (2H, m, H-5a, H-5b), 1.06--1.30 (12H, m, CH~2~), 0.68 (3H, t, *J* = 6.7 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 136.4.7--131.1 (C~Ar~), 76.1 (C-3), 75.4 (C-4), 66.2 (C-2), 64.8 (C-1), 34.5, 33.0, 30.7, 30.3, 26.6, 23.6 25.3 (C-5--C11), 16.7 (C-12); HRMS calcd for C~26~H~33~N~3~O~5~\[M+Na\]^+^: 490.2318, found 490.2316. (2*S*,3*S*,4*R*)-2-Azido-3,4-di-*O*-(benzoyloxy)-pentadecan-1-ol **24** {#s0125} ----------------------------------------------------------------------- 4.21 ^23^\[α\]~D~ = +157.2 (*c* 1.00, CHCl~3~). ^1^H NMR (CDCl~3~): *δ* 7.25--8.11 (10H, m, Ar--H), 5.48--5.5.4 (2H, m, H-3, H-4), 4.82 (1H, dd, *J*~1a,2~ = 7.3, *J*~1a,1b~ = 14.6 Hz, H-1a), 3.67--4.03 (2H, m, H-1b, H-2), 177--1.97 (2H, m, H-5a, H-5b), 1.09--1.46 (18H, m, CH~2~), 0.85 (3H, t, *J* = 6.6 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 162.5 (CO), 135.8--130.6 (C~Ar~), 75.4 (C-3), 75.1 (C-4), 65.3 (C-2), 64.3 (C-1), 34.0, 32.6, 32.4, 31.5, 29.7, 25.2, 23.8, 23.2, 22.8 22.4 (C-5--C14), 16.2 (C-15); HRMS calcd for C~29~H~39~N~3~O~5~\[M+Na\]^+^: 532.6370, found 532.6373. General procedure for glycosylation reaction (f) {#s0130} ------------------------------------------------ 4.22 To a solution of the persilylated galactose (3 equiv) in anhydrous CH~2~Cl~2~ (20 mL), iodotrimethylsilane (3 equiv) was added and the reaction mixture stirred at room temperature under argon for 30 min. The mixture was then concentrated and azeotroped twice with dry benzene (5 mL). The yellow residue was dissolved in dry benzene and kept under argon. Meanwhile, activated 4 Å molecular sieves, tetrabutyl ammonium iodide (6 equiv), respective glycosyl acceptors **22**, **23** and **24** (1 equiv) and diisopropylethylamine (4.5 equiv) were added to a two-necked flask fitted with a condenser. Benzene (10 mL) was added and the solution stirred at 70 °C for 20 min. The solution of the glycosyl iodide **25** in benzene was then cannulated into the two-necked flask and stirred at room temperature for two days. Upon completion of the reaction, the mixture was filtered through Celite and the Celite pad was washed with CH~2~Cl~2~ (10 mL). The filtrate was washed with saturated sodium thiosulfate (10 mL), brine (10 mL), dried over anhydrous Na~2~SO~4~ and concentrated in vacuo. The residue was dissolved in MeOH and Dowex 50WX8-200 ion exchange resin was added and the mixture stirred at room temperature overnight. The latter was then filtered and the filtrate concentrated to give a residue which was purified by flash chromatography using EtOAc/hexanes (7:1) to give the glycosylated product. (2*S*,3*S*,4*R*)-2-Azido-3,4-di-*O*-(benzoyloxy)-1-(α-[d]{.smallcaps}-galactopyranosyl)-nonane **26** {#s0135} ----------------------------------------------------------------------------------------------------- 4.23 Prepared following general procedure (f) using per-*O*-pentamethylsilyl-α-[d]{.smallcaps}-galactose (0.27 g, 0.49 mmol), iodotrimethylsilane (0.09 g, 0.07 mL, 0.49 mmol), TBAI (3.65 mg, 0.99 mmol), DIPEA (97.5 mg, 0.13 mL, 0.74 mmol) and **22** (70 mg, 0.16 mmol) to afford **26** as a pale yellow oil (43 mg, 0.07 mmol) in 46% yield. ^23^\[α\]~D~ = +36.1 (*c* 1.00, CHCl~3~). ^1^H NMR (CDCl~3~): *δ*7.38--7.96 (10H, m, Ar--H), 5.59--5.61 (1H, m, H-3^Cer^), 5.40--5.50 (1H, m, H-4^Cer^), 4.80 (1H, s, H-1), 4.10--4.13 (1H, dd, *J*~6a,~ ~6b~ = 7.9, *J*~5,6a~ = 3.0 Hz, H-6a), 3.91--3.94 (2H, m, H-2^Cer^, H-4), 3.80--3.84 (1H, dd, *J*~2,3~ = *J*~3,4~ = 6.0 Hz, H-3), 3.70--3.75 (4H, m, H-2, H-5, H-1a^Cer^, H-1b^Cer^), 3.62--3.67 (1H, m, H-6b), 1.81--1.91 (2H, m, H-5a^Cer^, H-5b^Cer^), 1.15--1.41 (6H, m, CH~2~), 0.83 (3H, t, *J* = 6.4 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 167.1, 167.0 (CO), 127.4--133.4 (C~Ar~), 101.4 (C-1), 73.2 (C-4^Cer^), 72.1 (C-3^Cer^), 70.9 (C-3), 70.6 (C-5), 70.2 (C-4), 69.7 (C-2), 68.1 (C-6), 62.4 (C-1^Cer^), 61.0 (C-2^Cer^), 21.9--33.3 (4 × CH~2~^Cer^), 13.9 (CH~3~^Cer^); HRMS calcd for C~29~H~37~N~3~O~10~\[M+Na\]^+^: 610.6183, found 610.6188. (2*S*,3*S*,4*R*)-2-Azido-3,4-di-*O*-(benzoyloxy)-1-(α-[d]{.smallcaps}-galactopyranosyl)-dodecane **27** {#s0140} ------------------------------------------------------------------------------------------------------- 4.24 Prepared following general procedure (f) using per-*O*-pentamethylsilyl-α-[d]{.smallcaps}-galactose (0.85 g, 1.58 mmol), iodotrimethylsilane (0.32 g, 0.23 mL, 1.58 mmol), TBAI (1.17 g, 3.16 mmol), DIPEA (1.05 g, 1.42 mL, 8.16 mmol) and **23** (0.25 g, 0.53 mmol) to afford **27** as an off-white solid (110 mg, 0.17 mmol) in 33% yield.. ^23^\[α\]~D~ = +92.4 (*c* 1.00, CHCl~3~). ^1^H NMR (CDCl~3~): *δ*7.38--7.98 (10H, m, Ar--H), 5.61--5.64 (1H, m, H-3^Cer^), 5.48--5.52 (1H, m, H-4^Cer^), 4.79 (1H, s, H-1), 4.11--4.13 (1H, dd, *J*~6a,~ ~6b~ = 8.0, *J*~5.6a~ = 2.9 Hz, H-6a), 3.90--3.93 (2H, m, H-2^Cer^, H-4), 3.82--3.85 (1H, m, H-3), 3.71--3.75 (4H, m, H-2, H-5, H-1a^Cer^, H-1b^Cer^), 3.63--3.67 (1H, dd, *J*~5,6b~ = 6.0 Hz, H-6b), 1.81--1.94 (2H, m, H-5a^Cer^, H-5b^Cer^), 1.18--1.42 (12H, m, CH~2~), 0.79 (3H, t, *J* = 6.7 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 166.8, 166.3 (CO), 128.7--134.2 (C~Ar~), 100.2 (C-1), 73.7 (C-4^Cer^), 72.6 (C-3^Cer^), 71.4 (C-3), 70.6 (C-5), 70.2 (C-4), 69.5 (C-2), 67.9 (C-6), 62.1 (C-1^Cer^), 61.2 (C-2^Cer^), 22.9--32.1 (6 × CH~2~^Cer^), 14.2 (CH~3~^Cer^); HRMS calcd for C~32~H~43~N~3~O~10~\[M+Na\]^+^: 652.2846, found 610.2845. (2*S*,3*S*,4*R*)-2-Azido-3,4-di-*O*-(benzoyloxy)-1-(α-[d]{.smallcaps}-galactopyranosyl)-pentadecane **28** {#s0145} ---------------------------------------------------------------------------------------------------------- 4.25 Prepared following general procedure (f) using per-*O*-pentamethylsilyl-α-[d]{.smallcaps}-galactose (1.61 g, 2.97 mmol), iodotrimethylsilane (0.59 g, 0.43 mL, 2.97 mmol), TBAI (2.19 g, 5.94 mmol), DIPEA (0.58 g, 0.78 mL, 4.45 mmol) and **24** (0.50 g, 0.99 mmol) to afford **28** as an off-white solid (175 mg, 0.26 mmol) in 26% yield.. ^23^\[α\]~D~ = +81.4 (*c* 1.00, CHCl~3~). ^1^H NMR (CDCl~3~): *δ*7.37--8.02 (10H, m, Ar--H), 5.62--5.66 (1H, m, H-3^Cer^), 5.48--5.55 (1H, m, H-4^Cer^), 4.80 (1H, d, *J*~1,2~ = 2.4 Hz H-1), 4.12--4.16 (1H, m, H-6a), 3.90--3.95 (2H, m, H-2^Cer^, H-4), 3.83--3.85 (1H, m, H-3), 3.71--3.74 (4H, m, H-2, H-5, H-1a^Cer^, H-1b^Cer^), 3.63--3.69 (1H, m, H-6b), 1.82--1.92 (2H, m, H-5a^Cer^, H-5b^Cer^), 1.15--1.30 (16H, m, CH~2~), 0.82 (3H, t, *J* = 6.9 Hz, CH~3~); ^13^C NMR (75 MHz, CDCl~3~): *δ* 135.1--139.5 (C~Ar~), 106.5 (C-1), 79.5 (C-4^Cer^), 78.1 (C-3^Cer^), 71.4 (C-3), 70.8 (C-5), 70.0 (C-4), 69.5 (C-2), 67.7 (C-6), 63.4 (C-1^Cer^), 62.6 (C-2^Cer^), 22.9--36.8 (11 × CH~2~^Cer^), 19.5 (CH~3~^Cer^); HRMS calcd for C~35~H~49~N~3~O~10~\[M+Na\]^+^: 694.7686, found 694.7683. General procedure for the removal of benzoate esters, hydrogenation of the azido group and subsequent *N*-acylation (g) {#s0150} ----------------------------------------------------------------------------------------------------------------------- 4.26 Compounds **26**, **27** and **28** were, respectively, dissolved in MeOH and a 1 M solution of NaOMe (5 mL) was added. The mixture was stirred at room temperature for 2 h, after which time TLC analysis indicated that the reaction was complete. The reaction was neutralised by the addition of Dowex 50WX8-200 resin. The latter was then filtered and the filtrate concentrated to give a residue which was purified by flash chromatography using EtOAc/MeOH (7:1) to give compounds **29**, **30** and **31** in quantitative yields. Compounds **29**, **30** and **31** were, respectively, dissolved in MeOH (10 mL) and stirred with Pd--C (5 mg) under H~2~ overnight, after which time the mixture was filtered through a pad of Celite, which was subsequently washed with MeOH. The combined filtrates were concentrated to give the respective amines as white solids. The crude amine was then dissolved in a 1:1 mixture of THF/NaOAC (saturated solution) (5 mL) and the freshly prepared acid chloride of hexacosanoic acid was added. The reaction was allowed to stir vigorously overnight at room temperature after which the organic phase was removed. The aqueous phase was further extracted with THF (2 × 1 mL), and the combined organic phases were concentrated. The residue was finally purified by flash chromatography (gradient from CHCl~3~ to 15% MeOH in CHCl~3~) to give target compounds **2**, **3** and **4**. (2*S*,3*S*,4*R*)-1-(α-[d]{.smallcaps}-Galactopyranosyl)-2-hexacosanoylamino-3,4-nonanediol **2** {#s0155} ------------------------------------------------------------------------------------------------ 4.27 Prepared following general procedure (g) using **26** (50 mg, 0.12 mmol) and hexacosanoic acid (71 mg, 0.18 mmol, 1.5 equiv) to give **2** as a white solid (31 mg, 0.04 mmol) in 35% yield.. ^23^\[α\]~D~ = +12.6 (*c* 1.00, CHCl~3~/CH~3~OH 2:1). ^1^H NMR (CDCl~3~/CD~3~OD 2:1): *δ* 4.85 (1H, d, *J*~1,2~ = 3.4 Hz, H-1), 4.13--4.16 (1H, m, H-2^Cer^), 3.86--3.89 (1H, m, H-3), 3.81--3.86 (1H, dd, *J*~1a,2~ = 4.6, *J*~1a,1b~ = 10.5 Hz, H-1a^Cer^), 3.72--3.78 (2H, m, H-2, H-5), 3.66--3.72 (3H, m, H-4, H-6a, H-6b), 3.60--3.66 (1H, m, H-1b^Cer^), 3.47--3.55 (2H, m, H-3^Cer^, H-4^Cer^), 2.13--2.18 (2H, m, NHCOCH~2~C~24~H~49~), 1.53--1.67 (3H, m, CH~2~), 0.94--1.46 (49H, m, CH~2~), 0.79 (6H, t, *J* = 6.7 Hz, CH~3~); ^13^C NMR (75 MHz, (CDCl~3~/CD~3~OD 2:1)): *δ* 103.6 (C-1), 77.7 (C-3^Cer^), 75.4 (C-4^Cer^), 73.7 (C-5), 73.2 (C-4), 72.6 (C-3), 71.7 (C-2), 70.7 (C-1^Cer^), 65.2 (C-6), 64.7 (C-2^Cer^), 23.8--33.4 (CH~2~), 15.9 (CH~3~, CH~3~^Cer^); HRMS calcd for C~41~H~81~NO~9~ \[M+Na\]^+^: 754.5809, found 754.5812. (2*S*,3*S*,4*R*)-1-(α-[d]{.smallcaps}-Galactopyranosyl)-2-hexacosanoylamino-3,4-dodecanediol **3** {#s0160} -------------------------------------------------------------------------------------------------- 4.28 Prepared following general procedure (g) using **27** (100 mg, 0.16 mmol) and hexacosanoic acid (95 mg, 0.24 mmol, 1.5 equiv) to give **3** as a white solid (45 mg, 0.06 mmol) in 37% yield. ^23^\[α\]~D~ = +15.6 (*c* 1.00, CHCl~3~/CH~3~OH 2:1). ^1^H NMR (CDCl~3~/CD~3~OD 2:1): *δ* 4.85 (1H, d, *J*~1,2~ = 3.8 Hz, H-1), 4.12--4.16 (1H, m, H-2^Cer^), 3.87--3.89 (1H, m, H-3), 3.81--3.85 (1H, dd, *J*~1a,2~ = 4.8, *J*~1a,1b~ = 10.5 Hz, H-1a^Cer^), 3.72--3.77 (2H, m, H-2, H-5), 3.66--3.72 (3H, m, H-4, H-6a, H-6b), 3.61--3.66 (1H, m, H-1b^Cer^), 3.47--3.51 (2H, m, H-3^Cer^, H-4^Cer^), 2.14--2.18 (2H, m, NHCOCH~2~C~24~H~49~), 1.46--1.65 (3H, m, CH~2~), 1.16--1.34 (57H, m, CH~2~), 0.82 (6H, t, *J* = 6.6 Hz, CH~3~); ^13^C NMR (75 MHz, (CDCl~3~/CD~3~OD 2:1)): *δ* 101.8 (C-1), 76.9 (C-3^Cer^), 74.2 (C-4^Cer^), 73.2 (C-5), 73.0 (C-4), 72.6 (C-3), 71.8 (C-2), 70.1 (C-1^Cer^), 65.5 (C-6), 64.4 (C-2^Cer^), 24.8--36.0 (CH~2~), 17.0 (CH~3~, CH~3~^Cer^); HRMS calcd for C~44~H~87~NO~9~ \[M+Na\]^+^: 797.6289, found 797.6291. (2*S*,3*S*,4*R*)-1-(α-[d]{.smallcaps}-Galactopyranosyl)-2-hexacosanoylamino-3,4-pentadecanediol **4** {#s0165} ----------------------------------------------------------------------------------------------------- 4.29 Prepared following general procedure (g) using **28** (100 mg, 0.15 mmol) and hexacosanoic acid (90 mg, 0.22 mmol, 1.5 equiv) to give **4** as a white solid (44 mg, 0.05 mmol) in 36% yield. ^23^\[α\]~D~ = +26.7 (*c* 1.00, CHCl~3~/CH~3~OH 2:1). ^1^H NMR (CDCl~3~/CD~3~OD 2:1): *δ* 4.85 (1H, d, *J*~1,2~ = 3.7 Hz, H-1), 4.13--4.16 (1H, m, H-2^Cer^), 3.87--3.89 (1H, m, H-3), 3.81--3.86 (1H, dd, *J*~1a,2~ = 4.8, *J*~1a,1b~ = 11.0 Hz, H-1a^Cer^), 3.72--3.77 (2H, m, H-2, H-5), 3.66--3.71 (3H, m, H-4, H-6a, H-6b), 3.61--3.66 (1H, m, H-1b^Cer^), 3.46--3.51 (2H, m, H-3^Cer^, H-4^Cer^), 2.13--2.18 (2H, m, NHCOCH~2~C~24~H~49~), 1.46--1.65 (3H, m, CH~2~), 1.11--1.35 (70H, m, CH~2~), 0.83 (6H, t, *J* = 6.8 Hz, CH~3~); ^13^C NMR (75 MHz, (CDCl~3~/CD~3~OD 2:1)): *δ* 102.5 (C-1), 76.8 (C-3^Cer^), 74.9 (C-4^Cer^), 73.7 (C-5), 73.1 (C-4), 72.9 (C-3), 71.9 (C-2), 70.7 (C-1^Cer^), 65.0 (C-6), 64.5 (C-2^Cer^), 25.6--35.1 (CH~2~), 16.9 (CH~3~, CH~3~^Cer^); HRMS calcd for C~50~H~99~NO~9~ \[M+Na\]^+^: 838.6760, found 838.6755. Elisa {#s0170} ----- 4.30 C1R-hCD1d cells were pulsed for overnight with GSLs. 5 × 10^4^ cell-pulsed targets were incubated at 37 °C with 2 × 10^4^ *i*NKT cells in a final volume of 200 μL. After 36 h, the supernatants were harvested, and the concentrations of IFN-γ and IL-4 were determined by commercial ELISA (BD Pharmingen). GSB acknowledges support in the form of a Personal Research Chair from Mr. James Badrick, Royal Society Wolfson Research Merit Award, as a former Lister Institute-Jenner Research Fellow, the Medical Council and The Wellcome Trust (084923/B/08/7). V.C. is supported by Cancer Research UK grant C399/A2291. The Bruker spectrometers used in this research was obtained, through Birmingham Science City: Innovative Uses for Advanced Materials in the Modern World (West Midlands Centre for Advanced Materials Project 2), with support from Advantage West Midlands (AWM) and part funded by the European Regional Development Fund (ERDF).

PubMed Central

2024-06-05T04:04:19.326324

2011-1-01

PMC3052435

Introduction ============ 1 T-cells detect antigens in the form of peptides bound to major histocompatibility complex (MHC) molecules at the cell surface. This primary recognition event enables the orchestration of adaptive immunity and targeted destruction of transformed and pathogen-infected cells. T-cell specificity is determined by the highly variable complementarity determining regions of the T cell receptor (TCR). The TCR/peptide-MHC (pMHC) interaction is very weak and classically endures for no longer than a few seconds at physiological temperatures. However, multimerization of pMHC molecules results in cooperative interactivity at the cell surface and ensures that the binding avidity of pMHC tetramers far exceeds the sum of the contributing monomeric affinities ([@bib34]). This avidity effect extends the binding half-life of pMHC tetramers ([@bib15]) and enables stable coherence to the surface of T-cells bearing cognate TCRs ([@bib1 bib4]). Consequently, pMHC tetramers have transformed the study of antigen-specific T-cells by enabling their visualization, enumeration, phenotypic characterization and isolation from *ex vivo* samples ([@bib1 bib8]). Indeed, pMHC tetramers have been used in thousands of studies in the decade since their initial inception and have spawned the formation of several commercial companies. We have recently used a monoclonal T-cell system to examine T-cell activation and pMHC class I (pMHCI) tetramer binding with a series of altered peptide ligands that vary in their affinity for the cognate TCR by over 100-fold ([@bib16]). Importantly, cell surface topography, including TCR and CD8 density, remain constant in this system. In this controlled system, efficient staining with tetrameric pMHCI required a monomeric TCR/pMHCI affinity of K~D~ \< 35μM; below this threshold, there was a sharp drop off in the intensity of pMHCI tetramer staining ([@bib16]). A reasonable T-cell agonist in this system bound with a K~D~ \~ 250μM and a weak agonist bound with a K~D~ \> 500μM. However, TCR/pMHCI affinities of \> 200μM were not detectable by pMHCI tetramer. Thus, using normal staining procedures, pMHC tetramers do not necessarily detect all T-cells that can respond to a particular agonist; similarly, not all agonists for a particular T-cell can be identified physically with pMHC tetramers. These potential limitations of pMHC tetramer staining, which likely extend across a range of multimeric valencies, have important implications for data interpretation and present a particular problem for the detection of tumour-specific or autoreactive T-cells that tend to express low affinity TCRs ([@bib9]). In this study, we demonstrate that pre-treatment with a protein kinase inhibitor (PKI) enhances pMHC tetramer staining of antigen-specific CD8^+^ and CD4^+^ T-cells and describe the mechanism through which these effects operate. Importantly, these benefits apply only to T-cells that express specific TCRs; PKI treatment does not result in the staining of T-cells that bear non-cognate TCRs. In addition, we show that PKI treatment lowers the TCR/pMHCI affinity threshold required for pMHCI tetramer binding by as much as 5 fold, thereby allowing the binding of pMHCI tetramers to CD8^+^ T-cells that express TCRs with very weak affinities for pMHCI (\> 500μM). This simple and universally applicable procedure thereby enables the visualization of previously undetectable tumour-specific and autoreactive CD8^+^ T-cells with pMHCI tetramers through the preferential enhancement of low avidity interactions with TCRs at the cell surface. Materials and methods ===================== 2 Cells ----- 2.1 The ILA1 CTL clone is specific for the HLA A⁎0201 (HLA A2 from hereon) restricted human telomerase reverse transcriptase (hTERT) epitope ILAKFLHWL (hTERT~540-548~). Mel13 and Mel5 CTL clones are specific for the HLA A2 restricted Melan-A~26-35~ epitope ELAGIGILTV. ILA1, Mel5 and Mel13 CD8^+^ cytotoxic T lymphocyte (CTL) clones were generated and re-stimulated as described previously ([@bib30 bib15]). CTL were maintained in RPMI 1640 (Gibco) supplemented with 100 U/ml penicillin (Gibco), 100μg/ml streptomycin (Gibco), 10% heat inactivated fetal calf serum (FCS; Gibco), 2.5% Cellkines (Helvetica Healthcare, Geneva), 200 IU/ml IL-2 and 25 ng/ml IL-15 (Peprotech). CTL lines specific for the influenza matrix protein M1~58-66~ (GILGFVFTL) and Melan-A~26-35~ (ELAGIGILTV) epitopes, both restricted by HLA A2, were generated by pulsing 6 x 10^6^ PBMC from an HLA A2 individual with cognate peptide at concentrations of 1μM and 100μM, respectively, for 1 hour at 37 °C; cells were subsequently washed and resuspended in RPMI 1640 supplemented with 100 U/ml penicillin (Gibco), 100 μg/ml streptomycin (Gibco) and 10% heat inactivated FCS (Gibco) only. After 3 days, increasing amounts of IL-2 were added to the media reaching a maximum concentration of 20 IU/ml by day 14; lines were then tested by pMHCI tetramer staining. Patient samples were collected by leukapheresis; mononuclear cells were isolated by standard Ficoll-Hypaque density gradient centrifugation and stored by cryopreservation. For autoimmune studies, blood was obtained from two HLA A2 patients with type 1 diabetes; both were adults, aged 27 and 31 years, and were studied within 3 months of diagnosis. Short-term lines from these diabetic patients were established as described above using the PPI~15-24~ autoantigen preproinsulin peptide (ALWGPDPAAA) ([@bib33]); this peptide binds HLA A2 with high affinity ([@bib3]). Naïve murine CTL were obtained by harvesting splenocytes from transgenic F5 Rag^+^ mice. A significant percentage of CD8^+^ T-cells within the splenic populations of these mice express the F5 TCR, which recognizes the H-2D^b^-restricted influenza H17-derived nucleoprotein peptide epitope (ASNENMDAM) ([@bib20]). The HLA DR⁎0101-restricted CD4^+^ T-cell clone C6 recognizes the influenza virus A HA~307-319~ epitope (PKYVKQNTLKLAT). Protein kinase inhibitors ------------------------- 2.2 Dasatinib was synthesized as described previously ([@bib17]). Biological activity was tested in a cell death titration assay on BA/F3 bcr-abl cells as described previously ([@bib19]). Dasatinib was dissolved in DMSO to a concentration of 1 mM and stored in aliquots at − 20 °C. Once thawed, these stocks of dasatinib were stored at 4 °C and used within 7 days. The 1 mM DMSO stock was diluted 1/10,000 in PBS on the day of experimentation to achieve a working solution of 100 nM; subsequent 1/2 dilution yielded a final concentration of 50 nM in cellular assays unless stated otherwise. The 100 nM working solution was always made up on the day of experimentation as the shelf-life of this solution is short (\~ days). Staurosporine (Biomol), Lck inhibitor II (Calbiochem), genestein (calbiochem), herbimycin A (Calbiochem), PP2 (Calbiochem) and PP3 (Calbiochem) were dissolved and stored at − 20 °C in DMSO. PKIs were dissolved in PBS prior to use and tested at concentrations of 1 nM, 3 nM, 5 nM, 10 nM, 20 nM, 50 nM, 100 nM, 250 nM, 500 nM and 1μM. pMHCI tetramer manufacture -------------------------- 2.3 Soluble biotinylated pMHCI monomers were produced as described previously ([@bib31]). Tetrameric pMHCI reagents were constructed by the addition of either R-Phycoerythrin (PE)-conjugated streptavidin (Molecular Probes, Invitrogen) or Alexa 488-conjugated streptavidin (Molecular Probes, Invitrogen) at a pMHCI:streptavidin molar ratio of 4:1. Conjugated streptavidin was added to a solution of soluble pMHCI in 5 equal aliquots at 20 min intervals and subsequently stored in the dark at 4 °C. pMHCI tetramer staining and flow cytometry: clones and splenoctyes ------------------------------------------------------------------ 2.4 10^5^ Mel13, Mel5 or ILA1 CTL were pre-treated at 37 °C with dasatinib at a range of concentrations (0--50 nM) for a series of durations up to 1 h. Mel13/Mel5 or ILA1 were then stained with either PE-conjugated HLA A2/ELAGIGILTV or HLA A2/ILAKFLHWL tetramer, respectively, at a final concentration of 10μg/ml for 20 min at 37 °C. The HLA DR⁎0101-restricted clone C6 was stained with HLA DR⁎0101/PKYVKQNTLKLAT PE tetramer for 20 min at 37 °C. After initial experiments to determine the optimal conditions of use, all subsequent experiments were performed by incubating T-cells ± 50 nM dasatinib for 30 min at 37 °C prior to tetramer staining. Unless otherwise stated dasatinib remained present throughout the staining procedure. Subsequent to tetramer staining, CTL clones were stained with anti-human CD8-FITC (clone SK1; BD Pharmingen) and 7-AAD (Viaprobe; BD Pharmingen) for 30 min on ice then washed twice with phosphate buffered saline (PBS); the HLA DR⁎0101-restricted clone C6 was stained with 7-AAD (Viaprobe; BD Pharmingen) only. For murine CTL, 5 x 10^5^ splenocytes were pre-treated with 50 nM dasatinib for 30 min at 37 °C, stained with H2-D^b^/ASNENMDAM PE-conjugated tetramer for 20 min at 37 °C and then anti-murine CD8-Cy5.5 for 30 min on ice, washed twice and re-suspended in PBS. Data were acquired using a FACSCalibur flow cytometer (BD) and analyzed using FlowJo software (Treestar Inc., Ashland, OR, USA). pMHCI tetramer staining and flow cytometry: human peripheral blood mononuclear cells ------------------------------------------------------------------------------------ 2.5 Frozen peripheral blood mononuclear cells (PBMCs) were thawed in a 37 °C water bath until a small clump of ice remained and then transferred into RPMI medium containing 10% FCS, 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM L-glutamine (Gibco) and 100 U DNase/ml (Roche Diagnostics Corporation, Indianapolis, IN, USA). PBMC were washed twice in this medium and then left to rest for 2 h at 37 °C. After 2 washes with PBS, 2 × 10^6^ PBMC were stained with live/dead^®^ fixable violet amine reactive dye (Invitrogen Corporation, Carlsbad, California, USA), washed and incubated for 30 min at 37 °C in PBS alone or PBS containing 50 nM dasatinib. Subsequently, PBMC in 50μl of PBS alone or PBS containing 50 nM dasatinib were stained for 20 min with pHLA A2 tetramers refolded around either CMV pp65~495-503~ (NLVPMVATV), EBV BMLFI~259-267~ (GLCTLVAML) or Melan-A/Mart-1~26-35~ (ELAGIGILTV) peptides. After 2 washes in PBS containing 1% FCS and 0.02% sodium azide (Sigma-Aldrich, St. Louis, MO, USA), cells were stained with a selection of the following cell surface monoclonal antibodies (mAbs): (i) anti-CD3-APC-Cy7 and anti-CD8-APC (BD Biosciences, San Jose, CA, USA); (ii) anti-CD4-PE-Cy5.5 (Caltag Laboratories, purchased through Invitogen Corporation, Carlsbad, California, USA); and, (iii) anti-CD8-quantum dot (QD)705, anti- CD14-Pacific Blue and anti-CD19-Pacific Blue, conjugated in-house according to standard protocols ([*http://drmr.com/abcon/index.html*](http://www.drmr.com/abcon/index.html)). The latter two mAbs were used to exclude CD14^+^ monocytes and CD19^+^ B cells, which can bind tetramer non-specifically, from the analysis. Finally, cells were washed and resuspended in PBS containing 1% paraformaledehyde (PFA). Stained PBMC were acquired on a BD LSR II (Becton Dickinson Immunocytometry Systems, San Jose, CA, USA) and analyzed using FlowJo software (Tree Star, Inc., Ashland, OR, USA). TCR downregulation and tetramer on-rate experiments --------------------------------------------------- 2.6 HLA A2-expressing C1R B cells ([@bib13]) were either pulsed with ELAGIGILTV peptide at a concentration of 10^--6^ M for 60 min at 37 °C or incubated in medium alone. After two washes with RPMI 1640 supplemented with 100 IU/ml penicillin and 100μg/ml streptomycin, 60,000 HLA A2^+^ C1R cells (pulsed or unpulsed) were incubated for 4 h at 37 °C with 30,000 Mel13 CTL that had been pre-treated with PBS ± 50 nM dasatinib for 30 min at 37 °C. Cells were then stained with anti-TCR-FITC (clone BMA 031; Serotec) and anti-CD8-APC (clone RPA-T8; BD Pharmingen) for 30 min on ice, washed twice and resuspended in PBS. Data were acquired using a FACSCalibur flow cytometer and analysed using FlowJo software (Treestar Inc., Ashland, OR, USA). Tetramer on-rate experiments were performed as previously described ([@bib16]). Fluorescence microscopy ----------------------- 2.7 10^5^ ILA1 CTL were treated with PBS ± 50 nM dasatinib for 30 min at 37 °C, then stained with Alexa 488-conjugated (Molecular Probes) HLA A2/ILAKFLHWL tetramer at a final concentration of 20μg/ml for 15 min at 37 °C. Following two washes with PBS, each sample was fixed in 2% paraformaldehyde. After fixing, ILA1 CTL were re-suspended in 100μl of 2% FCS/PBS and then spun on to a microscope slide at 550 rpm for 5 min using a cytospin. Samples were subsequently analysed on a Leica DM LB2 (Leica Microsystems) fluorescence microscope. IFNγ ELISpot assays ------------------- 2.8 CD8^+^ T-cell responses to islet autoantigen were detected by IFNγ ELISpot as described previously ([@bib7]) with the following modifications. PBMCs were pre-cultured at 37 °C/5% CO~2~ in single wells of 48-well plates at a density of 1 × 10^6^ cells in 0.5 ml TC medium (RPMI 1640 supplemented with antibiotics (Invitrogen) and 10% human AB serum (PAA, Somerset, UK)) containing the test peptide at a final concentration of 10μM. Control wells contained TC medium with an equivalent concentration of diluent (DMSO). After 24 h incubation, non-adherent cells were re-suspended using pre-warmed TC medium (2% AB serum), washed, brought to a concentration of 10^6^ cells/300μl, and then dispensed in 100μl aliquots into wells of 96-well ELISpot plates (Nunc Maxisorp; Merck Ltd., Poole, UK) pre-blocked with 1% bovine serum albumin in PBS and pre-coated with monoclonal anti-IFNγ (U-Cytech, Utrecht, NL). Assays were then developed according to the manufacturer\'s instructions; plates were dried and spots were counted using a BioReader 3000 (BioSys, Karben, Germany) and reported as total responder cells per 10^6^ PBMCs. Intracellular cytokine staining assays -------------------------------------- 2.9 10^6^ CTL were stimulated with specific peptide at a concentration of 10μg/ml for 6 h; brefeldin A (10μg/ml; Sigma-Aldrich) was added for the final 5 h. Unrelated peptide (10μg/ml) was used as a negative control. Briefly, the cells were fixed in paraformaldehyde (2%; Sigma-Aldrich), permeabilized with saponin (0.5%; Sigma-Aldrich), and labeled with APC-conjugated anti-TNFα mAb (BD Pharmingen). The cells were evaluated using a FACSCalibur flow cytometer (BD Biosciences). At least 10,000 events gated on forward and side scatter were analyzed using FlowJo software (Tree Star, Inc., Ashland, OR, USA). Corresponding isotype control antibodies were used to establish the quadrants for analysis. Mathematical modelling of the dasatinib effect on pMHCI tetramer staining ------------------------------------------------------------------------- 2.10 A mathematical model that relates staining intensity to tetramer binding kinetics has been described previously ([@bib16 bib28]). Briefly, in this model, the high binding avidity of pMHCI tetramers is accounted for by assuming that tetramers can engage up to three TCR molecules, forming a cluster of one, two, or three TCRs in which association and dissociation between the TCRs and the tetramer pMHCI sites occurs at a much higher rate than diffusive escape of temporarily unbound TCRs, tending to stabilize the cluster until the tetramer happens to become disassociated with TCR molecules at all four sites. Moreover, in singlet and duplet clusters, a free TCR can associate contacts with free pMHCI sites to form, respectively, a duplet or triplet cluster. The dasatinib effect is incorporated into this model by assuming that dasatinib alters the rate at which singlet and duplet clusters recruit free TCRs. Thus, the expression for the relative staining intensity I becomes:$$I = I_{\min} + \Delta_{I}\left( {\left( {K_{1}/K_{D}} \right)r_{0} + \delta_{D}\left( {K_{2}/K_{D}} \right)^{3}r_{0}^{2} + \delta_{T}^{2}\left( {K_{3}/K_{D}} \right)^{6}r_{0}^{3}} \right)\text{.}$$where K~D~ is the single-site dissociation constant; K~1~, K~2~ and K~3~ are kinetic parameters; I~min~ and Δ~I~ relate the read-out to the number of surface bound tetramers (I~min~ is a nuisance parameter, representing the background level); δ~*D*~ is the duplet recruitment enhancement factor and δ~*T*~ is the triplet recruitment enhancement factor. In the absence of dasatinib, we have δ~*D*~ = δ~*T*~ = 1, whereas these factors are greater than 1 if dasatinib promotes TCR recruitment. The scaled free TCR density, *r*~0~, is implicitly defined by the scaled conservation law:$$1 = r_{0} + \left( {K_{1}/K_{D}} \right)r_{0} + 2\delta_{D}\left( {K_{2}/K_{D}} \right)^{3}r_{0}^{2} + 3\delta_{T}^{2}\left( {K_{3}/K_{D}} \right)^{6}r_{0}^{3}$$(from [@bib16]). Association kinetics has been found empirically to be described very well by the biphasic exponential model:$$I\left( t \right) = I_{\min} + I_{\max,\;\text{fast}}\left( {1 - \exp\left( {\lambda_{\text{fast}}t} \right)} \right) + I_{\max,\;\text{slow}}\left( {1 - \exp\left( {\lambda_{\text{slow}}t} \right)} \right)$$where *I*(*t*) is the staining intensity at time *t* and *I*~max,\ fast~, I~max,\ slow~, *λ*~fast~, and *λ*~slow~ are positive parameters ([@bib28]). The goodness of fit is fair but not perfect ([Fig. 2](#fig2){ref-type="fig"}B); a better curve fit might be achieved with a more sophisticated model. However, we have taken a minimalistic approach to modelling the dasatinib effect because of the mechanistic uncertainties surrounding its mode of action. The duplet recruitment enhancement factor δ~*D*~ was estimated to equal 6.53 ± 2.93, whereas the triplet recruitment enhancement factor δ~T~ was estimated to equal 13.7 ± 7.38 (by simultaneous non-linear least- squares). The fit of the model indicates that the data are consistent with the hypothesis that dasatinib makes free TCRs more readily available to pMHCI tetramers for recruitment into duplet and triplet clusters. Results ======= 3 PKI treatment enhances pMHC tetramer staining of CD8^+^ and CD4^+^ T-cells -------------------------------------------------------------------------- 3.1 Incubation of ILA1, a CTL clone specific for the HLA A2-restricted epitope hTERT~540-548~ (ILAKFHWL), with the PKIs dasatinib or 3-(2-(1H-benzo\[d\]imidazol-1-yl)-6-(2-morpholinoethoxy)pyrimidin-4-ylamino)-4-methylphenol (Lck inhibitor II; Calbiochem) resulted in a \> 10 fold increase in pMHCI tetramer staining intensity ([Fig. 1](#fig1){ref-type="fig"}A). Pre-incubation with the PKI PP2 (Calbiochem) resulted in a moderate increase in tetramer staining; however, no significant enhancement was observed when CTL were pre-treated with herbimycin, PP3, genestein (Calbiochem) or staurosporine (Biomol) (data not shown). Importantly, PKI inhibitor treatment did not enhance staining with non-cognate pMHCI tetramer ([Fig. 1](#fig1){ref-type="fig"}B). Identical results were obtained for the HLA A2-restricted Melan A~26-35~ ELAGIGILTV-specific CTL clone Mel13 (data not shown). Dasatinib is an ATP-competitive, dual Src/Bcr-Abl kinase inhibitor that has recently entered the clinic for the treatment of chronic myelogenous leukaemia (CML) ([@bib27]). In addition, recent data shows that dasatinib can inhibit wild type and mutant forms of KIT, a class III receptor tyrosine kinase. As a result, dasatinib may also have a role to play in diseases associated with KIT activation loop mutations such as systemic mastiocytosis, acute myelogenous leukaemia (AML), seminoma, gastrointestinal stromal tumours and anal melanomas ([@bib26 bib2]). Dasatinib has been shown to be a potent inhibitor of the Src protein kinase Lck (IC~50~ = 0.4 nM) ([@bib5 bib29]). Furthermore, while dasatinib reversibly inhibits antigen-specific T-cell effector functions, it is not toxic to T-cells in the short term at concentrations \< 100 nM ([@bib29]). Indeed, T-cell clones incubated in 50 nM dasatinib for 24 h were able to regain responsiveness to antigen within 1 h of drug removal (data not shown). Dasatinib can also be used in flow cytometry-based applications without loss of cell viability ([@bib29]). These properties prompted us to select dasatinib for further investigation in the current study. The enhancement of pMHCI tetramer staining following dasatinib treatment was highly dose dependent ([Fig. 1](#fig1){ref-type="fig"}C). Maximal effect was achieved by exposing CTL to 50 nM dasatinib for 1 h, which resulted in an 89% increase in pMHCI tetramer staining intensity ([Fig. 1](#fig1){ref-type="fig"}C). Unexpectedly, pre-incubation of CTL with 50 nM dasatinib for as little as 30 s resulted in a 60% increase in pMHCI tetramer staining ([Fig. 1](#fig1){ref-type="fig"}D). Furthermore, incubation with 50 nM dasatinib for 30 min significantly enhanced the staining of both ILA1 and Mel13 CTL over a wide range of pMHCI tetramer concentrations ([Fig. 1](#fig1){ref-type="fig"}E&F). Pre-incubation with dasatinib also enhanced pMHCI tetramer staining of naïve murine F5 TCR CTL directly *ex vivo* ([Fig. 1](#fig1){ref-type="fig"}G) and pMHCII tetramer staining of a HLA DR⁎010-restricted CD4^+^ T-cell clone ([Fig. 1](#fig1){ref-type="fig"}H). Thus, pre-incubation with 50 nM dasatinib for 30 min provides a quick and easy way to enhance pMHC tetramer staining efficiency in both human (CD4^+^ and CD8^+^ T-cells) and murine systems. These effects are highly specific; increased pMHC tetramer binding only occurs in the presence of a cognate TCR/pMHC interaction ([Fig. 1](#fig1){ref-type="fig"}B&H). Dasatinib preferentially enhances pMHCI staining of T-cells bearing low affinity TCRs ------------------------------------------------------------------------------------- 3.2 In order to dissect further the effects of dasatinib, we examined pMHCI tetramer staining using several altered peptide ligands for the ILA1 CTL clone that differ in their binding affinity for the ILA1 TCR by \> 100-fold ([Table 1](#tbl1){ref-type="table"}). Pre-incubation with dasatinib enhanced staining efficiency with all variant pMHCI tetramers ([Fig. 2](#fig2){ref-type="fig"}A). The percentage increase in tetramer staining afforded by pre-incubation with dasatinib for 8E, 5Y, 4L, ILA index, 3G8T and 3G pMHCI tetramers was 675%, 1825%, 324%, 111%, 75% and 26%, respectively. Thus, the benefits of dasatinib pre-treatment in terms of enhanced tetramer staining intensity are greater for peptide variants that exhibit weaker interactions with the ILA1 TCR. The intensity of pMHCI tetramer staining in the presence and absence of dasatinib treatment was plotted against the monomeric TCR/pMHCI dissociation constants and a curve fitted according to the mathematical model outlined in the Materials and methods ([Fig. 2](#fig2){ref-type="fig"}B). The data demonstrate that, in the absence of dasatinib, there is a sharp reduction in tetramer staining intensity for ligands with TCR/pMHCI K~D~ \> 35μM; this is consistent with previous observations ([@bib16]). In the presence of dasatinib, however, the TCR/pMHCI affinity threshold for this sharp drop-off did not occur until the KD exceeded 200μM. In fact, dasatinib treatment allows detectable staining of the ILA1 clone even when the agonist TCR/pMHCI K~D~ exceeds 500μM. Dasatinib treatment therefore enables the physical detection of CTL bearing TCRs with low affinity for the cognate pMHCI ligand that would otherwise be undetectable using pMHCI tetramer staining alone. Dasatinib reduces pMHCI tetramer-induced cell death --------------------------------------------------- 3.3 Previous studies have shown that soluble pMHCI tetramer-induced signaling can trigger cell death ([@bib24 bib32 bib12 bib6]). This can reduce the number of live cells that remain after pMHCI tetramer staining under normal conditions. Dasatinib blocks antigen-specific signaling and subsequent T-cell effector functions ([@bib29]). Consequently, we hypothesized that dasatinib could prevent pMHCI tetramer-induced cell death. Indeed, the percentage of tetramer-positive cells that died when PBMCs were stained directly *ex vivo* with pMHCI tetramers representing epitopes derived from cytomegalovirus (CMV) and Epstein-Barr virus (EBV) was reduced in the presence of dasatinib ([Fig. 3](#fig3){ref-type="fig"}). Therefore, dasatinib exerts three beneficial effects: (i) it increases the intensity of pMHCI and pMHCII tetramer staining; (ii) it preferentially enhances pMHCI tetramer staining of T-cells bearing low affinity TCRs; and, (iii) it reduces pMHCI tetramer-induced cell death. Substantial improvements in the detection of antigen-specific CD8^+^ T-cells directly ex vivo --------------------------------------------------------------------------------------------- 3.4 The above findings suggest that dasatinib treatment might enable the identification of low avidity antigen-specific CD8^+^ T-cells directly *ex vivo* that cannot be 'seen' in the absence of the drug. To test this idea, we first examined the staining of CTL lines in the presence or absence of 50 nM dasatinib. Staining improvements were observed in three different CTL lines raised against the Melan-A/Mart-1~26-35~ epitope (ELAGIGILTV) and three different CTL lines stimulated with the influenza matrix M1~58-66~ epitope (GILGFVFTL), all derived from HLA A2^+^ individuals. Representative data are shown in [Fig. 4](#fig4){ref-type="fig"}A. In all cases, dasatinib treatment substantially enhanced the staining intensity of cognate CD8^+^ T-cells without concomitant increases in the tetramer-negative population. In accordance with the results above, CD8^+^ T-cells that stained poorly with pMHCI tetramer exhibited the greatest benefit from dasatinib treatment. The staining intensity of all cognate CD8^+^ T-cells increased by at least 2-fold after dasatinib treatment, but T-cells that bound tetramer weakly exhibited increases of \> 20-fold in their fluorescence intensity. In many cases, larger populations of cells that stained with the corresponding pMHCI tetramer were detected after dasatinib treatment. This increase in tetramer^+^ cells after dasatinib treatment likely reflects the combined effects of a lower detection threshold in terms of TCR/pMHCI affinity and the fact that dasatinib reduces pMHCI tetramer-induced cell death ([Fig. 3](#fig3){ref-type="fig"}). Subsequently, we examined whether dasatinib could enhance pMHCI tetramer staining of cognate CD8^+^ T-cells in direct *ex vivo* PBMC samples and enable the detection of antigen-specific CD8^+^ T-cells that are \'invisible\' with routine staining procedures. Indeed, a substantial increase in both pMHCI staining intensity and the percentage of antigen-specific CD8^+^ T-cells was observed at both 4 °C and 37 °C in PBMC samples stained with HLA A2 tetramers specific for antigens derived from CMV, EBV and Melan A ([Fig. 4](#fig4){ref-type="fig"}B). pMHCI staining of functional autoimmune CTL following dasatinib treatment ------------------------------------------------------------------------- 3.5 We next examined pMHCI tetramer staining of IE6, a preproinsulin (PPI~15-24~)-specific HLA A2- restricted autoreactive CTL clone isolated from a patient with type 1 diabetes. This CTL clone produces TNFα, IFNγ and MIP1*β* on stimulation with target cells pulsed with cognate PPI-derived peptide antigen ([Fig. 5](#fig5){ref-type="fig"}A), and does not stain with cognate pMHCI tetramer using conventional staining procedures ([Fig. 5](#fig5){ref-type="fig"}B). An identical result was obtained for 2D6, a different PPI-specific CTL clone isolated from a patient with type I diabetes (data not shown). Dasatinib treatment allowed both CTL clones to bind cognate tetramer without affecting staining with non-cognate tetramer ([Fig. 5](#fig5){ref-type="fig"}B & data not shown). In keeping with these findings, dasatinib treatment allowed the identification of a HLA A2/PPI~15-24~ tetramer-positive population directly *ex vivo* from a type I diabetic patient, consistent with a corresponding IFNγ ELISpot response to PPI~15-24~ peptide of 13 responder cells per 10^6^ PBMCs ([Fig. 5](#fig5){ref-type="fig"}C & data not shown). Dasatinib did not increase direct *ex vivo* HLA A2/PPI~15-24~ tetramer staining in healthy HLA A2-matched control subjects ([Fig. 5](#fig5){ref-type="fig"}C). A seven-fold increase in the percentage of autoreactive CTL was observed when short-term lines expanded from two type I diabetic patients were stained in the presence of dasatinib ([Fig. 5](#fig5){ref-type="fig"}C & data not shown). Thus, dasatinib treatment allows the detection of functional autoreactive CTL that are otherwise undetectable with standard staining conditions. How does dasatinib exert its beneficial effects on pMHC tetramer staining? -------------------------------------------------------------------------- 3.6 Previous studies have demonstrated that incubation with Src kinase inhibitors results in enhanced TCR and CD8 expression at the cell surface ([@bib18 bib10]). Consistent with these observations, we have recently demonstrated that increased levels of TCR and CD8 are seen at the cell surface following incubation with dasatinib for 4 h ([@bib29]). Initially, therefore, we investigated this increase in TCR and CD8 levels as a possible mechanism for the observed effects on tetramer binding. The beneficial effects of dasatinib on pMHCI tetramer staining were observed within seconds of dasatinib treatment ([Fig. 1](#fig1){ref-type="fig"}D), whereas significant increases in TCR and CD8 levels were not observed until \> 30 min ([Fig. 6](#fig6){ref-type="fig"}). Therefore, this time dependent accumulation of TCR and CD8 at the cell surface cannot explain the effects of dasatinib on tetramer binding. We next investigated whether the mechanism of PKI action operates through TCR- or CD8-mediated effects. Dasatinib treatment enhanced pMHCI tetramer staining of the HLA A2-restricted ELAGIGILTV-specific Melc5 CTL clone and a CTL line raised against the Melan-A/Mart-1~26-35~ epitope (ELAGIGILTV) with both wildtype and CD8-null (DT227/8KA) tetramers ([Fig. 7](#fig7){ref-type="fig"}A&B), thereby demonstrating that PKIs can exert their effects in the absence of a pMHCI/CD8 interaction. Thus, consistent with effects on pMHCII tetramer binding ([Fig. 1](#fig1){ref-type="fig"}), dasatinib does not enhance pMHCI tetramer binding via CD8-mediated effects. TCR expression levels are not static and TCRs are constantly being down-regulated from the cell surface ([@bib14]). TCR internalization is thought to be mediated by three different mechanisms: (i) constitutive recycling of the TCR between intracellular compartments and the plasma membrane in resting cells by an unknown mechanism ([@bib11]); (ii) protein kinase C activation ([@bib22 bib11]); and, (iii) lck-mediated tyrosine phosphorylation following TCR ligation by specific pMHCI ligand. Dasatinib has been shown to target lck and therefore has the potential to inhibit the latter pathway of TCR down-regulation. Indeed, dasatinib treatment was found to block antigen-induced TCR downregulation from the CTL surface ([Fig. 8](#fig8){ref-type="fig"}A). pMHC tetramers are rapidly internalized under normal staining conditions ([@bib30]) and therefore we reasoned that dasatinib might exert its beneficial effects by blocking this process. To this end, fluorescence microscopy was performed in the presence and absence of dasatinib ([Fig. 8](#fig8){ref-type="fig"}B). HLA A2/ILAKFLHWL-Alexa488 tetramer capping and internalization was blocked in the presence of dasatinib and remained on the cell surface where it formed a ring that was visibly brighter than tetramer that had been internalized ([Fig. 8](#fig8){ref-type="fig"}B). Thus, by preventing TCR downregulation, PK inhibition acts to drive the system towards a higher number of surface TCRs and a higher number of potential productive engagements with pMHC tetramer. This has the effect of increasing tetramer on-rate, at least in pMHCI systems ([Fig. 9](#fig9){ref-type="fig"}). Discussion ========== 4 pMHC tetramer technology has revolutionized the study of antigen specific T-cells. However, one major limitation of this technique is that pMHCI, and most likely pMHCII, tetramer staining is dependent on a distinct TCR affinity threshold ([@bib16]). Consequently, pMHC tetramers fail to identify T-cells that express TCRs with low affinity for cognate antigen; such low affinity interactions characterize TCR/pMHCI binding in tumor-specific and autoreactive CD8^+^ T-cells ([@bib9]). Here, we demonstrate that a short incubation with a reversible PKI such as dasatinib results in three major benefits in terms of pMHC tetramer staining. First, substantial improvements in pMHC tetramer staining intensity are observed. This effect applies to both CD4^+^ and CD8^+^ T-cells ([Fig. 1](#fig1){ref-type="fig"}). Indeed, the beneficial effects are so striking even at low pMHC tetramer concentrations that dasatinib treatment could be used to conserve reagent. Second, dasatinib treatment reduces tetramer-induced cell death that has been previously reported to be an issue with pMHCI tetramer staining protocols ([@bib24 bib32 bib12 bib6]). Third, the benefits of dasatinib treatment are greater for TCR/pMHCI interactions of weak affinity and, as a result, dasatinib enhances the detection of low avidity CD8^+^ T-cells; this effect increases the number of CD8^+^ T-cells that can be detected directly *ex vivo*, particularly in the setting of tumor-specific and autoreactive CD8^+^ T-cell populations. Such effects are also likely to apply to CD4^+^ T-cells, which typically bind cognate pMHCII antigens with affinities lower than those reported for pMHCI systems ([@bib9]). Importantly, no increase in background staining was seen in any of the systems tested here ([Figs. 1B, 1H, 5B & 5C](#fig1 fig5){ref-type="fig"}). In fact the tetramer negative background was actually seen to decrease with dasatinib treatment in some staining experiments ([Figs. 4C & 5C](#fig4 fig5){ref-type="fig"}). Positive staining of PBMC in the presence of dasatinib only ever coincided with a positive ELISpot result to the relevant peptide. Dasatinib proved to be a particularly powerful tool in the detection of autoreactive CTL from type I diabetic patients. No increase in the PPI tetramer positive population was observed in healthy donors and indeed staining was only ever seen if a functional response to the preproinsulin peptide was evident. Our results suggest that the benefits of dasatinib treatment apply only to CTL that express TCR specific for the pMHCI tetramer in use and capable of recognizing cognate pMHCI (although function may be impaired in some low avidity CTL populations). This conclusion is further strengthened by the finding that the beneficial effects are TCR mediated and do not involve the CD8 coreceptor. Therefore dasatinib facilitates the specific TCR/pMHCI interaction rather than the non-specific pMHCI/CD8 interaction. Dasatinib prevents TCR downregulation and tetramer internalization from the cell surface. How does this effect result in faster tetramer on-rates and the beneficial effects described above? When an individual pMHC molecule in a pMHC tetramer engages a cell surface TCR, this engagement can be either 'productive' or 'non-productive' in terms of capturing the tetramer from solution ([Fig. 10](#fig10){ref-type="fig"}). A productive engagement requires a second pMHC in the tetramer to bind a second TCR before the first pMHC dissociates. As a result, the main factor that determines whether a pMHCI tetramer exhibits stable binding is likely to be the duration of the primary monomeric TCR/pMHCI interaction. 'Non-productive' engagements are more likely to occur for low affinity ligands as they dissociate rapidly from cell surface TCR. Dasatinib would act to prevent TCR downregulation after 'non-productive' engagement thereby maintaining TCRs at the cell surface where they are available for future interactions with pMHCI. In addition, dasatinib treatment is likely to block the internalization of non-triggered TCRs that has been shown to occur at the same time as the internalization of pMHCI-engaged TCRs ([@bib23 bib25]). Dasatinib and other effective PKIs prevent TCR downregulation, acting to maintain surface TCRs and enabling a higher number of potential productive engagements with pMHC tetramer as indicated by the red arrows ([Fig. 10](#fig10){ref-type="fig"}). Increased TCR availability at the T cell surface would increase the likelihood of a 'productive engagement' for low affinity ligands. This effect would be less obvious for higher affinity ligands, where slow TCR/pMHCI off-rates already increase the likelihood that a second pMHCI arm can bind before dissociation of the first monomeric TCR/pMHCI interaction. The beneficial effects observed do not involve significant contribution from the CD8 coreceptor suggesting that CD8 availability at the T cell surface must be sufficient for pMHCI tetramer staining in the absence of dasatinib treatment. In summary, we have demonstrated that a short incubation with reversible PKIs such as dasatinib substantially improves the staining intensity of cognate T-cells with pMHC tetramers and can expose concealed antigen-specific T-cells that bear low affinity TCRs. These benefits are restricted to cognate T-cells and are not accompanied by concomitant increases in background staining. As such, dasatinib has many potential uses; (i) enhancement of routine pMHCI tetramer staining in the lab, (ii) conservation of this valuable reagent as the amount required is significantly reduced in the presence of dasatinib, (iii) detection of low avidity CTL populations which will be of particular importance for researchers and clinicians studying diseases characterized by CTL of this type such as chronic viral infection, auto-immune and neoplastic disease. The beneficial effects of this reagent also extend to MHCII tetramers, with which staining is often very poor or not visible at all. We conclude, that this simple and universally applicable technique is likely to be beneficial in all studies of antigen-specific T-cells. AL and AKS are funded by the Cardiff University Link Chair scheme. LW and MC are funded by The Wellcome Trust. DAP is a Medical Research Council (MRC) Senior Clinical Fellow. We would like to thank Awen Gallimore, James Matthews and Sarah Lauder for the provision of mouse spleens and murine pMHCI tetramers used in this study. We are grateful to Sefina Arif for performing the preproinsulin cytokine ELISpot assays. Studies on patients with type 1 diabetes in this report were supported by the Juvenile Diabetes Research Foundation grant 7-2005-877. MP and AS acknowledge financial support from the Department of Health via the National Institute for Health Research (NIHR) Comprehensive Biomedical Research Centre award to Guy\'s & St Thomas\' NHS Foundation Trust in partnership with King\'s College London. We are extremely grateful to Mario Roederer for the provision of custom-conjugated monoclonal antibodies. This research was supported by the Intramural Research Program of the National Institutes of Health, Vaccine Research Center, National Institute of Allergy and Infectious Diseases. ::: {#fig1 .fig} Fig. 1 ::: {.caption} ###### Dasatinib substantially improves pMHC tetramer staining intensity. A. 10^5^ ILA1 CTL were re-suspended in 40μl of PBS ± 50 nM dasatinib or Lck inhibitor II (Calbiochem), then incubated at 37 °C for 30 min. Cells were then stained with cognate HLA A2/ILAKFLHWL-PE tetramer at a final concentration of 10µg/ml for 20 min at 37 °C, washed twice in PBS and analyzed on a FACSCalibur (BD) flow cytometer. A \> 10-fold increase in median fluorescence intensity (MFI) was observed after treatment with 50 nM dasatinib (blue) or LcK inhibitor II (red) compared to staining without PKI pre-treatment (green line). B. 10^5^ ILA1 CTL were treated with various concentrations of dasatinib for 30 min at 37 °C, then stained with either HLA A2/ILAKFLHWL tetramer or the non-cognate HLA A2/ELAGIGILTV tetramer for 20 min at 37 °C before washing with PBS. C. 10^5^ ILA1 CTL were resuspended in 40μl of PBS ± the indicated concentration of dasatinib and incubated for 60 min at 37 °C. Cells were then stained with cognate HLA A2/ILAKFLHWL-PE tetramer at a final concentration of 10 μg/ml for 20 min at 37 °C and washed twice in PBS prior to flow cytometric analysis. D. As (A), but ILA1 CTL were incubated with 50 nM dasatinib for various times prior to staining with pMHCI tetramer. For this experiment the drug was washed off prior to staining. E. As (A), but tetramer concentration was varied to stain CTL pre-treated ± 50 nM dasatinib for 30 min. F. 10^5^ Mel13 CTL were stained with various concentrations of HLA A2/ELAGIGILTV tetramer following incubation ± 50 nM dasatinib for 30 min. G. 5x10^5^ splenocytes from an F5 TCR transgenic Rag^+^ mouse were resuspended in PBS ± 50 nM dasatinib and incubated for 30 min at 37 °C. Cells were subsequently stained with H2-D^b^/ASNENMDAM-PE tetramer for 20 min at 37 °C followed by anti-CD8 Cy5.5 for 30 min on ice prior to two washes in PBS and analysis by flow cytometry. H. 10^5^ cells of the HLA DR⁎0101-restricted, influenza virus A HA~307-319~ PKYVKQNTLKLAT-specific CD4^+^ clone C6 were incubated with PBS ± 50 nM dasatinib for 30 min at 37 °C, then stained with cognate PE-conjugated tetramer for 20 min at 37 °C. Samples were washed with PBS before flow cytometric analysis. Irrelevant tetramer was used as a negative control in all cases. :::  ::: ::: {#fig2 .fig} Fig. 2 ::: {.caption} ###### Dasatinib treatment preferentially increases the ability of pMHCI tetramers to stain T-cells bearing low affinity TCRs. A. 10^5^ ILA1 CTL were stained with 10μg/ml PE-conjugated HLA A2 tetramer folded around the 8E, 5Y, 4L, index (ILAKFLHWL), 3G8T or 3G peptides for 20 min at 37 °C following incubation ± 50 nM dasatinib for 30 min at 37 °C. For all samples, data were acquired with a FACSCalibur flow cytometer (BD) and analyzed using FlowJo software. Irrelevant tetramer was used as a negative control. B. The MFI of tetramer staining for all of the variants in the presence and absence of dasatinib displayed in (A) are plotted against the monomeric affinity of TCR/pMHCI interactions previously measured for each of these variants expressed as the dissociation constant (K~D~) ([Table 1](#tbl1){ref-type="table"}). Curves were fitted as described in the Materials and methods. :::  ::: ::: {#fig3 .fig} Fig. 3 ::: {.caption} ###### Dasatinib reduces pMHCI tetramer-induced cell death. Increased tetramer staining in the presence of dasatinib appears to be due partly to reduced cell death. It is known that pMHCI tetramer-induced signaling can trigger cell death ([@bib24 bib32 bib12 bib6]). Cell death induced by tetramer staining was assessed using the amine-reactive viability dye GrViD at the end of the staining procedure, which is spectrally distinct from ViViD. PBMC were stained with ViViD to identify and allow exclusion of dead and dying cells prior to the addition of pMHCI tetramer; GrVID staining was performed after pMHCI tetramer and surface antibody staining. Data were acquired on a BD LSR II flow cytometer and analyzed using FlowJo software. ViViD^+^, CD14^+^ and CD19^+^ cells were excluded from the analysis and the frequency of GrViD-positive cells was assessed in the tetramer-positive CD3^+^CD8^+^ T-cell populations. Representative flow profiles are shown here for CD8^+^ T-cells specific for the HLA A2-restricted epitopes CMV pp65~495-503~ (NLVPMVATV) and EBV BMLFI~259-267~ (GLCTLVAML). The frequencies of dead cells varied depending on the tetramer used, but the frequency of GrViD-positive dead cells within the tetramer-positive population was always lower in the presence of 50 nM dasatinib. These data, together with comparable results in other systems (data not shown), suggest that the cumulative cell death over the time course of a staining experiment could be substantially reduced by treatment with dasatinib. :::  ::: ::: {#fig4 .fig} Fig. 4 ::: {.caption} ###### Dasatinib enhances the visualization of antigen-specific CD8^+^ T-cells in mixed cell populations. A. Staining of HLA A2-restricted CTL lines expanded from PBMC by one round of stimulation with the influenza matrix M1~58-66~ peptide (GILGFVFTL) or the Melan-A/Mart-1~26-35~ peptide (ELAGIGILTV). Lines were stained with cognate tetramer ± pre-treatment with 50 nM dasatinib for 30 min at 37 °C. B. Flow cytometric profiles of live CD3^+^ lymphocytes stained with HLA A2 tetramers folded around the EBV BMLF1~259-267~ (GLCTLVAML), CMV pp65~495-503~ (NLVPMVATV) or Melan-A/Mart-1~26-35~ (ELAGIGILTV) peptide epitopes. 2x10^6^ PBMC were stained with the amine-reactive viability dye ViViD, then stained with tetramer (1μg in minimal staining volume) ± pre-treatment with dasatinib for 30 min at 37 °C. Cells were then stained with cell surface markers as described in the Materials and methods; a dump channel was used to exclude dead cells, CD14^+^ and CD19^+^ cells from the analysis. Boolean gating was carried out to exclude aggregates. Data were acquired with a BD LSR II flow cytometer and analyzed using FlowJo software. :::  ::: ::: {#fig5 .fig} Fig. 5 ::: {.caption} ###### Dasatinib allows detection of autoreactive CTL. A. CTL clone IE6, specific for the HLA A2-restricted epitope PPI~15-24~, was activated with either CMV pp65~495-503~ or PPI~15-24~ peptide for 6 h at 37 °C and then assayed for TNFα production by intracellular cytokine staining as detailed in the Materials and methods. B. Staining of CTL clone IE6 with either an irrelevant or HLA A2/PPI~15-24~ tetramer ± pre-treatment with 50 nM dasatinib for 30 min at 37 °C. C. Representative stainings with HLA A2/PPI tetramer ± pre-treatment with 50 nM dasatinib for 30 min at 37 °C. Left panels: control subject PBMCs; middle panels: type I diabetic patient PBMCs; right panels: a short-term line expanded by one round of peptide stimulation from a type 1 diabetic patient. :::  ::: ::: {#fig6 .fig} Fig. 6 ::: {.caption} ###### Dasatinib results in a time dependent increase in TCR and CD8 expression levels at the CTL cell surface. The ILA1 CTL clone was treated with PBS ± 50 nM dasatinib at 37 °C and 10^5^ CTL were removed from the medium at 0, 10, 30, 60, 180 and 250 min. CTL were subsequently stained with anti-CD8 FITC (clone SK1; BD, Pharmingen; left panel) or anti-TCR FITC (clone BMA 031; Serotec; right panel) for 30 min on ice, washed twice and resuspended in PBS. Data were acquired on a FACSCalibur flow cytometer (BD) and analyzed using FlowJo software. :::  ::: ::: {#fig7 .fig} Fig. 7 ::: {.caption} ###### Beneficial effects of dasatinib are not CD8-mediated. A. Melc5 CTL were pre- treated with PBS ± 50 nM dasatinib for 30 min at 37 °C, then stained with HLA A2 DT227/8KA cognate tetramer for 20 min at 37 °C. After washing twice, data were acquired on a FACSCalibur flow cytometer (BD) and analyzed using FlowJo software. B. Staining of HLA A2-restricted CTL lines expanded from PBMC by one round of stimulation with the Melan-A/Mart-1~26-35~ peptide (ELAGIGILTV). Lines were stained with either wild type or CD8 null cognate tetramer ± pre-treatment with 50 nM dasatinib for 30 min at 37 °C. :::  ::: ::: {#fig8 .fig} Fig. 8 ::: {.caption} ###### Dasatinib blocks antigen-induced TCR downregulation and tetramer internalization from the cell surface. A. Mel13 CTL were pre-treated with PBS ± 50 nM dasatinib and exposed to C1R-A2 B cells previously pulsed with 10^-6^M ELAGIGILTV peptide or medium alone for 4 h at 37 °C. Cells were subsequently stained with anti-TCR-FITC (clone BMA 031; Serotec) and anti-CD8-APC (clone RPA-T8; BD Pharmingen) mAbs for 30 min on ice, washed twice and resuspended in PBS. Data were acquired on a FACSCalibur flow cytometer (BD) and analyzed using FlowJo software. B. 10^5^ ILA1 CTL were pre-treated with PBS (i & ii) or PBS + 50 nM dasatinib (iii & iv) for 30 min at 37 °C, then stained with 20μg/ml HLA A2/ILAKFLHWL-Alexa488 tetramer for 15 min at 37 °C. Microscopy was performed as described in the Materials and methods. :::  ::: ::: {#fig9 .fig} Fig. 9 ::: {.caption} ###### Dasatinib enhances pMHCI tetramer on-rate. Rate of HLA A2/hTERT~540-548~ (ILAKFLHWL) tetramer recruitment to the cell surface of clone ILA1 is substantially enhanced following treatment of CTL with 50 nM dasatinib for 30 min at 37 °C. Subsequent to treatment with dasatinib, on-rate experiments were performed and analyzed as described previously ([@bib16]). Curves represent the following rate estimates: fast rate 0.14/min, slow rate 0.04/min (tetramer only); fast rate 0.42/min, slow rate 0.06/min (dasatinib + tetramer). :::  ::: ::: {#fig10 .fig} Fig. 10 ::: {.caption} ###### Dasatinib prevents down-regulation of 'empty' TCRs. Proposed model for the mechanism by which dasatinib enhances cognate tetramer staining. Dasatinib treatment prevents TCR and coreceptor down-regulation and maintains these receptors at the cell surface, thereby increasing molecular availability for further capture of pMHC tetramer from solution. :::  ::: ::: {#tbl1 .table-wrap} Table 1 ::: {.caption} ###### Affinity measurements of the interaction between the ILA1 TCR and hTERT~540-548~ pMHCI variants ::: LIGAND 8E 5Y 4L Index 8Y 3G8T 3G ---------------------------- -------- ----- ----- ------- ------ ------ ----- K~D~ ILA1 TCR Binding (μM) \> 500 242 117 36.6 22.6 4.04 3.7 Summary of the results obtained by nonlinear analysis of surface plasmon resonance binding equilibrium experiments as detailed in ([@bib16]) and ([@bib21]). K~D~ values were determined by analyzing the data in nonlinear curve fittings to the equation AB = B x AB~max~ / (K~D~ + B) assuming 1:1 Langmuir binding. ::: [^1]: 1 These authors contributed equally to this manuscript.

PubMed Central

2024-06-05T04:04:19.330373

2009-1-01

PMC3052437

Published online: November 18, 2010 Introduction {#sec1} ============ PPARγ is a nuclear hormone receptor activated by oxidized fatty acids and regulating many aspects of lipid metabolism and inflammation ([@bib14 bib17 bib18 bib24]). The major functions include regulation of adipocyte differentiation ([@bib40]) and lipid metabolism in macrophages ([@bib24 bib32 bib41]). The expression and activity of PPARγ in various cell types are strictly regulated ([@bib16 bib36 bib37 bib40]). However, expression of the receptor and the presence of appropriate ligands are usually not sufficient to elicit optimal or maximal responses. Further transcriptional mechanisms contribute to facilitate or restrict responsiveness leading to cell type- or condition-specific gene expression pattern ([@bib1 bib3 bib4 bib19]). Here, we aimed to understand the impact of extracellular signals on PPARγ activity in macrophages and DCs. Both cell types represent highly specialized but heterogeneous cell populations of the immune system. Macrophages originate from bone marrow progenitors committed to the monocytic lineage ([@bib7]). Monocytes are recruited to sites of inflammation and turn into macrophages. The immune phenotype of macrophages depends on the cellular environment and presence of various activator molecules ([@bib9]). In addition to pathogen clearance they also regulate resolution of inflammatory responses. These opposing or polarized activities are initiated and maintained by immunomodulatory factors such as cytokines and microbial products and manifest in distinct activation states. Proinflammatory molecules, such as interferon-γ (IFN-γ) and tumor necrosis factor (TNF) or activators of Toll-like receptors (TLRs), result in classical activation of macrophages. In contrast, alternatively activated macrophages, which differentiate upon IL-4 stimulus ([@bib34]), exhibit a different phenotype provoking tolerance or T helper 2 (Th2) immune responses ([@bib6]). IL-4 induces both *Pparg* and 12/15-lipoxygenese, which synthesizes a potential ligand for PPARγ ([@bib13]). Similarly to macrophages, dendritic cells (DCs) are also capable of inducing both inflammatory and anti-inflammatory responses. DCs are sentinels of the immune system and connect innate and acquired immunity ([@bib35]). Human DCs can be modeled by monocytes exposed to granulocyte-monocyte colony stimulating factor (GM-CSF) and IL-4. This cell type has been shown to be exquisitely responsive to PPARγ activation ([@bib10 bib26 bib37]). This shared requirement of IL-4 invokes an intriguing similarity between alternatively activated macrophages and DCs. PPARγ activity has been analyzed irrespective of the inflammatory state of macrophages and DCs and prior reports focused on downstream effects of PPARγ on inflammatory reactions. Based on these, PPARγ is considered a negative regulator of macrophage activation ([@bib15 bib33]). This is believed to be mediated by the failed induction of inflammatory genes by proinflammatory transcription factors ([@bib21 bib29]). However, in adipocytes, but also in DCs, PPARγ induces as well as represses hundreds of genes ([@bib11 bib38]). These observations suggest that PPARγ responses are stringently controlled and determined by cell type and condition-specific factors. The identification of such factors could explain differences in PPARγ-evoked responses in subtypes of macrophages and DCs. We used gene-specific and global transcriptomics approaches in mouse and human macrophage subtypes and DCs to show that proinflammatory molecules inhibited, whereas IL-4 augmented, both PPARγ expression and ligand-induced transcriptional activity. Pharmacological and genetic evidence showed that this effect was mediated by the STAT6 transcription factor, which acted as a facilitator of PPARγ-mediated transcription. In addition, we proposed a mechanism by which STAT6 interacted with PPARγ and the cooperative binding of the two factors led to increased PPARγ responsiveness. Thus, these findings provide the molecular mechanism for robust PPARγ-regulated gene expression in these cell types. Results {#sec2} ======= Expression of *PPARg* Is Determined by the Activation State of Macrophages and DCs {#sec2.1} ---------------------------------------------------------------------------------- To define conditions of maximal *PPARg* expression and responsiveness, human monocyte-derived macrophages were activated either with IL-4 or with the proinflammatory cytokines IFN-γ, TNF, or lipopolysaccharide (LPS). AMAC-1, CD206, CD209, and CD23 were used as markers to define alternative or CD80, CD83, CD86, and HLA-DR to define classical activation of macrophages. Immature DCs were differentiated from monocytes with GM-CSF+IL-4, and LPS was used to induce maturation. CD1a and CD209 were used as markers of DC development (data not shown) ([@bib8 bib31]). *PPARg* was induced during monocyte-macrophage transition ([Figure 1](#fig1){ref-type="fig"}A) and its expression was further increased by IL-4 but decreased by IFN-γ. Similarly, *PPARg* was induced upon differentiation of immature DCs and was modestly upregulated upon DC maturation with LPS ([Figure 1](#fig1){ref-type="fig"}A). The induction of *PPARg* by IL-4 was rapid and specific and translated into increased levels of PPARγ protein in macrophages (brown nuclear staining) ([Figure 1](#fig1){ref-type="fig"}B). Neither *PPARa* nor *PPARd* showed similar expression ([Figures S1](#app3){ref-type="sec"}A--S1C available online). In contrast, PPARγ was essentially missing from IFN-γ-stimulated classically activated cells ([Figure S1](#app3){ref-type="sec"}B). In order to assess the in vivo expression distribution of PPARγ in macrophages, we surveyed tissues via immunohistochemistry ([Figures S1](#app3){ref-type="sec"}E--S1Y). PPARγ-positive macrophages were identified in tissues such as Peyer\'s patch ([Figures S1](#app3){ref-type="sec"}E--S1G), lamina propria of normal small intestinal villi ([Figures S1](#app3){ref-type="sec"}H--S1J), reactive lymph node ([Figures S1](#app3){ref-type="sec"}K--S1M), lymphoepithelial tissue of the tonsil ([Figures S1](#app3){ref-type="sec"}N--S1P), perivascular macrophages of lymph node ([Figures S1](#app3){ref-type="sec"}Q--S1S), and the lung ([Figures S1](#app3){ref-type="sec"}T--S1Y), showing expression preferentially in alternatively activated macrophages as determined by DC-SIGN staining. Next, we analyzed the ligand-induced transcriptional activity of PPARγ by treating cells with the agonist Rosiglitazone (RSG). *FABP4* (encoding FABP4, also known as aP2) ([@bib39]), a well-established PPARγ target gene, was strongly induced at both mRNA and protein levels in macrophages activated with IL-4 and in GM-CSF+IL-4-induced immature DCs ([Figures 1](#fig1){ref-type="fig"}C and 1D). Modest induction of *FABP4* by RSG was observed in nonactivated control and TNF-treated macrophages and upon IL-4 treatment. Importantly, IFN-γ, IFN-γ+TNF, and/or LPS inhibited *FABP4* induction in both macrophages and mature DCs. These effects were specific for PPARγ target genes as shown by the fact that neither LXRα nor PPARα, δ activity was affected ([Figure S1](#app3){ref-type="sec"}D and data not shown). Therefore, diverse signals could induce *PPARg* expression (e.g., LPS, GM-CSF), but only IL-4 could also augment transcriptional responsiveness as determined by target gene induction. This finding led to the hypothesis that IL-4 is a facilitator of PPARγ via a mechanism we termed IL-4-induced augmentation of PPARγ response. IL-4 Is an Enhancer of PPARγ Transcriptional Responses {#sec2.2} ------------------------------------------------------ To test the extent of IL-4 augmentation of PPARγ response, we performed microarray analysis of RSG-treated human macrophages ([Figures 1](#fig1){ref-type="fig"}E and 1F). In the absence of IL-4, RSG induced 88 and repressed 32 genes ([Figure 1](#fig1){ref-type="fig"}E). IL-4 treatment resulted in an increase in the number of both the RSG-induced (336) and -repressed (274) genes. IL-4 not only enabled PPARγ to regulate a larger set of genes (635 versus 120) but also increased the magnitude of responses on individual genes. 665 genes (out of 730 regulated by RSG in control or IL-4-treated macrophages) were more responsive to RSG in the presence of IL-4 ([Figure 1](#fig1){ref-type="fig"}F; [Table S1](#app3){ref-type="sec"}). These data suggest that IL-4-augmented PPARγ response applies to the vast majority of PPARγ-regulated genes. IL-4 Induces Augmented PPARγ Response in Both Mouse Macrophages and DCs {#sec2.3} ----------------------------------------------------------------------- IL-4 increased the expression of *Pparg* in monocytes isolated from bone marrow of C57Bl/6 mice ([Figure 2](#fig2){ref-type="fig"}A). *Fabp4* was induced by RSG and this change was augmented by IL-4 ([Figure 2](#fig2){ref-type="fig"}A). Alternative activation markers such as found in inflammatory zone 1 (*Fizz1*), mannose receptor (*Mr*), chitinase 3-like 3 (*Ym1*), and arginase 1 (*Arg1*) were induced by IL-4 (data not shown). RSG treatment barely induced the expression of PPARγ target genes, PPARγ angiopoietin-related protein (*Angptl4*), or *Fabp4* in mouse macrophages whereas the addition of IL-4 elicited responsiveness to RSG ([Figure 2](#fig2){ref-type="fig"}B). Induction of PPARγ target genes showed a similar pattern in mouse bone marrow-derived DCs ([Figure S2](#app3){ref-type="sec"}A). These data demonstrate that human and murine macrophages and DCs behave similarly with respect to the effects of IL-4 on PPARγ expression and responsiveness. PPARγ Is Dispensable for IL-4 Signaling {#sec2.4} --------------------------------------- Next, we assessed the relationship between PPARγ-IL-4 in macrophage-specific PPARγ-deficient mice carrying loxP-flanked and null alleles *Pparg*^fl/−^ ([@bib12]) and a lysozyme (*Lys*) *Cre* transgene ([@bib5]). PCR assessment demonstrated near complete recombination ([Figure S2](#app3){ref-type="sec"}B). This was confirmed by the complete loss of *Fabp4* induction by RSG ([Figure 2](#fig2){ref-type="fig"}C). No major differences in the induction of *Arg1* or *Ym1* markers of alternatively activated macrophages were detected in IL-4-treated bone marrow-derived macrophages from wild-type, *Pparg*^+/−^ *LysCre*, or *Pparg*^fl/−^ *LysCre* mice ([Figure 2](#fig2){ref-type="fig"}D). Similar results were obtained in peritoneal macrophages (data not shown). We performed microarray experiments to address the general contribution of PPARγ to IL-4 responses by using macrophages from *Pparg*^+/−^ and *Pparg*^fl/−^ *LysCre* mice. The vast majority of IL-4-regulated genes showed similar expression pattern. Neither the number of IL-4-regulated genes nor the magnitude of responses was affected by the absence of *Pparg* ([Figures 2](#fig2){ref-type="fig"}E and 2F; [Table S2](#app3){ref-type="sec"}). Importantly, no difference was detected in the induction of alternative activation markers by IL-4 ([Table S2](#app3){ref-type="sec"}). The majority of changing genes showed less than 2-fold difference between the two genotypes (*Pparg*^+/−^ and *Pparg* ^fl/−^ *LysCre*) ([Figures 2](#fig2){ref-type="fig"}E and 2F). Notably, this is also true for the nonoverlapping genes ([Figure 2](#fig2){ref-type="fig"}F). These turned out to be the lowest responders to IL-4 (1.5- to 2.5-fold) and although not significantly, the majority of them are regulated by IL-4 in both genotypes and no differentially expressed cluster could be detected in the heatmap or in the gene lists ([Figure 2](#fig2){ref-type="fig"}E; [Table S2](#app3){ref-type="sec"}). Very little influence of PPARγ was found when we compared the effects of RSG on IL-4-regulated genes and analyzed the coregulated ones ([Figure 2](#fig2){ref-type="fig"}E; [Figure S2](#app3){ref-type="sec"}C). These data indicate only a modest contribution of PPARγ to IL-4 signaling and are inconsistent with PPARγ being required per se for initiation of alternative activation. The induction of markers of alternative macrophage activation *Ym1*, *Arg1*, or *Fizz1* by IL-4 were not affected by RSG treatment ([Figure 2](#fig2){ref-type="fig"}G). This was in agreement with the global gene expression analysis. We obtained similar results in Th2 cell-type response-prone BALB/c mice and in peritoneal macrophages (data not shown). Taken together, these data suggest that PPARγ is largely dispensable for IL-4-regulated gene expression in macrophages. STAT6 Is Required for PPARγ-Induced Gene Expression {#sec2.5} --------------------------------------------------- To identify which downstream effectors of IL-4 impact PPARγ expression and/or activity, we used pharmacological inhibitors that distinguish between STAT6 and insulin receptor substrate-2 (IRS-2) pathways. Janus Tyrosine Kinase 3 (JAK3) inhibitor WHI-P131, but not JAK2 inhibitor TYRPhostin (or AG490) or phosphatidilinositol-3 kinase (PI3K) inhibitor wortmannin, inhibited induction of an IL-4-regulated gene, *AMAC1*, and of *PPARg*, indicating that these effects of IL-4 are mediated by JAK3 ([Figure 3](#fig3){ref-type="fig"}A). Similarly, JAK3 but not the JAK2 and PI3K inhibitors inhibited IL-4-induced augmentation of *FABP4* induction by RSG ([Figure 3](#fig3){ref-type="fig"}A). These results are consistent with the report that JAK3 is the major JAK isoform in myeloid cells ([@bib43]) and implicated its substrate STAT6 as the downstream IL-4 effector. Next, we derived bone marrow-derived macrophages from wild-type and STAT6-deficient mice. Induction of the alternative activation marker *Ym1* and *Pparg* by IL-4 shows STAT6 dependence ([Figure 3](#fig3){ref-type="fig"}B). IL-4-augmented induction of PPARγ target genes *Fabp4* and *Angptl4* by RSG was detected only in wild-type but not in STAT6-deficient macrophages ([Figure 3](#fig3){ref-type="fig"}B). To test whether STAT6 is required for PPARγ-mediated gene expression on a global scale, we performed microarray analyses and identified RSG-regulated genes in the absence or presence of IL-4 in wild-type and STAT6-deficient mice ([Figures 3](#fig3){ref-type="fig"}C and 3D). The absence of STAT6 had a major impact on PPARγ-regulated gene expression. Substantially more genes were induced in the presence of IL-4 ([Figure 3](#fig3){ref-type="fig"}D, upper panels 45 versus 158) and the fold inductions were larger ([Figure 3](#fig3){ref-type="fig"}C). Moreover, the vast majority (82%) (45+158 genes) of the 225 RSG-induced genes were regulated by PPARγ in a STAT6-dependent manner and 99 out of the 225 RSG-induced genes showed higher expression upon IL-4 treatment ([Table S3](#app3){ref-type="sec"}). 125 RSG-induced genes were repressed by IL-4, indicating that the expression pattern of *Angptl4* (i.e., repression by STAT6 and activation by PPARγ; [Figure 3](#fig3){ref-type="fig"}B) is not a gene-specific phenomenon but rather characteristic of a set of genes. Interestingly, IL-4 did not increase the number of genes repressed by RSG: 173 in the control versus 176 in the IL-4-treated cells. However, out of the 318 genes RSG repressed, 70% showed STAT6 dependence ([Figure 3](#fig3){ref-type="fig"}C). *Stat6* deletion decreased the number of RSG-silenced genes in the absence of IL-4 ([Figure 3](#fig3){ref-type="fig"}D, lower panels 148 versus 48), whereas they were increased in the presence of IL-4 (162 versus 228), indicating the existence of further STAT6-independent silencing mechanisms. Taken together, these data suggest that IL-4 augments PPARγ activity via STAT6. Furthermore, STAT6 is required for induction of the majority of PPARγ target genes (82%). These data are consistent with a general, facilitating role for STAT6. STAT6 Augments PPARγ Activity on Target Gene Promoters {#sec2.6} ------------------------------------------------------ We took several possible mechanisms of STAT6-mediated PPARγ facilitation into consideration. An obvious one is to see whether new protein synthesis was required for the enhancing effect or if it is purely transcriptional. IL-4-dependent induction of a STAT6-regulated gene, *AMAC1*, and that of *PPARg* was not affected by cycloheximide (CXM), suggesting a direct transcriptional event ([Figure 4](#fig4){ref-type="fig"}A). As expected, RSG could activate PPARγ independently of new protein synthesis as reflected in *FABP4* induction ([Figure 4](#fig4){ref-type="fig"}A). In the presence of CXM, IL-4 could still enhance ligand-induced *FABP4* expression, suggesting that STAT6-augmented PPARγ response did not require new protein synthesis. However, *FABP4* induction slightly decreased, indicating the contribution of new protein synthesis, most probably PPARγ protein production upon IL-4 treatment. In order to evaluate the contribution of the increased expression of the receptor, we transiently transfected cells to overexpress PPARγ. PPARγ target gene expression was enhanced by STAT6 when cotransfected and activated by IL-4, which should be independent of the induction of PPARγ ([Figure 4](#fig4){ref-type="fig"}B and data not shown). To prove that IL-4 directly influences PPARγ, we excluded some obvious indirect mechanisms. IL-4 was shown to increase the production of a PPARγ activator, 15d-PGJ~2~, via inducing 12/15-lipoxygenase ([@bib13]). IL-4 augmented PPARγ response well before the induction of 15-lipoxygenase in human macrophages, suggesting that STAT6 is unlikely to act via ligand generation ([Figure 4](#fig4){ref-type="fig"}C). We also excluded that STAT6 would generate an activator for the retinoid X receptor, the permissive dimerization partner for PPARγ, by using an RXR antagonist ([Figure 4](#fig4){ref-type="fig"}D). We also tested trichostatin A (TSA), a histone deacetylase inhibitor, but no difference could be observed in the nonactivated cells, suggesting that IL-4 does not act via suspension of histone deacetylation ([Figure 4](#fig4){ref-type="fig"}E). Further possible mechanism could be that STAT6 induces degradation of a repressor for *PPARg* or synthesis of an activator. By using a proteasome inhibitor, MG132, and translation inhibitor CHX ([Figures 3](#fig3 fig4){ref-type="fig"}A--4A), we excluded these possibilities as well. Interestingly, MG132 not only did not inhibit the enhancement but it further increased it, suggesting the presence of an activating factor that is degraded upon activation of PPARγ-STAT6. Such factor could be either the transcription factor itself or its coactivators. We cannot exclude the possibility that other inhibitory factors exist, which are activated through the proteasomal pathway. These experiments left us with the likely possibility that STAT6 acts on the promoter of PPARγ target genes. STAT6 Facilitates PPARγ Signaling at the Transcriptional Level {#sec2.7} -------------------------------------------------------------- We chose the prototypic target gene, *FABP4* ([@bib39]), to study PPARγ response at the promoter level. A 5 kb fragment of the human promoter responded to PPARγ activators and also to IL-4 and this latter could augment the effect of RSG in a reporter assay in two different cell lines, RAW264.7 and 293T ([Figure 5](#fig5){ref-type="fig"}A). This was surprising, because this fragment did not contain the human ortholog of the originally identified PPARγ response element, which we term here AdipoPPRE ([Figure 5](#fig5){ref-type="fig"}B; [@bib39]). By using deletions and mutations, we identified a response element for PPARγ:RXR ([Figures 5](#fig5){ref-type="fig"}C and 5D and data not shown). We termed this element MacPPRE referring to macrophages ([Figure 5](#fig5){ref-type="fig"}D). Electrophoretic mobility shift assays (EMSA) and reporter assays were carried out to show preferential activation and binding of PPARγ to the enhancers ([Figures 5](#fig5){ref-type="fig"}E and 5F). This specificity disappeared when we mutated the PPARγ binding site to the consensus AGGTCA ([Figure S3](#app3){ref-type="sec"}A). The human ortholog hAdipoPPRE ([@bib39]) exhibited similar enhancer activities as the hMacPPRE ([Figures S3](#app3){ref-type="sec"}B and S3C). Interestingly, the two most conserved regions in the entire promoter region in mammals are the Adipo and MacPPREs along with the core promoter indicating their functional importance ([Figure 5](#fig5){ref-type="fig"}G; [Figure S3](#app3){ref-type="sec"}C). Unexpectedly, we found a consensus conserved STAT6 binding site downstream to MacPPRE ([Figure 5](#fig5){ref-type="fig"}D; [Figure S3](#app3){ref-type="sec"}C), which was not present in the proximity of AdipoPPRE. This STAT6 response element was functional and as efficient ([Figure 5](#fig5){ref-type="fig"}H) as a known STAT6 enhancer from the *CCL11* gene ([Figure 5](#fig5){ref-type="fig"}I; [@bib23]). A short promoter fragment that contained the composite element (MacPPRE and the STAT6 element) ([Figure 5](#fig5){ref-type="fig"}J) behaved similarly as the original 5 kb fragment ([Figure 5](#fig5){ref-type="fig"}A), indicating that this fragment is responsible for the STAT6-augmented PPARγ response. Mutation of the STAT6 binding site resulted in the loss of responsiveness to IL-4 ([Figure 5](#fig5){ref-type="fig"}K) without affecting induction by RSG. Mutation of DR1 abolished RSG-induced activation and also almost completely eliminated the effects of IL-4 ([Figure 5](#fig5){ref-type="fig"}L). The hAdipoPPRE did not show IL-4 responsiveness ([Figure 5](#fig5){ref-type="fig"}M). When isolated DR1s (consensus or MacPPRE from FABP4) were tested, STAT6 was ineffective in enhancing transcriptional activity ([Figures S5](#app3){ref-type="sec"}E and S5F). Similarly, activity of a Gal-fusion PPARγ could not be augmented by STAT6 ([Figure S5](#app3){ref-type="sec"}G). Thus these results indicate the requirement for the STAT6 binding site in the composite element to augment PPARγ response. In Vivo Binding of STAT6 to PPREs {#sec2.8} --------------------------------- These in vitro and transfection-based analyses established the presence of two functional PPREs in the FABP4 promoter, one of which (MacPPRE) is a complex element conferring IL-4-augmented PPARγ responsiveness. Next, we extended our studies to observe in vivo occupancy of this element by endogenous PPARγ and STAT6, and also to see how widespread the interaction of two factors is on known PPREs and STAT6 binding sites. Chromatin immunoprecipitation (ChIP) showed IL-4-induced binding of PPARγ and STAT6 to PPREs in 293T cells transfected with PPARγ and STAT6 ([Figure 6](#fig6){ref-type="fig"}A). In addition, both PPARγ and STAT6 were enriched on AdipoPPRE in wild-type macrophages when compared to *Stat6*^−/−^ cells ([Figure 6](#fig6){ref-type="fig"}B). Furthermore, with quantitative ChIP analysis on several known PPREs and STAT6 binding sites, we could detect PPARγ binding to all the tested elements except the negative control *Hoxa1*, and importantly this binding was markedly enriched in wild-type animals when compared to *Stat6*^−/−^ macrophages ([Figure 6](#fig6){ref-type="fig"}C), indicating that the two transcription factors are likely to be in the same DNA binding complex in vivo. Importantly, MacPPRE was more sensitive to the presence of STAT6 than AdipoPPRE. Thus, STAT6 seems to be enriched on several PPREs in vivo, suggesting a functional interaction between STAT6 and PPARγ. STAT6 Facilitates PPARγ\'s DNA Binding and Interacts with the Receptor {#sec2.9} ---------------------------------------------------------------------- The binding of STAT6 to PPREs in PPARγ target genes\' promoter and the close proximity of STAT6 and PPARγ binding sites in the human *FABP4* promoter raised the possibility of physical interaction between the two transcription factors. We first performed oligoprecipitation experiments in a monocytic leukemia cell line, THP-1, and both endogenously expressed PPARγ and STAT6 could be pulled down with wild-type MacPPRE ([Figure 6](#fig6){ref-type="fig"}D). STAT6 binding was detected only after IL-4 administration. Importantly, mutations in either PPRE or STAT6 binding sites diminished PPARγ binding, whereas STAT6 binding was eliminated only if STAT6 site was mutated. Finally, the interaction of PPARγ and STAT6 could be also detected by coimmunoprecipitation with tagged, expressed proteins ([Figures 6](#fig6){ref-type="fig"}E and 6F). STAT6 could be pulled down with PPARγ ([Figure 6](#fig6){ref-type="fig"}E) and vice versa ([Figure 6](#fig6){ref-type="fig"}F). STAT6 could also be pulled down with purified PPARγ protein ([Figure 6](#fig6){ref-type="fig"}G). Although there is a bit of inconsistency regarding the ligand dependency of PPARγ and STAT6 interaction, three out of four experiments ([Figures 6](#fig6){ref-type="fig"}D, 6F, and 6G) support that their interaction is ligand dependent. These data suggest that PPARγ and STAT6 bind the response element in *FABP4* promoter in vivo. Additionally, STAT6, by interacting with PPARγ, facilitates and is required for efficient endogenous PPARγ binding. Discussion {#sec3} ========== A key issue in immunology is to generate specific cell types often with opposing activities. It is of importance to understand the molecular details of the transcriptional mechanisms leading to the development of such subtypes. We found a diverse pattern of PPARγ expression and activity among the various macrophage and DC subtypes, equipping the cells with differential ability to respond to certain lipid signals. Although several agents could induce the transcription of *PPARg* (IL-4, LPS, transforming growth factor beta), only IL-4 was capable of augmenting its activity. This enhancement was reflected in the number of genes and the magnitude of responses. These data are in agreement with previous reports documenting very few positively regulated genes and a lower level of responses under proinflammatory conditions induced by LPS or IFN-γ ([@bib42]) and suggested the existance of both positive and negative interactions between PPARγ and cytokine signaling. PPARγ has been described as a negative regulator of macrophage activation by transrepression ([@bib30]). However, this mechanism is unlikely to play a role here. PPARγ was also reported to be required for maturation of alternatively activated macrophages and disruption of the gene impaired alternative macrophage-linked functions in mice ([@bib28]). Formally, these data suggested that PPARγ acts upstream of IL-4 signaling. We note that Odegaard et al. used BALB/c mice with *Mx-Cre* and rather focused on secondary effects of IL-4 signaling. Their data might reflect strain-specific differences and/or involvement of additional complex feedback mechanisms. However, our results are in agreement with another report also using C57Bl/6 mice ([@bib22]) and suggest that PPARγ is dispensable for alternative activation per se. Although direct IL-4 responses are barely altered in PPARγ-deficient macrophages, STAT6 appears to be required for maximal PPARγ activation. Therefore, PPARγ might be more appropriately considered as a downstream effector in the hierarchy of IL-4-STAT6-PPARγ signaling. This scenario does not formally rule out that in vivo a liganded receptor can contribute to IL-4-regulated events in a more complex way. Therefore, the role of PPARγ in macrophages besides lipid handling remains to be mapped, whereas in human DCs it is linked to lipid metabolism and lipid antigen presentation ([@bib37 bib38]). IL-4 and PPARγ signaling appear to be connected at multiple levels. IL-4-mediated induction of *Pparg* and 12/15-lipoxygenase that could generate endogenous activators for PPARγ ([@bib13]) provides two plausible mechanisms for enhanced response. Induction of *PPARg* itself is partly responsible for the enhancement, but IL-4-enhanced PPARγ activity appears much earlier than the lipoxygenase mRNA could be detected. Our results, presented here, point to STAT6 as the regulator of PPARγ response. Our global expression analyses showed that whereas STAT6 is required for maximal PPARγ response, PPARγ was largely dispensable for IL-4 signaling. This asymmetry is likely to be functionally important to provide specificity and allow STAT6 to act independently as well. The basis of this is an interaction by which STAT6 improves PPARγ activity via binding to the enhancer of PPARγ target genes. This is supported by three largely independent lines of evidence: the gene expression profile of STAT6-deficient macrophages, ChIP analysis of PPARγ and STAT6 target genes, and oligoprecipitation with the identified MacPPRE of *FABP4*. We propose therefore that STAT6 acts as a licensing factor for PPARγ to provide cell type-specific gene expression by enhancing DNA binding. Importantly, this interaction is specific; none of the other characterized receptors (PPARα, δ and LXRα) is influenced by STAT6. Additionally, probably as a special case of a robust IL-4-augmented PPARγ response, a conserved, complex enhancer (MacPPRE) exists in the *FABP4* gene. The mouse ortholog of this complex element behaves similarly to the human. We could detect PPARγ binding at both MacPPRE and AdipoPPRE in mouse adipocytes by analyzing recent ChIP-on-chip and ChIP-seq data ([@bib20 bib27]). Although this complex enhancer has a binding site for STAT6 and contributes to the robustness of the response, it is not clear whether other genes also use bone fide STAT6 sites or rather protein-protein interactions for the enhancement. The demonstration that the two proteins can physically interact provides support for the latter. To explore this, we have made a model of the PPARγ:RXRα heterodimer (pdb3DZY) and the STAT6 dimer (based on the STAT3B-DNA complex \[pdb1BG1\]) on a B-form MacPPRE DNA element, which suggests that the two protein complexes sit in close proximity on the DNA (data not shown). Limited rearrangement of this complex could easily bring the proteins into direct physical contact. Alternatively, an indirect interaction might be mediated by corecruitment of a shared coregulator or even by the DNA acting as an allosteric effector, facilitating communication between the complexes. A crosstalk between transcription factors in order to cooperatively orchestrate gene expression is not unprecedented. Estrogen receptor was reported to require the presence of another transcription factor, Forkhead box A1 (FoxA1), for efficient DNA binding and gene expression regulation ([@bib3 bib19]). Recently, C/EBPs were reported to bind to the vicinity of PPARγ response elements in adipocytes ([@bib20 bib27]). We suggest that STAT6 is likely to fulfill a similar role for PPARγ in macrophages and DCs. Although the molecular details of these crosstalks are still elusive, STAT6 might facilitate DNA binding of PPARγ or the two factors could synergistically recruit cofactors and chromatin remodeling enzymes. Alternatively, STAT6 itself might act as a coactivator to provide more efficient transactivation. It is intriguing to speculate that regulated and graded usage of licensing and facilitating factors (C/EBP and STAT6) could define specific responses of PPARγ leading to distinct gene expression programs in the various cell types or tissues. Experimental Procedures {#sec4} ======================= Materials {#sec4.1} --------- Ligands: LG268, LG1208, gifts from M. Leibowitz (Ligand Pharmaceuticals), WY14643, Rosiglitazone (RSG), T0901317, and MG132 (Alexis Biochemicals), GW501516, and GW9662 were gifts from T.M. Willson (GlaxoSmithKline). Cytokines were obtained from Peprotech. All other reagents were obtained from Sigma, consumables from Eppendorf, or as indicated. Isolation and Culture of Cells {#sec4.2} ------------------------------ Human monocytes were isolated from healthy volunteer\'s buffy coat, obtained with a Regional Ethical Board permit from the Regional Blood Bank, via CD14 MicroBeads (Miltenyi Biotec) and treated with vehicle (ethanol:dimethyl-sulfoxide 1:1) or as indicated. For activation we used IL-4 (100 ng/ml), IFN-γ (100 ng/ml), TNF (50 ng/ml), *E. coli* (O55:B5 serotype) LPS (100 ng/ml). Thioglycolate-elicited macrophages were harvested from the peritoneal cavity 4 days after injection of 3 ml 3% thioglycolate solution; bone marrow cells were isolated from the femur of mice. Mouse monocytes were isolated from bone marrow via negative selection method with magnetic separation (Miltenyi). Bone marrow cells were differentiated to macrophages by M-CSF (20 ng/ml) or to DCs by GM-CSF (20 ng/ml) and IL-4 (20 ng/ml) for 10 days. For activation we used mouse IL-4 (20 ng/ml), IFN-γ (20 ng/ml), TNF (20 ng/ml), *E. coli* (O55:B5 serotype) LPS (100 ng/ml). Real-Time Quantitative PCR {#sec4.3} -------------------------- Total RNA was isolated with Trizol Reagent (Invitrogen). RNA was reverse transcribed with High Capacity cDNA Archive Kit (Applied Biosystems). Transcript quantification was performed by quantitative real-time PCR via Taqman probes. Transcript levels were normalized to cyclophilin A or 36B4. Details of primers are in [Table S4](#app3){ref-type="sec"}. Immunoblotting {#sec4.4} -------------- Total cell lysates or nuclear extracts were resolved in SDS-PAGE and immunoblotted with FABP4 (Cayman Chemical), GAPDH (Abcam), Flag (M2, Sigma-Aldrich), V5 (Serotech), PPARγ (E8), or STAT6 (M-20-Santa Cruz) antibodies as indicated. Transient Transfection {#sec4.5} ---------------------- RAW264.7 cells were electroporated (300V for 15 ms); COS1 and HEK293T cells were transfected in triplicates with polyethyleneimine. Luciferase reporter activity was determined with Luciferase Assay System (Promega) and normalized to beta-galactosidase activity. Electrophoretic Mobility Shift Assay {#sec4.6} ------------------------------------ PPARα, γ, δ and RXR were in vitro transcribed and translated with T7 Quick TNT Kit (Promega). DNA was labeled in a random priming reaction (Fermentas) with radioactive \[^32^P\]dATP. For competition, nonlabeled cold DNA (2--10×), for supershift experiments PPARγ (Perseus) or RXR (Perseus) antibodies were used. Pull-Down Assays {#sec4.7} ---------------- Human PPARγ1 was tagged with streptavidin-binding protein and expressed in Rosetta BL21 (Novagen). After induction with 40 μM isopropyl-D-thiogalactopyranoside, PPARγ1 was purified with streptavidin-resin. Whole cell lysates of STAT6 or mock-transfected HEK293T cells were added to the resin and after washing analyzed by immunoblotting. Coimmunoprecipitation {#sec4.8} --------------------- HEK293T cells were transfected with *STAT6-V5* and *Flag-PPARg* expression vectors. V5 (AB Serotech) or Flag M2 (Sigma) antibodies were used for immunoprecipitation and subsequent immunoblotting. Chromatin Immunoprecipitation {#sec4.9} ----------------------------- ChIP was performed as described earlier ([@bib1]) with anti-PPARγ (CS-133-100, Diagenode and preimmune serum), anti-STAT6 (M20-Santa Cruz), or control immunoglobulin. Enrichment of genomic loci was quantitated with real-time PCR. Mice {#sec4.10} ---- Mice carrying null or floxed alleles of *Pparg* were described previously ([@bib2]). These mice were bred with *LysCre* transgene animals obtained from I. Förster (Univ. of Munich) ([@bib5]). *Stat6*^−/−^ mice were purchased from The Jackson Laboratory. Animals were housed under minimal disease conditions and the experiments were carried out under institutional ethical guidelines and licenses. Microarray Analysis {#sec4.11} ------------------- Microarray analysis was performed with Affymetrix microarrays (Human Genome U133 Plus 2.0 or Mouse Genome 430 2.0) and standard protocols. Microarray hybridizations were carried out at the Debrecen Clinical Genomics Center Microarray Facility. Analysis was performed with GC-RMA on the cel files in GeneSpring 7.3 (Agilent). For each condition, three biological replicates were analyzed. Signal intensities were normalized to the 50^th^ percentile (per chip), then to the median expression of the certain gene throughout the experiment (per gene), and finally each chip was normalized to its specific vehicle-treated control. Changing genes were called based on a t test (parametric, variances assumed to be equal, with Benjamini and Hochberg false discovery rate), p \< 0.05 and at least 1.5-fold changes. For annotation we used The Functional Annotation Tool at DAVID Bioinfomatics Resources 6.7 (National Institute of Allergy and Infectious Diseases \[NIAID\]). Immunohistochemistry {#sec4.12} -------------------- Immunostaining for PPARγ was carried out on paraffin-embedded cellular blocks with biotin-free Catalyzed Signal Amplification IHC detection kit (CSAII, Dako) and VIP substrate (Vector Labs). Sections were counterstained with methyl-green. Double immunofluorescence (IF) stainings were carried out on human tissues obtained from the archives of surgical tissue specimens of the Department of Pathology, University of Debrecen. For PPARγ staining, we used the red fluorescent tetramethyl-rhodamine (TMR)-tagged tyramide (Perkin-Elmer). All other IF for double stainings (CD68, DC-SIGN) were made sequentially via biotinylated secondary antibodies followed by a streptavidin-FITC development for green fluorescence (Dako). Sections were counterstained with 4\',6-diamidino-2-phenylindole (DAPI, Vector Laboratories) (blue nuclear fluorescence). Bioinformatic Analysis {#sec4.13} ---------------------- The PhastCons conservation scores for placental mammalian species were obtained from the UCSC site calculated from the MULTIZ (UCSC/Penn. State Bioinformatics) 44 vertebrate species whole-genome alignment. Oligoprecipitation Assays {#sec4.14} ------------------------- Nuclear extracts from THP1 cells treated with vehicle or RSG+IL-4 were prepared as described earlier ([@bib25]). Precleared extracts were incubated with annealed biotin-labeled oligonucleotides representing MacPPRE and streptavidin-agarose. Captured protein was analyzed by immunoblotting with PPARγ (E-8) and STAT6 (M-20) antibodies (Santa Cruz). Statistical Tests {#sec4.15} ----------------- All data are presented as means ± SD and based on experiments performed at least in triplicate. Statistical tests were performed on the fold changes via unpaired (two tail) t test, p \< 0.01. Accession Numbers {#app1} ================= The microarray data are available in the Gene Expression Omnibus (GEO) database (<http://www.ncbi.nlm.nih.gov/gds>) under the accession number GSE16387. Supplemental Information {#app3} ======================== Document S1. Supplemental Experimental Procedures and Three FiguresTable S1. Genes Regulated in Human Macrophages by PPARγMicroarray experiment was performed on human nonactivated (C) or alternatively activated (IL-4 100 ng/ml) macrophages ± 1 μM RSG. Genes regulated after 12 hr at least 1.5-fold significantly (p \< 0.05 with Benjamini-Hochberg multiple testing correction) are listed from Figures 1E and 1F. Subsets of up- and downregulated genes are listed in separate sheets. Functional annotation of RSG-regulated genes is also shown.Table S2. Genes Regulated by IL-4 in Ppar*γ*^+/−^*LysCre* and *Pparg*^fl/−^*LysCre* Mouse MacrophagesMicroarray experiment was performed on mouse bone marrow-derived nonactivated (C) or alternatively activated (IL-4 20 ng/ml) macrophages ± 1 μM RSG from wild-type and *Pparg*^+/−^*LysCre* and *Pparg*^fl/−^*LysCre*. IL-4-regulated genes from Figures 2E and 2F (\>2-fold, p \< 0.05 with Benjamini-Hochberg multiple testing correction) are listed in separate sheets. Functional annotation of IL-4-regulated genes is also shown.Table S3. Genes Regulated by PPARγ in Wild-Type and *Stat6*^−/−^ Mouse MacrophagesMicroarray experiment was performed on mouse bone marrow-derived noon-activated (C) or alternatively activated (IL-4 20 ng/ml) macrophages ± 1 μM RSG from wild-type and *Stat6*^−/−^ macrophages. RSG-regulated genes from Figures 3C and 3D (\>2-fold, p \< 0.05 with Benjamini-Hochberg multiple testing correction) are listed in separate sheets. Functional annotation of RSG-regulated genes is also shown.Table S4. Real-Time PCR AssaysSequence of in-house designed Taqman primers and probes and catalog numbers of Taqman assays ordered from Applied Biosystems are listed. We are grateful to L. Fariall (Univ. of Leicester, UK) for her help in EMSA experiments; M. Demeny for help with immunoprecipitation; and members of the L.N. laboratory for comments on the manuscript. We thank I. Furtos, M. Balogh, B. Menyhart, and A. Farkas for technical assistance. L.N. is an International Scholar of HHMI and holds a Wellcome Trust Senior Research Fellowship in Biomedical Sciences. This work was supported by grants from the Hungarian Science Research Fund OTKA NK72730 to L.N., OTKA/61814 to A.S., and OTKA F-68254 to B.L.B.; Hungarian Academy of Sciences Bolyai Scholarships to A.S., E.B., and B.L.B.; from the University of Debrecen (Mecenatura) to A.S.; TAMOP-4.2.2/08/1 to L.N.; and NKTH-Baross EA KFI EPIGEN08 to L.N. and B.L.B. Supplemental Information includes Supplemental Experimental Procedures, three figures, and four tables and can be found with this article online at [doi:10.1016/j.immuni.2010.11.009](10.1016/j.immuni.2010.11.009). ::: {#fig1 .fig} Figure 1 ::: {.caption} ###### Expression and Activity of PPARγ Is Dictated by Cytokines, the Role of IL-4 (A--C) Expression of *PPARg* (A) and *FABP4* (C) was determined by real-time PCR. Human monocytes (Mo) were differentiated to macrophages for 24 hr by attachment or to DCs for 5 days with GM-CSF and IL-4. Macrophages were treated with vehicle (C), IL-4, IFN-γ, TNF, IFN-γ+TNF (IT), LPS, or GM-CSF (GM) for 24 hr. LPS was added on day 5 for 24 hr to induce DC maturation. In (C) cells were also treated with RSG (R) or vehicle (DMSO:ethanol). Means normalized to cyclophilin A ± SD. n = 3, p \< 0.01 are shown. \(B) PPARγ protein was determined by immunostaining with PPARγ antibody. The positive nuclear staining is indicated by purple color (arrows). Methyl-green was used as nuclear counterstain. \(D) FABP4 protein was detected by immunoblot in human macrophages treated as indicated for 24 hr. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is the loading control. (E and F) Human macrophages were cultured for 12 hr in the absence or presence of IL-4 and treated with vehicle (C) or RSG (R), gene expression was analyzed by Affymetrix Human Genome U133 2.0 Plus microarrays (n = 3). Signal intensities were normalized to the 50^th^ percentile and to the median expression of genes and to the vehicle-treated control. \(E) Venn diagrams of RSG-regulated genes (\>1.5-fold change, p \< 0.05 parametric t test, Benjamini-Hochberg false discovery rate correction). \(F) Heatmaps of 730 probe sets from (E) on the left, and the efficacy (RSG-induced fold changes in IL-4-treated versus control macrophages) is shown on the right. :::  ::: ::: {#fig2 .fig} Figure 2 ::: {.caption} ###### PPARγ Is Dispensable for IL-4 Signaling \(A) Expression of *Pparg* (left) and *Fabp4* (right) in mouse bone marrow-derived monocytes activated as indicated was determined with real-time PCR. \(B) Mouse bone marrow-derived cells (n = 3) were differentiated to macrophages for 10 days, activated as indicated, and expression of *Angptl4* (left) and *Fabp4* (right) was determined. (C and D) Expression *Fabp4* (C), *Arg1*, and *Ym1* (D) in mouse bone marrow-derived macrophages differentiated from *Pparg*^+/+^, *Pparg*^+/−^, and *Pparg*^fl/−^*LysCre* mice are shown. (E and F) Bone marrow-derived macrophages from *Pparg*^+/−^ and *Pparg*^fl/−^*LysCre* mice (n = 3) were differentiated with M-CSF or MCSF+IL-4 in the absence (C) or presence (R) of RSG for 10 days. Gene expression was analyzed by Affymetrix Mouse Genome 430 2.0 microarrays. Signal intensities were normalized to the 50^th^ percentile and to the median expression of genes and to vehicle-treated control. \(E) The heatmap of the expression of IL-4-regulated genes (\>2-fold, significant differences are plotted; n = 3, parametric t test, Benjamini-Hochberg false discovery rate correction, p \< 0.05). \(F) Venn diagrams of the IL-4-regulated genes from macrophages of *Pparg*^+/−^ and *Pparg*^fl/−^*LysCre* mice. \(G) Expression of alternative activation markers, *Ym1*, *Arg1*, *Fizz1*, and that of *Pparg* from mouse bone marrow-derived macrophages. Means normalized to cyclophilin A ± SD. n = 3, p \< 0.01 are shown. :::  ::: ::: {#fig3 .fig} Figure 3 ::: {.caption} ###### IL-4 Acts through STAT6 to Induce Augmented PPARγ Response \(A) Human nonactivated or alternatively activated (IL-4) macrophages were exposed to vehicle (C), proteasome inhibitor MG132, JAK3 inhibitor WHI-P131, JAK2 inhibitor TyrPhostin AGN490 (TyrPh.), or PI-3K inhibitor wortmannin for 6 hr. Simultaneously, cells were treated with vehicle (C) or RSG. Expression of *AMAC1*, *PPARg*, and *FABP4* are shown. \(B) Expression of *Ym1*, *Pparg*, *Fabp4*, and *Angptl4* was analyzed in bone marrow-derived macrophages isolated from wild-type C56Bl/6 and *Stat6*^−/−^ mice and treated for 10 days as indicated. (C and D) Mouse bone marrow-derived macrophages from wild-type C57Bl/6 and *Stat6*^−/−^ mice were cultured in the presence of vehicle (C), RSG (R), IL-4, or IL-4+RSG. Gene expression was analyzed by Affymetrix Mouse Genome 430 2.0 microarrays. Signal intensities were normalized to the 50^th^ percentile and to the median expression of the gene throughout the experiment and finally to its specific vehicle-treated control. \(C) The heatmap of RSG-induced and repressed genes in the control or IL-4-treated wild-type macrophages. \(D) Venn diagrams of RSG-regulated genes. Means normalized to cyclophilin A ± SD. n = 3, p \< 0.01 are shown. :::  ::: ::: {#fig4 .fig} Figure 4 ::: {.caption} ###### STAT6 Acts on PPARγ Response at the Transcriptional Level \(A) Human macrophages were treated with vehicle or CXM, activated with vehicle (C) or IL-4 for 6 hr, and simultaneously vehicle (C) or RSG (R) was added. Expression of *AMAC1*, *PPARg*, and *FABP4* were measured with real-time PCR. \(B) 293T cells were transfected with mock, *PPARg*, *STAT6*, or *PPARg*+*STAT6* expression vectors and treated as indicated for 24 hr and expression of *FABP4* was analyzed by real-time PCR. \(C) Human macrophages were treated as indicated. Expression of *FABP4* and 15-lipoxygenase were measured by real-time PCR. \(D) Human macrophages were treated with vehicle (C) or IL-4 and simultaneously with vehicle (C), RSG (R) or LG1208 RXR antagonist (ant.) for 24 hr. Expression of *FABP4* was analyzed by real-time PCR. \(E) Human macrophages were treated with vehicle (C) or 100 nM TSA, activated with vehicle (C) or IL-4 for 12 hr, and simultaneously vehicle (C) or RSG (R) was also added. Expression of *FABP4* (left) and *PPARg* (right) were measured by real-time PCR. Means normalized to 36B4 (A, B, C) cyclophilin A (D, E) ± SD. n = 3, p \< 0.01 are shown. :::  ::: ::: {#fig5 .fig} Figure 5 ::: {.caption} ###### Identification of a Composite PPARγ-STAT6 Response Element in the *FABP4* Promoter \(A) A reporter construct containing 5 kb fragment of human *FABP4* promoter was cotransfected into RAW264.7 and 293T cells with the indicated expression vectors, and normalized luciferase activity was determined 24 hr after cytokine or ligand treatment (control \[C\] or RSG \[R\]). \(B) Schematic structure of the mouse and human *FABP4* with the localization of Adipo and MacPPREs. \(C) Deletion mutants of the 5 kb promoter of the human *FABP4* were cotransfected into 293T cells with mock or PPARγ expression vectors. Normalized luciferase activity was determined 24 hr after ligand treatment (vehicle \[C\] or RSG \[R\]). \(D) Sequences of human and mouse MacPPREs of the *Fabp4* gene. \(E) The human MacPPRE was mutated at the 1^st^ or 2^nd^ half site and tested in transfection assays. Reporter constructs were cotransfected into 293T cells with the indicated receptor, *VP16-PPARg* (VP-Pγ), *VP16-RXR* (VP-RXR) expression vectors. Normalized luciferase activity was determined 24 hr after ligand treatment (vehicle \[C\], WY14643 \[WY\], RSG \[R\], GW501516 \[GW\], or LG268 \[LG\]). \(F) DNA binding of PPARα (Pα), γ (Pγ), δ (Pδ), and RXR (X) to the human MacPPRE was analyzed by EMSA. Cold competitors of consensus DR1 (DR1), wild-type MacPPRE, or its mutants (1^st^ half site-M1, 2^nd^ half site-M2, downstream sequence outside the DR1-M3) were used. \(G) Conservation of human *FABP4* gene was analyzed by PhastCons conservation scores for placental mammals. The plotted region corresponds to the hg18 chromosome 8 genomic position from 82555216 to 82568028 on the negative strand. The y axis shows the PhastCons conservation scores (in the range 0--1) for each position. (H and I) Reporter constructs with three copies of STAT6 response elements of human *CCL11* (H) or *FABP4* (I) gene were transfected into 293T cells with *STAT6* and normalized luciferase activity was determined 24 hr after IL-4 (100 ng/ml) exposure. (J--M) Composite response elements of *FABP4* gene containing both PPRE and STAT6 binding sites were transfected into 293T cells along with *PPARg* and *STAT6* expression vectors. Human MacPPRE (J), STAT6 (K), or PPARγ (L) binding site mutant of MacPPRE and human AdipoPPRE (M) were tested. Normalized reporter activities as means ± SD, n = 3, p \< 0.01 are shown. :::  ::: ::: {#fig6 .fig} Figure 6 ::: {.caption} ###### STAT6 Colocalizes and Interacts with PPARγ In Vivo and In Vitro \(A) 293T cells were transfected with *PPARg* and *STAT6* expression vectors and ChIP was performed with anti-STAT6 and anti-PPARγ. IL-4-induced enrichment of *FABP4* MacPPRE, AdipoPPRE, *LXRa*-PPRE, and *CD36*-PPRE elements were analyzed by real-time PCR. *GAPDH* promoter was used as negative control. Enrichment over IgG and over control samples is presented. \(B) PPARγ and STAT6 ChIP was performed on bone marrow-derived macrophages from wild-type and *Stat6*^−/−^ mice. Enrichment of *Fabp4* PPRE over the IgG and over the wild-type samples is presented. \(C) PPARγ ChIP was performed on bone marrow-derived macrophages from wild-type and *Stat6*^−/−^ mice. Enrichment of PPREs (*Acsl1*, *Angptl4*, *Cd36*, osteopontin, pyruvate dehydrogenase kinase-4 \[*Pdk4*\], stearoyl-Coenzyme A desaturase-1 \[*Scd1*\], *Fabp4* Adipo-, and MacPPRE), STAT6 response elements of *Fizz1*, *Tarc1*, and *Ym1*, and the negative control *Hoxa1* over the IgG and over the wild-type samples is presented. \(D) hMacPPRE coprecipitates endogenous STAT6 and PPARγ from THP-1 cells. Biotin-labeled oligonucleotides corresponding to wild-type (WT), PPRE (M-PPRE), STAT6 site (M-S6), or double mutant (M-PPRE-S6) MacPPRE were incubated with equal amounts of nuclear extracts harvested from vehicle (−) or IL-4+RSG (IL-4+R)-treated (+) cells. Pulled down proteins were immunoblotted with the indicated antibodies. Input represents 1.67% of oligoprecipitated material. (E and F) 293T cells were transfected with mock, *V5-STAT6* (V5-S6), or *V5-STAT6*+*Flag-PPARg* (V5-S6+Flag-Pγ) as indicated. Coimmunoprecipitation from whole cell lysates was performed with V5 or Flag antibodies and presence of V5-STAT6 (E) and Flag-PPARγ (F) were analyzed by immunoblotting. \(G) PPARγ was expressed in bacteria and purified with streptavidine-resin, and STAT6 was pulled down from whole cell lysates of mock or V5-STAT6-transfected 293T cells. Proteins were analyzed by immunoblotting with V5 antibody. Means normalized to input ± SD. n = 3, p \< 0.05 are shown. :::  ::: [^1]: 7 Present address: Department of Molecular Biology, Massachusetts General Hospital, Department of Genetics, Harvard Medical School, Howard Hughes Medical Institute, Richard Simches Research Building, 185 Cambridge St., Boston, MA 02114, USA

PubMed Central

2024-06-05T04:04:19.333877

2010-11-24

PMC3052438

Published online: January 20, 2011 Results and Discussion {#sec1} ====================== KT-MT interaction develops in a step-wise manner \[[@bib1]\]. The KT initially interacts with the MT lateral surface (lateral attachment) and slides along the MT towards a spindle pole ([Figure 1](#fig1){ref-type="fig"}A, i, ii). Then, the KT is tethered at the end of the MT (end-on attachment) and transported further as the MT shrinks (end-on pulling) ([Figure 1](#fig1){ref-type="fig"}A, iii). Subsequently both sister KTs interact with MTs, and aberrant KT-MT interactions are removed by error correction ([Figure 1](#fig1){ref-type="fig"}A, iv, v) until sister KT biorientation (i.e., sister KTs attaching to MTs from opposite spindle poles) is established ([Figure 1](#fig1){ref-type="fig"}A, vi). The Ndc80 complex is an outer (i.e., closer to the MT) KT component, composed of four proteins ([Figure 1](#fig1){ref-type="fig"}B), and has a central role in comprising the KT-MT interface \[[@bib1 bib5]\]. The Ndc80 complex binds directly to the MT lateral surface in vitro, at the calponin-homology (CH) domain within Ndc80 protein (also called Hec1) \[[@bib8 bib9 bib10]\], and the complex is indeed required for the lateral KT-MT attachment in vivo \[[@bib11]\]. Presumably the Ndc80 complex is also involved in the end-on KT-MT attachment. Consistent with this, an injection of an antibody against the Ndc80 CH domain changed the dynamics of KT-MT interactions in metaphase \[[@bib12]\]. Moreover, the Ndc80 complex can couple a microsphere at the end of a dynamic MT in an in vitro reconstituted system \[[@bib13]\]. Thus, it is likely that the Ndc80 complex is directly involved in both the lateral and end-on KT-MT attachment. Given this, the Ndc80 complex may play a key role in the conversion from the lateral to end-on attachment. Mutations within the Ndc80 Loop Region Lead to Cell Lethality or Temperature-Sensitive Cell Growth {#sec1.1} -------------------------------------------------------------------------------------------------- The Ndc80 complex forms a long rod-shape structure with two globular domains at each end \[[@bib8 bib9 bib10]\] ([Figure 1](#fig1){ref-type="fig"}B). While one globular domain interacts with a MT, the other binds the Mtw1 complex (Mis12 complex in metazoa), a relatively inner KT component, i.e., closer to the centromere. These two globular domains are connected by long coiled-coil motifs. Peculiarly, this coiled-coil shaft is interrupted in the middle of Ndc80 protein \[[@bib6]\] by a region of 50--60 amino acid residues that does not conform to the coiled-coil structure \[[@bib10]\] ([Figure 1](#fig1){ref-type="fig"}C), thus presumably looping out from the coiled-coil shaft and hence called the loop region ([Figure 1](#fig1){ref-type="fig"}B). Indeed, electron microscopy revealed that the coiled-coil shaft of the Ndc80 complex showed a kink or flexible bend at the position of the loop region \[[@bib14]\]. Intriguingly, the loop region contains several evolutionarily conserved amino acid residues ([Figure 1](#fig1){ref-type="fig"}D) and probably forms a β-sheet structure that may be involved in protein-protein interaction \[[@bib14]\]. To address the role of the Ndc80 loop region, we constructed yeast strains whose only *ndc80* harbors a deletion of 20--40 amino acid residues within the loop region, i.e., *ndc80*Δ*480-520*, *ndc80*Δ*480-510*, *ndc80*Δ*490-520*, and *ndc80*Δ*490-510* ([Figure 1](#fig1){ref-type="fig"}C). Deletions *ndc80*Δ*480-520*, *ndc80*Δ*480-510*, and *ndc80*Δ*490-520* could not support cell viability at any temperature tried (data not shown), whereas *ndc80*Δ*490-510* cells showed growth at 25°C but not at 35°C ([Figure 1](#fig1){ref-type="fig"}E). We also constructed strains whose only *ndc80* had substitution of alanines for seven conserved amino acid residues within the loop region (and thus called *ndc80-7A*; [Figure 1](#fig1){ref-type="fig"}D). The *ndc80-7A* mutant cells showed growth at 25°C but not at 35°C ([Figure 1](#fig1){ref-type="fig"}E). Such temperature-sensitive growth of *ndc80*Δ*490-510* and *ndc80-7A* cells was not due to reduced expression of mutant Ndc80 proteins or a defect in interaction with Nuf2, another component of the Ndc80 complex ([Figure 1](#fig1){ref-type="fig"}F). Mutations in the Ndc80 Loop Region Support Initial KT-MT Interaction Normally but Sister KT Biorientation Is Defective {#sec1.2} ---------------------------------------------------------------------------------------------------------------------- To address possible roles of the Ndc80 loop region in KT-MT interactions, we visualized MTs and a selected centromere (*CEN5*) by live-cell imaging and compared their behavior in wild-type, *ndc80*Δ*490-510*, and *ndc80-7A* mutant cells at 35°C. *spc24-1* mutants are defective in KT-MT attachment \[[@bib11 bib15]\] and were used as a control. In wild-type cells, *CEN5* detached from MTs (upon KT disassembly resulting from centromere DNA replication \[[@bib16]\]) and moved away from a spindle pole. Within 2--3 min, *CEN5* interacted again with MTs when the KT was reassembled on *CEN5* \[[@bib16]\] ([Figure 2](#fig2){ref-type="fig"}A, i). In *ndc80*Δ*490-510* and *ndc80-7A* mutant cells, *CEN5* detached from MTs and subsequently reattached to MTs with similar timing to wild-type cells ([Figure 2](#fig2){ref-type="fig"}A, ii). The duration for *CEN5* dissociation from MTs was also similar between wild-type and the loop-region mutants ([Figure 2](#fig2){ref-type="fig"}A, iii). On the other hand, *spc24-1* mutants showed earlier and longer *CEN5* dissociation from MTs, compared with wild-type ([Figure S2](#app2){ref-type="sec"}A available online). In conclusion, mutations at the Ndc80 loop region had no effect on the efficiency of the initial interaction of KTs with MTs. However, a subsequent step was inefficient in the Ndc80 loop-region mutants. Wild-type, *ndc80* loop-region, and *spc24-1* mutants established a bipolar spindle at the end of S phase (data not shown), and wild-type cells showed separation of sister *CEN5*s immediately afterwards ([Figure 2](#fig2){ref-type="fig"}B, i, ii), indicative of sister *CEN5* biorientation on the spindle \[[@bib17 bib18]\]. In *ndc80-7A* and *ndc80*Δ*490-510* mutant cells, sister *CEN5*s were on the spindle but their separation was delayed ([Figure 2](#fig2){ref-type="fig"}B, i, ii). In most *spc24-1* cells, sister *CEN5*s remained unseparated and did not localize on the spindle ([Figure S2](#app2){ref-type="sec"}B). Thus, in mutants of the Ndc80 loop region, the establishment of sister KT biorientation is defective although KTs are caught on the spindle. Meanwhile, the *ndc80-7A* and *ndc80*Δ*490-510* mutants also showed failure to satisfy the spindle-assembly checkpoint \[[@bib19]\] ([Figure S2](#app2){ref-type="sec"}C). We also compared the nature of the biorientation defect in these mutants with that found in *ipl1* and *mps1* mutants \[[@bib1]\] ([Figure S2](#app2){ref-type="sec"}D). The Ndc80 Loop Region Is Required for the Efficient Conversion from Lateral to End-on KT-MT Attachment {#sec1.3} ------------------------------------------------------------------------------------------------------ To analyze the KT-MT attachment of Ndc80 loop mutants in more detail, we next used an engineered assay system, in which the assembly of the KT was delayed on a particular centromere (*CEN3*) by the activation of transcription from an adjacently inserted promoter ([Figure 3](#fig3){ref-type="fig"}A) \[[@bib11]\]. This procedure prevented *CEN3* from localizing on the mitotic spindle. While cells were arrested in metaphase, we reactivated *CEN3*, which led to KT reassembly and interaction with MTs extending from a spindle pole (spindle-pole MTs). This assay allowed observation of the individual KT-MT interaction with high spatial resolution because *CEN3* moved away from the spindle prior to its reactivation \[[@bib11]\]. In agreement with the results in [Figure 2](#fig2){ref-type="fig"}, in *ndc80*Δ*490-510* and *ndc80-7A* mutant cells, *CEN3* was captured by the lateral surface of a spindle-pole MT at 35°C with similar kinetics as wild-type cells; by contrast, subsequent sister *CEN3* separation on the spindle proceeded more slowly compared with wild-type cells, indicative of a delay in sister *CEN3* biorientation ([Figure 3](#fig3){ref-type="fig"}B). On the other hand, in *spc24-1* cells, the initial *CEN3* capture by MTs was defective \[[@bib11]\]. By using live-cell imaging, we investigated *CEN3*-MT interaction in further detail. In wild-type cells, after the initial *CEN3*-MT interaction, *CEN3* moved by sliding along a MT lateral surface toward a spindle pole \[[@bib11]\]. While *CEN3* was on the MT lateral surface, this MT often underwent depolymerization at its plus end and shrank until its plus end caught up with *CEN3* ([Figure 3](#fig3){ref-type="fig"}C, i). When this happened, either of the following two events occurred in wild-type cells \[[@bib3]\]: (1) *CEN3* was tethered at the MT end (end-on attachment) and pulled toward a spindle pole as the MT shrank further (end-on pulling) (40% of cases) or (2) the MT showed regrowth (MT rescue at *CEN3*) (60% of cases) ([Figure 3](#fig3){ref-type="fig"}C, i). In *ndc80*Δ*490-510* and *ndc80-7A* mutant cells, *CEN3* sliding occurred almost normally, except for a small number (\<5%) of *ndc80*Δ*490-510* cells showing *CEN3* pausing on a MT during an extended period (data not shown). Remarkably, in both *ndc80*Δ*490-510* and *ndc80-7A* mutants, the end-on attachment was rarely established at 35°C ([Figure 3](#fig3){ref-type="fig"}C, i), thus making subsequent end-on pulling infrequent ([Figure 3](#fig3){ref-type="fig"}C, ii), compared with wild-type cells. Thus, Ndc80 loop region is required for the efficient conversion from lateral to end-on KT-MT attachment. Notably, defects in end-on attachment correlate well with defects in sister KT biorientation. For example, the *ndc80-7A* mutant showed milder defects in both end-on attachment and biorientation, compared with *ndc80*Δ*490-510* (see Figures [2](#fig2){ref-type="fig"}B, [3](#fig3){ref-type="fig"}B, and 3C). We speculate that end-on attachment might be a prerequisite for biorientation. Consistent with this, it is suggested that end-on attachment is necessary to sustain KT-MT attachment when sister KT biorientation is established and tension is applied on the KT-MT interaction \[[@bib2 bib4]\]. Indeed, the end-on attachment seems to be more robust than the lateral attachment \[[@bib1 bib3]\]. The Ndc80 Loop Region Is Required for Ndc80-Dam1 Interaction and for Dam1 Loading on the KT {#sec1.4} ------------------------------------------------------------------------------------------- The Dam1 complex (also called DASH complex), composed of 10 proteins including Dam1 protein, also has an important role in end-on KT-MT attachment \[[@bib1 bib20]\]. In contrast to the Ndc80 complex, the Dam1 complex is not a part of the KT during the lateral KT-MT attachment and is loaded on the KT only upon end-on attachment \[[@bib3]\]. The Dam1 complex has the ability to track the plus end of a shrinking MT \[[@bib3 bib7]\] and, once loaded on the KT, it mediates the end-on pulling of the KT by a shrinking MT \[[@bib3]\]. During this process, the Dam1 complexes form oligomers and/or a ring structure encircling a MT \[[@bib21]\]. Thus, the Ndc80 loop and the Dam1 complex may work together to support end-on KT-MT attachment. In this regard, it is intriguing that the Ndc80 and Dam1 complexes showed a physical interaction \[[@bib22 bib23 bib24 bib25]\]. It was difficult to detect this interaction conclusively via coimmunoprecipitation or a protein pull-down (data not shown), but it could be detected with a yeast two-hybrid assay \[[@bib23]\]. We therefore addressed whether the interaction between Ndc80 and Dam1 was dependent on the Ndc80 loop region by using a yeast two-hybrid assay. We first confirmed that all the wild-type Ndc80 and its mutants Ndc80Δ490-510 and Ndc80-7A showed interaction with Nuf2 ([Figure 4](#fig4){ref-type="fig"}A, right), consistent with the result in [Figure 1](#fig1){ref-type="fig"}F. We also found that wild-type Ndc80 showed a positive interaction with Dam1, as reported previously \[[@bib23]\]. However, Ndc80Δ490-510 and Ndc80-7A mutants showed very little interaction with Dam1 ([Figure 4](#fig4){ref-type="fig"}A, left). Thus the loop region indeed facilitates interaction between Ndc80 and Dam1. What is the functional consequence of the Ndc80-Dam1 interaction? The Ndc80 complex is required for loading of the Dam1 complex on the KT \[[@bib22 bib26]\] and an Ndc80-Dam1 interaction may facilitate this process. If so, the Ndc80 loop region might be required for Dam1 complex loading on the KT. We tested this possibility by using chromatin immunoprecipitation. In wild-type cells, centromere DNA (*CEN3*) was clearly precipitated with the Dam1 protein and also with the Nuf2 protein ([Figure 4](#fig4){ref-type="fig"}B, i, ii). Remarkably, in *ndc80*Δ*490-510* and *ndc80-7A* mutants, *CEN3* precipitation with Dam1 was considerably reduced ([Figure 4](#fig4){ref-type="fig"}B, i, ii), although *CEN3* precipitation with Nuf2 was similar between the mutants and wild-type. This result suggests a defect in Dam1 loading on KTs in these mutants. We also compared the localization pattern of Dam1 and Mtw1 in metaphase ([Figure 4](#fig4){ref-type="fig"}C; [Figure S4](#app2){ref-type="sec"}A). Mtw1 is a component of the KT \[[@bib20]\] and should represent the position of KTs. Dam1 and Mtw1 showed almost perfect colocalization in wild-type cells. In *ndc80*Δ*490-510* cells, the total amount of Mtw1 and Dam1 on the spindle was not altered ([Figure S4](#app2){ref-type="sec"}A), but Dam1 signals were often present between two globular Mtw1 signals ([Figure 4](#fig4){ref-type="fig"}C; [Figure S4](#app2){ref-type="sec"}A). Results in [Figures 4](#fig4){ref-type="fig"}B and 4C suggest requirement of the Ndc80 loop region for Dam1 loading on the KT. The Ndc80 Loop Region Facilitates Interaction with the Dam1 Complex to Anchor the KT at the Dynamic MT Plus End {#sec1.5} --------------------------------------------------------------------------------------------------------------- Our study has revealed that the Ndc80 loop region mediates the interaction with the Dam1 complex to ensure proper KT-MT attachment ([Figure 4](#fig4){ref-type="fig"}D). With Ndc80 loop-region mutants, the lateral KT-MT attachment is still largely normal; consistently, this process does not require the Dam1 complex \[[@bib11]\]. On the other hand, the Dam1 complex has an important role in the end-on KT-MT attachment and subsequent end-on pulling of the KT by a MT \[[@bib3 bib7]\]. With Ndc80 loop region mutants, the Ndc80 and Dam1 complexes cannot interact properly, leading to the failure in the end-on attachment. It was recently demonstrated that the Dam1 complex is able to enhance MT binding of the Ndc80 complex (e.g., its cosedimentation with MTs) in vitro \[[@bib24 bib25]\]. Given this, by using a condition reported in \[[@bib24]\], we evaluated MT cosedimentation of the purified Ndc80 complex with loop mutants; its enhancement by the Dam1 complex was similar to that of the wild-type Ndc80 complex ([Figure S4](#app2){ref-type="sec"}B). We reason that the loop-dependent Ndc80-Dam1 interaction in vivo was not recapitulated in this particular condition in vitro. Alternatively, an additional factor, which is missing in the in vitro reaction, may be necessary for the interaction between Dam1 and the Ndc80 loop region. In this regard it is intriguing that, in fission yeast, Dis1 (an ortholog of Stu2 in budding yeast and XMAP215/chTOG in vertebrates) showed interaction with the Ndc80 loop region \[[@bib27]\]. However, in contrast to fission yeast, Ndc80 and Stu2 showed no interaction in budding yeast ([Figure S4](#app2){ref-type="sec"}C) and Ndc80 loop mutants did not alter Stu2 localization at KTs ([Figure S4](#app2){ref-type="sec"}D). Nonetheless, Stu2 shows interaction with Dam1-complex components in a two-hybrid assay (\[[@bib23]\]; data not shown). Thus we cannot exclude the possibility that Stu2 (possibly at the end of a shrinking MT) is involved in the Ndc80-Dam1 interaction. Our finding that the Ndc80 loop region mediates the interaction with the Dam1 complex is consistent with nanometer-scale mapping of KT components in metaphase \[[@bib28]\]. The Ndc80 complex bridges between the inner KT and a MT, and its Ndc80/Nuf2 globular head locates further outside (away from the inner KT) of the Dam1 complexes (see [Figure 4](#fig4){ref-type="fig"}D). In this configuration, the location of the Ndc80 loop region along the KT-MT axis approximately corresponds to that of the Dam1 complex \[[@bib28]\]. The lateral KT-MT attachment has advantages for the initial KT-MT interaction because the MT lateral surface provides a large contact surface, whereas the end-on attachment ensures more robust KT-MT interaction \[[@bib2 bib3 bib4]\], which is presumably required for sister KT biorientation. Thus the conversion from lateral to end-on attachment is an inevitable vital step in developing a proper KT-MT interaction. The Ndc80 and Dam1 complexes play central roles in comprising the KT-MT interface. Our study has identified the Ndc80 loop as an important mediator of the Ndc80-Dam1 interaction, whose role is to facilitate the crucial maturation step of the KT-MT interaction. Supplemental Information {#app2} ======================== Document S1. Supplemental Experimental Procedures and Two Figures We thank all T.U.T. lab members for helpful discussions; Adel Ibrahim for construction of *ndc80* mutants; T. Toda for personal communication; C. Allan, N. Kobayashi, E. Kitamura, and S. Swift for technical help for microscopy and computing; and K. Nasmyth, E. Schiebel, S. Harrison, K. Bloom, J.E. Haber, Y. Watanabe, S. Biggins, R. Tsien, K. Sawin, J.V. Kilmartin, and Yeast Resource Centre and EUROSCARF for reagents. This work was supported by Human Frontier Science Program, Cancer Research UK, Medical Research Council, Lister Research Institute Prize, and Association for International Cancer Research. S.K. was supported by a CRUK PhD studentship. T.U.T. is a Senior Research Fellow of Cancer Research UK. Supplemental Information includes Supplemental Experimental Procedures and two figures and can be found with this article online at [doi:10.1016/j.cub.2010.12.050](10.1016/j.cub.2010.12.050). ::: {#fig1 .fig} Figure 1 ::: {.caption} ###### Deletions and Mutations within the Ndc80 Loop Region Cause Cell Lethality or Temperature-Sensitive Cell Growth \(A) Step-wise development of kinetochore (KT)-microtubule (MT) interaction during prometaphase (i--v) and metaphase (vi). See more detail in \[[@bib1]\]. \(B) Structure of the Ndc80 complex, which consists of four proteins \[[@bib8 bib9 bib10]\]. The position of the loop region is indicated. \(C) The probability of forming coiled-coil motifs along amino acid residues of Ndc80 protein in *Saccharomyces cerevisiae*. Top: Full length of Ndc80. Bottom: Amino acid residues 420--550. Thick red lines indicate the positions of deletions in Ndc80 mutants, constructed in this study. ts, temperature-sensitive cell growth. \(D) Multiple sequence alignment of the Ndc80 loop region from different organisms. The regions with coiled-coil probability \<0.5 were selected for alignment. Some residues (numbers in parentheses) showed less conservation and were not shown here. Conserved residues are highlighted in colors: hydrophobic (light blue), acidic (purple), and basic (red) residues, asparagine (green), proline, and glycine (yellow). The positions of alanine substitution for the *ndc80-7A* mutant are shown in red rectangles. \(E) *ndc80*Δ*490-510* and *ndc80-7A* mutants show temperature-sensitive cell growth. 10-fold serial dilutions of wild-type (T6500), *ndc80*Δ*490-510* (T6566), and *ndc80-7A* (T7881) cells were spotted onto YPD plates and incubated at the 25°C (top) and 35°C (bottom) for 48 hr. \(F) Wild-type and *ndc80*Δ*490-510*, *ndc80-7A* cells show similar Ndc80 expression levels and similar Nuf2 association with Ndc80. *NUF2-myc* cells with *NDC80* wild-type (T7082), *ndc80*Δ*490-510* (T7085), or *ndc80-7A* (T8357), tagged with *HA*, were treated with α factor, released to fresh YPD medium at 35°C, and harvested after 70 min from the release (at which time the majority of cells were in metaphase). Wild-type cells without *HA* or *myc* tags (T7084, T6981) were also treated in the same way as controls. Total proteins (top) and the proteins immnunoprecipitated with a myc antibody (bottom) were detected by western blotting with an HA antibody. :::  ::: ::: {#fig2 .fig} Figure 2 ::: {.caption} ###### In Mutants of the Ndc80 Loop Region, the Initial KT-MT Interaction Occurs Normally, but Sister KT Biorientation Is Established Inefficiently \(A) The initial KT-MT interaction: wild-type (T7848), *ndc80*Δ*490-510* (T7862), and *ndc80-7A* (T8397) cells with *CEN5-tetOs TetR-3×CFP Venus-TUB1* were treated with α factor and released to fresh YPD medium at 35°C. CFP and Venus images were acquired every 10 s at 35°C. To avoid photo-bleaching of fluorescence signals during image acquisition, the field of observation was changed every 10 min. (i) Representative live-cell images, in which a wild-type cell showed *CEN5* detachment from, and subsequent reattachment to, MTs. The cell shape is outlined in white. *ndc80*Δ*490-510* and *ndc80-7A* mutants showed similar behavior of *CEN5* (data not shown). (ii) Timing of *CEN5* detachment from MTs, shown as the percentage of cells per field showing detachment during each 10 min time window. (iii) Duration of *CEN5* dissociation from MTs in individual cells (means and standard errors). n.s., difference is not significant. \(B) Establishment of sister KT biorientation. T7848, T7862, and T8397 cells (see A) were treated as in (A). CFP and Venus images were acquired every 1 min at 35°C. (i) Representative images of wild-type and *ndc80-7A* mutant cells, which showed separated and unseparated sister *CEN5*s, respectively. (ii) The percentage of cells showing separation of sister *CEN5*s on the bipolar spindle for at least two consecutive time points, until indicated time points (0 min: establishment of bipolar spindle). Sister *CEN5*s were scored as "separated" when two signals were discernible. See also [Figure S2](#app2){ref-type="sec"}. :::  ::: ::: {#fig3 .fig} Figure 3 ::: {.caption} ###### Mutants of the Ndc80 Loop Region Show Inefficient Conversion from Lateral to End-on KT-MT Attachment \(A) Experimental system to analyze KT interaction with individual MTs. See details in \[[@bib11]\]. \(B) Mutants of the Ndc80 loop region show normal initial KT interaction with MTs, but subsequent establishment of sister KT biorientation is inefficient. *NDC80* wild-type (T6803), *ndc80*Δ*490-510* (T6690), and *ndc80-7A* (T7955) cells with *P~GAL~-CEN3-tetOs TetR-GFP Venus-TUB1 P~MET3~-CDC20* were treated with α factor in methionine drop-out medium with 2% raffinose for 2.5 hr and released to YP medium containing 2% galactose, 2% raffinose, and 2 mM methionine at 25°C to inactivate *CEN3* and arrest cells in metaphase. After 3 hr, the culture temperature was changed to 35°C. After 15 min, cells were suspended in the same medium but containing 2% glucose instead of galactose/raffinose to reactivate *CEN3* (defined as 0 min). Cells were collected at indicated time points and fixed with paraformaldehyde. GFP and Venus images were acquired and *CEN3*-MT interaction was scored as indicated in the schematic drawing. In most of *spc24-1* cells analyzed in this assay, *CEN3* remained uncaptured by MTs for 60 min (Figure 2c in \[[@bib11]\]), in contrast to the *ndc80* loop-region mutants. \(C) Mutants of the Ndc80 loop region show inefficient conversion from the lateral to end-on KT-MT attachment. T6803, T6690, and T7955 cells (see B) were treated as in (B), except that cells were suspended in synthetic complete medium containing 2% glucose and 2 mM methionine to reactivate *CEN3*. Cells were immobilized and GFP and Venus images were acquired every 20 s for 30 min at 35°C. (i) When the plus end of a shrinking MT caught up with *CEN3*, the MT subsequently showed either regrowth (MT rescue) or tethering of *CEN3* to its plus end while shrinking further (end-on attachment and end-on pulling). Representative images of the events in wild-type cells and a graph showing frequency of the two events; these events happened in two mutants as in wild-type cells, albeit with very different frequencies. (ii) Frequency of each mode of *CEN3* transport by a MT toward a spindle pole. Modes were classified as indicated by the schematic drawing. Sliding and end-on pulling were scored only when *CEN3* moved for 1 μm or longer by each mode of the transport. The pink bars represent the cases where the end-on attachment was established before *CEN3* moved along the MT lateral side more than 1 μm. :::  ::: ::: {#fig4 .fig} Figure 4 ::: {.caption} ###### The Ndc80 Loop Region Is Required for Ndc80-Dam1 Interaction and for Dam1 Loading on the KT \(A) The Ndc80 loop region is required for Ndc80-Dam1 interaction in a two-hybrid assay. The same amount of cells expressing indicated proteins, fused with a DNA binding domain or an activation domain, were spotted on histidine drop-out plates and incubated at 35°C for 48 hr. Cell growth suggests interaction between the two relevant proteins. Ras and Raf were used as controls for the assay. \(B) The Ndc80 loop region is required for Dam1 loading on the KT. *DAM1-myc* cells with wild-type (T8761), *ndc80*Δ*490-510* (T8762), and *ndc80-7A* (T8763) were treated with α factor, released to fresh YPD medium at 35°C, harvested after 70 min from the release (at which time the majority of cells were in metaphase), and treated with formaldehyde to crosslink. *NUF2-myc* cells with wild-type (T8777), *ndc80*Δ*490-510* (T8778), and *ndc80-7A* (T8779) were treated in the same way. Wild-type cells without *myc* tags (T6500) were also treated in the same way, as a control. (i) Gel images of PCR products, amplified at *CEN3* region and at a noncentromere locus (*MPS1* locus, 45 kb from *CEN4*), with total DNA from whole cell extract (WCE) or immunoprecipitated DNA (IP) as a template. (ii) The percentage of recovered DNA was first quantified as a fraction of corresponding WCE in individual samples. Then, these percentage values were standardized, relative to that in *NDC80* wild-type cells (at *CEN3* region). Mean and standard errors were obtained from three independent experiments. \(C) The Ndc80 loop region is required for Dam1 colocalization with the KT. Wild-type (T7868) and *ndc80*Δ*490-510* (T7866) cells with *DAM1-3×GFP MTW1-3×CFP SPC42-RFP* were cultured and harvested as in (B). Representative images are shown here. Spindle pole bodies (SPBs) were visualized with Spc42-RFP. Other representative images and the quantification of total Dam1 and Mtw1 signals in individual cells are shown in [Figure S4](#app2){ref-type="sec"}A. \(D) Summary for the role of the Ndc80 loop region (shown in red) in the conversion of lateral to end-on KT-MT attachment. During lateral attachment, the Ndc80 complex (blue) binds a MT, presumably at its Ndc80/Nuf2 CH domains and the N-terminal region of Ndc80 \[[@bib8 bib9 bib10]\]. To convert lateral attachment to end-on attachment, it is crucial that the Ndc80 loop region mediates the interaction with the Dam1 complex (yellow), which localizes at the MT plus end and forms an oligomer and/or a ring encircling the MT \[[@bib3 bib7 bib21]\]. The Ndc80-Dam1 interaction could be direct or indirect, and more factors might be involved in this interaction. See also [Figure S4](#app2){ref-type="sec"}. :::  ::: [^1]: 3 These authors contributed equally to this work

PubMed Central

2024-06-05T04:04:19.337951

2011-2-08

PMC3052439

Setting the Stage ================= The Dietary Guidelines for Americans (DGA) were first formally introduced to the public in 1980, in an attempt to give consumers the science-based nutrition recommendations they need to build a healthy diet and prevent diet-related chronic disease. In the intervening decades since 1980, dietary lifestyles have not noticeably improved in the United States. Moreover, so-called lifestyle diseases, including diabetes, heart disease, cancer, osteoporosis, and especially obesity, have become more prevalent in the population, with dramatic increases in some conditions such as obesity and overweight. Although dietary guidance has become increasingly science based, there seems to be an ever-widening gap between the scientific evidence and consumer behavior. The 2010 Dietary Guidelines Advisory Committee (DGAC) report offers a new approach to making dietary recommendations and sounds a new note of urgency: the 2010 DGA are the first to call for a modification of the *food environment* and there is a new chapter on "translating and integrating the evidence, a call to action." The role of physical activity as part of the energy balance equation to reach public health goals has also been given a higher priority. The need for translating the evidence into real behavior change has never been greater, as has the need for appropriate communications to the public. As should be clear after 6 DGA reports in the past 3 decades, dietary change is not easy to achieve. It requires 2 key components to succeed: (1) food scientists working within industry or in academia need to reformulate product offerings or create new food products to help balance food choices available to consumers, and (2) consumers need to be truly motivated by nutrition science around human health. These concepts are similar to a push/pull scenario: industry creates the push by developing foods in line with the DGA and consumers pull by demanding healthier foods or, alternatively, by adopting new behaviors that will drive innovation and product reformulation. This paper focuses on the first component to address how food science can modify the food supply and the realities and challenges that pose barriers to change. Although Americans describe themselves as being relatively familiar with the DGA ([@b7]), this familiarity has yet to translate into meaningful modification of dietary lifestyles ([@b11]). There is even evidence that Americans are confused by the past 6 iterations of the DGA. A 2010 Centers for Disease Control and Prevention (CDC) report suggests that since the issuance of the 2000 and 2005 DGAs, vegetable consumption has shown zero change and average fruit intake has actually declined---with levels well below DGA targets ([@b4]). One might surmise that something needs to change. In the words of Linda Van Horn, a professor of preventive medicine at the Northwestern Univ. and chairman of the 2010 DGAC: "What has been done till now isn\'t working. To do nothing more effective than we have, means that five years from now we\'ll be in an even worse situation. And that would be unconscionable ([@b2])." The challenges are substantial. After so many years of concerted efforts to guide the public to dietary behavior in line with recommendations supported by a growing body of scientific evidence, how do we proceed at this point? In its "call to action," the DGAC offers an explicit prescription on how we are to proceed: > "A coordinated strategic plan that includes all sectors of society, including individuals, families, educators, communities, physicians and allied health professionals, public health advocates, policy makers, scientists, and small and large businesses (e.g., farmers, agricultural producers, food scientists, food manufacturers, and food retailers of all kinds), should be engaged in the development and ultimate implementation of a plan to help all Americans eat well, be physically active, and maintain good health and function. It is important that any strategic plan is evidence-informed, action-oriented, and focused on changes in systems in these sectors ([@b12])." The overall communication challenges around the DGA are well known. What is less well known are the challenges around modifying the food supply to aid in dietary behavior change. This paper takes those challenges as its focus in the hope that clarification will build a basis for understanding among all stakeholders in the food chain that are alluded to in the DGAC report. A Little History ================ U.S. government food and dietary guidance began in the early 20th century; food groups were identified and recommendations were issued to the public about how to choose foods from the different groups to achieve a healthy diet. A more rigorous process was initiated in 1977 with the "Dietary Goals for the United States," which was issued by the U.S. Senate Select Committee on Nutrition and Human Needs. A 1979 White House conference led to the formation of a panel of scientists to study the relationship between diet and health. The panel\'s report, *Healthy People: the Surgeon General\'s Report on Health Promotion and Disease Prevention*, was published in 1979, and the first DGA was enacted into law in 1980 ([@b1]). Since then, a committee of experts has been convened every 5 y to update the guidance, which is then duly reported to the Dept. of Health and Human Services (DHHS) and the United States Dept. of Agriculture (USDA), the 2 government agencies charged with promulgating the recommendations. Just as regularly as they are updated, the guidelines have been more or less ignored by the public. A recent Natl. Cancer Inst. statistical study of dietary intake patterns in the United States led to the following rather grim, but unsurprising, statement: "In conclusion, nearly the entire U.S. population consumes a diet that is not on par with recommendations. These findings add another piece to the rather disturbing picture that is emerging of a nation\'s diet in crisis ([@b8])." In compiling its report, the current DGAC had access to an unprecedented quantity of scientific research devoted to diet, which was collected through the USDA\'s continually updated Nutrition Evidence Library. The 2010 DGAC report was the first dietary guidance to explicitly recognize the importance of the environment, ubiquity of abundant calorie-dense food, and limited opportunities for physical activity. In calling for a multipartnered approach to make the new guidelines effective, the committee explicitly recognized the complexity of the process. Clearly, along with the communication challenges, there must be changes in the food environment. *T*he food supply needs to be modified such that it makes available healthful choices that are in alignment with the DGA accessible and desirable to the public. Prescription for Change ======================= The prescription for change poses a major challenge for food scientists. If food products are renovated or modified in ways that impinge negatively upon consumer perception, the modifications will fail to win consumer acceptance and will have no beneficial effect on dietary behavior. Common examples of this effect offered by food scientists are reduced-sodium products, which typically fail in the market because consumers perceive them as tasting badly (see the Burns roundtable abstract in the Appendices). In addition, there are significant financial barriers to reformulation. For example, expensive items include: product development costs, consumer research, higher cost of alternative ingredients, loss-of-opportunity costs, promotional expenses, and so forth. If food companies are simply to offer new products with healthier nutrient profiles, there are also significant hidden expenses for increased inventories and production inefficiencies, as well as for consumer confusion or lack of acceptance, among other reasons. If food products are fortified with healthier ingredients to better align them with the DGA, there may be unforeseen nutrition challenges. For example, if extra fiber or other plant nutrients are added to some foods, increased phytate levels may cause iron and other mineral absorption issues (see the Almeida roundtable abstract in the Appendices). Some of these concerns can be addressed ([@b9]) through higher intakes of foods containing vitamin C, as well as food preparation techniques such as soaking beans, grains, and seeds; leavening bread; using fermentation processes, and so forth ([@b5]). In addition, there is always the concern when adding beneficial nutrients that the processing required does not adversely affect the desired health benefits. The point is that these are not simple changes, but rather they are often complex and expensive modifications that are uncertain to appeal to consumers---not to mention the communication challenges around such reformulations or innovations. Underlying all of these challenges is the industry imperative to stay in business. To continue producing the nation\'s food, the food industry must take consumers to a large extent as they are, complete with desires, needs, and likes and with financial constraints (bearing in mind the importance of price points for food products). Marketing and the funding of continuing operations, including research and development, are critical to the bottom line. Furthermore, the DGA affords the food industry an opportunity to innovate and reformulate products to accomplish something extra beyond satisfying consumer palates and economic requirements. The DGA affords the industry an opportunity to confer health benefits to consumers and to embrace the role of acting as a *provider*, not simply as a merchant. Other opportunities may also exist because food and packaging technology can improve the quality, taste, and nutrition of packaged food products, and it can also help reach the goals of controlling portion sizes and reducing the environmental footprint. Call to Action: A Scientific Synergy ==================================== In 2010, a group of 4 science and science communications organizations undertook efforts in response to the DGAC\'s call to action to create synergy between food scientists and dietetic professionals through a robust scientific dialogue, with the ultimate goal of helping to integrate and translate DGA evidence into true behavior change. The 2 audiences, food scientists and dietitians/nutrition communicators/counselors, are key to achieving synergistic solutions for making dietary guidance effective, thereby reaching public health goals. Food scientists are tasked with innovating and renovating or reformulating food products, whereas nutritionists counsel clients and communicate dietary guidance to the public. The American Dietetic Assn., Inst. of Food Technologists, Intl. Food Information Council (IFIC), and the North American branch of the Intl. Life Sciences Inst. convened 2 expert roundtables of rigorous discussions, whose purpose was to enable the 2 key scientific audiences to interact, innovate, and close the knowledge gaps that are crucial to integrating and translating the DGA. As stated at the outset, the content of this paper is formed from the proceedings of the roundtables held in early October 2010 in Chicago, Illinois, and in Washington, D.C. The key DGA audiences---food scientists, dietitians/nutrition communicators, and government representatives---gathered with a select panel of speakers at each event. Guided by the DGAC report\'s new chapter urging translation and integration of the nutrition evidence, participants presented and critically analyzed ideas raised as the principal action themes of the report: - Reduce the incidence and prevalence of overweight and obesity in the U.S. population by reducing overall caloric intake; - Shift food intake patterns to a more plant-based diet that emphasizes vegetables, cooked dry beans and peas, fruits, whole grains, nuts, and seeds; and - Reduce intake of foods containing added sugars, solid fats (SoFAS), refined grains, and sodium ([@b12]). The 2010 DGAC physical activity recommendations were not specifically addressed because they fell outside of the primary expertise of the participating food scientists/nutritionists. The roundtables, in addition to the speaker panels, had other distinguished participants called "discussants," whose job was to draw out both speakers and other discussants on the opportunities and challenges in achieving real DGA change. The tone of the dialogues was proactive, and all present were urged to focus on success stories or paths deemed promising in accomplishing DGA and public health goals. Two well-known successes cited were the consumer switch to whole-wheat products and the move away from trans fats. In August of 2010, for the first time, sales of whole-wheat bread products surpassed those of refined wheat breads ([@b3]). There has also been a move away from products containing trans fatty acids in recent years. Food industry reformulations, following recommendations in the 2005 DGA that trans fat consumption be as low as possible, have resulted in the substantial replacement of trans fats in the American food supply with mono/poly unsaturated fatty acids ([@b10]). The Roundtable Presentations ============================ Each daylong roundtable began with an overview of DGA history, with special attention to the 2005 gaps and successes, and included a review of food supply changes since 2005. In the interest of brevity, abstracts of the presentations will be excerpted in this article. The complete presentation abstracts are included in the Appendices. 2005 gaps and successes and changes in the food supply since 2005 ----------------------------------------------------------------- Connie Weaver, 2005 DGAC member of the Purdue Univ., spoke of the emphasis in the latest DGAC report on obesity and on the excess of solid fats and added sugars (SoFAS) in typical diets, as well as on the recommendation that Americans shift to a more plant-based diet---more fruits and vegetables, dry beans, low-fat dairy, and less-lean meat than currently consumed. She said that a major challenge will be "to provide foods in affordable quantities that compete with foods rich in SoFAS and sodium for the palate of consumers." Challenges and opportunities for implementation of the 2010 guidelines ---------------------------------------------------------------------- Laina Bush, U.S. DHHS, presented results of government research into means of overcoming some of the barriers to compliance with the DGA among certain low-income and ethnic subpopulations. Generally, the research shows that significant barriers are sociocultural, culture-based food preferences, lack of readiness to change dietary behavior, and lack of personal or family preference for fresh fruits and vegetables. Food preparation customs were also cited as barriers to compliance with the DGA, as well as psychological distress and psychosocial stress. In addition, the high cost or perceived high cost of food was the most often cited barrier for 3 of the 5 groups studied. Joe Derochowski of the NPD Group, a global research firm dealing in consumer behavior and retail sales and marketing data, offered his take on consumer behavior in the face of dietary guidance. One of the hallmarks of American society\'s evolution over the past few decades has been the entrance of women, en masse, into the workforce. That event has altered American lifestyles in a major way: convenience rules, with its prepackaged meals, ready-to-eat food products, fast food, and casual restaurants as replacements for home-prepared meals, and the other trappings of our busy, multitask-laden, 24/7 modern lives. Cooking is becoming a largely forgotten skill. > "The growing need for convenience has given rise to quickly prepared meals, like frozen and ready-to-eat; appliances that enable food to be prepared quickly or with little or no effort; and drive-thru windows. While the percent of women working appears to have reached its peak since 2000, convenience remains at the center of this country\'s day-to-day lifestyles." Derochowski stated that the fundamentals of marketing apply also to the DGA. Food scientists and dietitians need to make it "convenient" for the consumer and for mom to integrate the DGA in the midst of all of their responsibilities. This includes the nature of each of the meal occasions---breakfast being about health, routine, and mobility; lunch being about speed; and dinner being about convenience. > "Mom is the key to fully integrating the Dietary Guidelines into her family\'s lifestyle. In order for her to accomplish this, we need to make it easy and seamless for her. The guidelines need to become part of the daily routine, quick and convenient to apply throughout the day, everyday. Since home is the primary source of meals, how can she easily implement the Dietary Guidelines into meal planning for her family?" The path forward: addressing 3 key concepts of the 2010 DGAC report ------------------------------------------------------------------- ### "Reduce the incidence and prevalence of overweight and obesity in the U.S. population by reducing overall caloric intake." Patricia Crawford, DrPH, Univ. of California at Berkeley, is a child obesity specialist and offered the group her view that the best strategy for curbing the prevalence of childhood obesity is to tackle the prevention of it, that is, the incidence---with education and behavior change as well as a healthy food environment, even among the youngest consumers. She pointed out that the 2010 DGAC, for the first time, examined the impact of the food environment on dietary intake and body weight---the environmental focus included restaurants, especially fast food restaurants, portion sizes both at home and away, food access at schools, and supermarket location, particularly as it relates to access to nutrient-dense foods. Richard Black, PhD, a nutrition scientist at Kraft Foods, offered the group some of his company\'s insights about the obesity issue: > "It has been proposed that, 'self-monitoring, including knowing one\'s own calorie requirement and the calorie content of foods, helps make individuals conscious of what, when, and how much they eat,' can be an effective approach to limit excessive energy intake ([@b12])." > > "...the large majority of the American public is unaware of its own caloric needs, and is equally naïve about the caloric content of most foods ([@b7]). Furthermore, daily variances in activity level compound the issue insofar as people are generally unable to estimate energy expenditure for most daily tasks, and for more intense bouts of exercise. And while it remains true that with a great deal of guidance (often from a registered Dietitian), or with the help of a commercial computer program, individuals may have some success in gauging their energy needs and their energy intake, counting calories is a tedious behavior and difficult to maintain in the absence of intense support." > > "While simply 'selling fewer calories' might seem the most obvious approach for the food industry, this is not a viable approach. As a thought experiment, consider the following: Company A decides to reduce the calories in its beverage product by 25%, and so aim to 'sell fewer calories.' However, when the public becomes aware of this lower calorie version of the drink, demand for the drink increases, and within a year, sales have increased by 35%. In effect, the Company A is now selling 10% more calories as a result of the calorie reduction. However, if consumers have switched from other higher calorie beverages to consume this lower calorie beverage, then the total calories sold by the food industry would be reduced." > > "Portion control is another successful approach to reduce energy intake. In this case, the food industry is ideally situated to provide products in portion-controlled servings, though consideration must be given to the potential for increased packaging and its environmental impact." > > "Interestingly, some efforts undertaken by the food industry to significantly reduce intake of calories and saturated fats seem to be dismissed as insufficient in the report of the DGAC. To whit, when a natural cheese is made from 2% dairy, it can be up to 45% reduced in saturated fat and over 30% reduced in calories. Similarly a slice of American processed cheese can be up to 50% reduced in saturated fat, and 30% reduced in calories. These are certainly behaviorally and health relevant changes, but the DGAC report recommends that Americans consume dairy only when it is either 1% or 0% milk fat. Therefore, 2% cheese is not seen as a viable dietary alternative to full fat cheese, despite the fact that it is palatable, affordable, available, and requires little if any change in diet patterns, whereas cheese made from 1% or 0% dairy is generally reported as unpalatable, somewhat more expensive, and requires significant commitment to incorporate into a diet plan." > > "Clearly, options to modify the diet do exist. However, communications about those options, and people\'s motivation to utilize those options, remains a challenge. Recommending dietary change that is so extreme as to be only aspirational rather than achievable will not serve the greater public need for dietary guidance to address the obesity epidemic. When all sectors work together, and acknowledge the limitations faced by each, it is possible to imagine the development of a dietary guidance plan that provides options for small changes, which over time can build one upon the next, and gradually lead the American public to healthier and more sustainable eating patterns. After all, we are asking people to fundamentally change how they think about food, shop for food, prepare food, and eat food. This will take time, patience, commitment and trust from everyone." ### "Shift food intake patterns to a more plant-based diet that emphasizes vegetables, cooked dry beans and peas, fruits, whole grains, nuts, and seeds." Nancy Keim, USDA, Agricultural Research Service, and Lindsay Allen, PhD, USDA, Agricultural Research Service Univ. of California at Davis, collaborated on a presentation: among other points, they suggested that consumer taste likes and dislikes, some of which are genetically based, are a major challenge to vegetable acceptability. Food preparation time is also a major constraint to increasing consumption of dry beans and peas. They had a suggestion for the food scientist participants: if the food industry could process quicker-to-cook forms of dry legumes, consumers might find them attractive. Nelson Almeida, PhD, FACN, a food scientist with Kellogg Co., summarized a strategic global study of potential industrial earnings for plant-based foods (across 200 markets, there is a projected growth of 40% from 2009 to 2014, with sales potentially increasing from \$39.1 to \$54.7 billion). > "The Mintel International Group analyzed attitudes towards food in a 2009 internet consumer research study of 2,000 demographically representative U.S. adults aged 18+, and Experian Consumer Research data from July 2007 to September 2005. Given the sharp rise in overweight Americans and various government, media and food company efforts to slow, stop or reverse this trend, consumers are more aware of their diet and the link of diet to health. 86% of U.S. adults said that healthy eating is very or somewhat important to them. Consumers of meat and dairy substitutes wish to improve their nutrition and health, manage their risk for heart disease, lose weight, and increase the safety of the food they eat..." > > "Currently, the wheat flour tortilla is the fastest growing product line of all grain-based products. This might be indicative of a growing interest in ... whole grains and/or fiber intake for health. Plant-based meals that contain more fiber and consumption of higher fiber plant foods over all meals throughout the day also allow for a higher intake of vitamins, minerals and bioactive phytochemicals, as well as promotion of satiety, intestinal health and reduction of chronic disease risk..." > > "It is the complement of the wide array of whole grains, fiber-based products, legumes, fruits, vegetables, nuts and seeds that makes a plant-based diet appealing, nourishing, tasty and healthful. Additions of fiber and bioactive portions of plants to the current set of food offerings is an important goal for food company research and development. It is that endless combination of predominantly plant-based meal selections made with simple, single ingredient foods as well as processed multi-ingredient foods that allow for balance, variety and moderation throughout the day..." > > "Desserts can also provide more plant-based appeal with inclusion of more fruit in the more traditional product offerings, like fruit streusels, but also with fruit as fat replacers, and additives such as dried fruit in cookies, muffins and quick-breads." > > "As Americans become better informed about healthy food combinations and practice these learnings, plant-based foods can become a greater part of the diet. As children and adolescents learn more about plant foods and plant-based diets through educational efforts in schools including cooking classes and community-based gardens with better communication on the Internet and advertising, it is hoped that future generations have a better understanding of food\'s connection to health, disease risk-prevention and well-being." ### "Reduce intake of foods containing added sugars, solid fats, refined grains, and sodium." Penny Kris-Etherton, PhD, The Pennsylvania State Univ., sees the gap between dietary recommendations and current consumer behavior along with heightened interest in diet and health as an opportunity to develop diet-improving interventions. She told participants that on an individual and group basis, cognitive-behavioral strategies have proved effective in behavior change---notable among these is motivational interviewing, with its well-ordered feedback and monitoring. Robbie Burns, formerly of Cadbury, represented the food scientist perspective on this issue and pointed out that "time delays between changes in dietary intake and health outcomes confound the ability to draw strong conclusions about the overall health impact of the DGA." > "Since their inception in 1980, Dietary Guidelines for Americans (DGA) have included, in some form, advice to decrease dietary intakes of added sugars, solid fats, refined grains and sodium. As a result and to meet consumer desires for more healthful products, the food industry has developed alternatives where all these negative components are reduced and in some cases eliminated." > > "In order to remain viable, food manufacturers must make products that meet consumer desires for taste, price and convenience (portable, easy to prepare, etc.; [@b7]). In most product categories, health aspects of the food have a lower priority than taste or price. Much of the progress towards meeting the DGA has been through decreasing saturated fat, trans fat, refined grains and/or sodium in ways that are not obvious to the consumer. However, some of the positive changes in the composition of food have been offset by increased food consumption." > > "The 2010 DGAC recommendations includes some notable changes from those of 2005; the goal for dietary saturated fat is decreased from less than 10% to less than 7% of energy; cholesterol targets are decreased from less than 300 mg/day to less than 200 mg/day and the sodium target is decreased from less than 2300 mg/day to less than 1500 mg/day. Simply raising the goal is unlikely to effect a greater rate of change in dietary habits unless accompanied by better ways to induce long-term changes in consumer behavior and/or specific engagement by food manufacturers in education efforts, creation of new products (e.g. to shift intake patterns; addressed elsewhere in this workshop) and/or reformulations of existing products." > > "Education efforts include information on websites and product packaging to highlight positive health messages such as the food pyramid. Food manufacturers can also modify marketing strategies to focus promotions on products that more closely adhere to the DGAC recommendations; e.g. by changing the relative advertising spend on diet or low-calorie beverages compared to sugar-sweetened beverages..." > > "Product reformulations within the realms of available technologies will not be sufficient to achieve the DGAC recommendations without changes in the pattern of foods consumed. However, for many foods that contribute significant quantities of added sugars, solid fats, refined grains, and sodium, reformulations can help meet goals when applied judiciously over a period of time. Note that it has taken over 20 years for consumers to change from full fat milk to lower fat options." > > "Food technology continues to provide tools to enable manufacturers to reduce undesirable components in certain foods. In some cases, the component can be replaced by a more desirable component such as polyunsaturated fat for solid fat and whole grain for refined grain. However, in other instances food additives are required. In these situations it is essential that the product not only meets consumer taste preferences, but also addresses their desire for simplicity and, in some cases, their fears of novel technologies." Strategic Priorities ==================== The roundtable discussions each day focused on the real-world challenges and opportunities of product innovation/renovation as well as education and communication in implementing the guidelines. Participants raised several main themes that were echoed and elaborated on throughout the sessions, as described below. - There needs to be a coordinated strategic plan with the active involvement of all sectors to achieve effective implementation. - Trust is a key factor in communicating dietary goals and modifying the food supply; mutual understanding and trust among all stakeholders is critical to implementation and requires input from industry and academic food scientists; public health professionals; dietitians, nutrition communicators, and counselors; and others in the food chain, as well as government. - The DGA should be viewed as aspirational with the bar set high. If the guidelines are seen as all-or-nothing goals, there would be no room to embrace or celebrate small changes and incremental dietary progress. - Consumer messages around nutrition and especially weight loss need to be even simpler and more targeted than the past communications of the DGA. - Sociocultural factors should be a major consideration in composing messages for consumers of different ethnic and demographic groups. - In considering the best evidence for what works and what does not work to improve consumer dietary choice, the best available evidence should point the way. - Children\'s nutrition education seems an optimal starting point for changing adult dietary patterns. - Behavioral science needs to be employed in drafting messages to influence consumer dietary choices ([@b13]). - Environmental modifications need to be part of any overall strategy in altering dietary patterns and need to be reinforced in messaging so as to enable the healthy choice to be the easy choice ([@b14]). - Food scientists, in striving to innovate and reformulate products, should employ both gradual modifications, or "stealth" methods, where consumers would perceive no change whatsoever in their favorite foods as they became healthier, and also education, or transparent means, whereby consumers would be encouraged to understand healthful modifications in their food. - There is a need to communicate more fully to the public the complexities inherent in enhancing nutrient profiles of their accustomed foods. Although the DGAC raised the issue of sustainability in its report, the roundtable participants acknowledged the issue but did not address it explicitly. Similarly, although the roundtable participants discussed the environmental focus of the DGAC, this topic was not addressed in depth. The broad thematic ideas listed above inspired a rich discussion in both roundtables, with participants producing a variety of ideas, suggestions, and insights that will be highlighted in the following paragraphs. Within these somewhat broad thematic concepts, there were numerous points offered in greater specificity. A coordinated strategic plan for all sectors -------------------------------------------- Food scientists working with industry emphatically made the point that there has been a lack of trust among the various stakeholders in the food chain and this must change if the country is to achieve a sustained dietary behavior change. To quote the DGAC, "A coordinated strategic plan that includes all sectors of society, including individuals, families, educators, communities, physicians and allied health professionals, public health advocates, policy makers, scientists, and small and large businesses (for example, farmers, agricultural producers, food scientists, food manufacturers, and food retailers of all kinds), should be engaged in the development and ultimate implementation of a plan to help all Americans eat well, be physically active, and maintain good health and function" ([@b12]). Participants in both roundtables overwhelmingly concurred, and some suggesting adding stakeholders that had not been listed by DGAC such as behavioral scientists and consumer advocates. Trust and mutual understanding ------------------------------ An often-repeated refrain in both roundtables was that: "Public health advocates don\'t trust industry; industry does n\'t trust government; there is too much mistrust currently for all interested groups to work together effectively." One food scientist participant pointed out that some policy-making scientific groups specifically prohibit industry-employed scientists from advisory panels, which is a prohibition he said he did not understand because there is an opportunity to provide industry expertise on such panels to widen and deepen the scientific knowledge base being drawn upon. Bringing together industry and academic scientists in a common public health cause would also create a growing basis for trust between food scientists and public health professionals. One approach might be to establish collaborative research teams of nutrition and food scientists to document that health benefits are still attained after the "form" of the nutrient component has been changed to meet food production and consumer preferences. Roundtable participants pointed out that an excellent example of trust building was provided in the discussions and sharing of viewpoints during the roundtables themselves. Food scientists and dietitians/nutrition educators/counselors were able to forge mutual understanding of their different perspectives as they strategized together around implementing the DGA. One major conclusion at both the Chicago and Washington meetings was the enormous, positive potential of collaboration: if all stakeholders can be helped to understand the demands on other players in the space and work together, they can attack the challenges from many angles, employing different approaches to achieve the same end. Aspirational goals, but reachable incrementally ----------------------------------------------- There was considerable discussion in both roundtables about taking a longer term view toward dietary change by taking account of intermediate dietary successes, rather than an all-or-nothing view. Industry food scientists pointed out that their companies had already reformulated a number of products to better align with the DGA, and much work along those lines is still being done. They suggested that given the key requirement of market acceptance, it was perhaps impractical to expect immediate and radical changes in dietary behavior. Intermediate dietary objectives and opportunities for interim behavior modification might be a more productive way to achieve broader public health goals. The incremental approach enables consumers to celebrate small improvements and motivates them to "stay the course" for greater change. As Voltaire observed in the 18th century, "the perfect is the enemy of good." Many roundtable participants agreed with this view; some made the point that reevaluating currently employed language to communicate nutrition messages with an eye toward simplification might also be productive. It was also suggested that dietary recommendations should not only be simple and realistic, but also few in number, with no more than 3 overarching messages at a time. Participants of both roundtables agreed that the public health community would benefit by promoting small steps in behavior change and developing ways to measure incremental dietary improvements. They also agreed that although product reformulations could achieve and are, in fact, already achieving small changes in consumer behavior, more can and still needs to be done. Taking account of culture ------------------------- Food scientists and others who communicate to consumers should take sociocultural factors into account when innovating/renovating products and composing messages. It is well known that different subpopulations have different taste preferences and dietary customs. Some participants pointed out that because of deep-rooted sociocultural traditions, shifting people\'s eating habits is extremely difficult. It has been noted that ethnic groups will often pay more for food with which they are familiar than for recommended healthier alternatives. Food scientists will be constrained in their development work by some of these attitudes; however, there is also an opportunity to tailor products, marketing, and nutrition messages to specific consumer groups. Considering the Hispanic American subpopulation, one participant pointed out that in communicating with these subpopulations, the messenger may be as important as the message (for example, *abuelas* or grandmothers, who are trusted advisers in food and health matters among Hispanic populations). The most effective dietary change strategies should be highly targeted dietary messages that are geographically and socioculturally targeted, and messages that do not link physical activity and dietary changes. It was also acknowledged that sociocultural conditions and consumer habits are realities for all populations, not just ethnic or underserved subpopulations. Dietary behavior patterns are set early in life ----------------------------------------------- Although obesity in general and childhood overweight in particular are among the most challenging issues raised in the 2010 DGAC, roundtable participants believe that there are some key opportunities for addressing obesity. Discussants made related points, suggesting that some of the most successful behavior-modification campaigns have started with school-age children (recycling, seatbelt safety, antismoking, and so on). One participant argued that in a highly structured environment, such as a calorie-regulated summer camp, research shows that given a healthy food and activity environment, "every biomarker for children\'s health improves." Participants observed that nutrition education at the very earliest educational stages could yield the most profound results, especially when linked with a healthy food and activity environment. By also targeting the parents of young children with consistent dietary-behavior messages, there may also be a chance to reach children through their parents. One participant made the point that involving mothers in children\'s education outreach is critical, and that any education of the child should be transferable to his or her mother either through sharing learning materials or inviting mothers to participate in their children\'s education. Another discussant referred to a new cell phone "app" that is capable of tracking energy balance, stating that such devices could well help educate children through their "cool factor." Because celebrity chefs have become popular invitees into schools, a similar opportunity may exist for dietitians and nutrition educators to follow suit. There was wide agreement among participants to support the DGAC\'s recommendation to stress cooking skills, both in family settings and in schools. Using Strategies That Work ========================== Along the lines of messaging, some participants drew on an [@b6] recommendation that when definitive research conclusions are not available, policy makers should work with the best evidence available, instead of delaying policy. The roundtable participants agreed that using the best available evidence about what works and what does not work should be the basis for action, while recognizing that the American public cannot be fitted with a one-size-fits-all dietary strategy. This is an opportunity for dietitians, nutrition communicators, and counselors to personalize guidance based on the needs and desires of individual consumers. Calling psychologists and behavioral economists ----------------------------------------------- Roundtable participants argued on multiple occasions that behavioral science must be part of the dietary change process, both in modifying preferences and in drafting messages to influence consumer dietary choices. Some of this work has already begun at the [@b13]. There was considerable discussion of cognitive-behavioral strategies in improving consumer diets, although these tend to be highly labor intensive and expensive. Food scientists formulating new and existing products already employ strategies keyed to consumer preferences (see the discussion of "stealth" modifications in the next paragraph). Industry representatives among the roundtable participants made the point that although products can easily be launched, they will not succeed in the marketplace without consumer acceptance, which involves meeting consumer expectations of taste, convenience, and cost. Stealth versus persuasion ------------------------- Consumers have been resistant to dietary change, partly because of established food preferences: "stealth" methods of change are potentially effective because they do not upset those established preferences. There are existing examples of food products that have been successfully modified to have a healthier nutrition profile without consumers\' perception of any change. The highest-profile example is the industry-wide conversion from trans fatty acids to mono- and polyunsaturated fats in product formulations, as noted earlier in this article. And of course, food innovations or reformulations would still be transparent through the product labels, both on the ingredient listing and the nutrition facts panel. Food scientist participants of the roundtables informed their dietitian colleagues that food companies typically spend 60% to 70% of their research and development budgets on renovation and only 30% to 40% on innovation of new food products. Reformulations are therefore the most efficient way to produce more healthful foods. ::: {#fig01 .fig} Figure 1 ::: {.caption} ###### Consumer Healthy Eating Trends 1992 to 2010 (proprietary marketing research, NPD Group \[<http://www.npdgroup.com>\]). :::  ::: Discussants made the point that consumer resistance to dietary change may be an opportunity for both the dietetic and food science communities. As indicated earlier in this paper, dietary change could be effected on a push/pull basis, with nutrition communicators persuading consumers to demand foods in line with the DGA and food scientists innovating and reformulating products to meet that demand and create a marketplace of healthier alternatives. Changing Nutrient Profiles: No Simple Task ========================================== Roundtable participants heard that 80% to 90% of new food products fail to achieve market acceptance. Some dietary change advocates have argued that the food industry, with its highly persuasive and well-financed marketing departments, can simply produce products with healthier nutrition profiles and then "sell them" to consumers. The 80% to 90% failure rate of new products is a sobering reality check. Both roundtable groups reached a similar conclusion: there is a knowledge gap between the production of food and the selection and consumption of food. Consumers could better understand the challenges of food manufacturing both in terms of reformulation of products without changing taste, appearance, and cost, and in terms of aligning products with the DGA. A public preference for extremely short ingredient lists on processed food products also poses major challenges to food scientists in renovating/reformulating food. About the Environment ===================== The 2010 DGAC raised a new issue in its report: the food-related physical environment, from the ubiquity of restaurants and food markets to the "built environment," which limits opportunities for physical activity. Roundtable participants considered the environment a key consideration, and urged all stakeholders to develop strategies recommending modification of environmental barriers to adopting the DGA. They stressed that in the absence of such strategies, efforts to effectively translate the DGA into behavior change are unlikely to succeed. In the words of a World Health Organization report, the goal should be to enable the healthy food choice to be the easy choice ([@b14]). Consensus Findings ================== The most important consensus findings from the 2 roundtables, and those themes enjoying the greatest consensus, were as follows: - There is a critical need for a coordinated strategic plan with the active involvement of all sectors, including industry, academia, public health professionals, and government, to achieve effective implementation of the DGA. - Both goals---modification of the food supply and more impactful communication strategies---are critical to achieving desired public health outcomes. - In order for the DGA to have its maximum impact on public health, new approaches need to be employed, such as setting strategic priorities, realistic public health objectives, and placing greater emphasis on practical solutions (for example, food product renovations to achieve incrementally changed dietary behavior). - Behavioral science needs to be brought to bear on the challenges. - Dietary messages need to be positive, very simple, few in number, and targeted to subpopulations, both to inform and motivate consumers. - Messages need to take account of sociocultural factors, consumer habits, and the realities of today\'s lifestyles. - To address the obesity epidemic, a key focus should be on very young children and on their parents. - Above all, trust and collaboration are essential among all stakeholders, including food industry scientists, public health community representatives, government agencies, nutrition communicators, retail food industry organizations, environmental planners, and others. Perhaps an overarching theme of the roundtables, not adequately captured in the above list, was "practicality." Those interested in dietary behavior change need to be practical about it by accepting intermediate successes and incremental gains, being patient with change (one food industry scientist pointed out that it has taken more than 20 y for consumers to accept lower fat dairy products \[see the Burns abstract in the Appendices\]), and combining food reformulations with consumer education (because both strategies are more successful when combined). Patience with making these changes is important from the food scientist\'s perspective. One participant pointed out that substituting noncaloric sweeteners for sugar in products can be moderately successfull in reducing consumer intake of added sugars, but there have been consumer acceptance issues with "artificial" sweeteners. Furthermore, "natural" sweeteners, such as Stevia with its bitter aftertaste, may require further additives to gain widespread consumer acceptance. And then there is also the issue of consumer preference for few additives. Similarly, lower fat reformulations may work well in some products, but not all. One scientist observed that: "in products such as cheese and chocolate the physical and sensorial properties of the saturated fats are such that significant reductions are not acceptable to consumers." (See the Burns abstract in the Appendices.) In addition, standards of identity for some products, such as certain cheeses, limit reformulation. Patience is a key virtue in guarding against excessive expectations, and food industry scientists need to be wary of unintended consequences. In the world of consumers where choice is the currency, in most product categories, the health aspects of the food have a lower priority than taste or price. Patience and perseverance are clearly virtues as stakeholders pool their efforts and abilities to bring the food supply more in line with the DGA. Regarding messaging, the following question was brought up at the roundtables and bears some further thought: "Is there some way to articulate the advantages of good nutrition than simply saying it is 'for good health'?" The last point in the above list of roundtable conclusions was clearly the one garnering greatest consensus and was regarded by participants at both roundtables as an essential component of any communication and dietary behavior change program. For a host of reasons, mutual trust and understanding of complementary roles and responsibilities have eroded over the years and need to be rebuilt. Understanding between dietitians/nutrition communicators/counselors and food scientists is decidedly crucial to restoring trust, and collaboration between these 2 groups is critical to pursuing the dietary changes necessary to reach public health goals. Small successes, small steps, and open, transparent processes will do much to build trust among the multiplicity of stakeholders critically interested in seeing Americans\' health and dietary regimes reach desired DGAC-recommended goals. We would like to acknowledge with gratitude the participation of the speakers, discussants, and staff for their contributions to the roundtables. A special thanks to Joe Derochowski of the NPD Group for his contributions to the roundtables and papers. We also would like to extend further thanks to Sylvia Rowe and Nick Alexander for organizing the abstracts into a concise and seamless whole that truly is greater than the sum of its parts, and to Ms. Rowe again for acting as moderator of the roundtables. The following individuals attended the roundtables. Speakers/Authors ================ Nick Alexander, SR Strategy LLC; Nelson G. Almeida, Kellogg Co.; Richard Black, Kraft Foods Inc.; Robbie Burns, Nutrition Implications, LLC; Laina Bush, DHHS; Patricia Crawford, Univ. of California, Berkeley; Joe Derochowski, NPD Group; Nancy Keim, USDA Agricultural Research Service; Penny Kris-Etherton, The Pennsylvania State Univ.; Sylvia Rowe, SR Strategy LLC; and Connie Weaver, Purdue Univ. Discussants =========== Jeanne Blankenship, American Dietetic Assn.; Dondeena Bradley, PepsiCo Inc.; Mary Christ-Erwin, Porter Novelli; Janet Collins, DuPont; Suzie Crockett, General Mills Inc.; Johanna Dwyer, Natl. Inst. of Health; Robert Earl, The Coca-Cola Co.; Cecilia Fileti, Latino Health Communications; Constance Geiger, Geiger & Associates; Marianne Gillette, McCormick & Co., Inc.; Jeanne Goldberg, Tufts Univ.; Cathy Adams Hutt, RdR Solutions; Barbara Ivens, ConAgra Foods, Inc.; Guy Johnson, McCormick Science Inst.; Michelle Matto, Intl. Dairy Foods Assn.; Kathy McMahon, Sara Lee Corp.; Melissa Musiker, Grocery Manufacturers Assn.; Jill Nicholls, Natl. Dairy Council; Susan Nitzke, Univ. of Wisconsin; Jessie Pavlinac, Oregon Health & Science Univ.; Mary Pat Raimondi, American Dietetic Assn.; Judith Rodriguez, Univ. of North Florida; Leila Saldanha, NutrIQ LLC; Marilyn Schorin, Schorin Strategies; and Pamela Starke-Reed, Natl. Inst. of Health. Observers ========= Carole Davis, USDA Center for Nutrition Policy and Promotion; Kathryn McMurry, HHS Office of Public Health and Science; and Rob Post, USDA Center for Nutrition Policy and Promotion. Staff ===== Will Fisher, Inst. of Food Technologists; Eric Hentges, ILSI North America; Esther Myers, American Dietetic Assn.; Sarah Ohlhorst, Inst. of Food Technologists; Wendy Reinhardt Kapsak, IFIC; Marianne Smith-Edge, IFIC; Lisa Spence, American Dietetic Assn.; and Heather Steele, ILSI North America. This paper is the product of a series of roundtables held in October 2010 on translating and integrating the science in the 2010 Dietary Guidelines. The roundtables focused on the opportunities and challenges for 2 key audiences: dietitians and food scientists. This was a collaborative effort between the American Dietetic Assn., the Inst. of Food Technologists, the IFIC, and the North American branch of the Intl. Life Sciences Inst. Supporting Information ====================== The following supporting information is available for this article. Abstracts of Roundtable Speaker Presentations are available at <http://www.ilsi.org/NorthAmerica>. Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. [^1]: Authors Rowe and Alexander are with SR Strategy LLC, Washington, DC 20036, U.S.A. Author Almeida is with Kellogg Co., Battle Creek, MI 49017, U.S.A. Author Black is with Kraft Foods Global, Inc., Glenview, IL 60025, U.S.A. Author Burns is with Nutrition Implications, LLC, Montville, NJ 07045, U.S.A. Author Bush is with U.S. Dept. of Health and Human Services, Science and Data Policy, Department of Health and Human Services, Washington, DC 20201, U.S.A. Author Crawford is with Univ. of California, Berkeley, CA 94720, U.S.A. Author Keim is with U.S. Dept. of Agriculture, Agricultural Research Service, Davis, CA 95616, U.S.A. Author Kris-Etherton is with The Pennsylvania State Univ., University Park, PA 16802, U.S.A. Author Weaver is with Purdue Univ., West Lafayette, IN 47907, U.S.A. Direct inquires to author Rowe (E-mail: <rowe@srstrategy.com>).

PubMed Central

2024-06-05T04:04:19.340139

2011-1-01

PMC3052441

Introduction ============ In recent years, the relationship between fruit and vegetable intake and health has been the focal point of much scientific investigation, in attempts to identify the specific plant components that convey health benefits. Fruits and vegetables, apart from being good sources of vitamins, minerals, and fiber, are also rich sources of potentially bioactive compounds known as phytochemicals. Antioxidants are abundant phytochemicals that prevent some of the processes involved in the development of cancer and cardiovascular disease ([@b14]). Evidence for their role in the prevention of other diverse degenerative diseases is continuously emerging. The bioaccessibility and bioavailability of each antioxidant differs greatly, and the most abundant antioxidants in ingested fruit are not necessarily those leading to the highest concentrations of active metabolites in target tissues ([@b39]). Several factors interfere with the bioavailability of antioxidants, such as food source and chemical interactions with other phytochemicals and biomolecules present in the food ([@b46]). Fruit antioxidants are commonly mixed with different macromolecules such as carbohydrates, lipids, and proteins to form the food matrix. In plant tissue, carbohydrates are the major compounds found, mainly in free and conjugated forms ([@b38]). Dietary fiber, the indigestible cell wall component of plant material, is considered to play an important role in human diet and health ([@b1]; [@b50]; [@b43]; [@b7]; [@b12]). However, there is evidence indicating that these complex carbohydrates directly interact with the food antioxidants and interfere with the adequate assimilation of these compounds ([@b19]; [@b46]; [@b48]; [@b8]; [@b47]). Most studies on antioxidant bioavailability are focused on foods and beverages from which antioxidants are easily released ([@b21]; [@b20]; [@b37]; [@b36]). Research concerning the bioaccessibility of phenolic compounds and other antioxidants from solid matrices are important, since only the compounds released from the food matrix and/or absorbed in the small intestine are potentially bioavailable and in a condition to exert their beneficial effects ([@b62]). Previous studies have reviewed the evidence indicating that the food microstructure affects the bioaccessibility and bioavailability of several nutrients, referring mostly to antioxidants ([@b46]). Nevertheless, the specific role of dietary fiber in the absorption of antioxidants has not been widely discussed. In this context, the purpose of the present review is to compile and analyze evidence concerning the association of dietary fiber with antioxidants during ingestion of fruit and vegetables, through physical and chemical interactions that modulate their release from the chyme to the gastrointestinal tract. Link between fruit and vegetable intake and human health -------------------------------------------------------- The link between diet and health has been recognized, since ancient times when physicians treated their patients with herbs and foods believed to have medicinal properties ([@b61]). Several epidemiological studies have shown an association between the consumption of diets rich in fruits and vegetables and a lowered risk for chronic diseases such as cancer ([@b61]; [@b64]), heart disease ([@b27]; [@b31]), and stroke ([@b23]; [@b30]). A reduced risk of obesity and better control of diabetes are some additional benefits that are likely to follow from increased consumption of plant foods ([@b61]). While the specific patterns of plant components responsible for this association remain to be established, there is evidence for a greater effect at higher levels of fruit and vegetable consumption, in line with current recommendations to increase fruit and vegetable intake to a minimum of 400 g/d ([@b45]; [@b8]). Fruits and vegetables, apart from being good sources of vitamins, minerals, and fiber, are also rich sources of potentially bioactive compounds known as phytochemicals. These compounds are not considered as nutrients, but much of the disease prevention potential of fruits and vegetables in human health is thought to be provided by these compounds ([@b14]; [@b8]; [@b24]). A popular explanation among scientists, and more recently in the mass media, has been that food components with antioxidant properties (including vitamins C and E, selenium, flavonoids, and ß-carotene) present in these foods may prevent some of the processes involved in the development of cancer (for example, by protecting DNA from oxidative damage) and in the development of cardiovascular disease (for example, by inhibiting oxidative damage to LDL-cholesterol) ([@b11]). Generalities of phenolic compounds and carotenoids -------------------------------------------------- Several thousand known molecules having a phenolic structure (that is, several hydroxyl groups on aromatic rings) have been identified in higher plants, and several hundred are found in fruits and vegetables ([@b39]; [@b11]). These molecules are secondary metabolites of plants and are generally involved in defense against ultraviolet radiation or physiological damage by pathogens ([@b11]). These compounds may be classified into different groups as a function of the number of phenol rings that they contain and of the structural elements that bind these rings to one another ([@b39]). Distinctions are thus made between the phenolic acids, flavonoids, stilbenes, and lignans. The flavonoids, which share a common structure consisting of 2 aromatic rings (A and B) bound together by 3 carbon atoms forming an oxygenated heterocycle (ring C), may themselves be divided into 6 subclasses as a function of the type of heterocycle involved: flavonols, flavones, isoflavones, flavanones, anthocyanidins, and flavanols (catechins and proanthocyanidins). In addition to this diversity, polyphenols may be associated with various carbohydrates and organic acids as well as with one another ([@b38]; [@b63]). Polyphenols are probably the most investigated molecules of nutritional interest. Expansive prospective studies, as well as small cross-sectional observations and interventions, have scientifically linked this class of molecules to demonstrated beneficial effects in terms of human health. On this basis, numerous studies have tried to evaluate through which mechanisms polyphenols exert their health benefits ([@b3]; [@b66]; [@b38]; [@b39]; [@b8]). These molecules have demonstrated several biological effects, as tested *in vitro* or *ex vivo*. They can inhibit the proliferation of cancer cells, reduce vascularization, protect neurons against oxidative stress, and stimulate vasodilation and improve insulin secretion ([@b20]; [@b58]). On the other hand, carotenoids and related compounds are the colors of nature. They consist of a group of over 600 naturally occurring colored pigments that are widespread in plants, but only about 24 commonly occur in human foodstuffs ([@b9]). In the plant, they serve 2 essential functions as accessory pigments in photosynthesis and photoprotection. These roles are achieved through the polyene structure of carotenoids, which allows the molecules to absorb light and to quench, or inactivate, singlet oxygen, and free radicals ([@b59]). Dietary intake of carotenoid-rich fruits and vegetables has been associated with reduced risk for a variety of common diseases including multiple types of cancer, cardiovascular disease, macular degeneration, and cataract formation ([@b54]). Carotenoids possess antioxidant properties that have been associated with cellular protection ([@b41]), regulation of cell growth, differentiation, and apoptosis ([@b63]). However, it is now well established that antioxidants undergo substantial metabolism after being ingested by humans in dietary relevant amounts and that concentrations of plasma metabolites after normal dietary intake rarely exceed nanomolar levels ([@b39]). Phenolic compounds and carotenoids bioaccessibility and bioavailability ----------------------------------------------------------------------- The *in vivo* effects of antioxidants depend not only on their concentrations in fruits and vegetables, but also on their bioaccessibility and bioavailability after ingestion. Many studies have focused on the bioavailability of these compounds after their ingestion ([@b39]; [@b46]; [@b47]). Understanding the concept of *bioavailability* is essential to all persons involved in food production and nutritional assessment as well as for determining diet and health relationships ([@b38]; [@b39]). Bioavailability is defined as the proportion of an antioxidant that is digested, absorbed, and utilized in normal metabolism; however, measurement of bioavailability relies heavily upon estimates of amounts of antioxidant absorbed. On the other hand, *bioaccessibility* is a commonly used term defined as the amount of an ingested nutrient that is available for absorption in the gut after digestion ([@b26]). In these terms, the bioavailability strictly depends on the bioaccessibility. In general, the absorption and transport processes of many of the potentially bioactive components of fruits and vegetables are complex and not fully understood; thus, prediction of their bioavailability is problematic. In this review, 2 major groups of antioxidants are considered: the phenolic compounds and carotenoids. Initially, the absorption of polyphenols from the diet was believed to be negligible, given that the majority of food flavonoids are bound to glycosides ([@b53]). It was expected that only the aglycones could pass freely into the blood stream from the gut wall, because no specific enzymes are secreted in the gut that can cleave glycosidic bonds ([@b39]). Recent studies, however, have demonstrated that the bioavailability of specific flavonoids and phenolic compounds is much higher than previously believed. The overall process is shown in [Figure 1](#fig01){ref-type="fig"}. ::: {#fig01 .fig} Figure 1 ::: {.caption} ###### General human bioabsorption of phenolic compounds contained in beverages or similar foods poor in fiber. :::  ::: The metabolism of phenolic compounds from beverages and food lacking dietary fiber practically starts in the lumen of the small intestine and postabsorption modifications also occur in the liver and other organs ([@b39]; Mullen and others 2006). In the case of polyphenols, as in the flavonoids, the majorities of these compounds are absorbed in some form from the intestine after consumption from food and beverages and can pass through the gut wall into the blood stream ([@b36]). In order for the absorption of flavonoids from the small intestine to occur, the sugars (glycosides) that are attached to the flavonoid skeleton must first be removed ([@b47]). This process is controlled by the action of enzymes manufactured in the small intestine (for example, mammalian β-glucosidase), resulting in the release of the flavonoid skeleton (the aglycone) from its sugar ([@b14]). Furthermore, there is also a growing appreciation for the fact that a significant portion of ingested dietary flavonoids and related compounds is not absorbed in the small intestine but rather passes to the large intestine. Here, they are degraded by the colonic microflora to simple phenolic acids, which can be absorbed into the circulatory system or exert antioxidant activity in the intestine environment ([@b8]). Scientists have found that the type of sugar attached to the flavonoid skeleton determines the site and the extent of absorption of glycosylated flavonoids ([@b15]). The location at which the sugar is attached to the flavonoid skeleton affects the mechanism by which glycosylated flavonoids are absorbed ([@b15]). Once the sugars have been removed from the skeleton for absorption, flavonoids are further metabolized in the gut and subsequently in the liver and kidneys to produce a vast number of modified molecules, known as flavonoid secondary metabolites. This process involves further conjugation of the flavonoid through the joining of glucuronate, sulfate, or methyl groups. In the intestine, this process is controlled by enzymes produced by the gut bacteria that play an important role in the metabolism of plant bioactive compounds ([@b14]). Flavonoid secondary metabolites can be detected in the blood and urine following the consumption of flavonoid-containing fruits and beverages, but only very small quantities of nonconjugated flavonoids in their original form can found in the same specimens ([@b34]). This implies that flavonoid secondary metabolites enter the circulation, and evidence suggests that it is these secondary metabolites, rather than the native flavonoids found in fruits and vegetables, that exert biological effects in the body ([@b34]; [@b14]). Research concentrated on the polyphenolic compounds found in fruits, vegetables, and beverages indicates that these molecules are not those that are transported through the human body in the circulatory system, or reach body tissues to elicit bioactive effects ([@b16]). Instead, it is their metabolites, formed in the small intestine and hepatic cells, and low molecular weight catabolic products of the colonic microflora that elicit antioxidant effects. Understanding transport and modification of these compounds certainly carries interest for drug discovery but also for dietary prevention of disease ([@b8]). On the other hand, the phenolic acids, the main phenol compounds in the human diet, are categorized in 2 large classes: benzoic acid derivatives and cinnamic acid derivatives ([@b33]). Both classes of molecules have been found to be present in fruits and vegetables at varying concentrations ([@b31]; [@b57]). However, improving our knowledge of phenolic acid bioavailability is an essential step in understanding their health effects. Even if phenolic acids are the main phenol compounds consumed, the bioavailability of these molecules has not received as much attention as that of flavonoids ([@b36]). Phenolic acids in the aglycone form are generally absorbed in the upper part of the gastrointestinal tract ([@b57]). Recently, it was shown that the stomach constitutes an active absorption site of numerous phenolic acids as gallic caffeic, ferulic, coumaric, and also chlorogenic acids can be absorbed from the stomach ([@b33]; [@b37]; [@b36]). This fact explains the rapid absorption of these compounds ranging from 1 to 2 h after intake of fruits and vegetables. The small intestine constitutes another absorption site. However, aglycone phenolic acids can be absorbed to different degrees---19.1% compared with 56.1% for caffeic and ferulic acids, respectively. In general, when the phenolic acids are esterified, the bioavailability decreases, reaching only 0.3% to 0.4% absorbed from the original intake. This is because esterified phenolic acids must be hydrolyzed in the enterocytes before reaching the blood circulation and the enzymatic machinery of these intestinal cells is not efficient enough to hydrolyze the ester bonds (Adam and others 2002; [@b37]). Carotenoids are hydrophobic and their absorption depends upon efficient release from the food matrix and subsequent solubilization by bile acids and digestive enzymes, culminating in their incorporation into micelles ([Figure 2](#fig02){ref-type="fig"}). Carotenoids are disassociated from their native environment in the plant tissue during food processing and digestion (acidic conditions and enzymatic hydrolysis) in the stomach ([@b32]). Also, carotenoids must also be released from their native environment during mastication, where mechanical forces and saliva enzymes disrupt tissues and cellular compartments ([@b46]). The main part of carotenoid metabolism occurs in the small intestine where they must also be dissolved in dietary lipids before they can be absorbed. Dietary lipids have been considered to be an important factor for stimulation of bile flow into the intestine and micelle formation ([@b21]; Roodenburg and others 2000). Lipids are capable of attaching to carotenoids due their hydrophobic character, but also may attach to water molecules that, in conjunction with bile salts, result in the formation of micelles and solubilized carotenoids in the system ([@b28]). This allows carotenoids to be absorbed passively from the micellar phase trough the lumen of the intestine to the lymphatic and blood circulatory system. However, it is unknown if all the carotenoids present in a mixed micelle are absorbed, or whether some are left behind in association with unabsorbed bile salts and cholesterol to be absorbed more distally or lost to the large intestine where they can exert their antioxidant properties. ::: {#fig02 .fig} Figure 2 ::: {.caption} ###### General human bioabsorption of carotenoids contained in food poor in dietary fiber. :::  ::: Because it is recognized that the carotenoids are not actively absorbed by the gut but are passively absorbed along with lipids, the efficiency of absorption of carotenoids is dependent on dissolving lipophilic molecules into dietary lipids ([@b28]). This may happen during food preparation and during the digestive processes. It is now recognized that this is a key process in absorption and may well be the single most important factor governing the rate and limit of absorption. It is not surprising, therefore, to find greater bioavailability from heat-treated foods that have also been coprocessed with oils ([@b60]; [@b22]). However, limited information exists on the bioavailability of dietary lipids and carotenoids entrapped in the food matrix. Soluble dietary fiber in the gut could attenuate the absorption of dietary fats and may, therefore, also inhibit the absorption of carotenoids as lipid soluble compounds ([@b54]; Unlu and others 2005). Antioxidant releasing and bioaccessibility ------------------------------------------ The first physical transformation of fruit and vegetable matrices during eating occurs in the mouth, and mastication is considered the initial step in the digestion of foods. Mastication consists of grinding food into small pieces and impregnating these pieces with saliva to form a bolus that is able to be swallowed. Decreasing the particle size enlarges the surface area available for action by digestive enzymes, thus increasing the overall digestion efficiency and the gastrointestinal absorption of antioxidants ([@b35]). As stated before, bioaccessibility is defined as the amount of a food constituent that is present in the gut, as a consequence of the release of this constituent from the solid food matrix and that may be able to pass through the intestinal barrier. Only antioxidants released from the fruit and vegetable matrix by the action of digestive enzymes (small intestine) and bacterial microflora (large intestine) are bioaccessible in the gut and therefore potentially bioavailable ([@b57]). Bioaccessibility is not taken into account in studies regarding the bioavailability of polyphenols. Moreover, most studies on polyphenol bioavailability use mainly pure single molecules (isolated from food or chemically synthesized), some beverages, and single foods; however, the bioavailability from whole foods may be substantially different ([@b39]). The influence of dietary fiber on the absorption of carotenoids and phenolic compounds -------------------------------------------------------------------------------------- There is ample evidence that the physical state of the food matrix plays a key role in the release, mass transfer, accessibility, and biochemical stability of many food components ([@b2]; [@b46]). Antioxidants are often located in natural cellular compartments or within assemblies produced during processing. In either case, they need to be released during digestion so that they can be absorbed in the gut ([@b46]). Furthermore, less is known about the interactions of antioxidants with other food components, such as dietary fiber. It is known that dietary fiber can reduce the bioavailability of macronutrients, especially fat, and some minerals and trace elements in the human diet. Because it was demonstrated in humans that pectin strongly decreased the bioavailability of β-carotene ([@b55]), dietary fiber is suspected to also affect the absorption of other carotenoids and probably that of α-tocopherol and polyphenols compounds. In general, the 2 main effects of dietary fiber in the foregut are to prolong gastric emptying time and to retard absorption of nutrients. Both are dependent on the physicochemical form of the fiber, and in particular, on its influence on digesta viscosity. Dietary fiber can act in the small intestine in 3 main physical forms: as soluble polymer chains in solution, as insoluble macromolecular assemblies, and as swollen, hydrated, sponge-like networks ([@b17]). The principal physiological effect of dietary fiber in the small intestine is to reduce the rate (and in some cases the extent) of release of nutrients or antioxidants ([@b5]). The dominant factors involved in the influence of dietary fiber on antioxidant digestion are (i) physical trapping of antioxidants within structured assemblies such as fruit tissue, and (ii) enhanced viscosity of gastric fluids restricting the peristaltic mixing process that promotes transport of enzymes to their substrates, bile salts to unmicellized fat, and soluble antioxidants to the gut wall ([@b43]). Secondary factors may include binding of bile salts (and perhaps enzymes) to specific fiber components and inhibition of diffusion across the unstirred layer ([@b1]; [@b17]). The rate of release of antioxidants from fibrous particles into the surrounding intestinal fluid is inversely proportional to particle size and is directly proportional to solute gradient ([@b1]; [@b7]). It is also affected by the following factors: the physical state of the solute (for example, whether it is present in solid form or is already dissolved in water trapped within the particle); the physical structure of the particle (for example, whether it is readily deformed, like a sponge, so that dissolved solids can be squeezed out by peristaltic contractions, or rigid, so that solutes must diffuse out); and the surface properties of the particle (for example, surface-tension effects) ([@b12]). The concentration of antioxidants within the continuous aqueous phase is constantly depleted by enteric absorption and replenished, as outlined above, by the release of material from food particles. The progress of these sequential release processes is, of course, also influenced by transit time (that is, the duration of exposure to a particular absorptive surface or digestive environment) ([@b17]). In addition to increasing the viscosity of the luminal contents, dietary fiber may reduce rates of antioxidant absorption mainly by physically trapping the antioxidants within the fiber matrix in the chyme. The chyme may be considered a 2-phase system with a discontinuous particulate phase dispersed in a continuous liquid phase. Antioxidants trapped within the particles must first be released into the continuous solution phase before they can be absorbed through the gut wall ([@b44]; [@b17]). Chemical interactions between polar groups from polyphenols and fiber polysaccharides may occur, but this has not been well studied ([@b17]; [@b47]). A significant number of studies concerning carotenoid absorption have been published in recent years. Most of these studies discuss the modulating factors previously reported. All of these data demonstrate the significant role exerted on carotenoid bioavailability, for example, by the physical properties of the food matrix that are able to influence digestive processes and thus absorption. Zhou and others (1996) suggested that the matrix, probably pectin-like fibers, and the crystalline form of carotenoids in carrot chromoplasts were the primary factors that reduced the relative bioavailability of carotenoids from carrots (so-called "incomplete release"). The effect of dietary fiber as pectin on β-carotene response after supplementation is known from earlier work. Studies with chicks and humans suggested that the addition of pectin to chow or test meal reduced the bioavailability of β-carotene ([@b29]). [@b52] also showed that pectin decreased the bioavailability of β-carotene by 42%. In addition, Torronen and others (1996) showed a 70% difference in the bioavailability of β-carotene from raw carrots compared with carrot juice consumed by well-nourished adult females for 6 wk, although the difference was not significant. In another study, the relative bioavailability of β-carotene from vegetables compared with purified β-carotene ranged between 3% and 6% for green leafy vegetables, 19% and 34% for carrots, and 22% and 24% for broccoli ([@b42]; [@b6]; [@b29]; [@b48]). This has also been widely demonstrated for lycopene ([@b49]) in tomato products (for example, tomato puree and paste are more bioavailable sources of lycopene than raw tomatoes). The same is true for β-carotene in spinach ([@b54]) in which the plasma response after the ingestion of pureed and thermally processed spinach was observed to be higher than that measured after the intake of raw vegetables. The presence of dietary fiber in vegetables and fruits may explain in part the lower bioavailability of carotenoids from plant foods. A general overview of the interference of dietary fiber in the absorption of carotenoids is shown in [Figure 3](#fig03){ref-type="fig"}. The release of carotenoids from plant foods occurs only when the cells in the fruit and vegetables matrix are disrupted, as is usually the case during food preparation, processing, and/or mastication, but not during digestion, at least in the human ileum ([@b29]; [@b28]; [@b65]; [@b18]; [@b19]). The extent of release from the fruit and vegetable matrix is highly variable depending on whether carotenoids are noncovalently bound to protein or fiber, dissolved in oil (as in corn, avocado, or palm oil), or in crystalline form (carrots), making their optimal absorption difficult to achieve ([@b10]; [@b66]; [@b67]). The role of fiber in fruit matrices where carotenoids are imbibed plays a critical role in adequate absorption ([@b48]). The interactions of carotenoids and specific components of dietary fiber are not clear. The bioaccessibility of carotenoids is interrupted probably because of micelle formation, necessary for the absorption of lipophilic substances, due the disturbance by viscous polysaccharides. It is suggested that fiber interferes with micelle formation by partitioning bile salts and fat in the gel phase of dietary fiber. In other words, the fiber may entrap the lipids and bile salt molecules, thereby avoiding micelle formation with carotenoids, which may block the passive absorption in the small intestine. Furthermore, dietary fiber increases the viscosity of the intestinal content. This results in reduced absorption of antioxidants because of slowed enzymatic activity in the pancreas and increased difficulty in contacting intestinal enterocytes. All the nonabsorbed carotenoids and dietary fiber along with entrapped lipids and bile salts pass to the large intestine, where the polysaccharides are hydrolyzed by bacterial enzymes and the carotenoids may exert their antioxidant activity in the large intestine environment. ::: {#fig03 .fig} Figure 3 ::: {.caption} ###### General human bioabsorption of carotenoids contained in foods with high contents of dietary fiber. :::  ::: Phenolic compounds appear as quantitatively and qualitatively important constituents of indigestible polysaccharides, such as dietary fiber present in fruits and juices ([@b3]) and other beverages such as beer and wine ([@b56]; [@b13]). Phenolic compounds associated with soluble dietary fiber may present different structures, including soluble flavonoids and phenolic acids. The main phenolic compounds associated with dietary fiber in wine are flavan-3-ols and benzoic acids, ([@b57]), while in beer are flavonoids, followed by hydroxycinnamic acids linked to arabinoxylans ([@b13]). Phenolic compounds therefore appear as important constituents of insoluble dietary fiber, which is mainly due to their ability to chemically interact and form complexes with protein and polysaccharides previously generated in the fruit maturation process or the chyme of the gastrointestinal tract ([@b47]). In the case of beverages such as wine, phenolic compounds contained in the liquid matrices are promptly bioaccessible and in a condition to exert their beneficial effects on the gastrointestinal tract, but this is not the case for phenolic compounds contained in solid matrices such as those in fruits and vegetables. Phenolic compounds must first be extracted to be bioaccessible and then potentially bioavailable ([@b62]), but dietary fiber can interfere with their bioavailability during digestion processes as shown in [Figure 4](#fig04){ref-type="fig"}. The gastrointestinal tract may be considered as an extractor where both the mechanical action during mastication in the mouth and the chemical action during the digestive phase in the stomach and intestine contribute to the extraction of phenolic compounds from solid matrices such as fruits and vegetables ([@b36]). In particular, the mechanical action of mastication mediates the breakdown of fruits cells with the release of the phenolic compounds contained in vacuoles and those linked weakly to the cell wall. The polyphenols linked more closely to the cell wall, especially in the skin cells, are released during the digestive gastro-pancreatic phase as a consequence of the actions of the acidic environment of the stomach and of the alkaline environment of the intestine ([@b62]; [@b8]). Only a minor portion of low molecular weight phenolic compounds (monomers or oligomers) are able to pass through the gut wall into the small intestine. Unabsorbed low molecular weight phenolic compounds and polyphenols associated with dietary fiber, which account for a major part of dietary polyphenols, are not bioavailable in the human upper intestine and reach the colon, where they become fermentable substrates for bacterial microflora along with indigestible carbohydrates and protein ([@b39]). Phenylacetic, phenylpropionic, and phenylbutyric acids, urolithin A, and urolithin B are absorbable metabolites of polyphenol colonic fermentation that may exert systemic effects ([@b51]). Nonabsorbable metabolites and nonfermented phenolic compounds remain in the colonic lumen, where they may contribute to a healthy antioxidant environment by scavenging free radicals and counteracting the effects of dietary pro-oxidants ([@b25]). On the other hand, some polyphenols may be excreted in the feces ([@b4]). ::: {#fig04 .fig} Figure 4 ::: {.caption} ###### General human bioabsorption of phenolic compounds contained in foods rich in dietary fiber, such as fruits and vegetables. :::  ::: Polyphenols bound to dietary fiber need to be hydrolyzed by enzymes in the upper area of the intestine; otherwise, these compounds will not be bioavailable for absorption in the human intestine. Considering that dietary fiber acts as an entrapping matrix and restricts the diffusion of the enzymes to their substrates, most of the polyphenols bound to dietary fiber may end up in the large intestine ([@b47]). From a nutritional point of view, it is apparent that digestible enzymes do not release completely polyphenols associated to dietary fiber; this suggests that this important fraction of polyphenols will not be bioavailable in the gut and only after colonic bacterial fermentation could be absorbed ([@b56]). The type of chemical interactions between phenolic compounds and dietary fiber includes the formation of ordered junctions stabilized by arrays of noncovalent bonds between hydroxide groups from phenolic compounds and polar groups from polysaccharide molecules (hydrogen bonds, electrostatic and dipolar interactions, van der Waals attractions) ([@b17]). The possible interactions and chemical bonds are illustrated in [Figure 5](#fig05){ref-type="fig"}. Because these bonds are individually weak, the interactions are stable only above a minimum critical length, and their formation and disruption often occur as sharp, cooperative processes in response to comparatively small changes in, for example, pH or solvent quality in the gastrointestinal tract (that is, the nature and concentration of dissolved solids in the chyme) ([@b44]). The precisely type of dietary fiber-polyphenols association (electrostatic bonds, hydrogen bonding, van der Waals forces, covalent bonds) remains to be elucidated. ::: {#fig05 .fig} Figure 5 ::: {.caption} ###### \(A) Native flavonoid structure bond to saccharides found in fruits and vegetables. Enzymatic hydrolysis between both molecules during digestion is needed for free absorption of flavonoid in the intestine. (B) Electrostatic and van der Waals interactions between flavonoids (F) and other phenolic compounds with polysaccharide chains (P). :::  ::: Polyphenols bound to dietary fiber need to be hydrolyzed by enzymes in the upper area of the intestine; otherwise, these compounds will not be bioaccessible for absorption in the human intestine but will be susceptible to degradation by the colonic microflora in the large intestine ([@b47]). Considering that dietary fiber acts as an entrapping matrix and restricts the diffusion of the enzymes to their substrates, most of the polyphenols bound to dietary fiber may end up in the large intestine. On the other hand, there is another example of interaction between phenolic compounds and fiber. In bran and aleurone, ferulic acid is mainly bound to arabinoxylans and other cell wall polysaccharides that are able to resist digestion in the upper gastrointestinal tract (Kern and others 2003; [@b68]). The ferulic acid esters might be hydrolyzed to free ferulic acid *in vivo*. Whether these ferulic acid esters will permeate the mucus layer and reach the esterases in the small intestinal mucosa depends on their molecular size and structure ([@b40]). Molecules larger than 30 kDa are not expected to diffuse through the mucus, while diffusion of lower molecular weight molecules is size dependent and influenced by the existence of an electric field ([@b3]). [@b68] showed how differences in the molecular size of ferulic acid sugar esters influences the degree of absorption and the absorption site of ferulic acid within the gut of the rat. When ferulic acid was either esterified to one arabinose or to several arabinoses and xyloses, most of the ferulic acid (60% to 70%) was not absorbed in the small intestine, leading to the conclusion that the major cleavage of ferulic acid esters takes place in the large intestine ([@b68]). This is in accordance with a previous study ([@b34]), which reported that the major cleavage of esters bound in hydroxycinnamates such as ferulic acid occurs in the colon by bacterial enzymes. In summary, the main reasons why phenolic compounds and carotenoids are not bioaccessible due the presence of dietary fiber include the following: (i) they are not well released from fruit and vegetable matrices, (ii) dietary fiber entraps the phenolic compounds during digestion in the upper intestine, and (iii) some antioxidants may be bound to polysaccharides and therefore require enzymatic hydrolysis to be absorbed, which is restricted by the action of dietary fiber matrices formed in the chyme. Finally, all nonabsorbable antioxidants reach the large intestine and remain in the colonic lumen where they may contribute to a healthy antioxidant environment. Conclusion ========== The limited bioavailability of antioxidants present in food from fruit and vegetable matrices is determined by their low bioaccessibility in the small intestine due to the physical and chemical interactions of the antioxidants with the indigestible polysaccharides of cell walls. Even if released during processing and digestion, antioxidants may interact with other food components in the gut by binding to macromolecules such as fiber and forming chemical complexes and colloidal structures that reduce or improve their bioavailability, a subject that needs urgent research. This has remarkable consequences in assessing the nutritional role and real impact of many fruit and vegetable phytochemicals in the prevention and therapy of some chronic human diseases. It is necessary to take serious steps to understand at a higher level the types of interactions and the real repercussions of dietary fiber in the bioabsorption of carotenoids and phenolic compounds in the gastrointestinal tract. Future research --------------- The bioavailability of nutrients and bioactive compounds present in fruits and vegetables is presently an extremely important area of food and nutrition research. However, future research is needed to improve the real contribution of fruits consumption to the well-being of consumers. This may be using different process to improve the releasing of antioxidants, especially polyphenols, from solid matrices. Bioavailability i*n vitro* and *in vivo* studies are needed from different fruit matrix with different fiber composition. The role of fiber as a control-released system of bioactive compounds must be studies deeper. On the other hand, considering the fact that dietary fiber interferes with the adequate absorption of antioxidants, an efficient and economic way to improve the antioxidant bioavailability may be to study of new human diet regimens to ensure the right assimilation of antioxidants from meals. These studies could establish times between meals when the absorption of antioxidants is more effective from beverages rather than from solid food. Also establishing times of fiber flow through the digestive system, to estimate when is more adequate to ingest a rich source of antioxidants such as beverages or antioxidants supplements. This simple information would be of great relevance for consumers. [^1]: Authors are with Coordinación de Tecnología de Alimentos de Origen Vegetal, Centro de Investigaciones en Alimentación y Desarrollo, Hermosillo, Sonora, México. Direct inquiries to author González-Aguilar (E-mail: <gustavo@ciad.mx>)

PubMed Central

2024-06-05T04:04:19.345498

2011-1-01

PMC3052469

INTRODUCTION {#sec1-1} ============ Synovial sarcoma (SS) is a rare soft tissue tumor comprising 5--10% of soft tissue sarcomas and less than 1% of all malignancies.\[[@CIT6][@CIT7][@CIT16][@CIT30]\] SS affects mainly adolescents and young adults with a peak incidence in the third decade. Approximately, 30% of cases occur before the age of 20, and 90% before 50.\[[@CIT7][@CIT30]\] Despite its name, SS does not originate and/or differentiate toward synovium (less than 5% of cases arise from joints or bursa). Although SS can occur in any part of the body, more than 80% of tumors arise in the deep soft tissue of the extremities, especially around the knee. The etiology of SS is unknown but they appear to have an epithelial phenotype. Although it is known for being particularly aggressive, SS often grows slowly, forming a circumscribed, multinodular tumor without a capsule.\[[@CIT6]\] More than 90% of SS show a consistent, balanced reciprocal translocation t(X:18)(p11:q11) presumably relevant in its pathogenesis. This translocation involves the fusion of the SYT gene at 18q11 to either homologous genes SSX1 or SSX2 at Xp11.\[[@CIT6][@CIT9][@CIT14][@CIT25][@CIT30][@CIT33]\] This allows for several histopathological variants, including monophasic, biphasic and poorly differentiated forms.\[[@CIT6][@CIT14][@CIT33]\] Although there is no significant correlation among tumor location, metastases at time of diagnosis, age, sex, or the type of transcript, in patients with localized tumors, SYT-SSX2 fusion transcripts seem to predict significantly longer metastasis-free survival than SYT-SSX1 fusion transcripts.\[[@CIT14]\] Surgical excision with wide, negative margins is the currently recommended treatment with adjuvant radiotherapy and/or doxorubicin-based chemotherapy.\[[@CIT19]\] While this is a mainstay of treatment, there is no consensus on the optimal treatment strategy. Local recurrence occurs in up to 50% of cases, usually within 2 years, although some studies have shown the 5 year local and distant recurrence rates to be 12% and 39%, respectively.\[[@CIT6][@CIT17]\] Lungs and bone are frequent sites of metastases, but regional lymph nodes can also be involved in 20% of cases.\[[@CIT6][@CIT17]\] Patients with favorable prognostic factors (calcifying variants and SSX2 involved fusions) have shown 10 year survival rates of 43--63%.\[[@CIT6]\] Reports of SS arising from, near, or metastatic to the spine are rare and difficult to find in the journals. In this article, we present the experience at our institution with these cases and review the available literature. MATERIALS AND METHODS {#sec1-2} ===================== We searched the patient and surgical pathology databases at our institution using the keywords *synovial sarcoma, spine, paraspinal, cervical, thoracic, or lumbar* and identified three patients diagnosed and treated for spinal SS since 2004. We also searched PubMed at <http://www.ncbi.nlm.nih.gov/pubmed/> using the same keywords and found 14 cases reported in the available literature. RESULTS {#sec1-3} ======= Case 1: Thoracic Dumbbell SS {#sec2-1} ---------------------------- A 59-year-old woman presented with a two-year history of constant, progressively worsening, left-sided upper thoracic pain. In the month prior to presentation, she also experienced ascending paresthesias from her toes to the T5 dermatome along with gait weakness and instability. Imaging studies elsewhere revealed a dumbbell-shaped upper thoracic mass emanating from the T5 foramen with extensive encasement and compression of the thoracic cord from approximately T4 through T6. Additionally, the tumor appeared to extend into the T5 vertebral body. A CT-guided biopsy of the mass was performed and was diagnosed as a "spindle cell tumor consistent with schwannoma." Since her symptoms continued to progress despite treatment with steroids and radiation, she was referred to our institution for further treatment including surgical resection. Following a lengthy discussion, the patient chose to undergo partial resection of her lesion. The tumor was exposed through a T3-T6 laminectomy and an extensive epidural mass was noted dorsally between T4 and T5 and extending out the T4-5 foramen, infiltrating the T4-5 facets and encasing the T4 nerve root. The tumor was grossly debulked, and the final pathology report revealed a high-grade SS \[[Figure 1](#F0001){ref-type="fig"}\]. The patient tolerated surgery well and there were no complications. Since the pathology showed a high-grade sarcoma, adjuvant therapy was planned to reduce her overall tumor burden before a second-staged surgical procedure. After she recovered from her surgery, the patient was started on three cycles of ifosfamide/adriamycin (7500 mg per sq. m/60 mg per sq. m, respectively) chemotherapy. She then returned home and received 46 Gy of total radiation to her tumor volume in addition to the previous 4 Gy received. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Case 1 (A--C) with low power (A) and high power (B) spindle cell appearance and focal epithelial membrane antigen immunoreactivity (C). Case 2 (D--F) illustrating the morphologic appearance of the paraspinal biopsy (D) and leptomeningeal infiltration (E). The RT-PCR (F) detects the chimeric fusion transcripts SYT-SSX2 using specific primers. Case 3 (G--J) Low power appearance (G) of the tumor with gaping vessels (so-called "hemangiopericytomatous vascular pattern") and high power (H). The tumor expresses both epithelial membrane antigen (I) as well as cytokeratin (J) immunoreactivity :::  ::: The adjuvant therapy had stabilized her disease and 7 months following her initial evaluation she underwent a second-staged operation. A one-piece gross total resection of T4, 5, 6, and 7 vertebra with a posterior instrumented spinal fusion from T1-L1 was performed. The tumor was removed with negative margins. The patient recovered well from the operation. She was seen on a 6-month basis by oncology with no recurrence of her tumor noted and improving symptoms. A year and a half later, she returned, complaining of new-onset back pain in her right scapular region and left mid back. This pain was found to be attributed to a fractured rod, which was surgically repaired. She continues to be followed and she has no evidence of tumor recurrence or metastases 67 months after the final resection. Case 2: Paraspinal SS with Leptomeningeal Spread {#sec2-2} ------------------------------------------------ A 54-year-old woman presented with several months of pain in her right buttock and hip, paresthesias of her right leg, and generalized right leg weakness. Her symptoms progressed, involving her left leg in a similar fashion and a sensory level at the costal margin was found bilaterally. Over the last month, she noted saddle numbness and loss of bladder and bowel control. Spine MR showed a large, right-sided, paraspinal mass centered around T10 with adjacent leptomeningeal enhancement---no spinal cord compression was noted \[[Figure 2](#F0002){ref-type="fig"}\]. A CT-guided biopsy was consistent with a spindle cell tumor, but the specific tumor type was indeterminate. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Preoperative T1-sagittal MRI with contrast shows (A) a paraspinal mass extending through the foramen as well as (B) leptomeningeal spread :::  ::: There was some question if the leptomeningeal enhancement was associated with the paraspinal mass or was an unrelated disease process. The location of maximum leptomeningeal enhancement was at the T12-L1 junction. A T12-L1 laminectomy was performed and the underlying thickened arachnoid was biopsied. The leptomeninges demonstrated the presence of a spindle cell tumor similar in morphology to the one seen in the epidural CT-guided biopsy. The tumor cell morphology suggested the diagnosis of monophasic SS, a diagnosis confirmed by immunohistochemical and RT-PCR studies \[[Figure 1](#F0001){ref-type="fig"}\]. The tumor arose from a paraspinal mass and extended through the dura to the leptomeninges, a very unusual presentation for SS. There were no metastatic findings at that time and systemic chemotherapy (ifosfamide and doxorubicin) was recommended. The patient returned home to receive treatment at another institution. We learned through correspondence with her home institution that the patient died 4 months later most likely from a right-midbrain lesion that appeared consistent with a glioma on MR imaging studies. No autopsy was performed. Case 3: T5-6 Paraspinal SS {#sec2-3} -------------------------- This 32-year-old woman presented with 8 years of progressive right-sided subscapular thoracic pain incited by lifting her arm above her head. This pain became more frequent and began to occur spontaneously, disrupting her sleep and causing her significant difficulties. Two MRI studies were performed during this time, the first in 2004 and the second in 2009, and she was told that the findings were unremarkable. When she presented to our institution, she had severe, mechanically reproducible pain located several inches to the right of her mid-thoracic spine. The pain also radiated both up to the axilla as well as down the right leg. A review of outside MRI studies revealed a lobulated, T1 isointense, mildly T2 hyperintense, 3 cm enhancing mass involving the right T5 nerve root through the foramen with extension into the paraspinal muscles \[[Figure 3](#F0003){ref-type="fig"}\]. These imaging characteristics were most consistent with a schwannoma or another nerve sheath tumor since there was little change on interval exams that were 5 years apart. ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Preoperative T1-axial MRI with contrast shows an enhancing paraspinal mass with foraminal extension :::  ::: As it was felt that this symptomatic lesion was likely benign and easily accessible, resection was advised. Intraoperatively, the mass was identified in the erector spinae muscles of the T4-7 region on the right side. The tumor was mobilized and gross total resection achieved, including the foraminal component. However, intraoperative pathology revealed a spindle cell sarcoma rather than a benign process, necessitating extension of the surgical resection. A T5-6 laminectomy was then performed, the dura incised, and the entire nerve root in the foramen of presumed origin was resected until negative dural margins were achieved. After the dural closure, all grossly visible tumor appeared to have been removed from the region and surrounding tissues. Final pathology showed a high-grade, fibrous-type, monophasic SS. The patient tolerated the surgery well and was neurologically intact. Concern for residual tumor remained as postoperative imaging studies demonstrated abnormal T5-6 transverse processes. To achieve a complete resection with negative margins a second surgery was undertaken where a one-piece gross total resection was performed of the entire tumor bed. Both transverse processes were resected and gross residual tumor was found in the T6 transverse process. Wide margins were excised, and the pathologist entered the OR to correlate the negative margins with the visualization of the previous tumor bed. Based on these findings, it appeared that the resection had removed all residual viable tumor present. The patient tolerated the second operation well. At 6-month follow-up, no local tumor recurrence was found, but several small lung nodules were noted on chest CT. These nodules were consistent with metastases and the largest was measured at 7 mm. She was seen in oncology and systemic chemotherapy to shrink the tumors and resection of these nodules has been planned. At 1-year follow-up, her spine continues to be disease free, while her lung nodules have slightly increased in size. Therapy is ongoing at this time to address the lung nodules. DISCUSSION {#sec1-4} ========== SS is a rare, aggressive neoplasm of uncertain origin predominantly affecting adolescents and young adults.\[[@CIT6][@CIT30]\] It is associated with a balanced reciprocal translocation t(X:18)(p11:q11). SS presents in a variety of histopathological forms, from monophasic, being uniformly comprised of spindle cells, to biphasic, with epithelial and spindle cell components.\[[@CIT6][@CIT9][@CIT14][@CIT25][@CIT30][@CIT33]\] Due to histological variety, SS can be mistaken for numerous other mesenchymal or nonmesenchymal tumors, making immunohistochemical and molecular studies important in achieving the correct diagnosis.\[[@CIT13]\] Definitive characteristics making SS completely distinguishable on radiologic examination have not been seen, making diagnosis difficult.\[[@CIT20]\] Plain radiographs may be normal in up to 50% of patients with SS, making it difficult to visualize unless the tumor is eroding adjacent bony structures. However, tumor calcification may be seen in approximately 30% of patients and these calcifications become readily apparent on routine radiography or computerized tomography.\[[@CIT29]\] On CT imaging, SS may appear as a well demarcated, hypodense mass with homogenously or heterogeneously enhancement, making it easy to confuse with other benign or malignant tumors.\[[@CIT20]\] MRI has proven to be the superior modality for detecting SS. In cases localized to the head and neck, it has been shown that these tumors possess a signal intensity similar to fat on T2-weighted images, and isointense as compared to gray matter on T1-weighted images.\[[@CIT11]\] It is widely accepted that the current most effective treatment for SS is a wide surgical excision with negative margins.\[[@CIT6][@CIT13][@CIT19]\] Use of adjuvant radiation therapy decreases the local recurrence rate.\[[@CIT1]\] Radiation proved to be superior to chemotherapy alone as adjuvant therapy to primary surgical excision.\[[@CIT10]\] A single chemotherapy protocol has not yet been proven most effective in treating SS, but two agents, doxorubicin and ifosfamide have demonstrated meaningful activity in the treatment of soft tissue sarcomas.\[[@CIT19]\] Specifically, high dose ifosfamide has been associated with improved disease-specific survival in adult patients with high-risk primary SS and should be considered a standard part of the chemotherapy regimen for this disease.\[[@CIT5][@CIT23]\] Most of the available information on SS has come from tumors localized to the extremities. However, up to 5% of these tumors are encountered in the body axis, including the spine, mediastinum, retroperitoneum, and head and neck region. Reports of SS located within the spinal axis have been rare and are limited to 13 case reports on 14 patients published in the literature \[[Table 1](#T0001){ref-type="table"}\]. Of these 14 patients with spinal SS, 8 were within the paraspinal musculature (with most having foraminal extension), 3 were intradural and associated with spinal nerve roots, 2 were metastatic lesions (one to bone, the other intramedullary), and 1 was a bony/lytic lesion. The cases we present here all involved the paraspinal muscles with two of them being associated with spinal nerve roots, the other having significant leptomeningeal extension, this finding not previously reported in the literature. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Review of spinal SS at our institution and reported cases in the literature by year ::: Our institution Age Imaging Treatment Outcome ---------------------------------------------------- --------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Case 1: Thoracic dumbbell**SS** 59 F Imaging studies revealed a dumbbell-shaped upper thoracic mass emanating from the T5 foramen with extensive encasement and compression of the thoracic cord from approximately T4 through T6. Additionally, the tumor appeared to extend into the T5 vertebral body. Two stage operation performed. First, a T3-5 laminectomy with debulking of tumor was performed. Three cycles of ifosfamide/ adriamycin (7500 mg per sq. m/60 mg per sq. m, respectively) chemotherapy performed. The patient received 46Gy of total radiation to tumor volume. The second operation consisted of en bloc resection of T4, 5, 6 and 7 vertebra with a posterior instrumented spinal fusion from T1-L1. The tumor was removed to negative margins. Patient continues to be followed and has no evidence of tumor recurrence or metastases at 67 months from the final resection. Case 2: Paraspinal**SS**with leptomeningeal spread 54 F Spine MRI showed a large, right-sided, paraspinal mass centered around T10 with adjacent leptomeningeal enhancement---no spinal cord compression was noted. CT-guided biopsy findings were consistent with a spindle cell tumor. The tumor cell morphology suggested the diagnosis of monophasic**SS,**a diagnosis confirmed by immunohistochemical and RT-PCR studies. The tumor was arising as a paraspinal mass and extending through the dura to the leptomeninges. There were no metastatic findings at that time and systemic chemotherapy (Ifosfamide and Doxorubicin) was recommended. The patient returned home to receive care at another institution. The patient died 4 months later. Case 3: T5-6 paraspinal**SS** 32 F MRI studies revealed a lobulated, T1 isointense, mildly T2 hyperintense, enhancing mass involving the right T5 nerve root through the foramen with extension into the paraspinal muscles. The mass measured 3.1 × 2.0 × 2.5 cm^3^ and there was no extension into the spinal canal. Intraoperatively, the mass was identified in the erector spinae muscles of the T4-7 region on the right side. The tumor was mobilized and removed. A T5-6 laminectomy was performed, the dura was incised and the entire nerve root in the foramen of presumed origin was resected including the dural attachments until the dural margins were negative. A second operation was performed to resect gross residual tumor found in the T6 transverse process. At six-month follow-up, no tumor recurrence was found, but several small lung nodules were noted on chest CT. These nodules were consistent with metastases and the largest was measured at 7 mm. Systemic chemotherapy to shrink the tumors and resection of these nodules was been planned after her most recent visit. **Literature cases by year** **Age** **Imaging** **Treatment** **Outcome** Arnold*et al.,*2010\[[@CIT2]\] 26 F CT showed destruction of odontoid and C2 body. MRI revealed tumor in posterior C2, ventral epidural space from C2-5 with narrowing of spinal canal. Widely metastatic disease confirmed. Spinal cord decompression with C2-3 laminectomy and posterior C2 corpectomy. Occipital-C7 fusion. Scheduled for palliative chemotherapy. Developed fever, leukocytosis and acidosis, leading to sepsis. Multiple metastases to liver and abdomen. Patient died 6 months later Barus*et al.,*2009\[[@CIT3]\] 14 F MRI showed a large lesion in lumbar paraspinal muscles from L2 to sacrum with extension into spinal canal and L3-4 thecal sac. Initial biopsy showed SS. Marginal resection to preserve neural elements performed. A 310 g lobulated mass removed with fibrous pseudocapsule. Two cycles of ifosfamide/ doxorubicin chemotherapy and total dose of 5940 cGy administered followed by four cycles of chemotherapy Patient developed chronic kidney disease. Almost 6 years postop., patient is free of local recurrence or metastatic disease. Koehler*et al.,*2009\[[@CIT15]\] 60 M MRI showed large right-sided paraspinal mass from T7-9. Calcifications and enhanced signal intensity of the central tumor segment. Right-sided thoracotomy and biopsy of tumor mass performed. Upon malignant confirmation, wide resection with negative margins performed. Vascular and neuro supply ligated and resected. Radiation therapy given. 9 month postoperative imaging showed no recurrence. Ravnik*et al.,*2009\[[@CIT21]\] 32 M Myelography showed a T12/L1 Contrast Block. CT/MRI showed intraspinal epidural T12-L2 tumor. Immediate surgical decompression. Three level laminectomy with epidural mass removal. Second debulking surgery. Six cycles (ifosfamide/doxorubicin) and 50.4 Gy of radiation. Local recurrence after 12 months. Further treatment refused. **Literature cases by year** **Age** **Imaging** **Treatment** **Outcome** Scollato*et al.,*2008\[[@CIT27]\] 59 M MRI showed intramedullary lesion at C3-5. Right pulmonary mass seen. Surgery performed after chemotherapy was refused by patient. Longitudinal myelotomy performed. Patient experienced postoperative pain relief but no neurological change. Patient died of lung and hepatic metastases 3 months later. de Ribaupierre*et al.,*2007\[[@CIT22]\] 11 F MRI showed intradural, heterogeneous mass in right C6-7 foramen. Right C6-7 foraminotomy with complete resection. Nerve roots found to be involved and were resected. Postop. chemotherapy administered (vincristine, actinomycine, and ifosfamide) as well as local radiotherapy. Disease free for 16 months then local reoccurrence found. Another operation performed and further chemotherapy/radiation given. Patient died 6 years after diagnosis. Greene*et al.,*2006\[[@CIT8]\] 11 F MRI showed a large intradural, extramedullary lesion at L2-4, filling canal and displacing nerve roots laterally. Multiple other nodules at C6, T2, T5, T8 and L1 levels noted Large intradural, extramedullary lesion from L2-4 seen. No dural involvement. Near total resection, small capsule remnant. Four courses of doxorubicin/ifosfamide delivered. Radiation given, 45 Gy to total spine, 54 Gy boost to lumbosacral region. Five months after diagnosis, MRI showed four intracranial metastases. Intracranial resection and radiation performed. Patient died 14 months later. Sakellaridis*et al.,*2006\[[@CIT24]\] 36 F MRI showed recurrent tumor of lumbar spine attached to dura at L2-3. Previous diagnosis of hemangiopericytoma. Extended laminectomy L1-3 with resection of tumor mass. Dura was opened and subarachnoid invasion was noted. 3500 rads given. 2 years postop. new C7 lesion and mediastinal tumor from T5-10. Third operation performed. Patient died 1.5 year later from metastases. Suh*et al.,*2005\[[@CIT29]\] 44 M MRI demonstrated large paravertebral and epidural mass displacing thecal sac at L4-5. No distant metastases seen. Hemilaminectomy/facetectomy at L4-5 performed. Mass was found to extend through right L4 foramen. Near-total resection performed. Patient refused chemotherapy. Radiation delivered to residual mass. At time of report, patient's symptoms showed improvement. No long-term follow-up. Morrison*et al.,*2001\[[@CIT18]\] 53 F MRI showed a large paraspinal mass from C7-T3 without dural or foraminal involvement. No metastases seen. Needle biopsy showed a spindle cell tumor. Complete surgical resection achieved. No outcome or follow-up reported. Wu*et al.,*2000\[[@CIT32]\] 30 M MRI showed an enhancing mass in the right paraspinal muscles from T12-L1 with extension into the spinal canal and displacement of the thecal sac. CT showed bony erosion of the pedicles, transverse process and posterior elements of the affected levels Needle biopsy showed a myxoid mesenchymal tumor. Immunohistochemistry confirmed SS with an open biopsy. No resection was reported No outcome or follow-up reported. Signorini*et al.,*1986\[[@CIT28]\] 59 M CT showed a lytic lesion involving the T2 vertebral body with right pedicle erosion. A myelogram showed a complete block at that level. Upfront radiation---48 Gy. 6 months later developed acute paraplegia. Underwent right transpleural subtotal resection of the T2 vertebral body. Pathology showed a biphasic SS Died 3 months later from disease progression. Treu*et al.,*1986\[[@CIT31]\] 21 M Patient 1 had a CT which showed a mass eroding the posterior arch of C1 with some dural compression. Patient 1 had a subtotal resection through a posterior approach. A monophasic SS was identified. Patient 1: 25 months later developed metastates to pancreas---no follow-up thereafter reported. 18 M Patient 2 had a CT scan which showed a lumbar paraspinal mass from L4-L5. Patient 2 had a gross total resection where the soft tissue component, transverse process of L4 and 5 and part of the psoas muscle was resected. Pathology confirmed a monophasic SS. Patient 2: No outcome or follow-up reported. ::: Clinical differential diagnosis for SS involving the spine includes primarily nerve sheath tumors, and most SS are assumed to be benign nerve sheath tumors preoperatively. Indeed, in two of our cases a simpler/smaller operation was performed first because the lesions were thought to be benign, however, when the final pathology identified SS, a larger more radical procedure was performed. With this limited data, the long-term outcomes are hard to quantitate for SS involving the spine. Based on the previous case reports, most patients died within 3 years following diagnosis, with the exception of a 14-year-old girl who remained disease-free at 6 years.\[[@CIT3]\] Over half of the reported spinal SS had dumbbell-shaped intraforaminal extension with a larger extraspinal and smaller intraspinal component. In most cases, the patients' symptoms were caused by the intraspinal component and resultant compression of neural elements. It was striking however, how large the extraspinal portion was in many of the cases. The most common tumor presenting in this fashion is a benign schwannoma, whereas, malignant nerve sheath tumors are very rare. In regards to SS, it is the rapid growth of the extraspinal portion that is most suggestive of a malignant process. However, in one of our cases, the patient had serial MRI scans 5 years apart and there was minimal growth during this time, which is most unusual for a sarcoma. Some of these unique properties of SS are just now being understood at a molecular level. It has been recently reported that microRNAs (miRNAs) may play an expanded role in the tumorigenesis of some cancers, and when deregulated, depending on their mRNA targets, they can act to down regulate tumor suppressor genes and give a growth advantage to tumor cell lines.\[[@CIT4]\] In SS, a unique pattern of deregulated miRNAs has been found that is distinct from muscle tissue and a wide range of other sarcoma types.\[[@CIT12]\] Specifically, the overexpression of a microRNA, miR-183, has been found to act as an oncogene through down regulation of EGR1 translation, a tumor suppressor that is correlated strongly with tumor formation and transformation processes when its levels are depleted.\[[@CIT26]\] Another overexpressed miRNA, let-7e, has been shown to down regulate expression of HMGA2, a transcription factor which works in concert with the SS18-SSX fusion product to decrease levels of SNAI1, a transcriptional repressor, ultimately causing epithelial differentiation and transition in SS.\[[@CIT12]\] Importantly, when let-7e was inhibited by a miRNA inhibitor, the proliferation of the SS cells were suppressed. While these findings are very new, they provide a better understanding of the underlying molecular basis for SS tumorigenesis and also present possible future targets for pharmacological therapies. Experts agree that the cornerstone of treatment for SS is wide surgical excisions with negative margins. Based on our experience, if a spinal dumbbell tumor has any characteristic that would be unusual for a benign nerve sheath tumor, a needle-guided biopsy should be performed first to obtain a diagnosis. Ideally, this is followed by a definitive operation to try and remove the tumor en bloc with negative margins. Traditionally, a 5 cm margin defines a negative margin in sarcoma surgery, however, in many of these cases this would encompass critical structures such as the spinal cord. Thus, in many cases, the best that can be hoped for is a gross total resection with "marginal margins" in which there is no pathologic specimen, but critical structures are not injured. CONCLUSION {#sec1-5} ========== SS of the spine can be challenging to diagnose and even harder to treat. Best available evidence suggests that a multimodal treatment strategy starting with aggressive surgical resection followed by radiation and chemotherapy offers the patients the best chance for a cure. The underlying molecular genetics for SS tumorigenesis is just now being elucidated and hopefully will lead to a better understanding in regards to tumor pathology and lead to novel therapeutics in the future. Available FREE in open access from: <http://www.surgicalneurologyint.com/text.asp?2011/2/1/18/76939>

PubMed Central

2024-06-05T04:04:19.349118

2011-2-21

PMC3052509

INTRODUCTION ============ Naturally acquired human brucellosis remains common worldwide. However, the incidence in the United States and much of Europe is now very low, so brucellosis probably would not be suspected as the cause of a febrile illness in a US or European citizen with no travel history ([@r17]). Additionally, definitive diagnosis of brucellosis has historically proven to be time-consuming and uncertain. Blood culture may be unrewarding, as viable bacteria are often undetectable in blood, and even when present, growth is slow, with blood cultures requiring up to 3 weeks of incubation before *Brucella* is detectable ([@r15]). Cultures on solid medium must be maintained for 3 or more days before growth is detected ([@r7]). These factors may combine to reduce the probability of timely diagnosis of a point-source outbreak. When detected in the acute stage, brucellosis is much more treatable. Therefore, rapid, non-invasive diagnostic methods can be of great benefit to provide optimal opportunity for appropriate treatment and to facilitate rapid investigation in case of intentional exposure ([@r8]). PCR assays have been designed that are specific for the *Brucella* genus. Speciation by PCR is possible, but it is not essential for initial diagnostics, especially for outbreak detection. We chose to model *B. suis* for a variety of reasons that are discussed later. Naturally acquired brucellosis usually involves a low-exposure dose. In such cases, disease progression is very slow, and an association with a point source of exposure would not be expected. We did not attempt to model natural brucellosis, but instead endeavoured to model high-dose, aerosol exposure such as might occur in a bioterrorism incident. Clearly, in such a case, there would be a continuum of actual inhaled dose. We focused on the high end of this spectrum, i.e. those that would be expected to present with clinical signs soon enough for a group of illnesses to be considered a single incident. METHODS ======= *Brucella* strain and culture. ------------------------------ *Brucella suis* 1330 was cultured on tryptose blood agar base slant tubes for 48 h. The slant tube was washed with 1 ml *Brucella* broth and the wash was added to a flask containing 200 ml *Brucella* broth and incubated for an additional 48 h. Animals, aerosol exposure, sampling, necropsy and tissue collection. -------------------------------------------------------------------- Twenty-four adult rhesus macaques (*Macaca mulatta*) were surgically implanted with Data Sciences International TA-D70 temperature and activity telemetry transmitters. Twelve macaques were assigned to this study and 12 were 'historical controls' from a previous study in which all handling, conditions and time points were identical. Experimental animals were exposed to *Brucella* organisms diluted in normal saline solution whereas historical controls were exposed to saline solution only. Control animals were necessary to serve as a baseline for haematology, blood chemistry and telemetry (body temperature and activity) as well as PCR, i.e. to ensure that the assay did not result in false-positives or -negatives when used on a given tissue type. Temperature and activity data were sampled every 15 min from 1 week before exposure until time of euthanasia. Complete blood counts (CBCs) and blood chemistries were performed 7, 14 and 30 days before exposure. Swab samples were taken from the face, conjunctiva, buccal mucosa, pharynx, nares and external auditory meatus and broncho-alveolar lavage (BAL) was performed immediately before exposure in order to obtain animal-matched negative control samples for PCR analysis. Extensive experience at the US Army Medical Research Institute of Infectious Diseases indicates that BAL does not affect susceptibility or disease progression in aerosol-exposed animals. Experimental animals were exposed to approximately 1 μm mass median aerodynamic diameter aerosolized particles of *B. suis* in a manner that standardizes the number of c.f.u. inhaled. Animals were anaesthetized in accordance with institute policy. Respiratory minute volumes were estimated immediately prior to exposure using head-out plethysmography (Buxco Research Systems). Respiratory minute volume was assumed to be constant over the exposure period. Each animal was exposed separately in a well-characterized dynamic airflow exposure chamber ([@r5]). Small particle aerosols were generated using 10 ml agent diluted in normal saline in a 3-jet Collison nebulizer (BGI). The generated aerosol was sampled using all-glass impingers attached to the exposure chamber. The contents of the impinger were assayed post-exposure to estimate the mean agent concentration in the chamber during the exposure. The estimated inhaled dose was calculated as the product of the chamber aerosol concentration, the respiratory minute volume and exposure duration. In order to achieve a targeted dose, the exposure duration was varied from animal to animal since both the chamber aerosol concentration and estimated respiratory minute volume are assumed to be constant throughout the exposure period. Exposure durations ranged from 5 to 15 min. The aerosol respiratory deposition fraction was assumed to be 100 %. Following aerosol exposure, the head of each monkey was wiped with a soap solution to remove deposited aerosol, and monkeys were housed individually under BSL-3 conditions. Animal care was provided in accordance with established guidelines ([@r4]). Three monkeys from each group were euthanized on days 1, 3, 5 and 7 post-exposure. On the day of euthanasia, CBC and blood chemistries, swab sampling (face, conjunctiva, buccal mucosa, pharynx, nares and external auditory meatus) and BAL were performed. During necropsy, urine was sampled by needle puncture of the exposed urinary bladder, specific organs and tissues were examined grossly, and representative samples were collected for histological evaluation. Tissue samples collected for histological evaluation were from the mandibular lymph node, liver (right caudal lobe), kidney (left), spleen, heart (ventricle), lung (right cranioventral lobe), hilar (tracheobronchial) lymph node, mesenteric lymph node, epididymis/ovarian tube, testis/ovary, prostate/uterus, ileocaecal junction, large intestine, bone marrow and brain (cerebrum). The following testing was conducted in order to assess tissue distribution of *Brucella* and ability to detect the organism or its DNA in various samples in order to improve diagnostic methods. Bacterial culture. ------------------ For bacterial culture, tissues were ground with a manual tissue grinder and diluted to a 10 % concentration in PBS. Swabs were immersed in 1 ml PBS. EDTA anti-coagulated blood, serum, BAL wash fluid and urine were used without further manipulation. All samples were serially diluted up to 10-fold. Serial dilutions were plated on *Brucella* agar and incubated in 5 % CO~2~ at 37 °C in a humid incubator for 3 days. Samples were plated in triplicate and the three plate reads were averaged to generate the reported value. Any colony with the phenotypic characteristics of *Brucella* was counted as a *Brucella* organism. Histopathology. --------------- Tissue samples were immersion-fixed in 10 % neutral-buffered formalin and prepared for histopathology. Sections were embedded in paraffin, sectioned, and cut at 5--6 μm, mounted on glass slides, and then stained with haematoxylin and eosin in preparation for examination by light microscopy. PCR. ---- Swab diluents, BAL wash fluid, 10 % suspensions of ground tissue samples, blood, serum and urine were extracted using the Qiagen DNA blood kit according to the manufacturer\'s instructions. PCR was performed on a Roche LightCycler 1.5 real-time PCR instrument as previously described using primers and probes for the *Brucella* *omp2A* gene ([@r3]). The forward and reverse primer and minor groove binder probe sequence for this assay are CCAggCgTACCggTTATCTC, AgACCCTTTTgAggTCTACTCCCTTA and TggTCgAAggCgCTC, respectively. The limit of detection of this assay is approximately 30 genome copies. Extracted DNA from *Brucella melitensis* strain 16M was used as a sample positive control. Samples were run in triplicate. Statistics. ----------- Paired *t*-tests for CBC and chemistry laboratory test values at day of bleed were compared to baseline values for each day with stepdown Sidak adjustments for multiple comparisons. Three pre-bleed values for chemistry and CBC tests were averaged to obtain the baseline value for each subject. All tissue bacterial load values were log~10~ transformed for analysis. All temperature and activity data were log~10~ transformed for analysis. After transformation, variables were better fitted to the assumptions required for time series analysis. Temperatures and activity levels from the first 72 h were taken to be baseline values and were compared to temperatures or activity levels from the 72 h immediately before euthanasia. Telemetry of subjects euthanized on a given day (1, 3, 5 and 7) was analysed separately. Data for subjects euthanized on day 1 could not be made to fit the model due to insufficient number of time points immediately before euthanasia and were not analysed in this manner as a result. A time series model for temperature and activity was developed to examine differences between baseline temperatures or activity and temperatures or activity 72 h before euthanasia. The baseline series was used to identify an Auto-Regressive Integrated Moving Average (ARIMA) model. ARIMA is necessary to compare telemetry values obtained with a short lag between measurements, because subsequent values are influenced by previous values; regression or similar models cannot be used because independence of errors cannot be assumed. Because telemetry data were obtained every 15 min, differences were calculated at lags 1 and 96 in order to convert the raw non-stationary data to a stationary form that allows for comparison. For direct assessment of fever spikes, fever was defined as a repeated measurement of a body temperature greater than or equal to 39.5 °C, in accordance with the institute animal care standard. RESULTS AND DISCUSSION ====================== Exposure dose ------------- Actual inhaled dosages of *B. suis* averaged 5.60×10^8^ c.f.u. (standard error of the mean: 1.84×10^7^; range: 4.90×10^8^--6.48×10^8^). Demonstration of *Brucella* organisms or DNA in various samples --------------------------------------------------------------- Nasal and/or pharyngeal swabs are common samples used to diagnose uncomplicated viral and bacterial aetiologies of febrile respiratory illnesses, whereas BAL may be ordered in cases with obvious pneumonia. Respiratory illness is a less common manifestation of naturally acquired brucellosis in humans; however, in light of the clear signs of bronchiolitis in the rhesus model under these conditions (see below), human cases occurring as a result of an intentional aerosol exposure may present with respiratory signs in addition to fever and a clinician might take nasal and pharyngeal swabs and, more rarely, BAL, without considering brucellosis to be a differential diagnosis. *Brucella* organisms were detected by routine culture (see Methods) and *Brucella* DNA was detected by real-time PCR of non-invasive diagnostic samples (pharyngeal and nasal swabs) and in BAL immediately after exposure and for 7 days after exposure. This is a novel finding and it is key in that, although this was a high-dose aerosol exposure not likely to be replicated in nature, obtaining both positive cultures and PCR detection from such samples argues for including PCR for *Brucella* in the diagnostic algorithm when an apparent point-source outbreak of respiratory illness occurs and cannot otherwise be diagnosed. *Brucella* was never isolated from or detected by PCR of blood, serum or urine. Failure to detect *Brucella* in blood components or urine correlates well with human clinical diagnostics experience. Culture and PCR showed a progression of the pathogen from lung to spleen and liver, and finally to the bone marrow over the 7 day course of the experiment (Tables [1](#t1){ref-type="table"} and [2](#t2){ref-type="table"}). In human brucellosis, it is recognized that *Brucella* quickly disseminates to the lymphatics as well. Notably, PCR was positive in the gonads by day 5 in all cases, in both sexes (though, by chance, the animals remaining in the experiment by day 5 included only one male animal). This correlates well with the propensity for *B. suis* to localize to the testes in humans. PCR was not well correlated with gross or histopathology findings in the mesenteric lymph nodes, but did accord well with tracheo-bronchial lymph node pathology and became positive in a progressively larger number of mandibular lymph nodes throughout the 7 day course of the study. There was also frequent detection by PCR in the large intestine and kidney later in the study, perhaps indicating wide-ranging tissue dissemination. It quickly became impossible to detect *Brucella* DNA in buccal swabs, indicating that it does not remain in the mouth. Surprisingly, PCR was also frequently positive in face, conjunctival and aural swabs throughout the 7 day course of the study; however, these would not be likely diagnostic samples in a human outbreak situation, and, at least in the case of face and aural swabs, this probably did not indicate the presence of viable organisms, but rather may indicate only the presence of pathogen DNA in the oily or waxy residues of the face or ear canal. These samples were not cultured. PCR and culture results matched in all cases for lung, liver and spleen. *Brucella* was cultured from one BAL sample on day 7, and from two pharyngeal swab samples on each of days 5 and 7, while DNA was not detected by PCR. *Brucella* DNA was detected by PCR in bone marrow in one case on day 5, in the pharyngeal swab in one case on day 1 and in the nasal swab in one case on day 3 and two cases each on days 5 and 7, in which culture was not positive. All diagnosis mis-matches were in cases where relatively lower amounts of bacteria were recovered or the PCR threshold crossing value was relatively higher, indicating lower levels of *Brucella* or *Brucella* DNA at the lower limit of detection, wherein both methods become unreliable. Overall, these results indicate that a battery of tests on a maximum number of sample types should be conducted to optimize the chances of detecting *Brucella*. However, when *Brucella* is not suspected in an outbreak situation, yet no other diagnostic methodologies yield a diagnosis, screening for *Brucella* by PCR may prove rewarding. Based on these results, neither culture nor PCR can be considered ideal for diagnosis of individual cases; however, the discrepancies in no way reduce the validity of including PCR in the diagnostic regimen for diagnosing an outbreak. It is never normal to either detect *Brucella* DNA or recover it by culture in a human being. Therefore, any positive result would be cause for further investigation. PCR is the ideal method of the two for screening because it is vastly faster and largely obviates biosafety problems that result from *Brucella* culture. Telemetry data -------------- The intervention parameter, which distinguished baseline temperatures from temperatures immediately prior to euthanasia, was not statistically significant for those subjects euthanized on day 3 (*P*=0.9861) but was significant for those subjects euthanized on days 5 (*P*=0.0270) and 7 (*P*=0.0037). Fever spikes were detected in single monkeys on days 1 and 2 and in three monkeys on day 3 (Fig. [1](#f1){ref-type="fig"}). Fever spikes were not noted on any other days. Fever was never detected in control monkeys. Although the rhesus model does not allow assessment of fever-related manifestations of brucellosis that occur in humans (e.g. night sweats), waxing and waning fever is consistent with human brucellosis and indicates that the rhesus macaque is a good model for human brucellosis. No change in activity levels was detected. No significant CBC or blood chemistry aberrations were detected. This is also consistent with human brucellosis because clinical pathology indicators are indistinct in human brucellosis ([@r6]). Gross pathology --------------- In terms of gross pathology, from day 1 post-exposure through day 7, there were varying degrees of enlargement of the mesenteric, tracheobronchiolar and mandibular lymph nodes. In all cases there was mild oedema. There were no significant gross findings in the remaining organs. On day 1 post-exposure, the most significant gross findings were enlarged and slightly oedematous mesenteric lymph nodes (two to three times normal). By day 3, mesenteric lymph nodes were enlarged by up to five times normal and remained mildly oedematous. At that time point, the tracheobronchial lymph nodes were enlarged up to three times normal and the mandibular lymph nodes were also slightly enlarged in two of the three monkeys sampled on that day. On day 5, the mesenteric lymph nodes were still oedematous and enlarged, but only up to three times normal. The tracheobronchial lymph nodes were enlarged three to four times normal and congested. The mandibular lymph nodes were slightly enlarged. By day 7, the mesenteric lymph nodes were enlarged two to three times normal and oedematous. The tracheobronchial lymph nodes were enlarged up to four times normal and oedematous. The mandibular lymph nodes were still slightly enlarged. Enlarged lymph nodes were occasionally noted in the controls; however, oedema was not noted. Lung congestion was noted in two of three exposed monkeys euthanized on days 1 and 3 after exposure, and in one of three monkeys euthanized on days 5 and 7 after exposure. The mesenteric lymph nodes were the first to show any significant macroscopic changes. By day 3, they were their largest over the span of 7 days. The tracheobronchial lymph nodes did not show any significant changes until day 3 and then reached the maximum noted size by day 7. These findings were non-specific, and are consistent with varying degrees of antigenic stimulation that corresponds to the lymphoid hyperplasia noted histologically (see below). There was no corresponding pattern of lymph node enlargement in the controls as was noted in the exposed animals. There was more oedema of the lymph nodes in the exposed animals than in the controls. Oedema is often an early (acute) manifestation of inflammation and may occur prior to any other grossly detectable indication of inflammation ([@r10]). The lung congestion noted was likely a terminal event related to the method of euthanasia. All gross and histological findings were non-specific and would not be attributable to any specific agent this early in the course of the disease. Histology --------- In terms of histological findings, the lymphoid hyperplasia and lymph node oedema corresponded with the gross findings of lymph node enlargement and oedema. There were no other significant histological findings specifically attributable to exposure to aerosolized *B. suis.* On day 3, there was multifocal bronchiolar epithelial degeneration and necrosis, with mild lymphohistiocytic and neutrophilic bronchiolitis and peribronchiolitis with alveolar oedema. The inflammation worsened by day 5 and the oedema was more pronounced. By day 7, the inflammation was similar to that observed on day 5 but the perivascular oedema was slightly more pronounced. By day 7, two out of three rhesus macaques showed mild lymphohistiocytic hepatitis. All tissues were stained with haematoxylin and eosin only and there was no attempt to visualize *Brucella* organisms in tissues. This corresponds with the approach that would be expected to be taken in diagnosing a biopsy specimen from an unknown case of acute febrile illness. There is virtually no information in the literature concerning histological findings in the lymph nodes of human patients. Occasionally, brucellosis has manifested as hepatitis, so there is a slight possibility that a liver biopsy might be a primary diagnostic sample for brucellosis, but there is almost no chance that a biopsy would be taken early in the course of infection because hepatic enlargement or changes notable on ultrasound are not reported to occur early in the course of exposure ([@r1]). Based on the limited information available, our data are consistent with reports of liver histology in human brucellosis, i.e. that parenchymal necrosis and lymphocytic infiltration are common. Histological findings in pulmonary brucellosis cases vary, and the vast majority of information available in the literature is associated with chronic cases, but our data are consistent with the limited amount available, i.e. that inflammation is often lymphohistiocytic and/or neutrophilic ([@r13]). We did not observe progression to granulomatous inflammation, presumably because of the short duration of the study. However, in fact, the histology findings in our study are meaningful for their indistinctiveness; that is, in contrast to later stage *B. suis* infection, in this model of acute *B. suis* infection, no major histological abnormalities were observed. Therefore, alternative diagnostic methods such as PCR would be expected to be preferable due to their higher speed and throughput. In other words, when faced with an unknown, the diagnostician may use PCR as a screening method. Conclusions ----------- Taken together, these data indicate that the rhesus macaque is a good model of human brucellosis. If this is the case, then an intentional human exposure would not be associated with any distinctive clinical, haematological or pathological signs. There would be little basis to suspect any aetiology, let alone brucellosis. For this reason, a rapid, reliable screening test is essential. Our data indicate that an intentional human exposure by aerosol can be detected by PCR of non-invasive samples, i.e. nasal and/or pharyngeal swabs. We detected *Brucella* DNA in nasal swab samples in all three monkeys sampled at each time point (days 1, 3, 5 and 7 after exposure -- with progressively declining apparent concentration based on the mean PCR threshold crossing points) and in pharyngeal swabs in two cases on day 1, three cases on day 3 and one case on day 5. DNA concentrations were not standardized; all methods were conducted in a manner similar to that which would be expected in a clinical diagnostic laboratory, i.e. qualitative. In an intentional exposure scenario, it is reasonable to expect a wide range of exposure doses, such that a subset of individuals would inhale a high dose; the far end of this spectrum was modelled in this study. Heavily exposed humans would probably experience upper respiratory illness and potentially enlarged regional lymph nodes. Physicians commonly attempt diagnosis of such cases with nasal and/or pharyngeal swabs, which are submitted for culture. Culture could also be rewarding in high-dose exposure cases, but PCR would provide a diagnosis within hours whereas culture would require days. The PCR assay used here is highly specific for the *Brucella* genus and *Brucella* is never normal flora for humans, so a PCR-positive sample would provide a high degree of confidence that a diagnosis had been made. The primary importance of these data is to show that adding PCR assays for selected agents can result in a diagnosis in an otherwise confusing outbreak situation, in a timely enough fashion that successful treatment and attribution may be possible. It would be reasonable in this scenario to empirically treat symptomatic individuals that may also have been exposed, but for which no definitive diagnosis could be obtained. Diagnosis of individuals exposed to a lower dose may still require invasive methods such as lymph node or bone marrow biopsy. In this study, neither culturable organism nor bacterial DNA could be found in blood, urine or any other non-invasive sample except certain swabs. In contrast to the study conducted by [@r12], this study shows for the first time that high-dose *Brucella* exposure of rhesus macaques can result in rapid illness and early dissemination to the entire haematopoietic system, liver and gonads. The diagnostic methods are expected to be applicable to other pathogenic *Brucella* species. *B. melitensis*, *Brucella abortus* and *B. suis* cause the vast majority of human disease. *Brucella canis* has been associated with human disease only in immunosuppressed patients, and marine mammal *Brucella* species have been associated with isolated cases of human disease but speciation continues to be debated ([@r2]; [@r11]; [@r14]). *B. abortus* was once common in the United States and caused hundreds of cases of human brucellosis annually, before being largely eradicated toward the end of the last century. *B. melitensis* is generally regarded as the most important species because it is associated with most severe human cases; however, this is due in part simply to its greater prevalence in key animal reservoirs. *B. suis* also causes severe disease in humans, but *B. melitensis* causes more obvious, 'classical' undulant fever while *B. suis*-associated disease is often associated with abscess formation ([@r16]). Laboratory-acquired infection is a common cause of brucellosis and historical evidence of laboratory exposure and infection indicates that the infectious dose of *Brucella* is extremely low. *B. suis* was weaponized by the US, former Soviet Union and China (Hoover & Friedlander, 1997) and this was an important factor in our decision to model *B. suis* infection as opposed to any other species of *Brucella*. A further reason for choosing to perform these studies with *B. suis* as opposed to *B. melitensis* or *B. abortus* is that the main surface antigen of *B. abortus* (so-called 'A' antigen) and that of *B. melitensis* (so-called 'M' antigen) are both present in *B. suis*. Therefore, *B. suis*-infected tissues archived during the course of this study can serve as a test set for future diagnostic assays specifically designed to detect either antigen. - BAL, broncho-alveolar lavage - CBC, complete blood count The opinions, interpretations, conclusions and recommendations are those of the authors and are not necessarily endorsed by the US Army. Work at USAMRIID was funded by DOD grant (Project\# 02-4-KK-032 8.10030\_07\_RD\_B). Research was conducted in compliance with the Animal Welfare Act and other federal statutes and regulations relating to animals and experiments involving animals and adheres to principles stated in the Guide for the Care and Use of Laboratory Animals, National Research Council, 1996. The facility where this research was conducted is fully accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International. We thank Dave Dyer for extensive support for aerosol exposure and telemetry and Diana Fisher for statistical support. ::: {#f1 .fig} Fig. 1. ::: {.caption} ###### Temperature trends in four monkeys that showed fever spikes (black, red, yellow and green lines). Lines terminate on days on which animals were euthanized according to the protocol schedule. :::  ::: ::: {#t1 .table-wrap} Table 1. ::: {.caption} ###### Log~10~ c.f.u. of bacteria (g tissue)^−1^ or (ml fluid)^−1^ detected from *B. suis*-infected rhesus macaques Numbers in parentheses denote the number of animals out of a total of three per time point in which *B. suis* was detected, if less than three. Blood and serum were also cultured but no organism was detected from these fluids. Data are reported as mean±standard error of the mean. ::: **Tissue/sample type** **Day** ------------------------- ----------- --------------- --------------- --------------- Lung 3.16±0.86 5.19±0.79 4.66±0.57 5.31±0.49 Liver 0.75±0.16 (2) 1.60±0.15 Spleen 1.85±0.59 (2) 3.32±0.41 Bone marrow 0.98±0.26 Pharyngeal swab 3.91±0.43 4.14±0.35 3.66±0.17 2.84±0.15 (2) Nasal swab 2.27±0.16 3.93±0.43 (2) 1.85 (1) 5.04 (1) Broncho-alveolar lavage 5.70±1.84 2.50±0.26 2.25±0.40 1.31±0.01 ::: ::: {#t2 .table-wrap} Table 2. ::: {.caption} ###### PCR results from tissues or other samples from *B. suis*-infected rhesus macaques Numbers denote the mean threshold crossing point of a triplicate for a given animal (three per time point, except in the case of reproductive organs, where the number per time point is indicated in parentheses). Brain, blood, serum, urine and small intestine were never positive. ::: **Tissue/sample type** **Day** ----------------------------- ------------------ ------------------ ------------------ ---------------------- Pharyngeal swab 37.2, 38.0 33.8, 36.7, 36.8 38.0 Nasal swab 25.8, 29.2, 29.3 30.0, 34.6, 36.2 32.9, 34.4, 36.0 34.8, 36.3, 37.6 Face swab 27.7, 27.8, 28.7 32.1, 30.3 30.5, 34.0, 34.4 32.6, 33.7, 34.2 Conjunctival swab 30.0, 30.7, 32.5 31.0, 33.6, 36.4 31.7 33.7, 34.3, 36.1 Buccal mucosal swab 35.5 36.7 Aural swab 29.1, 29.5, 29.8 30.4, 30.9, 32.7 29.6, 31.7, 33.9 33.0, 33.2, 34.4 Heart 37.2 37.5 Kidney 34.2, 36.9, 37.9 38.0 Large intestine 34.1 31.1, 33.5, 34.5 31.9, 34.6 Testes (1) (1) 35.6 (1) Epididymis (1) (1) 33.0 (1) Prostate (1) (1) 34.3 (1) Ovary (2) (2) 36.2, 36.4 (2) 33.7, 35.5, 35.7 (3) Uterine tube (2) (2) (2) (3) Uterus (2) (2) (2) (3) Mesenteric lymph node 36.4, 36.8 Tracheobronchial lymph node 29.2, 32.2, 32.8 24.6, 29.2, 29.7 26.3, 27.0, 27.1 26.8, 29.7, 34.1 Mandibular lymph node 29.6 27.5, 34.1 30.6, 31.8, 33.4 Lung 23.5, 26.7, 35.7 22.3, 24.6, 25.7 23.3, 24.2, 24.9 24.3, 24.7, 27.5 Liver 34.6, 35.5 32.3, 33.8, 34.1 Spleen 32.3, 38.4 30.8, 34.3, 35.3 Bone marrow 35.7 34.2, 35.9, 36.1 Broncho-alveolar lavage 29.0, 31.0, 31.3 32.7, 33.6, 35.1 34.0, 35.6, 36.2 34.9, 36.4 :::

PubMed Central

2024-06-05T04:04:19.352225

2010-6-01

PMC3052515

INTRODUCTION ============ The identification of viral aetiologies for hepatic ([@r4]) and cervical ([@r14]) cancers has made antiviral vaccination a relatively simple and effective means of disease prevention. Human gammaherpesviruses -- Epstein--Barr virus (EBV) and Kaposi\'s sarcoma-associated herpesvirus (KSHV) -- are also oncogenic, but the lack of single, unifying features of the associated cancers has made it unclear how directly infection and disease are linked and so what vaccination might achieve. The robust persistence of herpesviruses in immunocompetent hosts also makes vaccination a considerable challenge. EBV transforms B cells *in vitro* and, in immunocompromised patients, the viral genes responsible for transformation can cause disease ([@r7]). However, EBV-infected cancers in immunocompetent hosts tend to express the same viral genes as non-transformed cells. They differ in also carrying oncogenic host mutations; indeed, Burkitt\'s lymphoma is associated more strongly with c-myc translocation than with EBV infection ([@r44]). Thus, viral genes seem mostly to have triggering or accessory roles in disease, with host oncogenes being the main drivers. The hit-and-run hypothesis proposes that viral genomes initiating disease can be lost entirely to obscure a cancer\'s viral origin ([@r2]). Early on, viral genes are likely to be essential for cancer-cell survival ([@r17]). However, cancers accumulate vast numbers of host mutations ([@r27]), some of which will inevitably promote more autonomous growth. Thus, it seems inevitable that a cancer will, with time, evolve increasing independence from viral gene functions that could allow viral genome loss. The main problem with the hit-and-run hypothesis has been a lack of experimental support. Analyses of gammaherpesvirus-induced cancers have focused on African Burkitt\'s lymphoma, nasopharyngeal carcinoma and Kaposi\'s sarcoma, because their high frequencies of viral genome retention make plausible a causal link between infection and disease. However, focusing on virus-positive cancers tells us little about genome loss, as here most presenting cancers would be virus-negative. Instead, it is necessary to track prospectively the fate of viral genomes in transformed cells. *In vitro*, B-cell cancers tend to maintain gammaherpesvirus genomes, whereas Kaposi\'s sarcoma and nasopharyngeal carcinoma tend to lose them ([@r15]; [@r10]). *In vivo*, murid herpesvirus-4 (MuHV-4) infection increases the incidence of virus-negative cancers ([@r42]; [@r43]). However, the difficulty of analysing spontaneous cancers, where the molecular changes driving transformation are almost always unknown, makes firm functional conclusions hard to draw. To ensure that the host factors contributing to cancer remained known, we used Cre--*lox* recombination in a well-established conditional mouse cancer model (reviewed by [@r10a]) to transform virus-infected cells, and then analysed the emerging cancers for viral genome retention. RESULTS ======= Generation of Cre^+^ MuHV-4 --------------------------- We inserted a human cytomegalovirus (HCMV) IE1 promoter-driven Cre expression cassette between the 3′ ends of MuHV-4 ORFs 57 and 58 (Fig. [1a, b](#f1){ref-type="fig"}). We used an HCMV IE1 promoter because this can be active in latently infected cells ([@r31]; [@r34]). Thus, Cre could be expressed without MuHV-4 lytic genes killing the infected cells. Two functionally indistiguishable mutants were obtained. Both showed Cre expression by excising spontaneously their *loxP*-flanked bacterial artificial chromosome (BAC) cassettes, and immunofluorescence showed Cre expression in infected-cell nuclei (Fig. [1c](#f1){ref-type="fig"}). (The Cre coding sequence used incorporates an N-terminal nuclear-localization signal.) *In vivo* *loxP* recombination by Cre^+^ MuHV-4 ----------------------------------------------- We tested whether viral Cre expression could recombine *loxP* sites in the host genome by infecting mouse embryonic fibroblasts derived from ROSA26-*lacZ*^flox/flox^ reporter mice (Fig. [2a](#f2){ref-type="fig"}). *β*-Galactosidase assays were strongly positive, indicating *loxP* recombination. Such recombination was also achieved by infecting ROSA26-*lacZ*^flox/flox^ mice intraperitoneally (i.p.) with Cre^+^ MuHV-4 (Fig. [2b](#f2){ref-type="fig"}): widespread *β*-galactosidase expression was evident on the diaphragm, a site commonly infected by i.p. MuHV-4 ([@r23]). We then infected p53^flox/flox^K-ras^LSL-G12D/+^ mice i.p. with Cre^+^ MuHV-4 (Fig. [2c, d](#f2){ref-type="fig"}). More than 90 % of infected mice developed cancers within 3 months, compared with 0 % of uninfected or wild-type MuHV-infected controls. Cancers occurred most frequently on the diaphragm. Disease was rare within 30 days, and most cancers were single lesions. In contrast, virus replication was widespread: 3 days after inoculation, spleens yielded (2.1±1.2)×10^4^ and peritoneal washes (1.7±1.2)×10^3^ infectious centres per mouse (mean±[sd]{.smallcaps} titres, *n*=6, with lytic titres \<1 % of infectious centre titres); even 2 months later, spleens yielded (2.2±1.5)×10^2^ infectious centres per mouse (*n*=6). Therefore, cancer growth was much more restricted than viral latency and functional Cre expression. Analysis of virus-triggered cancers ----------------------------------- All of the cancers analysed (*n*\>12) were histological sarcomas (Fig. [3a](#f3){ref-type="fig"}). *In situ* hybridization (Fig. [3b](#f3){ref-type="fig"}) showed surprisingly little expression of the MuHV-4 tRNAs normally abundant in lytic and latent infections ([@r6]). At most, a few positive cells were scattered around the main cancer mass. Real-time PCR (Fig. [3c](#f3){ref-type="fig"}) established that sarcomas contained lower copy numbers of viral genomes than latently infected spleens of the same mice. Fresh sarcoma explants included lymphocytes, macrophages and fibroblasts (Fig. [4a, b](#f4){ref-type="fig"}), but only fibroblasts grew out. Thirteen of 20 explants yielded infectious virus. Viral spread soon overwhelmed these positive cultures, consistent with fibroblasts being highly permissive for MuHV-4 lytic replication. The others remained virus-negative. At 2 days post-explant, titres were low in all cultures (\<1 p.f.u. per 10^4^ cells), and \<5 % of fibroblasts cloned at this time (39 of 744 clones from eight mice) yielded infectious virus. Clones lacking infectious virus also lacked viral genomes by PCR (Fig. [4c](#f4){ref-type="fig"}) and Southern blotting (Fig. [4d](#f4){ref-type="fig"}). Nevertheless, all sarcomas showed the expected patterns of Cre-induced p53 disruption and k-ras(G12D) expression (Fig. [5](#f5){ref-type="fig"}). Therefore, the vast majority of cancer cells showed genetic changes consistent with previous virus infection but, by the time of presentation, were not virus-infected. A trivial explanation for the lack of viral genomes in transformed cells would be that Cre uptake from infected-cell debris was sufficient for transformation. However, infecting p53^flox/flox^K-ras^LSL-G12D/+^ mice (*n*=24) with herpes simplex virus (HSV) expressing Cre from an HCMV IE1 promoter caused no disease. Also, Cre^+^ HSV similarly shows no spread of Cre signal *in vivo* ([@r28]), and Cre^+^ MuHV-4 plaque assays on ROSA26-*lacZ*^flox/flox^ fibroblasts showed no obvious spread of *β*-galactosidase expression to uninfected cells. Even when virus was recovered from cancer cells, it might have come from infiltrating, non-transformed cells rather than being that responsible for the original oncogenic hit. We examined this possibility by infecting mice with a mix of Cre^+^ and Cre^−^ MuHV-4 and typing the virus recovered from sarcomas for Cre expression. Cre^+^ MuHV-4 showed approximately 30-fold lower latent titres than Cre^−^ virus, so we used an input Cre^+^/Cre^−^ mixture of 30 : 1. Only one of 18 virus-positive sarcoma explants was Cre^+^ by immunofluorescence. PCR and DNA sequencing of the ORF57/58 junction showed that the Cre^−^ viruses were wild-type. This did not cause sarcomas (Fig. [2](#f2){ref-type="fig"}), so even when virus infection was observed in sarcoma explants, it appeared rarely to be that responsible for transformation. Vaccination against virus-triggered cancers ------------------------------------------- The high efficiency of virus-triggered oncogenesis in our model suggested that vaccine-induced protection might be difficult to achieve. However, when Cre was substituted for ORF50 to make a replication-deficient Cre^+^ MuHV-4, both i.p. and intranasal (i.n.) infections gave no disease in p53^flox/flox^K-ras^LSL-G12D/+^ mice over 5 months (*n*=30). This lack of disease without lytic spread suggested that vaccination might still work -- for example, the cells first encountered by incoming virions might not be transformed by k-ras. We therefore immunized p53^flox/flox^K-ras^LSL-G12D/+^ mice either i.n. or i.p. with ORF73^−^Cre^−^ MuHV-4, which lacks episome maintenance and so fails to persist *in vivo* ([@r13]; [@r24]). This protected completely against Cre^+^ virus challenge (Fig. [6](#f6){ref-type="fig"}). As a further test of vaccine efficacy, we established an i.n. Cre^+^ virus challenge model (Fig. [7](#f7){ref-type="fig"}). This caused a more rapid illness than i.p. infection, with weight loss and respiratory difficulties as early as 7 days post-inoculation. The lungs of infected mice became grossly enlarged, and histological examination (Fig. [7a](#f7){ref-type="fig"}) showed extensive cell proliferation obliterating the alveolar air spaces. p53^flox/flox^K-ras^LSL-G12D/+^ mice infected with Cre^−^ MuHV-4 and p53^flox/flox^ mice infected with Cre^+^ MuHV-4 remained clinically well, so disease again reflected k-ras activation. *In situ* hybridization (Fig. [7b](#f7){ref-type="fig"}) showed viral tRNA expression in acutely infected lungs and lymphoid tissue, but not in diseased lungs. Therefore, viral genomes were again lost rapidly from the transformed cells. Vaccination i.p. with Cre^−^ORF73^−^ MuHV-4 protected completely against both macroscopic and microscopic disease (Fig. [7c--e](#f7){ref-type="fig"}). It also protected against the milder histological changes induced by Cre^+^ MuHV-4 in p53^flox/flox^ mice (Fig. [8](#f8){ref-type="fig"}). DISCUSSION ========== A viral aetiology is rarely considered for cancers that lack viral genomes. Our data show that cells driven to proliferate by host oncogenes readily lose gammaherpesvirus genomes *in vivo*. Relying on viral genome detection to establish aetiology could therefore underestimate the number of cancers to which gammaherpesviruses contribute. Most analyses of human cancers have focused on examples of genome retention. The hypothesis that these viral genomes contribute to disease ([@r17]) makes sense, as there must be a growth advantage to offset any immune recognition of viral antigens. Thus, whilst EBV genes seem not to drive the growth of EBV^+^ Burkitt\'s lymphoma directly ([@r20]), they may still provide important co-factors ([@r44]). However, the retention of viral genomes by some cancer types does not establish that viral genome retention is the norm. Interestingly, whilst EBV^+^ Burkitt\'s lymphoma is associated strongly with immunosuppressive malaria infection, EBV^−^ Burkitt\'s lymphoma occurs later and shows no such association. Thus, in immunocompetent hosts, EBV genome loss may be required for cancers to evolve. Viral antigen recognition ([@r30]) provides a context for understanding both genome-positive and genome-negative cancers. Cells driven to proliferate by the EBV growth programme are normally killed by antiviral T cells, so EBV-driven cancers are limited to the immunocompromised. In contrast, host mutations drive non-immunogenic cell proliferation even when the viral growth programme is turned off. This creates a new balance: viral genes are now required only for accessory roles, allowing viral antigen recognition to be reduced. However, some immune control may still occur -- for example, the evasion of antigen presentation by gammaherpesvirus episome-maintenance proteins ([@r46]; [@r3]) can fail at high proliferation rates ([@r25]). Also, the accumulation of host mutations is unlikely to stop. If host mutations alone remain insufficient to maintain transformation, cancer cells losing viral genomes will themselves be lost; however, if host mutations become sufficient, then antiviral T cells can select for viral genome loss. The predominance of sarcomas in our model was surprising, as MuHV-4 classically persists in B cells ([@r41]). However, stromal cells may also be an important site of persistence ([@r39]; [@r40]) -- consistent with such an idea, ORF50^−^ MuHV-4 genomes were well-maintained over 3 weeks in both BHK-21 and p53^−/−^K-ras^LSL-G12D/+^ fibroblasts (data not shown). Stromal cells may also be more sensitive than B cells to transformation by k-ras ([@r26]; [@r19]). A key point is that known viral tropisms do not necessarily predict the cell type of virus-triggered cancers. Thus, hit-and-run oncogenesis may be more relevant to rarely EBV^+^ cancers such as gastric adenocarcinoma ([@r9]; [@r32]) than to those of B cells. Even in transformed fibroblasts, MuHV-4 (unlike HSV) is far from uniformly lytic ([@r22]), and productive MuHV-4 spread is strongly constrained *in vivo* by host immunity. Therefore, it would seem quite feasible for a virus-positive cancer to develop in a cell type permissive for lytic replication. There is no certain way to identify a human cancer as previously virus-positive once it becomes virus-negative, so human gammaherpesvirus disease burdens may only be revealed by vaccination. This is not necessarily straightforward: subunit vaccines have so far failed to limit gammaherpesvirus persistence ([@r35]; [@r38]). However, live-attenuated vaccines can reduce MuHV-4 latent loads ([@r45]; [@r5]; [@r12]; [@r29]). Here, we extended this protection to a high-penetrance cancer. Latency-deficient EBV and KSHV vaccines therefore deserve serious consideration. The possibility that gammaherpesviruses contribute to more cancers than simply those remaining viral genome-positive argues that such vaccines might greatly benefit human health. METHODS ======= Mice. ----- p53^flox/flox^ ([@r21]), K-ras^LSL-G12D/+^ ([@r18]) and ROSA26-*lacZ*^flox/flox^ ([@r36]) mice were infected with MuHV-4 either i.n. under general anaesthesia (10^4^ p.f.u.) or i.p. (10^6^ p.f.u.). All experiments conformed to local and national ethical regulations. Mice were killed when they showed macroscopic cancers or other signs of ill health. All mice were examined post-mortem for clinically inapparent cancers. The PCR primer sequences for detecting *loxP* recombination were: p53 -- 5′-CACAAAAACAGGTTAAACCCAG and 5′-GAAGACAGAAAAGGGGAGGG to detect only the recombined locus (612 bp); and k-ras -- 5′-CCATGGCTTGAGTAAGTCTGC and 5′-CGCAGACTGTAGAGCAGCG to detect the 'floxed' (flanked by *loxP* sites) G12D k-ras cassette (550 bp) before but not after recombination, or 5′-GTCTTTCCCCAGCACAGTGC, 5′-CTCTTGCCTACGCCACCAGCTC and 5′-AGCTAGCCACCATGGCTTGAGTAAGTCTGCA to amplify from the floxed G12D k-ras cassette a 500 bp band before recombination and a 650 bp band after recombination. Cells. ------ For *ex vivo* explants, tissues were minced finely and digested with trypsin before culture. Embryonic fibroblasts were derived from 14 day embryos. All cells were grown in Dulbecco\'s modified Eagle\'s medium supplemented with 10 % fetal calf serum, 2 mM glutamine, 50 μM *β*-mercaptoethanol (Sigma), 100 U penicillin ml^−1^ and 100 μg streptomycin ml^−1^. All media and reagents listed here except *β*-mercaptoethanol were from PAA Laboratories GmbH. Viruses. -------- ORF73^−^ MuHV-4 has been described previously ([@r13]). To make Cre^+^ MuHV-4, an HCMV IE1 promoter-driven Cre expression cassette was excised from pGS403 ([@r33]) with *Sal*I/*Sac*II, end-repaired and cloned into the intergenic *Mfe*I site (genomic co-ordinate 77176 of GenBank accession no. U97553) of a *Bgl*II MuHV-4 genomic clone (co-ordinates 75338--78717). All other genomic co-ordinates are also given relative to GenBank accession no. U97553. The Cre expression cassette plus genomic flanks was then subcloned with *Sph*I/*Sca*I (78413--75785) into the *Sph*I/*Sma*I sites of pST76K-SR and recombined into an MuHV-4 BAC ([@r1]). Infectious virus was recovered by transfecting BAC DNA into BHK-21 cells. The BAC cassette was removed by virus passage through NIH-3T3-CRE cells ([@r37]) and virus stocks were grown in BHK-21 cells ([@r8]). Replication-deficient, Cre^+^ MuHV-4 was made by digesting a *Hin*cII genomic fragment (63844--70433) in pUC9 with *Bsm*I (67792) and *Cla*I (69177) to remove most of ORF50 exon 2 (67661--69376). The Cre coding sequence plus a 3′ poly(A) site from pGS403 was ligated in its place in frame with the ORF50 AUG. The Cre coding sequence plus genomic flanks (66120--70433) was then subcloned with *Kpn*I into pST76K-SR, and recombined into the MuHV-4 BAC. ORF50^−^Cre^+^ virus was recovered by transfecting BAC DNA into NIH-3T3-TET50 cells and inducing ORF50 expression with doxycycline ([@r23]). Virus assays. ------------- Virus stocks were titrated by plaque assay on BHK-21 cells ([@r8]). Latent virus was measured by infectious centre assay ([@r8]). Plaque titres of freeze--thawed spleen cells were always \<1 % of infectious centre assay titres. Viral genome loads were measured by quantitative PCR ([@r23]). Briefly, MuHV-4 genomic co-ordinates 4166--4252 were amplified from 50--100 ng DNA and quantified by hybridization with a Taqman probe (genomic coordinates 4218--4189) (Rotor Gene 3000; Corbett Research), in comparison with a standard curve of cloned plasmid template amplified in parallel. Cellular DNA was quantified in the same way by amplifying part of the adenosine phosphoribosyltransferase gene (forward primer, 5′-GGGGCAAAACCAAAAAAGGA; reverse primer, 5′-TGTGTGTGGGGCCTGAGTC; probe, 5′-TGCCTAAACACAAGCATCCCTACCTCAA). To quantify viral DNA by Southern blotting, DNA was extracted from cells (Wizard Genomic DNA purification kit; Promega), digested with *Pst*I, electrophoresed, transferred to Hybond nylon membranes (Roche Diagnostics), then probed with a \[^32^P\]dCTP random-primed 1.2 kb *Pst*I genomic fragment corresponding to the MuHV-4 terminal repeat unit ([@r11]), washed (65 °C, 0.2 % SSC, 0.1 % SDS) and exposed to X-ray film. Recombinant viruses were analysed qualitatively for genomic structure in a similar way, except that viral DNA was digested with *Bgl*II or *Hin*dIII and probed with a *Bgl*II-restricted genomic fragment (co-ordinates 75338--78717) or the HCMV IE1--Cre construct. Cells expressing viral tRNAs 1--4 were detected by *in situ* hybridization of formaldehyde-fixed, paraffin-embedded spleen cell sections, using a digoxigenin-labelled riboprobe transcribed from pEH1.4 ([@r6]). Hybridized probe was detected with alkaline phosphatase-conjugated anti-digoxigenin Fab fragments (Roche Diagnostics). *β*-Galactosidase assay. ------------------------ *In vitro* samples were fixed in 4 % formaldehyde (30 min), then washed in PBS and incubated (3 h, 37 °C) in PBS with 0.01 % sodium deoxycholate, 0.02 % Nonidet P-40, 2 mM MgCl~2~, 4.5 mM potassium ferricyanide, 4.5 mM potassium ferrocyanide, 1 mg X-Gal ml^−1^, before washing. *In vivo* samples were fixed in 4 % formaldehyde (18 h) then frozen in OCT medium, sectioned, washed in PBS and developed as described above before washing and mounting. Immunofluorescence. ------------------- Cells were plated onto glass cover slides, then fixed (4 % formaldehyde, 30 min), permeabilized (0.1 % Triton X-100, 15 min), blocked (3 % BSA in PBS, 15 min) and stained for syndecan-1, CD44, VCAM-1 (all mAbs from BD Biosciences) or with the macrophage-specific mAb F4/80 (AbCam) plus Alexa Fluor 568-conjugated goat anti-rat IgG pAb (Invitrogen), for the MuHV-4 ORF75c using mAb BN-6C12 ([@r16]) plus Alexa Fluor 568-conjugated goat anti-mouse IgG pAb (Invitrogen), for MuHV-4 antigens using a polyclonal rabbit serum ([@r41]) and for Cre recombinase using a polyclonal rabbit serum (AbCam) plus goat anti-rabbit IgG pAb (Invitrogen). The cells were mounted in ProLong Gold anti-fade reagent with DAPI (Invitrogen) and imaged using an Olympus IX70 microscope plus a Retiga 2000R camera line (QImaging). We thank Heather Coleman for generating ORF50^−^Cre^+^ MuHV-4, Dave Tuveson and Doug Winton for providing mice, and Barry Potter and Ming Du for help with histology. P. G. S. is a Wellcome Trust Senior Clinical Fellow (GR076956MA). This work was also supported by the Medical Research Council (G0701185) and the Wellcome Trust (WT089111MA). ::: {#f1 .fig} Fig. 1. ::: {.caption} ###### Characterization of Cre^+^ MuHV-4. (a) An HCMV IE1 promoter-driven Cre expression cassette was inserted between MuHV-4 ORFs 57 and 58. Relevant restriction sites are shown. (b) Viral DNA was digested with *Hin*dIII or *Bgl*II and probed with either a genomic *Bgl*II clone or the HCMV IE1--Cre construct, as shown in (a). WT, Wild-type; Cre^+^, recombinant; Cre^+^ind, independently derived recombinant. (c) BHK-21 cells were infected with wild-type or Cre^+^ MuHV-4 (1 p.f.u. per cell, 16 h), then fixed, permeabilized and stained for Cre recombinase or for MuHV-4 antigens using polyclonal rabbit sera. Nuclei were counterstained with DAPI. :::  ::: ::: {#f2 .fig} Fig. 2. ::: {.caption} ###### Cre recombinase-triggered cancers in MuHV-4-infected mice. (a) Embryonic fibroblasts from ROSA26-*lacZ*^flox/flox^ mice were infected (0.3 p.f.u. per cell, 16 h) with either wild-type or Cre^+^ MuHV-4, then fixed and incubated with X-Gal to reveal *β*-galactosidase expression, indicating Cre-mediated recombination. Arrowheads show examples of positive staining. (b) ROSA26-*lacZ*^flox/flox^ mice were infected i.p. with Cre^+^ MuHV-4. Three days later, diaphragms were stained post-mortem for *β*-galactosidase expression with X-Gal. Representative images from two mice are shown. (c) p53^flox/flox^K-ras^LSL-G12D/+^ mice were infected i.p. with wild-type or Cre^+^ MuHV-4. All of the former mice remained healthy; all but one of those infected with Cre^+^ MuHV-4 developed cancers within 3 months. Equivalent results were obtained in three further experiments. (d) A typical i.p. cancer. :::  ::: ::: {#f3 .fig} Fig. 3. ::: {.caption} ###### MuHV-4-triggered sarcomas. (a) Representative haematoxylin/eosin-stained sections from p53^flox/flox^K-ras^LSL-G12D/+^ mice infected with Cre^+^ MuHV-4. Bar, 100 μm. (b) Cancer or spleen sections of Cre^+^ MuHV-4-infected p53^flox/flox^K-ras^LSL-G12D/+^ mice were probed for MuHV-4 tRNAs 1--4. Representative images are shown. Arrowheads show positive cells. Bar, 100 μm. (c) DNA samples from paired cancers and spleens were analysed for viral genome copy number by quantitative PCR. Each viral copy number is expressed relative to the cellular DNA copy number in the same sample. :::  ::: ::: {#f4 .fig} Fig. 4. ::: {.caption} ###### Analysis of explanted cancer cells from Cre^+^ MuHV-4-infected p53^flox/flox^K-ras^LSL-G12D/+^ mice. (a) A typical phase-contrast image of a primary cancer culture 1 day post-explant. (b) Immunostaining of a primary cancer culture at 3 days post-explant shows typical VCAM-1^+^CD44^+^CD138^+^ fibroblasts, and some F4/80^+^ macrophages. Occasional fibroblasts (\<1 %) were viral antigen-positive, shown here by staining for the ORF75c tegument protein. (c) Cloned cancer cells were analysed for viral genomes by quantitative PCR. Viral DNA copy numbers are expressed relative to cellular DNA copy numbers. Only clone 10 yielded infectious virus; below the dashed line (\<1 viral genome per 100 cell genomes), clones were considered virus-negative. (d) A subset of the clones in (c) was further analysed by probing *Pst*I-digested DNA (1 mg per lane=500 000 cells) for the MuHV-4 1.2 kb terminal repeat (approx. 30 copies per genome) by Southern blotting. One picogram of plasmid DNA=200 000 copies, so no detectable viral genomes implies \<1 copy per 75 cells. :::  ::: ::: {#f5 .fig} Fig. 5. ::: {.caption} ###### PCR detection of Cre-mediated recombination in samples from Cre^+^ MuHV-4-infected p53^flox/flox^K-ras^LSL-G12D/+^ mice. (a) PCR analysis of the p53 locus of two p53^flox/flox^K-ras^LSL-G12D/+^ mice, their primary cancers and fibroblast clones derived from them. The primers amplify the floxed p53 locus only after recombination (612 bp). Identical data were obtained for a further 10 mice. Negative images of ethidium bromide-stained PCR products are shown. (b) PCR analysis of the floxed G12D k-ras cassette of the same samples. The primers amplify the cassette (550 bp) before but not after recombination. (c) Multiplex PCR analysis of the ras locus of the same samples plus additional controls. The primers amplify from the wild-type k-ras locus a 622 bp band, and from the floxed G12D k-ras cassette a 500 bp band before recombination and a 650 bp band after recombination. The clones lack the 500 bp band of the parental cancers because they contain no cells with unrecombined G12D k-ras. WT, p53^flox/flox^G12D k-ras^−/−^ littermate; mut, purified 500 bp band; control DNA, non-transgenic mice. :::  ::: ::: {#f6 .fig} Fig. 6. ::: {.caption} ###### Vaccination against MuHV-4-triggered sarcomas. (a) p53^flox/flox^K-ras^LSL-G12D/+^ mice were not vaccinated or vaccinated i.p. with ORF73^−^Cre^−^ MuHV-4, then 2 months later challenged i.p. with Cre^+^ MuHV-4 and followed for cancer incidence. At 4 months, the vaccinated mice showed no disease. The data are from one of two equivalent experiments. (b) p53^flox/flox^K-ras^LSL-G12D/+^ mice were not vaccinated or vaccinated i.n. with ORF73^−^Cre^−^ MuHV-4, then 2 months later challenged i.p. with Cre^+^ MuHV-4 as in (a). The data are from one of two equivalent experiments. (c) In an equivalent experiment to (b), spleens were were analysed for viral DNA content by quantitative PCR 1 month after Cre^+^ virus challenge. Viral genomes per cell genome are shown for each mouse (means of three replicate reactions). The dashed line shows the sensitivity limit of one viral genome per 500 cell genomes. :::  ::: ::: {#f7 .fig} Fig. 7. ::: {.caption} ###### Vaccination against i.n. Cre^+^ MuHV-4 challenge. (a) p53^flox/flox^K-ras^LSL-G12D/+^ or p53^flox/flox^ mice were infected i.n. with Cre^+^ MuHV-4. Lungs were examined by haematoxylin/eosin staining at 15 or 35 days post-infection. The p53^flox/flox^ mice showed moderate abnormalities but remained clinically well. Bar, 100 μm. Sections are representative of at least six mice per group. (b) p53^flox/flox^K-ras^LSL-G12D/+^ mice were not infected or infected i.n. with Cre^+^ MuHV-4. Lungs and mediastinal lymph nodes (MLN) were analysed for viral tRNAs by *in situ* hybridization. The sections are each representative of at least five mice per group. The arrows show examples of positive cells. Bar, 100 μm. (c) p53^flox/flox^K-ras^LSL-G12D/+^ mice were vaccinated i.p. with ORF73^−^Cre^−^ MuHV-4, and 2 months later challenged i.n. with Cre^+^ MuHV-4. Mice were killed when they showed \>20 % weight loss or progressive respiratory difficulties. The vaccinated mice remained entirely well. Equivalent data were obtained in one further experiment. (d) *Ex vivo* p53^flox/flox^K-ras^LSL-G12D/+^ lungs (three per group) are shown 1 month after i.n. Cre^+^ MuHV-4, after the same challenge but vaccinated i.p. with Cre^−^ORF73^−^ MuHV-4 2 months earlier, or without infection. Equivalent results were obtained in three further experiments. (e) Lungs of p53^flox/flox^K-ras^LSL-G12D/+^ mice were examined by haematoxylin/eosin staining 35 days post-infection with Cre^+^ MuHV-4. The lungs of vaccinated mice were macroscopically and histologically normal. Three representative images are shown for each group. Equivalent results were obtained in two further experiments, each with five mice per group. Bar, 100 μm. :::  ::: ::: {#f8 .fig} Fig. 8. ::: {.caption} ###### Protection of p53^flox/flox^ mice against i.n. Cre^+^ MuHV-4 by an ORF73^−^Cre^−^ vaccine. p53^flox/flox^ mice were not vaccinated or vaccinated i.p. with ORF73^−^Cre^−^ MuHV-4, then 3 months later challenged i.n. with ORF73^+^Cre^+^ MuHV-4. Lungs were examined histologically at 1 month post-challenge. Equivalent p53^flox/flox^K-ras^LSL-G12D/+^ lungs are shown in Fig. [7](#f7){ref-type="fig"}. Bars, 100 μm. The results are representative of \>15 mice per group from three independent experiments. :::  :::

PubMed Central

2024-06-05T04:04:19.354909

2010-9-01

PMC3052518

*Vaccinia virus* (VACV) is the best-studied member of the family *Poxviridae*; it is a double-stranded DNA virus that replicates in the cytoplasm of host cells ([@r17]). VACV was used as the vaccine to eradicate smallpox ([@r7]) and thereafter has continued to be studied due to its development as a vaccine for other infectious diseases ([@r21]), and because it is an excellent model for studying virus--host interactions. Although immunoprophylaxis with VACV led to eradication of smallpox, VACV had an imperfect safety record ([@r11]) and it was recognized that the safety profile needed improvement if VACV was to be used again widely for human vaccination. One strategy to improve safety was to passage VACV strains repeatedly in cell culture until attenuation was achieved, the classical manner in which live-attenuated vaccines were developed for yellow fever, polio, measles, mumps and rubella. In this way, the LC16m8 strain was derived from VACV Lister in Japan ([@r9]), and modified virus Ankara (MVA) was derived from VACV chorioallantois virus Ankara (CVA) in Germany by more than 570 passages in chicken embryo fibroblasts (CEFs) ([@r13]; [@r23]; [@r14]). During this serial passage, MVA incurred six major deletions ([@r16]) and many smaller mutations ([@r1]) compared with the parental CVA ([@r15]), and developed a severe restriction in host range so that it is unable to replicate in most mammalian cells ([@r16]; [@r24]). In the later stages of the smallpox-eradication campaign, MVA was used as a smallpox vaccine in over 120 000 individuals without complications ([@r23]; [@r14]; [@r12]). MVA lacks many immunomodulators that are present in other strains of VACV ([@r1]; [@r2]), but other immunomodulators remain and deletion of two of these, B15 and A41, enhanced MVA immunogenicity ([@r22]; [@r5]). This paper concerns MVA protein 183, the counterpart of VACV strain Western Reserve (WR) protein B14 ([@r3]). VACV WR protein B14 is expressed early after infection in the cytoplasm and is non-essential for virus replication in cell culture ([@r3]). However, a *B14R* deletion mutant (vΔB14) was attenuated in a mouse intradermal (i.d.) infection model and affected the inflammatory response to infection ([@r3]). Further study demonstrated that B14 inhibits nuclear factor *κ*B (NF-*κ*B) activation by binding to the inhibitor of *κ*B kinase (IKK) complex via the IKK*β* subunit ([@r4]). The crystal structure of B14 was solved and revealed a B-cell lymphoma (Bcl)-2 protein fold, despite B14 lacking sequence similarity to Bcl-2 proteins, a family of proteins that regulate apoptosis. Anti-apoptotic Bcl-2 proteins contain a surface groove that binds the BH3 peptide of pro-apoptotic Bcl-2 family members and thereby neutralizes pro-apoptotic activity. In comparison, B14 lacked a surface groove and was not anti-apoptotic ([@r8]). Given that deletion of *B14R* from VACV WR increased the recruitment of leukocytes to the site of infection ([@r3]), it was hypothesized that deletion of gene *183R* might increase MVA immunogenicity and increase its potency as a vaccine. B14 and 183 share 95 % amino acid identity, and there are three sequence differences (Fig. [1a](#f1){ref-type="fig"}). First, in B14 *α*-helix 1 (*α*1; Fig. [1b](#f1){ref-type="fig"}), there is a conservative R27K substitution in MVA 183 and the same change is seen in CVA, rabbitpox virus Utrecht (a VACV strain) and VACV strains Acambis 3 and 3737. Second, in B14 *α*-helix 5, there is another conservative substitution (A84V) in MVA. In different VACV strains and other orthopoxviruses, alanine, valine or threonine is found at this position (Fig. [1b](#f1){ref-type="fig"}). Third, and most notably, there is a 6 aa deletion in MVA 183 corresponding to B14 *α*-helix 6 (*α*6; Fig. [1b](#f1){ref-type="fig"}), and this deletion is not found in any other VACV strains or orthopoxviruses. The central position of this helix in the B14 structure suggests that this deletion in 183 might affect the overall fold of the protein. In contrast, the other amino acid substitutions are unlikely to do so. To examine whether protein 183 behaves in an analogous manner to B14 as an inhibitor of NF-*κ*B activation, the *183R* gene including DNA encoding an N-terminal FLAG tag was cloned into the mammalian expression vector pCI (Promega) downstream of a human cytomegalovirus immediate-early promoter. HEK 293T cells ([@r4]) were transfected with a reporter plasmid containing firefly luciferase attached to an NF-*κ*B-responsive promoter ([@r4]), an internal control containing *Renilla* luciferase linked to the thymidine kinase promoter ([@r4]), and either pCI or pCI encoding FLAG-tagged B14 or 183 (Fig. [2a](#f2){ref-type="fig"}). After 18 h, the transfected cells were stimulated with tumour necrosis factor alpha (TNF-*α*; Peprotech) for 8 h. NF-*κ*B activation was assessed by firefly luciferase activity and normalized by *Renilla* luciferase activity. Triplicate samples were collected and data are expressed as the mean fold induction relative to pCI. Whilst B14 inhibited NF-*κ*B activation as expected, MVA 183 had no such activity (Fig. [2a](#f2){ref-type="fig"}). To compare levels of B14 and 183 protein expression, cell lysates from the reporter assays were analysed by immunoblotting with anti-FLAG (Sigma) and anti-*α*-tubulin (Upstate Biotechnology) mAbs (Fig. [2b](#f2){ref-type="fig"}). B14 was easily detectable, whereas 183 was not. However, after TNF-*α* treatment, the level of 183 increased for unknown reasons. Blotting with an anti-*α*-tubulin mAb confirmed equal loading of samples. It appeared, therefore, that 183 was unstable, but stability was increased by activation of the signalling pathway leading from TNF-*α* to NF-*κ*B activation. Nonetheless, despite 183 expression being detectable after TNF-*α* stimulation, under the conditions tested it was unable to inhibit NF-*κ*B activation. MVA does not replicate in many mammalian cell lines and is avirulent in mice. Therefore, to investigate the role of MVA 183 in virus virulence, the *183R* gene was inserted into VACV strain WR lacking *B14R*, vΔB14 ([@r3]), and the virulence of the recombinant virus, vΔB14-MVA183, was measured in a murine i.d. model ([@r25]). A pSJH7-based plasmid ([@r10]) containing the *183R* open reading frame flanked by 250 bp of WR genomic DNA upstream and downstream of *B14R* was transfected into RK-13 cells (European Collection of Cell Cultures; ECACC) that were infected with vΔB14. A recombinant virus containing MVA 183, vΔB14-MVA183, was selected by transient dominant selection ([@r6]) and PCR analysis confirmed that vΔB14-MVA183 contained the correct genome structure at the *B14R/183R* locus (see Supplementary Fig. S1, available in JGV Online). vΔB14-MVA183 was purified in parallel with viruses vB14 (wild-type, WR) and vΔB14 ([@r3]) by sedimentation through a sucrose cushion, and virus infectivity was titrated on BS-C-1 cells (ECACC). Groups of female C57Bl/6 mice, between 6 and 8 weeks old (Harlan), were anaesthetized and inoculated with each virus into each ear pinna ([@r25]; [@r26]). Titres of inocula were confirmed by plaque assay. Animals were examined daily and the diameter of lesions at the inoculation site was measured using a micrometer. vΔB14 induced a smaller lesion size than wild-type virus and reinsertion of *B14R* into vΔB14 restored lesion size, as reported previously ([@r3]). In contrast, insertion of *183R* into vΔB14 did not alter lesion size (Fig. [3a](#f3){ref-type="fig"}). Therefore, 183 does not contribute to virulence in this model, a conclusion consistent with the protein being unstable and unable to inhibit NF-*κ*B activation in cell culture. To examine 183 expression in infected cells, lysates from CEFs infected with MVA or vΔB14-MVA183 were analysed by immunoblotting using the anti-B14 polyclonal antibody ([@r3]) and mouse mAbs against D8 (a VACV late protein) ([@r19]) and *α*-tubulin (Fig. [3b](#f3){ref-type="fig"}). As before, although B14 produced by VACV WR was detected easily, MVA 183 was not detected in cells infected with MVA or vΔB14-MVA183. This difference was unlikely to be due to altered transcription of gene *183R*, as the nucleotide sequence upstream of MVA *183R* was the same in WR *B14R*, and so the difference probably reflected protein instability as noted above (Fig. [2b](#f2){ref-type="fig"}). A major pathway for the removal of labile proteins in eukaryotic cells is via ubiquitination and degradation by the proteasome. To assess whether protein 183 was being degraded by this mechanism, a proteasome inhibitor (MG132; [@r20]) was added to cells prior to and throughout infection. MG132 is a short oligopeptidic sequence with an aldehyde electrophilic structure at the C terminus. It blocks multiple active sites in the proteasome by forming transition-state analogues via interacting with a catalytic hydroxyl/thiol group to form a reversible hemi (thio)acetal. In the presence of MG132, protein 183 expression from MVA or vΔB14-MVA183 was detected and was similar in each case (Fig. [3b](#f3){ref-type="fig"}). This showed that 183 was labile and degraded by the proteasome. It also indicated that additional factors expressed by VACV WR were unable to stabilize protein 183. Analysis of VACV protein D8 expression levels in these cells showed a substantial decrease in the presence of MG132. This observation was in accord with the report that MG132 blocks a post-entry step in VACV replication ([@r20]). After virus entry and uncoating, viral DNA synthesis is prevented so that virus intermediate- and late-gene expression is inhibited ([@r20]). D8 is a late protein and is therefore reduced considerably by MG132. This reduction in D8 expression also served as a control, showing that MG132 was inhibiting proteasome activity. Several lines of evidence indicate that MVA 183 is a labile protein. First, following transfection of mammalian cells with plasmids encoding 183, the protein was not detected, whereas B14 was. Second, after infection of avian cells by MVA or a VACV WR virus encoding MVA 183, the 183 protein was not detectable unless the proteasome was inhibited by MG132. Third, attempts to express protein 183 in *Escherichia coli* under conditions that yielded stable, soluble B14 protein that was used for protein crystallography ([@r8]) gave only low yields of insoluble 183 (data not shown). Therefore, 183 is a labile protein and the probable cause is deletion of the REISAI motif in the B14 *α*6 helix. This deletion is found only in MVA and in no other VACV strain, including the parental virus CVA, and so arose during MVA passage in CEFs. The observation that MVA 183 does not inhibit NK-*κ*B activation may explain, at least in part, why MVA infection activates NF-*κ*B in contrast to other VACV strains ([@r18]). Interestingly, protein 183 was stabilized after treatment of cells with TNF-*α*. Possible explanations for this are alterations to the proteasome-mediated degradation of proteins after TNF-*α* stimulation, or that a component of the TNF-*α*-induced signalling pathway leading to NF-*κ*B activation can somehow stabilize the 183 protein once the pathway is activated. However, if the latter were true, it is notable that, even when 183 is present, at the levels tested it did not inhibit NF-*κ*B activation. It is also formally possible, but less likely, that TNF-*α* treatment is affecting 183 mRNA levels. The original bioinformatic analysis of the MVA genome recorded open reading frames that were retained, deleted or disrupted by mutation compared with other VACV strains ([@r1]), but the functional consequences of smaller mutations, such as those noted here for protein 183, were unknown. Here we demonstrate that, although MVA protein 183 shares 95 % amino acid sequence identity with protein B14 from VACV strain WR, it is unstable and lacks the functions assigned to protein B14, such as inhibition of NF-*κ*B activation and contribution to virus virulence. Other MVA genes encoding immunomodulatory proteins with small deletions or mutations relative to counterparts in other VACV strains are as yet uncharacterized and these changes may have important functional consequences. Attempts to increase the immunogenicity of MVA by engineering or deletion of genes encoding immunomodulators should be informed by knowledge of whether the encoded protein is functional, and priority should be given to engineering genes known to encode functional proteins. In summary, MVA protein 183 is a non-functional counterpart of VACV WR protein B14 and neither inhibits NF-*κ*B activation following addition of TNF-*α*, nor contributes to VACV strain WR virulence. The loss of function is probably attributable to a 6 aa deletion within the *α*6 helix of the Bcl-2 fold and it is unlikely that removal of this gene from the MVA genome will enhance MVA immunogenicity. Supplementary Material ====================== ::: {.caption} ###### \[Supplementary figure\] ::: This research was conducted with grant support from the Wellcome Trust and as part of the Poxvirus T-Cell Vaccine Discovery Consortium (PTVDC) under the Collaboration for AIDS Vaccine Discovery with support from the Bill and Melinda Gates Foundation. G. L. S. is a Wellcome Trust Principal Research Fellow. A supplementary figure showing PCR analysis of the *B14R* locus is available with the online version of this paper. ::: {#f1 .fig} Fig. 1. ::: {.caption} ###### B14 structure. (a) Alignment of proteins WR B14 and MVA 183. Positions of divergence are shown in bold and underlined. The positions of the *α*-helices of the B14 crystal structure are shown above the alignment and are coloured from N to C terminus as shown in the three-dimensional structure in (b). (b) Ribbon representation of the crystal structure of B14. *α*-Helices are marked by sequence from N to C terminus and are coloured from blue (N terminus) to magenta (C terminus). The changes in 183 compared with B14 are shown underlined and in bold, including the REISAI motif that is deleted in 183. :::  ::: ::: {#f2 .fig} Fig. 2. ::: {.caption} ###### Protein 183 does not inhibit NF-*κ*B activation by TNF-*α*. (a) HEK 293T cells were transfected with either pCI or pCI containing FLAG-tagged B14 or 183 (pB14f and p183f) and a luciferase NF-*κ*B reporter plasmid for 18 h. Cells were then stimulated with 50 ng TNF-*α* ml^−1^ for 8 h and luminescence was assessed as described previously ([@r4]). The assay was carried out in triplicate and the error bars represent [sd]{.smallcaps} from the mean. (b) A small aliquot of the cell lysates was retained and analysed by immunoblotting with anti-FLAG and anti-*α*-tubulin mAbs. The positions of molecular mass markers in kDa are indicated. :::  ::: ::: {#f3 .fig} Fig. 3. ::: {.caption} ###### \(a) Measurement of virulence of VACV WR strains with or without MVA 183. (a) Groups of eight C57Bl/6 mice were infected i.d. with 10^4^ p.f.u. vB14 (▪), vΔB14 (▴) or vΔB14-MVA183 (•) in 10 μl PBS and the lesion diameter was recorded daily. Error bars represent [sem]{.smallcaps}. (b) Expression of B14 and 183 proteins after infection of CEFs. CEFs were infected at 10 p.f.u. per cell with VB14, vΔB14-MVA183 or MVA. Where indicated, cells were treated with MG132 (10 μM) for 1.5 h before and throughout infection. Infected cells were incubated for 8 h before lysis and analysis by SDS-PAGE and immunoblotting with anti-B14, anti-D8 and anti-*α*-tubulin antibodies. The positions of molecular mass markers in kDa are indicated. :::  ::: [^1]: †**Present address:** MRC/UCL Centre for Medical Molecular Virology, Division of Infection and Immunity, University College London, 46 Cleveland Street, London W1T 4JF, UK. [^2]: ‡**Present address:** Cancer Research UK Gurdon Institute, University of Cambridge, Tennis Court Road, Cambridge CB2 1QN, UK. [^3]: §**Present address:** Oxford University Medical School, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DU, UK.

PubMed Central

2024-06-05T04:04:19.357039

2010-9-01

PMC3052519

INTRODUCTION ============ *Vaccinia virus* (VACV) is the prototypical member of the genus *Orthopoxvirus* (OPV). As with all poxviruses, VACV is a large dsDNA virus that replicates within the cytoplasm of infected cells ([@r26]). This replication strategy results in the production of immunostimulatory nucleic acids that have the potential to activate the innate immune response via pattern-recognition receptors (PRRs). Not surprisingly, VACV therefore encodes many intracellular and extracellular modulators of innate immunity that serve to prevent the host from mounting an effective immune response to infection (for examples see [@r2]; [@r6]; [@r10]; [@r14]; [@r29]; [@r33]; for reviews see [@r27]; [@r32]). Self and non-self cytoplasmic dsDNA induces the expression of type I interferon (IFN) and NF-*κ*B regulated genes independently of Toll-like receptor 9 (TLR9) ([@r19]). However, the signalling pathways involved in this process are poorly understood. The DNA-dependent activator of IFN regulatory factors, DAI (also known as ZBP1), was one candidate for the activation of IFN signalling in response to dsDNA in certain cell types ([@r34]). DAI contains two Z-DNA-binding domains that bind dsDNA in a sequence-independent manner. Once bound to dsDNA, DAI is thought to recruit TANK-binding kinase 1 (TBK1) and IFN regulatory factor 3 (IRF3). IRF3 is then activated and translocates to the nucleus, resulting in the induction of type I IFN ([@r34]). In addition, DAI also recruits receptor-interacting protein kinases, RIP1 and RIP3, through RIP homotypic interaction motif-dependent interactions, resulting in the activation of NF-*κ*B ([@r20]). However, other dsDNA sensors must exist because murine embryonic fibroblasts (MEFs) lacking DAI still respond normally to intracellular dsDNA ([@r35]). Indeed, recent work has identified absent in melanoma 2 (AIM2) ([@r5]; [@r13]; [@r18]; [@r31]) and RNA polymerase III ([@r1]; [@r11]) as two further dsDNA sensors. As a consequence of its cytoplasmic replication it is probable that VACV will encode inhibitors of cytosolic dsDNA sensing, and one candidate is E3. E3 is an intracellular protein that is highly conserved among OPVs and is expressed early during infection localizing to both the cytoplasm and nucleus ([@r36]; [@r39]). E3 is important for VACV resistance to IFN, and can be split into distinct N- and C-terminal halves. The C terminus of E3 contains a dsRNA-binding domain that serves to sequester dsRNA produced during viral replication, thereby preventing the activation of protein kinase R (PKR) and RNase L, as well as, the activation of IRF3 ([@r7]; [@r8], [@r9]; [@r36]; [@r37]). Consequently, the dsRNA-binding domain is important for the IFN resistant phenotype of VACV ([@r3]; [@r9]). The E3 N terminus, on the other hand, contains a region with amino acid similarity to the Z-DNA-binding domain of DAI ([@r34]). The biological significance of this similarity is poorly understood; however, the N terminus is required for full VACV virulence ([@r4]; [@r3]). Interestingly, the Z-DNA-binding domain of DAI and another Z-DNA-binding protein, adenosine deaminase 1 (ADAR1), can substitute for the E3 N terminus in promoting VACV virulence ([@r22]). Furthermore, mutation of residues in the putative Z-DNA-binding domain of E3 analogous to those involved in contacting Z-DNA in ADAR1 and DAI reduced VACV virulence ([@r22]), suggesting that the E3 N terminus functions by binding dsDNA. Intriguingly, expression of full-length E3 was reported to reduce IFN-*β* expression in response to the dsDNA species poly(dA--dT) in MEFs ([@r35]). However, the role of the N terminus in this phenotype was not investigated. These observations, coupled with reports that DAI functions as a PRR prompted us to investigate whether the E3 Z-DNA-binding domain antagonizes intracellular dsDNA PRRs by sequestering dsDNA. RESULTS ======= Transfection of poly(dA--dT) induces IFN-*β* in a dose-dependent manner independently of TLR9 --------------------------------------------------------------------------------------------- The dsDNA, poly(dA--dT), is a potent inducer of type I IFN following the transfection of MEFs ([@r19]), and E3 can inhibit IFN-*β* expression in response to poly(dA--dT) ([@r25]; [@r35]). For consistency with previous reports we therefore used poly(dA--dT) in this study. Initially, a suitable reporter assay system in which the role of the E3 N terminus in blocking dsDNA signalling could be tested was developed by transfecting 293T cells with an IFN-*β* reporter and titrating increasing doses of poly(dA--dT). In accordance with previous reports, poly(dA--dT) induced IFN-*β* reporter activity in a dose-dependent manner (Fig. [1a](#f1){ref-type="fig"}). 293T cells express low levels of TLR9, which detects extracellular unmethylated CpG DNA ([@r17]). Although, it seems unlikely that poly(dA--dT) would be a natural ligand for TLR9 it was necessary to demonstrate that induction of IFN-*β* by poly(dA--dT) was TLR9 independent. Accordingly, 293T cells were transfected with poly(dA--dT) or, alternatively, poly(dA--dT) was added to the culture medium in the absence of transfection reagent (polyethylenimine, PEI) and the effects upon IFN-*β* reporter activity were determined. Transfection of poly(dA--dT) led to a 75-fold activation of IFN-*β* reporter activity relative to the mock control. However, this effect was completely ablated when poly(dA--dT) was added to the culture medium in the absence of transfection reagent (Fig. [1b](#f1){ref-type="fig"}). These results, therefore, confirm that poly(dA--dT) is a potent stimulator of an intracellular dsDNA PRR. The E3 dsRNA-binding domain inhibits IFN-*β* expression in response to poly(dA--dT) independently of the E3 Z-DNA-binding domain -------------------------------------------------------------------------------------------------------------------------------- Next, the ability of E3 to block IFN-*β* reporter activity in response to the transfection of poly(dA--dT) was tested. 293T cells were co-transfected with the IFN-*β* reporter together with expression plasmids encoding N-terminally FLAG-tagged E3 and deletion mutants lacking either the N-terminal 83 aa (Δ83N) or the C-terminal 26 aa (Δ26C) (Fig. [2a](#f2){ref-type="fig"}). Transfection of poly(dA--dT) and empty vector control (pcDNA4) resulted in a 13-fold induction of IFN-*β* luciferase activity, and in agreement with earlier reports ([@r25]; [@r35]), expression of full-length E3 led to an almost complete ablation of IFN-*β* reporter activity. However, contrary to what was expected, Δ83N lacking the Z-DNA-binding domain was almost as efficient as full-length E3 in blocking IFN-*β*. In contrast, removal of the C-terminal 26 aa, which prevents E3 binding dsRNA ([@r7]), removed the inhibitory activity and was comparable to the empty vector control (Fig. [2b](#f2){ref-type="fig"}). Similar results were also obtained in HeLa cells confirming these effects were not cell type specific (Fig. [2c](#f2){ref-type="fig"}). Furthermore, the inability of the Z-DNA-binding domain to block poly(dA--dT) signalling was confirmed using a different set of vectors expressing either the Z-DNA-binding (aa 1--83) or dsRNA-binding (Δ83N, aa 84--190) domains of E3 (Fig. [2a, d](#f2){ref-type="fig"}) both of which are expressed at equivalent levels (Fig. [2e](#f2){ref-type="fig"} and data not shown). The failure of Δ26C to inhibit IFN-*β* reporter activity was not due to an absence of protein because immunoblotting with anti-FLAG monoclonal antibody (mAb) confirmed protein expression (Fig. [2e](#f2){ref-type="fig"}). Furthermore, despite being relatively unstable in comparison to full-length E3 and Δ26C, the Δ83N deletion mutant is still a potent inhibitor of poly(dA--dT) signalling. The ability of wild-type and mutant E3 to block IFN-*β* secretion from 293T cells induced by poly(dA--dT) was tested by ELISA. In agreement with the luciferase data transfection of poly(dA--dT) induced IFN-*β* secretion in those cells transfected with pcDNA4 and this was blocked following expression of either E3 or Δ83N (Fig. [3a](#f3){ref-type="fig"}). However, this inhibitory effect was lost following the expression of Δ26C, confirming that the dsRNA, but not Z-DNA-binding domain, is important for the inhibition of poly(dA--dT) signalling. The ability of E3 to block IFN-stimulated response element (ISRE) reporter activity in response to poly(dA--dT) was examined next. It was hypothesized that if the E3 dsRNA-binding domain was inhibiting IFN-*β* production in response to poly(dA--dT), an indirect inhibition of ISRE activity should be observed. In agreement with the ELISA data (Fig. [3a](#f3){ref-type="fig"}), transfection of poly(dA--dT) led to a modest threefold stimulation of ISRE activity, whereas both full-length E3 and Δ83N produced a statistically significant inhibition (Fig. [3b](#f3){ref-type="fig"}). This effect was completely ablated in those cells expressing E3 Δ26C, consistent with the IFN-*β* reporter data. Moreover, E3 was unable to inhibit ISRE reporter activity in response to IFN-*α* stimulation, confirming that the E3-mediated reduction in ISRE activity was not a consequence of E3 directly modulating the Jak/STAT signalling pathway (Fig. [3c](#f3){ref-type="fig"}). To extend these observations we tested whether E3 associated with cytoplasmic biotinylated dsDNA by co-precipitation and found no interaction by either immunoblotting or silver staining (Fig. [4](#f4){ref-type="fig"}). In contrast, the interaction of several cytoplasmic dsDNA-binding proteins with biotinylated DNA was detected and this was unaffected by the expression of E3 (Fig. [4](#f4){ref-type="fig"} upper panel). 293T cells transfected with a plasmid expressing FLAG-tagged VACV protein B14 served as a negative control. The B14 protein binds IKK*β* and inhibits NF-*κ*B activation ([@r10]) and would not be expected to bind dsDNA. In line with this, no binding of B14 to biotinylated dsDNA was detected (Fig. [4](#f4){ref-type="fig"}). Therefore, contrary to what was expected, the E3 dsRNA-binding domain is an inhibitor of the poly(dA--dT) signalling pathway, whilst the E3 Z-DNA-binding domain is entirely dispensable. The E3 dsRNA-binding domain inhibits NF-*κ*B activity in response to poly(dA--dT) --------------------------------------------------------------------------------- The transfection of dsDNA induces not only IFN-*β* expression but also NF-*κ*B activation ([@r20]; [@r34]) and, therefore, the ability of E3 to block NF-*κ*B reporter activity in response to poly(dA--dT) was investigated. Transfection of poly(dA--dT) resulted in a fivefold activation of luciferase activity in cells transfected with the empty vector control plasmid and this was blocked in cells expressing either full-length E3 or Δ83N. In contrast, Δ26C was not inhibitory (Fig. [5a](#f5){ref-type="fig"}). This is consistent with the E3 dsRNA-binding domain being crucial for E3 to block signalling in response to poly(dA--dT), whilst the Z-DNA-binding domain is not required (Figs [2](#f2){ref-type="fig"} and [3](#f3){ref-type="fig"}). Several pro-inflammatory signalling pathways converge on the IKK complex, resulting in the phosphorylation of I*κ*B*α* and the translocation of p65 and p50 homo- and heterodimers to the nucleus ([@r16]). Since E3 can block NF-*κ*B activity in response to poly(dA--dT) it was important to determine whether E3 was a broad spectrum NF-*κ*B inhibitor or whether inhibition was specific to the poly(dA--dT)-sensing pathway. To address this, 293T cells were co-transfected with the NF-*κ*B reporter, and E3 expression plasmids, and then treated with tumour necrosis factor-alpha (TNF-*α*). Treatment of the control cells with TNF-*α* induced a strong induction of NF-*κ*B reporter activity. However, in contrast to poly(dA--dT) stimulation, neither full-length E3 nor Δ83N caused statistically significant repression of this activity (compare Fig. [5a](#f5){ref-type="fig"} against [5b](#f5){ref-type="fig"}). Similar results were obtained when using IL-1*β* to activate NF-*κ*B (Fig. [5c](#f5){ref-type="fig"}). These data confirm that E3 is not a broad spectrum NF-*κ*B inhibitor. E3 is an inhibitor of the RNA polymerase III cytoplasmic dsDNA signalling pathway --------------------------------------------------------------------------------- Recently, RNA polymerase III was described as a novel cytoplasmic PRR for AT-rich dsDNA ([@r1]; [@r11]). Specifically, RNA polymerase III transcribes poly(dA--dT), resulting in the production of 5′-triphosphate poly(A-U) RNA. This RNA then serves as a pathogen-associated molecular pattern (PAMP) for the PRR, retinoic acid-inducible gene I (RIG-I), resulting in IFN-*β* production ([@r38]). Consistent with these reports, transfection of total RNA harvested from poly(dA--dT) transfected cells resulted in IFN-*β* mRNA upregulation. Treatment of the RNA preparation with RNase A under low salt conditions, which results in the cleavage of dsRNA, ablated this induction (Fig. [6a](#f6){ref-type="fig"}), whereas treatment with DNase I did not (Fig. [6b](#f6){ref-type="fig"}). The specificity of the enzymic reactions was confirmed by treating the dsRNA analogue, poly(I : C), or the dsDNA, poly(dA--dT), with either RNase A or DNase I, respectively, under the same conditions as described above (Fig. [6c](#f6){ref-type="fig"}). The binding of 5′-triphosphate dsRNA by RIG-I results in the recruitment of the adaptor protein, IFN-*β* promoter stimulator 1 (IPS-1), which then acts as a scaffold for the activation of both IRF3 and NF-*κ*B ([@r21]). Indeed, transfection of the total RNA, which induced IFN-*β*, resulted in both ISG56.1 (Fig. [7a](#f7){ref-type="fig"}) and NF-*κ*B (Fig. [7b](#f7){ref-type="fig"}) stimulation in 293T cells. ISG56.1 is a transcriptional target for IRF3 and is, therefore, a good readout for IRF3-specific activation ([@r15]). However, this luciferase activity was ablated by E3, and the C-terminal dsRNA-binding domain was sufficient to mediate this inhibition (Fig. [7a, b](#f7){ref-type="fig"}). Furthermore, the level of luciferase activity obtained following transfection of the total RNA and subsequent inhibition by E3 was in line with that obtained when transfecting the synthetic dsRNA, poly(I : C), which is known to engage RIG-I (Fig. [7c](#f7){ref-type="fig"}). These results, therefore, support the conclusion that the immunostimulatory RNA engages the RIG-I signalling pathway and that E3 inhibits this process. DISCUSSION ========== The innate immune response represents the first line of defence against pathogenic micro-organisms. PRRs serve to detect conserved PAMPs. Once engaged PRRs signal via adaptor proteins to activate NF-*κ*B and IRF3, resulting in the secretion of type I IFN and pro-inflammatory cytokines and chemokines ([@r28]). The VACV E3 protein is a well known inhibitor of dsRNA PRRs and antiviral molecules. It was proposed that E3 may also function as an inhibitor of intracellular dsDNA PRRs by sequestering dsDNA via the N-terminal Z-DNA-binding domain ([@r35]). However, this hypothesis is based on amino acid similarity to the cytoplasmic dsDNA sensor, DAI, and has never formally been demonstrated. In agreement with recent reports we confirm that E3 is a potent inhibitor of type I IFN expression in response to the dsDNA, poly(dA--dT) ([@r25]; [@r35]). Using E3 deletion mutants we show that the assumption that this inhibition is mediated via the Z-DNA-binding domain is untrue. Indeed, the ability of E3 to block signalling in response to poly(dA--dT) maps entirely to the C-terminal dsRNA-binding domain. The inability of the E3 Z-DNA-binding domain to bind cytoplasmic dsDNA is also supported by pull-down assays in which biotinylated dsDNA could recruit other DNA-binding proteins, but not E3. This indicates that the virulence promoting function of the E3 N terminus is not due to binding of cytoplasmic dsDNA. Given the amino acid similarity of the E3 and DAI Z-DNA-binding domains these results also question whether the primary function of DAI is as a cytoplasmic PRR. Whilst this study was ongoing, RNA polymerase III was described as a novel cytoplasmic PRR for dsDNA. RNA polymerase III was found to transcribe AT-rich DNA, resulting in the production of 5′-triphosphate poly(A-U) RNA, which then served as a PAMP for RIG-I ([@r1]; [@r11]). These reports, therefore, perfectly explain both the observations from our group, and others, that the E3 dsRNA-binding domain can inhibit poly(dA--dT) signalling ([@r25]). Indeed, we have confirmed that the transfection of poly(dA--dT) results in the production of an immunostimulatory RNA, which activates both IRF3 and NF-*κ*B, thereby inducing type I IFN. Furthermore, expression of the E3 dsRNA-binding domain is sufficient to antagonize this response, suggesting that E3 is specifically binding the immunostimulatory RNA preventing activation of divergent signalling pathways downstream of the RIG-I receptor complex. In addition to the RNA polymerase III-sensing pathway, RIG-I may bind directly to dsDNA ([@r12]). However, we found no evidence to support E3 associating with dsDNA, ruling out the possibility that the E3 dsRNA-binding domain associates directly with DNA. We conclude that E3 is a novel inhibitor of the RNA polymerase III AT-rich DNA-sensing pathway and that this is mediated via the well characterized function of E3 binding to dsRNA. This might be important for VACV replication since the VACV genome is AT-rich and, therefore, one would envisage VACV being particularly sensitive to sensing via RNA polymerase III ([@r26]). This is difficult to test directly, however, due to the essential function of RNA polymerase III within cells. Whilst data presented in this report helps further clarify why VACV lacking the C terminus of the E3 protein presents with such an avirulent phenotype, the function of the N terminus in virulence remains unclear ([@r3]; [@r9]). Although our studies suggest that E3 does not bind immunostimulatory dsDNA, which is likely to reside within a right-handed B-DNA conformation, we cannot rule out that the E3 N terminus associates with left-handed Z-DNA. The formation of Z-DNA occurs when RNA polymerase II moves along DNA, resulting in negative supercoiling behind the transcription complex. Furthermore, sequences that favour the formation of Z-DNA may be a common feature of most gene transcription start sites ([@r30]). It is, therefore, interesting to speculate that the importance of the E3 N terminus and other Z-DNA-binding proteins might lie in regulating cellular transcription. Indeed, E3 has been reported to upregulate the transcription of certain genes ([@r23]), although, this remains controversial ([@r24]). The mechanism by which the E3 N terminus promotes virulence remains unknown. In summary, our results extend our knowledge of E3 as an inhibitor of innate immunity and provide an example of a dsDNA virus encoding an inhibitor of the recently described RNA polymerase III-dependent sensing of dsDNA in innate immunity. METHODS ======= Cell culture. ------------- Human embryonic kidney 293T and HeLa cells were cultured in Dulbecco\'s modified Eagle\'s medium (DMEM; Gibco) supplemented with 10 % fetal bovine serum, 100 U penicillin ml^−1^ and 100 μg streptomycin ml^−1^ and 2 mM [l]{.smallcaps}-glutamine (Sigma-Aldrich). Plasmids. --------- N-terminal FLAG-tagged E3, Δ26C and Δ83N were generated by PCR using the *E3L* gene from VACV strain Western Reserve (WR) as template. The PCR products were then cloned into the multiple cloning site of pcDNA4 TO (Invitrogen). The IFN-*β*, ISRE, IFN-stimulated gene 56.1 (IRF3) and NF-*κ*B reporters, which drive firefly luciferase expression, and the *Renilla* luciferase control plasmid (pRL-TK) have been described previously ([@r10]). Reporter assays. ---------------- 293T cells were seeded into 96-well tissue culture plates overnight prior to transfection with 70 ng of the indicated reporter plasmids, 10 ng pRL-TK and 70 ng of the appropriate E3 expression plasmid or empty vector control (pcDNA4) as indicated using PEI (Park Scientific). Following incubation overnight cells were stimulated as follows. For dsDNA stimulation, 293T cells were transfected with the indicated amounts of poly(dA--dT) (Sigma-Aldrich) for 24 h. For IFN-*α*, TNF-*α* and IL-1*β* (Peprotech) stimulations the cells were treated at the indicated concentrations for 7 h. Once the stimulation was complete cell lysates were harvested using passive lysis buffer and dual luciferase reporter assays were performed following the manufacturer\'s instructions (Promega). Relative luciferase activity was determined by normalizing the firefly luciferase data against the respective *Renilla* luciferase control. In all cases, data are from one of two to four independent experiments with similar qualitative results. Data from experiments performed in triplicate are expressed as means±[sem]{.smallcaps}. Statistical analysis was performed by an F-test to determine equal or unequal variance followed by a Student\'s *t*-test where appropriate. ELISA. ------ Levels of IFN-*β* in the supernatant of 293T cells were measured using a human IFN-*β*-specific ELISA kit (PBL Biomedical Laboratories) following the manufacturer\'s instructions. DNA pull-down assay and immunoblotting. --------------------------------------- DNA pull-down assays were performed by transfecting 293T cells with biotinylated dsDNA (Integrated DNA Technologies) using PEI. Following 4 h incubation, cells were lysed on ice using buffer containing 10 mM Tris/HCl, pH 8, 0.1 % NP-40, 10 mM MgCl~2~ and the cytoplasmic fraction was isolated. This was incubated with streptavidin agarose for 2 h and the beads were then washed three times with chilled PBS prior to analysis by both immunoblotting and silver staining (Invitrogen). Immunoblotting was performed as described previously ([@r10]) using mouse anti-FLAG (Sigma-Aldrich) and mouse anti-*α*-tubulin primary mAbs (Upstate). RNA isolation and enzyme treatment of nucleic acids. ---------------------------------------------------- Total RNA was isolated using TRIzol reagent (Invitrogen) following the manufacturer\'s instructions. Nucleic acids were treated either with RNase A using the conditions described previously ([@r11]) or with amplification grade DNase I following the manufacturer\'s instructions (Invitrogen). Real-time qPCR. --------------- RNA was reverse transcribed using Superscript III Reverse Transcriptase (Invitrogen) following the manufacturer\'s instructions. Real-time qPCR was performed using FAST SYBR green master mix (Applied Biosystems) on a 7900HT thermocycler (Applied Biosystems) and data were analysed using the relative quantification manager software. Real-time qPCR for GAPDH was used to normalize all data. This work was funded by the Medical Research Council. G. L. S. is a Wellcome Trust Principal Research Fellow. ::: {#f1 .fig} Fig. 1. ::: {.caption} ###### Poly(dA--dT) induces IFN-*β* reporter activity in a dose-dependent and TLR9-independent manner. 293T cells were co-transfected with pRL-TK and IFN-*β* reporter plasmids overnight. (a) The cells were then transfected with the indicated amounts of poly(dA--dT) for 24 h prior to the harvesting of cells for dual luciferase reporter assay. (b) 200 ng poly(dA--dT) was added to the cells in the presence, or absence, of the transfection reagent PEI, as indicated, and the cells then lysed 24 h later for dual luciferase reporter assay. Errors bars indicate the mean±[sem]{.smallcaps}. :::  ::: ::: {#f2 .fig} Fig. 2. ::: {.caption} ###### E3 inhibits IFN-*β* production in response to poly(dA--dT) via its C-terminal dsRNA-binding domain. (a) Schematic of full-length E3 and mutants in the C-terminal dsRNA-binding and N-terminal Z-DNA-binding domains. Amino acid positions are indicated at the top of full-length E3. (b) 293T and (c) HeLa cells were co-transfected with pRL-TK and IFN-*β* reporter plasmids together with pcDNA4, E3, Δ26C and Δ83N plasmids overnight. These cells were then transfected with 200 ng poly(dA--dT) and lysed for dual luciferase reporter assay after 24 h. (d) HeLa cells were co-transfected with pRL-TK and IFN-*β* reporter plasmids together with pcDNA4, E3, E3 dsRNA-binding domain and E3 Z-DNA-binding domain plasmids overnight and treated as described for (b) and (c). (e) 293T cells were transfected with pcDNA4, E3, Δ26C and Δ83N plasmids overnight. Cell lysates were then resolved by SDS-PAGE (12 % gel) and immunoblotted using anti-FLAG mAb (1 : 1000). Anti-*α*-tubulin mAb (1 : 5000) served as a protein loading control. The positions of molecular mass markers in kDa are indicated. :::  ::: ::: {#f3 .fig} Fig. 3. ::: {.caption} ###### E3 inhibits the secretion of IFN-*β* and subsequent ISRE activation in response to poly(dA--dT). (a) 293T cells were transfected with pcDNA4, E3, Δ26C and Δ83N plasmids overnight. The cells were then transfected with 750 ng poly(dA--dT) for 24 h and IFN-*β* in the cell supernatant was measured by ELISA. (b, c) 293T cells were co-transfected with pRL-TK and ISRE reporter plasmids together with pcDNA4, E3, Δ26C and Δ83N plasmids overnight. The cells were then either (b) transfected with 200 ng poly(dA--dT) for 24 h or (c) stimulated with 500 U IFN-*α* ml^−1^ for 7 h prior to the harvesting of cells for dual luciferase reporter assay. Error bars indicate the mean±[sem]{.smallcaps} (\**P*\<0.05, \*\**P*\<0.01). :::  ::: ::: {#f4 .fig} Fig. 4. ::: {.caption} ###### E3 does not bind cytoplasmic biotinylated dsDNA. 293T cells were transfected with plasmids expressing either FLAG-tagged B14 or E3 or an empty vector control (pcDNA4). The following day the cells were transfected with 20 μg biotinylated dsDNA and 4 h post-transfection the cells were lysed and the biotinylated dsDNA was pulled down from the cytoplasmic fraction using 100 μl streptavidin agarose. The pulled down fractions were then resolved on NuPAGE 4--12 % Bis-Tris pre-cast gels and then either silver stained (upper panel) or immunoblotted using anti-FLAG mAb (1 : 1000) (lower panel). The positions of molecular mass markers in kDa are indicated. :::  ::: ::: {#f5 .fig} Fig. 5. ::: {.caption} ###### The E3 dsRNA-binding domain inhibits NF-*κ*B reporter activity in response to poly(dA--dT). 293T cells were co-transfected with pRL-TK and NF-*κ*B reporter plasmids together with pcDNA4, E3, Δ26C and Δ83N plasmids overnight. The cells were then either (a) transfected with 200 ng poly(dA--dT) for 24 h or stimulated with (b) 250 ng TNF-*α* ml^−1^ or (c) 100 ng IL-1*β* ml^−1^ for 7 h prior to the harvesting of cells for dual luciferase reporter assay. Error bars indicate the mean±[sem]{.smallcaps} (\**P*\<0.05). :::  ::: ::: {#f6 .fig} Fig. 6. ::: {.caption} ###### Transfection of total RNA harvested from poly(dA--dT) transfected cells results in IFN-*β* expression. 293T cells were transfected with 12 μg total RNA extracted from 293T cells that had been transfected with 12 μg poly(dA--dT) for 24 h. Induction of IFN-*β* mRNA was confirmed to be specific for an immunostimulatory RNA species by treating total RNA with either (a) 0.1 mg RNase A ml^−1^ or (b) 0.1 U amplification grade DNase I μl^−1^. Total RNA was then extracted 24 h post-transfection and real-time qPCR performed for IFN-*β*. These data were normalized against GAPDH mRNA levels. Error bars indicate the mean±[sem]{.smallcaps} (\**P*\<0.05). (c) Cleavage of dsRNA or dsDNA by treatment with either RNase A or DNase I, respectively, was confirmed by treating poly(I : C) and poly(dA--dT) under the same conditions as for (a) and (b), respectively. The samples were then resolved on a 1 % agarose gel and stained using ethidium bromide. :::  ::: ::: {#f7 .fig} Fig. 7. ::: {.caption} ###### The E3 dsRNA-binding domain blocks IRF3 and NF-*κ*B activation in response to poly(dA--dT) transcribed RNA. 293T cells were co-transfected with pRL-TK and either (a) ISG56.1 or (b) NF-*κ*B reporter plasmids together with plasmids expressing either E3 or the E3 dsRNA-binding domain or the empty vector control (pcDNA4) overnight. The cells were then transfected with 1 μg total RNA extracted from 293Ts that had been transfected with 12 μg poly(dA--dT) for 24 h. At 24 h post-transfection, cells were lysed for dual luciferase reporter assay. (c) As a positive control 293T cells were co-transfected with the NF-*κ*B reporter plasmid as described for (b) and the following day were transfected with 800 ng poly(I : C) for 24 h prior to harvesting of cells for dual luciferase reporter assay. Error bars indicate the mean±[sem]{.smallcaps} (\**P*\<0.05, \*\**P*\<0.01, \*\*\**P*\<0.001). :::  :::

PubMed Central

2024-06-05T04:04:19.358785

2010-9-01

PMC3052527

*Vaccinia virus* (VACV) is the prototypical member of the genus *Orthopoxvirus* of the *Poxviridae.* It replicates in the cytosol and produces multiple types of infectious virions ([@r30]; [@r3]; [@r24]). The first infectious progeny is the intracellular mature virus (IMV), which is surrounded by a single-lipid envelope ([@r4]; [@r10]) and remains in the cell until cell lysis. However, some IMV are transported via microtubules to the early endosomes or *trans*-Golgi network where they are wrapped by two cellular membranes containing several VACV proteins. The resulting intracellular enveloped virus (IEV) is then transported on microtubules to the cell surface where the outer membrane fuses with the plasma membrane to externalize a double-enveloped virus by exocytosis. This virion is called cell-associated enveloped virus (CEV) if it remains on the cell surface, or extracellular enveloped virus (EEV) if it is released from the cell. The CEV/EEV outer membrane contains at least six viral proteins: A33 ([@r26]), A34 ([@r5]), A56 ([@r29]), B5 ([@r8]; [@r35]), F13 ([@r1]) and K2 ([@r32]; [@r33]). A34 is a type II transmembrane protein with different glycoforms between 23 and 28 kDa and its extracellular part contains a C-type lectin-like domain ([@r5]). A K151D point mutation in the VACV strain Western Reserve (WR) A34, which is present naturally in the VACV International Health Department (IHD)-J strain, caused an increase in EEV release ([@r2]). Similarly, deletion of the *A34R* gene (vΔA34R) from VACV WR caused a 25-fold increase in EEV, but such EEV had a fivefold reduction in specific infectivity ([@r19]). Deletion or suppression of the *A34R* gene caused a small plaque phenotype ([@r5]; [@r19]), inability to form actin tails ([@r36]; [@r28]) and severe attenuation ([@r19]). B5 is a 42 kDa type I transmembrane glycoprotein ([@r8]; [@r11]) with an extracellular domain composed of four short consensus repeats (SCRs) characteristic of complement control proteins ([@r31]), although there is no evidence that B5 regulates complement activity. After the SCRs B5 has an acidic stalk region (ST) before the transmembrane domain (TM) and a short cytoplasmic tail (CT). Both the SCRs and CT are dispensable for targeting B5 to the EEV membrane ([@r9]; [@r15]; [@r17]), although the latter affects its transport to the cell surface ([@r18]) and recycling via endosomes ([@r34]). B5 is needed for IMV wrapping to form IEV ([@r7]; [@r35]). B5 and A34 interact ([@r27]; [@r6]; [@r22]; [@r25]) and in the absence of A34, the amount of B5 incorporated in EEV is decreased markedly ([@r6]; [@r22]; [@r25]). B5 and A34 each affect the glycosaminoglycan (GAG)-dependent rupture of the EEV outer membrane during EEV entry ([@r14]; [@r25]). Although B5 expressed on its own displays a cellular localization profile very similar to the one observed in the context of viral infection ([@r12]; [@r16]), this is not the case for A34. In infected cells, A34 is found at the Golgi, on the cell surface and in CEV/EEV. In contrast, when it is expressed alone it accumulates in the perinuclear region and does not go to the plasma membrane ([@r16]). In addition, attempts to express A34 on its own from classical eukaryotic expression vectors (pcDNA3, pCI) and in several recombinant expression systems yielded poor levels of expression. For example, in a system where soluble forms of EEV proteins A56, B5 and A33 were expressed in CHO cells and secreted into the medium, the yield obtained for A34 was about 20-fold lower than for B5 ([@r13]; [@r23]; M. Law unpublished data). Here, we present data showing that in the absence of B5, the level of A34 is markedly decreased, most probably because of misfolding and consequential degradation. In a recent study on EEV entry, we generated several VACV mutants with alterations in the B5 stalk acidic residues ([@r25]), for the structure of these mutants see Fig. [1(a)](#f1){ref-type="fig"}. Using those viruses, we analysed lysates from infected RK13 cells by immunoblotting with mouse monoclonal antibodies (mAbs) against B5 (36-6; [@r25]), A34 (34-1; [@r25]) and the IMV protein D8 (AB1.1; [@r21]) as an infection control (Fig. [1b](#f1){ref-type="fig"}). Mouse mAbs against the A34 and B5 proteins were produced by immunization of mice with purified recombinant protein expressed from mammalian cells ([@r13]; [@r23]). As noted previously, these mutations affected the electrophoretic mobility of B5 ([@r25]). In addition, this analysis showed that in the absence of B5, the amount of A34 in the infected cells was reduced considerably, and with some of the B5 mutants the glycosylation profile of A34 was different. When B5 was deleted, one distinct band at about 20 kDa (A34\*) was observed instead of the 23--28 kDa bands made by wild-type (WT) virus. In contrast, deletion of all the B5 SCRs (vSCR0), had no effect and A34 retained the WT profile (note the remaining B5 fragment was not visible due to its small size). However, substitution of the acidic residues of the stalk with alanines (vST2-35ala) had the same effect as deleting B5. Moreover, substitution of the five acidic residues closest to the membrane (vST23-35ala) also led to the A34\* profile. Interestingly, vST2-16 and vST28-35 showed a mixed profile with the A34\* band evident together with higher molecular mass forms (Fig. [1b](#f1){ref-type="fig"}). Since vST23-35ala and vST28-35ala differ only at residue 23 and display distinct A34 glycosylation profiles, we wondered if that amino acid could by itself influence A34 glycosylation. To address this, a recombinant VACV in which the aspartic acid 23 of the stalk region was mutated to alanine (vST23ala) was constructed by using transient dominant selection as described previously ([@r25]). When tested by immunoblotting as above, vST23ala showed the same A34 profile as the WT virus and therefore, mutating the aspartic acid 23 is not sufficient to alter A34 glycosylation (Fig. [1c](#f1){ref-type="fig"}). Overall, analysis of these mutants suggested that the B5 stalk is important for correct glycosylation of A34. To determine more precisely the nature of the A34 20 kDa isoform, we used drugs that affect glycosylation: namely tunicamycin, an inhibitor of *N*-acetylglucosamine transferase and kifunensin, an inhibitor of *α*-mannosidase I (Fig. [2a](#f2){ref-type="fig"}). RK13 cells were infected with WR or vΔB5R at 5 p.f.u. per cell for 90 min and then incubated overnight in Dulbecco's modified Eagle's medium containing 2.5 % fetal bovine serum with or without 1 μM tunicamycin or 5 μM kifunensin. Cell lysates were then prepared and analysed by immunoblotting (Fig. [2b](#f2){ref-type="fig"}). Treatment of WR-infected cells with tunicamycin produced a single A34 band, corresponding to the unglycosylated polypeptide (A34~ug~). The A34~ug~ is predicted to be 19.6 kDa ([@r5]), and we observed a slightly smaller band of about 17 kDa. In vΔB5R-infected cells, A34 is slightly larger than A34~ug~ and was still reduced in size in the presence of tunicamycin, indicating that A34\* is a partially glycosylated form. Notably, levels of B5 and A34 both decreased in the presence of tunicamycin, indicating that glycosylation is required for stability of these proteins. In the presence of kifunensin, the A34\* pattern was observed in WR-infected cells, suggesting that A34\* represents an intermediate with nine mannose residues (Man~9~), before trimming by the *α*-mannosidases. Another interesting observation was that the amount of A34 in the absence of B5 was increased by kifunensin treatment. This is in agreement with reports showing that processing by *α*-mannosidases acts as a signal to target misfolded proteins for proteasomal degradation and that inhibition of these enzymes by kifunensin treatment leads to accumulation of misfolded (Man~9~)-glycoproteins ([@r20]). To address this further, the stability of A34 with time was investigated (Fig. [2c](#f2){ref-type="fig"}). RK13 cells were infected with WR or vΔB5R as before and cell lysates were prepared at 4, 8, 12 and 24 h p.i. and analysed by immunoblotting. Up to 8 h p.i., A34 was easily detected in cells infected by either virus, although it was less abundant in vΔB5R-infected cells, but thereafter A34 declined substantially in vΔB5R-infected cells and was barely visible at 24 h. This suggests that synthesis and accumulation of A34 starts normally without B5, but as the rate of synthesis decreases later during infection, the level of A34 declines. Collectively, those data suggest that in the absence of B5, A34 is misfolded and degraded. Next, the effect of A34 glycosylation status on the subcellular localization was investigated. BSC-1 cells were infected with viruses at 2 p.f.u. per cell for 8 h, fixed with PBS--4 % paraformaldehyde (PFA) for 10 min on ice and then in PBS--8 % PFA for 20 min at room temperature. Fixed cells were permeabilized with 0.2 % Triton X-100 and incubated with anti-A34 mAb and a rabbit anti-protein disulphide isomerase Ab (anti-PDI; Abcam) to stain the endoplasmic reticulum (Fig. [3](#f3){ref-type="fig"}). Consistent with Fig. [2(c)](#f2){ref-type="fig"}, significant levels of A34 were present in both WR- and vΔB5R-infected cells at 8 h p.i. In WR-infected cells, the anti-A34 mAb labelled the Golgi as well as punctate structures corresponding to virions in the periphery, as described previously ([@r16]), but no significant co-localization with PDI was observed (Fig. [3a](#f3){ref-type="fig"}). In contrast, in vΔB5R-infected cells A34 was present throughout the cell in a reticular pattern co-localizing with PDI (Fig. [3a](#f3){ref-type="fig"}), similar to that seen when A34 was expressed from a Semliki Forest virus vector ([@r16]). No staining of VACV particles was observed, and this may be explained by the wrapping defect of vΔB5R ([@r7]; [@r35]). Cells infected with vΔB5R, vST23-35ala and vST28-35ala all showed a significant amount of A34 in the ER and nuclear envelope, but vST23-35ala and vST28-35ala also showed some staining of viral particles (Fig. [3b](#f3){ref-type="fig"}; enlargement of the punctate staining representing virions is shown for the vST28-35 image). Cells infected with vST23-35ala generally had fewer particles than WR- or even vST28-35ala-infected cells. This is consistent with the fact that vST23-35 produced much less EEV than WR, whereas EEV production by vST28-35ala was only slightly reduced ([@r25]). Overall, these data show that the A34\* band in SDS-PAGE correlates with the presence of A34 in the ER. This is consistent with A34\* being a (Man~9~)-A34 that accumulates in the ER. Data presented here indicate that in the absence of B5, or in presence of some mutated forms of B5, A34 is not correctly folded and accumulates in the ER as a partially glycosylated intermediate. Ultimately, at least in the case of vΔB5R, this would lead to proteasomal degradation. A hypothesis to explain these data would be that an interaction of the negatively charged acidic residues in the B5 stalk region with positive charges of A34 might help A34 to acquire the correct conformation. Alternatively, the B5 stalk could play a role in the trafficking of the B5/A34 complex. Taken together with the previous data showing that A34 is required for proper incorporation of B5 in the EEV membrane, this shows that there is a complex mutual interaction between these two VACV proteins making it difficult to unravel their respective roles in virus wrapping, egress and re-entry due to their inter-dependence. We thank members of our laboratory for their help and support, in particular Dr Gareth Morgan and Mr Mike Hollinshead for helpful discussions and assistance with microscopy. This work was supported by the Medical Research Council. G. L. S. is a Wellcome Trust Principal Research Fellow. ::: {#f1 .fig} Fig. 1. ::: {.caption} ###### B5 affects the abundance and size of A34. (a) Structure of the B5 stalk mutants used in this study. The amino acid residues of the stalk are numbered 1--39 and modified residues are shown in bold. (b, c). RK13 cells were infected with the indicated viruses and cell lysates were prepared at 16 h post-infection (p.i.) and immunoblotted with mAbs against B5, D8 and A34. The positions of molecular mass markers are shown on the left in kDa. :::  ::: ::: {#f2 .fig} Fig. 2. ::: {.caption} ###### In the absence of B5, A34 has aberrant glycosylation and is targeted for degradation. (a) Simplified diagram of the glycosylation pathway showing where tunicamycin and kifunensin act. Gluc, Glucose; GlcNAc, *N*-acetylglucosamine; Man, mannose. (b) RK13 cells were infected with WR and vΔB5R and incubated with or without tunicamycin (Tunic.) or kifunensin (Kifu.). At 16 h p.i., cell lysates were prepared and analysed by immunoblotting with anti-B5, anti-D8 and anti-A34 mAbs. Another image of the bottom section of the membrane is shown for clarity. A34~ug~ and B5~ug~, unglycosylated A34 and unglycosylated B5. (c) A34 is synthesized in the absence of B5, but is degraded over time. RK13 cells were infected with WR or vΔB5R and harvested at the indicated times p.i. Cell lysates were prepared and analysed by immunoblotting as in (b). The positions of molecular mass markers are shown on the left in kDa. :::  ::: ::: {#f3 .fig} Fig. 3. ::: {.caption} ###### Absence or alteration of B5 leads to accumulation of A34 in the ER. (a, b). BSC-1 cells were infected for 8 h with the viruses shown, fixed and processed for immunofluorescence using anti-A34 mAb followed by anti-mouse-Alexa 546 (red) and anti-PDI followed by anti-rabbit-Alexa 488 (green). Samples were viewed on a Zeiss 510 Meta confocal microscope using Zeiss LSM software. The right panel of each row shows the merged image of the left and centre panel. Boxes within individual panels show regions of the cell before and after magnification. Bars, 10 μM. :::  :::

PubMed Central

2024-06-05T04:04:19.361399

2010-7-01